UES

  

 

E

 

.V
CB.
_. m

A

 
 

.EEN

THE RELATIONSHIP

 

D R!

 

 

 

(R-

HT-RW

 

AB

I‘W.

TH

1m

SK

TFLOLO

AN Dr

 

 

  

 

   

e'dfi PhD.

 
 

 

hT-Ehm

 

are

uje

hhhhhh s:

Effo'rithe

$18]

The.

W
R

 

 

 

ERTi F0

“ROB

 

xraoxm

y...-.,

..,.4..

 

 

 

 

inesms

0-169

 

This is to certify that the

thesis entitled

The Relationship Between Land Values
and Flood Risk in the Wabash
River Basin

presented In]

Robert Fox Boxley, Jr.

has been accepted towards fulfillment

of the requirements for

Doctor of Philosoghy (hxpeein Agricultural Economics

WWW

Major professor

DateMW 3 L. /76?

WW3 °

uwwg $9 0.10 g?“
/ .

 

 

 

 

 

 

 

 

 

 

u
u

m.

 

 

ABSTRACT

THE RELATIONSHIP BETWEEN LAND VALUES AND

FLOOD RISK IN THE WABASH RIVER BASIN
By

Robert Fox Boxley, Jr.

This Study investigates the feasibility of estimating
Expected agricultural flood damages from land price data.

The use of relative price relationships for flood—free and

floOdplain lands has frequently been suggested as an alter—

native to current Federal Agency procedures. The rationale

for such an estimation procedure is based on land rent

theOPy Which implies that floodplain lands should command

lower prices than otherwise comparable flood—free lands by

the amount of the present value of the expected future
losses of earnings attributable to flooding. The primary
Objectives of this study were to determine if the expected
price differentials exist or can be measured, and to com—

pare the estimated monetary evaluation of the flood risk
Obtained from land price data With the estimates prepared
by the U98. Army Corps of Engineers (COE).
Eight counties, representing three distinct areas
With respect to flooding hazards and types of farming,

- 'ver
Were selected from the Indiana portion of the Wabash R1

 

 

Robert Fox Boxley, Jr.

Basin for empirical analysis. Farm sales data covering

both floodplain and upland farms were obtained from

mmlic sources in each county. The average discounting

fbr flood risk among the floodplain sales was estimated

by using multiple linear regression techniques to first
estimate the influence of independent variables common to
both upland and floodplain tracts from the sale price of
the upland observations only. This yielded value estimates
(regression coefficients) for the independent variables
free of any influence of flood risk which were then used

to estimate a "flood-free" sale price (per acre basis) for
each floodplain observation. The residual difference
between the actual sale price and the estimated sale price
was taken as the estimate of the flood risk discounting on
each farm. These residuals were further analyzed to obtain
estimates of average discounting for flood risk in each
area comparable to the estimates of expected flood damages
prepared by the Corps of Engineers.

In two of the three study areas, the estimates of
expected flood damages were reasonably close to the
estimates derived by the COE. In the Upper Wabash study
area, the best estimate of the average discounting for
flood risk by the land market was $56 per acre. This is
reasonably close to the Crops estimate (capitalized at
5.0 per Cent) of $UO per acre under present protection

levels. However, this agreement may be more apparent than

 

 

 

 

 

 

 

   

WV.

 

 

 

 

,.
w.

 

 

 

 

Robert Fox Boxley, Jr.

real because the land market failed to reflect any changes
attributable to a substantial reduction in flood risk that
occurred during the study period. Possibly, however, some
capitalization of the expected benefits from protection
had already occurred by the time of the study.

In the second study area along the lower reaches of
the Wabash and White Rivers the analysis revealed no sig—
nificant discounting for residual flood risks in areas
receiving levee protection. This was consistent with a
priori expectations. The estimated discount for farm
sales without levee protection was $118 per acre. At 5.0
per cent, this is equivalent to expected average annual
losses from flooding of $5.89 per acre and was not sig-
nificantly different from the COE estimate of $6.81 per
acre, after correcting for an error in COE computations.

In the third study area——Bartholomew and Jackson
Counties in the White sub—basin——statistically reliable
estimates of the flood risk could not be obtained from
land price data. In Bartholomew County, the average
(unadjusted) floodplain sale price per acre was very
close to the average sale price of the upland sales, but
there was evidence that the floodplains would command at
least a $35 per acre premium in the absence of flood risks.
In Jackson County, the floodplains commanded a $110 pre—
mium over the uplands even with the existing flood risk

and this relationship provided no means of estimating an

 

 

Robert Fox Boxley, Jr.

expected price for the floodplain sales in the absence of
flood risk. However, the existing price relationships in
both counties were widely inconsistent with the COE
estimate of expected flood damages. In attempting to
reconcile the estimates, several errors were discovered
in the Corps computations. The nature of the errors was
such that they could have been detected by the COE had
the relative land price information been available and
capitalization principles applied. Thus, the land market
analysis still provided a criterion for judging the accuracy
of COE techniques even though it could not provide an
independent estimate of the flood risk discounting.

