/:»

     

4/1,
/1/& 7’ / 4/7
/ {’7
,3 c 5’
RETURNING MATERIALS: /:3 49¢?
MSU P1ace in book drop to ‘

remove this checkout from
your record. FINES will
be charged if book is
returned after the date
stamped below.

 
 
 

LIBRARIES
“_

       

i grew. DEanulQ.
, r.‘ ;. hLNEW ’

1 1.

 

 

 

 

PESTICIDES AND POLICY:
RISK-BENEFIT ANALYSIS AT THE
ENVIRONMENTAL PROTECTION AGENCY

BY
Patty Teresa Skelding

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Agricultural Economics

1984

J, '\
V..- b
' . t
\‘ ‘~

ABSTRACT
PESTICIDES AND POLICY:

RISK BENEFIT ANALYSIS AT THE
ENVIRONMENTAL PROTECTION AGENCY

BY

Patty Teresa Skelding

There has been much criticism of the 0.8. Environmental
Protection Agency's regulation of pesticides, but very
little empirical work to describe that process. In this
research, the quality of EPA's information on risks and
benefits of pesticide use is quantified for eight case-study
pesticides, using EPA documents. Goodman and Kruskal's
coefficient of ordinal association measures correlations
between those data and five explanatory variables depicting
interest group incentives to influence the regulatory pro-
cess. The five explanatory variables are: value of pesti-
cide use to manufacturers, percent of crop treated, per acre
user losses, total user losses, agiégigg”S§:§BEWEESEEEIE§>
These variables are also quantified from EPA documents. T37
general, risk information was poorer than benefits informa-
tion, interest groups do not influence risk information,
manufacturers and users influence benefits information, the
pesticide's relative risk does not influence benefit infor-

mation, and pesticide manufacturers and users impact EPA

decisions. I/

ACKNOWLEDGMENTS

I would like to thank several people who made the com-
pletion of this arduous project possible. My husband Mark
stayed by my side during the setbacks as well as the vic-
tories, and he suffered more than any spouse should. My
parents and several close friends, including Diane Smith and
Charles and Katy Abdalla, gave me the extra push I needed to
keep going.

My Major Professor and Thesis Advisor, Dr. Eileen van
Ravenswaay, was a constant source of creativity, knowledge,
incentive and support. I am intellectually indebted to her,
and I also owe her a great deal for her patience. The other
members of my thesis committee, Dr. Warren Samuels and Dr.
Lawrence Libby, also contributed a great deal to the final
product of this research.

I also deeply appreciate the generous financial support
for this project from the Michigan State University Agricul-
tural Experiment Station and the 0.5. Department of Agricul-
ture Economic Research Service.

Above all, I thank the people of Community Church of

Owosso and I thank God.

-11-

TABLE OF CONTENTS

LIST OF TABLES O...00....OOOOOOOOOOOOOOOOOOOOOOOOOOOOOCOOOOOOOOO Vi
LIST OF FIGURES .COCOOOCOOOCOOOOOOOO0..0.0...OOOIOOOOOOOIOOOOOOO XVi

CHAPTER 1: INTRODUCTION
OVERVIEW .................................................... 1
THE RESEARCH PROBLEM ........................................ 3
THE RESEARCH SETTING: EPA REGULATION OF PESTICIDES .......... 10
ORGANIZATION OF THE THESIS .................................. 20

CHAPTER 2: THEORIES OF THE REGULATORY PROCESS:
A REVIEW OF THE LITERATURE
INTRODUCTION ................................................ 22
THE NATURE OF REGULATION .................................... 22
THEORIES OF REGULATORY DECISION MAKING ...................... 29
Theories With Agency-Wide Objective Functions .......... 29

Agency survival ................................... 30
Budget-maximizing or size-maximizing agencies ..... 39
Summary -- theories with agency-wide objective

functions ........................................ 43

Theories Assuming Different Goals Within an Agency ..... 44
coalitions Of regulators 0......OOOCOOIOOOOOOOOOOOO 45
Unique objective functions for individual

regulators ....................................... 47
THE USE OF INFORMATION IN REGULATORY DECISION MAKING ........ 52
Is More Information Better? ............................ 52
Characteristics of Information Search .................. 53
The Use of Information ................................. 57
'Information leads to rational decisions ........... 57
Information is used by conflicting interests
to influence decision makers ..................... 57
Information legitimizes decisions ................. 58
Information clarifies issues and potential
solutions ........................................ 60
Multiple uses of information ...................... 61

Summary: The Nature and Use of Information in

Regulatory Decision Making ............................ 62
EPA'S REGULATION OF PESTICIDES: DETERMINANTS OF
DECISION MAKING AND USES OF INFORMATION .................... 63

HAPTER 3: CONCEPTUALIZING THE IMPACTS OF PESTICIDE USE
INTRODUCTION ................................................ 66
OVERVIEW OF EPA'S ANALYSIS OF PESTICIDES .................... 66
PROBLEMS IN CONCEPTUALIZING RISKS AND BENEFITS .............. 68

-iii-

The Amount of Information and Difficulties
in TraCAng Impacts OOOOOOOOOOOOOOOOOOOOOOOOOOOOOVOOOOOOO
Value Judgments in Risk-Benefit Analysis ...............
CONCEPTUALIZING THE IMPACTS OF PESTICIDE USE ................
. R18k Measurement OOOOOOOOOOOOOOOOOOOOOOOO'OOOOOOOOOOOOOOO

TOXiC1ty OOOOOOOOOOOOOOOOOOOOOOOOOO‘OOOOOOOOOOOOOOOO

EnVironmental fate O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O

exposure OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

The risk analysis .................................

The measurement 0: Benefits O O O O O O O O O O O O O O O O O O O O O O O O O O O O
Theoretical discussion of the net benefits

OE pestiCide use O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O

Real impaCts O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O

Distribution of impacts ...........................

PROBLEMS IN MEASURING IMPACTS ...............................

 

CHAPTER 4:.METHODOLOGY -
INTRODUCTION OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 91
CHOICE OF DATA BASE OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 91
MEASURING EXPLANATORY VARIABLES ............................. 94

wniCh variableS? OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 94

How "eaaured? OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 98

Value of pesticide use to manufacturers ........... 99

Value of pesticide use to producers ............... 100

Percentage of producers affected .................. 104

Losses to risk-bearers ............................ 105

Problems in measuring explanatory variables ....... 108

MEASURING DEPENDENT VARIABLES OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 109

mien measured? OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 109

80" measured? OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 113

Problems in Measuring Dependent Variables .............. 122

Specific comments on risk coding .................. 122

General comments on risk coding ................... 122

. General comments on benefits coding ............... 126

MANIPULATING DATA FOR COMPARISON BETWEEN PESTICIDES ......... 127
MEASURING CORRELATION BETWEEN DEPENDENT AND EXPLANATORY

VARIABLES OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 129

SUMMARY OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO I133

CHAPTER 5: RESULTS: QUALITY OF INFORMATION ON
RISK OF PESTICIDE USE ‘

INTRODUCTION OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 137-

OVBRVIEW OF EPA'S RISK ASSESSMENT PROCESS eeeeeeeeeeeeeeeeee'e 137

QUALITY OF EPA'S RISK INFORMATION OOOOOOOOOOOOOOOOOOOOOOO‘OOOO 1‘2

CORRELATION BETWEEN EXPLANATORY VARIABLES AND .

QUALITY OF RISK INFORMATION VARIABLES ...................... 146
Explanation of Tables .................................. 146
Findings for Value of Pesticide Use to Manufacturers ... 150
Findings for Percent of Crop Treated .................... 158
Findings for Annual Per Acre User Losses ............... 160
Findings for Annual Total User Losses .................. 161
Findings for EPA' s Risk Ranking ........................ 163

—v-

Summary: Correlation Between Explanatory Variables
and Quality of Information on Risk ....................

CHAPTER 6: RESULTS: QUALITY OF INFORMATION ON
BENEFITS OF PESTICIDE USE ......................................
GENERAL OBSERVATIONS ON EPA'S BENEFITS ANALYSIS .............
THE QUALITY OF EPA'S BENEFITS INFORMATION ...................
CORRELATION BETWEEN EXPLANATORY VARIABLES AND
THE QUALITY OF INFORMATION ON BENEFITS .....................
Findings for Value of Pesticide Use to Manufacturers ...
Findings for Percent of Crop Treated ...................
Findings for Annual Per Acre User Losses ...............
Findings for Annual Total User Losses ..................
Findings for EPA's Risk Ranking ........................
Findings for All of the Explanatory Variables
and EPA's Decisions on Pesticide Registration .........
Summary: Correlation Between Explanatory Variables
and Quality of Information on Benefits ................

CHAPTER 7: CONCLUSIONS

OVERVIEW OF RESEARCH ........................................
SUMMARY OF RESULTS ..........................................
ANALYSIS OF RESULTS .........................................
Value Judgments in Pesticide Regulation ................
Correlation Between Variables ..........................
EPA: Vote-Maximizer or Budget-Maximizer? ...............
Predicted results for vote-maximizing agency ......
Predicted results for budget-maximizing agency ....

Summary: EPA Incentives and Recommendations to
Achieve Change ........................................

THE ROLE OF INFORMATION: IMPLICATIONS FOR ECONOMIC

POLICY ANALYSIS ............................................
IMPLICATIONS OF FINDINGS FOR PROPOSED REGULATORY REFORMS ....
The Bush Task Force Reforms ............................
National Academy of Sciences Recommendations ...........
DIRECTIONS FOR FUTURE RESEARCH ..............................

APPENDICES
APPENDIX A: RANKING OF PESTICIDE USES ACCORDING
TO FOUR EXPLANATORY VARIABLES ..............................
APPENDIX B: CORRELATIONS BETWEEN EXPLANATORY
VARIABLES AND QUALITY OF INFORMATION .......................

BIBLIOGRAPHY ......OOOOOOOOO......OCOOOOOOOOOO.........OOOOOOOOOO

165

167
167
170

174
177
180
182
184
187

189

191

194
195
197
197
198
201
201
202

204
206
208
208

212
215

217
225
294

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

4.1

4.2

4.14

4.15

LIST OF TAB; ES

Value to Manufacturers of Uses of DBCP. .

Total User Losses from Pesticide Cancel-
lation for DBCP (Annual)O OOOOOOOOOOOOOOO
Annual Per Acre and Total User Losses
from Pesticide Cancellation for DBCP. ...
Percentage of Acres Treated with DBCP
(Per Crop Cycle).

Measuring Relative Risk for Amitraz. ....
EPA's Pesticide Risk Ranking: Criteria
Used to Determine Risk and Score for Each
PestiCideO OOOOOOOOOOOOOOOO0.000000000000
Typical Risk Variables for Which Quality
of Information was Measured. ............
Benefits Variables for Which Quality of
Information was Measured. ...............
Measurement Scale for Quality of Informa-
tion: TOXiCitYe .....OOOOOOOOOOOOOOOOOOO
Measurement Scale for Quality of Informa-
tion: Exposure.

Measurement Scale for Quality of Informa~

tion: Dermal and Inhalation Exposure. ..
Measurement Scale for Quality of Informa-
tion: Risk Assessment. .................
Measurement Scale for Quality of Informa-
tion: Risk of Alternative Pest Control.

Measurement Scale for Quality of Informa-
tion: Acute Toxicity to Wildlife. .....

Measurement Scale for Quality of Informa-
tion: Significant Decreases in Nontarget
Populations.

.....O...’

100

102

103

105

106

107

111

112

115

115

116

116

116

117

118

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

4.16

4.17

4.20

4.21

5.3

5.4

5.5

-vii-

Measurement Scale for Quality of Informa-

tion:

Fatality to Endangered Species. ..

Measurement Scale for Quality of Informa-
All Changes in Pest Control Costs
and Changes in Value of Production Data
Except Price.

tion:

Measurement Scale for Quality of Informa-

tion:

Price.

OOOOOOOOOO........OOOOOOOCOOOOOOOOOO.

Measurement Scale for Quality of Informa-
Per Acre Change in Pest Control

tion:

Costs,
Costs,
duction,
duction,

Costs,
GOOdSI

Nonusers,
facturers),
economic Impacts,

Total

Change

in

Pest

Control

Per Acre Change in Value of Pro-

Change

Impacts
Distribution of

Marketers,
Geographic

on

in

Social Impacts,

Other
Producers

of
Impacts

Consumers,

Impacts,

Total Change in Value of Pro-
Production

Other

(Users,

Manu~

Macro-
Degree

of Compliance, Costs of Restrictions. ..

Measurement Scale for EPA Decision.

Specific Problems in Measuring Dependent
Risk Variables.

Calculation

Chlorobenzilate:

Calculation
EPA's Pesticide Risk Ranking and Consumer

Risk of Oncogenicity.

Quality of Risk Information.

of

Gamma

Coefficient

for

Value of Pesticide Use
to Manufacturers and Consumer Risk of
Oncogenicity.

of

Gamma

Coefficient

for

Number of Uses Measured for Each Pesti-
cide According to Explanatory Variables.

Summary of Correlations Between

Quality

of Risk Data and Value of Pesticide Use

to Manufacturer.

Summary of Correlations Between Quality

of Risk Data and Percent of Crop Tested.

Summary of Correlations Between Quality
of Risk Data and Annual Per Acre User

LOSSOS.

......OCOOOOOOOOOOO000......OOOOOO

119

120

120

121

121

123

134

135

144

151

152

153

-viii-

Table 5 .6 Summary of Correlations Between Quality
of Risk Data and Annual Total User
Losses. ......OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 155

Table 5.7 Summary of Correlations Between EPA's
Risk Rankimg and the Average Quality of
Risk Information. I.........OIOOOOOOOOOOOOOO0...... 164

Table 6.1 Summary of Quality of Benefits Informa-

tion. ......OIOOOOO.........OOOOOOOOOOOOI0.00...... 172

Table 6.2 Summary of Correlations Between Quality
of Benefits Data and Value of Pesticide
Use to Manufacturer. .............................. 178

Table 6.3 Summary of Correlations Between Quality
of Benefits Data and Percent of Crop
Treated. ......OOOO........OOOOOOOOOOOOOOOOO0...... 181

Table 6.4 Summary of Correlations Between Quality
of Benefits Data and Annual Per Acre User
Losses. OOOOOOOOOOOOO0.000000000000000000000000000. 183

Table 6.5 Summary of Correlations Between Quality
of Benefits Data and Annual Total User
Losses. ......OOOOOOO0.0000IOOOOOOOOIOOOOOOCOOOOOO. 185

Table 6.6 Summary of Correlations Between EPA's
Pesticide Risk Ranking and the Average
Quality of Benefits Information. .................. 188

Table 6. 7 Summary of Correlations Between EPA
Pesticide Registration Decisions and Four
Explanatory variables. .OOOOOOOOOOOOO OOOOOOO O OOOOOO 190

Table 7.1 Changes in Pesticide Regulation Since
President Reagan's Inauguration. .................. 213

Table A1. Ranking of Uses for Each Pesticide
According to Value of Pesticide.................... 217

Table A2. Ranking of Uses for Each Pesticide
According to Annual Total User Losses
from Pesticide Cancellation. ...................... 219

Table A3. Ranking of Uses for Each Pesticide
According to Annual Per Acre User Losses
from Pesticide Cancellation. ...................... 221

Table A4. Ranking of Uses for Each Pesticide
According to Percent of Crop Treated. ............. 223

Table

Table

Table

Table

Table

Table

Table

Table

.Table

Table

Table

El.

82.

B3.

B4.

85.

86.

B7.

88.

B9.

810.

311.

...-ix...
Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Consumer Risk, as Measured

by Gamma. ..................................0......

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Ground Applicator Risk, as
Measured by Gamma. ................................

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Mixer and Loader Risk. as
Measured by Gamma. ................................

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Pilot Risk, as Measured by

Gamma. ........................................0...

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Risk to Persons Exposed to
Drift. as Measured by Gamma. ......................

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Farmworker Risk, as
Measured by Gamma. ................................

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Risk to Persons Exposed to
Spills, as Measured by Gamma. .....................

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Risk to Aerial Ground
Crews. as Measured by Gamma. ......................

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Animal Risk, as Measured

by Gamma. .....................................0...

Correlations Between Percent of Crop
Treated and Quality of Information on
Consumer Risk, as Measured by Gamma. ..............

Correlations Between Percent of Crop

Treated and Quality of Information on
Ground Applicator Risk, as Measured by

“mma. ........................................0...

225

226

227

228

229

230

231

232

233

234

235

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

812.

813.

814.

815.

816.

817.

818.

819.

820.

821.

822.

-x-

Correlations Between Percent of Crop
Treated and Quality of Information on
Mixer and Loader Risk. as Measured by

Gamma.

Correlations Between Percent of Crop
Treated and Quality of Pilot Risk Infor-
mation. as Measured by Gamma. .....................

Correlations Between Percent of Crop
Treated and Quality of Information on
Risk to Persons Exposed to Drift, as
Measured by Gamma. ................................

Correlations Between Percent of Crop
Treated and Quality of Farmworker Risk
Information, as Measured by Gamma. ................

Correlations Between Percent of Crop
Treated and Quality of Information on
Persons Exposed to Spills, as Measured by

Gamma.

Correlations Between Percent of Crop
Treated and Quality of Information on
Aerial Ground Crew Risk, as Measured by

Gamma.

Correlations Between Percent of Crop
Treated and Quality of Information on
Animal Risk] as measured by Gamma. 0000000000000...

Correlations Between Annual Per Acre User
Losses from Cancellation and Quality of
Information on Risk to Consumers. as
Measured by Gamma. ................................

Correlations Between Annual Per Acre User
Losses from Cancellation and Quality of
Information on Ground Applicator Risk, as
Measured by Gamma. ................................

Correlations Between Annual Per Acre User
Losses from Cancellation and Quality of
Information on Risk to Mixers and Load-
ers, as Measured by Gamma. ........................

Correlations Between Annual Per Acre User
Losses from Cancellation and Quality of
Information on Pilot Risk, as Measured by

Gamma o

236

237

238

239

240

241

242

243

244

245

246

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

823.

824.

825.

826.

827.

828.

829.

830.

831.

832.

833.

Correlations Between Annual Per Acre User
Losses from Cancellation and Quality of
Information on Risk to Persons Exposed to
Drift. as Measured by Gamma. ......................

Correlations Between Annual Per Acre User
Losses from Cancellation and Quality of
Information on Risk to Farmworkers. as
Measured by Gamma. ................................

Correlations Between Annual Per Acre User
Losses from Cancellation and Quality of
Information on Risk to Persons Exposed to
Spills. as Measured by Gamma. .....................

Correlations Between Annual Per Acre User
Losses from Cancellation and Quality of
Information on Risk to Aerial Ground
Crews. as Measured by Gamma. ......................

Correlations Between Annual Per Acre User
Losses from Cancellation and Quality of
Information on Animal Risk. as Measured

by Gamma. ...00............OOCOOCOOOOOOCCO0.00....O

Correlations Between Annual Total User
Losses from Cancellation and Quality of
Information on Risk to Consumers. as
Measured by Gamma. ................................

Correlations Between Annual Total User
Losses from Cancellation and Quality of
Information on Risk to Ground Applica-
tors. as Measured by Gamma. .......................

Correlations Between Annual Total User
Losses from Cancellation and Quality of
Information on Risk to Mixers and
Loaders. as Measured by Gamma. ....................

Correlations Between Annual Total User
Losses from Cancellation and Quality of
Information on Risk to Pilots. as
Measured by Gamma. ................................

Correlations Between Annual Total User
Losses from Cancellation and Quality of
Information on Risk to Persons Exposed to
Drift. as Measured by Gamma. ......................

Correlations Between Annual Total User
Losses from Cancellation and Quality of
Information on Risk to Parmworkers. as
Measured by Gamma. ................................

247

248

249

250

251

252

253

254

255

256

257

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

834.

835.

836.

837.

838.

839.

840.

841.

842.

843.

844.

Correlations

Between

-xii-

Annual Total User

Losses from Cancellation and Quality of
Information on Risk to Persons Exposed to
Spills. as Measured by Gamma. .....................

Correlations

Information

Crews.

Correlations

Between Annual Total User
Losses from Cancellation and Quality of

on

Between Annual Total

Risk
as Measured by Gamma. ......OOOOOOOOOOOOO...

to

Aerial Ground

User

Losses from Cancellation and Quality of
Information on Animal Risk. as Measured

by Gamma.

Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Consumer Risk.

as Measured by Gamma. ..............

Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Pilot Risk. as Measured by Gamma. .................

Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Risk to Persons Exposed to Drift. as

Measured by Gamma.

Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Farmworker Risk. as Measured by Gamma. ............

Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on

Aerial Ground Crew Risk.

Gamma.

as Measured by

Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Risk to Persons Exposed to Accidents and
Spills. as Measured by Gamma. .....................

Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Ground Mixer and Loader Risk. as Measured

by Gamma.

Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Pesticide Toxicity. as Measured by Gamma. .........

258

259

260

261

261

262

262

263

263

264

264

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

845.

846.

847.

848.

849.

850.

851.

852.

853.

854.

—xiii-

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Changes in Pest Control
Costs from Pesticide Cancellation. as
Measured by Gamma. ................................

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Changes in Value of Pro-
duction from Pesticide Cancellation. as
Measured by Gamma. ................................

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Other Socio-Economic
Impacts. as Measured by Gamma. ....................

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Distribution of Impacts of
Pesticide Cancellation. as Measured by

Gamma. ......OOOOOOOOOO0.0......OOOOOOOOOO000......

Correlations Between Value of Pesticide
Uses to Manufacturers and Quality of
Information on Compliance with Proposed

Restrictions. as Measured by Gamma.

Correlations Between Value of Pesticide
Uses to Manufacturers and Pesticide Reg-
istration Decisions. as Measured by

Gamma. 0.0.0.000.........OOOOOOCUOOOOOOOO0.0.000...

Correlations Between Percent of

Crop

Treated and Quality of Information on
Changes in Pest Control Costs from Pesti-

cide Cancellation. as Measured by Gamma.

Correlations Between Percent of

Crop

Treated and Quality of Information on

Changes in Value of Production

from

Pesticide Cancellation. as Measured by

Gamma. O.....O0.000000000000000.........OOOOOOOOOOC

Correlations Between Percent of

Crop

Treated and Quality of Information on
Other Socio-Economic Impacts. as Measured

by Gamma. O......OOOOOC......OOOOCOO00.00..........

Correlations Between Percent of

Crop

Treated and Quality of Information on
Distribution of Impacts of Pesticide

Cancellation. as Measured by Gamma.

265

266

267

268

269

270

271

272

273

274

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

855.

856.

857.

858.

859.

860.

861.

862.

863.

864.

-xiv—
Correlations Between Percent

of Crop

Treated and Quality of Information on

Estimated Compliance with

Proposed

Restrictions, as HeaSUIEd by Gamma0 000000000000000

Correlations Between Percent

of Crop

Treated and Pesticide Registration

Decisions. as Measured by Gamma.

Correlations Between Annual Per Acre User
Losses and Quality of Information on
Changes in Pest Control Costs from Pesti-
cide Cancellation. as Measured by Gamma. ..........

Correlations Between Annual Per Acre User
Losses and Quality of Information on
Changes in Value of Production from
Pesticide Cancellation. as Measured by

Gamma. ................0...........................

Correlations Between Annual Per Acre User
Losses from Pesticide Cancellation and
Quality of Information for Other Socio-
Economic Impacts. as Measured by Gamma. ...........

Correlations Between Annual Per Acre User
Losses from Pesticide Cancellation and
Quality of Information on Distribution of
Impacts of Pesticide Cancellation. as
Measured by Gamma. ................................

Correlations Between Annual Per Acre User
Losses from Pesticide Cancellation and
Quality of Information on Compliance with
Proposed Restrictions. as Measured by

Gamma. ......O.....................................

Correlations Between Annual Per Acre User
Losses from Pesticide Cancellation and
Pesticide Registration Decisions. as
Measured by Gamma. ................................

Correlations Between Annual Total User
Losses from Pesticide Cancellation and
Quality of Information on Changes in Pest
Control Costs from Pesticide Cancella-
tion. as Measured by Gamma. .......................

Correlations Between Annual Total User
Losses from Pesticide Cancellation and
Quality of Information on Changes in
Value of Production from Pesticide Can-

cellation. as Measured by Gamma.

275

276

277

278

279

280

281

282

283

284

Table 865. Correlations Between Annual Total User
Losses from Pesticide Cancellation and
Quality of Information on Other Socio-
Economic Impacts of Pesticide Cancella-
tion. as Measured by Gamma. ....................... 285

Table B66. Correlations Between Annual Total User
Losses from Pesticide Cancellation and
Quality of Information on Distribution of
Impacts from Pesticide Cancellation. as
Measured by Gamma. ................................ 286

Table B67. Correlations Between Annual Total User
Losses from Pesticide Cancellation and
Quality of Information on Compliance with
Proposed Restrictions. as Measured by

Gamma. O......OOOOOOOOOOOOOOOC......OOOOOOOOOOOOOOO 287

Table 868. Correlations Between Annual Total User
Losses from Pesticide Cancellation and
Quality of Information on Pesticide
Registration Decisions. as Measured by

Gamma. ......OOOOOOOOOO......OOOOOOOOOOOOOOOO...... 288

Table B69. Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Changes in Pest Control Costs. as
Measured by Gamma. ................................ 289

Table B70. Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Changes in Value of Production. as
Measured by Gamma. ................................ 290

Table B71. Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Other Socio-Economic Impacts. as Measured
by Gamma. ......................................... 291

Table B72. Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Distribution of Impacts. as Measured by

Gamma. ......OOOOOOOOOOO......COOOIOOOOOOOOOOO0.... 292

Table B73. Correlations Between EPA's Pesticide Risk
Ranking and Quality of Information on
Compliance. as Measured by Gamma. ................. 293

Figure

Figure

Figure
Figure

Figure

Figure

Figure

Figure

Figure

Figure

1.1

LIST OF FIGURES

Idealized Timetable for the RPAR Process. ......... 14

Steps in the RPAR process and van Ravens-
waay's four stages of regulatory decision

making. ......OOOOOOOOOOOOOO......OOOOOOOOOOOOOOOOO 18

Groups and Routes of Pesticide Exposure. .......... 76
Mathematical Models to Determine Risk. ............ 79

Different Conceptualizations of the Real
Benefits of Pesticide Use. ........................ 81

Value of Increased Quantity or Quality of
Output. 0.000.000.0000.........OOOOOOOOOOOOOOOO.... 83

Value of Reduced Production Costs. ................ 83

Summary of Conceptualization of Benefits
and RiSks. ......OOOOOIOOO......OOOOOOOOOOOOOOO.... 89

Components of Pesticide Risk Assessment
and Principal Organizational Responsi-
bilities in the Office of Pesticide
Programs. ........................................ 143

Components of Pesticide Benefit Analysis
(for a given site) and Principal Organi-
zational Responsibilities in the Office
of Pesticide Programs. ............................ 171

-xvi-

CHAPTER 1
INTRODUCTION
OVERVIEW

A variety of chemicals are used in the U.S. food system
to obtain a variety of economic benefits. Pesticides are
used to improve crop quality and yields. Pharmaceuticals!
are used to improve the health of livestock. Preservatives
are used to reduce food spoilage and food poisoning risks.

While chemical technologies provide benefits. they also
pose risks to human health and the environment. Governments
have responded to this situation by enacting regulatory
statutes which set general policy on the use of these tech—
nologies. The responsibility for carrying out these laws
has been assigned to regulatory agencies. The manner in
which the laws have been carried out by the agencies has
created a great deal of controversy.

One of the major points of criticism has focused on the
way information about the tmmefits and risks of chemicals
has been collected. analyzed. and used for choice-making
within regulatory agencies. Agencies have been accused of
being less than thorough in collecting data. arbitrary in
their analysis of data. and vague about how they use infor-
mation to weigh benefits and risks. A variety of reforms.

such as cost-benefit analysis. have been proposed as

@

AV»

pa.

1.-

.n.
A...

.5-

‘4‘

M

b'u

H.\h

.‘\

'5

m

l'.

F\
nlfi

f3?”
potential remedies for the perceived problems. And the
proposed reforms have themselves become the targets of
criticism. For the most part. however. criticisms and
proposed reforms have not been based on systematic.
empirical information about regulatory decision processes.

The objective of this research is to provide systematic.
empirical information about the regulatory process of col-
lecting and analyzing data (N1 the consequences of chemical
technologies. Theories of regulatory decision—making are
used to explain the types and quality of information ob-
served in case studies of actual regulatory decisions. The
research focuses on pesticide regulation. Concentration~on
gone regulatory area rather than on several allows us to
study the use of risk-benefit analysis between cases within
a regulatory program without having to study decision making
under two or more different legislative mandates and/or
within two or more different agencies. Therefore. the
complexities of inter—agency and inter-program behavior are
avoided.

The remainder of this chapter is organized under three
subheadings. Under the first. the broad topic of pesticide
regulation is narrowed down into a statement of the research
problem emu} several research questions. thus answering the
question of what it is about pesticide regulation that is to

be investigated. The second section (us this chapter de-

scribes the research setting by providing an overview of the

-3-
Environmental Protection Agency's (EPA) regulation of pesti—
cides that are suspected of posing a risk to humans or other
organisms in the ecosystem. The final section is a brief

description of the organization of the body of the thesis.

THE RESEARCH PROBLEM

Research on regulation by economists tends to take one
of two directions: analysis of the impacts of regulation on
different groups in the economic system and analysis of the
process of regulatory decision making. Although the vast
majority of work by economists is of the former variety.
this research is concerned with the process of regulatory
decision making on pesticides. This process is essentially
one of conflict-resolution.

Why be concerned with such a complex process? Many
people see no point in examining the process if the outcome
of the process is known. However. there are at least six
compelling reasons for having a knowledge of the process.
First. in order to justify the resources devoted to public
policy analysis. it is necessary to understand the utility
of such analysis in the conflict-resolution process. Sec-
ond. knowledge (M5 the process highlights points of uncer—
tainty in the regulatory problem. which may help to focus on
relevant issues and prioritize the use of limited policy

analysis resources.l Third. desired changes in regulatory

 

lSteven Maynard-Moody and Charles C. McClintock. ”Square
Pegs in Round Holes: Program Evaluation and Organizational
Uncertainty." Policy Studies Journal 9. no. 5 (Spring 1981):
644-666. ‘

 

 

_—-:——Ib-.

 

.- .
R, in

,n

-hD-

a...

'5 .

’1

L1

. EU

I!“ i

nieTl‘.

4.

performance can be identified (assumedly through the polit-
ical process). but these changes cannot be implemented
unless the variables affecting the agency decision-making
process are known. .As George Stigler puts it. "Until the
basic logic of political life is developed. reformers will
be ill-equipped to use the state for their reforms
..."2 Fourth. a great deal of the controversy over regula—

tion focuses on the process rather than the outcome per se.

For example. regarding risk~related regulation such as pes—

ticide regulation. there is much disagreement in the scien-

tific community over how to determine risk. how to extrapo-“

late risk findings in other organisms to humans. how to

translate risk into numbers of deaths or injuries. how to U

weigh risk factors in decision making. and other questions
of process. For all types of regulation. one often hears
the argument that bureaucratic processes are inherently
inefficient and counterproductive. In order to answer these
charges. policy analysts need to know how decisions are
being made in the face of uncertainty and limited resources.
A fifth reason for investigating the process of pesticide
regulation is that knowledge of the regulatory process al-
lows us to predict. as well as explain. agency decisions.

Predicting decisions allows more accurate analysis of

 

2George Stigler. ”The Theory of Economic Regulation."

Bell Journal of Economics and Management Science 2. no. 1
(Spring 1971): I8.

-5-

program impacts. leading to more precise policy design.
Finally. knowledge of the process enables individuals or
groups affected by the regulation to have a greater impact
on the decision-making process (i.e.. to articulate their
preferences or ideas on regulatory reform).

Gaining insights about regulatory' decision making is
extremely complex due to the intricacies of human psychology
and sociology. Things are not always what they appear to
be: Majone warns us that '[RJegulators have sought legit-
imacy for their decisions by wrapping them in a cloak of

3 Samuels and Shaffer (1982)

scientific respectability.“
tell us that policymakers and interested parties often
invoke symbols. myths and ideology to convince others of
their views. Similarly. Edelman (1977) describes how
”political language” in bureaucracy is used to justify
decisions to the various conflicting groups that can impact
the agency. Language is used to shape behavior within an
agency and to evoke favorable perceptions of agency perfor-
mance outside of the agency: ”It is not facts that are

crucial. but language forms and socially cued percep-

tions."4 Regulatory decisions are not cut-and-dried choices

 

3Giandomenico Majone. ”Process and Outcome in Regulatory
Decision-Making.” American Behavioral Scientist 22. no. 5
(May/June 1979). p. 561.

4Murray Edelman. Political Language: Words that Succeed
and Policies that FailVTNew York: Academic Press. 1977). p.
85.

 

-6-

based on objective calculations of net social benefits.
Ultimately. they are political choices between conflicting
interests. In addition. various subprocesses of risk-
related decision making are characterized by what van
Ravenswaay (1983) calls a "science-policy interface." That
is. there are points in decision making and its subprocesses
where choices are based on policy considerations rather than
science. Choices can be politicized due to the nature of
the choice (e.g.. choices made by weighing benefits and
costs. or choices made under uncertainty and risk) or due to
an inability or unwillingness to expend the funds necessary
to make completely informed decisions. Whatever the reason.
there is discretion in regulatory agency decision making
which cannot be explained by scientific knowledge or legis-
lative mandates.

Adding to the complexity of the decision-making process
is the fact that a number of different subprocesses combine
to form the entire decision-making process. These subpro—
cesses or stages include problem detection. identification
of alternative solutions to the problem. identification of
consequences associated with alternatives and choice between
alternatives. These stages of decision-making do not neces-
sarily occur in the order presented. and the extent to which
each occurs varies. All four of these stages and their sub—
stages contain policy (as well as science) aspects.

It is the policy aspects of the third stage. the identi-
fication of consequences. that this research investigates.

There is an attempt to answer the following questions:

-7-

1. Which consequences of pesticide use and regulation

 

are identified? Somehow. decisions are Imuke to evaluate

 

some impacts and not others in EPA's RPAR risk-benefit
analyses.5 (hue of the objectives of this research is to
determine which impacts are evaluated and to suggest reasons
for the choice of impacts.

2. To what extent are the consequences evaluated? This

 

is the heart of the empirical work contained in this thesis.
The question can be worded in another way: How good is the
information on consequences? The question is answered by a
careful examination of EPA's "position documents" which con-
stitute one of the outputs of the pesticide regulatory
process.

3. Why are consequences evaluated to that extent?

 

There is also an attempt to explain variations in the qual-
ity and amount of information obtained for a particular
consequence (mostly variations between uses of a pesticide).
Theme is a complex array of factors that could cause this
variation. but the focus here is on the influence of ex—
ternal interest groups on EPA's information-gathering
behavior. Chapter 2 reviews the literature on factors
affecting the regulatory decision-making process.

4. Which consequences are ignored? EPA does not con-

 

sider some of the impacts of pesticide use and regulation

 

5See Chapter 3 for a discussion of the value judgments
involved in conceptualizing and choosing impacts for analy-
sis.

-8-
for budgetary and other reasons. In order to determine
which impacts are excluded. the risks and benefits of pesti-
cide use and/or regulation must be conceptualized. This
taxonomy of risks and benefits is found in Chapter 3.

There are several reasons for investigating this partic-
ular stage of decision making. The most obvious is the need
to narrow the scope of the research in some manner in order
to make it more manageable. Examining one subprocess or
stage of regulatory decision making allows a reduction in
scope without resulting in a superficial investigation of
the entire process. Second. and perhaps most important.
most of the criticisms and attempts at reform of pesticide
regulation have focused on the identification of conse-
quences. In fact. there have already been two major studies
of pesticide regulation since the regulatory reforms of

1972.6

In both studies a policy aspect of regulatory deci-
sion making on pesticides is acknowledged, but the emphasis
in both is on obtaining additional and better information on
the consequences of regulatory alternatives rather than on
understanding how such policy choices are made or how more
information affects decisions that are essentially policy

choices. Many of the reforms of the pesticide regulatory

process are attempts to ”objectively” reconcile the

 

6Environmental Studies Board Committee on Prototype
Explicit Analyses for Pesticides, Regulating Pesticides
(Washington. DC: National Research Council. 1980).

National Research Council, Pesticide Decision-Making_
(Washington. DC: National Academy of Sciences, 1977).

 

 

-9-
uncertainty inherent in risk-benefit analysis: examples
include a Scientific Advisory Panel (SAP) established to
review assessments of the risks of pesticide use. the
creation of a role for the United States Department of
Agriculture (USDA) and state experts in the benefits analy-
sis. and the opportunity for members of the general public
to comment on EPA actions at various points in the regu-
latory process. In addition. a number of proposed reforms
relate to the identification of the consequences in pesti-
cide regulation. In order to evaluate the validity of
criticisms and the effectiveness of reforms it is essential
to understand the constraints on regulatory decision mak-
ing7 which limit the extent of consequence identification.
even in the absence of uncertainty. A final reason for
studying this stage of decision making is that knowledge
about one stage of decision making is useful in understand-
ing the other stages of decision making since they are
interdependent. In the case of consequence identification.
some theories of decision making state that decisions are
(or should be) made on the basis of information on the con-
sequences of alternative courses of action. If this
describes actual decision making. then the identification of
consequences is crucial to the choice between alternative

policies. Although this is probably an overstatement of the

 

7See Chapter 2 for a theoretical discussion of these
constraints.

role of information in regulatory decision making and an
understatement of the influence of other factors. there is
undoubtedly a relationship between the two stages. as well
as between the identification of consequences and the other
two stages of regulatory decision making.

Thus. this research attempts to answer two general
questions —- how extensively are consequences identified for
each alternative. and why? The research consists of both

descriptive and explanatory work on one stage of regulatory

decision making.

THE RESEARCH SETTING: EPA REGULATION OF PESTICIDES

The above research questions are examined in the context
of the process developed by Congress and the Environmental
Protection Agency (EPA) to determine how pesticides that
pose a risk to human health and the environment should be

used. The Rebuttable Presumption Against Registration

, ......o at!
M“. .'_ . .' ,ufisz-J “n...“ru“~fi ...- v‘O-n' “avg-«.... _..',"I . Mug“ . '9.
N. .W'
W

(RPAR) process was developed by the EPA in response to the
Federal Environmental Pesticide Control Act (FEPCA). which
was a 1972 amendment to the 1947 Federal Insecticide. Fungi-
cide and Rodenticide Act (FIFRA). The RPAR process is the
major mechanism with which EPA makes regulatory decisions on
pesticides which are suspected of causing harm to humans or
other non-target organisms.

The following is a brief overview of the RPAR process.
Although the focus of the research is on the identification
of consequences of alternative regulatory actions. a general

description of the entire process is appropriate in order to

:11;

provide a setting for the research. given the inter—
dependence of the four stages of regulatory decision making.

Before 1970. while USDA still administered FIFRA. the
focus of the Act was on the registration and accurate
labeling of efficacious (as opposed to safe) chemicals. In
response to increasing knowledge and concern about the
safety of pesticides. the Act was amended in 1964 to include
chemicals which injure man. other vertebrates and other
organisms valuable to man in the definition of a "mis-
branded" chemical. The 1964 amendments also ended "protest
registration." a practice which allowed the registration of
a suspected misbranded chemical and placed the burden of
proving that a: chemical was ndsbranded (Hi the USDA. In-
stead. the registrants now had to prove that their chemicals
were safe. efficacious and/or properly branded in order to
obtain registration.

In 1970. the administration of FIFRA was transferred
from USDA to EPA. Two years later. major changes were made
in FIFRA when FEPCA was adopted. The most widely-quoted

phrase in the new FIFRA summarizes the major change in

philosophy contained in this amendment: EPA can register or.,/*

reregister only those pesticides which "... when used in
accordance with widespread and commonly recognized practice
... will not generally cause unreasonable adverse effects on

. 8
the env1ronment ..." where unreasonable adverse effects on

 

8Federal Insecticide. Fungicide and Rodenticide Act. 7
US Code Annotated. Sec. 136(a)(5)(D).

 

-12-
the environment are "[aJny unreasonable risk to man or the
environment. taking into account the economic. social and
environmental costs and benefits of the use of any pesticide
... '5) These phrases signified a breaking of ties between
FIFRA and the interests of farmers. No longer was the Act
to solve conflicts between farmers and pesticide producers
-- the arena for conflict now contained pesticide producers
and farmers (Hi the one luuui versus those exposed involun-
tarily to the health risks of pesticides on the other.
Additional amendments to FIFRA in 1975 seemed intended
to promote accurate and balanced consideration of the bene-
fits and risks of pesticide use in EPA decisions on pesti-
cide registration. First. the amendments provided a: role
for the USDA in the benefits analysis. Also. the 1975
amendments authorized the creation of a Scientific Advisory
Panel (SAP) to review the risk analysis for each pesticide.
The RPAR process itself was created by EPA regulation in
1975. Initially. the process was meant to be a mechanism
for finding problem pesticides. with administrative hearings
used to make regulatory decisions on the problem chemicals.
However. it soon became evident that EPA would like to
replace lengthy hearings with the RPAR process. which was an
informal and non-adversarial regulatory mechanism when

compared with the administrative hearings.

 

9Federal Insecticide. Fungicide and Rodenticide Act. 7
US Code Annotated. Sec. 136(bb).

 

"I- ,.

 

v'

ya

11‘)

..J

 

b
\-

 

-13...

Figure 1.1 summarizes the steps in the RPAR process in
the form of an idealized timetable. This timetable has
never been followed for any chemical so far in FEPCA's brief
history. but it does provide a general chronology of the
process.

First. there is the pre-RPAR review. during which the
EPA decides whether or not to begin the RPAR process for the
particular pesticide in question. The decision is suppos-
edly based entirely on risk. In order to make this de-
cision. EPA obtains data from registrants and from the open
literature. and can also require additional information
(e.g.. tests on toxicity of the chemical) from registrants.
From this information. EPA determines whether or not one or
more specific "triggers" for various health and environmen—

10 Supposedly. EPA considersu/’//

tal risks are met or exceeded.
exposure levels and margins-of—safety in addition to infor—
mation (N1 the toxicity (us the chemical in making the pre-
liminary risk assessment. It is generally EPA procedure not
to share much information with registrants during this phase
of the process.

Next. if EPA feels that a risk trigger has been met (as
it almost always does once pre—RPAR review has begun). a

Notice of Rebuttable Presumption Against Registration is

published in the Federal Register. often with the supporting

 

10U.S. Environmental Protection Agency. "Regulations for

the Enforcement of the Federal Insecticide. Fungicide and
Rodenticide Act." 40 Code 2E Federal Regulations. part
162.11 (July. 1983).

 

 

29.x

by

by

by

by

by

by

Sources:

+105

+180

+210
+240

+270

+300

1.

2.

-14..

Activity

Pre-RPAR Review.

EPA lets USDA and registrants know of its inten-
tions.

USDA may begin work on joint EPA/USDA/States
Benefits analysis if it feels that the EPA‘s
Notice of Presumption Against Registration will
not be rebutted.

Position Document 1 (Preliminary risk assessment)
and Notice of Presumption Against Registration
published in Federal Register.

Development of rebuttals to PDI and Notice of
Presumption Against Registration by pesticide
registrants and perhaps USDA. If rebuttals are
successful. then EPA publishes Position Document
2 (which terminates the RPAR process) in the
Federal Register.

Benefits Analysis by USDA/EPA/States. This report
is used in EPA‘s risk-benefit analysis.

EPA completes risk-benefit analysis.

Position Document 2/3 becomes available (discusses
risks. rebuttals. benefits. regulatory alterna-
tives and recommended alternative). Call for
public comment. Availability of PD 2/3 and
Preliminary Notice of Determination published in
Federal Register.

Comments due from public and from the Scientific
Advisory Panel (SAP evaluates the risk analysis
only). Comments also due from the Secretary of
Agriculture on impacts on the agricultural
economy.

Position Document 4 (analysis of comments and
final Agency decision) published in Federal
Register with Notice of Intent to Cancel.

Meeting with Dr. Fred Tschirley on April 15. 1981.

Environmental Studies Board Committee on Prototype
Explicit Analyses for Pesticides. Regulatigg Pesti-

 

cides. (Washington. DC: National Research Council.

-15-

3. U.S. Environmental Protection Agency Office of
Pesticide Programs. ”Status Report on Rebuttable
Presumption Against Registration (RPAR) or Special
Review Chemicals. Registration Standards Program.
and Data Call-In Program“ (Washington. D.C.: U.S.
Environmental Protection Agency. March 1984).

Figure 1.1 - Idealized Timetable for the RPAR Process

document. Position Document One. which is a preliminary risk
assessment of the pesticide. This Notice of RPAR and Posi-
tion Document One are subject to rebuttal attempts by
registrants and other interested parties. The official time
period allowed for rebuttal responses is 45 days from pub-
lication of the Notice of RPAR. but EPA generally grants 60-
day extensions. Rebuttal can be accomplished by proving
that the pesticide doesn't meet the trigger(s) or by showing
that exposure is low enough so that risk is not great. In
addition. EPA also seems to consider rebuttal comments that
show that the benefits of pesticide use are so great that
the risk is worth it. Rebuttals are rarely successful.

The next step. if the notice of RPAR is successfully
rebutted. is the issuance of a Position Document 2 explain-
ing why PDl was rebutted and the return of the pesticide in
question to the registration process. If the rebuttal

attempts are unsuccessful. then a risk-benefit analysis of

 

. l

b l

-16-

the status—quo use of the pesticide is developed. alterna—
tive solutions t1) risk reduction are generated. the risk-
benefit analysis is extrapolated and embellished upon for
each alternative solution. and a preliminary decision is
made between the alternative solutions (generally for each
use of the pesticide). All of this. as well as the chemical
background and regulatory history of the pesticide. and an
analysis of the rebuttal comments. is contained in Position
Document 2/3. which supports the Preliminary Notice of
Determination published in the Federal Register.

There is opportunity for external review of and comment
on the Position Document 2/3 and the proposed regulatory
action. Interested parties have 30 days to submit comments.
but sometimes extensions are granted by EPA. In addition.
the risk-benefit analysis and proposed regulatory decision
are reviewed by the Secretary of Agriculture and the
Scientific Advisory Panel (SAP). These two parties have 60
days from the publication of the Preliminary Notice of
Determination in the Federal Register to respond. If their
comments are received by EPA within 30 days. they must be
published in the Federal Register with EPA's reply and the
Final Notice of Determination.

The Final Notice of Determination is published in the
Federal Register. often with its supporting Position
Document 4n Position Document 4 replies to comments from
the Secretary of Agriculture. the SAP and others. and ex-

plains the rationale for the final Agency decision. Omce

-17-
this document is published. cancellation activities can

begin.11

After this step. the RPAR process is complete and
the only recourse for interested parties is administrative
hearings.

The details of the process are still very much in a
developmental stage. as can be evidenced by reforms proposed
by EPA in the August 7. 1980. Federal Register and more
recent proposals to attempt to negotiate solutions to risk
problems with pesticide manufacturers instead of undertaking
the benefits analysis.

Figure 2 shows how the steps of the RPAR process corre-
spond with the four stages of decision making. There are
some additional aspects of the identification of conse—
quences stage which should be noted since that stage is the
focus of the research. The EPA obtains much of its infor-
mation from parties outside of the agency. such as the USDA.
state experts and the registrants. Many of these external
parties have a vested interest in the outcome of the regula-
tory process. and they may have an influence on the process
due to their control over information. Two different divi—
sions of the EPA's Office of Pesticide Programs collect and
analyze data for the risk and benefit assessments. The risk
analysis is performed by the Hazard Evaluation Division and

the benefits analysis is performed by the Benefits and Field

 

11"Conditional Cancellation“ means that a particular use
of the chemical will be banned if changes in labeling and/or
use practices are not made. "Unconditional Cancellation"
means that the chemical is banned.

Steps in the RPAR Process

 

Pre-RPAR Review

Notice of Presumption
Against Registration
and Position Document 1

Rebuttal attempts and
public comments

If successful rebuttal.
Position Document 2 and
return to registration

USDA/EPA/States Benefits
Analysis and EPA
risk-benefit analysis

Position Document 2/3 and
Preliminary Notice of
Determination

Public Comments

Position Document 4

 

Source:

-13-

Stages of Decision Making

 

Problem Detection: Identifi-
cation of Consequences

Problem Detection: Identifi-
cation of Consequences

Problem Detection: Identifi-
fication of Consequences

Choice Between Alternatives
(alternatives = return to
registration and initiate
RPAR Process)

Identification of Consequences

Generation of Alternative
Solutions: Identification of
Consequences: Choice Between
Alternatives

Identification of Consequences:
perhaps Problem Detection and
Generation of Alternative
Solutions

Choice Between Alternatives

Compiled by the author.

Figure 1.2 — Steps in the RPAR process and van Ravenswaay's
four stages of regulatory decision making.

 

-19-
Studies Division (in conjunction with USDA and State ex-
perts). A third division. the Special Pesticide Review
Division (SPRD). has overall responsibility for coordinating
”fl-ww4m-Mwm.

implementation of the RPAR process and also supervises the
weighing of risks and benefits and recommends a regulatory
Option tx> the Deputy Assistant Administrator for Pesticide
Programs (who ultimately answers to the Administrator of the
EPA). The common thread throughout the risk-benefit analy-
sis for a particular pesticide is the project manager. who
is from the SPRD. However. the final decision is technic-
ally the responsibility of the Administrator: thus. the
situation is one in which information on the consequences of
alternative regulatory options is not generated by the final
decision maker.

One of the outputs of the RPAR process. aside from the
regulatory decision. is time set of position documents for
each pesticide which goes through the process. These doc-
uments provide information supporting various actions taken
by the Agency. from the initial rebuttable presumption
through the final decision. These position documents con-
tain information on the consequences of full use of a pesti-
cide and of alternative regulatory actions. Thus. the
position documents for 13K; eight pesticides for which the
RPAR process had been completed at the commencement of the
empirical work serve as the data base for the work on conse-
quence identification. One limitation of the use of the

position documents as the data base is that all consequences

-20-
which are considered may not be in the position documents.
However. it is still of value (as discussed in the previous
subsection) to know which consequences are identified. why.
and whether or not the consequences identified appear to
have any bearing on the choice between regulatory alterna-

tives.

ORGANIZATION OF THE THESIS

The remainder of this thesis is organized as follows:

Chapter 2 consists of a review of the theoretical liter-
ature on regulatory decision-making processes. Some of the
works examined in this chapter serve as the basis for the
explanatory variables conceptualized and operationalized to
explain EPA's identification of consequences in pesticide
decision making.

Chapter(g:is a conceptualization of the risks and bene-
fits of pesticide use. The intent of this chapter is to
provide a comprehensive taxonomy of risks and benefits to
serve as a basis of comparison with EPA's taxonomy of risks
and benefits.

Chapter 4 describes the methodology used in coding data
and testing relationships about EPA identification of conse—
quences in the RPAR process. The data set consists of the
position documents for eight pesticides for which RPARs have
been completed.

Chapteq:§)is the first empirical chapter. consisting of
an analysis of the descriptive statistics and explanatory

work on the risks of pesticide use.

-21-
Chapter 6 presents the results of empirical work on the
benefits of pesticide use.
Chapterflcontains results. conclusions and directions

for future research.

CHAPTER 2

THEORIES OF THE REGULATORY PROCESS:

A REVIEW OF THE LITERATURE

INTRODUCTION

There is an immense body of literature from many differ-
ent disciplines on various aspects of regulatory agency de-
cision making. The objective of this chapter is to organize
and summarize some of this diverse literature to help in
describing. explaining and predicting the EPA's behavior
with regard to pesticide regulation. The literature review
is organized under four subheadings: the nature of regula-
tion. theories of regulatory decision making. the role of
information in decision making. and the usefulness of the
theories in explaining the quality of information obtained
by EPA to evaluate consequences of pesticide use or regula-

tion.12

THE NATURE OF REGULATION

Several authors have explored the nature of regulation

in general. as mandated and implemented. rather than details

 

12Other reviews of the literature on regulation are con-

tained in Noll (1976). Posner (1974). McCraw (1975). Fiorino
and Metlay (1977). Owen and Braeutigam (1978). Mueller
(1979) and Mitnick (1980).

-22-

56'

CO

OJ
C)

be

-23-
of regulatory decision making. The readings discussed below
do not constitute a random sample of this literature --
three of the four works are written by institutional econo-
mists. However. the articles suffice in presenting a pic-
ture of regulation which is not commonly considered in the
literature of neoclassical welfare economics. Regulation is
seen as the outcome of a power struggle for rights between
competing interests. the product of the value system of some
dominant class. or a response to an inappropriate balance
between equity and efficiency. rather than as a mechanism
for achieving greater economic efficiency within the status
quo system of rights. This literature provides a view of
regulation which supports the work of many of the theorists
in the next section. who concern themselves more with the
political realities of regulatory decision making and its
attendant impacts than with unconstrained decision making
with the goal of economic efficiency.

Paul Weaver (1978). a non-institutionalist (and also a
non-economist). feels that the nature of regulation has
changed over time. but that it is and always has been the
manifestation of the values of a particular dominant class.
The 'old" decision making (e.g.. airlines. public utili—
ties). characterized by the “iron triangle" -- a coalition
of the regulated industry. the regulatory agency or commis-
sion. and the Congressional subcommittees responsible for
legislative hearings and oversight -- was a reflection of

the values of the liberal democratic class which dominated

(*

£1)

"h

'V

r“

-24-
national thought at the time of the enactment of the enabl-
ing legislation. The "new" decision making. which is not
explained by the capture theories of regulation or by most
empirical work. is a result of the humanistic values charac-
teristic of a new dominant class. This regulation includes
much of the health and safety and income maintenance regula-
tion. and Weaver claims that the new iron triangle is com-
prised of the press. public interest groups and the Federal
government.
Weaver sees regulation as social policy rather than
economic policy. Indeed.
That is why all economists. whatever their
political views. end up being so critical of
government regulation. at least as it works
out in practice. They think regulatory policy

should make sense economicall31y -- which. of
course. it never quite does.

Weaver is wrong. Although many economists see the
purpose of regulation as correcting "market imperfections"
ix! an economically efficient manner. and judge regulation
accordingly. not all do. Among those who do not are the
following institutional economists. who attempt to describe
the regulatory process in a positive manner in terms of the
relationship between institutions (common and statutory law.
regulation. customs. morals. et cetera) and economics,

Reynolds (1981) criticizes some of his fellow economists

for judging regulations which are a response to efficiency

 

13Paul Weaver. "Regulation. Social Policy. and Class

Conflict." The Public Interest 50 (Winter 1978): 56.

 

 

r'

(ll

CT

4

V‘

(I)

'1

III

LA)

-25-
and/or equity problems in terms of efficiency alone. Like
Weaver. he notices variations in regulation depending on the
time of the enactment of the enabling legislation. The old
regulation (basically pre-l970) is industry-specific and
attempts to achieve the desired allocational or distribu-
tional goals by augmenting revenues of the regulated group.
Such regulation is typically demanded (and captured) by
firms. industry groups (n: trade associations. Function-
specific regulation (post-1969) is seen by Reynolds as often
being a response to the industry-specific regulation's
impacts on equity and/or efficiency. However. this type of
regulation affects costs rather than revenues of the firms
or other regulated groups.

When describing the industry-specific and function-
specific regulation. Reynolds is describing two types of
explicit regulation. Explicit regulation consisms of the
implementation and enforcement of statutory and common law
chosen in the political process. whereas implicit regulation
refers to informal "laws" such as habits. norms. ethics and
values.

The entire system of regulation in Reynolds' world is
affected by and affects technology. The system needs to be
a combination of implicit and explicit regulation which
exhibits some degree of companionship with the technological
and physical world. If this compatibility does not exist.
then the regulatory system will change. Compatibility is

essentially an appropriate (socially acceptable) balance

V“

T‘

...

U]

(”f

(D

(T

’1

T‘

U)

’k!

—26-
between equity and efficiency and an adequate level of
flexibility. It is a function of the legislation. sup-
porting institutions. and "escapements" in the regulation
(mechanisms which result in distorted perceptions of the
relationship between compliance with and impacts or costs of
the regulation). Sudden perceptions of market failure are
really responses to social and ethical changes. and/or the
loss of escapements for existing regulation.

Changes in the regulatory system comprise a life cycle.
In general. Reynolds sees implicit regulation evolving over
the long run. If the implicit regulation fails to produce
socially acceptable allocative and distributive impacts.
then revenue-augmenting (industry-specific) regulation is
established.14 In response to new allocations an distri-
butions. other groups may attempt to obtain their own
revenue-augmenting regulation to offset costs of the
revenue-augmenting regulation of others. Function-specific
regulation is often a response to perceived inadequacies of
industry-specific regulations. Finally. the current popular
support for "deregulation" is aa response to changes which
have made the function-specific regulation unacceptable.
Reynolds' regulatory system is dynamic and it determines and
is determined by technological progress.

Samuels. Schmid. and Shaffer (1981) also see the regula-

tory system as dynamic and as a dependent and explanatory

 

14Explicit regulation usually overrules implicit regula-

tion.

-2'7-

variable in larger systems. in) these economists. the cen-
tral issue in regulatory decision making is not which form
of regulation is 'best' or the efficiency of regulation. but
rather "whose interests are to be promoted by regulation and
how...,"15 for regulation restricts the opportunity set of
some individuals while expanding or improving opportunities
for others. As implemented. regulation (and changes in reg-
ulation) are rights in that both regulations and rights
distribute costs, benefits and risk among interdependent
parties.

In a broader context. regulation both exists in and
determines a power structure. It is only one of many forms
of social control which are intended to achieve a balance
between freedom and control. hierarchy and equality. and
continuity and change. Various groups within a power struc-
ture try to use government to achieve regulation in their
favor. Both "public" and “private" regulations free some
individuals at the expense of others (albeit different
individuals); they are merely alternative ways of achieving
social control. According to the authors. ”Arguments over

regulation concern the pattern of freedom versus exposure to

 

15Warren J. Samuels. A. Allan Schmid and James D.
Shaffer. “Regulation and Regulatory Reform: Some Funda-
mental Conceptions' in Law and Economics: An Institutional
Perspective. eds. Warren J. Samuels and A. Allan Schmid
(Boston: Martinus Nijhoff Publishing. 1981). p. 248.

 

 

-28-

the freedoms of others .... made by power-playing actors
participating in the drama of working out social con-
trol."16

Samuels and Shaffer (1982) continue this theme in their
paper on deregulation. In this paper. they attempt to show
that eight arguments commonly put forth in favor of deregu-
lation are normative arguments and are inconclusive. Some
of the major points in the paper are that deregulation and
regulation are funtionally the same in that both protect
rights (different people's rights). deregulation represents
a restructuring (n3 rights. only selective perception makes
deregulation distinct from rights. to say that regulation
and/or deregulation are good or bad requires normative judg-
ments. both regulation and deregulation require coercion of
those whose opportunity sets are restricted. and definitions
of output (which are determined by regulation and deregula-
tion) determine whether or not deregulation promotes produc-
tivity and efficiency along with the set of rights used to
determine the efficient or Pareto-optimal allocation of
resources (which is also determined by regulation and dereg-
ulation).

All of these writings on the nature of regulation are
based on the general theme that regulatory decisions are
political as well as scientific. Decisions do not simply

fall out of the analysis of benefit and cost information.

nor does the regulatory process revolve around the gathering

 

l6IbidoI p. 255.

-29-
of objective information. This has strong implications for
decision making in general and for the identification of
consequences of regulatory decisions in particular. The
next section of this chapter is an attempt to move from the
general to the specific by summarizing different theories on
the factors which influence the regulatory process and its

output.

THEORIES OF REGULATORY DECISION MAKING

Regulation theorists need to address two major questions
if they are to provide information on the regulatory
decision-making process: what are the goals (objective
functions) of regulators. and what factors constrain the
achievement of those goals? The theorists answer these
questions 511 different ways. In order ix) facilitate the
discussion of this diverse group of theories. the works are
organized according to the hypothesized objective functions
of the regulators. The theories can be classified under the
subheadings of theories with .agency-wide objective func-
tions. theories with different objective functions for
various coalitions within an agency and theories with dif-
ferent objective functions for individual regulators in the
agency.

Theories With Agency-Wide Objective Functions. Some

 

theorists assume that agency goals dominate the behavior of
individual regulators within the agency. This behavioral
assumption has some validity. and perhaps good explanatory

power since agency leaders (as well as managers in the

-30-
private sector) have mechanisms to encourage individuals to
adopt agency goals. Also. such assumptions simplify the
study of agency decision making by ignoring individual
aberrations from agency goals.

Agency survival. Several theorists assert that

 

there is a link between voter behavior and agency behavior.
That link is the elected politician whose fate is determined
by the voter and who. in turn. determines the fate of the
agency. Legislators and other elected officials have power
over agency officials via appropriations. program authoriza-
tion. appointments and legislative oversight (see Thurber
(1976)).17 Thus. the agency officials try to make decisions
which improve the number of votes. size of the campaign
chest. and perhaps personal wealth of the politician.

One of the more interesting theories of this type was
formulated by one University of Chicago professor and ex-
tended and formalized by another. Nobel prize-winner George
Stigler (1971) is really explaining legislative creation in
his work cut economic regulation. but Sam Peltzman extends
Stigler's work to agency decision making by assuming that
voters will express dissatisfaction with agency decisions by
voting against the elected politicians who appointed the

agency officials. Voters and regulatory agencies are linked

 

17Thurber also hypothesizes that the agency-elected
official relationship may be one of cooperation rather than
confrontation.

-31-
because agency officials require elected politicians'
support in order to perpetuate their activities. It is con-
ceivable that this hypothesis applies to the EPA since the
environmental lobby influences a large number of voters and
is politically sophisticated. the environmental groups are
often opposed by industry groups that are also well organ—
ized and powerful. and legislators are often rather clearly
for or against government programs for environmental pro-
tection.

Stigler suggests that regulathmi is often obtained by
(and for) industry due to the nature of the "market" for
regulation. The regulatory market is different from the
economic market in that the output must be adhered to (in
principle. anyway) by everyone. If all regulatory decisions
were made democratically (i.e.. by voting). then policies
which resulted in more gainers than losers would be adopted.
assuming perfectly informed and rational voting.

Why. then. do regulations often seem to benefit fewer
people than they hurt? The answer has to do with the nature
of the market for regulation. People do not vote on all
issues: instead they elect representatives to vote on
issues. These representatives and their political parties
(and. if extended one step further. the agencies) are the
suppliers of regulation. Regulation is demanded by industry
and other groups in Stigler’s model because it results in
the redistribution of wealth and income. and it is paid for

with votes and dollars for the politician. Thus. in order

-32-
to pay for regulation with the required number of votes
and/or dollars. voters must incur the costs of informing
themselves on an issue. organizing to articulate their
preferences and persuading others to support (or not oppose)

18 They will only

the politician supplying the regulation.
incur these costs if the benefits from obtaining the favor-
able regulation exceed the costs. As the size of the group
demanding regulation increases. the per capita benefits
decrease and the costs of organization and persuading other
voters increase. The free-rider problem also increases with
size -- firms may expect other firms within the industry to
pay for beneficial regulation from which no already-existing
firm can be excluded. Since costs increase and benefits de-
crease with size. smaller groups with an economic incentive
to precisely articulate their demand (such as industry
groups) may be the only demanders of regulation according to
Stigler.

Peltzman (1972) formalizes and generalizes Stigler's
theory. He explicitly establishes the voter-agency link by
making note of the power of the President and Congress to
appoint. agency' officials euui the resulting potential for
these elected officials tx> be held accountable for agency
performance. The goal of regulators is to maximize votes

and/or resources for elected representatives in order to

 

18Bartlett (1973) develops a framework which shows how
groups may subsidize information on the outcomes of policy
to other groups in order to influence perceptions of the
policy.

-33-
assure agency survival. Resources are assumed by Peltzman
to be used by politicians to diffuse opposition to regula—
tion.

Peltzman's mathematical model solves for the vote-
maximizing sizes of the groups to be benefited and taxed as
well as the vote-maximizing distribution of benefits and
costs among groups. The variables which explain the regu-
lators' decisions on distribution of benefits and costs
include the wealth of the winning and losing groups. the
groups' responses (in terms of votes and resources) to
taxation or benefits. and the sizes of the winning and
losing groups. Thus. in order to maximize votes for the
politician. the agency decision must reflect each group's
relative power to affect the politician and. ultimately. the
agency. If there is no opposition to a policy and if all
beneficiaries will vote for the politician. then size be-
comes the dominant factor. However. if there is opposition
anui if votes are difficult to deliver. beneficiaries must
expend resources to obtain votes (i.e.. organize and articu-
late preferences) and mitigate opposition. so wealth of the
groups becomes important.

The explanatory variables in the decision on the sizes
of benefiting and losing groups include the amount of sup-
port for the regulation. the amount of opposition. and the
costs of organization facing the different groups.

Several general conclusions can be drawn from Peltzman's

model. First. even if one group obtains all of the benefits

-34-
of the regulation. the groups would have obtained more
benefits if a cartel had been formed without help from the
government. The reason for this phenomenon in Peltzman's
model is that regulators must account for the political
power of opposing groups in order to maximize votes: the
vote maximizing distribution of benefits and costs of regu-
lation is determined by equating marginal political costs
(to the regulator) and marginal political benefits. Second.
Peltzman concludes from his model that there may be more
than one group of beneficiaries if regulators can supply
different levels of benefits to individual voters according
to their sensitivity to benefits or taxes. In fact. tax-
sensitive members of the losing group may actually benefit
from the regulation while insensitive members of the winning
group may be taxed. Stigler's theory. which proposes that
the regulated industry is the only winning group. is a
special case of Peltzman's general theory.

Other theories hypothesizing agency-wide goals of vote-
maximization are very similar to the Peltzman-Stigler
theories. although not as precisely stated.

Owen and Braeutigam also assume crucial links between
voters. Congress and regulatory agencies. According to this
theory. voters wish to obtain regulation to protect their
wealth in the face of technological and economic change
(Stigler and Peltzman. in focusing on economic regulation.
see regulation as a means of increasing wealth). In other

words. voters wish to take some of the risk out of the

-35-
marketplace. This desire is communicated to the agencies by
Congress and voters via administrative law. which slows
change. allows all interested parties to voice their con-
cerns. and protects the status quo.19

Therefore. if regulatory agencies want Congressional
support. they must protect voters from sudden changes in
income or wealth. Administrative law helps to ensure this
protection. This result is essentially equivalent to
Peltzman's result that the winners in the regulatory process
cannot receive all potential gains due to the necessity for
regulators to account for conflicting interests. Owen and
Braeutigam weaken the link between the agency and the voter
by suggesting that other theories of regulation may be
usefully interjected into their framework to explain discre-
tionary agency behavior.

John Baldwin's (1975) theory likewise makes a connection
between the agency. elected officials and voters. Bureau-
crats are assumed to be self-interested with a dominant goal
of agency perpetuation. which is accomplished by helping to
ensure the re-election of the incumbent government. Since
regulation imposes costs on some groups. the agency's role
is to establish an agreement between conflicting interests.
Regulators thus attempt to choose the regulatory alternative

with costs that can be disguised from the losers (resulting

 

19This is an example of ”voice" as described in Hirsch-
man (1970).

-36-
in fewer lost votes to the incumbent government). This con-
clusion is consistent with the Peltzman and Stigler models.
which predict decisions resulting in diffuse costs and
concentrated benefits. Baldwin acknowledges the existence
of agency behavior independent of the incumbent government
when he suggests that some agency support may come from
outside of the government.

A goal which is related to vote-maximization is that of
conflict-minimization. The conflict-minimization theorists
take a somewhat broader view of external pressures on agen-
cies than (k) the vote-maximization theorists. and conflict
minimization may imply efforts to avoid opposition rather
than to seek support. In fact. efforts to disguise costs as
suggested by Baldwin may actually point more toward conflict
minimization than vote maximization. However. the general
idea is still the same. The agency is seen as being subject
to pressures from the courts. the media. other agencies and
interest groups. as well as from elected officials. In
order to maintain the agency. regulators attempt to minimize
conflict from these sources.

Paul Joskow (1974) points out that agencies are gener—
ally given a good deal of flexibility in decision making
with regard to the intent of a mandate. the method of
achieving the intent and procedures for implementing the
method. Constraints of due process from the courts and the
legislature limit this discretion somewhat. Regulatory

behavior is also constrained by the agency environment.

-37-
which consists of interactions between the agency and groups
affected by regulatory decisions. Regulators experience
pressures from other participants in the regulatory process
-- consumer groups may act as "intervenors" during the
regulatory process or exert pressure on the agency outside
of the formal process. and regulated firms are also in
contact with the agency during the formal process as well as
on an informal basis.

According to Joskow. an agency can be in equilibrium
with its environment if procedures and regulatory instru-
ments have been developed to balance the conflicting
interests. or the agency can tmezhi the "innovation mode."
during which a search for procedures and instruments to
restore equilibrium occurs. Factors which create an impetus
for change in the regulatory process include economic
factors (e.g.. industry changes. inflation) and political
factors (e.g.. the environmental movement. civil rights
movement). These changes modify the relative power and the
nature of the various interests in the regulatory process.
Thus. the theory is similar to Peltzman's in terms of ex-
planatory variables.

Richard Posner (1971) implicitly assumes some sort of
conflict-minimization or vote-maximization goal for agencies
when he describes the rationale for the cross-subsidization
forfll of ‘regulation. Internalr- or’ cross-subsidization. is
regulation which results in the production of goods or ser-

vices at lower prices or in larger amounts than what would

_38-
have been produced in the absence of regulation. The result
is that some people pay (are taxed) so that others can
obtain goods at subsidized prices.

Posner hypothesizes that cross—subsidization is popular
with regulators as an instrument for redistributive policy
due to its low visibility relative to direct taxes (again.
the Stigler-Peltzman theme of dispersed costs). the diffi-
culty of judicial review. relatively low administrative
costs and implementation without need for Congressional
approval. This type of regulation scores high marks with
regulated firms because it often results in entry control.
and comprehensive enforcement of prices and quantities of
goods is difficult. In Posner's theory. regulators appear
to make regulatory decisions which balance conflicting
interests in response to environmental pressures.

All of the preceding theories contain some common ele—
ments. All. of course. retain the agency goal of survival
via vote-maximization or conflict-minimization. In addi-
tion. they all suggest that constraints on goal attainment
include pmessures 1J1 the agency's external environment —-
namely. pressures from conflicting interests. courts. legis-
lators and/or the executive branch. Thus. if these theories
represent a realistic abstraction of regulatory agency
behavior. relevant explanatory variables of regulatory
decision making would include the power of groups in the

regulatory process to impact the agency.

 

.vu
Cu

in.

St}

-3g-

Budget-maximizing, or size-maximizing, agencies.

 

Another important set of theories with agency-wide goals is
based on the assumption that regulators want to maximize
agency size. budget or influence. There are two rationales
for assuming this goal: some theorists feel that agency
survival is assured not by maximizing votes for legislators
or balancing conflicting interests. but rather by more
actively trying to increase agency size over time: and other
theorists believe that agency size and/or scope is tied to
goals of individual regulators. such as salary. prestige.
power or job security. These theories do not necessarily
invalidate the theories described in the previous section:
instead. they may better explain regulatory decision making
at certain points in an agency's history. For example. it
may be that the EPA was a "growth" agency during much of the
19703 when there was a great demand for environmental pro-
tection. but that it is currently a "conflict-minimization"
or "survival" agency in the face of recession.

Warren Samuels (1973) believes that regulators seek to
increase agency authority or power as a means of achieving
their own individual goals. such as maximization of income.
status or power. The main constraint in his model is the
power of others to affect regulators' decisions. However.
the power structure not only determines decisions but is
also determined by previous decisions.

Thus. the system of regulatory decision making is a

struggle between interested parties to obtain favorable

-4Q-

legislation or regulation. .A group's strategy is often to
attempt to make its particular goals the goals of the
agency:

Rational decision—making would seem to
require well-defined. precise. agreed—
upon objectives and goals of regulation.
and the absence of same in regulation has
been widely lamented. But a heterogen-
eous institution leads ix) ambiguous
statements of purpose and goals because
of the continuing jockeying for position
as parties-in-interest compete to have
their particular goals become the goals
of the institution and to revise the
statutory goals. the intermediate ends
and the working rules in their interests.
With multiple. competing goals and
criteria advanced by competing parties-
in-interest. clear euui consistent goals
or solutions are not permanently possi-
ble. as all are subject to continual
revision inzfipe corner or another of the
institution.

In furthering its own goals. then. the agency must be re-
sponsive to the power and goals of conflicting interests.
Other variables in Samuels' general equilibrium system
which affect and are affected by decision making include
working rules (institutions which control the distribution
and use of power). the system of values (regulatory impacts
may be capitalized into asset values while those values are
used in making decisions). the structure of opportunity sets
of various interested groups. ideology. preferences. indi—

vidual choice. resource allocation and distribution of

 

20Warren J. Samuels. "Public Utilities and the Theory of

Power" in Perspectives in Public Regulation: Essays on
Political Economy. ed. Milton Russell (Carbondale. I11:
Southern Illinois University Press. 1973). p. 10.

 

 

-41-
income and wealth. These variables. in defining the power
structure. help to explain regulatory decisions.

There are similarities between Samuels' approach and
that of Peltzman. Both theorists stress the importance of
economic and political power and the ability of groups to
organize. given the structure of benefits and costs of
action. Both recognize the advantage of regulated indus-
tries in the power struggle. But Samuels' theory is a much
more comprehensive conceptualization of regulatory decision
making. Peltzman. by his own admission. does not attempt to
describe the determinants of the power which he suggests
explains regulatory decision making. Samuels. on the other
hand. provides a general equilibrium model of the determi-
nants of power as well as its impacts on regulatory decision
making.

William Niskanen (1975) develops a theory of the supply
of government output based on two major assumptions: agen—
cies strive to maximize their budgets. and both suppliers
(agencies) and demanders (government review groups) of
regulation are monopolistic. The latter assumption provides
a bilateral monopoly situation in the market for regulation
and establishes the agency-elected official link favored by
so many other theorists. However. Niskanen does not feel
that the link is so complete that the goals of the elected
officials are directly transferred to the agency. Instead.
the power of elected officials serves as a constraint on

agency' attainment (M5 the budget-maximization. goal. along

-42-
with other factors which determine the reward structures of
bureaucrats. Niskanen's major conclusion is that monopolis-
tic agencies overproduce and overspend relative to the
demand for the good or service produced by regulation when
the agency goal is budget maximization and when a monopolis-
tic review committee is the demander of regulation.

Shapiro and Shelton (1977) suggest that agency size is
related to personal goals of regulators. particularly
salary. so they assume that the agency-wide goal is to in-
crease agency size. Agency discretion in the attainment of
this goal is constrained by relationships with other groups
in the regulatory process. such as legislators. taxpayers.
the executive branch and beneficiaries of regulation.
According to this theory. agency officials attempt to make
decisions which garner support from beneficiaries of the
decision. facilitate re-electhmi of elected officials and
disperse costs of the regulation so that taxpayers perceive
the costs as being lower than the costs of Opposing the
regulatory decision. These are tflue same factors that are
emphasized in the theories of Stigler and Peltzman.

Downing and Brady (1979) also assume that bureaucrats
are interested in increasing agency responsibility in order
to increase their own income and power and to further speci-
fic agency goals. They suggest that politicians have goals
of re-election. upward mobility and the furthering of some
concept of the public interest: firm managers try to in—

crease their own real incomes by increasing firm profits and

-43-
stock values: and consumer group leaders desire changes in
property rights which reflect their ideas of equity.

Choices by these groups are constrained by income and by
the costs of action. Rational behavior involves bargaining
with other groups for votes. contributions and/or resource
transfers. Trades affecting the benefits and costs of
policy decisions are made between groups in order to mutu—
ally improve their situations. There also may be attempts
to change the costs of action faced by groups favoring or
opposing policies which affect a particular group's welfare.

McKenzie and Macaulay (1980) assume a primary agency
goal of increasing size and domain. Regulators attain this
goal kn: making decisions which decrease efficiency in the
private sector in order to make the public sector look
relatively efficient. For example. monopolization of the
private sector raises the price of privately-produced goods
relative to public sector substitutes. Constraints on this
type of behavior by the agency are not discussed in the
article. The implication is that costs of action to one
group are related to the power of other groups in the reg-
ulatory process.

Summary -- theories with agencijide objective

 

functions. The major contribution of the preceding theories

 

to this research is to develop the concept of interest group
power as an explanatory variable for regulatory decision
making. The Peltzman model is particularly precise in its

description of how various characteristics of groups in the

-44-
regulatory process affect the nature of the regulatory
process. and thus proves very useful in hypothesizing rela-
tionships between explanatory and dependent variables.

Theories Assuming Different Goals Within an Agency3

 

Another group of theorists hypothesizes that individual reg-
ulators or coalitions of regulators have unique. multiple-
argument objective functions. These theorists believe that
attempts to aggregate individual behavior on the basis of
some dominant goal do not result in accurate explanations of
decision making. However. this argument is not adopted in
this research. Individual or coalition goals which vary
from an agency-wide goal may account for unexplained vari—
ations between observed agency behavior and behavior
predicted by one of the theories in the previous section.
but it is assumed here that good explanatory power is ob-
tained from models with agency-wide goals (especially when
considering the relative simplicity of these models). for
many of the same reasons that firm—wide objective functions
have proven useful in the theory of the firm. Furthermore.
there is no reason why individual goals couldn't be incor-
porated into one of the theories with agency-wide goals to
improve its explanatory power in future research -- the two
views are not necessarily mutually exclusive. Since the
choice in this research is to place more emphasis on agency-
wide objective functions. the following theories will be

described with more brevity than the preceding theories.

-45-

Coalitions of regulators. Wilson and Downs are the

 

major proponents of a group of theories in which individual
regulators are assumed to have objective functions which are
similar enough for the regulators to be categorized into
discrete groups of individuals with common goals.

In James Q. Wilson's theory (1980). there are three

types of bureaucrats. Careerists look to agency success for

 

their own personal success. Their primary goal is to main-
tain the agency. which is accomplished by cultivating

21 An example of

political support and avoiding scandals.
the latter strategy appears in Wilson's book: ”In regula-
ting pesticides. EPA is keenly aware that if a product it
has registered is later shown to produce cancer on a large
scale. the agency will be crucified and the careers of all

"22 Careerists in

concerned blighted. if not destroyed.
agencies learn through reinforcement to accommodate people
who will use errors against them.

Politicians in an agency have long-term goals of upward

 

mobility outside of the agency which will result in finan-
cial rewards and desirable career patterns. Therefore. they
want to be as visible as possible to potential future em-

ployers.

 

2J'Careerists appear to be the pure form of bureaucrats
described in the theories with agency-wide objective func-
tions.

 

2James Q. Wilson. The Politics of Regulation (New
York: Basic Books. Inc.. 1980). p. 375.

 

-46-

Finally. there are _professionals. in ‘Wilson's typical

 

agency. who obtain utility from others in the same profes-
sion. Individuals in the same professions have similar ways
of looking at the world. so they are likely to form coali-
tions within an agency. When there are individuals from
several different professions in an agency. such as in the
EPA. decision making may be greatly influenced by the pro-
fession which dominates.

The behavior of the three types of regulators is influ-
enced by certain key characteristics of the environment
within which regulatory decisions are made. Wilson believes
that technological enui economic change. institutions. and
politics and ideas (e.g.. access to the gmflitical system.
the media and friendly legislators) are three important
characteristics of the regulatory environment.

Anthony Downs (1965) also delineates several different
groups of regulators according to their goals. Climbers and

Conservers behave only according to personal goals. Climb-

 

ers attempt to maximize their own power. income and prestige
by obtaining promotions. movimg to better jobs outside of
the agency or increasing the prestige and power of their
current positions. Conservers. on the other hand. maximize
security and convenience. and thus are very much interested
in protecting the status quo. Other regulators pursue both
personal and collective goals: mixed-motive officials in-

clude zealots. advocates and statesmen. Zealots devote

 

 

-47-
their collective energies ix) narrow issues. whereas advo-
cates fight for particular organizations or policy areas.
Statesmen support the broadest collective group -- they
maintain goals relevant to society as a whole in their
objective functions.

In considering variables affecting agency decision
making. Downs feels that the internal structural variables
of an organization cannot be considered separately from
variables describing the external environment within which
an agency operates. Some of the internal and external
variables contained in his hypotheses on agency behavior
include information costs. capability of decision makers.
degree of uncertainty. degree of information distortion and
conformity of lower-level bureaucracy to agency goals (both
affected by the degree of hierarchy in the agency). time
constraints. bias of decision makers. degree of coordination
between agencies with overlapping duties. degree of goal
homogeneity between individual regulators. the age (M5 the
agency and the agency's rate of growth. Thus. decision
making is influenced by the internal structure of the
agency. which determines and is determined by the agency's
external environment.

Unique objective functions for individual regula-

 

tors. The models which hypothesize unique objective
functions for individual regulators are more complex and
less definitive than those which assume an agency-wide

objective function. However. they may result in greater

~48-
explanatory power. The following is a brief summary of some
of these theories. The theories are not necessarily in
conflict with the Stigler-Peltzman and other theories:
instead. they may explain some of the deviation of actual
decision making from decision making predicted by theories
with agency-wide objective functions.

Roger Noll (1971a) assumes that individual regulators
maximize unique objective functions which may contain the
regulator's welfare. agency maintenance and/or growth.23 and
the welfare of groups and individuals impacted by the reg-
ulators' decisions. Noll downplays somewhat the Congress—
agency link. Although he feels that Congress is able to
influence decision making. he also believes that the budget
of a single agency is such a small portion of the total.
complex federal budget that Congress is not particularly
informed or concerned about individual decisions in an
agency when making budget decisions.

Like Stigler and Peltzman and many others. Noll sees the
power of interest groups. dependent on benefits and costs of
group activity. as affecting decisions. The tendency to
overcompromise so that all interested parties "gain" some-
thing (as in Peltzman's theory) results in a sacrifice of
the "social optimum" iJI‘Noll's mind. The degree of con-

troversy is hypothesized to be directly related to the

 

23It is conceivable that an agency could be growing dur-

ing one time period (e.g.. EPA in early 19705) and struggle
for maintenance during another time period (EPA now?).

_49-
length of time it takes an agency to reach a decision. the
amount of information search. the shifting of responsibility
for the decision and the costs of access to the agency by
interested parties. The latter strategy for diffusing con-
troversy may backfire if interest groups bypass the agency
anui go directly to Congress. thus increasing Congressional
pressure on the agency. Many of these ideas are reiterated
in Noll (1971b).

Marc Roberts (1975) has formulated a framework of the
variables which explain the decision outcomes of private and
public organizations. In an effort to explain how individ—
ual decisions are translated into an cmganization's deci-
sions. Roberts classifies explanatory variables into fcmr
groups: variables describing the agency's external environ-
ment. the internal structure of the organization. the set of
rewards and punishments which control an individual's behav-
ior within the organization and individual beliefs and
goals.

Roberts feels that (flue "external variables" are less
important than "internal variables" in explaining organiza-
tional decision making. which is counter to most of the
other theories examined thus far. However. this hypothesis
may be less true for government agencies requiring legisla-
tive and taxpayer support than for organizations which do

not have to account to external constituencies.

-5o-

Shaffer (1979) identifies several groups as participants
in the regulatory process: government (including the legis-
lature and the bureaucracy). consumers. the media and pro-
ducers of ruwhmegulation goods. The individuals in these
groups have their own sets of goals and operate in an en-
vironment constrained tar the institutional structure.
ideology. physical constraints and uncertainty. These con-
straints on decision making both determine behavior and are
determined by prior behavior in the political-economic
system.

Government suppliers of regulation have different goals
and face unique environments. and may not be held responsi-
ble for decisions since much government output is unmeasur-
able. Regulators seek information on the preferences of
those groups which demand regulation. As Stigler. Peltzman
and others recognize. it may be costly for some individuals
or groups to make their preferences known to regulators. but
Shaffer extends this concept by suggesting that there is a
market for information on preferences. Thus. the suppliers
of this information are able to influence the regulatory
process.

Ross Eckert (1973) attempts to explain decision making
with alternative internal structures of regulatory bodies
(an agency structure and a commission structure) in an
empirical study of taxicab regulation. He specifically
examines how different incentive structures facing commis-

sioners and agency officials result in different choices.

-51_

Eckert assumes that regulators rationally maximize
utility. with utility a function of multiple variables such
as wealth. prestige and convenience. Commissioners usually
have a legislatively-determined income independent of hours
worked or difficulty of duties. whereas agency officials may
obtain salary increases as the scope and complexity of their
tasks increase. Therefore. commissioners may generate more
utility by simplifying tasks while agency officials may max-
imize utility by increasing the agency budget or the number
of firms regulated. Eckert presents data on commission and
agency regulation of taxicabs which tentatively support his
ideas.

Kohlmeier (1973) sheds some light on the informal regu-
latory process -- specifically. with regard to private
meetings between regulators and the regulated. If the
regulators' objective functions include future» employment
within the regulated industry or avoidance of adverse
industry-wide publicity against the agency. then private
meetings may impact regulatory decisions.

Theories assuming unique individual objective functions
usually result in fewer definitive conclusions than the
theories assuming individual conformity to agency-wide
goals. Although the former group of theories may have
greater explanatory power. their use in applied research may
be difficult due to problems in measuring the variables.
determining individuals' objective functions. and aggrega-
ting individual behavior to make a statement about agency

behavior.

-52-

THE USE OF INFORMATION IN REGULATORY DECISION MAKING

If it is acknowledged that regulators operate in a
political setting and have goals other than maximizing net
benefits to society. then it becomes necessary to examine
the role of policy analysis and its resultant benefit-cost
information in regulatory decision making in order to under-
stand why the information is of a particular quality. This
section of the chapter examines the issues of whether more
information leads to better decisions. the characteristics
of regulators' information searches anui the various hypo-
thesized roles of information in regulatory decision making.

Is More Information Better? Many economists believe

 

that with enough information. a decision will "fall out" of
an economic analysis.24 but this viewpoint is not shared by
all. Still. many people do believe that more information
enhances decision making. even if it doesn't explicitly
point to decisions.

Schmid (1978) points out that no matter how good the
information is on costs and benefits. there is still no
objective decision rule for choosing winners and losers in
regulatory decisions. The concept of potential Pareto-
optimality in neoclassical welfare economics is based on an
assumption that winners reimburse losers: however. this does

not occur in the real world. Schmid feels that a careful

 

24This conclusion is reached after adopting the norma-

tive premises that utility can be summed across individuals
and that objective values can be obtained for benefits and
costs.

-53-
taxonomy and analysis of impacts can lead to better-informed
decisions. and therefore is very useful. but such informa-
tion does not tell regulators what to decide.

Lowrance (1976) specifically' addresses. risk..analysis.
such as the type used in pesticide regulation. He disting-
uishes between two subprocesses in regulatory risk analyses:
the determination of risk. which has the potential of being
a scientific process. and the judgment of the acceptability
of that risk. which is a value judgment. Like Schmid. he
feels that information can provide the regulator with a
determination of policy impacts. but it does not provide an
answer as to the acceptability of those impacts.

Majone (1979) also feels that risk evaluation is not
scientific or factual due to uncertainty and the necessity
of value judgments in regulatory decisions. He believes
that changes in the decision process may result in the
channeling of disagreement and conflict into better infor-
mation and better decisions.

Characteristics of Information Search. Several authors

 

stress the need for regulators to make uncertainty and
information overload more manageable in decision making.
ALl of these authors suggest that decision makers develop
standard operating procedures to achieve this manageability.

van Ravenswaay and Hull (1981) look at the information
requirements of food safety regulation which would result if
the goal of a regulator was maximization of net social bene—

fits. They imply that there is a market for information.

_54-
with supply determined by costs of the information and
demand determined in part by the agency's information
budget. Both information costs and the information budget
are affected by the regulatory context (including methods
for calculating the maximum amount of a contaminant to allow
in food. legal and technical constraints on information
production. and the total budget and other tasks facing the
agency) and by characteristics of the regulatory problem
(including complexity. familiarity and urgency).25 Since
the agency is unable to "buy" perfect knowledge. strategies
are developed to deal with uncertainty and conflict. Strat-
egies for obtaining information on risks from food contami-
nants include using high-dosage animal tests to determine
toxicity. extrapolating results from animals to humans and
from high-dose to low-dose. and making assumptions about
human exposure. Strategies for dealing with benefits in-
clude assuming full compliance and ignoring changes in the
price of food (and of other inputs in the production of
food) due to impacts of the proposed regulation. Thus. reg-
ulators simplify their information search in order to make
it more manageable.

Fiorino and Metlay (1977) also hypothesize that agency
strategies for coping with uncertainty are a crucial aspect
of decision making. When uncertainty is present in a prob-

lem. standard Operating procedures for dividing issues into

 

25Many of these variables are controlled by interested

parties. including the regulated firms and Congress.

-55-
sets of simpler problems are developed. These strategies
tend to be cybernetic. with previous outcomes partially
determining current strategies. Sometimes these strategies
fail to ensure rational decision making under uncertainty:
major issues may not be noticed. the decision process may
not be sensitized to signs of failure. subjective judgments
may have unintended outcomes. there may be imprecision in
implementing desired solutions and standard operating proce-
dures may not be useful when a case-by-case approach is
needed.

Edmunds (1980) also stresses the need for agencies to
simplify problems to make them more manageable. He specif-
ically' addresses the complexity and information overload
which accompany environmental policy decisions. One way in
which information is reduced to manageable levels is by
excluding consideration of some of the impacts of environ-
mental activity. or by ”bounding the problem rationally.”
Thus. complex interrelationships are reduced to simpler
concepts which are comprehensible to humans. Furthermore.
regulatory agencies and decision making processes are
actually structured to deal with issues within the bounded
area of a problem. and to ignore issues outside of that
domain. Information search goes no further than the or-
ganization's boundaries. In the case of toxic substances.
where uncertainty is present in both problem definition and
solution. EPA definitely must limit the amount of informa-

tion obtained on interactions between chemicals and the

—56-
ecological and economic systems. Congress has helped the
agency to some extent in this bounding process.

Sabatier (1978) provides a framework to conceptualize
the acquisition and provision of technical information in
regulatory decision making. He is essentially describing an
information market. Provision of information depends on
resources. characteristics of the problem. the legal and
political setting. and the expected reaction of decision
makers to the information. Acquisition of information
varies between agencies. within an agency. and between
branches of government. Between branches of government.
acquisition is affected by agency resources relative to
other branches. sources of legitimization of decisions.
agency mandates and court review of agency decisions. Dif—
ferences in acquisition Of information within an agency can
be explained by the degree of risk from decision impacts.
the social class of affected groups. the degree of conflict
and the potential influence of the information on decisions.
Variance between agencies depends on personnel. the degree
of risk from decision impacts. the capacity and willingness
of agencies to deal with technical issues and the amount Of

conflict in the political environment.26

Sabatier shows
that information search can be limited by many factors other

than the complexity of the problem.

 

261n the presence of conflict. agencies may seek in-

formation. on political preferences rather than technical
information.

-57-

In summary. the authors in this section suggest that
information acquisition is not simply a matter of deter-
mining what data are needed and then obtaining them.
Information costs. complexity of the problem. and political.
legal. budgetary. technical and time constraints are just a
few of the factors which determine the search for and ac-
quisition of information.

The Use of Information. After information is acquired.

 

how do agencies use it in the decision making process? The
literature addressing this question is categorized according
to the way it answers the question.

Information leads to rational decisions. Neoclass-

 

ical welfare economics deserves mention under this category.
It is widely believed among economists that policy analysis
involves obtaining information on net benefits of alterna-
tive policies. and that once enough information is acquired
the only remaining task is to determine which policy alter-

27 This view assumes

native maximizes net social benefits.
that the goal of regulator is (or ought to be) to maximize
net social benefits. an assumption which has been discred-
ited by the theories of regulatory decision making presented

earlier in this chapter.

Information is used by conflicting interests to

 

influence decision makers. Several authors believe that

 

 

27Accordingly. much criticism from these economists
centers on the failure of regulators to Obtain the ”correct”
information and choose the most efficient solution.

~58-
analysis is used by interested parties to influence decision
making. Johnson (1973) feels that information contained in
economic analysis is not used as much as information from
interested parties in making decisions. According to
Johnson. the Federal Communications Commission (and assumed—
1y. other regulatory bodies) sees itself as a quasi—judicial
body which gathers information from interested parties and
then makes decisions based on that information rather than
on formal economic analysis.

Randall Bartlett (1973) prOposes that uncertainty in
decision making results in information subsidies to agencies
from interest groups and individuals who control access to
information. Shaffer (1979) also feels that parties who
control information are able to influence regulatory deci-
sion making. since they provide information on preferences
of various groups. Stigler and Peltzman both acknowledge
the potential use of information to influence decision mak-
ing by including lobbying and the mitigation of opposition
to a policy as variables in their theories.

Information legitimizes decisions. ‘The idea that

 

information serves to legitimize previously-made agency
decisions is accepted by several theorists.

Majone (1979). in trying to persuade analysts to examine
the process by which decisions are made rather than the

outcomes. sees analysis as a legitimization of decisions:

-59-
”Regulators have sought legitimacy for their decisions by
wrapping them in a cloak of scientific respectability."28

Kelman (1974) also thinks that information justifies.
rather than aids. decision making when he studies the risk-
based regulatory process of the Consumer Safety Product
Commission (CSPC). Regulators must make a velue judgment
regarding safety (see Lowrance (1976)). One of the apparent
biases in CSPC is to at least partially decrease risk when a
product is investigated.

In order to cope with uncertainty. the CPSC has devel-
oped an information system to obtain data on injuries from
products and accompanying decision rules to determine which
products to regulate. Kelman suggests that the CPSC puts
energy and resources into the justification of product
investigation. the determination of the expected success of
the chosen policy (as Opposed to all alternative policies)
and estimates of the worth of averted risks and the costs
associated with the chosen policy.

Finally. Andrews (1980) suggests that information is
Obtained and documented in order to satisfy superiors within
an agency (or external reviewers. such as the Congress) and
to support previously-made decisions. Complex regulatory
decisions are too value-laden to be solved by information
alone. Andrews takes this idea a step further by hypothe-

sizing that information may actually result in worse

 

28Majone. p. 561.

-60-
decisions by confusing affected parties and diverting
attention from crucial issues.

Information (clarifies issues and Jpotential solu-

 

tiggg. Another widely—held impression is; that information
aids the decision maker in determining what the issues in a
regulatory problem are. and what the potential solutions to
the problem are.

Schwerin and Coyer (1979). like many other theorists.
recognize that regulatory decisions must be political
choices. but they attempt to show that the state of knowl-
edge and/or the amount of information in a particular policy
area affects the amounts of uncertainty and disagreement in
decision making. which in turn determine the outcomes of
decision rules based on voting. bargaining or hierarchy.

Knowledge is "paradigmatic" when there is a generally
accepted way of analyzing the problem and "preparadigmatic"
when there is no clear theoretical basis for analyzing the
problem. Schwerin and Coyer hypothesize that. given the
amount of information. paradigmatic knowledge narrows the
range of alternative solutions available to decision makers
since there is less uncertainty as to what the issues are
and how they can be solved. Decision outcomes are also
predicted based on the state of the knowledge. decision-
making rules and the power structure involved.

Maynard—Moody and McClintock (1981) feel that applied
policy research plays a large role in helping decision

makers grasp issues and limit possible solutions. Policy

-6l-
analysis provides general background information for prob-
lems rather than specific answers.

Multiple uses of information. Many authors

 

acknowledge various uses of policy research without placing
special emphasis on any particular use. They generally
agree with the preceding authors on the nature of the uses
of information.

Paul Sabatier (1978) provides a framework for informa-
tion use along with the previously described framework on
information acquisition and provision. In this framework.
information is used to influence decision makers: to conform
with mandates. case law. ethics. standard operating proce-
dures and court review (i.e.. to justify decisions): and to
facilitate long-term changes in agency perspective. There
are many determinants Of how information is used in
Sabatier's framework -- some include the resources of the
information source (credibility. communication. skills.
power). the content of the information message. the
political context and the resources and perspective Of
decision makers. An implication is that the greater the
degree of conflict in a regulatory issue the more informa-
tion is obtained (all other variables equal). but the less
influence it has on decision making.

A work similar in scope to Sabatier's is the article by
Peter H. Schuck (1979). Schuck outlines five general groups
of factors which may affect the implementation of a regula-

tory program. Information is only one of the five groups.

—62-
implyimg that many other factors may determine decisions.
Those other factors include the structure of the regulated
parties. the nature of the regulatory Objective. the en-
forceability of the regulation and the political support for
the regulation. Although Schuck does not specifically
predict the impact on regulation of the fifth group Of
factors. supply of and demand for information. the implica-
tion is that information use in regulatory decision making
depends on the other groups of factors as well as on the
sources and availability of information. the quality of
information. the quantity of information and other deter-
minants Of information supply and demand.

Finally. Park (1973) suggests several alternative uses
of information when he concludes that policy analysis does
not point to a clear decision. Information may be used by
conflicting interests to support their viewpoints. it may be
used by the agency and its supporters to justify previously-
made decisions. or it may provide regulators with a common
framework for discussing the policy issues of specific
problems.