From the results of the analysis it was concluded
that the land market apparently does discount for flood
risk in a systematic manner but that the large standard
errors associated with estimates obtained from land price
data and the uncertainties in determining an appropriate
capitalization rate make it impractical to place sole
reliance on relative land prices for estimation of expected
flood damages. However, information on relative land
price relationships and the application of capitalization
principles should add substantially to the accuracy of
conventional federal agency flood damage estimation pro—
cedures. It is therefore recommended that land price data

for areas slated to receive flood protection be collected

 

 

 

y- .-..

" ":or rr‘r
"Z--Z:.-..

‘v

 

 

v- ....._,‘,
1A ‘
h. u..g.,~

:1..-
’v—

 

 

Robert Fox Boxley, Jr.

m1a regular basis and that a land value check, based on
capitalization principles, be made a part of each project

justification study.

 

 

'2. ~-
-,... '>-‘~

_. an“
-uVV“

if .

THE RELATIONSHIP BETWEEN LAND VALUES AND

FLOOD RISK IN THE WABASH RIVER BASIN
By

Robert Fox Boxley, Jr.

 

A THESIS

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

DOCTOR OF PHILOSOPHY

Department of Agricultural Economics

1969

I .K a _ . S .n ... _ L a 2.. .L
a . 5 .J .C C . y .w.\ C Mm .. .
.7. WV: n6 - 3 a: e ‘ A a n; n» V»
C LL 5 .nu .vW F. n. _ e l. u. . «b
e r '1 _ e w n W J AR «\u e l. “U C»
W. A: A: s .n e w." in“ n: J x.
D. 5b h e . V . v h w. .n .L n .t
D. s _ v .nln VA A... v at at. 4‘ 3 bu
7n 8 LL n7. Dc .nb S .. w h. nub $1.. at s:
. . _ I. A v .hlu .7. A» A: V.“ .7. iv
QM .\v Av «V «\v a» Au» uW . h u A.»
l - I. t: _ "A: R\U W - A—V H . «D H . h u Aux.
n. C. .r. 5“ . . . . .nn I... a: ”a A.» . .
a. ..y . ._ C. C. .. :. 3“ {a . . n».

 

 

ACKNOWLEDGMENTS

Appreciation is extended to Dr. A. Allan Schmid and
Dr. Lester V. Manderscheid who shared, at different inter—
vals, the task of guiding the study and the preparation of
this thesis; and to Dr. John Brake who served as the chair—
man of the author's guidance committee. Appreciation is
also extended to Dr. Milton Steinmueller and Dr. Roger
Strohbehn who read and made many helpful comments on the
thesis. Dr. Strohbehn was also involved in the study as
director of the North Central Resource Group, Economic
Research Service, with which the author was associated
during the course of the study.

Finally an expression of gratitude is due to my wife,
Jennie, for making the time spent on this study as

pleasant as it was.

 

 

iC’NOWLEDGI'EN IE

INTRODW

Deadw-
Ian;
tcnl
Bem
Lap:

.1 A
16:88:
“- ~l ”'1 a!
i“ AlHLOw
“' do. '\
AWN/t} p

HI. PREVIOU

Verdi

TABLE OF CONTENTS

Page
ACKNOWLEDGMENTS . . . . . . . . . . . . ii
LIST OF TABLES . . . . . . . . . . . . vi
LIST OF FIGURES . . . . . . . . . . . . viii
Chapter
I. INTRODUCTION . . . . . . . . . . 1
Objectives . . . . . . . . . . A
Definitions . . . . . . . . . 8
Land Value Model . . . . 8
Conventional or Agency Procedures . . 8
Benefits versus Damages . . . . . 8
Land Value Model Applications . . 9
Research Backgroud and Area of Study . . 12
II. A THEORETICAL FRAMEWORK FOR ESTIMATING
FLOOD CONTROL BENEFITS . . . . . . 14
Agricultural Benefits and the Role of
Land Values . . . . . . 18
Land Values in Practice . . . . 32

Evaluating Flood Hazards . . . . . 32
Valuing Noncomparable Lands . . . .