Summary: The Nature and Use of Information in Regula-

 

tory Decision Making. ‘The above theorists all agree that

 

there is much more to regulatory decision making than ob-
taining objective information to make economically efficient
decisions. Regulators have discretion in decision making:

thus. decision making is influenced by factors such as the

-63-
goals of the agency and the power of other groups to influ-
ence the agency. Information may be used to justify deci—
sions. influence decision makers or clarify issues and
potential solutions. and this use is also affected by a
variety'<mf variables. Which variables affect information
acquisition and use depends on the regulatory situation. so
it is necessary to attempt to describe which aspects of the
above theories apply to EPA in its regulation of pesticides.
EPA'S REGULATION OF PESTICIDES: DETERMINANTS OF
DECISION MAKING AND USES OF INFORMATION

The many uses of information in the regulatory process
were outlined in the last section. Information may be used
to identify the consequences of regulatory alternatives for
different groups. It may be used to clarify issues. pref-
erences and solutions. It may be used to justify regulatory
choices or to influence decision makers. But regardless of
how information is used. it can generally be said that
information is considered important in decision-making and
that information is often costly. These two characteristics
of information make it important and interesting to examine
the types and quality of information that EPA and other
participants get into the RPAR process.

The empirical work presented in Chapters 5 and 6 will
show the types and quality of information on the conse-
quences Of pesticide use found in eight RPAR cases. The
data on quality of information. obtained from formal EPA

position documents. reflect how the RPAR participants in

-64—
each of the eight cases allocated their limited resources in
order to get information into EPA's decision process. The
position documents also contain measures of explanatory
variables -- that is. variables suggested in theories of
regulatory decision making to explain the quality of EPA's
information.

Thus. the empirical findings will be examined to see
how well the theories of regulation explain the quality of
information in the RPAR cases. Virtually all Of the
theories reviewed in this chapter lead tO the prediction
that the pattern of information will be related to the
preferences and power of groups affected by the EPA's
decisions. But the two major theories -- vote-maximization
and budget-maximization —- yield somewhat different pre-
dictions about the relative influence of interest groups in
the regulatory process.

Theories which assume that EPA's goal is to maximize
votes (n: political support (or ndnimize conflict) suggest
that interest group power is strictly a function of size.
wealth and organization costs. As a result. costly infor-
mation which supports a group's position should be found
only if the group is large. wealthy and well-organized.
When information is too costly to obtain. EPA should make
assumptions which favor the most powerful groups. But
Peltzman points out that vote-maximizing agencies make some
concessions to groups that Oppose powerful interests. so

there should be some information reflecting the opposition.

—65-
However. the opposing information would not be expected to
appear if it was costly to provide. EPA would not go out of
its way to perform exposure studies or make assumptions
favoring the opposition.

What should be expected if EPA is a budget-maximizing
agency? Interest group power would be related to the degree
to which a group furthers EPA's budget-maximization goal. as
well as the group's size. wealth and organization costs.
Accordingly. less powerful groups could become more powerful
if they were EPA allies. and stronger groups could be weak-
ened if they did not help to expand EPA's regulatory
authority. The latter groups could expect challenges to
their information from EPA. Agency assumptions should favor
agency growth. even ii? the assumptions were challenged by
powerful (but anti-EPA) groups. In addition. EPA could be
expected to expend resources on information increasing its
authority. and it could force powerful groups to obtain pro-
EPA information (or make it more costly for them to provide
anti-EPA information).

The ability of the vote-maximizing and budget-maximizing
theories to explain the empirical results will be analyzed

in Chapter 7.

CHAPTER 3
CONCEPTUALIZING THE IMPACTS OF PESTICIDE USE
INTRODUCTION
The objective of this chapter is to conceptualize the
risks and benefits of pesticide use. and to point out the
difficulties and value judgments inherent in conceptualizing
these impacts. Such a conceptualization provides a base-
line with which tO compare EPA's assessment of impacts for
various uses of different pesticides. in terms of both
excluded impacts and the quality of information obtained for

included impacts.

OVERVIEW OF EPA'S ANALYSIS OF PESTICIDES
EPA. in analyzing the impacts of pesticide use. adOpts a
risk-benefit framework based on modern toxicological princi-
ples and neo-classical welfare economics. EPA focuses on
the risks and benefits of pesticide 2.5-2 rather than pest-

icide regulation or pesticide manufacture and use. It is

 

 

difficult to determine all of the impacts of pesticide use.
but the choice of a framework provides some direction with
regard to the identification of impacts. The decision to
focus on pesticide HES narrows the scope of EPA's analysis.

but it also reduces the emphasis on the impacts associated

-66_

W‘Vi

-67-

with various regulatory alternatives. In pesticide regula—
tion. EPA does consider the impacts of alternative regula-
tory Options. but in an incremental manner that is very much
dependent on the analysis of pesticide use impacts. Final-
ly. the emphasis on use enables (even assures) EPA to ignore
the regulatory impacts (n1 manufacturers. formulators. and
marketers of pesticide chemicals: the effects of compliance
costs and non-compliance with regulatory decisions on reg-
ulatory outcomes: and the effects of other costs of regula-
tion on society.

The outcome of these problem-reducing techniques is that
some impacts of pesticide use and regulation are not con—
sidered. so the interests of some affected parties are not
considered in the regulatory process. This is not complete-
ly true of manufacturers. although the impacts on pesticide
producers are not explicitly considered in EPA's risk-
benefit analysis. because manufacturers are able to make
themselves heard in the rebuttal phase and at other points
in the RPAR process (including the assessment of benefits).
Although the process is supposedly open to all interested
parties. other groups with a stake in the outcome often do
not participate. perhaps due to high transactions costs or
low per capita benefits. Such groups may include formula-
tors of the pesticide. marketers of both the pesticide and
the goods for which the pesticide is an input. some users of
the pesticide and certain risk-bearers (although the EPA and
the Environmental Defense Fund play the role of political

entrepreneur in some cases).

\/

~68-

PROBLEMS IN CONCEPTUALIZING RISKS AND BENEFITS

The difficulties in conceptualizing the impacts of an
action fall into two general categories —— information
overload and the need for value judgments. Information
overload refers to the overwhelming number and complexity of
interrelationships in the system under analysis. and the
need for value judgments results from the aggregation of
welfare measures across individuals in policy analysis
techniques.

The Amount of Information and Difficulties in Tracing
Impacts. Many policy analysis textbooks tend to overshadow
problems in conceptualizing impacts while focusing on
measurement problems. The implication is that quantitative
description of impacts is au1 Objective process totally
dependent for its accuracy (n1 the resources and skills of
the analyst (see. for example. Anderson and Settle (1979)).

However. given the complexity of social. economic and
ecological systems. there is a problem of information over-
load in attempting to determine the impacts of an action
(Edmunds. 1982). much less the direction of the impacts.
For that reason. a benefit-cost or risk-benefit analysis
must be limited in scope. The impression from reading the
public documents from the EPA's regulatory actions on pest-
icides is that an incremental approach is adopted: current
risk-benefit analyses tend to be similar in structure and

scope (i.e.. impacts examined) to past analyses. Thus. one

-69—

criterion for limiting the impacts identified in an analysis
(aside from limits already imposed by the choice Of a frame-
work Of analysis and the focus on pesticide use) may be
whether or not the impacts have been studied in past anal—
yses. Other criteria may include choosing the most impor-
tant impacts for study in terms of dollars of benefits or
lives at risk. choosing the most obvious impacts for study.
or choosing those impacts which are the focus of the most

public input or controversy.

Value Judgments in Risk-Benefit Analysis. Some skeptics\\\
\\

of policy analysis seem to feel that there is a way to make
risk-benefit or cost-benefit analysis objective. but that we
lack the technology. resources or expertise to accomplish
this objectivity (Hapgood (1979). Anderson and Settle
(1978)). However. others see policy analysis as being
comprised Of value-laden elements which can never be avoided
and which are best dealt with in the open.

Scientists have noticed the value judgments inherent in
risk-benefit analysis. probably because they are often
requested to determine the acceptability of the risk of
certain actions. Lowrance (1980) points out that there are

two components of a risk analysis: measuring risk (which

 

can theoretically be done in an objective manner) and judg—

 

‘m‘-ww -..—... “my

MM

ing themggceptability of risk (which is always a value
v/M .....- .---.--“...m-w.

judgment). Wessel (1980) also describes value judgments in

risk—benefit analysis -- he notes that although science

itself is valueless. the use of science involves values.

9"

-70-

Other scientists continuing this theme include Gusman (1980)
and Sieck (1980).

Economists have also noted the value judgments in policy
analysis. especially in trying to aggregate individual
welfare into societal welfare. Assumptions must be made
regarding the marginal utility of a unit of income or wealth
to individuals at different points in the societal income
distribution. and about the ability to develop a single
social welfare function. Most economists feel that these
problems can either be solved or are not serious enough to

hinder the practice Of policy analysis. The usual method Of

...-
_.

aggregating individual welfare is to adopt the Hicks—Kaldor
criterion cflf potential Pareto-optimality. This criterion
assumes that the marginal utility of money or wealth is
equal for all individuals regardless of income. and that
costs and benefits can be summed across individuals to
Obtain a net benefit calculation for society. The value-
laden rationale for this economic efficiency calculation of
social welfare is that an action is good for society if
those who benefit from the action would be able to compen—
sate the losers and still have some benefits left over.
Thus. it assumes that efficiency is the primary goal of
public policy.

In literature vflfixfix is mostly theoretical. economists
have suggested ways of designing a social welfare function
incorporating more than economic efficiency. Haveman and

Weisbrod (1977) suggest attaching weights to different.

-71-

benefits and costs to incorporate equity in calculations of
net benefits. Many others. including Steiner (1977) and
Freeman (1977). suggest the possibility of obtaining soci-
etal welfare functions.

Others reject the view of a social welfare function.
Schmid (1979) believes that a: social welfare function can
never be objectively determined. and it is not the analyst's
place to make the necessary value judgments in an attempt to
do so. Such value judgments are questions of public choice.
and as such belong in the political process rather than
buried in a benefit-cost analysis.29

Schmid also points out some Of the more subtle value
judgments involved in policy analysis. The choice of im—
pacts to be examined in an analysis involves decisions on
which are to be included and how they are to be categorized.
These choices affect the issues addressed. measurements of
costs and benefits. and solutions to the problems. Schmid
also articulates a pOint that is rarely noticed by policy
analysts and welfare economists -- in order to calculate
costs and benefits. one must assume a particular distribu—
tion of property rights. In order for a policy to result in
the accrual of benefits to certain individuals. those in-
dividuals must be endowed with an initial set Of rights and
the rights to possess the benefits. Likewise. a cost to an

individual is a cost only if the rights of others allow them

 

For a similar viewpoint. see Hanke and Walkey (1977).

-72-

to impose costs on that individual. In other words. bene-
fits and costs are defined by a system of rights. Analysts
generally assume the status quo distribution of rights in
assessing the impacts Of a policy. but this is a value
judgment which implies that the current set of rights is the
best set of rights. If the desire of policymakers is to
change the system of rights. then analyzing the economic
efficiency of a policy is not terribly relevant. especially
given that the new set of rights will probably result in a
different efficient solution than the current set of rights

(which is used in the analysis).30

CONCEPTUALIZING THE IMPACTS OF PESTICIDE USE

In the last section. the choices involved in concep-
tualizing the impacts of an action were outlined. In this
section. those choices must be made in order to conceptual-
ize the benefits and risks of pesticide use. The choices
made will be the same choices that EPA makes. for the objec-
tive is to measure the quality of EPA's information against
some standard: it is the standard that is to be developed

here.

 

30A widely quoted "axiom" of the new welfare economics.

known as the Coase Rule. postulates that ownership of re-
sources (as determined by prOperty rights) has no effect on
resource allocation. given no transactions costs in Oper—
ating in the market. However. Schmid (1978) points out that
the Coase Rule only holds true if the change in rights does
not affect real income of the rights-holders. transactions
costs are zero. resources are perfectly mobile and the
rights of other parties are not affected by the exchange
between two parties. The result is that the Coase Rule is
295 widely applicable.

-73-

One choice has already been made -- to consider impacts
from the viewpoint of pesticide BEE (versus pesticide regu-
lation. pesticide manufacture. etc.). Some of the reasons
for EPA's choice of this viewpoint were previously dis-
cussed. The result Of the decision is that some impacts are
excluded from consideration.

Another decision is made regarding the analytical
method. with contingent value judgments. to be used. The
EPA has chosen risk-benefit analysis as its method for
assessing impacts. There is an implicit assumption of a set
of property rights t1) be used as a: basis for determining
costs and benefits: specifically. the status quo distribu-
tion of rights. EPA is also assuming that benefits and
risks can.lxa aggregated across individuals. Finally. the
choice of risk-benefit analysis as a methodology for deter-
mining impacts means that health risks are compared to bene-

fits net 93 all other costs. rather than total benefits

 

compared to total costs.

After choosing the method of analysis. EPA must classify
risks and benefits into Operational categories. decide how
extensively to measure impacts. and choose methods of
measurement. All of these decisions require value judg—
ments. as does the final regulatory decision.

Since the empirical work in this research is intended to

describe and explain the quality of information obtained for

-74-

various impacts Of pesticide use. some of EPA's value judg-
ments will be adopted in the following specification of im-
pacts in order to facilitate comparison and develop depen-
dent variables.

Risk Measurement. Lowrance (1980) divides risk analysis

 

inmo two separate subprocesses -u- the measurement of risk
(which can be objective) and an evaluation of the safety of
the risk (which is subjective). For the purposes of this
research. greater emphasis is placed on EPA's measurement of
risk than on the judgment of the acceptability of that risk.

Toxicity. Toxicity of a pesticide to human or»

 

nonhuman organisms refers to the disease- or injury-causing
capability of the pesticide. Ideally. scientists would like
to establish a dose-response curve for each affected organ-
ism which would describe the correlation between different
levels of exposure to a chemical and the incidence of dis-
ease. However. due to limited resources and technological
constraints. scientists must make do with group dose-
response data from nonhuman organisms which are Often based
on high-dosage experiments. Therefore. estimates for humans
usually are extrapolated from rodents or other test animals
to humans and from high doses to lower doses. On the rare
occasions when human exposure data is available (e.g.. as a
result of known accidental exposure). there is generally
only one dosage level which is sometimes unknown. In such
instances. the most that scientists can objectively discern

is that exposure to a particular compound is or is not

-75-

associated with an increased occurrence of disease. Some—
times data are not even available for test animals: in these
cases. EPA must rely on more economical cellular tests. data
for compounds with similar chemical structures and other
less desirable methods of estimated toxicity.

Environmental Fate. .A very complex aspect of

 

pesticide risk is the environmental fate of a pesticide.
Environmental fate generally refers to where and how long a
pesticide persists in an environment. This depends on a
myriad of chemical and physical processes that a pesticide
is exposed to after it is introduced into the environment.
Ideally. EPA would have a reliable model to determine
environmental fate for all of the possible use conditions of
a pesticide» Obviously. if a: pesticide dissipates or'
degrades rapidly. exposure will be lessened and so will
risk.

EPA's difficulties in measuring environmental fate arise
from the difficulties in conceptualizing and measuring the
many possible mechanisms for pesticide degradation. These
mechanisms include: photolysis. hydrolysis. oxidation.
drift. and runoff. The degree to which any of these or
other processes occur depends on field conditions. climate.
application methods. pesticide characteristics. and other
criteria. Data for all of these criteria are difficult. if
not impossible. to obtain. Just conceptualizing the phys—-
ical and chemical processes to which a pesticide will be

exposed is a major challenge.

r—w.

\-./

-76-

Exposure. A third major component of risk measure—
ment is the determination of levels of exposure of individ-
uals coming into contact with a pesticide. First. the
various groups exposed to the pesticide must be identified.
Then. the exposure of the individuals in these groups to the
pesticide via various routes (ingestion. inhalation. dermal
absorption) must somehow be estimated. Figure 3.1 lists
some of the groups that may be exposed to a given pesticide

and the potential routes of exposure.

/'

 

Group Potential Route of Exposure
Mixers and Loaders Dermal Absorption
and Aerial Ground Inhalation of Spray
Crews Ingestion of Food Residues
Aerial and Ground Dermal Absorption
Applicators Inhalation of Spray

Ingestion of Food Residues

Persons Exposed to Dermal Absorption
Spray Drift Inhalation of Spray

Ingestion of Food Residues

Farmworkers Dermal Absorption
Inhalation of Spray
Ingestion of Food Residues

Consumers of Contaminated Ingestion of Food Residues
Food

Persons Exposed Via Dermal Absoption

Accidents Inhalation of Spray

Ingestion of Food Residues

Figure 3.1 - Groups and Routes of Pesticide Exposure

-77-

Exposure would ideally be measured for each individual
under various conditions. but this is beyond EPA's means.
In descending order of accuracy. other methods of estimating
exposure would include estimating average exposure for
subgroups within a larger group (e.g.. female applicators.
infant consumers). estimating average exposure under "nor-
mal" conditions for a random sample of individuals in a
group. estimating average exposure for a non-random sample
under experimentally-imposed conditions. using exposure data
from similar pesticides. and guessing or assuming levels of
exposure. All of these methods have been used by EPA in

pesticide risk analyses.

For exposure to be meaningful across groups. it must be L////,a

put into common units of measurement. For example. the
average U.S. consumer may ingest 50 milligrams of endrin per
year throughout his Or her lifespan while the average appli-
cator is dermally exposed to 500 milligrams per year for
forty years. ten percent of which is absorbed through the
skin. Ln order to compare the two types of exposure. the
estimates are recalculated in terms of average body weight
and a particular period of time. Therefore. the final
estimate would be expressed as milligrams of pesticide per

kilogram of body weight per year.

§

\

 

The risk analysis. In order to estimate the prob— \\

 

ability of disease. or risk. the EPA needs to bring the

/!

estimates of toxicity together with the estimates of en- /

(

vironmental fate and exposure. When animal data are used \\

‘.

I
I

—78-

instead of human data for estimating human exposure. the
data must be extrapolated. Often. EPA assumes that the same
relationships between dose and response hold for humans as

for the test animals. In addition. extrapolation from high

dose to low dose must be accomplished if the test animals f

were exposed to higher dosages than humans. There are a5,

number of mathematical models for accomplishing this: EPA
generally chooses the linear or one-hit model. which is
based on the assumption that the smallest doses of a chemi-
cal result in some increased probability Of disease. Other
models. such as probit and logit models. predict less risk
at lower levels of exposure. (See Figure 3.2 for a compari-
son of the models.)

EPA always multiplies the exposure estimates by a "mar-
gin of safety." allegedly to insure that potential errors in
extrapolation leading 13) low risk estimates are corrected
for and to provide for sensitive members of a population.
The effect Of a margin of safety is to overstate the risk
estimate relative to a risk estimate calculated without the

margin of safety. There is no generally-accepted margin of

safety. EPA used a margin of safety of 100 for seven of the:

eight pesticides studied in this research. and used a margin
of safety of 500 for the remaining pesticide. Endrin. In
the case of Endrin. the margin of safety of 500 allowed a
definitive finding of human risk based on animal test re~
sults. However. EPA was able to provide sketchy evidence

suggesting that humans are more sensitive to Endrin than

-79-

 

 

 

 

Model Relationship of Relative Prediction of Risk
-—__— Dose to Risk At Low Levels of‘Exposure
Probit1 Normal Low Risk

Logit2 Logarithmic Medium Risk
One-hit3 Linear High Risk

Source: Compiled by author from information in Epp. et a1
(1977).

1The probit model is based on the assumption that dif-
ferent individuals' sensitivities to disease-causing chemi-
cals can be described with a bell-shaped or normal curve.

2Logit models assume that there are a given number Of
sites available for chemical bonding: when those sites are

filled the risk curve levels off. resulting in logarithmic
curves.

3The one-hit model assumes that a single exposure could

cause disease and that there is a linear relationship be-
tween dose and response. ‘

Figure 3.2 - Mathematical Models to Determine Risk

other mammals as a justification for the larger margin of
safety.

Carcinogenic risk is generally reported in probabilistic
terms. since it represents the probability of a particular
level 6f exposure resulting in illness or death. This
probability is based on the results of chemical exposure to
samples of test animals. and so reflects probabilities for
groups of individuals rather than specific individuals.

Teratogenic and fetotoxic risk is often reported as a
margin of safety rather than in probabilistic terms. Margir:
of safety is estimated by dividing human exposure into the.

lowest test animal exposure which produced a significant

 

-80-

number Of birth defects or fetal deaths or into the highest
no effect level of exposure to test animals. A value judg-
ment then has to be made on whether the estimated margin of
safety is adequate or safe: for seven of the eight case
study pesticides. a figure of less than 100 is considered
representative of substantial risk. and for Endrin anything
below 500 was considered unacceptable.

The Measurement of Benefits. The concept of the bene—

 

fits Of pesticide use in risk-benefit analysis is far
broader than the concept of benefits in cost-benefit anal-
ysis. Figure 3.3 compares the two different concepts.
Essentially. benefits in risk-benefit analysis are really
net benefits. including costs other than the risk of injury
or death. In benefit—cost analysis. gross benefits are
compared with costs. To facilitate discussion. benefits
will hereafter be referred to as net benefits.

Theoretical discussion of the net benefits ofgpes-

 

ticide use. Principles from benefit-cost analysis provide a

 

framework for examining the net benefits of pesticide use.
Net benefits can be discussed in terms of real net benefits.
which represent true net gains to society. and pecuniary net
benefits. which are redistributions of resources between
individuals which result in no net gains to society.

Real impacts. Real impacts or net benefits can

 

more specifically be described as the net benefits to
society of using the pesticide over the next best pest

control method. These real run: benefits take the form of

 

 

-82-

changes in output. quality. and/or production costs of the
final product for which the pesticide is an input. Some
real net benefits Of pesticide use are not normally captured
in the market for the output: these benefits might include
improved health and safety resulting from better nutrition
(due to less expensive food). fewer automobile accidents
from view obstruction as a result of herbicide use. the
avoidance of costs of transferring resources in the event of
pesticide restriction. and the avoidance of regulatory
costs.

The value in a competitive market of increased output or
quality of a pmeduct that is attributable to the use of a
pesticide can be measured by the willingness of society to
pay for the increased output or quality. Assuming that the
pesticide increases output or quality. Figure 3.4 illus-
trates the willingness of society to pay as being equal to A
+ B + C. That area is equal to the product of the change in
quantity or quality and the new market price. plus the value
of the gensumers' surplus associated with the new produc-
tion.

The decrease in production costs can also be depicted
graphically. by D + E - C in Figure 3.5. The area describ-
ing the reduction in total costs is equal to the reduced
costs resulting from a: decrease in production without the
pesticide (D + E). less the increased costs incurred in
producing the additional output (C). Figure 3.5 is based on
the assumption that production costs would fall: in reality.

W

total costs could increase with the use of the pesticide.

-33-

ply with Alternative

Price Sup 1y with Pesticide
per
Unit of |////// A///’///9

 

 

 

 

 

Good pA -_ A
PP / 'B
C
I 1 , Emmami
bA 6P Quantity of Good

Figure 3.4 - Value of Increased Quantity or Quality of
Output '= ‘

Price
per 1 .
Unit of . Mpp y W‘lth Alternative
Good

 

 

 

7v—

l
! Demand
Q .

A Q? Quantity of Good

Figure 3.5 - Value of Reduced Production Costs

-84-

Since the value of increased output and decreased pro-
duction costs is negotiated in the marketplace for the pest
control inputs and the product. EPA may be able to obtain
data on supply. demand and prices. although such data may
not be easy to obtain. The market provides key information
on real impacts. Input markets may also provide information
on other real net benefits of pesticide use. Restriction or
cancellation of pesticides would change the relative price
structure and productivity of inputs which are substitutes
for or complements to pesticides. Although the price
changes themselves are pecuniary impacts (see discussion on
pecuniary impacts below). the avoidance of any capital
losses to input users. unemployment or underemployment of
resources (due to immobility. noncompetitive labor markets.
etc.) and any other costs of transferring resources in the
event of pesticide cancellation in order to re-employ them
are real net benefits of pesticide use. In the same manner.
the avoidance of the real costs incurred in regulating the
pesticide via the RPAR process also could be classified as
real net benefits.

Also included under real net benefits. as the term is
used in risk-benefit analysis. would be "non-risk" costs Of
pesticide use. The costs of purchasing and applying the
pesticide are already included under changes in production
costs. In addition. there may be increases in the costs of
related production processes due to acute or chronic toxic-

ity. phytotoxicity. air or water pollution. or soil erosion.

-85—

Some of the products for which production costs may increase
include livestock. crOps. recreation..and :fish. Finally.
there are the environmental impacts other than those affect-
ing human health and production processes. including the
loss in value of environmental amenities resulting front
death or injury of wildlife or environmental pollution.

Pecuniary impacts. Pecuniary impacts refer tc>

 

those impacts which are transfers of income. wealth or util-
ity between individuals or groups rather than net societal
increases or decreases in income. wealth or utility. Exam—
ples of pecuniary impacts that may result from pesticide use
include:

1. Decreases in output prices to producers who don't
use the pesticide and to producers of substitute products.
offset by savings to consumers.

2. Impacts on input markets (e.g.. changes in relative
prices) other than transfer costs. offset by impacts on
buyers of inputs.

3. Impacts on markets for insurance and health care
due to more or less illness. offset by impacts on consumers
of those services.

Other examples of pecuniary impacts could be found.
depending on the specific pesticide and its circumstances Of
use.

Distribution of impacts. The benefits and risks of

 

pesticide use are distributed among many participants in and

-86-

out of the marketplace. Voluntary risk (e.g.. risk to pest-
icide applicators) may be better known to the risk-bearer.
and thus may be borne in part by the risk-imposer in finan—
cial terms and by the risk-bearer in health and financial
terms. 'If risk is involuntary and unknown. then the risk—
bearers will probably be faced with all of the risk. Risk
distributed among the risk-bearers may also be partly passed
on to employers in the form of lower productivity. and to
families and friends in the form of suffering. reduced
longevity. etc.

Real net benefits are also distributed among various
parties. The changes in value of production and production
costs are enjoyed by producers who use the pesticide. pro-
ducers who don't use the pesticide (who actually may suffer
a decline in income if product prices fall) and consumers of
the product. Improved safety and nutrition are enjoyed by
the general public and are evidenced by improved health and
productivity. although health care industries may experience
reduced revenues. The avoidance of capital losses and
transfer costs is of benefit to the owners of inputs and
users of inputs. and probably consumers Of the final product
of those inputs. Taxpayers and regulated firms profit from
the avoidance of regulatory costs. When the costs of re-
lated production processes increase due to pesticide
toxicity. producers and consumers of those goods generally
bear those costs. Finally. environmentalists and related

groups suffer from environmental impacts of pesticide use.

-87-

The distribution of pecuniary impacts depends on the
impact in question. As already mentioned. competitive
market producers who don't use a pesticide benefit if the
price of the product increases and pay if it falls (at least
in terms of returns per unit). This is a pecuniary impact.
ans it is offset by prices paid by consumers. These same
effects hold true for producers of substitutes for the final
product. Producers Of complementary products Observe the
opposite impacts: they benefit if the product price falls
and pay if it increases. Market prices for other products.
such as insurance and health care. may be affected by pest-
icide use and are distributed among the producers and the
consumers of the product. Non-transfer cost impacts on
owners of inputs. such as changes in relative and/or abso-
lute prices of inputs. are offset by impacts on input buyers
and ultimately on consumers of the product for which the
inputs are used. Other pecuniary impacts are distributed
similarly.

Impacts may be distributed disproportionately on a
geographic basis. as well as among groups. Individuals in a
particular region of the country may use more of a pest-
icide. consume greater amounts of pesticide residue. or own
resources with less mobility. In addition. if pesticide use
has macroeconomic implications. other individuals may be
affected in different ways than outlined above. It is
conceivable that the use of pesticides could have such a

profound effect (Hi the production of unufli of the nation's

-88-

supply of a given commodity that employment. price levels.
balance of payments. and other macroeconomic variables would
change. Finally. the relationship of the economic system to
political and social variables may result in societal

changes.

PROBLEMS IN MEASURING IMPACTS

A conceptualization Of the impacts of pesticide use was
accomplished in the previous discussion and is summarized in
Figure 3.6. The first section of this chapter pointed out
some of the many difficulties in conceptualizing impacts.
There are additional problems involved in measuring impacts
once they have been conceptualized.

Risk is difficult to measure due to a lack of informa-
tion on the response of humans to various doses and on human
exposure to the pesticide. Extrapolation from high dose to
low dose. from nonhumans to humans. and from a test system
to actual environmental conditions is necessary in order to
estimate risk. In addition. some method of estimating human
exposure must be devised. In some instances. technology is
not available for accurate measurement: in other cases.
accurate measurement is too costly to perform. The result
is that EPA either omits the information on risk from the
regulatory decision or implements lower cost and technolog-
ically feasible strategies for developing dose response and

exposure data. These strategies are discussed in Chapter 5.

I. Risks

A.
B.
C.
D.

-89-

Exposure to affected groups via various routes.
Toxicity of chemical.

Risk analysis.

Distribution among risk bearers. risk imposers.
others.

II. Net Benefits

A.

Real

1. Value of increased output of product.

2. Value of increased quality of output.

3. Decreases in costs of production of original
output.

4. Costs of producing new output.

5. Improved health and safety resulting from
better nutrition. fewer automobile accidents.
etc.

6. Avoidance of: Capital losses to users of in-
puts. unemployment. or other costs of trans-
ferring resources.

7. Avoidance of costs of regulating pesticide.

8. Increases in costs of related production
processes.

9. Environmental impacts other than risk and
production process impacts.

Pecuniary

1. Changes in output prices to non—using pro-
ducers.

2. Impacts on input markets.

3. Other.

Distribution of Benefits

hWNH
0

Among groups.

Geographic.

Macroeconomic benefits.
Social/Political benefits.

Figure 3.6 - Summary of Conceptualization of Benefits and

Risks

-90-

'The measurement Of net real benefits can also present

problems for regulators. At best. information on changes in ////””
yield due to pesticide use are Obtained from agricultural
test plots. These sample plots may not coincide with actual
environmental conditions in which crops are grown. It also
may be difficult to obtain pesticide price and use data. im—
pacts Of quantity changes on product price. impacts of pest- t/“
icide use (N1 other inputs 1J1 the production process. etc.
Net real benefits not accounted for in market prices are
even more difficult to quantify. as are net real benefits
accruing to groups other than the users of the pesticide and
pecuniary impacts. The problem in measuring net benefits is
considered by most economists to be one of costs rather than
one of technology. As a result of limited budgets. EPA must
again develop strategies for incorporating information on
net real benefits in risk-benefit analyses. Some of these
strategies are presented in Chapter 5.

Finally. distributions of impacts are difficult and
costly to measure. Determining who gets what share of real
and pecuniary impacts demands a knowledge of the general
equilibrium nature of the economic system -- knowledge which

is complex and expensive.

CHAPTER 4

METHODOLOGY

INTRODUCTION
The next step in the research. after examining theoreti-
cal literature on regulatory decision making and the risks
and benefits of pesticide use. was to attempt to empirically
describe the quality of information obtained by the EPA and
to use measures of some of the variables suggested by the
regulatory decision making literature to explain the quality
of information. This chapter describes the choice of a data
base. the measurement Of explanatory variables. the measure-
ment of dependent variables. and the measurement of correla—

tion between the explanatory and dependent variables.

CHOICE OF DATA BASE

The decision was made to use case studies as the source
Of data on explanatory and dependent variables. It was felt
that case studies would best describe the quality of infor-
mation used by EPA in pesticide decision making. Using case
studies from the same agency and governed by the same stat-
ute most likely eliminated some of the variation between
cases. thus providing more certainty in the empirical re—

sults. Lave. in The Strategy pf Social Regulation. conducts

 

 

an analysis of regulatory processes with case studies. In

-91-

-92-
the words of Crandall and Lave. "It is ironic that the
evidence for the enhanced role of scientific data and
analysis is a set of case studies ~- soft. impressionistic
material ~- which is being used to suggest the value of hard
analysis! But we know of no better way to illuminate the
role of scientific evidence in the world of regula-
tion.”31

The cases chosen for study were the eight pesticides for
which the RPAR process had been completed when the empirical
work began in March of 1982: Chlorobenzilate. DBCP. Prona-
mide. Amitraz. 2.4.5-T. Silvex. Endrin and Dimethoate.32

There was also some discretion as to where to obtain the
data for each case pesticide. The decision was to use the
sets of printed public position documents generated as an
output of the RPAR process. There is justification for the
use of so-called "paper trails" in previous research.

33 explain that different types

Maynard-Moody and McClintock
of research should be used when analyzing different regula-
tory processes. They cite two variables. ”outcome
standards“ and "causal understanding." that should be used
to evaluate a regulatory process to determine how to conduct

research of that process. In organizations theory. outcome

standards refer to agency goals and causal understanding

 

31Lester Lave. The Strategy of Social Regulation (Washing-
ton. D.C.: The Brookings Institution. 1981). p. 16.

 

32Since that time. 10 additional RPARs have been completed.

33Maynard-Moody and McClintock. pp. 644-666.

-93-
refers to the relationships between agency actions and
results. If both are ambiguous. as one could definitely
argue in the EPA's regulation of pesticides. then re—
searchers need a methodology that does not assume rigid
agency goals. The authors' suggestions? Case studies.
among other things. and qualitative measurement that relies
on observations. paper trails. and enabling legislation.
“Where quantitative measurement is used.“ they write on page
651. ”its purpose is to illuminate patterns of behavior in
the program.” The authors acknowledge that 'By relying
solely on reports. judgments about the evaluations were
based only on written comments. and there may be differences
between what actually happened and what was reported. It is
possible that insights and problems that were important ...
were not described. Nevertheless published reports are an
accessible source of information that reflect. however
imperfectly. the issues discussed.” The authors' main point
is that carefully designed studies of the "efficiency” of a
regulatory process are not relevant when it is not even
clear what the agency goals are.

Many other researchers have used paper trails in studies
of agency decisions. The studies are Often descriptive in
nature: the researchers generally’ acknowledge that paper
trails may constitute an incomplete picture of the regula—
tory process. but it is felt that the documents could

improve understanding of the process.

-94-

The position documents describe information sources and
measures of explanatory variables in a relatively consistent
manner between pesticides (as well as between uses of a
single pesticide). which made them ideal for comparison.
They are also interesting from the standpoint of analyzing
EPA's public record.

There is justification for using other sources of data.
especially interviews with EPA officials and/or interest
group representatives. Such interviews would have reflected

participants' perceptions of the quality of EPA's informa-

 

tion (which might not have always been accurate). as well as
allowing the researcher to obtain more details on the EPA's
data sources. However. time and financial constraints did

not allow the exercise of this option.

MEASURING EXPLANATORY VARIABLES

Which Variables? The literature on regulatory decision

 

making described in Chapter 2 pointed to a variety of vari-
ables that might explain decision making. It was decided to
investigate the explanatory power of variables reflecting
the incentives and power of different interest groups over
the information-gathering portion of the idecision making
process.

Three major groups stand out as having a stake in pesti-

cide decision outcomes: pesticide manufacturers. pesticide

-95-
users. and risk-bearers. The principles of the Peltzman—
type theories suggest that pesticide manufacturers might be
able to influence the regulatory process because the major
manufacturers are relatively few in number (thus reducing
organization costs) and in many cases the stakes appear to
be rather high due to the sale of large quantities of a
pesticide by individual corporations.

Pesticide users are usually more dispersed in terms of
numbers and distribution of impacts. although there are
incidents of oligopoly (such as strawberry nursery stock
growers in Delaware and Maryland. and pineapple growers in
Hawaii).

Risk-bearers are likely to be even more dispersed.
There is no formal organization to express the views of the
millions of people consuming food with pesticide residues:
there is not even a vocal organization for workers exposed
to harmful pesticides. Furthermore. the risks are not at
all well known to exposed sectors of the population and it
is not apparent what they could do if they decided that the
risk of being exposed exceeded the benefits.

To reflect the power of these interest groups over the
decision making process. the following explanatory variables
were chosen.

1) The stake of pesticide manufacturers in the outcome

of the regulatory process. measured by the product
of the retail price per unit of the pesticide and

the units of pesticide used for a particular crop.

2)

3)

-96-
Thus. the researcher attempted to measure this
variable for each use of each pesticide (86 uses in
all). This measure was intended to proxy the
relative value of different uses of a pesticide tc>
the manufacturer. It is hypothesized that greater'
value of a pesticide use to manufacturers leads tc>
more and better information on that use.

The stake of pesticide users (producers) in the
outcome of the regulatory process. measured two
ways: by the annual per acre user losses from
pesticide cancellation for each use and by the
annual total user losses from pesticide cancella—
tion for each use. Again. this variable was
measured for as many of the 86 uses as possible.
It is hypothesized that a greater economic stake to
producers leads to more and better information.

The percentage of total acres of a crop treated
with the pesticide was also used as a measure of an
explanatory variable. The measure was intended to
proxy the number of users affected for each pesti-
cide use. Stigler and Peltzman hypothesized that
the more users. the greater the difficulty in
organizing the political interest group.

But growers often have already-organized user
groups. including the USDA. to represent them in

the political process. Therefore. the hypothesis

4)

-97-
tested was that a larger percentage of the partic-
ular crop or use would result in better information
because it would serve as a larger incentive for'
user groups to become involved. This hypothesis is
not necessarily restricted to information on bene—-
fits. A similar argument can be made for risk
information. since widespread use may spring risk—
bearers' groups into action. In addition. risk-
bearers may react to better information from users
with increased information of their own.

The amount of risk is hypothesized to be positively
related to the quality of information. However.
this hypothesis is very difficult to measure for a
particular use of a pesticide. or even between uses
of a pesticide. because risk is usually constant
per unit of exposure for a pesticide. and data on
exposure and number of exposed persons are poor.
Instead. the decision was made to use risk tO

predict the quality of information between the

 

different_pesticides. So. unlike the other ex-

 

planatory variables. this variable was measured for
each of the 8 pesticides rather than for each of
the 86 uses. EPA's own ranking of the RPAR pesti-
cides was used to measure risk. particularly in
light of the following comment by Edwin Johnson.

Director of EPA's Office of Pesticide Programs:

-98-
"Pesticide law requires that both risks
and benefits be considered in all deci-
sions. But risk drives the process in
terms of depth of analysis and allocation

of agency resources. The greater the

 

apparent or potential for risk. the more
in-depth the analysis of both risks and
benefits. the more resources devoted to
the evaluation. and the greater the de-
gree of public involvement and scrutiny

accorded the decision.“ (Emphasis

 

 

added).34
So according to the people who make the decisions.
more and better information is gathered for pesti-
cides with greater actual or perceived risk. 13:
addition. the public (interest groups?) becomes
more involved when there is a higher degree of
actual or perceived risk.

How Measured? As mentioned previously. the EPA's pub-

 

lished position documents served as the data base for the
eight case study pesticides. The data for the explanatory
variables did not always just ”fall out" of the position
documents. Rather. measures for the variables often had to

be calculated.

 

34Edwin L. Johnson. ”Risk Assessment in an Administrative
Agency.“ The American Statistician 36. rum. 3. part 2
(August 1982): p. 233.

-99-

Value of pesticide use to manufacturers. The

 

measure used to proxy the value of a pesticide use to the
manufacturer was the quantity of pesticide used times the
pesticide's retail price per Ludln This measure was used
purely for ranking the pesticide uses: it was assumed that
the ratio of wholesale price to retail price is relatively
constant between uses of a pesticide and that manufacturer
costs per unit Of the pesticide are constant for all uses.
If these assumptions are correct. then the data provides an
accurate picture of the relative economic importance of the
different uses of the pesticide to manufacturers.

The time period chosen for measuring the quantity of
pesticide used was one year. This caused some problems in
the case of DBCP. because this nematocide is often applied
on a three-year cycle. For those uses. the quantity was
determined as the one-year average. In addition. amounts of
pesticide used were sometimes described as "minor."
"little." or "negligible." In such instances. the pesticide
use was assigned an annual value to manufacturer of 100.
which placed that use below the uses for which the values
were calculated. Position documents 2/3 and supplementary
benefits analyses were the sources of this data.

Retail prices for the pesticides were fairly easily
found in the position documents 2/3. Occasionally. dif-
ferent forms of the pesticide were used for different uses.

so there was more than one price for the pesticide.

-100-
Table 4.1 shows the value of uses of DBCP to manufactur-
ers. as an example. Data for the other pesticides can be

found in Appendix A.

Value of pesticide use to producers. Two measures

 

were used to describe the economic stake of producers in
uses of a pesticide: annual total user losses of pesticide
cancellation and annual per acre user losses of pesticide
cancellation. Again. data for both measures were obtained
from the position documents (primarily PD 2/3s). Pesticide
uses with qualitative descriptions such as ”minor.” “negli-
gible.“ “little." ”insignificant.” or 'small' were assigned
a value of 100 for total costs and .25 for per acre costs.
placing them in a lower ranking than those uses for which

costs to producers could be calculated.

Table 4.1 Value to Manufacturers of Uses of DBCP

 

 

 

Price of DBCP Pounds used Value to
233 per pound ($) Annually Manufacturer
Soybeans .67 12.378.000 8.293.260
Grapes 1.00 3.200.000 3.200.000
Almonds .67 3.417.000 2.289.390
Vegetables/Melons/

Strawberries .66 3.392.000 2.238.720
Peanuts .67 3.195.000 2.140.650
Cotton .66 2.700.000 1.782.000
Peaches/Nectarines .66 1.823.000 1.203.180
Citrus .62 112921000 8011040
Commercial Turf .66 550.000 363.000
Plums .67 452.000 302.840
Pineapple .83 302.000 250.660
Home Lawns .67 200.000 134.000
Other Berries .67 92.000 61.640
Strawberry Nursery

Stock .67 16.000 10.720
Home Gardens .67 5.000 3.350
Bananas - negligible 100
Apricots/Cherries/Figs - negligible 100
Ornamentals — unknown ?

-101-

The total losses measure was intended to depict the
industry—wide net economic impact Of pesticide cancellation.
whereas per acre losses provided information on the approx-
imate net impact of cancellation per producer. The former‘
measure. if correlated with the quality Of information.
would lend credence to the hypothesis that producer groups
provide information according to total dollars at stake.
The latter measure would go along with Peltzman and
Stigler's hypothesis that per capita economic stake helps
drive the political process. The original intent was to
report costs of cancellation per average-sized farm for each
use of each pesticide. but such data were simply not avail—
able. so the per acre cost figures had to be used instead.

Again. the problem: of' multi-year' application cycles.
namely for DBCP. had to be solved. Annual averages were
calculated when losses from pesticide cancellation were
reported for multiple-year cycles.

Data were not always available for per acre user losses
from pesticide cancellation because there were not always
data on acres treated with the pesticide. When there were
no quantitative or qualitative data. the use was not in—
cluded in the analysis of correlation between per acre costs
and quality of information. Table 4.2 and 4.3 show annual
total and per acre costs of cancellation of DBCP. respec-
tively. Data for other pesticides can be found in Appendix

A.

-102-

Table 4.2 Total User Losses from Pesticide
Cancellation for DBCP (Annual)

Use

 

Peaches/Nectarines

Soybeans

Grapes
Vegetables/Melons/Strawberries
Citrus

Almonds

Peanuts

Pineapple

Commercial Turf

Strawberry Nursery Stock

Plums

Home Lawns

Cotton

Other Berries

Apricots. Figs. Cherries. Walnuts
Bananas

Home Gardens

Ornamentals

 

* per year. after three years.

** annual averages based on three-year

Total Annual Costs (5)

 

$ 26.890.000**
23.500.000
21.670.000**
14.500.000

8.950.000**
8.834.000**
6.800.000
6.200.000*
2.200.000-5.600.000
l.500.000-5.600.000
4.600.000**
2.750.000
2.600.000
1.000.000
"negligible” (100)
”negligible" (100)
"negligible" (100)
9

crop cycles.

-103-

Table 4.3 Annual Per Acre and Total User Losses
from Pesticide Cancellation for DBCP

 

 

Annual Per Acre
Total Acres Costs of
£33 Costs (S) Treated Cancellation ($)
Strawberry Nursery $1.500.000 $2.500.00
Stock to to
5.600.000 600 9.333.00*
Pineapple 6.200.000 5.000 l.240.00*
Peaches/Nectarines 26.890.000 42.000 640.24*
Plums 4.600.000 9.000 511.00*
Commercial Turf 2.200.000 118.92
to to
5.600.000 18.500 302.70*
Citrus 8.950.000 31.200 286.86*
Grapes 21.670.000 83.000 261.08*
Ornamentals ? ? 88.08
to
172.55
Almonds 8.834.000 71.000 124.42*
Home Lawns 2.750.000 31.250 88.00*
Vegetables/Melons/

Strawberries 14.500.000 374.000 38.77*
Soybeans 23.500.000 1.133.000 20.74
Peanuts 6.800.000 355.000 19.15*
Cotton 2.600.000 225.000 11.56*
Other Berries 1.000.000 600 .25
Home Gardens ”negligible” ? .25
Apricots/Figs/

Cherries/Walnuts "negligible“ ”negligible” .25
Bananas “negligible” ”negligible" .25

 

* Calculated by author

~104-

Percentage of producers affected. As previously

 

mentioned. it was hypothesized that a higher percentage of
affected producers would serve as an incentive for producer
groups to involve themselves in the regulatory process. The
best measure of percentage of producers affected was the
percentage of total acres in the nation treated with the
pesticide. The assumption was that the percentage of acres
treated was directly related to the number of producers who
use the pesticide.

If the percentage of acres treated was described as
”very minor“ or ”little.” a value of .01% of total acres
treated was assigned to the use of the pesticide. which
ranked the use below those uses for which quantitative
estimates had been obtained.

In the case of endrin. the percentage of acres treated
for ”other vegetable seeds" was described as "like water-
melon and conifers.” The percentage for watermelon and
conifer seeds was 100% and 90% respectively. so a value of
90% was assigned to other vegetable seeds.

If there were no qualitative or quantitative data on the
percentage of acres treated. the use was not included in the
data base. If crOp cycles exceeded or were less than one

year. the percent of the crop treated per cycle was used.

 

Table 4.4 shows the percentage Of acres treated. intended to
proxy the percentage of affected producers. for DCBP. Data

for other pesticides can be found in Appendix A.

-105-

Table 4.4 Percentage of Acres Treated with DBCP
(Per Crop Cycle)

 

 

Use Percent of Crop Treated
Strawberry Nursery Stock 100%
Vegetables/Melons/Strawberries 8.1% to 95%
Plums 70%
Almonds 54%
Pineapple 46%
Peaches/Nectarines 44%
Grapes 31%
Peanuts 23%
Citrus 7.9%
Soybeans 2.1%
Cotton 2.0%
Home Gardens <0.5%
Other Berries <0.1%
Apricots/Cherries/Figs/Walnuts <0.l%
Bananas <O.l%
Commercial Turf ?
Ornamentals ?

'0

Home Lawns

Losses to risk-bearers. Although pesticide manu-

 

facturers and users gain from the use of a pesticide. some
individuals lose —- namely. risk-bearing individuals and
possibly nonusing producers. Data on losses to nonusers
were not available. but there were meaningful data on risk
for the eight pesticides. In fact. the data were EPA's own
rankings of the eight pesticides according to various risk
variables. which lended more credence to the testing of
Edwin Johnson's statement that more risk leads to more
information.

The ranking for Amitraz had to be calculated by the
researcher. since it was not rated by EPA. Table 4.5 shows
the relative risk calculation for Amitraz. EPA's ranking

of the eight pesticides can be found in Table 4.6.

Table 4.5 Measuring

Criteria

Total Production

Oncogenicity (Mammals)

Other Chronic Toxicity
(Mammals)

Acute Toxicity (Mammals)
Fish and Wildlife Toxicity
Persistence
Biomagnification and

Bioaccumulation

Environmental Mobility

~106-

Relative Risk for Amitraz

Measure

up to 162.715 lbs
active ingredient

positive

negative

positive
(p.1 of PD 1)

negative
(p.1 of PD 1)

no evidence of
persistence

no evidence
drift

TOTAL Risk Factor

Score

.Auocusm ecu >9 UOOMHsonO mm3 :OHSB .NmuuHsd mow bewoxm- ehmH ..O.o .COumchan
[Hm-Es: men-9800.5 Lemmas: Dow-Home: .msz>wm mpHoHumwm 130QO no 03qu .<.m.m.m.D "wousom

.meu mwumwum wumoHch mmmoom Hmuou no“ Ucm mHumuHuu Hm5©H>H©cH co nmuoum Lou enemas: mmcmHm«

 

 

 

 

e s n 5 NH MH MH Hm mmoom q<eoe
H H H H H H H H wuHHHboz HeucwecouH>cm
o o O o m o m w coHumHsEsoomon
Ucm coHumoHuHcmmEOHm
H o m m m H N v mocwuanmme
-h o o o o H m H m OMHHGHHZ
m can anm 0» >3on8.
-
o m o o H m H m nHmEEmz
Ou wuHOone wusu<
o o H H o o o o mHmssmz 0»
quxoe OHcOLLO uwcuo
v v o o v v v v 388.3808
0 o m m o m H N coHuuseoum Hmuoa
weHcmEoum NmuuHE< alm.w.m xw>HHm wumocquHo some wumHHNcmb cHuecm mHmwuHuU
IouoHcU
wuoom
«mUHOHunwm

comm uOm mmoum mam mem mcHEumuwo Cu new: mHmwuflHu umcchmm mem meHOHumOm n.¢mm ©.v wHQmH

~108-

Problems in measurigg explanatory variables. There

 

were some problems in measuring the explanatory variables.
Specifically. the "value to manufacturers" variable was
proxied by using retail price times the quantity Of the pes-
ticide used annually. Within a pesticide. this measure
presumed a constant relationship between wholesale and
retail prices and a constant rate of return between uses of
the pesticide. For comparison between pesticides. it pre-
sumed a constant rate of return.

For "annual total user losses from pesticide cancella—
tion." data were often available for crop or application
cycles rather than on an annual basis. In those instances.
averages were calculated to reflect annual values. Averages
may not have reflected the loss picture in a particular
year. but they did provide a basis for overall comparison
between uses of a pesticide.

For one pesticide. there was difficulty in calculating
”annual per acre user losses." These data had generally
been calculated by EPA. but in the case of DBCP. they had to
be calculated by the researcher. This was accomplished by
dividing ”annual total user losses" by the estimated number
of acres treated.

In the case of "percent of acres treated." the data were
fairly attainable for crop or application cycles. There was
no need to calculate the average annual percentage of acres
treated. since the intent was to capture the proportion of
producers affected in general rather than the proportion

affected in a single year.

-109-

The EPA's risk ranking for the pesticides was considered
to be relatively reliable. It reflects EPA's own perception
of risk. which is most relevant since it was being used to
test Johnson's hypothesis that more risk led to more and
better information. The calculation of relative risk for'
Amitraz was not difficult. although the researcher was not
completely sure of the values for biomagnification/
bioaccumulation and persistence. based on the information in
the position documents. Phone calls to EPA indicated that
Amitraz never had been ranked according to risk.

In general. there was a problem in measuring the ex-
planatory variables with data from EPA's own risk benefit
analyses. If EPA assigned qualitative or low values in-
accurately. simply because there was no information. there
may have been a bias introduced into the empirical work.
For example. EPA may have described annual total costs of
DBCP cancellation to banana producers as being “negligible“
because there was no information rather than because there
was information that DBCP use on bananas really was negli-
gible. However. EPA often made a distinction between an
unknown value and a negligible value for explanatory vari-

ables.

MEASURING DEPENDENT VARIABLES

Which Measured? The dependent variables in this re-

 

search refer to the quality of information obtained by EPA
to describe the various impacts of pesticide use. Chapter 3

was devoted to a description of the potential risks and

-llO-
benefits of pesticide use: this conceptualization was the
basis for delineating and organizing the impacts which EPA
calculated and could have calculated. EPA's consistency in
measuring impacts across pesticides helped the researcher
break the impacts apart into variables for which the EPA
collected data. It was the quality of information obtained
for these variables which served as the dependent variables.
EPA's consistency was also important for the one explanatory
variable (EPA's risk ranking) used to compare the quality of

information between pesticides rather than between uses.

 

Tables 4.7 and 4.8 describe typical sets of variables per-
taining to risk and benefits for a pesticide. The goals in
designing these sets were tO cover the impacts that EPA
measured. to make them broad enough to allow for comparisons
between pesticides. and to make them specific enough to
describe variation between uses of a pesticide.

These tables can serve as a comparison with Figure 3.5.
which outlines the theoretical conceptualization of benefits
and risks. In general. EPA spent virtually all of its
resources measuring what most people would consider to be
the major impacts -- risk analyses for the major groups
(applicators and consumers. in particular). the benefits
from increased value (HE production. and the benefits from
reduced (or increased) pest control costs. Most of the-
pecuniary and distributional impacts were ignored for the
eight pesticides. and some of the risk categories (espe—

cially farmworkers and exposure to spills and drift) also

-111-

Table 4.7 Typical Risk Variables for Which
Quality of Information was Measured

Toxicity
Exposure. Human

Consumers
Residues in food
Amount of food in diet
Percent Of crOp treated
Risk

Ground Applicators

Dermal
Amount of pesticide exposure
Absorption of pesticide through skin
Duration of exposure

Inhalation
Amount of pesticide exposure
Absorption of pesticide
Duration of exposure

Dietary
Residues in food
Amount of food in diet
Percent of crop treated

Risk
Oncogenicity
Fetotoxicity/Reproductive Effects
Mutagenicity

Aerial Applicators. Mixers and Loaders. Farmworkers.
Persons exposed via drift. Persons exposed via spills.
Flaggers

Same as for Ground Applicators

Exposure. Wildlife

Decreases in Nontarget Species
Potential exposure
Lethal dose required to kill 50% of test animals

(L050)

Likelihood of exposure
Risk

Acute Toxicity to Wildlife
Residues in food

LD50

Amount of food consumed
Risk

Fatality to Endangered Species
Exposure

LD50
Risk

-112—

Table 4.8 Benefits Variables for Which Quality of
Information was Measured

Changes in pest control costs due to pesticide cancellation
Cost of pesticide per unit
Amount of pesticide used per acre
Acres treated with pesticide
Cost of alternative pest control per unit
Amount of alternative used per acre
Acres that would be treated with alternative
Acres that would be abandoned
Per acre changes in pest control costs
Total changes in pest control costs

Changes in yield due to pesticide cancellation
Yield per acre with pesticide
Yield per acre with alternative pest control
Quality of product per acre with pesticide
Quality of product per acre with alternative pest
control
Price per unit of product
Per acre change in value of production
Total change in value of production

Changes in other production costs
Impacts on producers of other goods
Distribution of impacts

Users Of pesticide

Nonusers

Marketers of product

Consumers of product

Pesticide manufacturers
Geographic distribution of impacts
Macroeconomic impacts
Social impacts

Probability of compliance with EPA restrictions

Cost of restrictions to users

-ll3-
were ignored for virtualLy all of the pesticides. In ad-
dition. the dietary exposure route was usually ignored for
the non-consumer risk categories.

In summary. EPA never met the standard set forth in
Chapter 3 for measuring the impacts of pesticide use -- many
Of the impacts were ignored or. at best. qualitatively
discussed (mentioned might be a better word). How then. was
the transition made from a theoretical conceptualization of
impacts to a pragmatic description in Tables 4.6 and 4.7 of
which "bits" EPA actually looked at 1J1 its position docu-
ments?

Mainly. the transition was made by carefully reading and
rereading the position documents. Detailed study of the
documents in chronological order (according to the date of
publication Of the PD 2/3s) revealed an incremental approach
by EPA in determining which impacts of the different pesti-
cide uses to measure. how to measure them. and sources of
data.

How Measured? By far. the largest block Of time in the

 

empirical work was spent measuring the dependent variables
(i.e.. the quality of information for the variables which
made up the impacts of pesticide use). Position documents
were painstakingly read and reread to obtain first a broad
picture of the impacts for each pesticide use and how they
were measured and then a detailed picture of how the impacts

were measured.

-ll4-

Impacts were initially listed on separate sheets of
paper. with details on how each impact was measured by EPA.
Based on this information. and after impacts were outlined.
measurement scales were developed for each impact. These
scales are depicted in Tables 4.9 to 4.19. The measurement
scales were then applied to each impact for each use of each
pesticide. The data were obtained by once more carefully
reading the PD 2/3s and benefits analyses. and were coded by
hand onto large sheets of paper. On the first reading. data
were coded only for those impacts for which the quality of
information (i.e.. the information source) was fairly-well
documented. This comprised approximately 2/3 of the data.
The documents were scoured again to obtain data on the less
well-documented impacts. Risk data were completely checked
twice. due to the researcher's relative unfamiliarity with
risk analysis and due to poor documentation of sources of
information. Benefits data were thoroughly checked once:
many of the data were Obtained from the benefits analyses
rather than tflua position documents. All together. nearly
8.000 pieces of data were obtained for 86 uses Of the eight
pesticides. Copies of the data are on file at Michigan

State University Department of Agricultural Economics.

-115-

Table 4.9 Measurement Scale for Quality of Information:
Toxicity

Each Unrebutted Animal Study 2 pts.

Each Unrebutted Epidemiological Study
(Study of a Human Population) 5 pts.

Suggestive Evidence (e.g.. Mutagenicity
or Chemical Structures as Suggestive
of Oncogenicity) 1 pt.
for each category
of evidence

Table 4.10 Measurement Scale for Quality of Information:
Exposure

Residues in food. amount of food consumed. and percent Of
crOp treated:

Ignored or not specifically mentioned = 0 pts.
Qualitatively mentioned = 1 pt.

Quantitative. assumed = 2 pts.
Quantitative. expert opinion = 3 pts.
Study. quantitative range = 4 pts.

Study. quantitative point = 5 pts.

-116-

Table 4.11 Measurement Scale for Quality of Information:
Dermal and Inhalation Exposure

Amount of exposure. absorption rate. and duration of
exposure:

Ignored or not specifically mentioned = 0 pts.
Qualitatively mentioned = 1 pt.

Quantitative. assumed = 2 pts.
Quantitative. expert opinion = 3 pts.
Study of similar pesticide. quantitative range = 4 pts.
Study of similar pesticide. quantitative point = 5 pts.
Study. quantitative range = 6 pts.
Study. quantitative point = 7 pts.

Table 4.12 Measurement Scale for Quality of Information:
Risk Assessment

Ignored or not

specifically mentioned = 0 pts.
Qualitatively mentioned = 1 pt.

Quantitative. range = 2 pts.
Quantitative. point = 3 pts.

Table 4.13 Measurement Scale for Quality of Information:
Risk Of Alternative Pest Control

Ignored or not
specifically mentioned

0 pts.

Qualitatively mentioned 1 pt.

Quantitative 2 pts.

Table 4.15

-118-

Measurement Scale for Quality of Information:
Significant Decreases in Nontarget Populations

Potential Exposure

 

Ignored or not specifically mentioned =

Qualitatively mentioned =

Quantitative.
Quantitative.
Study
Study
Study
Study

assumed =

expert Opinion =

of similar pesticide. quantitative range =
of similar pesticide. quantitative point =
of pesticide. quantitative range =

Of pesticide. quantitative point =

\lO‘Uluwal-‘O

Lethal Dose Required to Kill 50 Percent of Test Animals

Ignored or not specifically mentioned =

Quantitative.

LD50

LDso

assumed =

for similar species =

for species =

Likelihood of Exposure

 

Ignored or not specifically mentioned =

Qualitatively mentioned =

Quantitative.

Quantitative.

assumed =

expert opinion =

Theoretical model. quantitative estimate =

Study
Study
Study
Study

Risk

of similar pesticide. quantitative range =
of similar pesticide. quantitative point =
of pesticide. quantitative range =

of pesticide. quantitative point =

Ignored or not specifically mentioned =

Qualitatively mentioned =

Quantitative.

range =

Quantitative. point =

O

1
2
3

CDQO‘U‘ubOJNi-‘O

NNHO

pts.
pt.

pts.
pts.
pts.
pts.
pts.
pts.

pts.
pt.

pts.
pts.

pts.
pt.

pts.
pts.
pts.
pts.
pts.
pts.
pts.

pts.
pt.

pts.
pts.

-ll9-

Table 4.16 Measurement Scale for Quality of Information:
Fatality to Endangered Species

Concentration of Residues

 

 

Ignored or not specifically mentioned = 0 pts.
Qualitatively mentioned = 1 pt.
Quantitative. assumed = 2 pts.
Quantitative. expert opinion = 3 pts.
Theoretical model. quantitative estimate = 4 pts.
Monitoring similar species. quantitative range = 5 pts.
Monitoring similar species. quantitative point = 6 pts.
Monitoring species. quantitative range = 7 pts.
Monitoring species. quantitative point = 8 pts.
Lethal Doses Required to Kill 50 Percent of Test Animals
Ignored or not specifically mentioned = 0 pts.
Quantitative. Assumed = 1 pt.
LDSO for similar species = 2 pts.
L050 for species = 3 pts.
Ignored or not specifically mentioned = 0 pts.
Qualitatively mentioned = 1 pt.
Quantitative. range = 2 pts.
Quantitative. point = 3 pts.

-120-

Table 4.17 Measurement Scale for Quality of Information:
All Changes in Pest Control Costs and Changes

in Value of Production Data Except Price

Ignored or not specifically mentioned
Qualitatively mentioned

Quantitative. assumed

Quantitative. expert opinion

Study or published data. quantitative range

Study or published data. quantitative point

U'lnb

pts.
pt.

pts.
pts.
pts.

pts.

Table 4.18 Measurement Scale for Quality of Information:

Price

Ignored or not specifically mentioned
Qualitatively mentioned

Quantitative. assumed

Quantitative. expert Opinion

Study. quantitative range

Study. quantitative point

Quantitative study. varies with quality
of commodity

Quantitative study. varies with supply
of commodity

Quantitative study. varies with quality
and supply

pts.

pt.

pts.
pts.
pts.

pts.

pts.

pts.

pts.

Table 4.19

-121-

Measurement Scale for Quality of Information
Per Acre Change in Pest Control Costs. Total
Change in Pest Control Costs. Per Acre Change
in Value of Production. Total Change in Value
of Production. Change in Other Production
Costs. Impacts on Producers of Other Goods.
Distribution of Impacts (Users. Nonusers.
Marketers. Consumers. Manufacturers).
Geographic Impacts. Macroeconomic Impacts.
Social Impacts. Degree of Compliance. Costs of
Restrictions

Ignored or not specifically mentioned = 0 pts.

Qualitatively mentioned

Quantitative.

1 pt.

range = 2 pts.

Quantitative. point = 3 pts.

Table 4.20 Measurement Scale for EPA Decision

Decision

Register pesticide = 0 pts.
Register with restrictions = 1 pt.
Cancel pesticide = 2 pts.

-122-

Problems in Measuring Dependent Variables. The depend-

 

ent variables were in some ways easier and in some ways
harder to measure than the explanatory variables. They were
easier because data were being generated rather than sought.
If data were unavailable for the explanatory variables. they
were unavailable -- the researcher could not go out and
measure the acres of apples treated with Amitraz. Measure-
ment of the dependent variables was up to the researcher. so
there was more control over the quality Of the data.

But measuring the dependent variables was difficult
because EPA's documentation of information sources was Often
poor. there were about 8.000 pieces of data to collect. and
the measurement scale for each dependent variable was good
but not foolproof. which sometimes led to questions over the
value to be assigned to a variable. In general. the bene-
fits variables were easier to measure because of better
documentation and the researcher's training in economics.

Specific comments on risk coding. Specific prob-

 

lems in measuring the risk variables can best be depicted by
presenting some of the notes made during the measurement
process. These notes are summarized in Table 4.21.

General comments on risk coding. The basic

 

approach in coding the dependent risk variables was to
strive for consistency within a pesticide over consistency
between pesticides. Instead of being straightforward. the
risk coding Often ended up as a judgment call after all the
information had been compiled. The goal in these situations

was to make consistent judgment calls.

-123-

Table 4.21 Specific Problems in Measuring Dependent
Risk Variables

Chlorobenzilate

 

0

Assume that ground application exposure includes ex-
posure from mixing and loading.

Since absorption was estimated and does not change
between routes of exposure. should it be coded for all
routes? No. because in many cases those routes of
exposure were not even mentioned.

When there are inconsistencies between the text and
tables. which should be followed? In case of farm-
workers. go by text.

Is there ever aerial application for cotton. fruits and
nuts and "other?” Is there ever farmworker exposure for
”other”? Assumed not.

Assumed no dietary exposure from "other" uses.

National Cancer Institute study on oncogenicity not
reported because it had not been finalized.

Toxicity (Adverse Reproductive Effects) was coded as 11
due to 5 studies and suggestive work (1 pt.).

Residues in food coded as 2 because detection level was
assumed even in the absence of any detected residues.
Residues for Florida citrus also given 2 because an
assumption was the limiting factor (even though other
subcomponents of the data were of a higher quality than
a 2).

The data source for "food in diet" was simply undocu—
mented.

Duration of exposure was given a 3 because. although it
was a worst-case assumption. it was based on USDA opin-
ion.

For drift. aerial applicators. noncitrus applicators and
farmworkers. there was no mention of absorption and
duration. The text simply said that there ”could be
exposure.” Therefore. duration and absorption were
coded as 0. with exposure getting a l.

Assumed that there could be aerial application for
Arizona citrus even though no mention. because the text
does not say that there is no citrus rust mite in
Arizona (only citrus pest for which aerial application
was used).

-124-

Amitraz

O

Amitraz was exceptionally' well-documented. The only
note was that a 2 was assigned to absorption of the
chemical. even though the position document mentioned a
study on toxicological properties. because there was no
documentation of how absorption rates were Obtained.

S-T

For toxicity. when studies only had preliminary results.
they were assigned a 1. Position document 1 was used to
code toxicity.

 

O For the percent of crop treated under dietary exposure.
a 2 was coded. because EPA appeared to assume that all
milk and meat had 2.4.5-T residues.

0 Even though there appeared to be expert opinion on the
amount of inhalation exposure to range and handsprayers.
a 2 was coded because the expert appeared to have no
basis for his statement. It appeared that his statement
was an assumption.

0 The amount of dermal exposure to range and handsprayers
was given a 1 because. although Dow Chemical provided
data. the data was not used by EPA and was difficult to
interpret.

o The researcher was less confident of the risk data for
2.4.5-T than for other pesticides. Much of the data had
to be obtained from position document 1 instead of the
PD 2/3.

Pronamide

o Toxicity received a 3 due to one study and suggestive
evidence (on nitrosamines).

Endrin

0 Potential exposure to nontarget populations was assigned
a 0 in spite of a reference to "environmental effects."
because the reference was not specific enough to use as
data.

O For Endrin use on bird perches and tree paint. exposure

to bystanders was coded under exposure via drift.

DBCP

-125~

 

Mixer and loader risk was often included in ground
applicator exposure.

Food residues for alfalfa reported for secondary prod-
ucts in the food chain (meat. milk. eggs).

Reproductive effects risk analysis for some uses re—
ported as margins-Of-safety for individuals consuming
quantities of residues. rather than as numbers of re-
productive effects per 1000 population.

Although quantitative risk analyses were obtained for
many exposed groups and uses. there was no mention
anywhere of the information used to determine the degree
of absorption of inhaled pesticide. Therefore. zeroes
were coded.

For exposure from spill cleanup under the ”other
berries" use. the quantitative risk analysis may have
only pertained to raspberries.

Dimethoate

 

O

Mixer and loader risk was either determined for aerial
spray mixers and loaders. or else was included in ground
applicator exposure.

The risk analysis for reproductive effects on consumers
was coded as 3 because the margin-Of-safety for individ-
ual consumers was quantified even though the number Of
adverse reproductive effects per 1,000 population was
not.

For alfalfa. food residues in secondary products (meat.
milk. eggs) were reported.

For lettuce ground applicator duration of dermal ex—
posure. the separate benefits analysis was used to
determine the quality Of information. It appeared that
EPA used an expert opinion on the number of lettuce
ground applicators to estimate duration.

Silvex

Risk information very sketchy in general.

For dietary exposure in rice ("food in diet" variable).
there was no indication of the source of the data. so
the quality of information could not be measured.

—126-

Another general rule was that if a variable relevant to
pesticide use was mentioned qualitatively. but without
regard to a specific use of the pesticide. then a zero was
coded for that use. This cut down on attempts to disting—
uish between the quality of qualitative data -- the
measurement scale became cruder. but actual measurement
became less crude. It was reasonable to code a zero because
information without reference to use was of little or no
help to EPA in decision making.

The rest of the measurement scale was more straight-

forward. The difficulty came in two major areas:

1) Finding EPA's data sources in the position docu-
ments.

2) Interpreting the quality of information when there
was more than one component to a particular vari-
able. For example. "durathmi of exposure" some-

times consisted of data on acres treated. time
needed to spray one acre. and number of applica-
tors. The rule in cases such as this was to use
the ”limiting factor" for coding: that is. to code
the lowest-quality piece of information for ”dura-
tion of exposure.”

General comments on benefits coding. Benefits
coding was easier because it was more familiar. but also
because there tended ix) be better documentation of infor-
mation sources. For each pesticide except Chlorobenzilate
and Silvex. there was a benefits analysis which supplemented
the Position Document 2/3. The additional documents pro-

vided a good deal of information that otherwise would not:

have been obtained.

~127-

The same basic rules used in the risk coding were
applied tO the benefits coding. There was one small vari-
ation in the "limiting factor" rule: if EPA obtained a
number for a benefits variable. the researcher coded the
quality of that number. even if it only represented those
states or regions responding to a USDA/EPA survey. This
decision was made because EPA generally reported the number
as representative of the nation in Position Document 2/3.
and because generally the number represented most of the
product treated with the pesticide.

Before ending the discussion of dependent variable
coding. it should be mentioned that the initial coding
included all of the risk and benefits variables for each
regulatory alternative considered by EPA. Since there were
as many as seven alternatives for each use of a pesticide.
this added several thousands of pieces Of data to the ini-
tial coding. The idea had been to determine variations in
the quality of information between regulatory options. but
the data were terrible and often nonexistent. It did not
appear that such an analysis would be meaningful. so the

data were not checked in the secondary stages of coding.

MANIPULATING DATA FOR COMPARISON BETWEEN PESTICIDES
Testing Johnson's hypothesis that greater risk resulted

in better information required ranking the pesticides (as

 

opposed to their uses) according to the EPA's definition of
risk and then comparing the quality of information between
pesticides. (Nu; first task was certainly easier than the

second.

—128—

The ranking according to risk was described earlier in
this chapter. But how was the information on dependent
variables put into a form that would be comparable between
pesticides? The only answer was to average the data across
uses for each risk and benefit variable to come up with one
number per pesticide per dependent variable.

There were several problems with this approach. First.
the number of uses varied from 28 (Dimethoate) to 2 (Ami—
traz). This meant that for Amitraz. the quality of infor-
mation for one use carried far more weight than it did for
Dimethoate. It is also possible that EPA had a far more
manageable task in Amitraz than it did in Dimethoate. so it
was able to obtain better information.

1\ second problem was that a few of the risk variables
varied between pesticides. For example. Endrin was the only
pesticide for which data on risk to nontarget organisms.
wildlife and endangered species were obtained. The Dimeth-
oate. Pronamide. Amitraz. Silvex and 2.4.5-T documents
measured risk to two or more different types Of ground
applicators. and there was no way to know whether to compare
handsprayers. tractor .applicators. boom .applicators. com-
pressed air applicators. DBCP application did not even
involve spraying. The decision was made to leave ground
applicators out (u? the analysis of quality of information
between pesticides.

A third general problem was that there was probably less

consistency between pesticides than within each pesticide.

-129-
In short. more confidence should be placed in the results of
the analysis of quality of information within a pesticide.
MEASURING CORRELATION BETWEEN DEPENDENT
AND EXPLANATORY VARIABLES

For each pesticide. correlations were measured between
four of the explanatory variables and the quality of each
"bit" of risk and benefit information for each HES of the
pesticide. As a result. each pesticide has as many poten-
tial correlations as there were dependent variables for each
Of the four explanatory variables (value of pesticide use to
manufacturers. percent of crop treated. annual per acre user
losses from pesticide cancellation and annual total user
losses from pesticide cancellation.

For the fifth explanatory variable. EPA's pesticide risk
ranking. the correlation between the variable and the qual-
ity of each "bit" of risk and benefit information for each

pesticide was measured. As a result. there were only as

 

many correlations as there were dependent variables.

Why were the four "interest group" explanatory variables
correlated with the quality of risk and benefit information
for pesticide pggg? There were several reasons for attempt—
ing to explain variation in the quality of information
between pggg of a single pesticide. First. it is a matter
of control over the research setting. Variation between

pesticides could be explained by a wide variety of varia—

 

bles. such as different levels of resources. different

researchers. different costs of information. and. Of course.

-130-
different levels of interest group pressure. But within a
single pesticide. most of those variables do not vary.
Thus. there are fewer non-hypothesized reasons for variation
in the quality of information between uses of a pesticide.
In short. the research setting is more controlled.

Another reason for examining variation between uses
rather than between pesticides is that there are some major
difficulties in measuring dependent variables (quality Of
information) for a pesticide. These problems were discussed
earlier in reference to the EPA's pesticide risk ranking
explanatory variable. In general. the average quality of
information across all uses of a pesticide may not accur-
ately reflect the overall quality of information for the
pesticide.

Finally. examining variation between uses of a pesticide
is interesting from a research perspective. Interest group
incentives vary at least as widely between uses as between
pesticides. But constraints (”1 decision-making presumably
do not. .As a result the research question becomes one of’
how EPA and interest groups allocate their scarce informa-
tion resources between pesticide uses within the particular
regulatory environment that is characteristic of an indi-
vidual pesticide.

There are two widely-accepted statistics used to measure
correlation between explanatory and dependent variables when

one variable is measured cardinally and the other ordinally.

-131-
The first statistic to measure the relationship between an
explanatory and a: dependent variable is called multiserial
correlation. This statistic assumes that one variable is
measured ordinally and the other on an interval scale. It
also assumes that the ordinal scale variables (in this re-
search. the quality of information variables) would approx—
imate a normal distribution if they could be measured more
precisely on an interval scale and that there is a linear
relationship between the explanatory' and dependent vari—
ables. Basically. multiserial correlation accurately trans-
forms the ordinal scale into an interval scale if the or-
dinal scale variables are normally distributed. Once this
is done. an adapted version of "Pearson's r" is calculated
to show the degree of correlation.

Goodman and Kruskal's coefficient of ordinal association
(called the "gamma coefficient") can be used either when
both variables are measured ordinally or when one is meas-
ured cardinally and the other ordinally. The gamma coeffi-
cient was chosen for this research for several reasons.
First. the explanatory variables were measured crudely. and
actually ordinally when the EPA described them qualitative-
ly. Given that. it would probably be risky to consider them
as being measured cardinally (which would be a requirement
of multiserial correlation). Second. there is no reason to
believe that the data for each dependent variable would
approximate the normal distribution needed for accurate

measures of correlation using multiserial correlation. The

-l32-
quality of information for a particular impact of pesticide
use tended to be incremetal between uses of a pesticide (and
thus. fairly constant). so there was not always a clustering
of data points towards the center of the scale.

A third reason for choosing gamma is that the measure of
correlation was intended to serve as a guide to the nature
of the relationships between the quality of information and
the various explanatory variables. Gamma serves that pur-
pose: multiserial correlation provides a more precise
measure that may not be justified by the data. Fourth. when
testing Edwin Johnson's statement that risk drives the risk-
benefit analysis. the explanatory variable (“risk of pesti-
cide“) and dependent variables were measured ordinally. so
multiserial correlation could not be used. The use of gamma
to measure the other hypotheses provided consistency in the
research. In short. use of the more complex and precise
multiserial correlation probably could not have been de-
fended because of the quality of the data collected for
explanatory and dependent variables.

A programmable calculator was used to calcuate the gamma
coefficients in this research. The data for dependent and
explanatory variables were entered by hand. and each gamma
coefficient was checked at least once (at least three times
if the first and second answers did not agree). To calcu-
late gamma. pesticide uses were ranked from highest value to

lowest value of a particular explanatory variable.

~133-

The formula to calculate the gamma coefficient is:
(number of agreements in direction of changes for explana-
tory and dependent variables - number of inversions) + (num-
ber of agreements + number of inversions). Gamma can range
from -l.O to +1.0: if there is no variation in the dependent
variables. gamma is indeterminant and cannot be calculated.
Examples of gamma coefficient calculations to explain varia-
tion between 3222 of a pesticide and to explain variation

between pesticides can be found in Tables 4.22 and 4.23.

 

Once all of the gammas were calculated. they had to be
evaluated. Ini general. non-zero gammas of less than .250
and greater than -.250 were considered to show very weak
correlations. given the quality of the data. Tables in
Chapters 5 and 6 show the frequencies of gammas with differ-
ent values for each dependent variable. For the four vari-
ables explaining variation in the quality of information
between pesticide 2335. the tables also show the number of
pesticides with primarily positive and negative gammas for
risk and benefit categories. These summaries of the vast
number of gammas calculated will serve as the basis for the

evaluation and interpretation of the results.

SUMMARY
The methodology described in this chapter is complica-
ted. not because of complex statistical techniques. but
because of the amount of data. the measurement difficulties
and the shortage of similar methodologies in previous re—
search. What follows is a brief summary of the methodology

used in this research.

-l34-

Table 4.22 Calculation of Gamma Coefficient

for Chlorobenzilate:

Value of

Pesticide Use to Manufacturers
and Consumer Risk of Oncogenicity

Value of Use

Consumer Risk —
Oncogenicity

 

 

233 to Manufacturers
Florida Citrus $3.220.000
Texas Citrus $406.000
Cotton $212.550
California Citrus $30.000
Arizona Citrus $24.000
Other $100*
Fruits and Nuts ?

 

* Value assigned for "little use”

Number of decreases in consumer risk of
oncogenicity as value of use decreases:

Number of increases in consumer risk of
oncogenicity as value of use decreases:
Gamma = 0-3 = = -1

_- :2
0+3 3

2
3

not applicable
3
3

not applicable
3

-l35-

Table 4.23 Calculation of Gamma Coefficient for
EPA's Pesticide Risk Ranking and Consumer
Risk of Oncogenicity

 

 

EPA Risk Consumer Risk -
Pesticide Ranking Oncogenicity
Endrin 21 not applicable
Chlorobenzilate 13 2.8
DBCP 13 3.0
Dimethoate 12 1.0
Silvex 7 0.8
2.4.5-T 7 1.0
Amitraz 7 3.0
Pronamide 6 2.7

Number of decreases in consumer risk of
oncogenicity as EPA risk ranking decreases: 10

Number of increases in consumer risk of
oncogenicity as EPA risk ranking increases: 5

Gamma = 10-5 = 5 = .33

10+5 l

-l36-

Explanatory variables were defined to reflect the incen—
tives of interest groups to involve themselves in the poli—
tical process. Four of the variables were measured for each
use of each pesticide. using EPA's Position Document 2/3s.
The fifth variable. ”EPA's Risk Ranking“. was measured for
each pesticide with EPA data.

The dependent variables reflected the quality of EPA's
information on risks and benefits for each use of each
pesticide. Measurement scales were devised by reading EPA's
position documents and benefits analyses. and values were
then assigned to the variables.

Correlations between the explanatory variables and the
dependent variables were measured by Goodman and Kruskal's
coefficient of ordinal association (“gamma”). In order to
make the dependent variables comparable with EPA's pesticide
risk ranking which was only obtained for each pesticide. the
quality of information was averaged across all uses of each
pesticide. Thus. one gamma was calculated for each depend-
ent variable. For the other four explanatory variables.
gammas were calculated for each dependent variable and each
pesticide. as correlation between explanatory variables and
the quality of information for uses of a pesticide were

estimated.

CHAPTER 5

RESULTS: QUALITY OF INFORMATION
ON RISK OF PESTICIDE USE

INTRODUCTION
There are three main Objectives of Chapter 5n First.
there are some general comments and qualitative examples of
EPA's risk assessment process. Second. the quality Of EPA's
risk information is summarized. And finally. the relation-
ships between the explanatory variables. as measured. and

the quality of EPA's risk information are summarized.

OVERVIEW OF EPA'S RISK ASSESSMENT PROCESS

EPA's risk assessments tend to be incremental in nature
from one pesticide to another (although not to the same
extent as the benefits assessments. due to differences in
risk between pesticides). As a result. there are some
general trends in the risk assessment process.

First. it is important to point out that EPA has a
definite stake in protecting the findings of its preliminary
risk analysis. The preliminary risk analysis initiates the
RPAR process and often serves as the basis for the final
risk assessment contained in Position Document 2/3. If the
preliminary analysis is not correct in its projection of
"unacceptable risk." then why was so much time (often two
years or more at the rebuttal stage of the process) and
money invested in studying the pesticide? Perhaps this

-137-

-138-
accounts for the apparent antagonistic relationship between
EPA and the registrants and EPA's own defensiveness. which
are evident even on the pages of the public documents re-
sulting from the RPAR process.

However. EPA's commitment to its preliminary risk analy-
sis does not extend to doing in-house toxicity and exposure
studies. at least for the eight case-study pesticides. No
doubt this is due to a lack of appropriated funds. but it
means that EPA has to rely on data from outside groups.
When registrants submit exposure or toxicity data that would
lower EPA's risk calculations. it does not seem to be
greeted by EPA with the same enthusiasm as data that
increases the risk calculations or benefits data from users.
For example. when Dow Chemical Company provided information
on applicator exposure to the herbicide 2.4.5-T during
forestry applications. EPA supposedly used the information
to revise its risk analysis. However. no evidence of re-
vision could be found in the public documents. One of EPA's
reservations in using the data was that the forestry use was
not being considered in the 2.4.5-T RPAR. so extrapolation
of the exposure data across uses would have to occur.
However. EPA often uses exposure data from other pesticides.
as well as other uses. in its risk analysis. The 2.4.5-T
risk analysis was no exception. It may be relevant to point
out that the 2.4.5-T RPAR was the most hotly contested of

the eight case-study chemicals.

-l39-

What about data from the groups which bear the risk?
Risk-bearing groups are. in general. very diffuse. unin-
formed and unorganized. Information on exposure and
toxicity of pesticides is very costly to Obtain (often
because such information has not yet been compiled). In
addition. risk-bearing groups often contain large numbers of
geographically-dispersed members. which results in high
organization costs. The theories of Stigler and Peltzman
would predict that these groups would gain little (or lose a
lot) in the decision-making process due to their inability
to impact regulators. However. as van Ravenswaay (1982)
points out. these groups' political costs may be subsidized
by ”political entrepreneurs.“ (see also Wilson (1980)) who
are able to influence decision-making. In the RPAR process.
the two major political entrepreneurs for risk-bearing
groups appear to be the Environmental Defense Fund (EDF) and
the EPA itself. Part of the EPA's activism is built into
the process -- since a determination of risk is needed to
initiate the RPAR process. EPA has a stake in defending its
risk findings. And while the environmental groups such as
EDF are not highly visible during all regulatory activities
for all pesticides. they may very well reserve their own
limited resources for controversial. high-risk pesticides --
the very pesticides which they should have the most success

in affecting regulatory outcomes.

\
_..-.‘ y
{I

/ ., j

Another characteristic of EPA's risk assessment process p////
is the need to bound the problem and develop standard oper-
ating procedures in order to cope with the potentially
infinite information requirements of pesticide regulation.
van Ravenswaay and Hull (1981) describe many of EPA's
standard Operating procedures when they discuss strategies
for meeting information requirements in food safety regula-
tion. In estimating risk. EPA usually determines toxicity .&’////
of a chemical from high-dosage animal tests rather than
human studies. Margins of safety are used when extra—
polating animal tests to humans. in case humans are more
sensitive to a chemical than the test animals. Exposure to
different groups of humans is estimated by using the same
sources all of the time -— Food and Drug Administration
”Market Basket Surveys" for pesticide residues in food. a
small group of studies on applicator exposure (Often data
from other pesticides are used). data from registrants and
user groups. and assumptions on exposure. Extrapolation of ,1
risk from high doses given to test animals to relatively low
doses experienced by humans is accomplished by using math-
ematical models of dose-response. In all of the case
studies for this research. EPA uses the model which usually
results in the highest prediction of risk. The “one-hit”
model assumes that very minimal doses of a chemical present
some risk of chronic illness.

It is difficult to estimate the direction of the bias of

L/'

these strategies on risk estimates. Risk estimation is

,* ffl\}‘)x
{/-l4l;/;J
\xw/fl/
characterized by such uncertainty that it is impossible to
know whether EPA's estimates are high or low. EPA's inclin-
ation is definitely toward erring on the side of safety. but
what if humans are much more sensitive to a chemical than
animals? What if exposure estimates are too low? What
about more sensitive members of the population or people
with above-average exposure? It is conceivable that a risk
estimate is too low even with EPA's bias toward safety.
Along with simplifying the information search. EPA also
reduces or "bounds" problems. as suggested by Edmunds
(1980). Problems are reduced by excluding certain impacts
from consideration -— EPA excludes the impacts of pesticide
use (N1 chemical manufacturing workers (who are supposedly
covered under the Occupational Safety and Hazard Act).
exposure to other chemicals which may enhance or lower the
impacts of the pesticide in question. certain relationships

ix: the ecosystem. auui other risk variables. All of this

represents an effort by EPA to make the problem of pesticide

regulation comprehensible and manageable. However. as willyJ/,x“‘

be pointed out in the next section. EPA's existing risk

information still leaves a lot to be desired. \\,//”‘

The internal structure of EPA's Office Of Pesticide
Programs (OPP) also reveals a lot about which impacts are
included in risk-benefit analysis and which are not. 'The
Hazard Evaluation Division within OPP. largely responsible
for the risk assessment. is composed of a Toxicology Branch.

an Environmental Fate Branch. a Residue Chemistry Branch

-l42-
and an Ecological Effects Branch. Figure 5.1 shows what is
ideally accounted for by each branch. but in reality many

impacts of pesticide use are excluded.

QUALITY OF EPA'S RISK INFORMATION

Table 5.1 summarizes the quality of more than 5.500 EPA
risk data. as measured by the researcher. In addition to
the data summarized. there are many other risk character—
istics that EPA did not attempt to measure for any of the
eight pesticides. as pointed out in Chapters 3 and 4.

Table 5.1 does not show high—quality information under-
lying EPA's risk assessments. In the three major measure—
ment categories. EPA ignores risk variables in the great
majority of cases. For the dietary exposure category. where
76.812 percent of the variables were not even mentioned.
most of the zero scores resulted from EPA's failure to
acknowledge dietary exposure as an additional residue source
for ground applicators. mixers and loaders. pilots. ground
crews. farmworkers. and persons exposed to spills or drift.
But in the second and third categories. the zero scores
resulted from EPA's failure to mention aspects of dermal and
inhalation exposure and risk to various groups.

The minor categories do not vary much from the three
major ones. Most of the measures are congregated around the
zero and one scores.

Another tendency reflected in Table 5.1 is EPA's avoid—
ance of quantitative range estimates. even for the risk

measures. When quantitative estimates were made. they were

-l43-

 

Component

HED Branch

 

Background
Chemical and physical properties
Environmental fate and persistence
Human exposure analysis
Dermal
Respiratory
Dietary (food and water)
Inhalation. penetration and
absorption rates
Human health risk
Cancer
Acute toxicity
Other chronic toxicity
Ecological hazard

Residue Chemistry
Environmental Fate

Environmental Fate
Environmental Fate
Residue Chemistry

Toxicology

CAG.* Toxicology
Toxicology
Toxicology
Ecological Effects

 

 

*Not a branch of HED.

Source: Environmental Studies Board Committee on Prototype
Explicit Analyses for Pesticides. Regulating Pesti-
cides (Washington. D.C.: National Research Council.

1980). p. 38.

Figure 5.1 Components of Pesticide
Principal Organizational Responsibilities in the

 

Risk Assessment

Office of Pesticide Programs

-l44-

 

 

 

 

 

Table 5.1 Quality Of Risk Information
Measure-

Variables ment Meaning of Fre- Percent

Measured Scale Measure quency Of Total

-Dietary Exposure 5 quantitative. 106 6.679%
(Residues. Food point
in Diet. Percent 4 quantitative. O 0%
Crop Treated) range

3 expert Opinion 94 5.923%
2 assumption 134 8.444%
1 qualitative 34 2.142%
0 ignored 1219 76.812%

-Dermal & Inhala- 7 study of pesti- 88 3.039%
tion Exposure cide. point
(Amount. Absorp- 6 study of pesti— 4 .138%
tion. Duration) cide. range

5 study Of similar 130 4.489%

-Residues & Amount situation. point
Of Food Consumed 4 study Of similar 4 .138%
for Acute Toxi- situation. range
city to Wildlife 3 expert Opinion 187 6.457%

2 assumption 273 9.427%

—Potential Exposure 1 qualitative 201 6.941%
for Nontarget Or- 0 ignored 2009 69.372%
ganisms

-Risk (Consumers. 3 quantitative. 172 14.576%
Ground Applica- point
tors. Mixers & 2 quantitative. 12 1.017%
Loaders. Ground range
Crews. Drift. l qualitative 429 36.356%
Spills. Pilots. 0 ignored 567 48.051%
Farmworkers.

Wildlife. Non-
target Organ-
isms. Endangered
Species)

-LD (Wildlife. 3 L05 . species 3 11.538%
Nontarget Or- 2 L050. similar 3 11.538%
ganisms. Endan- species
gered Species) l assumed 0 0%

0 ignored 20 76.923%

 

-145-

 

 

 

Measure-
Variables ment Meaning of Fre- Percent
Measured Scale Measure quency of Total
-Risk of Alterna- 2 quantitative 0 0%
tive Pest Control 1 qualitative 42 55.263%
Method 0 ignored 34 44.737%
-Likelihood of 8 study of pesti- 0 0%
Exposure for cide. point
Nontarget Or- 7 study of pesti- 2 12.500%
ganisms cide. range
6 similar situation. 0 0%
-Concentration of point
Residues for 5 similar situation. 0 0%
Endangered range
Species 4 theoretical model 1 6.250%
3 expert opinion 0 0%
2 assumed 0 0%
l qualitative 7 43.750%
0 ignored 6 37.500%

 

~146-
almost always point estimates. This does not necessarily
reflect greater precision in EPA's estimates and supporting
data -n- it could possibly reflect pressure from within and
without the agency to come up with definite numbers to aid
or justify decisions.

In general. it can be said that EPA's risk data for the
eight case-study pesticides were not complete. and when
present. not usually a result of quantitative studies Of the
pesticide. Instead. measurements tended to be based on
expert opinions. assumptions. or studies of other chemicals.

CORRELATION BETWEEN EXPLANATORY VARIABLES
AND QUALITY OF RISK INFORMATION VARIABLES

This portion of the Chapter examines the extent of cor-
relation. as measured by gamma (G). between the hypothesized
explanatory variables and the quality of information used by
EPA to assess the risk of the 86 uses of the 8 pesticides.
One of two results was found for each potential gamma. If
there was no variation in the quality of risk information
for all of a pesticide's uses that could be measured for a
particular explanatory variable. then there was no G -- by'
definition. it was undefined. Otherwise. a G in the range

of -l to +1 was found.

Explanation Of Tables. Although it would facilitate the

 

analysis of the research findings. it would be misleading to
use a single number to summarize all of these correlations.

Instead. Tables 5.3 through 5.7 were devised from Tables 1

-l47-
through 44 in Appendix B to summarize for each explanatory
variable:

O the numbers and values of gamma coefficients that
could be calculated for exposure and risk. within
each Of nine risk categories.

0 when gamma coefficients could not be calculated
(i.e.. when there was no variation in the quality
of risk data between uses of a pesticide). the
quality of risk information for exposure and risk.
within each of the nine risk categories.

O the total possible numbers of gamma coefficients
for risk and exposure. within each of the nine risk
categories.

O the number of pesticides for which one or more
gamma coefficients could be calculated. for each of?
the nine risk categories. Unless otherwise noted.
the number is the total out of a possible eight
pesticides.

o for each of the nine risk categories. the number Of
pesticides with primarily positive gammas. pri-
marily negative gammas. and equal numbers of posi-
tive and negative gammas.

The nine risk categories are consumers. ground applica-
tors. mixers and loaders. pilots. persons exposed to drift.
farmworkers. persons exposed to spills. aerial ground crews
and animals. Exposure refers to dietary exposure for con~

sumers: dermal. inhalation and dietary exposure for the

—l48-
seven other human risk categories: and animal exposure.
Risk refers to oncogenic. mutagenic. and fetotoxic and
reproductive risk to humans. and animal risk.
In describing the values of the gammas that could be
calculated. the following measurement scale was devised:
-1 to .-750
-.749 to —.500
—.499 to —.250
-.249 to O
.001 to .250
.251 to .500
.501 to .750
.751 to 1
This particular measurement scale was chosen in part because
the researcher felt that gammas between the values of -.249
and +.250 could not be considered strongly positive or nega-
tive. given the limitations in measuring explanatory and
dependent variables. Tables 5.3 through 5.7 show the fre—
quency of gamma values for exposure and risk within each of
the nine risk categories.
Gammas that could not be calculated were designated with
a "--". Since this reflects no variation in the quality of
information between uses of a pesticide. the number in
parentheses shows the quality of information for all uses of
a pesticide that can be measured for the particular explana-
tory variable. When a "--(x)" is left blank for a risk

category (as opposed to showing a zero). the number in the

—149-
parentheses is not included in the measurement scale for
those particular risk or exposure variables. In addition.
"—-(?)" means that the quality of information for at least
half of the uses of the pesticide could not be determined.

The "total possible G" line in each table is the sum of
the gammas that could be calculated and those that could not
be calculated. It is useful because it shows how Often (or
perhaps. how seldom) there was variation in the quality of
information between uses an) that gammas could actually be
calculated.

The lower half of each table reflects the number of
pesticides that showed positive. negative or ambiguous
relationships between the explanatory variable and the
quality of information for each of the nine risk categories.
This was determined by simply summing the positive and
negative gammas and comparing the two sums. These figures
show whether a large number of positive gammas in a risk
category reflect many positive gammas for one pesticide. or
a positive trend for several pesticides. It is the positive
trend for several pesticides that will reflect the kind of
correlation that was hypothesized between the explanatory
variable and the quality of information for a particular
risk category.

In the body of the table. the numbers in parentheses
denote the frequency of "lower range" gammas in each of the
eight ranges for the value of gamma. Lower range and upper

range gammas occur because for some pesticides. the uses

-150-
could only be measured in ranges for the explanatory vari-
ables. Using the high point of the range to rank uses
within a pesticide according to one of the explanatory
variables led to different values for gamma than if the low
point of the range was used. Therefore. "lower range” and
“upper range” gammas must be reported.

Table 5.2 is also useful in interpreting the results
found in Tables 5.3 through 5.6. This table shows the total
number of uses of each pesticide. and the number of uses
that can be measured for each of four explanatory variables.
Gammas based on very few uses of a pesticide are not as
reliable as gammas based on most uses. For example. for the
”value of use to manufacturer” variable only 3 of 11 uses of
Endrin could be measured. Therefore. all gammas measuring
the correlations between the value of Endrin uses to man-
ufacturers and the quality of Endrin risk/benefit informa-
tion are based on only three uses of the pesticide. On the
other hand. 17 of 18 DBCP uses could be measured according
to value of uses to manufacturers. Presumably. the DBCP

gammas would be more reliable than the Endrin gammas.

Findings for Value of Pesticide Use to Manufacturers.
The explanatory variable ”value of a pesticide use to manu-
facturer” is intended to proxy the likelihood of manufac-
turers to become vigorously involved in the information—
gathering phase of the regulatory process. Pesticide man-
ufacturers seem to fit the Stigler/Peltzman mold for a grOUp

that can impact the regulatory process. There are generally

-151-

.mms one as mchum
HHS-n you summon-“se.. o)» ecu sou-500 umcuumonmu ecu .oHcmHug amour-Baum come now owumngm
mum: moms hams soc mchHsuwumc no memonusa ecu mom .mumwe LOMus can named nocHs oucH ems gmchum
HHS-m: ecu DHHem mes-Huge cem- .monomusm uHumcmc mom £9639.- .mwms HH mmc fine-B 0303mm.“ ecu.

 

 

N
.wms oco oucH somemHHoo
mums hmcu cmsocu cw>w .moms wusuumH m HHm emucsoo umcoumwmwu ecu .mecmHum> 03D mnocu ou
mcHeuooom consumes meemcoum no men: we noses: ecu mcHucsoo cmcz .mco an uncommon condone who:
moms mosuuwH m HHm Keen-m8”. mono no accesses can sums-3035:8- ou ems no 8H8? mcHusmme-e CH

.xcdmumowm.>c1>Hco mums coch wosuuwH :0 men: one cOch uo m .nwms oH mac meEmcomm meHuHunwm ace H

me vs Os no mo chmusnmoz memo Hmuoe

N N N N N nmuuHE¢

N m N m m 91m .e .N

m o m N h xm>HHm

m m h w h mumHHNcwcouoHco

o m Hm HOH oH HmeHsmcoum

h s OH m HH NcHuecm

vN 5N oN vN mN wumocumEHo

5H mH mH 5H mH mono

.cmo «censuses couumaflmocmo emanate mumusuomuscmz ou saw >9 pmumsHm>m monoaumma

some nwmmoq mcHOHummm nouu no on: no wsHm> new: no uwcssz

ammo Hence scum monmoa ucmuuwm
mono shoe use

”cu mchuooud concern: memo mo umcEsz

neHcmHum> wuoumcdexm ou mcHeuooo< mcHuHumem comm now enhance: memo no uwcssz N.m chmH

It.

m
m

l:- Aerial
gun madam.
hpmnluk hpflnIuk

utuath-aflubninpnMMuahmuluurupthan)
Mun hum-u-
Nhu pudtswut
upmnluklwauaut human:

tum
Gnu-n Amuauu mdudus

Iwunluklwaueud annuluk

 

tfle&3tunneluuuuuubuuQOHBQOtudnua-dunsdlaud.Uheehmhauu

“6

«000000.!