 

Components of Land Value . . . . 50

Sources of Land Value Data . . . . 55

Land Values in Practice: Summary . . . 57

III. PREVIOUS APPLICATIONS . . . . . . 61
Verdigris River, Kansas . . . . . . 61

Grand River Reach, Missouri . .. . .. . 6A

Ray, Clay and Platte Counties, Missouri . 68
Washita River Basin, Oklahoma . . . . 7“
Conclusions . . . . . . . . . . 77

iii

Chapter

IV. SELECTIC
COLLECTI

FlOOdi
Select
The
The
The
Data E

Pro;
Asse
Maps
Data C
Flood;
Jppe
Lowe
Whit

J,»
>4<4 a (2 (J O
"5

—3 '3 :3: <1 '1) fr] '5} '
C) ‘-<.‘ C) m (D
'9 ’9 C 5—14 (/1 "5 ’(J '(J

I

a
o
2‘

Lam

Typ|

Flo

Prese

REgj

VI- REGRESS
Gener‘
F10
Combi

 

Chapter Page
IV. SELECTION OF STUDY AREAS AND DATA

 

COLLECTION . . . . . . . . . . . 79
Flooding Problems . . . 83
Selection of Areas for Investigation . . 90

The Lower Wabash Area . . . . . 93
The Upper Wabash Area . . . . . . 97
The White Subarea . . . . . . . 99
Data Sources . . . . . . . . . . 100
Property Records . . . . . . . . 100
Assessment Records . . . . . . . 101
Maps and Photographs . . . . . . 102
Data Collection . . . . . . 10A
Floodplain and Upland Comparisons . . . 109
Upper Wabash Area . . . . . . 110
Lower Wabash Area . . . . . . . 112
White Subarea . . . . . . . . . 112
V. REGRESSION ANALYSIS: GENERAL . . . . . 118
Land Price Models . . . . . . . . 118
Variables . . . . . . . . . . 122
Sale Price . . . . . . . . . 122
Date of Sale . . . . . . . . 123
Total Acres Sold . . . . . 123
Acres of Cropland by Grade . 123
Acres of Pasture and Acres of Woodland. 125
Acres of Other Land . . . . . . . 125
Type of Instrument . . . . . . . 125
Expansion . - - 126
Per Cent of Mineral and Oil Rights
Sold . . 126
Value of Farm Improvements . . . . 127
House Grade . . . . . . . 128
Type of Road . . . . . . . . 129
Topography . . . . . . . . 129
Township - - - . - - 130
Land Capability Unit GrOups . . . . 130
Type of Farm . . . . . . . . . 131
Flood Risk . . . - ' 132
Presentation and Interpretation of
Regression Results . . . . - 135
VI. REGRESSION ANALYSIS: UPPER WABASH AREA . 140
General ' ' ' ' $38
Floodplain Sales I952- ~1956 - - ' ' 1A2

Combined Analysis, All Sales . . .

iv

 

Chapter

Upland
Year

I .3
-< r).

() n)
: ,

Ii. COMfgilf

11“..

-'..' :11

 

Chapter page

Upland and Estimated Floodplain Sales . 146
Year of Sale Adjustments . . . 1A7
Price and Flood Risk Adjustments . . 151
VII. REGRESSION ANALYSIS: LOWER WABASH AREA . 157
Upland and Estimated Floodplain Sales . 159
VIII. REGRESSION ANALYSIS: WHITE SUBAREA . . . 165
Upland and Floodplain Values,

Bartholomew County . . 167

Upland and Floodplain Values, Jackson
County . . . . . . . . . . . 171
IX. COMPARISONS OF LAND VALUE AND FLOOD RISK . 17A
Capitalization Rates . . . . . . 175
Upper Wabash Area . . . . . . . . 177
Lower Wabash Area . . . . . . . . 183
White Subarea . . . . . . . . . 186
X. CONCLUSIONS AND RECOMMENDATIONS . . . . 19A
BIBLIOGRAPHY . . . . . . . . . . . . . 20A
APPENDICES . . . . . . . . . . . . . 210

 

 

l\)

(A)

4:4

Farms ar
Wabash I

Study AI
Wabash I

Selecte<
Upper We

Selec:e<
Lower We

Selects:
Barthcl<

Selects
fers, J;

tiOOS,

EStimat
tions,

10.

ll.

12.

13.

LIST OF TABLES

Farms and Farm Characteristics by Sample Areas,
Wabash Land Value Study, 1956 and 1965 . . .

Study Areas, Reaches, and Floodplain Acres,
Wabash Land Value Study . . . . . . .

Selected Characteristics of A06 Farm Transfers,
Upper Wabash Area, 1952—1966 . . . . . .

Selected Characteristics of 33“ Farm Transfers,
Lower Wabash Area, 1962—1966 . . . . . .