000—00000

00000000

0000000!~

#0000000

#0000000

00000000

00000000

M0000NON

“00000.00

d0000fl00

00000000

H000000fl

Hoooeooq

00000609

00-002000“

GOOOOOGA

NOOOHOOA

agaofi§§
.. a s all.

{3333333
3 H
.iiifiiai

 

 

-152-

0000I-O0000

000000050-0

HOF0=0

00000006!‘

”----

AAAAAAAAA
U
N 00-00 «N ”8°

00000

ooHooHN3—o

H0000

-00N00'030

0100-000

N0d00fi0:0

H0—0N0

HOH0000a0

fl0h00

.00400040

---”---U-

1 1 2 1
hiapuwhilhtapuuhll

fi‘um‘hflkuu
htamuhh6)htapuuu7)

M

 

 

 

uninuuuc at
arummccmu
hedwhud
“tuna“
uMaruk
Raga-nu
mdunch
ducnnuflnn
upmnueud
maunch

“wan-nu

mpuncm

m

human-
flhu padtokut

Mun
manna: ufluwhu

Inauaua upmnluklwaaaud hmunfluklmuuent

Idmulnln(mmcsmpunmuudmaakarma-awn»

tad

tflehdlunqwtauuuu.bun—ODHQ1IM‘InI-lhuntd1upuuhl
Gnu-n

 

OHOHOHOO

OHHnHNOG

oooeoooo

00000d00

00000fld0

#0000000

00000000

000000dfl

00000Hdfl

000000d0

00000NNN

"FOOHOHOCO

EOOBOHOH

“#00fl000

Aha“

'oe'voo'
N 3 °

NO0HOH00

 

 

-153-

00N00

OONOONQRH

d0dz0

00000003H

---”-

oonoonnsu

«ONVO

00N000fla0

HOMOO

NOHO0F030

HOHOO

d0d000030

fl0h00

------

”U”- ---

 

 

M

hulhuuhc It

 

mhuothukuu

S

C

3

3
Hi 0

C

26.
Ni N

3

i

3.
0 N
B n
0 N

saga-nu

u’uuch

2

dueunnfluu
«manna-a
U’flflfl?

-154-

o u o o o a a a o ..o .58...

 

 

 

 

...... 9:38.. no
. is. «It 51.

o o a a o o o . o u so 338!
adult-— 5:.

a a o a n a a a v .6 8318
ashlar. 51.

233-85.3.3931-8. ...-3138352.! Inna)!-
~ « a n n a n o ... 2589.. .3
33.28 ...

3:8 o 62.. .8

13oz...- uo as]...

a o a 8. 3 2. 3 2. 2 8 «a 8 3 3 8 8 3 8 9.888.... 38
o o o I 3 a. a 3 a a s 8 3 2. v a a 3 g
o o o a o a 8. o 8. o o a o o o 8 o a o v E I
o o o B 3 6 a. n .3. a v 8 v a a 3 o 3 o a 3. I
o o o o u o .3 a .3 a a a a a a a a a a u 3 I
o o o o o a 8. o E a o a o v o a o c o v a. I
o o o o a o 5 a .3 n o o a o a o a a v a a. I
o o o .o. o o o o o o .3 I

o o o 5 a a a a n v a. I

o o o 8. o o o o o 3. I

o o o 3 a o o a a E I

a n » ... 2 v 3 a o 3 3 n o 33.3

n I a a n

 

3. n a 3 as.
3. O 80. 3 3m.

3. o 3. o ..N. N 2.. h 3. o 3. o 3. N 3. N AN. N 3. o 3. o 3. N 3. a 3. a 3. a an. A M u
w 3. 0 8n. 3 3N.
o

3. o 3. n 3. o 3. o 3. o 3. o 3. a 3. o 3. o 3. o 3. o 3. o 3. o 3. o .N. o 3. o

3. o 3. N 3. o 3. o 3. o 3. o .N. a 3. N AN. n .N. n 3. 0 AN. 0 3. c 3. o 3. n :4. a A
.N. N .N. n 3. o 3. o 3. o 3. a 3. o 3. o 3. o 3. a 3. u an. N 3. n 3. h 3. o 3N. nu a 3. n SN. 3 «8.
3. a 3. N 3. o 3. o 3. o 3. o 3. o 3. o 3. o 3. u 3. o .N. o 3. n 3. u 3. v 3n. «N 3. 3. N o 8 a...
3. o 3. o 3. n 3. o 3. u 3. u 3. o 3. o 3. o 3. o 3. n 3. «a 3. o 3. o 3. o 3. o 3. o 3. a 8N... on a”...
3. o 3. o 3. o 3. o 3. o 3. o 3. a 3.. n 3. o 3. o 3. o 3. o 3. o 3. o 3. o 3. o 3. N 3. u g... 3 2.:
3. o 3. o 3. o 3. o 3. o 3. o 3. n .N. N 3. o 3. o 3. o 3. o 3. o 3. o 3. o 3. 0 A... o 3. a 02.... 3 7

 

gull!!- 33 a; a; a; gill} a; gallon] a; 03
3|?- Ildoflfldgsoll 53833338.. 33: 2.8.8!- 88838- lilac .3!
3n.- uli nan-888 ...—I... I!!!

.§33§%5§.!35
igizgigaikigéfii «Jail

~155-

AH ognamwoa 9 ac.
a

3 3398a o ”3

a a n

N

as caauuooa a any Au can“

saoug
~

83888 no ~83:

o o 0 Be 2.3.9..
9... 338: no
Eillialfig

o N a 30,358
zquhfig

n v v -..o .2538.-
btlvu. .3:

138?:

n w n .iiuaha
. 0338708

3350fi53kfl

 

3

a

z

a

3 «0 ON a a on Osa'gnaa

 

O
O

9
2
8'
:1
33
a

h

‘ 8

On 1888

0
III
g
V
Ff
on
H
H

 

00000

0 000000000

sen
Rvfl
3v~
:vo
avw
Svo
ave
ave

OOOONHO O

CONDO

OOOMOOOO

8. o 8. o o
a... n 33 on v
.8 n .3 n
8. o .2 u
8. H .3 ~
.2 o
3 a
.8 o
3» n

O OONOONnR—a
HOHHO
an
H OOOOOMOBfl
00“

H
M
F
M

OOOGOOOO
OOOdOOOO
OOOHOOOO
HOOOOOON
MOOOOOO'
OOOOOOHN

'0 OOMOOMnso-o

GOOD-«OHM

HONVO

OOOMOOHO

3 bONOOQNgO

OOOGONNN

3*!
3?‘
3f]
3*l
37'
37'
3.-
37'

NOWDO

O noncoa~33
NONOO

9 HOmONDsz.
QOhOO
VOHVNHV

n
.‘9.
n
o

Asap.
omosnao
oaosnn.
SQOHSQ

Gama?
Qflfsmcf
aquamxf

afitsul

OOOOVOOH
OOOOflOOH
OOONFOOF

A

II.

o-O

U

n

d
000MHNOM

xfluegag {Hugsflu {HIBSQHxaugaafi {Hollauxflagaaa 1393899133838” ‘ansanu Duo
3.55 8209.38 3:..u3u88 gangs—88a 303.. 83

#324 raise; Ialsal

Eildil .332 .3538 .3,
83? Sixv

:8g3ogfluggc88033553n3g8i3.Ildixiii
J‘Sulaifw135u513a*§303330813a8327§899bllu0603'

-156-

few major manufacturers of a pesticide (meaning low organi-
zation costs) and the financial stakes can be very high.
Stigler and Peltzman have hypothesized that the more the
regulated parties have at stake, the more they will affect.
the regulatory process. In this research, it is hypothe-
sized that manufacturers of a pesticide will improve the
quality of information on the benefits and risks of the morte
valuable uses of that pesticide. Likewise, it is hypothe-
sized that the manufacturers will spend less time and fewer'
resources in attempts to gather information on risk for less
valuable uses of the pesticide.

The results do not, in general, support these hypo-
theses. For seven of the nine risk categories, a vast
majority of the gammas could not be calculated (that is, the
quality of information in those categories did not usually
vary between uses of the various pesticides). The two risk
categories for which a majority of gammas could be calcu-
lated, ground applicators and animals, showed two different
results.

The ground applicator category showed 25 of 42 positive
gammas in the .751 to 1 range, which indicates strong posi-
tive correlations. Only 14 gammas were not positive, of
which 13 were in the -.249 to 0 range. Furthermore, all six
of the pesticides for which one or more gammas could be
calculated had primarily positive gammas. Ground applicator
risk is the most important risk category in EPA's studies,

with the possible exception of consumer risk.

-157—

The animal risk category showed mixed results for tflue
one pesticide to which it applied, Endrin. There were equal
numbers of strongly negative and strongly positive gammas:
and the remaining gamma was in the -.249 to 0 range. As a
result, Endrin just barely ended up in the primarily nega-
tive group for the animal risk category.

About all that can really be said for the remaining risk
categories is that gamma could not usually be calculated.
When it could not, the quality of information across all
uses of a pesticide was usually zero (i.e., the risk was
ignored), except for consumers. This is partly because
dietary exposure for all non-consumer human risk categories
was usually ignored.

Thus, the hypothesized relationship was not supported by
empirical results. Quality of risk data tended not to vary
at all, and when it did, the results were ambiguous when
compared with the value of pesticide uses to manufacturers.
What could cause these results? One potential cause could
be that manufacturers do not want to see better information
in EPA's risk analyses. Vague or poor information could
result in a better regulatory outcome for manufacturers. A
second potential explanation is that pesticide manufacturers
are unable to impact EPA's risk analyses. This explanation
would seem to fly in the face of Peltzman and Stigler's
work. A third potential explanation is that the manufac-
turers allocated their scarce resources towards securing

better information on ground applicator risk because EPA

~158-
spends most of its resources on that risk category. This
would make even more sense if the manufacturers perceived
that EPA based its decisions on findings of ground appli-
cator risk. The manufacturers may simply feel that they get
the most value for their information expenditures by con-
centrating on the high-risk ground applicator category, and
then saving the rest of their resources for improving the

quality of benefits information.

Findings for Percent of Crgp Treated. The ”percent of
crop treated” explanatory variable is intended to proxy the
number of pesticide users affected by EPA's regulatory
actions. It is hypothesized in this research that as a
larger number or proportion of users are affected, the
quality of risk information will increase as those users
impact the regulatory process. Stigler and Peltzman might
argue that the organization costs of larger groups might
hinder the ability of the groups to affect the process, but
it is hypothesized that the presence of political entrepre-
neurs such as USDA and well-organized user groups eliminates
organization costs.

Once again, the results do not seem to support the
hypothesis. The only risk categories for which a majority
of the total possible gammas could be calculated were ground

applicators and animals. When G could not be calculated,

 

the quality of risk information was usually zero for every
risk category except consumers (that is, the particular

exposure or risk variable was usually ignored by EPA).

~159-

The numbers of pesticides with primarily positive, nega-
tive or mixed gammas did not suggest a particular relation-
ship between the percent of crop treated and the quality of
risk data for any of the nine risk categories.

What could cause these results? Perhaps pesticide users
do not necessarily want to improve risk information,
especially if better information is perceived as leading to
more restrictions on pesticides. Pesticide users may also
be ground applicators and the profit motive may overrule any
desire to see stronger information and regulations to reduce
ground applicator risk. Users certainly have a better
handle on benefits information, so they may focus their
efforts in that direction, instead of trying to influence
risk information. Finally, EPA probably attempts to defend
its risk analysis from external information that undermines
the findings. 'This should be more true for risk than for
benefit information because USDA has major responsibility
for preparing the benefits analysis.

Although the overall results showed no clear trend when
G could be calculated, that was not the case for individual
pesticides. Some pesticides showed primarily positive Gs
for a risk category, others showed primarily negative Gs and
still others showed equal numbers of positive and negative
Gs. The two pesticides with the most measurable uses for
percent of crap treated showed opposite results. The per-

cent of DBCP uses treated was positively correlated with the

-160-
quality of DBCP risk information for the three risk cate-
gories for which DBCP Gs could be calculated. On the other
hand, Dimethoate showed negative Gs for the five risk cate—
gories for which Gs could be calculated. Apparently, the
percent of crop treated influenced risk information within
some pesticides, but not others.

But the overwhelming result for this explanatory vari-
able was that there usually was no variation in the quality
of risk information between pesticide uses. In general,
when there was no variation, the quality of information was

zero. Thus, EPA tended to ignore many aspects of risk.

Findings for Annual Per Acre User Losses. The "annual

 

per acre user losses” variable was intended to proxy poten-
tial per capita user losses from pesticide cancellation. In
accordance with the Stigler and Peltzman theories, it is
hypothesized that attempts to influence the regulatory
process, and thus the quality of risk information, will
increase as per acre user losses increase.

The results, as shown in Table 5.5, are somewhat more
positive than for the previous explanatory variables. For
all risk categories except persons exposed to spills, there
appears to be some relationship between per acre user losses
and quality of risk information. One good reason for these
results could be that EPA gathers good information to defend
the risk of those pesticide uses with higher potential per
capita user losses. Greater economic benefits presumably

mean fewer restrictions for a pesticide use, unless EPA can

-161-
show a greater likelihood of risk. Another potential ex-
planation which is more in keeping with Stigler and Peltzman
is that pesticide users become more involved in the regula-
tory process as per capita stakes increase, resulting in
better risk information for the uses with higher annual per
acre user losses.

But still, in most cases, there was no variation between
pesticide uses in the quality of risk information. More
often than not, EPA ignored risk variables -- the quality of
risk information for pesticide uses that could be measured

according to annual per acre user losses was not good.

Findings for Annual Total User Losses. The final explan-
atory variable to describe variation in the quality of
information between BEE of a pesticide was "annual total
user losses." This variable is intended to measure the
potential industry-wide impacts of pesticide cancellation.
The hypothesis is that as the aggregate value of a pesticide
use to users increases, the quality of risk information will
increase as pesticide users or their political entrepreneurs
become involved in the regulatory process. It should be
noted that there is a difference between per acre and total
user losses. It is often the case that a low-volume, low
total user losses use of a pesticide has high per acre user
impacts. Examining both variables can provide insight into
whether EPA and pesticide advocates react more to aggregate

or individual economic impacts.

~162-

The results in Table 5.6 are similar to the results for
annual per acre user losses. Based on numbers of pesticides
with primarily positive Gs, there appears to be a positive
correlation between annual total user losses and quality of
information for six of the nine risk categories, and for the
two major categories (consumers and ground applicators). In
addition, where there are pesticides with primarily negative
Gs, the negative Gs are almost always in the weakly negative
-.249 to 0 range.

Again, EPA may seek to obtain better risk information
when it knows the risk findings will be balanced against
high potential benefits of a pesticide use. Likewise, the
pesticide users themselves will encourage better information
if they feel that the information will result in findings of
lower risk. If the latter explanation is true, it appears
that the user groups can overcome any political organization
costs in many instances -— this could reflect the presence
of political entrepreneurs or already-organized user groups.
However, it is possible that organization costs prevented
user groups from having an even larger impact on the quality
of information: hence, the failure to see more positive Gs.

As for all of the other variables, G could not be calcu—
lated in most cases due to no variation between uses in the
quality of risk information. When G could not be calcu-

lated, the quality of risk information tended to be zero.

—163-

Findings For EPA's Risk Ranking. One explanatory vari—

 

able was used to predict differences in the quality of

information between pesticides, as opposed to uses of a

 

pesticide. That variable was the relative risk of each
pesticide as measured by EPA's own risk ranking method. The
intent of measuring the correlation between the EPA risk
ranking and the quality of risk information was to test EPA
Office of Pesticide Programs Chief Edwin Johnson's statement
that the quality of the agency's risk-benefit analyses is
determined by a pesticide's relative risk.

Table 5.7 shows the results for this explanatory vari—
able. There were some difficulties in averaging the quality
of risk information across pesticides, mainly because risk
categories were not consistent between pesticides. For
example, gamma could not even be calculated for the ground
applicator risk category, because several of the pesticides
had more than one type of ground applications. Supporting
tables for Table 5.7 can be found in Appendix B (Tables 37
through 44).

The results were a mixed bag. There were five risk
categories showing primarily positive gammas, and two show-
ing primarily negative. Across all risk categories, 25 of
45 positive gammas were in the weakly positive quartile
(.001 to .250). Very few gammas could not be calculated (13
of 78 total gammas). Seven of those 13 occurred in toxicity
and the risk portion of each risk category because only one
pesticide, Dimethoate, was considered to have a risk of

mutagenicity -- therefore, there could be no variance in the

. in? is. .3896 «as. 8:89“ gas. 3:898 .32 c.5896 is. 838..» .12 93896 .9339
A 33:. 8328

-164- ,

GOOHHOOO’H'

n
' a
n

O .

M

23.8 no :3» sun-Em

n

h

n

O
n

M

M

023008.

as. 2:308
no unis. ~86.
0:33.:

.33-on .

a 3838.. #38. ..

 

M 0.0 0,0

.2 u
.8 M

000000

A

O 0

000000
OOOON

o

o

OOOOO—IOO

OOOOHo—c—a

o

OOOUDN

o

o

OONOOOO—fl

OVOONHOOO

OOOONOO

o

OOOMDOOOO
OOOONOO—I

o

OOHHOHOOO

O

dHOOOOOo-O

o; 3 Sp.
92.. 3 Sn.
8a.. 3 3d.
8.... 3 H8.

o 3 2a.-
8a... 3 8...-
8a.- 3 at..-
as... 3 o4-

 

 

.8003

35632 8

a 3.5: 00395 2.030..

.8833 can 0.3qu no 35 9.008 a 30:3 030523 5 acid:

£02590

3»:

I:

undue 9.
ocean-nu 38!»...

3'83 ~35:

c. an 33,

838585 is. a... 33.3 888.. 1.» ...... 25.3. is. can. 5.38 .83-188 no Ella «.... .38

—l65-
quality of information. fNua other five that could not be
calculated under risk from accidents and spills represented
average quality of information of zero across all pesti-
cides.

In short, Johnson's hypothesis is not supported. There
could be several reasons for this. One is that he could be
just plain wrong. ,Another could txa the method of testing
his hypothesis or the data constraints in this research. A
third possibility is that EPA does the best it can in ob-
taining risk information, but it is forced to work under the
technological constraints hampering all risk assessments. A
final explanation could be that risk §£e_s_ drive EPA's in-
formation-gathering process -- within the constraints of
interest group pressure, a la Stigler and Peltzman. Johnson
may be right about the internal workings of EPA, but he may
be less than fully aware of the external constraints on

agency operations.

Summary; Correlation Between Explanatory Variables and

Quality of Information on Risk. In general, the hypo-

 

thesized relationships between the explanatory variables and
the quality of risk information did not hold up well. In
fact, some of the evidence even showed results opposite of
what was expected. And in many cases, the information was
uniformly bad.

Some of the major reasons why these general results may

have occurred include:

/

O

,/

1

fi 6 6 -/

,/ /
Interest groups, {BECK as pesticide manufacturers
and users, may not want better_r_i_s_l_<_ information.
Therefore, they will not try to influence the in-
formation gathering process, other than to try to
sabotage it.
The risk assessment methodology is often more art
than science. This could be one cause of the poor
risk information.
The EPA may not have the necessary budget to obtain
good data. In general, EPA only receives toxicity
data from pesticide manufacturers. Exposure data
are EPA's responsibility, and they can be very
costly to obtain. Even if manufacturers do provide
exposure data, EPA may discount the value of those
data.
Interest groups may prefer to allocate their scarce
resources toward improving benefits information, in
the hope that better information on a pesticide
use's benefits will lead to less stringent regula-
tions. The manufacturers and users may feel that
they have a much better chance of making an impact
(Hi the benefits side than on the risk side. The
results outlined in Chapter 6 would seem to support

this hypothesis.

CHAPTER 6

RESULTS: QUALITY OF INFORMATION
ON BENEFITS OF PESTICIDE USE

The EPA relies heavily on professionals from USDA, state
agencies, and land-grant universities in the preparation of
the benefits analysis. This chapter explores the quality of
the information contained in the benefits analyses, and
provides some insight into the power of different groups to

influence the quality of information.

GENERAL OBSERVATIONS ON EPA'S BENEFITS ANALYSES

Various groups provide input via the comment process and
other informal mechanisms on the benefits of pesticide use.
EPA often uses this input, especially when the groups are
pesticide user groups. Several examples can be cited to
illustrate this point.

131 the case of the pesticide Endrin, the agency stated
in position documents that undefended uses of the pesticide
were to be canceled -- the apparent assumption was that if a
use was undefended, there must be no benefits from that
particular use of the pesticide. This, of course, ignores
what Peltzman, Stigler and others have worked so hard to
point out -- that there are costs of using the political

process.

~167-

-l68-

In the case of Chlorobenzilate, a similar attitude was
present with the use of the pesticide on citrus fruit in
Arizona. Although Arizona citrus growers face similar pest
problems and growing conditions as growers in Texas, EPA
initially canceled Chlorobenzilate use on Arizona citrus,
while retaining its use, with restrictions, in Texas. EPA's
reason for the Arizona decision, as reported in position
document 2/3, was that there must be no benefits of use
since Arizona growers did not respond to a USDA survey on
the subject. Upon protest from USDA and the eventual re-
ceipt of information from Arizona, EPA reconsidered and
decided to allow restricted use of Chlorobenzilate in
Arizona. It is possible that EPA was attempting to appease
groups opposed to Chlorobenzilate by canceling the "use of
least resistance" -- in this case, use by a small group of
producers with potentially high costs of articulating
preferences to EPA. Only when USDA played the role of
political entrepreneur was the use of Chlorobenzilate on
Arizona citrus salvaged.

Another example of user impact on the regulatory process
involves a small group of users of the nematocide DBCP with
high per capita benefits of use and an already-existing
political organization called the Pineapple Growers Associ-
ation of Hawaii (PGAH). PGAH provided information on the
risks and benefits of DBCP use on pineapple, which EPA used
extensively in its risk-benefit analysis. While many uses

of the pesticide were canceled, the use on pineapple was

-l69-

registered with restrictions. Of course, the registration
may have been approved for reasons other than PGAH's dialog
with EPA. This example not only illustrates the potential
influence of a user group on the quality of information in a
pesticide risk-benefit analysis, but it is also a fine
example of EH1 already-organized interest group with access
to the regulatory process.

EPA simplifies its information search by bounding the
problem and the variables to be measured (as pointed out in
Chapter 3). For example, the economic impacts of pesticide
cancellation on pesticide manufacturers are either excluded
or only tmiefly touched upon ix! EPAHs benefits analysis.
Total compliance with proposed restrictions is always
assumed, and "expert opinions" are heavily relied upon as a
source of data. Changes in prices of agricultural products
and in related goods, resulting from pesticide regulation,
as well as changes in prices of inputs other than the pest-
icide, are often dismissed.

As with risk analysis strategies, it is difficult to
estimate the impact of these strategies on the accuracy of
EPA's benefits information. The assumption of total com—
pliance overstates the costs of pesticide regulation, but
this may be compensated for by EPA's failure to quantify the
effects of price changes. Also, EPA has a tendency to
implicitly attribute all potential losses in benefits due to
cancellation to pesticide users without acknowledging that

consumers and pesticide manufacturers will absorb some of

~170-
those losses. This overstates the benefits of pesticide use
to user groups although it does not result in an overstate-
ment of total benefits.

Again, the internal structure of EPA's Office of Pesti-
cide Programs sheds some light (”I the impacts included in
EPA's Benefits Studies. The Benefits and Field Studies
Division of OPP is responsible for the benefits analyses,
and contains an Animal Sciences and Index Branch, a Plant
Sciences Branch and an Economic Analysis Branch. Figure 6.1

shows what is ideally accounted for by each branch.

THE QUALITY OF EPA'S BENEFITS INFORMATION

Table 6.1 summarizes the quality of information obtained
by EPA for its benefits analyses for the eight case study
pesticides.

The data for the two major benefits categories measuring
changes in pest control costs and value of production were
noticeably clustered around the lower end of the quality of
information scale. Well over half of the variables were
either ignored or only qualitatively mentioned. Another
35.9 percent of the variables were quantified on the basis
of assumptions or, in most cases, expert opinions. Only 6.1
percent of the 773 pieces of data were based on some sort of
study. It is obvious that EPA does not usually rely on or
conduct studies on the benefits of pesticide use.

The prices of goods for which the pesticides are an
input were measured on a different scale than the other

variables. Prices were ignored more than half of the time,

-171-

 

BFSD Branch

 

 

Insect icides/ Herbicides/
Component Rodenticides Fungiches
Current use analysis
EPA registrations of RPARs ASIBa PSBb
and alternatives
Recommendations for use of ASIB PSB
RPAR and alternatives
Use of RPAR and alternatives EABC EAB
Performance evaluation of RPAR
and alternatives
Pest infestation and damage ASIB PSB
Comparative performance ASIB PSB
evaluation
Use impact analysis EAB EAB
(projected change in use)
Economic impact analysis
Impact on production cost EAB EAB
Impact on volume produced EAB EAB
Impact on consumer prices EAB EAB
Aggregated economic impact EAB EAB
Limitations of analysis EAB EAB

 

a Animal Sciences and Index Branch.

b Plant Sciences Branch.

Economic Analysis Branch.

Source: Environmental Studies Book Committee

on Prototype

Explicit Analyses for Pesticides, Regulating Pesti-
National Research Council,

cides (Washington, D.C.

1966), p. 38.

 

Figure 6.1 Components of Pesticide Benefit Analysis (for a
given site) and Principal Organizational Respon—
sibilities in the Office of Pesticide Programs

-172-

 

536 use 5386

 

 

 

 

3.84 a 5.5 832. 5.33358 a
banal-co uo 3&3
ammo; A fit. 833 .335358 A
53938 no 32390
38;... n 51. 32.3 .833353 e
83.8 8 e58 :63. .833356 m
8 o 8:8 :63. .9333; e
8 o .258 toes 5333.56 n
00 O 33 .2583an N
38..." n 3835.. 38332.8 a
33:3 8 e899 o 88 we 8:5 n
”.58 .36 02333 no
886 S .63.. 8 e83 .333358 a.
2.39532 53 £3.13 ..
.88 .38 33:8,. 8 .3338 53 {said ..
3 o 33- 8 ...-B 533358 e 358.32 51. £33» s
. . .3338 5a.. <23: n
v.34.” «3 .338 3.98 .833353 n 335.: .83 ..
. . . 33953:}. on»: 38¢ I
68% en also. £333.30 ~ 2.3.83? no <8: ..
usgnsnsfi we :8 a
.338 8a 3835.. 333323 a . 3.8:. 38s ..
. .3338 no {one u
. 2.2.8 a3 6883 e “.2388“qu «o .38 a
38. no 0 a” 2530: ac «mace-t 01.8 A 333...;
7 agent

Saul-85 3388 no 33.8 no i a.» .38 .

-173-

 

”abhomd

. 883

a~mm.m~

«mmm.mv

 

Heap? No a

«2 .

mam _

Hhm

303333.. no 380 I
85:98 -
.385 338 I
30:9: 0% I
3085 38388 I
389: no 333333“. ..
808 35o 30
8% :0 Bug I
380 8383.6
.350 S 8:20 ..
. 8383 no
358 ~23.33.35:..38

 

 

 

m 81> 5 «9.20 ”Boa ..
83838 we 83>
.368 633838 N 5 856 82 .38 .. .
38 33:8 33;
83338 H S 0936 ~38. ..
330 ”83:8 38
080:3" o 5 8:98 80¢ you I.
8:362 quflmmIoo: 300m 33039
uguao:

.3538. 83.88:: 3388 no 33.8 no Ella a.» .36

-l74-
but when they were not ignored, they were generally based on

some type of a study. However, variation of price with

 

quality and/or supply of the good was rarely considered.

Finally, several variables were measured on a 0-3 scale.
These variables were either summary variables (changes in
pest control costs, changes in value of production) or else
they were poorly documented so that all that could really be
said was that they were ignored, qualitatively measured or
quantitatively measured. Once again, nearly half of these
variables were ignored. Another 28.3 percent were only
qualitatively mentioned. Only 22.3 percent of the 1459
pieces of data were quantified.

CORRELATION BETWEEN EXPLANATORY VARIABLES
AND THE QUALITY OF INFORMATION ON BENEFITS

This section of the Chapter examines the extent of
correlations, as measured by gamma, between the hypothesized
explanatory variables and the quality of information used by
EPA to assess the benefits of the 86 uses of the eight pest-
icides. There were 244 potential gammas for each explana-
tory variable, but one of two results were found. If there
was no variation in the quality of benefits information
between all uses of a pesticide, gammas could not be calcu-
lated. Otherwise, a gamma within the range of -l to +1 was
calculated. Of the total 976 potential gammas for four
explanatory variables, 274 of them could not be calculated.

The following tables 6.2 through 6.5 are of the same

format as tables 5.3 through 5.6. Each table represents a

-175-
summary of the correlations between one explanatory variable
and the quality of EPA's benefits data. For each explana-
tory variable one of the tables shows:

0 The numbers and values of the gamma coefficients
that could be calculated for each of five benefits
categories.

0 The quality of information within each of the five
benefits categories when gamma coefficients could
not be calculated (i.e., when there was no varia-
tion in the quality of benefits data between uses
of a pesticide).

o The total potential numbers of gamma coefficients

 

within each of the five benefits categories.

0» The number of pesticides for which one or more
gamma coefficients could be calculated for each of
the five benefits categories. Unless otherwise
noted, the number is the total out of a possible
eight pesticides.

o The number of pesticides with primarily positive
gammas, primarily negative gammas, and equal
numbers of positive and negative gammas.

The benefits data are delineated according to five bene-
fits categories: changes in pest control costs, changes in
value of production, other socioeconomic impacts, distrib-
ution of impacts, and compliance costs. In the tables, they
are arranged from left to right in decreasing order of

importance in EPA's benefits analyses.

~176-

The same measurement scale used in tables 5.3 through
5.6 was used to describe the values of the gammas that could
be calculated for the quality of benefits data. That scale
is:

-1 to -.750

-.749 to -.500

-.499 to -.250

-.249 to O

.001 to .250

.251 to .500

.501 to .750

.751 to 1
Also, as in tables 5.3 through 5.6, gammas that could not be
calculated due to a lack of variation in quality of benefits
data were designated with "--". The number in parentheses
shows the quality of benefits information across all uses of
a pesticide that could be measured for a particular explana—
tory variable. The "total possible G” line is the sum of
the gammas that could be calculated and those that could
not.

The lower half of each table depicts the number of pest-
icides that showed positive, negative or ambiguous relation-
ships between the explanatory variable and the quality of
information for each of the five benefits categories. The
figures show whether a large number of positive gammas in a
benefits category reflect many positive gammas for one

pesticide, or a positive trend for several pesticides. The

-l77-
latter would reflect the kind of correlation hypothesized
for the explanatory variable and the quality of benefits
information.

Once again, Table 5.2 should be used to show the number
of uses of each pesticide that could be measured for each of
the four explanatory variables. Supporting tables for
Tables 6.2 through 6.5 can be found in Appendix 8 (Tables 45

through 67).

Findings for Value of Pesticide Use to Manufacturers.

 

The intent of using the value of a particular pesticide use
to the manufacturers as an explanatory variable was to proxy
the likelihood of manufacturers to get involved in the
regulatory process, due to their potential losses from
regulation (M5 the pesticide use. The hypothesis was that
the quality of information would improve as the value of a
pesticide use to the manufacturers increased.

The results shown in Table 6.2 seem to support this
hypothesis. For all of the benefits categories except
"compliance costs," a majority of the gammas could be cal-
culated. When there was no variation in the quality of
benefits information and G could not be calculated, the
predominant quality of information ranged from three (expert
opinion) for the major category of "changes in pest control
costs" to zero (ignored) for the minor categories of "other
socioeconomic impacts" and "distribution of impacts."

Most pesticides for which gamma could be calculated

showed predominantly positive gamma coefficients across all

-l78-

aunt:su-qwcauunumluuuqunwocumaununuuum-ucuuum-m-a»
hnhauu

“kid
6

QMNIQGI

manage-

Chnuiuhu cunninum- «hrhdr' ouuuuun uqmwu.
«Pumahn bmuulnuu «luau

 

.100 a .07”
-07. b .0”
-0” b .03
.0” :0 0
.miu.ao
.aiu.wo
.QID.”O
.731 a 10°

Rauuuoou

BOOQUUOQ

0P09HNNU

Conunwoa

”0000000

 

t

 

-(fl
—Wfl
—40
-19
-M)
—4M
-Wfl
—(U
-wm
-(fl

anuoUoo

N000~l00000

oUwou

ocoow

OUMON

 

u

G

8

 

bulb-who

 

flrHM$GGMd
boauuud

mama-nu
pdunch

«mad-nu
mpunch

‘mmqnnnnnn
otpuuns
mpdnch

unnotuuuu-

3E

-l79-

five risk categories. Only three pesticides ever showed
primarily negative or equal numbers of positive and negative
gammas: Endrin, Amitraz and Silvex. For Endrin, only 3 (of
a possible 11) uses could be measured according to value to
manufacturer. Amitraz and Silvex each had only two measur—
able uses. 80 all of the pesticides with larger numbers of
measurable uses, including Dimethoate and DBCP with 24 and
17 measurable uses respectively, showed primarily positive
gammas across all five benefit categories.

If these results occurred because pesticide manufac—
turers are able 1x) influence the regulatory process, then
manufacturers must be convinced that improving benefits
information will help them achieve more favorable regulatory
outcomes for high-value uses of a pesticide. But there
could be other reasons for the correlation between value to
manufacturers and quality of benefits information. One is
that benefits information is easy to obtain, relative to
risk information, and so EPA chooses to expend its limited
resources on good benefits information for higher-value
uses. This does not seem likely if EPA is serving as a
political entrepreneur for risk-bearers: in that case, the
agency would probably want to avoid extensive information on
benefits. However, USDA's control over the benefits analy-
ses may have some impact on what EPA would like to see. A
second alternative explanation for the positive correlation
is that better benefits information may already be developed

and available for the higher-value uses.

 

-180-

Findings for Percent of Crop Treated. The explanatory
variable "percent of crop treated” was intended to proxy the
number of users of a pesticide that would be affected by a
regulatory action. The hypothesis was that the quality of
benfits information would improve as the number of users
increased, because political entrepreneurs (namely, USDA and
user groups) would become more active in the process if a
greater proportion of their constituencies were affected.

Table 6.3 shows the results of the empirical testing of
this hypothesis. The results support the hypothesis except
in the ”other socioeconomic impacts” benefits category. In
that category, there are equal numbers of positive and nega-
tive gammas, and equal numbers of pesticides with primarily
positive and primarily negative gammas. For all five of the
benefits categories, the pesticides with primarily negative
Gs and equal numbers of positive and negative Gs range from
Dimethoate (20 measurable uses) to 2,4,5-T and Amitraz (2
measurable uses). They are not exclusively the two and
three measurable use-pesticides, as is the case for the
value of use to manufacturers explanatory variable.

Once again, a majority of the possible gammas can be
calculated in all benefits categories except for the "com-
pliance costs” category. And once again, the quality of
information ranges from zero to three when gamma cannot be
calculated.

The same types of alternative explanations described in

the last section could explain the correlation for this

-181-

tial-6.3 mummummuqamaumuwumm

Wotan-(mummmmmum)

 

 

 

 

 

 

Value of m in but any. in ma- Otlur acto- Distribution W
a (lateral m of mucus We tweets of nect- mots
4.0 a -.750 4 (4) 9 (9) 2 (2) 2 (2) o
-.749 to -.300 0 (O) 0 (0) 1 (1) 1 (l) o
«20 to o 1 (1) 2 (2) 4 (4) 1 (1) 0
.ml to .250 5 (4) 14 (11) J, (2) 4 (4) l
.251 to .300 10 (14) 9 (13) 4 (4) _ 4 (4) o
.301 to .7” I (6) 5 (5) 2 (2) 2 (1) O
.731 to 1.0 16 (15) 19 (13) 4 (4) 6 (7) 1
m1 6 9 22 23 3
- (a) o
- (7) o
— (6) o
- (5) o o
- (4) o o
- (3) 1,4 9 2 2 2
- (2) 0 O 0 0 0
- (1) 1 2 2 4 4
- (O) 1 4 14 ll 3
- (7) 10 2 O O 0
M 26 17 13 17 9
m1 mm. 6 72 an 0 0 12
line: of nautical-o
fa.- was: 6 and
b- alcuhud a a 7 8 2 of a
pan. 6
with My
‘ punitiv- 6'1 5 6 3 s l
with My
”use 0'. 1 2 3 2 O
with «3.1 M
of positive I
”the 6's 2 0 1 1 l

-182-
explanatory variable. First, the hypothesized relationship
could be true. Second, EPA (or USDA) may decide on its own
to pursue good information for those uses with a high pro-
portion of users. Third, better benefits information may

already be available for the high percentage uses.

Findings for Annual Per Acre User Losses. The "annual

 

per acre user losses” explanatory variable was intended to
be an approximate measure of per capita user losses. The
hypothesized relationship is that the quality of benefits
information will improve as per capita losses increase,
because individual users will have more at stake and will
therefore attempt to influence the regulatory process.

Table 6.4 summarizes the empirical results. The results
are similar to those in Table 6.2 for the value of a pesti—
cide use to manufacturers, but they are even more positive
in the two major benefits categories, "changes in pest
control costs” and "changes in value of production.” A
majority of the Gs that can be calculated are positive for
all five categories, and most of those are .251 or greater.
In addition, most of the pesticides for which one or more
gammas can be calculated show primarily positive gammas
across all benefits categories. When pesticides do show
primarily negative or equal numbers of positive and negative
gammas, they usually either have few measurable uses or few
measurable gammas, or both.

As for the previous two explanatory variables, a major-
ity of the possible gammas cannot be calculated for "compli-

ance costs.” Also, most gammas can't be calculated for

-183-

”1.6.4 mamumawmumuuum
uda—(med-nummugu)

 

 

 

 

 

Valu- ot Ouaga- 1n but any. in m Other Socio- cm W
a m m at We: We m of Inact- Oust-
-1.0 to -.750 2 (2) 7 (7) 2 (1) 1 (1) o (0)
-.249 to O 1 (O) 3 (2) 4 (4) 4 (5) 0 (1)
.001 to .250 5 (2) 9 (6) 2 (2) 2 (2) o (0)
.251 no .500 9 (11) 15 (14) 5 (5) 2 (3) 1 (0)
.301 to .750 7 (11) 3 (17) 1 (2) 3 (3) 1 (O)
.751 to 1.0 24 (21) 17 (14) 8 (7) 6 (5) 1 (1)
M 43 61 23 19 3
- (I)
- (7)
— (6)
- (5) 1 1
-- (4) o o
— (3) 14 9 2 2 y 2
- (2) 0 0 0 0 o
- (1) o 3 2 a 4
- (O) 1 2 13 11 3
- (7) a 4 o 0 0
M 24 19 17 21 9
m1 W G 72 an 40 0 12

 

Moth-tidal-
«mew

be amulet-d 8 a 7 7 2 of a
pan. 6
with My
positive 0': 6 6 5 5 1
with My
”tive 6'. 1- 2 2 0 O
with «:11 m
of positive I

uptic- 6'. l o 0 2 1

-184-

”distribution of impacts." The quality of information when
gamma cannot be calculated ranges from zero (ignored) to
five (quantitative point estimate based on a study or pub-
lished data).

The fact that there was a strong positive correlation
between the proxy for per capita user losses and the quality
of benefits information lends credence to the Stigler/

Peltzman theory that per capita economic impacts motivate

 

individuals to organize in an effort to affect regulatory
outcomes. Although the existence of the positive correla-
tion does not prove that (or any of the other) hypothesis,

it certainly supports the theory.

Findings for Annual Total User Losses. The last vari-
able used to explain variation between uses of a single
pesticide was ”annual total user losses.” This variable was
meant to describe the aggregate annual losses for all users,
as compared to the "per acre user losses" variable intended
to capture individual user losses. The relationship between
total user losses and quality of information was expected to
be a positive one, reflecting the expected participation of
pesticide users (or their political entrepreneurs) in the
regulatory process as potential economic losses increase.

Table 6.5 shows that, again, most gammas could be calcu-
lated except for compliance costs and distribution of
impacts. When gamma could not be calculated, the quality of
information ranged from zero to five. In the two major

categories, changes in pest control costs and changes in

-185-

Tam-6.5 mammmwamamummmw.
mac-(mmpmmmmmtm)

 

 

 

 

 

m ot Gunp- in that M in Nut 0th..- ado- W W
G W m of Wm We met. of m m
«70 to -.300 0 (O) O (O) 1 1 (1) o
-.249 to o 0 (O) 2 (2) 4 3 (3) 0
.m1 to .250 4 (3) 4 (5) 2 1 (2) o
.251 to .300 9 (12) 13 (14) 5 3 (5) 1
.501 to .750 6 (5) 1O (7) 3 4 (1) o
.751 to 1.0 21 (22) , 16 (19) 4 5 (5) 2
amen 44 56 21 16 3
- (6) O
- (7) O
- (6) o
- (5) 1 1
- (4) o o
- (3) 18 12 3 3 2
- (2) o o o o o
— (1) o 3 3 a 4
- (O) 1 4 13 11 3
- (7) a 4 o o 0
am 23 24 19 22 9
total loam. 6 72 so ' 40 4o 12

 

Moth-tida-
tormdlcomld

be clan-ted 8 6 7 7 2 of a
" ”a 6
with may
positive G'- 6 6 4 5 2
with print-117
”tive 6's 0 2 1 1 o

with 4931 m
of positive 4
”tin 6's 2 O 2 1 O

-186-
value of production, the quality of information was usually
three (expert opinion) or better when gamma could not be
calculated due to ru> variation between uses. But in the
three minor categories, the quality of information was
usually zero (ignored) or one (qualitative measurement).

The results for gammas that could be calculated were
similar to those for per acre user losses in Table 6.4.
Gammas were usually positive for all five benefits cate-
gories, and most pesticides showed primarily positive
gammas. At first blush, the correlation summarized in Table
6.4 for per acre user losses appears to be stronger. But
when one looks at the two major benefits categories, which
far outweigh the three minor categories in terms of EPA's
time and the importance of the information to the regulatory
process, the annual total user losses variable appears to
have slightly greater explanatory power. The pesticides
with primarily negative gammas and equal numbers of positive
and negative gammas in the two major benefits categories
were Amitraz and 2,4,5-T, both of which had only two measur-
able uses for annual total user losses.

This result does rMM: quite match the Stigler/Peltzman

theory that per capita economic stake determines attempts to

 

influence the regulatory process. But one possible explan—
ation for the slightly stronger explanatory power of
aggregate user losses is that USDA and already-existing user
groups are political entrepreneurs for pesticide users.

These political entrepreneurs may be more sensitized to

—l87-
react to an aggregate figure reflecting losses to all users,

rather than data on individual impacts.

Findings for EPA's Risk Ranking. Once again, Edwin

 

Johnson's statement that the risk of a pesticide drives the
regulatory process was tested. In order to test the hypo—
thesis, EPA's risk ranking for each pesticide was used as
the explanatory variable. The quality of benefits informa-
tion for a pesticide was measured by averaging the in-
formation quality across all uses of the pesticide.

The results of this test can be found in Table 6.6.
Averaging the quality of information data across uses was
easier for benefits than for risk, so the benefits results
are probably somewhat more reliable. Supporting tables for
Table 6.6 can be found in Appendix B (Tables 69 through 73).

To put it simply, there was absolutely no indication the
Johnson statement holds true for benefits information. All
of the possible gammas could be calculated, and most of them
were negative. The negative gammas were clustered around
the two least negative ranges for gamma (-.249 to 0 and
—.499 to -.250).

None of tflua five benefits categories showed primarily
positive gammas - (flue two major categories were very pre-
dominantly negative and the three minor categories showed
equal numbers of positive and negative gammas.

This is a very interesting result. It makes sense that
Johnson's statement would test out (1) be true, given the

strong correlations already' shown between interest group

-188-

Table 6.6 Sm of Oorrelatione between @A'e Peeticide Riel: mm and the
Average Quality of Benefite Inter-tion

Huber of. Galilee

 

 

 

Changee Charge in Other
in Peat Value of Socioeconaic Distribution Clo-pliance
Value of G Control Ooeta Production Igcte of 353:2: Ooete
-1.0 to -.750 o o o ' o o
-.749 to -.5(X) 1 3 O 1 O
-.499 to -.250 5 ' 3 1 o 1
-.249 to O 4 3 2 2 2 O
.001 to .250 O 2 0 1 O
.251 to .500 O 0 .‘ O” 0 1
.501 to .750 O O 2 O 0
.751 to 1.0 O O O 1 O
-—-- O O O 0 O
' mm Poeeible o 9 - 10 5 , s ' 2
Primary Sign of Gal—a .
poeitive
negative at at
equal mater

of positive and j _
negative at x x

—189-
"propensity to participate" and the quality of benefits
information. But why the negative results for the two major
risk categories? Could it be that interest groups are
pragmatic, and decide not to put as much pressure on EPA's
benefits analyses when pesticide risk is high?
Participation in the regulatory process can be costly, so it
would be wise of the interest groups (and their political
entrepreneurs) to save their scarce political and economic
resources for pesticide regulatory processes where they

think they can win.

Findings for All of the Explanatory Variables and EPA's

 

Decisions on Pesticide Registration. The ultimate goal for

 

interest groups trying to affect the regulatory process is
obtainimg a favorable pesticide registration decision from
the EPA. Table 6.7 summarizes the correlation between EPA
decisions and the four variables used to explain variation
between EESE of a pesticide. A supporting table for Table
6.7 is found in Appendix B (Table 68). The results bode
well for the interest groups and ill for the EPA.

Decisions varied between measurable uses for only four
of the eight pesticides, so the "decision-explanatory vari—
able" gamma could only be calculated for those four (DBCP,
Amitraz, Endrin, Chlorobenzilate). When decisions did not
vary, all uses were registered with restricted use or sent
to administrative hearings.

When gamma could be calculated, it was virtually always

positive, indicating a positive relationship between the

-l90-

Table 6.7 Su-ry of Correlations between EPA Pesticide Registration Decisions
and Four Explanatory Variables

Muster of Games
(“shore in parentheses denote lower range Gs)

Explanatory Variable:

Value of Use to Percent Annual per Acre Annual mm
Value of Gene Manufacturers Crop Treated User Losses User Losses

 

-1.0 to -.750 o o (0) o (0) o
-.749 to -.500 o o (0) o (0) o
-.499 to -.250 o o (0) o (a) o
-.249 to o 1 1 (o) o (0) 1
.001 to .250 o o (0) o (0) o
.251 to .500 1 ‘ o (15 1 (2) o
.501 to .750 o 1 (1) 2 (1) 1
.751 to 1.0 2 2 (2) 1 (1) 2
—— 4 4 (4), 4 (4) 4
Total Possible G 8 8 8 8
Meter of haticides
(mumpnntheaesdemtelomrrangeGe)
for which 6 oould A
be calculated 4 4 (4) 4 4
with positive 6 3 3 (4) 4 3
with negative 6 1 1(0) 0 o
with G . 0 O O (O) O 1

-l9l-
four explanatory variables and EPA decisions. For example,
the data show that as the value of a pesticide use to the
manufacturers increases, EPA is more likely to fully regis-
ter that use in three of the four pesticides for which gamma
could be calculated.

The apparent success of pesticide manufacturers and
users in influencing EPA's decisions is further emphasized
by the gamma calculated to show the correlation between
EPA's pesticide risk rankings and the average decisions made
for each pesticide. That gamma was calculated to be -.368,
indicating that the more risky pesticides, determined by

EPA's own standards, are not the most regulated pesticides.

Summary: Correlation Between Explanatory Variables and

 

Quality of Information on Benefits. The differences between

 

the empirical results for risk and benefit information are
like night and day. For some reason, the empirical results
1x1 this reseach indicate that pesticide umnufacturers and
users seem far better able (or willing) to affect EPA's
benefits analyses than its risk analyses. Some of the
reasons for this disparity in results might include:

o Pesticide users and manufacturers want risk infor-
mation to remain as incomplete and vague as pos-
sible, because they do not think they will be able
to rebut EPA's preliminary risk findings.

0 EPA serves as a political entrepreneur for risk
bearers, and so has managed to insulate itself from

user and manufacturer pressure to improve exposure

—l92-
and risk data for the most important use of a
pesticide.

o Pesticide users and manufacturers have decided to
channel their limited political and financial
resources toward the benefits side of the equation,
because they feel that they can accomplish more by
improving benefits information.

o USDA (and perhaps some well-organized user and
manufacturer groups) serves as a: political entre-
preneur for pesticide users, and possibly manufac-
turers. Since USDA seems to oversee the benefits
analysis, it channels RPAR resources into good
benefits data for the most important pesticide
uses.

The empirical results also indicate that EPA's pesticide
risk ranking does an ambiguous job of explaining the varia-
tion in the quality of risk data between pesticides, and a
terrible job of explaining the variation in the quality of
benefits data. Why is this so, when Edwin Johnson so con-
fidently claims that risk drives the information gathering
process? Again, there are several possibilities. Johnson
may simply be wrong. Or perhaps some flaw in the empirical
work has resulted in arguable findings. A third possibility

is that Johnson perceives that risk drives the process from

 

his perch within the agency. But maybe EPA's pesticide reg-
ulation is really driven by external constraints including

interest group pressure, and the process is only driven by

-l93-
risk within those constraints. In the case of benefits, the
USDA-pesticide user liaison may be so powerful that it
overshadows EPA's internal constraints.

In summary, the strong correlation between the explana-
tory variables and the quality of benefits information does
not necessarily prove that interest groups affect EPA's
information search. during pesticide regulation. But the
empirical results certainly lend a good deal of support to

various theories of interest group power.

CHAPTER 7
CONCLUSIONS
OVERVIEW OF RESEARCH

Zhi a broad sense, this research was an effort to gain
some insight into EPA's regulation of pesticides. The
specific objective, as noted in Chapter 1, was to provide
systematic, empirical information about EPA's information-
gathering activities ixx the Rebuttable Presumption Against
Registration (RPAR) process. The quality of EPA's risk and
benefit information was empirically described, and hypo-
theses on interest group pressure were tested.

To briefly review the structure of this thesis, theories
of regulatory decision making and a conceptualization of
pesticide risks anui benefits were described in Chapters 2
and 3. In Chapter 4, the methodology for the empirical work
was outlined. Scales were developed to depict the quality
of various pieces of risk and benefit information for each
use of each pesticide. EPA documents were used to measure
the quality of information variables, which served as the
dependent variables in the research. Explanatory variables
were also quantified from EPA documents. Four of the ex—
planatory variables measured the incentive for pesticide
users or manufacturers to become involved in the regulation

of a pesticide use: the fifth ranked the eight case study

-l94-

~195-

pesticides according to EPA's own determination of risk in
an effort to explain differences in the quality of informa—

tion between pesticides. Thus, correlations between the

 

four "interest group” explanatory variables and quality of
information were measured for each pesticide, while the EPA
risk ranking variable was used to explain variation across

all of the pesticides.

SUMMARY OF RESULTS

The results of the empirical work can be found in Chap-

ters S and 6. There were four major findings:

0 All five explanatory variables showed mixed results
when correlated with the quality of risk informa-
tion. The variables could not be considered to be
highly correlated with the quality of risk infor-
mation in either a positive or a negative way.

0 The four "interest group" variables (value of
pesticide to manufacturers, percent of crOp
treated, per acre user losses, and total user
losses) showed positive correlations with the
quality of benefits information. The strongest
showings came from the per acre user losses and
total user losses variables, intended to proxy the
per capita losses and industry-wide losses from
pesticide cancellation. The EPA pesticide risk
ranking variable showed positive and negative
correlations with the quality of benefits infor-
mation, so the relationships could not be deter-

mined.

-l96—

EPA failed to measure or even mention many of the
risk and benefits variables. When risk data were
quantified, the sources tended to be assumptions,
expert opinions, or studies of other pesticides
instead of studies of the pesticide in question.
For benefits data, theme was a great reliance on
expert opinions and assumptions. EPA usually
quantified the major risk and benefit categories,
such as consumer and ground applicator exposure and
risk, changes in pest control costs, and changes in
value of production. But the remaining categories
were generally ignored or only qualitatively men—
tioned. Overall, EPA's information was not good
and risk information was poorer than benefits
information.

It appears that interested parties were able to
affect EPA's decisions. The four "interest group"
variables did a good job of explaining EPA's de-
cisions. As the economic stake of users and manu-
facturers for a particular pesticide use increased,
so did the likelihood of an EPA decision to reg-
ister or conditionally register that use. It seems
that pesticide users and manufacturers were able to
obtain decisions that minimized their losses,
possibly at the expense of risk bearers (especially

given that a negative correlation exists between

 

stringency of decisions and EPA's pesticide risk

rankings.)

-l97-

ANALYSIS OF RESULTS

Value Judgments in Pesticide Regulation. There are many

 

assumptions made, but not highlighted, in EPA's risk-benefit
analysis. Some of these value judgments were carried over
into this research in order to analyze EPA's activities.
The first that comes to mind is the definition of benefits,
costs (subtracted from benefits to obtain net benefits) and
risk. EPA's definition of these parameters affects the
issues addressed, how they are measured, and ultimately, how
those issues are resolved. But perhaps more importantly,
this definition of net benefits and risk assumes a particu-
lar distribution of property rights. It is usually the
status quo distribution of rights that is assumed by EPA as
a gauge to measure net benefits and risk, and this assumes
that the status quo is the best set of rights.

Other value judgments used by EPA in its analysis in-
clude assumptions necessary 1x3 quantify human risk and to
judge the acceptability of that risk. As Wessel (1980)
suggests, science itself is valueless, but the use of
science requires value judgments.

On the net benefits side, EPA makes a value judgment as
soon as it aggregates user losses from pesticide regulation
(or disaggregates into per acre user losses). Assuming that
additional dollars are of equal value to all pesticide users
or that average per acre losses even begin to reflect indi-

vidual differences makes the analysis far less predictive of

-l98-

real human impacts. In addition, EPA often fails to con-
sider the value or distribution of certain economic impacts.

These value judgments are necessary for EPA to simplify
its analyses, or for technical reasons. However, the fail-
ure to acknowledge the assumptions can be misleading to
participants in and students of the agency's regulatory
process.

Correlation Between Variables. One of the most striking

 

results in the empirical work was the lack of correlation
between the "interest group" explanatory variables and the
quality of risk information, and the strong correlation of
the variables with the quality of benefits information. At
first glance, the benefits results are consistent with the
hypothesized relationships, but the risk results are not.
However, when considering the characteristics of EPA's
regulatory process, the risk results may not seem so incon-
sistent after all.

'The EPA is largely responsible for the preparation of
the risk analysis. Although theme is a great deal of re-
liance (N1 outside research to estimate exposure and toxi-
city, it is still up to the EPA to choose which information
to use and how to compile it into a risk analysis. A dif-
ferent group of players has responsibility for the benefits
analyses. USDA and land-grant university specialists team
up with EPA economists to obtain the data for the prelim-
inary benefits analyses. USDA seems to have primary

responsibility for the analyses, which end up as fairly

-199-

polished USDA publications. For the case study pesticides
for which they were pmblished, the benefits analyses were
used virtually verbatim in EPA's final position documents.

One of the theorists cited in Chapter 2, James Q.
Wilson, refers to the participation of "political entre-
preneurs" in the regulatory process. EPA appeared to assume
this role in representing risk bearers for the eight case-
study pesticides. Because of this agency role, EPA was
expected to have some control over the amount and quality of
risk data. The results seem to show this by reflecting no
relationship between pesticide users' and manufacturers'
incentives and the quality of risk information.

Another possible explanation for the risk results is
that the hypotheses may not have accurately reflected the
relationship between better risk information and regulatory
outcomes favorable to pesticide users and manufacturers. In
fact, manufacturers and users may have felt that better
information would be harder to challenge, and would lead to
greater restrictions on pesticide use.

The theories and hypotheses in Chapter 2 seemed to
adequately explain the quality of benefits information. In
fact, the relationships between per acre user losses and
total user losses and the quality of benefits data may have
been especially strong because the USDA served as a politi-
cal entrepreneur for pesticide users. USDA and state agri-
cultural experts provided most of the benefits information,

and there is no doubt that their constituencies were made Up

-200-

of pesticide users. And EPA's allegiance to pesticide users
seemed to decline with the quality of benefits information,
as was demonstrated for Arizona citrus growers using Chloro-
benzilate and the minor Endrin uses. Thus, there was a real
incentive for pesticide users and their political entre-
preneurs to provide good benefits information.

The fact that EPA's pesticide risk ranking did not
correlate positively with the average quality of risk and
benefit information does not conflict with the hypotheses at
all -- it simply suggests that EPA does not have complete
control over the regulatory process. In fact, based on the
results for the other four explanatory variables, it appears
that pesticide users and manufacturers have greater control
over the benefits information. The quality of risk informa-
tion did show a slightly greater correlation with EPA's risk
ranking than the benefits information, which indicates that
EPA is better able to influence risk information.

Finally, there was a strong positive correlation between
the four ”interest group” explanatory variables and EPA's
pesticide registration decisions, meaning that a pesticide
use was more likely to be fully registered as its importance
to manufacturers and users increased. But for the EPA
pesticide risk ranking variable, EPA restrictions increased
as risk decreased. This suggests that it is pesticide users
and manufacturers who determine regulatory outcomes. EPA
may be able to influence the quality of risk information to

a very small degree, but the quality of benefits information

-201-

appears to be influenced by interest group politics. More
importantly, so do pesticide decisions, which are the reg-

ulatory bottom line.

EPA: Vote-Maximizer or Budget-Maximizer? In Chapter 2,

 

there were several hypotheses on the expected differences in
results if EPA was a vote-maximizing or budget-maximizing

agency. This section addresses those hypotheses.

Predicted results for vote-maximizing agency.

 

1. Does costly information found in EPA's position
documents support powerful groups? The risk information
obtained by EPA did not appear to support pesticide users or
manufacturers who are most likely the politically powerful
groups. On the other hand, benefits information did appear
to support those groups. However, since EPA prepared the
risk analysis but not the benefits analysis, the answer to
this question must be no, contrary to what would be expected
if EPA were a vote-maximizing agency.

2. Did EPA's assumptions favor powerful groups? In
general, the answer would again have to be no. Risk assump-
tions tended to be worst cases, which certainly do not favor
pesticide manufacturers and users. When benefits data were
unknown, the impacts were usually ignored, resulting in
lower benefits estimates. There were some notable excep-
tions: the failure to account for dietary exposure in most

non-consumer risk categories, and assumptions of total

-202-

compliance with proposed regulations. But in most cases,
the assumptions did not favor the powerful groups.

3. Was there some information favoring Opposing
groups? Although we would expect otherwise if EPA was a
vote-maximizer, risk information. almost always seemed to
favor risk-bearers. Risk bearers would not be expected to
be powerful because of uncertainty of risk and high organi-
zation costs, yet EPA ignored most toxicity and exposure
information reflecting lower risk.

4. Were there assumptions or costly information favor-
ing opposing groups? Again, the answer is yes, even though
the vote-maximizing theories would suggest otherwise. EPA
often made assumptions favoring risk bearers. Examples
included using the one-hit model to determine risk and
assumptions of tolerance levels when residues in food could
not be found.

Based on the answers to these four questions, it would
be hard to claim that EPA maximized votes during its infor-
mation search. However, the final pesticide registration
decisions may have been a different story, since they seemed

to favor the powerful groups.

Predicted results for budget-maximizing agency.

 

1. Did EPA challenge information from powerful groups?
EPA definitely expended a great deal of energy and resources
in challenging pesticide manufacturers' risk rebuttal infor-
mation. On the other hand, EPA did not generally challenge

information showing extensive benefits of pesticide use.

-203-

2. Did EPA's assumptions favor its role in regulating
pesticides? Certainly some of the risk assumptions, such as
assuming maximum allowable residues in food crops and using
conservative "one-hit" risk models, resulted in higher risk
estimates and a stronger role for EPA in pesticide regula-
tion. EPA's own rules in triggering and implementing the
RPAR process also seemed to protect that role, requiring
toxicity (but not exposure) information to initiate the
process, challenging manufacturer's toxicity data, and
requiring manufacturer benefits of pesticide use to be
ignored.

3. Did EPA expend resources to obtain information to
protect its role? In general, EPA did not conduct its own
studies -- the agency usually relied on risk data from
government or academic studies or pesticide manufacturers,
exposure data from studies of other pesticides, and benefits
data from USDA. However, EPA definitely spent a lot of time
and money (Hi the RPAR process, which is structured to
strengthen EPA's role in pesticide regulation.

4. Did EPA force powerful groups to provide pro-EPA
information or make it costly for them to provide anti-EPA
information? The great bulk of the toxicity data, which
incriminated the pesticide and triggered the RPAR process,
was provided by pesticide manufacturers so that they could
register a chemical. Information to dispel preliminary risk

findings was costly for manufacturers because it had to be

—204-

provided in the adversarial rebuttal process. And informa-
tion on the benefits of pesticides to manufacturers was
almost infinitely costly since it could not be considered in
the RPAR process. However, information on benefits to
pesticide users appeared to be relatively easy to inject
into the RPAR process, in spite of the fact that this type
of information challenged stringent pesticide regulations.
Again, this could have been a result of EPA's lack of con-
trol over the benefits analyses.

These four indicators suggest that EPA might have been a
budget-maximizing agency during the information gathering
stage of the RPAR process. The regulatory program appeared
to be structured in a way that supported a role for EPA in
pesticide regulation.

Summary: EPA Incentives and Recommendations to Achieve

 

Change. If budget maximization is (”KB of the goals that
drives EPA's regulation of pesticides, operating within the
constraints of limited agency resources is certainly
another. Knowledge of these two agency goals, and of the
impact of pesticide producers, users and risk-bearers on the
attainment of these goals, can help in designing the regu-
latory program to achieve different results.

If EPA strives to obtain better information for higher-
risk pesticides, as Edwin Johnson intimates, then the agency
should look at mechanisms to improve risk information (if
that information shows higher risk) and ways of increasing

the relative power of risk-bearing groups.

-205-

Since EPA's constituency for purposes of justifying its
budget requests is probably risk-bearing groups, the goal of
budget maximization is consistent with dedicating more
resources to high-risk pesticides. This is evidenced by the

answers to the questions in the previous section. The

problem is not EPA's intentions: it is the financial andw

political constraints (ME regulatory life. Changing those
constraints is not easy. It would involve attempts to
obtain more funding for toxicity and exposure studies con-
ducted within the EPA or by another agency. Alternatively,
existing law could be amended to require pesticide manufac-:
turers to provide more data.

Increasing the relative power of risk-bearing groups
could take the form of reducing the organization costs of
such groups (i.e., subsidizing and encouraging the formation
of groups) and reducing the costs of articulating risk-
bearer preferences 1J1 the regulatory process (e.g., sur-
veying risk-bearers, collecting better exposure data, hold-
ing public hearings in areas where the pesticide is heavily
used). Another way to increase the relative power of risk-
bearers is to decrease the power of adversarial groups, such
as pesticide manufacturers and users. USDA is doubtless the
primary voice for pesticide users, and it appears that this
role prevents EPA from achieving more stringent regulation
of higher-risk pesticides. As a result, EPA could re-
evaluate its relationship with USDA. In general, reducing
opportunities to voice concerns will decrease the political

clout of pesticide users and manufacturers.

-206—

THE ROLE OF INFORMATION:
IMPLICATIONS FOR ECONOMIC POLICY ANALYSIS

The bottom line for participants in the RPAR process is
their impact on the decision and ultimate outcome of the
process. Given that pesticide users and manufacturers
apparently affect the quality of benefits information and,
to a lesser extent, risk information, how does this trans-
late into an impact on decisions?

There are a couple of general statements that can be
made about EPA's pesticide decisions. Once an RPAR has been
issued, EPA is unlikely to fully register a product, because
a decision outcome is needed which justifies the time and
resources expended (Hi the decision-making process. Also,
even the best information does not point to a clear decision
for EPA. The risk of pesticide use is too uncertain for a
calculation of the net benefits of alternative policies.
Even if net benefits could be calculated, there is no reason
to believe that EPA would find the answer, or the assump-
tions on which it is based, politically acceptable. It was
pointed out in Chapter 2 that EPA may have goals other than
maximizing some normative measure of net social welfare.

Given these two <general conclusions, can information
influence decision outcomes? It is evident from reading the
position documents that the risk-benefit information does
help to determine outcomes. The information helps EPA deter-
mine who is at most risk, who benefits the most, which
methods of application cause the most exposure, and other

characteristics of pesticide use. This is of use in helping

(9c/L

r:

-2o7-

EPA to develop alternative courses of action and choose
between them. Risk-benefit analysis then, helps the EPA to
clarify issues and determine effective risk—reduction poli-
cies. This explains the efforts of different interest
groups to provide risk and benefit information.

To some extent, information may also be used as a justi-
fication for previously-made decisions. Decisions in the
RPAR process are predetermined in the sense that once an
RPAR is issued, there is almost always an implicit decision
to restrict the pesticide to some extent. As a result, much
of EPA's information search is devoted to sustaining risk
findings. But this does not negate the conclusion that the
specifics of the regulatory alternatives, and to some degree
the choice between those alternatives, are determined by
risk and benefit information rather than vice versa.

In support of that conclusion, there was a strong corre-
lation between the explanatory variables and EPA's deci-
sions, as described in Table 6.7 in Chapter 6. Larger
values for the explanatory variables resulted in fewer
restrictions, with the exception of the EPA pesticide risk
ranking variable. That variable showed a substantial nega-
tive correlation with EPA decisions (i.e., as risk in-
creased, restrictions decreased), once again indicating that
the RPAR process is not driven by EPA's perception of risk.

Some students of regulation have proposed the theory
that regulation is just a facade -- that it actually pro-

tects the regulated by preventing more stringent regulation.

-208-

While the results of this research, taken out of context,
may support such a theory, the reality appears to be some—
what different. EPA does not appear to be working for
pesticide manufacturers and users. Instead, it appears that
these groups and their political entrepreneurs simply "out-
muscle" EPA and risk-bearers in the regulatory process.
Although EPA's regulatory style is severely cramped by the
regulated, it seems to be neither the goal of the agency nor
the intent of the regulatory design to protect pesticide
manufacturers and users. Likewise, the very presence of EPA
and organized risk-bearers no doubt tempers the style of the
regulated groups. Information on risk does impact regula—
tory outcomes, and as Peltzman suggests, the regulated
groups do not get everything they want from the regulatory
process.

In summary, there is a use for risk-benefit information
in pesticide regulation. Although decisions do not fall out
of the risk-benefit information, the information does pro-
vide some guidance to EPA on how to reduce risks. This is
not the extensive role for information envisioned by many
policy analysts, but it does seem to result in risk reduc-
tion and improved public safety.

IMPLICATIONS OF FINDINGS
FOR PROPOSED REGULATORY REFORMS

The Bush Task Force Reforms. There have been rumblings

 

in Washington, ever since the Reagan administration took

-209-

office, of implementing a major overhaul of the RPAR pro-
cess. The most concrete prOposals emanated from directives
from Vice President George Bush's task force on regulatory
reform. 11 new "Special Review" process has been prOposed.
The name change is the smallest difference between the
current and proposed processes.

Some of the major changes include:

o Notifying pesticide manufacturers before the pro-
cess begins, and utilizing their input to determine
which uses (as opposed to all uses) should be re-
viewed. EPA would continue to review the critical
minor uses, apparently because it is recognized
that manufacturers would have little interest in
defending those uses.

0 Relying on manufacturers for exposure data.

0 Attempting to negotiate use restrictions after ex-
posure data were obtained.

0 If no solution was reached, and if EPA's top man-
agement agreed that the risk exists, the Scientific
Advisory Panel would review the risk data, and USDA
would review and comment on a benefits analysis
prepared by manufacturers.

0 Assuming that the special review continued, nego-
tiations would be initiated between EPA and the
manufacturers to try to reach a risk—reducing

solution.

—210-

0 Only if negotiations failed would a USDA benefits
analysis and EPA document outlining regulatory
options be prepared, with USDA, SAP and public
comment.

0 Additional negotiations, with a final decision doc-
ument if negotiations failed.

0 Relying on risk assessments from other agencies and
offices when possible.

0 Revising RPAR "risk triggers" to account for expo-
sure and negative studies, with the effect of
making it more difficult to initiate the special
review.

Based on the results of this research, how do these
proposed reforms affect the process and the role of interest
groups in the process? EPA makes it clear that the reforms
are intended to allow more negotiated settlements, reduce
paperwork, and otherwise streamline the process. But this
also means less power for interest groups other than pesti-
cide manufacturers and less public scrutiny of the process.
Research such as the type contained in this thesis would
probably be precluded, since there would be no paper trails.

Giving pesticide manufacturers control over exposure
data would undoubtedly change the outcome of the process.
Whereas the quality of risk information was shown to be
poorly correlated with the value of the pesticide to
manufacturers, this would almost certainly change when

manufacturers gained control of the exposure information

-211-

search. And the current strong correlation between the
value of a pesticide use to manufacturers and EPA decisions
should become even stronger.

Consultations with manufacturers over which uses should
be reviewed, without using completed risk and benefits
analyses, may eliminate pesticide users from the process.
In many cases, uses that are valuable to manufacturers will
also be important to users, but for some uses this will not
be the case. As a result, lower correlations between pesti-
cide user variables and risk information should be expected
for some uses.

The use of informal contacts and negotiations between
EPA and pesticide manufacturers will also change the balance
of power in the process, particularly if EPA is not as com-
mitted to defending risk-bearers. Much of the process will
no longer be documented, and the nature of a negotiated
solution means that EPA will not have to defend its decision
on other than political grounds. Actually, most of the
defending will occur during the risk analysis.

The revision of the risk triggers should result in fewer
pesticides being considered for special review. The assump-
tion here by the Bush task force is that some pesticides
that do not pose a risk are being regulated due to broad
risk triggers.

If EPA is pro-manufacturer, as it appears to be (rela-

tively speaking) under the Reagan administration, then this

-212-

process should save time and money and be easier for pesti—
cide manufacturers. But if EPA is anti-manufacturer, as it
has been in the past and is certain to be at some point in
the future, attempts to negotiate may be futile and the
process may be as drawn out as it is now, albeit with less
documentation.

If decisions were made based on risk information only,
or on risk and benefits information provided by manufac—
turers, the regulatory outcome would probably be different
than it is now. Solutions would be biased toward manufac-
turers' perceptions of risk and benefits, and the final
choice for each pesticide use would most likely be highly
correlated with values of the uses to manufacturers.

To describe changes in pesticide regulation since 1981,
even without formal changes in the RPAR process, Table 7.1
shows activity before and after that date. The table shows
that since 1981, more pesticides have been returned to the
registration process without restrictions or with negotiated
restrictions, fewer RPAR's have been issued, fewer manufac-
turers have voluntarily agreed to cancel their product and
more RPAR's have been completed.

National Academy of Sciences Recommendations. In the

 

book Regulating Pesticides (Environmental Studies Board Com-

 

 

mittee on Prototype Explicit Analyses for Pesticides, 1980),
several recommendations are suggested to improve the RPAR
process. The panel of economists and natural scientists

recommended that:

-213-

Table 7.1 Changes in Pesticide Regulation Since President
Reagan's Inauguration

RPARs Issued
(according to date of PDl publication)

 

January 21, 1981 - September 1983 0
April 1978 - January 20, 1981 10

Volunta§y7Cancellations
(according to date of publication)

 

January 21, 1981 - September 1983 2
April 1978 — January 20, 1981 5

Pesticides Returned to Registration
(according to date returned)

 

January 21, 1981 - September 1983 14
April 1978 - January 21, 1981 5

RPARs Completed
(according to date of PD4 publication)

 

January 21, 1981 - September 1983 10
April 1978 — January 21, 1981 7

 

Source: Compiled from data in "Status Report on Rebuttable
Presumption Against Registration (RPAR) or Special
Review Chemicals, Registration Standards or Data
Call-In Programs" (U.S. Environmental Protection
Agency Office of Pesticide Programs, Washington,
D.C., September, 1983).

~214-

The OPP should prioritize pesticides according to
toxicity of active ingredients and extent of use,
and regulate pesticides in order of prioritization.
Alternative pest control methods should be identi-
fied early in the review process. If any of those
alternatives are potentially hazardous, they should
be removed from the prioritization and reviewed as
quickly as possible.

The OPP should not try to generate numerical risk
estimates ‘when there are ru: reliable human
epidemiological data. Instead, OPP should "rank"
pesticide risk relative to other pesticides, and
the risk of alternative regulatory schemes should
be presented in terms of dosage.

EPA staff should observe actual application,
formulation, and study of the RPAR pesticides.

OPP should estimate the "economic lifetime" of the
RPAR pesticides, and estimate risk and benefits
accordingly.

OPP should present exposure estimates as a range,
with the low point reflecting the most likely
degree of exposure, and the high point reflecting
the maximum potential exposure.

OPP should not include the effects of pesticide
regulations on net farm income resulting from crOp
price changes in its benefits analyses, because

these changes are offset by consumers.

-215-

o The OPP should establish a Benefits Review Panel of
experts to review the benefits analyses. Members
of that panel should be involved at the outset of
the benefits analysis.

Unlike the Bush task force recommendations, which focus
on ways to cut the costs and length of the RPAR process, the
NAS proposals are centered around a quest for better infor-
mation. The relevant question is: will these proposals
lead to improved decision making?

It has already been noted that information does not seem
to help choice-making, but it may help the EPA to develop
alternative solutions. Several of the NAS recommendations
would shed additional light on potential risk reduction
methods. Identifying alternative pest control methods that
may also be hazardous, ranking alternative regulatory
schemes according tn) dosage, and observing actual applica-
tion procedures may all help to clarify regulatory Options
for EPA. But some of the other recommendations, those that
call for refinement in aggregate risk or benefit calcula-
tions, may not be of as much value to the agency in its
decision-making. 1w: matter how precise the calculation of
net benefits and risks, EPA still must make value judgments

when weighing benefits against risks in decision making.

DIRECTIONS FOR FUTURE RESEARCH
There are many areas of pesticide regulation that can
still be investigated. An obvious topic of research sug-

gested by the pmevious section is to attempt to duplicate

-216-

the research for the new Special Review process if it is

implemented.

Other suggestions revolve around improving the current

research:

0

Include more case studies now that more RPARs have
been completed.

Improve measurement of quality of information vari-
ables by interviewing personnel from EPA, manufac-
turers, pesticide user groups, and risk-bearing
groups or their political entrepreneurs.

Improve measurement of explanatory variables by
going outside of the position documents for vari—
able specification and measurement, and possibly
develop additional variables for risk-bearing
groups.

Use econometric modeling if data is good enough to
justify such a methodology. Such a procedure would
tell more about the relationships between explana-
tory and dependent variables and would highlight
model specification problems, if any.

Focus on stages of the regulatory process other

than information search.

Given adequate time and resources, this research could

be expanded upon in order to find out more about EPA's reg-

ulation of pesticides. But the research has accomplished a

great deal in terms of describing the information search

portion of the RPAR process and providing estimates of

interest group influence on that portion of the process.

APPENDICES

APPENDIX A

RANKING OF PESTICIDE USES ACCORDING
TO FOUR EXPLANATORY VARIABLES

APPENDIX A
Ranking of Pesticide Uses According to Four Explanatory Variables

Table A1. Ranking of Uses for Each Pesticide According to Value of Pesticide
Use to Manufacturers

 

Pesticide g§g_ Value to Manufacturers
Pronamide LQttuce $2,091,000 to 4,182,000
Alfalfa $2,084,000

Ornamentals $420,750

Ber-ties $240, 570

Turf $205,425

Sugarbeet Seed > $64,350

Silvex Rangeland $750,000
Rice $11 [ 000

Apples ?

Noncrop Uses ?

Pears ?

Prunes ?

Sugarcane ?

2,4,5-T Rangeland $6,750,000
Noncrop Uses $800,000

Amitraz Pears $3,047,414
Apples $1,129,480

Chlorobenzilate Florida Citrus $3,220,000
Texas Citrus $406,000

Cotton $212,550

California Citrus $30,000

Arizona Citrus $24,000

Other "Little Use”

(assigned value of $100)

Fruits/Nuts ?

Endrin Small Grains $1,380,000
Orchards $152,800

Cotton $149,910

Alfalfa ?

Bird Perches ?

Conifer Seeds ?

Ornamentals ?

Sugarcane ?

Tree Paint ?

Vegetable Seeds ?

watermelon Seeds ?

-217-

DBCP

Dimethoate

-218-

Soybeans

Grapes

Almonds
vegetables/MeIons/Strawberries
Peanuts

Cotton
Peaches/Nectarines
Citrus

Commercial Turf

Plums

Pineapple

Home Lawns

Other Berries

Strawberry Nursery Stock
Home Gardens

Bananas

Apricots, Cherries, Figs
Ornamentals

Sorghum

Citrus

Corn

Cotton

Tomatoes (Fresh)
Dry Beans

Alfalfa

Grapes

Livestock Premises
Tomatoes (Process)
Pecans

Snap Beans (Fresh)
Apples

Snap Beans (Process)
Safflower

Lettuce

Soybeans

Brocolli

Tobacco

Citrus Blackfly
Forest Seed Orchards
Peppers

Wheat

Pears

Ornamentals
Turnips
Swiss Chard
Cabbage

$8,293,260

$3,200,000

$2,289,390

$2,238,720

$2,140,650

$1,782,000

$1,203,180

$801,040

$363,000

$302,840

$250,660

< $134,000

$61,640

$10,720

< $31 350

"Negligible"

(assigned value of $100)
"Negligible"

(assigned value of $100)
?

$2,831,370
$2,067,768
$1,920,000
$1,690,648
$1,323,576
$1,081,404
$627,750
$617,014
$552,552
$544,050
$226,533
$198,090
$162,000
$151,258
$105,507
$71,982

$51,000

$47,988

$3,648

< $3,203

$1,698

1 $837

"Minor Use"
(assigned value of $100)
"Negligible use"
(assigned value of $100)

‘0'0'0‘0

-219-

Table A2. Ranking of Uses for Each Pesticide According to Annual Total User
Losses from Pesticide Cancellation

Pesticide

Pronamide

2,4,5—T

Silvex

Amitraz

Chlorobenzilate

Endrin

922
Lettuce (Salinas)
Lettuce (Maria)
Alfalfa

Lettuce (Imperial)
Lettuce (Other CA)
Lettuce (AZ)
Berries
Ornamentals

Turf

Sugarbeet Seed

Rangeland
Rice
Noncrop Areas

Sugarcane
Prunes
Apples
Rangeland

Rice
Noncrop Uses
Pears

Pears
Apples

Florida Citrus
California Citrus
Texas Citrus
Cotton

Fruits & Nuts
Arizona Citrus
Other

Small Grains (Major Uses)
Orchards

Conifer Seeds

Cotton

Sugarcane

Alfalfa

Ornamentals

Watermelon Seeds

Other Vegetable Seeds
Tree Paint

Bird Perches

Small Grains (Minor Uses)

Total Annual User Lossfi

$10,097,068
$2,429,724
$2,333,775
$1,305,390
$784,1CK)
455,655

‘0'00000

$16,406, 000
515171000
?

$3,800,00 to $10,100,000
$1,800,000

$1,000,000

"Small"

(assigned value of $100)
"Not Significant"
(assigned value of $100)
"Not Significant"
(assigned value of $100)
?

$8,271,000
$-410,07O

$27,177,600
$2,280,200
$240,400
§_$69,000
?

?

$14,700,000 to 15,400,000
$2,645,000 to 513361000
$3,000,000

$717,850

$4,600

~ 0

~

DBCP

Dimethoate

-220-

Peaches/Nectarines
Soybeans

Grapes
Vegetables/Melons/Strawberries
Citrus

Almonds

Peanuts

Pineapple

Commercial Turf
Strawberry Nursery Stock
Plums

Home Lawns

Cotton

Other Berries

Apricots, Figs, Etc.

Bananas
Home Gardens
Ornamentals

Grapes

Corn

Tomatoes (Fresh)
Snap Beans (Fresh)
Dry Beans

Cotton

Alfalfa

Brocolli

Pecans

Citrus

Citrus Blackfly
Snap Beans (Process)
Apples

Safflower
Livestock Premises
Tobacco

Peppers

Lettuce

Wheat

Pears

Forest Seed Orchards
Soybeans

Tomatoes (Process)
Sorghum

Cabbage

Swiss Chard

Turnips

Ornamentals

$26,890,000

$21,670,000

$14,500,000

$8,950,000

$8,834,000

$6,800,000

$6,200,000

$2,200,000 to $5,600,000
$1,500,000 to $5,600,000
$4,600,000

$2,750,000

$2,600,000

$1,000,000

"Negligible"

(assigned value of $100)
"Negligible”

(assigned value of $100)
"Negligible"

(assigned value of $100)

9

$9,900,000
$8,030,000
$3,900,000
$3,600,000
$1,800,000
$1,730,000
$1,726,000
$1,270,000

$745,800

> $551,000

" $234,500

$130,800

$90,000

$34,000

$30,900

$5,600

3,$2,700 to $-l,700
> $400 to $-1211500
-' "Minor"
(assigned value of $100)
"Minor"

(assigned value of $100)
$-337.50 to $-675
$-21,6OO

$—371,000

$-608,000

M0000”)

-221-

Table A3. Ranking of Uses for Each Pesticide According to Annual Per Acre User
Losses from Pesticide Cancellation

Pesticide

Pronamide

2,4,5-T

Silvex

Chlorobenzilate

Amitraz

Endrin

Use

Lettuce (Salinas)
Lettuce (Maria)
Lettuce (Imperial)

Lettuce (Other California)

Alfalfa

Lettuce (Arizona)
Berries
Ornamentals

Turf

Sugarbeet Seed

Rice
Rangeland
Noncrop

Prunes
Sugarcane

Apples
Rangeland

Rice
Noncrop Uses
Pears

Citrus (California)
Citrus (Florida)
Citrus (Texas)
Cotton

Fruits and Nuts
Citrus (Arizona)

Other Uses

Pears
Apples

Orchards

Small Grains (Major Uses)
Conifer Seeds

Sugarcane

Cotton

Alfalfa

Ornamentals

watermelon Seeds

Other Vegetable Seeds
Tree Paint

Bird Perches

Small Grains (Minor Uses)

Per Acre User Losses

 

$54 to $408
$54 to $402
$50 to $55
$50 to $55
$8.05 to $38.14

'0 '0 '0 '0 '0

$14.00

$11.00

"Little"

(assigned a value of $.25)

$222.00

$40.00 to $-1.SO

$20.00

”Small"

(assigned a value of $.25)
”Insignificant"

(assigned a value of $.25)
"Little"

(assigned a value of $.25)
?

$90.00

$47.00

$4.72

< $3.20

3 $2.88

"Insignificant"

(assigned a value of $.25)
'3

$170.20
$-7.89

$34.00 to $69.00

$33.00 and $1.66 to $1.94
$20.00

$9.20

$4.70

NJ'O’O'U'UOO

DBCP

Dimethoate

-222—

Strawberry Nursery Stock
Pineapple
Peaches/Nectarines

Plums

Commercial Turf

Citrus

Grapes

Ornamentals

Almonds

Home Lawns
vegetables/MeIons/Strawberries
Soybeans

Peanuts

Cotton

Other Berries

Home Gardens
Apricots, Figs, Etc.

Bananas

Snap Beans (Fresh)
Brocolli

Tomatoes (Fresh)
Livestock Premises
Pecans

Corn

Lettuce

Apples
Dry Beans

Peppers

Cabbage

Grapes

Snap Beans (Process)
Citrus

Alfalfa

Tobacco

Citrus Blackfly
Swiss Chard
Turnips
Safflower
Cotton

Wheat

Pears

Sorghum

Soybeans

Forest Seed Orchards
Tomatoes (Process)
Ornamentals

$21500 to $91333

$1,240

$640.24

$511

$118.92 to $302.70
$288.70

$261.08

$88.08 to $172.55
$124.42

$88.00

$38.76

$20.74

$18.18

$11.56

"Little"

(assigned value of $.25)
"Negligible"

(assigned value of $.25)
"Negligible”

(assigned value of $.25)
"Negligible"

(assigned value of $.25)

$76.70

$74.15

$43.50

$19.70

$14.34

$12.52

$7.57 to $-.02
$7.00

$6.81

$6.70 to $-4.l4
$3.87 to $—2.77
$3.83

$3.60

Z.$3.58

$.70 to $3.51
$3.48

$2.15

$1.05 to $—4.37
$1.05 to $-4.37
$1.04

$.71

"Minor"
(assigned a value of $.25)
"Negligible"
(assigned a value of $.25)
$-.55

$-l.27

$-2.25 to $-4.50
$—12.37

V)

O

-223-

Table A4. Ranking of Uses for Each Pesticide According to Percent of Crop

 

Treated
Pesticide U§g_ Percent of Crop Treated
Chlorobenzilate Florida Citrus 67%
Texas Citrus 50%
Arizona Citrus 5%
California Citrus 1.6%
Fruits and Nuts 1.0%
Cotton .41%
Other ”Little”
(assigned value of .Ol%)
Endrin Watermelon Seeds 100%
Conifer Seeds 90%
of direct seeded acres
Other Vegetable Seeds "Like Conifer and Watermelon"
(assigned value of 90%)
Orchards 11.2%
Small Grains (Major Uses) 9.2%
Cotton < 2.0%
Sugarcane < 0.2%
Tree Paint "Confidential-Very Minor”
(assigned value of .Ol%)
Alfalfa " 0%
Ornamentals " 0%
Bird Perches ?
Small Grains (Minor Uses) ?
Pronamide Sugarbeet Seed 90.00%
Lettuce 55.00%
Berries 54.00%
Ornamentals 5.50%
Alfalfa 0.44%
Turf ?
Amitraz Pears 60.1%
Apples 10.0%
Silvex Prunes 80%
Sugarcane 15%
Apples 10%
Rice 0.08%
Rangeland 0.01%
Pears ?
Noncrop Uses ?
2,4,5-T Rice 12%
Rangeland 0.1%

Noncrop uses ?

DBCP

Dimethoate

-224-

Strawberry Nursery Stock
Vegetables/Melons/Strawberries
Plums

Almonds

Pineapple
Peaches/Nectarines
Grapes

Peanuts

Citrus

Soybeans

Cotton

Home Gardens

Other Berries

Apricots, Cherries, Figs
Bananas

Ornamentals

Commercial Turf

Home Lawns

Tomatoes (Fresh)
Brocolli

Grapes

Lettuce
Safflower

Beans

Pecans

Cotton

Citrus

Tomatoes (Fresh)
Forest Seed Orchards
Sorghum

Apples

Alfalfa

Tobacco

Citrus Blackfly
Corn

Pears

Wheat

Soybeans
Livestock Premises
Peppers

Cabbage

Swiss Chard
Turnips
Ornamentals

<

100%

.l% to 95%
70%

54%

46%

44%

31%

23%

7.9%
2.1%
2.0%
0.5%
0.1%
0.1%
0.1%
?

?
?

AAAA

83.00%
33.00%
30.60%
20. to 25.%
22%
21%
17%
14%
12%
10%

4%

3%
2.60%
1%

< 1%

< 1%
0.60%

< 0.50%
0.37%
0.03%

°O'\J°\J°\J°\J'\J

APPENDIX B

CORRELATIONS BETWEEN EXPLANATORY VARIABLES
AND QUALITY OF INFORMATION

.APPENDIXIB

Oorrelatims Between Explanatory Variables and Quality of Information

0090009 6303300 on 330 £5.00 0: use... 0:00... 13!... H0850 05 .m 56:09? 5 00.33 :0 non

{0.300.393 Do:- 0:00... :0}... H0850 05 5 $2034 5 0033 30 mos

N

402690 ..xL 00005:ng 5 L085: 05 an 6000:3006 3 000: 30 now 538535
no 335.6 05. 830302“ 0 no 0005 c0033 532585 «0 3.3090 0...... 5 83033 0: 003 0.0:...

 

 

 

 

 

H
2H3): Amy), AH... mo”. Amen: Amen, oooozuoehn
ax: ox: .mva, A~vun Amy), Amen: nouuaec
ox: A: o\c an Assn: H: canoes
o\: a. H, Amen, Assn: ANS): mamasuoooooofinu
4.. El 5... H _ E... El Boo
ox: 243:, .H.11 .031: 2H2): Aden). 91m.v.~
ax: «Han: .ans Aden: .H.un «ecu: nooahm
ax: «ox: H Assn: Assn, Alwyn: assessors

commas: chaououos oomoooo cougars uofla o3 woos oooowooz oehowuoom
\oosuoooouoom mono unmouom
#0? 0.30098
8300

 

Ilse an 60.5000: es in? ual-0:00 so
cowuesuoucu no 5330 use 30030335: 3 000: 0303000 «0 33> .0033 08330.38 .3 0.38.

-225-

-226-

Ona.

boo.

 

 

 

9H. 8H. So. So. So. So. 4.8. 21 So. oHH. 3on .12
RH. SH. SH. 08... 08... 08... 0.8... 08.- 08... .08... .108... :3 4.68
3850.8

ox: ox: .5! 8.!8.!8.! 8.! 3.! 8.! 8.! 8.! 3...! «82.5
ox: man. 4: 8.! 8.! 8.! o o mam. o 8 Han. 5.66
a). m. m. 8.! 8.! 8.! m. m. m. m. m. n. 3358836
4: 2.! H 8.!8.!8.! H 8.! E! 08.. 8.! 8.! some
a}. H H 8.!H H H 8.! H H H H .89 388...
ox: H H 8.!H H H 8.! H H H H Bfiflg
H.133

ox: 8.! 8.! 8.!8.! 8.! 8.! 8.! 8.! 8.! 8.! 8.! .838
4: ox: H ETEIH H 3.! H H E! H 83828..
. Heavens-sou

ex: ex: H 8.!8.!H H 8.! H H 8.! H nuisances.
8228..

:8 5.6» coo Homo... 33 3.6 5H... 83 2582 8H» 83 9382 «3038..

(3:: 600.133» .88 and ..H :83. :88 .982 :88 1.83..
. 18.6258 a coon .
L348!) !-ooHuoH£cH !Hefioou!
sew. 0.26095
DEED

8600 an 603000: on 3.0:. 583an 9596 so
839.585 uo 3330 are 090500355: 3 see: 030300.. no 03; c0030.. 053309900 .3 038.

wNH. emu. om.” .

 

 

 

 

8H. 8H. 8H. 8H. 8H. 8.! 8H. 8H. 8.! 3588
ex: o}. 8.! 8.!8.!8.! 8.! 8.! 8.! 8.! 8.! 8.! "8;!
ox: H: {a 8.! 8.! 8.! 8.! 8.! H- 8.! 8.! 8.! 5.65
a). 8.! 8.! 8.! 8.! 8.! 8...... 8.! 8.! 8.! 8.! 8.! 34223825
4.. 3.... H 8.... 8.! 8.! H 8.... E! a.) a... a... _ 88
a). 8.! 8.! 8.!8.!8.! 8.! 8.! 8.! 8.! 8.! 8.! 0.04....
o). 8.! 8.! 8.!8.!8.! 8.! 8...! 8.! 8.1» 8.! 8.! .838.
w 4.. ox: H 8.! 8.!H H 2.! H H 8.! H 858865:
n {a {a H 8.! 8.! H H 8.! H H 8.! H H3805
_ 8mg
ca... 5.80 ......H. goons 38 nose 83 83 0592 con. 8H» ”.592 663mm.
Imus: 1000...\0>3 100:0 @090 CH 139. 10.50 nah—00nd louse unkempt
1800.63. a room
!.CBoHHT!.. !!coHuoH£:H H058!
x02 gm
05.8
0530 an 6005000... as 3.0.? Levee.— oce noun: :0
838585 00 5:80 6:4 883835: 3 one: 8638.. 00 8H; :83... 0.833938 .8 Home

 

-23l-

 

 

8... 8... 8.... 8.18... 8... 8... 8... 8... 8... 8.-- 8.... 3858.3
«S. ox: 8.! 8.!8.!8.! 8.... 8... 8.... 8.... 8... 8... «85.5
a}. H- ax: 8...8.le.-- 8... 8... 8... 8... 8... H- 5.65
Q: 8... 8... 8... 8... 8... 8.l 8... 8... 8... 8... 8... ouuHHucBouoHfi
{a 2... 8... 8...8.|8... 8... 8... 2... 8.! 8... 8... 82
ex: 8... 8... 8...8.|8... 8.| 8.l 8... 8.| 8.| 8... .H..m.v.~
a). 8... 8... 8... 8...8.l 8.... 8... 8... 8... 8... 8... .258
a}. a). 8... 8.l8.l8.l 8.... 8... 8.! 8... 8.| 8... 8355....
ccm 5x0» :00 0003. yawn 080 . :03 :03 9.59.2 :03 :03 £582 003%.?!“
I35: .3883... I008 mono 5 Ignaz 10.5.5 .339. .859 .9634

.8338. w 80..

.....ESoHTI .....coHuoHflcHll. £3.5ch
.32 85396
2:8
5833993.... 6:3... 3880338qu 3 .328

839.585 uo 33.3 c5 3953855.: 8 no»: «3038.. we 83. 5253 33.3.88 .8 0H3

-229-

3...-

3....

3...-

3.!-

3....

3...-

3...-

3....

3...-

3.-..

3.]

8..- 3858.8

 

 

 

 

 

a}. a}. :T- 8..- Ho.- 8..- 8.- 8..- 8.- ANTI ST. ST- «835
a} H- «\c 8.- 8..- 8... 8..- 8... H 8..- H- H- 5.65
a}. m. m. 8... 8..- 8... 8..- 8... m. 8..- 8... m. 333.3826
a}. a). c\c a}. .m.\: c\c «\c «\c m\c n.\: o\: c\c mom—o
u\: H H 3.-- H H H 2..- H H H H 363..
a}. 3..- HH.- 8.- 8.- 8..- 8... 8..- HH.- 8.! 8..- HH.-- 53%
ox: ox: 8..- 8... 8..- 8... 8.! 8.| 8..- 8.| 8.l 8..- 8295.:
:00 5x0... cum avg ”.03 ago :03 SH... ”.522 :03 :03 ...—.5054 833mg
.32: -0u0m\o>Hu Iooco no.5 5 I33. I89: Lacuna I8.—.5 .9334
-2683: a v00...
ll-ifloHol-l 53225 Hanan-...!-
mem . . 9560me
nEEmU
8.56 .3 gnu: an 52.5 3 00898 9888 9. ..on :0

COHHEOHCH HO i0ng 05 FuguUOflHDH-fl: 00 D8: afiduflg HO 8Hfl> C0888 gdufldflhhop

.mm 0.38.

. ...—W ...-.... —"'..--v' ‘v-Pi

 

 

 

 

2... AH... AH... STE-.3... 8.! a... .H... 8..- S... 3.! 38583
Q: a}. AH... 8...8.|3... 8..I 8..- 8... 8..- 3... 2...- ”83:2
Q: 8..- Q: 8...8.|3... 8..- 3... 8..- 8... 8..- 3..- 5.65
a}. AH... AH... 3.....3....3.- 3.... 3..- o o 0 AH...- 33359835
Q: 2... .H.3.ls.l_.8...§l 8... 8...:.§....§I 3... 3... 68
Q: a... 8... 3....o.l.o.... 8T 8... 8.... 8... 3... 8.... 3......
a}. 8..- 8... “QT-3.13..- 3... 3.-- Ho... 3..- 3... 8..- .833
.
m .3. {a 8... 8....2l8... a... 8... 8T 8... 8.... 8... 8.5.8..
.
cmm :28» com. yoga. umHo ago :03 :03 “.592 SH... :03 “.522 «gonna
-39: .OuwggHu .88 no.5 ...H .33. -38 .9034 -88 .9022
b.6959. a 30..
...-..GSaHoII- ...-83225 Hanan-......
x92 09.5098.
28
g E 08:08: an 3.3... pawn—Dinah .3
SHugucn uo 3330 can announce—Eu: o» no: 030308 uo 3H!» c0253 08339.58 .8 033.

~233-

 

o\c o\c m\: a}. {a a}. a}. m.\: a}. o\: «\C 9.850 a:
H H _ H H- H E.- o H- H- H. H. 5.68
x92 omen $25096 33¢ 095950 Oman mascuumm 33m 0,508.6 mo ems 850098 Bandung
o8“. 33338.. H3238
“H595

 

uwuuomm 0809.85.—
ou x»?

330:: ca 53189 33¢ «cons—Eon uomuoucoz 5 0068000

2:50

93.00 3 con—Hana: no 5.02. inc: :0
cons—Bowen no >373 wen nuguoouacn: 0» an»: 8332..“ uo 81$ c0253 95339.58

.3 0.309

-234-

 

 

 

a? ome. «\c 91. H Re. 585
ox: H- H- 3... 2...- HNT. 33359836
HHT. El ST- momf E.- 3... o 8:5. ~93
3| 5.. ST. .- 3T. 2T. o no.5. noun:
”3.2523
ox: El HHT. H m.- o .836
{a «\c .2.- HNT. 3T. .3.- 3.33
ex: ex: 0 E.- El 3... 82208
HS. 3.. HHT. ST. ST. 2... all}
a}. El .3.- o E.- El 9 8.5. .333
a} El H2.- H E.- El o 8:5. 3&2
”Ban
ESE-@392. 385m auHoHcomooco Rum-a. 3.3 5 u8m 8:38". 03033..
2,3260qu 00.5 ucwuuwm
x3”. . Panacea
8:8

I530 an 00.530: on 3.0:.
nag-6:00 co coUEoucH no 5330 can @3088 99.6 no uncouom c093»!— ncouucnouuoo 63 038.

-235-

 

 

 

 

o8.- .._wmo..-....iono.-. H.- H.- H.- H.- H.- 80. H.-i H.-. 30. “v8.5.5.3
«no... . e8.- vnof H.- H.- H... H.- H... 8o. H.- H.- 80. moms». noun:
32.3 “.2
8o- wmof mg.- mva 2:.- mva ava mva ava «.3.- mva 2:.- 002.383..
33.- vnor v8.- vHH.- vHH.- vHH.- «HH.- «Si «Si vHH.- «HH.- vHH.- o cocoa 88..
«u: goggles
”3858.3
{a H H HoT-HSIHSI HEI 8T- H 8T- 8T- H .823
a} a} El 8.1.8.18...- El El E.- El E.- El 353
ex: ax: m.- EIEI H- H- E! H- 7 El 7 838:8:
~30ng
{c ox: 7 El 2T- 7 7 El 7 7 2T- 7 .HQHEHHES:
amid—Bum
ax: H- 7 8T- ? H- 7 8T- 7 H- H- H- 838:9?
.3339
ex: 7 7 8T- 7 H- H- 8..- H- H- H- 7 “£8.35:
a-m.v.~
a}. 3! H 8T- 8T- 3T- H Hovl 2.7- H 5! Helooocazug
a? E! H ST- 8T- 8T- H HoT- El 7 3T- 5! o 8.5. 9&9
.58
a). HHH. ox: 8T- 87- ST- mnmf nnnf HHH. mumf nnnr HHH. 5.65
a). H H 8T- 87- HoT- H H H H H H 8.223825
.600 5x3 :00 goons. uoHo 00:6 :3... :03 0.59.2 c3... :03 gig 033.30;
Liz: laumm\c>3 50:0 no.5 5 I33. no.5: nah—03¢ I939 £9034
gamma a 08...
Ila-tauoHc-Il-I IIIIICOmuEflEH agave-tilt!-
xufi. 8:896
0530
'36 an @8533: a 3.0:. yea-03;
H.566 co 839585 no 5230 can H.338. a8... Ho £888 5938 20338.30 HS .38.

-236-

 

 

 

mme 3.1- mme anf mme mme mme 8H? ST- mme maHf 8T- u omen: ~33

~H.H.- ~H.H.- NHL.- ~FH.- «Si «2.- .2.H.- «Bf ET- ~S.- «2..- 8.-- c macs. .398
_ 8852.3
{a mmm. {a 8T- 87- 8T- 8T- 8T- mmm. H8! 8T- 8T- 5.65
«\c ST. 8T- 8.-- 8T- H8! 87- ST- 87- ST- 87- 87! 33359825
a} 8T- 8T- 3T- 87- 8T- 87- 8T- 8T- ST- HoT- ST- H.433

{a :T- H 8T- 87- ST- H HoT- El 3T- El ST- omcs. 36..

a? El H ST- 3T- 8.3 H ST- ?T- ST- HSI ST- omcs. yuan:
. . noun
a}. BT- Ho.| 8T- 87- ST- 87- 8T- 8T- ST- 87- 8T- 333
{a a} .2..- §.-. ST- ST- E.- El .2.- El 5... El "825

a} «\c H- E! ET- H- 7 ET- H- H- E! H- ugflauocoz

ax: {a m.- 5% E...- H- 7 El. H- H- 2.7... H... . H3885
82.2.0.5
50 5.8» 5.6 gowns 33 ago :03 :03 ”H.592 :03 c3... peso—.2 333mg

..uusz ..Oumm\0>wu 50:0 85 5 133. I38 A5034 sou—.5 Inbound
$883,. a coo...
l-IEBoHoII- lines-HES Hanan-Ill
via madam L
9:8
258 E 88.5: 8 £qu 3.38

Ham .3tz so 83385 ac 3296 can Bug 8.5 Ho agony». 5958 9833830 .3. «Home

-237-

 

 

 

o o o o o o o o o o o o u ~95. ~93
m8- m~o.- m8? m~o.- m~o.- m~o.- m~o.- mmof m~o.- m~o.- Ref 98.- u 2.55. .68:
3853.8
m\c a}. o\c a). o\c a}. {a a}. a}. a}. a}. 0). age—gm
ax: 8T- 8T- 8T- 8T- 8T- 8T- ST- .8T- 8T- HoT- 3T- 333
{c a}. 8T- 8T- 8T- 8T-. ST- HNT- 3T- HSI RT- HmT- 5835
a}. 8T- 8T- ST- 87- 8T- 8T- 8T- Hovl HSI ST- HoT- .H.-m.v.~
o\: a}. o\: a}. a}. a}. m\: a}. «\C a}. a). «\c 0 00:5. “Sing
m\: o\c a}. «\c o\: n\: c\: a). o\c a}. 0).. o\: U 00:5. “Hung
mono
Q: c. ex: 8.! 8T- 87- H mam. m. H man. m. 5.65
H\: HHT- HHT- 8T- 8T- 3T- 2:! 8T- HHT- ST- 8T- .3! 32389825
5m 56» com Human. Has «96 8H» 8.: H.592 SH» SH» ”.522 «3338
I33: loumm\a>3 nooco no.6 5 I33. ..RHHE hue”? I950 £5034
688%,. a coon
ll-Huflch-ll ---coHuoH£:H HEB-ll-
xaHm 858%..
E8

...-8.5%

no 503385 H3”. 3H: Ho 3280 can 3289 no.6 Ho uncouom 5.38 333338 .3. 0H3

-238-

HOV-l

8T...

3T-

Hovl. ST- ST...

HOV-....

ST-

87...

ST-

Howl.

How-...

3852.3

 

 

.o\c man. a}. H3..- 87- H8..- 37- ST- mmm. How-I man. map. 5.65
Q: H H ST- HoT-ST- 8T- Hovl H 8T- HSI H 333:3
I835

{a 3T- 2T- 8T- HHVI HHT- 3T- 2T- 3T- 5|. 8T- HHT- a-m.v.~
a}. m\: a}. a}. a? o\c M\: o\c o\c a}. a). a}. mono
ax: HHVI HHT- HoT-HoT-HoT- 8T- 87- HHT- 8T- 8T- :T- .833
a). {a HHVI HoT-HoT-ST- Hovl ST- 8T... 3T- Hovl ET- "835
{a {a 8.-- HoT-HoT-ST- 8T- Hovl ST- 8T- 8T- Hal 8288..
:00 :onu 50 gowns 30.3 no.6 c3... c3... ”H.582 :03 cod..— uHBQZ 8:338
139: 53303”. I005 @090 5 Lama Imus 16034 I890 Iauoont

3050.8me a 80m

13339-3- ---:oHuoH£:H--.I Ill-Hafiuo-IIII
xnmm 950098..
2.5 3 H338: no 5qu 3 0825 2.033

0» “Hov— co 539.585 no 3330 one @3089 00.5 no ”=80qu c0253 95338.30 .39 03¢.

-239-

 

 

 

3! a... 3... 87:31.5! 8.... E... El ST: 8...: El 3852.3
ox: 8T. 8T. 8T.8.l 8.!. 8...: 8T: 8T: 8.!. ST. 8.... .H..mé.~
a}. H H “8.1.8! 8.!. 8T: 8T: H 8T: 8T: H .838
Q: Q: 3!. ASISVISVI 8T. 8T: 8T: .8! ST. E... «8&5
Q: {a 8T. 87.8.1.8... 8T. 8T- 8T. 8.! 8T. 8.... 8H;
{a :T: 2.88.... ST: ST: ST: .8... 8T: 2.82:: ST: ST: 8...... mum:
{a H- ex: Hal 8T: 3T. 8T. 8T. H- 8T. 8T. H.. 5.65
a). 37.. A3.... «3:187:37: A ST: “8|. HI H H. ST... 30323836
:00 5x3 cam among. “.30 noun :03 :03 H.582 :03 :03 £522 033.38
I35: IO»0&\0>3 1005 00.6 5 ..Hmom 3820 19334 18.5 Iguana?
suing a coo»
llbfloHeull :oHumHBcHll- ...IIIHEBXTIII
Xflfim Ogmaxm
gnu
5308:0033
...—035.535 9.3“ “Huang—non no >9:qu 0:0 ©3003. no.5 uo ...—830m gun 033309.80 6.3 0.33.

-240-

.37.

 

 

 

Q: 3... a} .8! .8! .8! .8! cm... .8! .8! man. 5.65
{a .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! 32355836
.8! .8! .8! .8! .8! .8! .8... .8... .8! .8.... .8! .8! 3852.3,
mx: .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! .H..m.v.~
Q: .3! .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! .68
a}. .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! xgHHm
Q: a}. .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! «8.52
ox: a} .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! 8238..
50 5.8» cm... ”.8.... ”.08 88 co: 8: HES: 8H» 8H”. ”.522 8338..
-32: acuwm\w>3 .005 00.5 5 .33. .820 g sou—.5 .9334
-Scouaum a coo...
. !.LG5mHo.!! --!coHuoH2cH .. !!!HnE.oa.!.!
xmfi. 3:898.
«5.8
228 E 0888: 8 6:3... 3 .883
38.5.. no 838585 uo .3230 can Yang no.6 uo €8qu 51.58 3033830 .an oHnua

8T .9!

-241-

 

 

 

 

{a .8! .8! .8! .8! .8! .8! .8! .8! .8! .633
{a {a .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! «EH5.
{a a}. a}. a}. a}. a}. o\: a}. a}. «\c a}. a). 82.0.5.5
o\c c\: o\: a}. a}. H..\c m.\: a}. M\: o\: m\: a}. mono
a). .3! .3! .8! .3! .3! .3! .8! .8! .8! .8! .3! Tm.v.~
ox: .8! a}. .8! .8! .8! .8! .8! .8! .8! .8! .8! 5.65
{a .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! .8! cuoHHucfiouoHfi
o o o o o o o o o o o o o 8:5. ~16;
m3... m~0.| mmo... m~0.- m8... m3.-. £5... 93... mmof m8... m~o.l m8... o 09.5. noun...

3858.3
50 :36... com goons ago 9.50 :03 :03 .8595 :03 :3... 9582 03039.8
33:: .39.};3 .005 980 5 Law: -025 .8034 -850 g

-2680.»— a moon
!!§oHo.!. 83225 358...!
wad. 0260me
2380
£38 an 08.63: on 3.0:. 39.6

@580 Hots. co 83.3.85 .8 3330 can 038.6 noun «0 ”.582. :83! 983368 ...B ~38.

-242-

m\: a}. a}. a}. a}. a}. a}. o\: n.\: a). a).

 

8050 A:
m... ~.I mi mom. mm. o mmw. 0 ~... nnn... mmH. 5.65
mem omen 9.50096. “.03. . «6.59.00 83 @333: x92. 0.2.60me uo Oman 8:398. 033.30..

coo... 52358.. H3238
“.582

 

8.8% 30985 «SHE? 3 5338. 3:2 2.2”.ng ammuflcoz S 8258.

3 .anm
95.8

. g 3 08:00.: on
3.02 1:55 :0 839502..— uo auzfld can 030099 00.5 no c.5090.“ 602.500 acouuanouuou

oonn Dung

-243-

 

 

 

Q: 3: .3! .3! .8! .8! ”8.2.:
.3! .8! .3! m. .8! .8! 08:... 080..
.3! .3! .3! o .8! .3! 08.... .33..
30850.3
3: 0%. 3: n3. 0 8... 8:9. 006..
ox: n3. ox: 8...- H an. 8:9. 8&0
280.0
ox: .....- m.- .8! .8! .3! 33350835
ax: .3! .8.- H .8! .3! 00:5. 026..
Q: .3! .8! H .8! .8! 8:0: 0.0....
. .000:
Q: .3! .3! H H- m. 8:5. 00.6..
3: .3! .3! H m.- o 8:3. 088
3:530
Q: .3! .3! .8! .3! .3! is;
Q: Q: .8! .8! .8! .3! 83.88..
330300032: 000300 auHoHcmmouco 0000009 003 ca 0000 0003000 0303000
. 05000000000 - 000w 0000.000
0.03. 0.000090...
0.5.00

...-:00 >0 0090000: 00 500.5080 3 x03. 00 0038005
uo 3:36 000 003030.008 00.6 00000.. 000: 0000 000 13000 0003000 0003309000 .03 0309

-244-

 

 

 

Q: Q: .Ql .Sl .8l .8! .Cl .NT .Nvl El .NT ....l N835
RH. 8H. 8H. N. N. N. N. N. 8H. N. N. :2. 03H: ...:
RH. 20. RH. «.8. m8. 3:. mm... 08. 8:. m8. «.8. n3. «\0 a :8:
. .0005. 0033
3:. So. :8. N. N. N. N. N. :2. N. N. EH. 3on 02
m8. m8. .08. 08.: 08.: 08.: 08... :No... :No... :No... 08.: 08.. Q0 0 :8:
”00000 0000:
"38:02.3
Q: nmm. Q: .8l .o.l .8l 0 HHH. 8.... o HHH. NNN. 8:5. 026..
Q: 5... Q: .8l .8l .8l 0 HHH. 5... o HHH. S... 8:5. 08::
«000000
Q: m. m. .SI .SI .8l m. m. .... m. .m. m. 32.28835
Q: :7 H .8l .el .Sl H .8l El ... .N.l El 8:5. :83
Q: .HT H .Sl .8l .8l H .8l ....l ... .N.l El 8:5. 08::
. «0000
Q: 58. $0. .8l .8l .8l .Sl .8l 8N... .Sl .Sl 8N... 8:5. 026..
Q:. NR. nnn. .8l .Sl .Sl .ovl .8l 2:. .8l .8l 9H. 8:5. 08::
uxgnum
Q: o 0 .8l 0 o 0 .8l 0 o o 0 80008...
Q: o o .o.l o o 0 .8l o o o o 888085.
£13}
Q: Q: .H.l ....l El 5| 3| ....l .3l .H.l .....l .....l 82.8835.
Q: Q: .Sl ....l ....l .N.l El .....l. .2..- .....l El .....l H3805
“03000000
:00 . Canon. :00 Hg 00:— 0006 :0“. 8a» 859—2 Sm... 8m“. 0592 saneuuua
I000: 0000\030 .005 0000 5 I000". I002. 10000.0( 30000 0000.2
63805. a 08..
lifloHall lSHuoHEE H88ulll
0.02 00000000
05.8

300 3 0000000: 00 3.03 000003004 000000 00 030000000.“
00 5:86 000 0030300000 0000 00000.. 000: 0004 000 H3004 0003000 0003300000 .80 0300.

-245-

 

 

 

0m~. RN. 0mm. 8w. own. 8w. Omm. 0mm. 2:... omw. emu. 3T: moan: .833
H8. H8. H8. H8. H8. H8. H8. H8. 8T: H8. H8. ST: 8.5. .88:
"3353.3
«\c a} :ul 8T: HSI Hovl .3... ST: ST: HSI 3T: 2...... any»?!
Q: ~. a} 8T: Svl 8T. 8T. 87.. a. 8T: 8T: 8T: coca: 3.64
ex: ~.u a}. ST: ST. 8T: 8T. 8.!. N... 8T: 8T. 8.!. coca... .88:
"5.65
a}. 3T: 3T: 8.!. 8T: HoTI 3T: 3T: 8T... .81... .8!. 3T... auoHHuconouoHpD
ox: HHH.... H 8.!. .8.... 8T: H .8.... El. 5.... 3...... 3T... 095. 3:3
a}. .3!. H ST: ST. ST: H 8.!. .Htl GT. .21. QT: 8:5. .58:
Evan
{a 8T: Hovl 8T. 8T: HSI 8T: ST: ST: ST: ST. 8.!. 538
ex: ST: HSI ST: 8.!. 8T: ST: ST: 3T: 8T: 8T: .8.... sum.e.~
{c ox: 2T. HS..- Ht.... 5.! .3!. El 8T: HHTI 2...... ST: 8&8?
.93:
a}. Q: :33 .3!. 2...... 37: ET. 2...... 3T: HSI 2T: 3T: H3835
cog
com 5.8... cam “.38. ”.93 $3 c3... :03 £582 :03 :03 “.595 .3338
I32: lauon\a>3 vouco menu 5 lama: lung nah—89¢ 38:5 nah—03¢
gum”. a coo...
[unabsoHH—ula unicoHuungcH {laugh—Ts
x»: 9.5.3me
8.56
2.50 an @833: no .Bmcuoa can 30tz o... xufi. no 838585
no 52qu can 5333056 60.6 cocoon you: 85¢ you H352 c0933 98338.30 .Hwn 038.

-246-

 

 

 

 

omo. omo. omo. one. omo. omo. omo. . cmo. omo. omo. omo. omo. coca: ua:oa

nmn.u nmn.a nmm.u nmn.: nmm.u mmm.a mmm.u mmm.u mmm.u nmn.u nmm.: mmn.u omega pang:
"ouoozuueHo
u\c ox: Hmvnu Hovuu Ho... Ho... HNVun HNHIu Hm... H~vui H~vnu .mvun nouuHez

ax: w. ax: Ho... Hovun Ho... o mmm. ~. o man. ~. omega umzoa

{a N. {c 8T. 8.!. 8T. H mam: N. H Raf ~. 8:8 noon...
”cHuonm
Q: .3... HHT. ST: 8T. 8T: 8T. 87.. .HT. .8! ST. ST. 30323826
a}. Q: a}. n\: u\: M\: Q: a). «\c a}. a}. o\c numo
u\c Howl: Hovsu Hovun Ho... Hovuu Hovnn .o..u Hovun Ho... Hovnu Hovuu xo>HHm
ox: .8.... 8T: 8.!.8! ST. ST. .8.... 8T: 8T: 8.!. .8l .Tmiw
a}. a}. as: a}. a}. a). o\: a}. Q? a}. «\c a}. age—scan.“
=8 5.6» cam H83. 38 8.6 8H» 83 235 83 SH”. £52... 8338..

33:: I0u0u\w>uu I005 00.5 5 ..Hnmm 18:5 lab—03¢ louse Inbound
nosoouama a coon
. .unnuuauuuoHouuunu coHuchch uauuunuHasuoouuunuu.
x3: 25009.n—
usage
9&8 S H.888: no 3.3.. 8:8 8 83.585

Ho 3230 as. SHuoHHoocoo .83 833 you: 82 ...8 H35... 59.58 3033038 .-n oHnnH.

-247-

«3|

.8!.

ST.

8T. HOT. 37..

8T!

8T.

3T.

3....

H8...

8... 3852.3

 

 

 

{a .3. 5.. 87.878... 8T 8T 8.... El .2! El «835
a? 98. ex: 8T. ST. ST. 8T. 8T. Q:.. HSI N. one. was». .533
{a .3... Q: ST. HSI HSI 8T. 8T. H ST. N... R... 8.5. 8&2
”5.68.
«\c .... m. 8T. 8.... H8... 8T. 8T. m. 8.! 8T. m. ouoHHNcBBoHfi
m\c H..\c a}. a}. n\c Q: a). a}. m\: o\: a). a}. name
a}. ST. ST. 8T. 8T. 8T. 8T. 8T. HHT. .8... ST. HHT. macs. 3.6..
a}. HHT. HHT- Howl 8T. 3T. 8T. 8T. HHH.... 8T. HSI HHH.... no.5. 8%:
«xgHHm
a}. H H 8.... H H H E... H H H H eun.v.~
a). o\: 3T. HoHIHoTuHoT. ST. 8T. 8T. 8.... 8T. 8T. 829.88
5.5 5.8» cmm ummua. ”.03 none :03 :03 £595 :03 :03 H.595 033qu
I33: Ioumm\0>Hu I005 Q95 :H .33. louse Iguana. Ion—.5 .9532
-8339. a coo...
......LGBoHoln... ......coHuoHBE- ...llHnanli
#3: 18:00me
DEB"...
......8 F 32.3! an .315 3 3898 3038 B so? co SHusuoufi
Ho 3280 can. SHuuHHoocno 5.: 883 3.5 83 .5. H352 $258 9833830 .3. .36

-248-

 

 

 

 

3.- 3! 2T- 3T- GT- ST- 8T- 3T- 3! 8T- 5l 3! mafia :38
Hal 2.! 2T- El HmT- RT- .8-- .2l 2T- .8-- ET- HHT- 8:5. noun:
“38:...23
Q: Q: .3! ST- ST- 8T- 8T- . ST- 8T- 8T- 8T- :T- «85?:
.Q: T Q: ST- ST- ST- 8T- 8T- 7 .8| ST- ? 8:3. :93
Q: T Q: 8T- 87- ST- HSI 8T- H- .8! 8T- 7 $5.: .58:
”5.65
Q: HHT- .HT- ST- ST- 8T- 8T- 87.... m.- m... :m. HHT- 33353835
Q: 3! 2.88- Hovl ST- SVI 8T- 87- HHuotl ST- 87- ST- oocn: 5.3
Q: .HT- .Huonvl ST- ST- .8! ST- .oT... Snot- 8T- 8T- .Sl 8:5. .68:
E8:
Q: mam. mam. 8T- 87- 2:...- 87- ST- mnn. ST- ST- nnn. coca: .83..
Q: H H 3T- .8! 8T- 87- ST- H. 8T- 87- H omen: noun:
uxgfiwm
Q: 8T- 87- ST- ST- 8....- .o...- 87- 8T- 8T- 8T- 3T- .H-n.v.~
Q: Q: 8T- .oT- ST- ST- ST... 8T- 8.-- ST- .oT- BT- 82.38:
cam 5.6g 50 amen. ”.03 no.6 SH... :03 956:2 :03 :3... 950:! 030308
I32: ..Ouwm\m>Hu 60:0 095 :H ..Huom Inna 10.534 louse A5034
$883: a coo...
. IICBwHo-II ---:oHuoH:::H Hanan-«III
x»? 95896
05.8

no 5330 0cm coHuuHHoucmo :53 0230.” you: 804 ...—0.. H352 c0153 08338500

8.56 an cannon: no .uumxuozfiom 3 ...—3.x :0 cons-boas
.33 0.309

-249-

How-I

3v...-

«8....

Hog-I Hovl- 8T...

A8..-

ST-

3V...-

ST-

H8..-

37-

BoBuEH:

 

 

 

 

Q: Q: .87- .oT-ST-ST- 8.! 8T- 87- 8.! 5| 87- "83::
Q: c Q: 3T- 8T- ST- ST- ST- 87- ST- 87- o 8:2 :26:
Q: nmmf Q: 8T- 8T- ST- ST- ST- .oT- Howl ST- nmnf 2.5... yoga:

EH65
Q: ST- ST- ST- 87- ST- HSI ST- 8T- 8T- 8T- 8T- 33352835
Q: El ST- ST-ST-ST- ST- ST- E! 3T- 5! 3T- :8:
Q: ST- ST- ST- ST-ST- ST- ST- 8T- 8T- 8T- ST- .838
Q: Hal 87- 8T- HoT-STF 8T- 8T- ST- Hovl ST- ST- is}
Q: Q: 8T- .oTST-ST- 8T- 8T- 8T- .8! ST- .8! 8228:
cam :18» 5o among. ”.30 no.6 :03 SH... £5054 :03 :03 H.522 «30.308
I39: -Ouwm\w>Hu 50:0 080 :H ..Hoaz I025 fig louse Luann?

goo: o :8:
. ll-EBoHo-II- 53225 .l-I-Hsnoall-I-
mem 0.50%wa
8:8
8:8 E H.828»: 8 .nHHHam 3 .5898 2.038 3 :3: :o 83385

:0 32:3 :5 833188 .83 :88: you: 82 you H35: $2.53 9533830 .3: «Hana

-250-

 

 

 

0:0. 0:0. 0:0. 0:0. omo. one. omo. 0:0. 0:0. omo. omo. omo. coca: nose:
nmn... mmn... mmm... mmm... nmm... nmn... mmmf mmm... mmm... mmn... nmn... nmn... 00:0: H000:

«030050.50
:\: :Q: .001: .ovuu onuu onun “0.1: onuu “ova: onuu .001: onun song‘s:
:Q: .00.: . ax: Ao.nu onnn A001: :0... 500:: onun «ovnu .0... .00.: :fiuocu
:\: “ova: .ovuu onnn onau “00:: “00:: A001: onnn onnu onun AoVan mundan:aoouo~:0
0\: :\: m\: o\: u\: 0\: n\: 0\: «\c o\: «\c a}. name
:\: onuu onnu .0::: onun .oVus A031: onun .ovun “syn: .0.nu .03.: xm>nfim
:Q: a H “00:: H H H .mvnu a H a H 91m.v:~
Q: a}. :\: .:\: :\: :\: Q: :\: o\: :\: o\: m.\: 03:30....—
:00 :30» :00 3:39 303 00:: :03 :03 350:2 :03 :03 3:005. 0303qu

13:: 030...\0>3 100:0 mono 5 I30: .025 nah—03¢ Iago Impound . ..
0.6.800: a 000m .
...»...lfiBoEllu ......383223 32:8!!! .
x02 0.3009: A
2:50

...-56 3 00.3000: an .0380 950.5 :33 0» 0.0.:— co 833535

no 33:00 0:: 83:38:00 603 0000?— you: 85¢ :00 35:5 5033!. 9.0339300 .03 0.33.

-251-

 

 

0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0).. 000:8 3.:
0.: o mmn... Hmm. nmm. mmm. Nav. 2.0. 0mm. N. man. 005800304
~.I m. 0 RN. man. 30. :3. >8. mmn. ~... mnm. 00:0: 0000:
«5.65
x02 83 0000096 33¢ 00:50:00 80.: 0000.30: x02 0000005 no omen 0000096 0303000
08.. 333385 13:80:
350:2
02800 080985 8:03: B 03006:. 38: 0:033:00: 0050082 5 8:808
0: x»::
95.00
9.50 an 0000000: 00 3.38 Hus—«:1 :0 Saul-hows
00 3.30.6 0:: 83:38:00 £03 00000.... 000: 80¢ 00.. 3052 000333 000330qu0 .hNfl 030a.

-252-

 

 

 

ST: 2T. ST: ~00. 3T: El 00.5. 0260
2T. 3!. El 02. El El .95. 0&8
385050
a} ax: 3T: 5.... El 3T: 20:5
«\c «.3. a}. as; H S... «was. .333
Q: Re. {a «S... H «av. macs. 3&2
2.00.0
ex: 7. 7 5| El El 303080836
ox: ST: El A El El .95. 026..
{c 3!. 5| H 5.... 8T: 8.5. .88:
:82
Q: ST- ST- H o o .0520
{a El. 3T: 8.!. El. 3!. annim
{a Q: 5.... El. 5... 5!. 83:80
xfiuficwmflaa .3830 5.808088 80309 ”.03 5 08... 8.033. 8833..
0300000900”. 00.00 0:00.000
x02 00:00qu
0.5.00

05 an 00.5000: 00 500.5080 3 x03. :0 :03
-0535 u0 5:30 0:0 53030008 503 00003 000: H309 131:: 50500 0.8338000 .30 0309

-253-

 

80 a3. moo. nnH. HHH. n3. n2. HHH. nnw. HHH.. n2. mnm. Hue:— u:
80 m8. 80. 8o- 8a.- 08.- .- e8- .- 89.- .- 8a.- {was
. «avocad—
ugauoaaa
"3333..
ex: ox..- 3| 87.8.... 8... El 5.. 2.... 5.. a... .3... ~85:
ox: poo. ax... 8T. 8T. 8... H..- 3.... >8. v.- HHH.. Go. omen: 333 .
{c ~. ox: . 8.! .8| 8... ~.- HHH.- ~. ~.- HHH.- H. .95. 8&2
. . EH85
ox: . H H 8T. HSI 3T. H H H H H H 33359825
ox: HHT. H 8T. 8T. 8T. H ST. E.. 98.- 5.. 3... was... ~26;
{a HHT. H 8... 8T. 8T. H ST El 89. 5.. 5.. 2.25. ....E:
. ‘ “82
{a o o 8... 8... ST. 8... a... HHH.- 8... ST. 3.- 5:3
a). H H §l H H H 8... H H H H 833.6
{a H H 8.... H H H 8... H H H H augatoocn:
£133
Q... ax: HHT. El 2.... 5.. .HT. 2.... ST. ST. 2.... GT. augouauucs.
«\c a). 8... 2.7.2.7. 3T. 2T. 2.... El El ST. 3... H3838
«easing .
cum 5x3 :00 008:. avg 003 :3... :03 ...—50.2 :03 :3» 95¢! .3032...—
.3:: .3955: .88 86 5 .33. -88 .989. -83 .983 .
é! o .88
..IlgoHT. ......coHuuHBEll. .lllHnBoQIII
0.589»

 

.32

. I5 .5 v8.63: no 6933331 9580 3 33.. :0 :03
£585 Ho 3:30 a... 83.3850 :8... 883 .88 H32. H35... 5:58 3033.800 dB .38.

 

-254-

 

 

 

 

on... $6. 98. mm... a8. mm... .3... 3... ....l 8... 98. ....l «was. .25..
30. 8... _ «8. 8o. 8... 3... 80. «8. 8.l «8. «8. 8.l «95. noon:
«30050.50
{a {a El ....l 8.l 8.l El . El El .~.l El El 2.5.5
ax: a. ax: .o... .o... .o... .o... .o.-- ~. .0... .o... .0... «menu umzoa
«\c N. a}. 8.l 8.l 8.l .o.l 8.l w. 8.l .o.l .o.l 8...... 8&3
”5.65
ax: .o.-- .o... .o... .o... .o... .o... .o... .o... .o... .o... .o... oaoHHuconouonu
{a El H 8.l .o.l .o.l H 8.l El .~.l 5| 5| omen... .26..
{c .H.l H .o.l .o.l .o.l H .8.l El .~.l El El omen... .68:
"none
ox: .o... .o... .o... .o... .o.-- .o... .o.-- .o... .o... .o... .o... xa>HHm
ox: .o... .o... .o... .o... .o... .o... .o... .o... .o... .o... .o... 91m.v.~
m\: «\c .H... .m... ..... .m... .H... ..... .m... .H... Ah... .m... unmanuanocaz
Q: ...}. El El .....l .~.l .....l El El El .....l .....l 2.39.80
“8338..
.56 5.6.. 5.. “ooh. uoHo no.6 83 83 “.582 .83 83 ....58¢ 8338..
-35: .Ouo....\o>3 .005 00.5 CH .33. -88 3.900.? 19.50 .9631
luv—60.59. . a 600....
rouge! lISHuchscu llllaflnuoolllll
302 890096
95.8
8.56 F 68.68: am .8334 can 8.0.3: on 30.2 no :03
-9585 uo .3380 can 83oHH8c8 .83 828.. you: H32. H85: :83! 30333.8 .02 3%...

-255-

 

 

 

0:... 03.- 0:... 0.3.- 03.- 0:... 0:... 03.- 0:... 03.- 0:... 0:... 00:5. 0030..

3.1- 02... 05.- 00H... 05... 03.- 0:... EH... 00H... 00H... 05... 0:... 00:5. 000.5
. «30050.50
a). a} .n.l 8.l .o.l .o.l .~.l .~.l .n.l .~.l .~.l .n.l se.»?

a}. .... {a .o.l .o.l .o.l H Rm. 0. H nan. .... 8...... .06..

0\: o. 0\: .3l 3.... ST. H mmn. o. H mmm. o. 00:0: 0000:
”5005
a}. El .H.l 8.l 8.l .o.l 8.l .o.l .H.l .o.l 8.l .H.l 32358036
0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: mumo
{a .o.l 8.... .o.l .o.l .o.l 8.l 8.l .o.l 8.l 8.l 8.l .038
a). 8.l 8.l 8.l .o.l 8.l .o.l 8.l 8.l 8.l .o.l 8.l 5.0....
0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0350.500
:00 :onu :00 0000.0 ”.03. 00:0 :03 :03 0:082 , :3... :03 £59.! 0303000

-30: 000003... .095 00.5 5 $00: .002: 1.00%? -0000 neg
$500000”. 0 0000
llEBoHT ...lcoHumHEcH llHQnonull
x03 0&5»ng
0.58
. §E§§3.30HE3§H¢883.

10.585 u0 00:30 0:0 53030056 .83 00000: 000: H509 H0052 c0050: 083300000 2:3 030a.

-256-

3...!

3...:

3...:

8.! 8.... 8T. _

ST:

ST!

STI

37...

3T...

8.!. 3858.3

 

 

 

a}. a}. 3| 8.|8T.8.| 8.!. 8.!. .8.... 5| 8...: GT: «835.
ox: one. {a .o.l 8.1.8.... 8.! 8.!. 2.... 8.1. ~. 08. 85:36..
{c 08. {a .8! 8T. .8.... 8T: 8.! 2.... 2...: ~. 08. 09.5. .88:
“5.6.0
o}. H H 8...: 8.1.8.1. 8T: 8.! H 8.... 8.! H Sofia—.3826
0\: 0\: 0\: 0\: 0\: 0\: 0}. 0\: 0\: 0\: 0\: 0\: 800
ox: 2T: 3.-.. 8.|8.|8T.. 8.!. 8T: 3.... 8.l 8.! .3... .836
ox: 3.... 3... 87:37:27.. 3.... El 3...: 5| El .3... sum.v.~
{a a). 8...: 8.l8.!.8.l 8T. 8.| 8T. .8.... 8.| ST. 8388..
50 5x3 50 aces 03.. 8:0 5H”. 8H» “.58: 8H» 8H» 058.... «3033..
-32. 500...)»: .88 :80 ..H :33. :88 .385. :88 .985. .
.80803 a 08..
lulEBoHollu ..llcoBmngcH Hafiwolul...
0.03 8000me
£5.00
gfigfififlssgucoflgfiifiicofl
......o..& «o 3:30 05 833388 :80 889. nan: H38. H85... 50.58 9833830 .NB 3...:

:TI .31.. :TI .3!. A3.... .3!.

8.! 8.-.. 3.... 8... 8.... 0....

385800

 

 

 

{a Q... 2...: ..oT. 8...: 8T: 8.| 8T. 8.!. 8T: 8T: 2...... 08.02

ex: H: H..\.. 5| 8.! 8.! 8.!. 8...: H: 8.| 8.!. .7 500.0

{a 2.... 3.... 8.... 8.!8... 8...: 8.... H.. H . H 2.130228835

ox: . El .Huon.| 8.!. 8...: 8.!. 8.| 8.!. 200...! 8...: 8T: 8.!. .68
r a}. H H 8.l 8.... 8...: 8.! 8.!. H 8.!. 8...: H .620...
4. {8 .8.I 8... 8.18.1.8... 8... 8.... 8.! _ 8.... 8.! 8.... .160

ox: Q: 3.! 8.1.8...- 8.!. .8.... .o.l 8.!. 8.1 8.!. 3.... 83.8..

:00 :33 :00 00000. 003 0030 :03 :03 0:003. :03 :03 0:09.! 0303000

I000: 0000\030 1005 0000 :0 I009. 10000 50000: I008 00000:

000500002 0 0000
100330.! £832...ch lullHnBoollln
0.03 0000096
2:8
5 >0 0000000: 00 508—003-000 00 80:. :0 :03
-8085 00 .3230 0.... 83330.30 :80 «88.. 08.. H38. H35... :83! 2833-0000 .20.. .30...

-258-

“BI

23'

3......

8... 8... 8...

8.... 8.!. 3.... 3...: 3...: 3......

3858.3

3.3 H5.

 

 

 

 

{a a). .8.... 8.1.8.33.-- 8.... 8.... 8...... 8.l 8.... 8....
a}. o a). 8.... 8.1.8.... 8.... 8... 8.... 8.... 8... o 5.2006.
ex: o ax: 8.|8.l8.l 8.| 8.... 8.... 8.l 8.... o .80. .88...
2305.
ox: 8.... 8... 8.... 8.1.8.... 8.... 8.| 8.l 8.... 8.| 8.... 33388036
a). 2.! 8... 8.1.8.... 8.... 8.... 8.... 2.... 3... 5| 3.... 88
a). 3.... 8.... 8... 3.| 8.... 8.... 8.... 8.... 8.... 8... 8.... 02.30
Q: 8.... 8.... 8.!8.!8.!. 8.... 8.... 8...... 8.... 8.... 8.!. .H..m.v.~
a}. a}. 8.... 8.l8.l8.l 8.... .8... 8.... 8.... 8.... 8.... 83.80..
:00 :30» :00 000.00. 003 0000 :03 :03 0:095 :03 :03 050.5 0033000
-000: 0000\33 I005 @000 :0 .4001 -0000 100009. 30000 g
-9808”. . m 08...
. .....LESSHTI. £83225 IllnHSnooIIII.
x03 800E000
05.8
0.80 3 0888: no 3308 3 088.0 .838 3 .80. 8 8380.0
.30. 0H8...

8 3380 8.. 83.83880 w 080 883 .8.... H32. H35... 82.53 83ng

-259-

0-.I man... 0nd:- 0.:..- 0nd... 0:.I 0H3I 0H3.- 0HH.I 0nd! and... 0.3.! 00.000 0030..

.5.- 2.H.- :0.- EH.- S.H.- 2.H.- .5.- EH- .5.- E.H.- £0.- cud-8.5.088

.38528
a). {a 8T- 8.- 8..- 8.- 8.l 8..- 8.- 8.....8T- 8T- «83.:
ex: 8.! ...x: 8.!8.!8.! 8T- 8.-- ST- 8.l 8..- 8.l c305
ex: 8.! 8T- 8.- 8.! 8...- 8.- 8..- 8.- 8..- 8.- 8.3 0033:3836
0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0x: 0\: 0\: 0000
a}. 8.-- .o.-.. 8.!-8.1.8...- .o..... 3.1. 8.- 8.... 3...- 8.-I .833
a). .H... 3.! 8.-.H.-.H.- 3..- ....T- 3.! ST,- GT- 2.! a-m.v.~
0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0\: 0030.800

 

:00 :30... :00 00000. 003 0000 :03 :03 059.2. :03 :03 050...: 0033000

 

 

 

I30: 0000\0>3 00:0 0000 :0 ..300 ..0000 ..00000: I0000 -00000<
0000000: 0 0000
Ill-00000005.: :030H0::H Manon-Ill
003 0000005.“.
. . 9.3.00

30030000000000 .03000950003313803583
I885 00 00330 0:0 83030800 0000 00000.. 0000 ~38. «00:5. :003000 0830000000 .30 030a.

0\: 0\: 0\: 0\: 0\: . 0\:. 0\: .0\: 0\: 0\: 0\:

 

 

8.28 3.
~. 0 man. men. 0. 0.8. «.0... 2b. m2... 0... w. no.5. 02.3
N. 0 man. man. «2.. n8. . :0. :N. o «... man. a?! 0&0:
«500.0
x03. 83 00000me 0.0.? 00.50000 one 0002002 0.02. 00000000 00 one 0000096 0303000
. 080 . 0023380 a: 0:38
. 050.2 .
03800 080988 00:03: 3 0:302. 3.6: 0 0830308 00838: :0 008.68
3 000. .
g
33830000000200.0000. 00:58:03 .
0.08.: «o 00:30 0:0 8:038:00 .800 383 08: :08. 03:5 5033 0833.008 .08 038.

-261-

 

 

 

 

 

 

 

! m. 0N0. N. ~. ~. 0 0 $0. 0 0:. >3. 0
:00 :06... :00 0000.0. 030 0000 :30 :30 0:082 :30 :30 0:092
I000: 0000\0030 00:0 0000 3 I300: 50000 100000< .0000 .0003:
luv—0,500.00 a 000..—
!..!>00030!J! 500323.!!! ~33!
:3: 00000000
0.0.00
00-00 >0 0000000: 00 3.3:
00:... co 830.0005 0.0 5:30 0:0 82.5. 000. 80303.. 0.00 5.508. 0.833.008 .08 .38.
! n8. . man. .8. MR... 83... o
c0032. c1330.. 5088 0380.. 003 5 08.. 08308.
. \90000000000 0000 0000000
03...
0.5.00
£5.00 00 0000000: 00 3.0:.
00528 8 830.085 no 00200 0.... 82:0. .00. 8.033.. 0.0.... :83... 08.03.88 ...8 .30...

-262-

II nun. . H n3. m3. 2:. II 0:. mmv. 000. m. H.

 

 

 

 

 

 

 

0
:00 3.60 :00 00009 030 0000 :30 :30 0:082 :30 :30 0:092
I000: 0000\030 I005 0000 3 I300: I0000 000002 I0000 I00000<
00000000 0 0000 .
1:00.337! Illco.0c.c..5 3.000ng
0.30. 00000000 .
0.5.00
0.0.00 an 0000000: 00 x30 08.003
L00... 8 8.02.805 00 000.00 0.... 050...... 0.0. 800.08.. a 0...... 51.08. 083-0808 .8... 00......
II 30f ~00? II 00m.I com? 80? II 000. 00m.I man. 0:. 0
50 5.60 .8.... 08...... 03.. on... 83 8.0 058... co... 8.0 0595.
I000: 0000\030 I005 0000 3 I300». I0000 000002 I030 000002
0.0000000 0 0000
. L003?! lul500002§ .EBIIIIII
0.0:. 000%
0.5.00
30030000000200 50000000000000 0000000
00 0.0:. :0 030000005 00 53000 0:0 030:0: 000. 0330000 0.200 0003000 0030300000 .000 030a

-263-

 

 

 

 

 

 

 

.l o. . a I. I. In .... II a 58. 58. n3. 0
:00 5.5a :05 gowns uouo ago :03 :03 uSQZ :03 :03 ”.5651.
I32: leuom\o>3 100:0 no.5 5 Lama Imus: nah—03¢ louse lagoon:
b.6833. a 600m
€30un Illttcogouozcn ngoculllll
x31 . 3:00me
9.56
2:8 3 @838: no .333 c5 3:832 9. 3398 .888
3 33x :0 839.585 uo 5330 can 0:358 go:— 353:?" 02mm goo ucozanouuou .NE 038.
I. «N... m. N..‘ N. N. N. N. . 8m. : N... .. N. N... . o
50 53» com gags ...03 no.6 :03 no: ”:38! :03 :03 ”.592
Ivan: IOuom\u>3 .605 @090 an luau". louse 19.3 18.5 Lab—03¢
lasagna: a coca
lib—53o!!! [Juncofiaaflfili nus—Noel!!!
gnu: ousoogfi
...-.30
2.30 3 09503: no :3:— 326 0.39.5
.90 0.38.

122 8 83.585 no 3230 as. 9.3.5.. .3: «3038.. nib 5158. 203358

-264-

I So... 8a.:

 

h. «3:339: gqu0369001\>u«3x30u0h 5 «030805
.5333.

 

 

EMU

9.56 an 08:03: as $303.09 030
1309.— :0 538535 no 3330 can @5ch 33¢ 030300.— 953 c.0300 guuunouuoo

.v: 0.30%

.8Eocoum new 5.38:9? 058... no was. 0:0 :93 0.3.: 585 053.60.» 003.3 0 ~33qu c

 

 

 

 

 

«02.! «:1:
I. Om~. $0.: 23. n3. mnmf new... at“... N... N-.I .23. new...
cow :13» can. Joana. use @030 :03 :03 ~59! :03 :03 ”.592
I32: nouom\o>3 :88 Q96 5 :33. I33 :90; :82. .9034

6.609sz a coon
Inuit—330'! £83325 «IE—RT!
33¢ 830qu
95.8 .
...-.8 .5 3.38: on 31:. ~88; can 99:: _
.23 0.33.

950.5 :0 539-985 uo 330:0 can 053:3. 33: 030309. n23 ......)qu guuanouuoo

-265-

 

:TI .3!. 3T! 3T... . El ST: .3!. A3.... .21.. x253
Ba. Be. 2e. «3. H H H «8. Ba. .68
...S. «2.. 8m. 2h. H «S. EH. H on... 3858.3
H H o H ST: ST: H .21. 3T: auméfi
El. 5... H .2... 5... H: 3T. 5.... .3!. «82.5
So. So. o8. H 3.... . E... E... El o 8288...
H- H.. H H.. .8.... o 5| H H.. 5.65
. m8. m8. .8.... 8m. m8. H H H El. 33H"
Icoaouogu
380 380 . 08853 33 83 3:39: Rug 8338.. 2.3 8833..
H9350 H9530 «98¢ .6532 $6.32 18:32 302 no <8: 8333..
...an 0.8.. 5? Ho <8: «0 H80 «0 #8
5 80¢ 000309
09:20 you 5
H38. 09.20
9.3.8
£530 3 08:03: on 5030200600 030308 Iona 3000 nonuso an..— 5 0093.6 :0
.23. Had.

83385 no 5330 can 8.85055ch 3 can: 0303.35- .3 33> 5253 9.0338500

-266-

 

HHT. HHT 8T. 8T. 2T. E... H.. H: El 2T. 33%
8m. mmo. . 8m. «8. 3%. H . ope. 2%. HS. 8m. mono
«mm. ~mm. H com. mum. 3H. 92. 8H. 8H. 8N. Sofiumfio
5.... .8.... H o H H mmm.: «mm... H H suméa
8T. 8T. H H 5!. 5| H.. H.. H.. H.. 833
H H H o8. H El. 8.! 8T: H 3.... 82.885
o o H.. H H.. . 5.... mmm... mm”. H.. H 5.65
H H 8T: 8y! 08. H QR. at. om~. on". 33H»
...:onouogo
8326 8396 88 383% «>38. «32 23...: «33 93m: «33 8338a
:95 Ho loam Ho Ho 3.2 LB 2 Lawn L32 Loom L32 Luca
8H2, . 8H5, 8?. 5H: 53 Ho Ho Ho «o
:H :H o3 8E. @3529 «x «x «x <\
«9.20 8:20 86¢ 83¢ 3:30 3:30 3o; 32»
H38. 83 no.“
233
935 an 08:30: no 503030950 03338 Scum 53% ac «30> 5 @0898 :0
833.85. Ho 5230 Be auguuousco: 3 «on: 8332: no SH; :33 9633968 6:. «Hana

-267-

 

ST. ST: H H: 8T. 336
m8... 3...: 02. RN. SN. .62
NS... 8a.- 5. 3m. 8w. Sofiuga
8T. 87.. H H H . 33%
87! ST: Anvil 3T! :41. N95?!
8T: 2:! H H. . 8m. 82.885
o H H H: H: 5.65
3T: ST: ST: H 8w. 32H»
....onouoch
fleas . . 33.9: 303.: 880 .350 no 326 830.68.. «3338
H308 38:88.89. 02.0338 9.89695 :0 30095 .350 :H 09.86

 

BORE—H 02.088538 .550
«...-mu

5 3 62503: no .3009: 369609.638 9050 :0
838535 uo 5330 can 3330355: 3 no.5 «30309— HO «30> c0253 3.033398 4.3 038.

-268-

3T...

3T:

3T1

3T...

xm>uqm

 

8T. 03... m8. 8H. «Hm. .62
ST. HHT. «.3. 8T: 2%. 3858.3
8T: H H o H .Tm.r.~
8.!. El .3!. H 5!. "8:5
8.!. H H Hal «8. 838..
H.. H.. H.. o H- 565
2:! H H .8] H 333.3835
8830328: 32695 uo mum—59.00 ”.2695. no 833.95.. 803.52 83: £503qu _

3503mm.“

 

muonEH no 333333
£530

38 3 H.838: an .coHuoHHoocao 8332.5 no 3035 no 8332...an no
832535 uo 3330 6:0 30.33323: 0... non: 833nm.— HO and; 52.53 95338.30 .mvm 033.

-269-

o\: M\: x0313

 

8T: 3.... moon
.3!. ST: 38533
«\c . a} anm.v.~
.3! H . . was?
3| ST: 82385
ST: H 565
ST. .8.... 323598016
8529.8 no Boon 8529.8 «o 8.53 8638..
95.8

g 3 09.33: an 5.833.509. 60039.5 53 0053030 :0.
:03—.3535 «0 5330 was 3033335: 3 non: 030.38.— u0 «3&— 5953 9.0330th

.mvm oHnnH. .

-270-

ll xgzm

 

 

9H... .68
l 385an
I. Sea
A wound—.2
.... . 83885
H 585.
com. ouoHHsanouoHfi .
8338 8338a
you 938

5330 an 0883: no 58.3300
SHufiunHmma 833,...8 Ba fissuouuacm: 3 no.5 8332: Ho 81> 59.53 ucofiaHohoo 6mm 03¢

-271-

Nos. .

 

8v. 8v. 8m. H H $5. 8m. com. a mafia .533
8m. 3m. 3‘. 8m H H H H 8m. o 8:9. 395.
"82
H H 58. H El El El El El xgHHm
El El H 5| 5| El H.. El El Hintw
El El H El El H: El El El «83.5
Ev. Ev. v8. a2. 2.5. 9%. H8. mR. 8.4.. o 3.5. no.3
8v. 8v. v8. ma. HE. now. So. at. «3. o amen: 3%:
“SmBHSHa
mam... mmmr . H.. H: El El El El El 8H3:
m~Hm. o. HHH. mmm. H Rm. «8. Hum. 3H. _ 5.65
So. So. Hovl nmm. 5m. . «Hp. H H El .35:N
. . ICOQQHOHfi
330 330 cococcfla 2,3 2:» 3:323 games. 8332: 3.5x 833nm.“
H828 H828 83: $832 :8932 .5832 «82 no <8: 8632...
“mam ”.3“. 53 Ho <8: Ho ”.80 Ho H30
5 .83 033.;
09.98 you 5 nouoc
H38 8:20
938

:58 E c838: no .SHHcHHaoSo 8338.. 8.6 380 H838 fig 5
@0958 :0 838535 uo 530:0 can 033.: 90.5 mo ago.— u0u 59.53 983309900

23m 0.38.

~272-

 

8.30

«3. «cm. 8m. . 8n. «2.. «8. man} Rm. 3.x... 2... o .95..
. kg
«3. n3. H 3.. o8. H HHH. HHH. «mo. «2.. o 8.5.
. you?
”88
H H .H 58. H El HHH.- nnmr H El .853
El ST.- H- H Hal 7 H H El El is}
El El H H El El 7 H- .H- 7 ~83!
8n. own. H :8. «.8. H8. 3H. 3H. HHH. we... a 85.
. 3.3
Ba. Ba. H «8. SN. 25. . 8H. 8H. SH. 8.. u .95.
. yuan:
«33583
nnmi «Sr. 7 H- 7 El 8.! 8T nan: El 8188..
31 28. man. HHH. Rn. 3... 2H.- Xor o2. Rn. . 5.65
H H 8T 8T «an. H. «8. «no. o8. 8w. 33?
gas
8393 83.8.. H.8o 3383. 833 .23 2:8 8H3 8:3 «33 .3338
6.... Ho .8... Ho Ho «82 3.3:. #8.. LS HH Lu... L3: Loom
8:5 83. 8?. 5H: 53 Ho Ho Ho Ho
5 5 v3.89 @883. «x < ¢\ ¢\
.890 855 8.6.. 8.3 3330 52.8 33» 3o:
H38. 83 you .

'6 an 68:33. an 53:28.8 .3033.— louu 83969.... no 31>

..H 39.90 8 83.83.: Ho 5330 H..... 3.8:. 8.6 Ho 33.8.. 8258 3833898. .«B .33.

-273-

H H m8. mam. H2... o «9.5. 33H
H H HHH.- 8N. m8.- o 09,5. 38:

 

 

..Homo
Hal Hovl o 8.»..- Hel 33?.
Hal HSI El . El . . H- 1;}
H8! HoT- El Hal . HHT- "8&5
mme mom- mom. Hovl ohm. o coca: 333
81- mcmf mom. Hovl 2h. 0 8:5. .58:

“38583
Hel 8T mmmr mmmf H- 8328..
o8. , mme man. man. 31 5.65
ST Hal Hal H , H ouoHHucBouoHfi
BSEH . 30§H 389.: 8.08 .350 no 380 839685 8333..

1308 03080833 35338 30960.5 :0 3009: . 9930 5 09.20

3009: 03538 350
935.8

g 3 won—Hana: no Gucci-H 339.com
noHuom nap—“Ho :0 539535 uo 3330 can @333. 00.5 no “.5809 502.3 9.0333900 .nmm mung

-274-

. onll . mow-I . hmo. 0mm. mvm. U macaw 90304

 

 

Hovl v8.- HmH. 8... 8.... o 8:9. gang:
Sumo
8T- . El 8...- Hovl H 33%
Ho.- H- H- H Hm.- 91m.v.~
Hal 3| 5| H El «83.2
HSI HHT. HHm. Hovl 8m. 0 was». .833
Hovl El 8.... Hovl How. u cocoa you...
«ougga
HOV-I mmm.l com. Hovll mmm.l consicoum
HHH.- mum. HmH. ..HN. «8. 5.65
8T H H Hal H 333.3836 .
UHQHDUOUUQE nuggturo pagan-8 #052 HO .numuflxunz 385:9.- flhQnD 8w0aufl0fi
833...... 3033 .0 SHuBHBaHH.
maiau

3.6 3 H.893... no .SfloHHoocao 8.838.. no 3025 .0
833.53.. 8 839585 .0. .HuHHSo 93 38.5 8.6 no .589. 5253 9333830 .Ea 3......

-275-

3T:

:7:

man—

 

95.8

a). a). .8sz

a}. . .o.\c. aum€~N

5.... H «8...?

HHT... ST. 33583

. 3| 8T 82.28..

2v... HHH. £65

ST. 8.!. 333.3826
8529.8 no 9:3 85:98 Ho 8.58 8Hoflumm

9330 an @8332 no 5833.388 wag 5H3 09330-00
c3838 co 839585 no 5230 can cannons 86 no €088 5953 3833830 .3. «Hana

-276-

m9. u macs. .633
HHH.: o 095. 3%:
Emma
l _ .838
I . 38:».
H . any»?!
I. 3858.3
I. SHSSE
«2.. 5.65
H BuHHucBouoHfi

 

 

8:302. @3338
now 9330

, 8330 3 09.530:
an 58:38 838333. 030302. 93 03.239 90.5 no H..—couch coin-on 9.0338900 .03 038.

-277-

«2.. HS. , 8m. 9095. .26..

 

HH... 2... n2. 8m. H H
3... 2H. 8H. . an... H H 8... Ha... 8.... o 3...... ..an
. .88..
2H. n8. H mm... .H El H H H o 095. .16..
H H ...... H H El H H H o .85. 3......
. . 3.03am ,
H H H H 5!. 5!. man. 5| .3... .....m.q.~
5.... .8!. H- 5!. 5| El 5!. El 2...... 83.8..
5...... 2.... 8H. Hum. 8~. m3. Hen. Sn. 8... o 8.5. .26..
HHH. Hen. SH. . «a. 8.... 08. So... no... . ea... c 8.5. 92...:
. .3353:
......m. 8m. nmnf A mm... H nan. o. m... on... o 8.5. .63..
3e. 8.. ..8. Hm... H m. o. H mm... o 8.5. 8......
“5.65.
. H 8T. H H w H El El 2...... 33.23835
5.... 5.... H _ El .3!. H.. 5| 5| ST. "and...
380 380. ......ch .2... 2:. 3:22,... 8.8.... 8.638.. 35$ 8.3.8..
H828 H838 38.. 5:32 .5832 .5832 3.2 .0 <8: «3033..
.8. .2... 5.: .0 <8: .0 .80 , . no .80 x
S 88.. no.8...
00:20 nan 5 no.2
H38. 00:98
Slow

«88... a... H.838: 3 .SHuaHHg 8333.. 8.. 38o H380 as... :H

. .‘ 882.0 8 8325.... .o 52.8 H..... 383 .8: 80.. ...... H35... 89.5.... «8338.8 4.2 .38

-278-

 

3m. 3.... 8... 3.. 8m. HHH.. .3. v3. 2.... 3... egos...
.333
a... a... 3.... 8H. «9.. HS. on... 8... 3... can. come...
...-H8:
.8....
HHH. HHH. H H ...9. H fl... m3. H 3| no?!
.333
H H H .8... H H man... man... H 2.... on?!
no.3:
. 2836
.5| .2... nan... . H. H nan. Han. nan. H H 3.3
5.. .2... 5...... H- El _ .....l H.. H. .2... ...... 81.8...
m8. ...8. 2.... «8. Hon. H3. 3.... 8.... man. 8.. 08......
8030.-
SH. SH. 2.... 3H. ..-. 98... n3. «.3. 9.... 2H. c 1.5....
...-H5...
.3858...
RM. 8H. 1.... Han... 8m. 8.. 3.... «8. 8o. :6. 08......
9030..
man. man. H ..8. HHH. 8o. 02. «H... 6... 3w. 0 8.5.
no.8:
......5
3| 3.... 8.! 8.! H 7...! H H 8H... 8H: 33:
155835
5| .2... H H .2! .2! H: H.. H.. H.. «8:!
828.. 838.. 88 3593.2 2.3.... .33 93.... 8.2 2:2 8.2 8.3%....
[out no Icha— UO HO .82 lulu ad la...- LOUA‘ lug lhlu u‘ Inn’—
33. 8H... 8H... . 5H: 5H: . no ... Ho ...
:H ..H 303...... H.325. ¢\ .H\ <\ <\
856 8.36 8.2 8.2 .3326 .513 3.5. 2....
H38. 80.. ...... .

g

g 3 nous-co:
an 5030:0230 0303-0.— lOuu .3309qu no can.) H: .0935 cc :03-Incas no
5:16 .33 Sauafig .303...- louu cocoa.— 93 ...-0‘ u..- Hg ...-3!. 38339—80 .3... 038. .

-279-

 

23.8 .

H H ca. 9.5. o 0 8:9. 02.3
H H 08. mm». 3.8.... 0 09.3 3&3
£08
Hal 8T: 0mm... mi 8.!. 0 3:9. 033
8T. 8T. 28. H.. ST. 0 09.5. yoga:
«xgdwm
8T. 8T... H H o almiw
8T. 87.. 5.... 3.... H: 8Heacoum
mva 08. «9. mos. own. 0 095. 333
8v... 2:... van. $5. 25. 0 amend 0&8
"3858.3
H o Ev. 8v. 8.... 0 8:5. 333
8m. 8v. . So. 0 Se. 0 09.5. 0&9
“5.65
HSI HSI H8! 0 com. mmm. 33359835
ST. .8.... 5| 8...: El. «8:5
302...: 3023. Bonus 880 .850 00 380 83088.. .8HoHuu8
H308 0258000005.. 02080000 9.50395 :0 .3009: 0050 H: 00:50

g 3 60.503: no .3009: 03053038 0050 .uOu cogs-nous mo

5330 can 8.3030055 £50308 Iona common nan: 80¢ pom H352 c0253 08339.58 .93 033.

-280-

 

 

HSI 2b... «8.- 03. ms... 0 $5. .333
ST 2b... m8... «2.. m2. 0 $5. 9&8
£08
Howl: HHH.: coo.Hr Howl: HHH. 0 once: umxoa
Hovn: HHH.: omh.n Hovun H 0 omega gang:
. . “xasHHm
8T. o o H H .me.v.~
8T. El H 8T: 5!. 83285
ST. 3| «8. Hel H8. 0 no.3. 333
H8! HHTI So. 8T. m-. 0 8:3. .38:
80858.3

o 2.0.. new. a3. EN. 0 09.5. .533

man... am. am. . HHH.. «He. 0 8:9. 3%:
"5.6.5
ST. HHH.! HHH.: . 8T. HHTI SoHHucflouoHso
HOT! 29!. «.3!... H 37.: . saunas—4
883033.25. 30:69.5 uo 8859.00 uuavoumfi mo 833.qu 803:0: macaw 030308

8638a Boa: no SHHBHBoHa .
2330
g

5 08:33 no .coHuoHHoocou 8638.. Ho Boon-H Ho SHuBHbaHn :0 83.385 00
5330 new 83339.8 «3238 83 833 you: 83 .5. Hung 5253 3.33830 .9! «Hana

-281-

23....

:7...

moan

 

0\: 0\: 003.3
a). {a eumim
El 8.!. 83280
El 8T. . 3358.3
H8... man... 0 00:5. 0033
2h. 50. 0 039. 0&3
"5005
2T 3T . 3523825
23!. a 0003.5
85.2030 00 330 85:93 0o 8003 8333..

3

0.3.00 an 009.0000: 00 300303000: 00000000 5m: 000039.00 :0 0030-08—5 00
5330 0:0 0030300000 030.3000 800 00000.. 000: 80¢ 00m H0052 000300 0003300000 .30 0309

-282-

men.
~00.

com...

H

 

00000000
000 000.00

mumo
x0300
elm . v . N
0300800
000050000
0 00000 00300
0 omens 0000:
«00005
0003000000030

00000.2

 

«00000000

03.00 >0 0000000: 00 ~000000000 0030000000: 0000
$0000 000 0030300000 0303000 8000 00000.0 0000 0000 000 H0005 0003000 00000300000 .N00 0.300.

-283-

~00. 02.. ~00... 0.0.0. 000. : 0000.. 000.10.300.00

HP!
HP!

 

02.. 000... ...,000... 000. . 000. 000. 80.. 0 09.00.0008
£000
0 0 80. 000. . 0 El 0 0 0 000000
El El 0.. El El . El 0 El El 01.0.00
El 5.... 80. El El E... El El El 0000080...
02.. 00.0. 000. 000. 000. 00.0. «B. 000. 000. 0 00:00. .0033
00.. 80. 00.0. 000. 000. 00.0. 08. 000. 00.. 0 00:00. 0000:
. «000050500
«00. SN. 0 000. 0 0.00. 80. 0 000. 0 00:00 0030.0
0.00.. 000. 0. 000. 0 80. 02.. 0 000. 0 00:00 0000:
"0000......
0 0 El 80. 0 El El El El 00000000000005
El El 0 El El 0.. El El El 0000000
0008 00000 00:00:30 0:00 0>00 00:32.00 0000000. 000000000 005\ 000000000
000080 08008 00000 -2000: -0500: -2000: 00.000 00 <00: 000000000
0000 0000 50: 00 <00: 00 008 00 008
:0 00:05 0000000.
000000 0000 00004
0000.0. 000
0.0.00

300
an 0000000: 00 0030300000 00003000 0000 00000 0000000 0000 00 0000000 00 003005005
00 00000000 0:0 80000000000 000000000 .800 00003 000: 00090 00:05 0003000 00000000008 .000 0000.0.

-284-

 

000. 000. 80. ~00. 00.0. 000. $0. 30. 00.0. 0%. 0 00000
.0033
man. 000. 000. 000. 000. 000. 000. 000. 000. 0~0. 0.00000
.0000:
0000
000. 000. 0 com. 000. 0 on~.u om~.- 0 000.. 000000
000.: 000-- 0 0- 000.: 0 0- 0- .00.: 000.: 010.0.~
Svl .El .2l 8N. Atl ASI Svl El “Bl Evl 000000000
~00. ~00. 02.. 00.0. 03. ~00. ~00. ~00. 0m~. mom. 0 00:00
kg”—
00~. 00~. 000. 000. 000. 0~o. 0~0. 0~0. 000. 00~. 0.00000
.0005:
«0000000050
000. 000. 0~0. 0 000. 000. 000. ~00. oo~. 000. 0.00000
- . .0038
000. 000. 000. 0 0m~. 000. 00~. 0~m. 000. 000. 0.00000
.0008
000.005
.00.. 000.: 00... .00.: com. 00... 0 0 o 0 000000
100000005
:ul :ul 0 0 AEl AEI 0.. 0: 0: 0: 00.0005
0000000 0000000 0000 000000000 0>0000 0000000 0>0000 00000 00000: 00000 000000000
.000 uo 5.00 no 00 00.000 1000? 00.000 0.00004 $000 1000.? 10000
0:000. 0305 000.00 0003 0003 - 0003 0003 0003
:0 00 0000000 0\ 0x . 0x. 0x
00:06 00:06 00.000 000008 000008 00000 00000
0008. 0.004 .0000
0.0000
0350 an
00.03000: 00 .00300000000 000000000 08.00 000000690... uo 0:0; 5 0000000 00 000.003.0805
00 000003 0:0 80000000000 000000000 .800 00008 0000 0000.0. 0005.0 00300 2000000008 .30 0000.0.

-285-

mNHf

 

mme 8m. H o o 8:3. 3.3
So... So... «8. H o o cocoa yang:
“mono ‘
8T: 8T: 96. 8w... ST: .838
ST. 87.. .8.... 3!. H anmaé
Hal 2:! ST. ST: 8N. maggot
Sn... 8%: ma. 2%.. 8v 6 3:3. .333
2...: ~21: awn. SN... o9. o omen... 93a:
“385an
08. oov. H 8v. CON. 0 00:5. .3304
08. 8... H 8N. 8v. 0 8:5. .69.:
. "5.65
2:! ST: HSI H ems. 333:3825
8T: 8T. El 8T. 5.... «82.2
3035 BoogH 303.: 88c 355 no 380 832685 03033..
H308 3880850: 0298008 8qu :0 3009: .350 5 0936

5
>n 03:30: no 503330.30 03338 no Bonn-H UHBQBoulonoom .350 :0 53385
Ho 3330 Han 833188 8638.. .88 833 you: H38. Hung 5853 9833858 .mmm oHan.

-286-

8.! o . ...HH. 8w. 0.8. u can»... .3

 

 

.8I o 9H. v9. 8... u was... 9&8
88..
.8!. .HT. 8...: 8.... H .833
8.!. H H H: .8.... aum.v.~.
8.!. 2.! 8m... 8...: .2... 8398....
8.! .3... SN. .8.... «mm. a ones. ...3 ‘

.8l 2.... RN. 8...: ....m. o 8:9. 3......
. «3358.3

o ~Hq. «H... H..... «H... u 095. .833

o 9...... 3m. 3... 8m. 9 8...... noon... .
EUR...
.8.... El 2.... .o.l 3.! 333.3836
8.!. .3!. 2...... H .8... «.53
3330325.: 3030.... no 9.8.5300 ”.260: uo 338.98.. 309.32 anon: 0303»?—

8Honu£ 3895 .0 833.33..
g

3.8 .5 .333: on ...oHaoHHoocfi 8338.. :8. 38...: no 83:353.. 8 83385
no 3230 93 832328 8.033.. :8... 883 you: H38. H35... .3258. «83338 .8: £3...

-287-

5.... 3|

mono

 

m}. a}. .835
a? fi\c. Humi~
ST. ST. 82.283
5.... 8T. 3858.3
m8: H 0 3.5. .333
3%. H 0 095. 0&3
_ "5.65.
3!. 8T. SaHHucfiouoHfi
28.... H No.53
85:38 no 330 8529.8 no 0868 «3303..
9:8

g >0 09.503: 00 530303008. 000000.!— 5.3 00.33030 :0 cons-nous

no 5330 can 830300.30 030303 loam 0003a 000: H38. 1352 50303 0303300000 4.8 033.

-288-

o ammo
II x9>dfim

II s1m.v.m

I 82958 .

I _ 3338.3
So. 565
H 333.3825

H «Bug

 

 

5:303 «Begum.—
.uoH £35

3 an 6.09.63: on 55.333 538.333. 03033.. :0 Squguca
no 5:30 93 83389.8 863:: loan «83.. non: ~38 ~35: 8258. 2033898 £8 cine

-289-

_ five: 38o. 33:8 anon 5 09.20 ~38 .

 

com... Bnoo 35:00 am?» 5 09.65 9.84 nan
8a... 353 86398 3 3% n82
m8... 233532 5:3 Banana 88¢
mmvr 833mg 5? Bug 834
2%... 933334 no <8: _
Sc... 03338 no axon:

o ..zégqumfioflc no 38

o . fiéogofiaom no ”.80
2.5a

Slaw an 68:53: no .330 35:00 anon 5"
ago :0 saga-nous no 3:30 05 mad—cum gav— agofinom ESE :09...qu 303309.30 .mom 0.38..

-290-

8%: 83808... no 83> 5 09.20 ~38.

 

8m... 83808.. Ho 83, E 335 82 08
8.6.... 080 no 8?—
8~.... 02:8 «30380 m: 093:8... 8.02 .
m8... £38032 5:: 03809 «82
Se... 86380 5:: 0325. 3.3
$1 832032 uo {3330
>2. «3338 no {3230 x
8m... 23203? no <32;
02... . «3338 No <33»
9.30

3 3 00.5000: an 5030.603 no «30> 5
000096 :0 0039535 no 33000 0:0 @0353. #02 030308 02mm 0003qu 000330.300 .05 038.

-291-

 

00. 303.: 0380
3.0. 3009: 00300000000:
Ommf. 302.2: 3.30.5000
90... 00000 0050 no 000308.... 00 3009:
o ‘ 330 5330000 h5.50. 5.00.0030

00:00

03:00 an 00.5000: 00 .3009: 03900000300. _
0050 00 0030-085 uo 33000 000 003000 0.03 0303000 0.30 0003000 0003300000 .HB 030a.

-292-

.Hmo.l

HHH.

oom.l

OhgfluUflUna—hfl: 9H.

HUG-8.3m HO pug-50:8 OH.
08080 00 833.3: on.
09000002 on.

“98: OH.

 

 

EMU

nuouaEH uo couuanquuawo

0.50 3 00380: no .303: no 5330503
no 332085 00 3:30 05 9.3.5. 0.0:. 030330 0:30 5058. 30338000 .20 03.0.

-293-

80....

«RN .

0000a H9300 no 3000
0000a H9300 no 000000

 

. g

05.00 an 00.5000: 00 500030

...80 co 830.085 00. 5:30 05 83:2 in. 80038.. 02.0 5308 0833330 .20 038.

BIBL IOGRAPHY

BIBLIOGRAPHY

Anderson, Elizabeth L. 'Carcinogen Risk Assessment in the
Environmental Protection Agency." Association of Food

 

and Drug Officials Quarterly Bulletin 44, No. 1 (January
1980): pp. 65-66.

 

Anderson, Lee G. and Settle, Russell P. Benefit-Cost Analy-
sis: A Practical Guide. Lexington, MA: Lexington
Books, 1977.

 

Andrews, Richard N.L. "Values Analysis in Environmental
Policy.“ Policy Studies Journal 9, No. 3 (Winter 1980):
pp. 369-378.

 

Aspelin, Arnold L., and Ballard, Gary L. ”Economic Aspects
of Current Pesticide Regulatory Programs and Outlook for
the Future." Paper presented at the American Meeting of
the Agricultural Economics Association, August 1981.

Baldwin, John R. The Regulatory Agency_and the Public Corpo-
ration. Cambridge, MA: Ballinger Publishing Co., 1975.

Bartlett, Randall. Economic Foundations of Political Power.
New York: The Free Press, 1973.

 

Brickman, Ronald and Jasonoff, Sheila. ”Concepts of Risk and
Safety in Toxic Substances Regulation: A Comparison of
France and the U.S.” Policy Studies Journal 9, No. 3
(Winter 1980): pp. 394-403.

"Cancer Risk: An Interview with Dr. Eflizabeth Anderson.”
EPA Journal 5, No. 3 (March 1979): pp. 8-9 and 40.

 

Cohen, Dave. "Guarding Against Cancer.” EPA Journal 4, No.
3 (March 1978): pp. 12-13.

 

"Congress Expedites Pesticide Program." EPA Journal 4, No.
8 (September 1978): p. 34.

 

Council for Agricultural Science and Technology. Impacts of
Government Regulation on the Development of Chemical
Pesticides for Agriculture and Forestry. Ames, IA?—
Council for Agricultural Science and Technology, January
1981.

 

~294-

-295-

. Social and Economic Impacts of Restricting Pesti-
cide Use in Agriculture. Ames, IA: Council for Agri-
cultural Science and Technology, December 1980.

 

 

Davis, Otto A. and Kamien, Morton I. "Externalities, Infor-
mation and Alternative Collective lAction," In Public
Expenditure and Policy Analysis, pp. 82-104. Edited By
Robert H. Haveman and Julius Margolis. Chicago: Rand
McNally College Publishing Co., 1977.

 

Downing, Paul B., and Brady, Gordon L. ”Constrained Self-
Interest and the Formation of Public Policy.” Public
Choice 34 (1979): pp. 15-28.

Downs, Anthony. "A Theory of Bureaucracy." American Eco-
nomic Review 55 (May 1965): pp. 439-446.

 

 

Dreyfus, Daniel A. ”The Limitations of Policy Research in
Congressional Decision-Making.” Policy Studies Journal
4, No. 3 (Spring 1976): pp. 269-273.

 

Ebner, Lawrence S. "A Concise Guide to RPAR, Cancellation
and Suspension Proceedings Under FIFRA." Washington,
October 1981.

. 'EDF v Costle: EPA's New Case for Pesticide Reg-
ulation.” Env1ronmenta1 Regulation Analyst 2, No. 1
(December 1980): pp. 6-8.

 

. ”Rebuttable Presumption Against Registration: How
EPA Has Improved Pesticide Review.” Environmental Reg-
ulation Analyst 1, No. 4 (March 1980): pp. 2-5.

 

 

Eckert, Ross D. ”On the Incentives of Regulators: The Case
of Taxicabs.” Public Choice 14 (Spring 1973): pp.

 

 

 

83-99.

Edelman, Murray. Political Langgage: Words That Succeed and
Policies That Fail. New York: Academic Press, Inc.,
1977.

Edmunds, Stahrl W. ”Environmental Policy: Bounded Ration-
ality Applied to Unbounded Ecological Problems.” Policy
Studies Journal 9, No. 3 (Winter 1980): pp. 359-368.

 

Environmental Studies Board Committee on Prototype Explicit
Analyses for Pesticides. Regulating Pesticides. Wash-
ington: National Research Council, 1980.

 

"EPA's Todhunter Tells NACA What Reforms the Agency Will

Deliver and When." Pesticide and Toxic Chemical News, 28
April 1982, pp. 12-15.

 

-296-

Epp, Donald J.: Tellefsen, F.R.: Shute, G.A.; Bear, R.M. and
Wilkinson, K.P. Identification and Specification of

 

Inputs for Benefit-Cost Modeling of Pesticide Use.
Washington: U.S. Environmental Protection Agency, August
1977.

"Federal Insecticide Fungicide and Rodenticide Act." 7 U.S.
Code Annotated, Section 136.

 

Feller, Irwin and Flanary, Patricia E. "Economic and Politi-
cal Aspects of the Benefits and Costs of Food Regula—
tion.” Association of Food and Drug Officials Quarterly
Bulletin 44, No. 1 (January 1980): pp. 47-59.

"FIFRA Changes, OPP Reorganization Prime Topics of Pesticide
Regulation Course.” Pesticide and Tbxic Chemical News,

 

Fiorino, Daniel J. and Metlay, Daniel S. ”Theories of Agency
Failure, or, Why Regulatory Agencies Continue to Be
Unreliable When the Solutions Seem So Obvious.” Paper
presented at Annual Meeting of the American Political
Science Association, 1977.

Fraenkel, R. and Hadwiger, D. 'The Agricultural Policy
Process.” Policy Studies Journal 4, No. 1 (Autumn

 

Freeman, A. Myrick, III. The Benefits of Environmental
Improvement: Theory and Practice. Baltimore, MD: Johns
Hopkins University Press, 1979.

 

Freeman, Linton C. Elementary Applied Statistics: For
Students in Behavioral Sciences. New York: John Wiley
and Sons, Inc., 1965

 

 

Gangstad, Edward O. “Benefit/Cost Analysis of Silvex Can-
cellation.” Journal of Aquatic Plant Management 20
(1982): pp. 45-49.

 

Green, M.B. Pesticides: Boon or Bain? Boulder, CO:
Westview Press, 1976.

 

Gusman, Sam; von Moltke, K.; Irwin, F. and Whitehead, C.
Public Policy for Chemicals: National and International
Issues. Washington: The ConservatiOn Foundation, 1980.

 

Hapgood, Fred. ”Risk Benefit Analysis: Putting a Price on
Life.” The Atlantic Monthly, January 1979, pp. 33-38.

 

-297-

Haveman, Robert H. and Weisbrod, Burton A. "Defining Bene-
fits of Public Programs: Some Guidance for Policy
Analysts." In Public Expenditure and Policy Analysis,
pp. 135-160. Edited by Robert H. Haveman and Julius
Margolis. Chicago, IL: Rand McNally College Publishing
Co. 1977.

 

Hildreth, R. James. "Economists, Regulation and Public
Policy.” American Journal of Agricultural Economics 61,
Number 4, Part 2 (November 1979): pp. 755-759}

 

Hirschman, Albert 0. Exit, Voice, and Loyalty. Cambridge,
MA: Harvard University Press, 1970.

 

Honeycutt, R.C., and Ballantine, L.G. ”Application of Mathe-
matical Modeling Results for Environmental Risk Assess-
ments.“ Paper presented an: American Chemical Society
Symposium, September 15, 1982.

Johnson, Edwin L. ”Risk Assessment in an Administrative
Agency.” The American Statistician 36, No. 3, Part 2
(August 1982): pp. 232-239.

 

Joskow, Paul L. "Inflation and Environmental Concern:
Structural Change in the Process of Public Utility Price
Regulation.“ Journal of Law and Economics 17, No. 2
(October 1974): pp. 291-327.

 

Kelman, Steven. ”Regulation by the Numbers: A Report on the
Consumer Product Safety Commission.“ The Public Inter-

 

est 36 (Summer 1974): pp. 83-102.

Kohlmeier, Louis M. “The Regulators and the Regulated." In
The Political Economy of Federal Policl, pp. 256-260.
Edited by Robert H. Haveman and Robert D. Hamrin. New
York: Harper and Row Publishers, 1973.

 

Lave, Lester B. The Strategy of Social Regulation. Washing-
ton: The Brookings Institution, 1981.

 

Lowrance, William W. Of Acceptable Risk: Science and the
Determination of Safety. Los Altos, CA: William Kauf-
mann, Inc., 1976.

 

 

Majone, Giandomenico. ”Process and Outcome in Regulatory
Decision-Making.” American Behavioral Scientist 22, No.
5 (May/June 1979): pp. 561-583.

 

March, James G. and Simon, Herbert A. Organizations. New
York: John Wiley and Sons, Inc., 1958.

 

Maynard-Moody, Steven and McClintock, Charles C. ”Square
Pegs in Round Holes: Program Evaluation and Organi-
zational Uncertainty.” Policy Studies Journal 9, No. 5
(Spring 1981): pp. 644-666.

 

-298-

McCraw, Thomas K. ”Regulation in America: A Review
Article." Business History Review 49, No. 2 (Summer
1975): pp. 159-183.

 

McFadden, Daniel. ”The Revealed Preferences of a Government
Bureaucracy: Empirical Evidence.” Bell Journal of

 

Economics 7, No. 1 (Spring 1976): pp. 55-72.

 

McKenzie, Richard B. and Macaulay, Hugh H. ”A Bureaucratic
Theory of Regulation.“ Public Choice 35 (1980): pp.
297-313.

 

Mishan, E.J. Cost-Benefit Analysis: New and Expanded Edi-
tion. New York: Praeger Publishers, 1976.

 

Mitnick, Barry M. The Political Economy of Regulation:
Creating, Designing and RemovingRegulatory Forms. New
York: Columbia University Press, 1980.

 

 

Moffitt, L. Joe and Farnsworth, R.L. 'Bioeconomic Analysis
of Pesticide Use.” Agricultural Economics Research 33,
No. 4 (October 1980): pp. 12-18.

 

Mueller, Dennis C. Public Choice. New York: Cambridge Uni-
versity Press, 1979.

 

National Research Council. Pesticide Decisiog;Making, Wash-
ington: National Academy of Sciences, 1977.

 

National Research Council. Decision-MakinLin the Environ-
mental Protection Agency, vols. 11a and 11b. Washington:
National Academy of Sciences, 1977.

 

 

Nelkin, Dorothy and Pollak, Michael. ”Problems and Proce-
dures in the Regulation of Technological Risk.“ In
Making Bureaucracies Work, pp. 259-278. Edited by Carol
Ii. Weiss and Allen H. Barton. Beverly Hills, CA: Sage
Publications, 1979.

 

Niskanen, William A. 'Bureaucrats and Politicians.” Journal
of Law and Economics 18 (December 1975): pp. 617-643.

 

. "Competition Among Government Bureaus." American
Behavioral Scientist 22, No. 5 (May/June 19795: pp.

 

Noll, Roger G. ReformingRegulation. Washington: The Brook-
ings Institution, 1971.

 

. 'The Behavior of Regulatory Agencies.” Review of
Social Economy 29 (March 1971): pp. 15-19.

 

 

-299-

Owen, Bruce M. and Braeutigam, Ronald. The Regulation Game:
Strategic Use of the Administrative Process. Cambridge,
MA: Ballinger Publishing Co., 1978.

 

 

Park, Rolla Edward, ed. The Role of Analysis in Regulatory
Decisionmaking: The Case of Cable Television. Lexing-
ton, MA: Lexington Books, 1973.

 

 

Peltzman, Sam. ”Toward a More General Theory of Regulation.”
Journal of Law and Economics 19, No. 3 (October 1976):
pp. 211-248.

 

Porter, Michael E. and Sagansky, Jeffrey F. ”Information,
Politics and Economic Analysis: The Regulatory Decision
Process in Air Freight Cases." Public Policy 24, No. 2
(Spring 1976): pp. 263-307.

 

Portney, Paul R., ed. Current Issues in U.S. Environmental
Policy. Baltimore, MD: Johns Hopkins University Press,
1978.

 

Posner, Richard A. “Taxation by Regulation.“ Bell Journalof
Economics and Managgment Science 2, No. 1 (Spring 1971):
pp. 22"50.

 

 

. “Theories of Economic Regulation.‘ Bell Journal
of Economics and Management Science 5, No. 2 (Autumn,
1974): pp. 335-358.

 

 

Reynolds, Larry. "Foundations of an Institutional Theory of
Regulation.“ Journal of Economic Issues 15, No. 3
(September 1981): pp. 641-656.

 

”Risks and Challenges: An Interview with Dr. John A. Tod-
hunter.“ EPA Journal 7, No. 10 (November/December 1981):
pp. 10-12.

 

Roberts, Marc J. “An Evolutionary and Institutional View of
the Behavior of Public and Private Companies.” American
Economic Review 65 (May 1975): pp. 415-427.

 

Rodricks, Joseph V. I'Risk/Benefit Assessment: Or Some Notes
on Cost-Benefit Analysis in Regulation of Food Chemi-
cals.” Association of Food and Drug_Officials Quarterly
Bulletin 43, No. 1 (January 1979): pp. 3-12. ii

Roos, Leslie L., Jr. and Liroff, Richard A. “Political
Dynamics and Environmental Policy.” Policy Studies
Journal 1, No. 4 (Summer 1973): pp. 226-233. “

 

”RPAR Process Reformed Almost Beyond Recognition.“ Pesticide
and Toxic Chemical News, 28 April 1982, pp. 9-11. ’fifi

 

 

-300-

Russell, Milton and Shelton, Robert B. 'A Model of Regula-
tory Agency Behavior.“ Public Choice 20 (Winter 1974):
pp. 47-62.

 

Sabatier, Paul. 'The Acquisition and Utilization of Techni-
cal Information by Administrative Agencies.‘ Adminis-
trative Science Quarterlyy 23 (September 19785: pp.
396-417.

 

 

Samuels, Warren J. “Legal-Economic Policy.“ Policy Studies
Journal 2, No. 1 (Autumn, 1973): pp. 12-15.

 

. ”Public Utilities and the Theory of Power." In
Perspectives in Public Regulation: Essays on Political
Economy, pp. 1-337 Edited by Milton Russell. Carbon-
dale, IL: Southern Illinois University Press, 1973.

 

, Schmid, A. Allan and Shaffer, James D., ”Regula-
tion and Regulatory Reform: Some Fundamental Concep-
tions." In Law and Economics: An Institutional Per-
spective, pp. 248-266. Edited by Warren J. Samuels and
A. Allan Schmid. Boston, MA: Martinus Nijhoff Publish-
ing, 1981.

 

 

and Shaffer, James D. ”Deregulation: The Principal
Inconclusive Arguments.” East Lansing, MI, 1982.

Schmid, A. Allan.'Analytica1 Institutional Economics: Chal-
lenging Problems in the Economics of Resources for a New
Environment.” American Journal of Agricultural Eco-
nomics 54, No. 5 (December 1972): pp. 893-901.

 

. I'Political Economy of Public Investment.” East
Lansing, MI, January 1980.

. Property, Power and Public Choice: An Inquiry
into Law and Economics. New York: Praeger Publishers,
1978.

 

Schuck, Peter H. "Regulation: Asking the Right Questions."
National Journal, 28 April 1979, pp. 711-717.

 

Schwerin, Don S. and Coyer, Brian W. ”Institutions and
Knowledge as Determinants of Regulatory Performance:
Toxic Chemicals and Electricity." Paper presented at
the Symposium on Regulatory Policy, Houston, TX, Novem-
ber 1979.

Shaffer, James D. "Observations on the Political-Economics
of Regulations.” American Journal of Agricultural Eco-
nomics 61, No. 4, Part 2 (November 1979?: pp. 721-731.i‘—

-301-

Shapiro, David L. and Shelton, Robert B. I‘The Application of
an Agency Decision-Making Model." Public Choice 32
(Winter 1977): pp. 51-65.

 

Starling, Grover. The Politics and Economics of Public
Policy: An Introductory Analysis with Cases. Homewood,
IL: The Dorsey Press,‘1979.

 

 

Steiner, Peter O. "The Public Sector and the Public Inter-
est.” In Public Expenditure and Policy Analysis, pp.
27-66. Edited by Robert H. Haveman and Julius Margolis.
Chicago, IL: Rand McNally College Publishing Company,
1977.

 

Stigler, George J. 'The Theory of Economic Regulation.”
Bell Journal of Economics and Management Science 2, No.
1 (Spring 1971): pp. 3-21.

 

Thurber, James A. 'Legislative-Administrative Relations.”
Policy Studies Journal 5, No. 1 (Autumn 1976): pp.
56-65.

 

U.S. Congress Office of Technology Assessment. Environ-
mental Contaminants in Food. Washington: U.S. Govern-
ment Printing Office, December 1979.

U.S. Department of Agriculture/U.S. Environmental Protection
Agency/State RPAR Assessment Team on 2,4,5-T. The Bio-
logic and Economic Assessment of 2,4,54T. Washington:
U.S. Department of Agriculture, February 1979.

 

U.S. Department of Agriculture/U.S. Environmental Protection
Agency/State RPAR Assessment Team on Amitraz. The Bio-
logic and Economic Assessment of Amitraz. washington:
U.S. Department of Agriculture, August 1978.

 

U.S. Department of Agriculture/U.S. Environmental Protection
Agency/State RPAR Assessment Team on DBCP. The Biologic
and Economic Assessment of DBCP. Washington: U.S.
Department of Agriculture, November 1977 and July 1978.

 

U.S. Department of Agriculture/U.S. Environmental Protection
Agency/State RPAR Assessment Team on Dimethoate. The
Biolggic and Economic Assessment of Dimethoate. Wash-
_ihgton: U.S. Department of Agriculture, June 1979.

U.S. Department of Agriculture/U.S. Environmental Protection
Agency/State RPAR Assessment Team on Endrin. The Bio-
logic and Economic Assessment of Endrin. Washington?
U.S. Department of Agriculture, November 1976 and Feb-
ruary 1977.

 

-302-

U.S. Department of Agriculture/U.S. Environmental Protection
Agency/State RPAR Assessment Team on Styrchnine,
Styrchnine Sulfate, 1080/1081. The Biologic and Boo-

 

nomic Assessment of Strychnine, Strychnine Sulfate,

 

108071081. Washington: U.S. Department of Agriculture,
March 1977.

 

U.S. Environmental Protection Agency. ”Draft Rules Governing
Rebuttable Presumption Against Registration (RPAR) Pro-
ceedings: Criteria for the Initiation of RPAR Proceed-
ings.” Washington, 1982.

. I'Health Risk and Economic Impact Assessments of
Suspected Carcinogens: Interim Procedures and Guide-
lines.“ Federal Register 41, No. 102 (May 25, 1976):
pp. 21402321405.

 

. I'Mutagenicity Risk Assessments: Proposed Guide-
lines.” Federal Register 45, No. 221 (November 13,

1980): pp. 74984-74988.

. ”Pesticide Registration Guidelines (40 CFR Part
163, Subparts A-P; New).' Federal Register 47, No. 8
(January 13, 1982): pp. 1826-1829.

 

. “Regulations for the Enforcement of the Federal
Insecticide, Fungicide and Rodenticide Act.” Code of

 

Federal Regulations 40, Chapter 1, Parts 162-164, July
1983.

 

U.S. Environmental Protection ,Agency’ Office of Pesticide
Programs. ”2,4,5-T: Position Document 2/3.” Washing-
ton: U.S. Environmental Protection Agency, 15 March
1979.

. "Amitraz (BAAM): Position Document 1.” Washing-
ton: U.S. Environmental Protection Agency, 4 March 1977.

. "Amitraz (BAAM): Position Document 3.” Washing-
ton: U.S. Environmental Protection. Agencyy 6 January
1979.

. "Amitraz (BAAM): Position Document 4.“ Washing-
ton: U.S. Environmental Protection Agency, 4 June 1979.

. "Chlorobenzilate: Position Document 3.“ Washing-
ton: U.S. Environmental Protection Agency, 11 July 1978.

. ”Chlorobenzilate: Position Document 4.“ Washing-

ton: U.S. Environmental Protection Agency, 13 February
1979.

-303-

. “Compound 1080 and 1081: Position Document 1."
Washington: U.S. Environmental Protection Agency, 27
October 1976.

. “Decision and Emergency Order Suspending Registra-
tions for Certain Uses of 2-(2,4,5-Trichlorophenoxy)
Propionic Acid (Silvex); Notice of Intent to Cancel
Certain Registrations of Pesticide Products Centaining
Silvex; Suspension Order for Silvex.“ Washington: U.S.
Environmental Protection Agency, 28 February 1979.

. 'Dibromochloropropane (DBCP): Final Position
Document.” washington: U.S. Environmental Protection
Agency, 6 September 1978.

. 'Dibromochloropropane: Suspension Order and Notice
of Intent to Cancel.” Federal Re ister 44, No. 219
(November 9, 1979): pp. 65135-65179.

. ”Dimethoate Position Document 2/3.‘ Washington:
U.S. Environmental Protection Agency, 19 November, 1979.

. “Endrin: Notice of Intent to Cancel Registrations
and Denial of Applications for Registration of Pesticide
Products Containing Endrin, and Statement of Reasons.”
Federal Register 44, No. 144 (July 25, 1979): pp. 43632-
43657.

 

. 1"Endrin: Position, Document 2/3.' Washington:
U.S. Environmental Protection Agency, 18 October 1978.

. ”Final Determination Concerning the Rebuttable
Presumption Against Registration for Certain Uses of
2,4,5-T; Notice of Intent to Hold a Hearing to Determine
Whether or Not Certain Uses of 2,4,5-T Should Be Can-
celed; Publication of Final Position Document Concerning
A11 Non-Suspended Uses of 2,4,5-T.‘ Federal Register

 

44, No. 241 (December 13, 1979): pp. 72316-72341.

. I'Fluoroacetamide (Compound 1081): Position Docu-
ment 2.” Washington: U.S. Environmental Protection
Agency, 28 February 1980.

. "Notice of Determination Not to Initiate a Rebut-
table Presumption Against Registration of Pesticide
Products Containing Carbaryl.‘ Federal Register 45, No.
241 (December 12, 1980): pp. 81869-81876.

 

. “Notice of Determination Not to Initiate a Rebut-
table Presumption Against Registration of Pesticide

Products Containing Triallate.” Federal Register 45,
No. 242 (December 15, 1980): pp. 82 49-82352.

-304-

. “Notice of Intent to Cancel Registrations and Deny

Applications for Registration of Pesticide Products

Containing Pronamide." Federal Register 44, No. 209
(October 26, 1979): pp. 61640-61651.

 

. "Notice of Presumption Against Registration and

Continued Registration of Pesticide Products Containing

Chlorobenzilate.” Federal Register 41, No. 103 (May 26,
1976): pp. 21517-21519.

 

. "Notice of Rebuttable Presumption Against Regis-
tration and Continued Registration of Pesticide Products
Containing 2,4,5-T.' Federal Register 43, No. 78 (April
21, 1978): pp. 17116-17157.

 

. "Notice of Rebuttable Presumption Against Regis-
tration and Continued Registration of Pesticide Products
Containing Amitraz.” Federal Register 42, No. 66 (April
4, 1977): pp. 18298-18303.

 

. "Notice of Rebuttable Presumption Against Regis-
tration and Continued Registration of Pesticide Products
Containing Dibromochloropropane (DBCP)." Federal Re -
ister 42, No. 184 (September 22, 1977): pp. 48026-48045.

 

. ”Notice of Rebuttable Presumption Against Regis-
tration and Continued Registration of Pesticide Products
Containing Dimethoate." Federal Register 42, No. 176
(September 12, 1977): pp. 45806-45816.

 

. ”Notice of Rebuttable Presumption Against Regis-
tration and Continued Registration of Pesticide Products
Containing Endrin.” Federal Register 41, No. 145 (July
27, 1976): pp. 31316-31337.

 

. ”Notice of Rebuttable Presumption Against Regis-
tration and Continued Registration of Pesticide Products
Containing Pronamide.” Federal Register 42, No. 98 (May
20, 1977) pp. 25906-25911.

 

. "Notice of Termination of Rebuttable Presumption
Against Registration, of ZPesticide iProducts Containing
Fluoroacetamide (Compound 1081).” Federal Register 45,
No. 41 (February 28, 1980): pp. 13189-13191.

 

. "Pesticide Products Containing Dibromochloropro-

pane (DBCP) and Statement of Reasons: Intent to Cancel

the Registrations or Change the Classifications."
Federal Register 42, No. 212 (November 3, 1977): pp.

 

57545-57548.

 

 

 

 

-305-

. “Pesticide Products Containing Dimethoate: Final
Notice of Determination Concluding Rebuttable Presump-
tion Against Registration and Continued Registration.”
Federal Register 46, No. 12 (January 19, 1981): pp.

 

5334-5369.

. ”Position Document #1 on Endrin." Washington:

U.S. Environmental Protection Agency, 12 July 1976.

. "Pronamide: Position Document 2/3.” Washington:
U.S. Environmental Protection Agency, 15 January 1979.

. ”Rotenone: Notice of Completion of Pre-RPAR

Review." Federal Register 46, No. 135 (July 15, 1981):

p. 36745.

. September 1983 Status Report on Rebuttable Pre-

sumption Against Registration (RPAR) or Special Review,

Registration Standards and the Data Call-In Program.
Washington: U.S. Environmental Protection Agency, Sep-
tember 1983.

. ”Silvex: Position Document 1/2/3.” washington:

U.S. Environmental Protection Agency, 9 July 1979.

U.S.

U.S.

van

Environmental Protection Agency' Office of Pesticide
Programs and Office of Water and Hazardous Materials. ‘5
Benefit-Cost System for Chemical Pesticides. Washing-
ton: U.S. Environmental Protection Agency, January 1976.

House of Representatives Committee on Governmental
Operations. Hearings on EPA's Implementation of the
Pesticides Control Act. Washington: U.S. Government
Printing Office, 1976.

Ravenswaay, Eileen 0. "Political Economics of Regula-
tion.” Paper presented to Michigan State University
Department of Entomology, East Lansing, MI, 30 October
1980.

. “The Science-Policy Interface." PCB's: Human and
Environmental Hazards. Edited by Frank D'Itri and
Michael Kamrin. Ann Arbor, MI: Ann Arbor Science Pub—
lishers, Inc., 1983.

 

and Hull, Beth A. ”Regulatory Decision-Making on
Toxic Substances in Food." Paper presented at the
Public Choice Society Meetings, New Orleans, LA, March
1981.

 

 

-306-

and Wallace, L. Tim. ”Changing Agricultural Market-
ing Programs.” In Federal Marketing Programs in Agri-

 

culture: Issues and Options, pp. 305-326. Edited by
Walter J. Armbruster, Dennis Henderson and Ronald D.
Knutson. Dansville, IL: The Interstate Printers and
Publishers, 1983.

 

Viscusi, W. Kip, and Zeckhauser, Richard J. “Optimal Stan-
dards with Incomplete Enforcement.” Public Policy 27,
No. 4 (Fall 1979): pp. 437-456.

 

Wade, Larry L. "Economics Applied to Policy Studies.”
Policy Studies Journal 2, No. 1 (Autumn 1973): pp.
15-180

 

Weaver, Paul IL. "Regulation, Social Policy and Class Con-
flixH:.' The Public Interest 50 (Winter 1978): pp.
45-63.

 

Wessel, Milton R. Science and Conscience. New York: Colum-
bia University Press, 1980.

 

Wilson, James O. The Politics of Regulation. New York:
Basic Books, Inc., 1980, pp. 357-394.

 

Wright, Charles L. 'A Note on the Decision Rules of Public
Regulatory Agencies." Public Choice 31 (Fall 1977): pp.