Selected Characteristics of 81 Farm Transfers,
Bartholomew County, White Subarea, 1964—1966

Selected Characteristics of 118 Farm Trans-

fers, Jackson County, White Subarea, 196A—1966
Flood Frequency Zones by Streams, Wabash Land
Value Study . . . . . . . . . . .

Estimated Regression Coefficients, 406 Observa—
tions Upper Wabash Area, 1952—1966 . . . .

Estimated Upland Regression Coefficients and
Mean Values for Upland and Floodplain Observa—
tions, Upper Wabash Area, 1952—1966 . . . .

Average Land Prices and Flood Risk Calcula—
tions, Upper Wabash Area . . . . . .

Estimated Regression Coefficients, 33A Observa—
tions, Lower Wabash Area, 1962-1966 . . . .

Estimated Upland Regression Coefficients and
Mean Values for Upland and Floodplain Observa—
tions, Lower Wabash Area, 1962—1966 . . . .

ices and Flood Risk Calcula—

Average Land Pr
tions, Lower Wabash Area . . . .

vi

Page

94

95

111

113

114

116

13“

1A3

1A8

152

158

160

162

 

 

Table

,,i.
J:

Estimate
Bartholo
Subarea,

Estimate
Mean Val
Observat

Average
tions, E

Estimate
Mean Val
tions, ;

Table

1A.

15.

16.

17.

l8.

19.

Estimated Regression Coefficients, All Sales,
Bartholomew and Jackson Counties White
SUbarea, 196u_1966 c a o a I o I I
Estimated Upland Regression Coefficients and
Mean Values for Upland and Floodplain
Observations, Bartholomew County, 1964—1966 .
Average Land Prices and Flood Risk Calcula—
tions, Bartholomew County . . . . . . .
Estimated Upland Regression Coefficients and
Mean Values for Upland and Floodplain Observa—
tions, Jackson County, l96A-1966 . . . . .
Average Rate Earned on Investment, Indiana
Farm Account Cooperators, Selected Years . .
Flood Stages, Acres Inundated, Natural Fre—
quencies, and Estimated Average Annual Crop
Damages; Reach EW—A, Jackson County, Indiana .

 

Page

166

168

170

172

178

191

 

Figure

1. Capital
Land .

a. Capital
Land .

. Actual L
Estimate

1966

 

LIST OF FIGURES

Page
Capital Investment and Rent to Floodfree
Land O l O O I I D l‘ I O I l I l 20
Capital Investment and Rent to Floodplain
Land I o o o u a c I o n O I I I 23
General Site Map, Wabash Basin . . . . . 80
Physiographic Regions, Wabash Basin . . . . 82
Major Land Resource Areas, Wabash Basin . . 84
Actual Land Price Indices, Indiana, and
Estimated Indices, Upper Wabash Area, 1955-
1966............ 150

viii

 

 

Federal i
has amounted to
ssntrol policy

expenditures ar

ream for imprOV
resu

-25 in eval

aid or flood pr

5701‘;ng body 0.1

 

CHAPTER I

INTRODUCTION

Federal investment in flood protection and prevention
has amounted to more than $7 billion since a national flood
control policy was adopted in 1936. Currently, Federal
expenditures are approximately $500 million per year and
the annual rate of investment is increasing.1 Wise
investments of this magnitude depend importantly on reli-
able benefit-cost analyses Of flood protection projects.
More than a decade ago Fred Clarenbach, in a contribution
to a Task Force Report on Water Resources and Power found
that: "The accumulated evidence clearly indicates much
room for improvement Of Federal agency procedures and
results in evaluations of water resource projects generally

"2

and of flood protection proposals in particular. The

growing body of literature and professional interest in

 

lU.S. Congress, House, Task Force on Federal Flood
Control Policy, A Unified National Program for Managing
Flood Losses, H. Doc. A65, 89th Cong-a 2d sess., 19 3 p- 3.

2Fred A. Clarenbach, "Reliability of Estimates of
Agricultural Damage from Floods," in Commission on Organi-
zation of the Executive Branch ofthe Government, Task
Force Report on Water Resources and Power (Washington,
D.C.: Government Printing Office, l954), III, p. 1277.

 

 

 

 

 

possible avenue

use of land val
benefits. The
flood protectio
other water res
orally, to any
fall gains can
or contractual
Federal
:le benefit est
flood control
'n'llfred Pinel a
Each used curre
strate that the
Aeas could not
the DPOJects .
”f errors in t]
both to Peoomm.

this field of project evaluation suggests that Clarenbach's
observations are still valid today.

The purpose of this study is to investigate one
possible avenue of improvement in agency procedures——the
use of land values as a means of estimating expected project
benefits. The applications Specifically considered are for
flood protection projects but the principles apply also to
other water resource development projects and, more gen-
erally, to any investment decision where substantial wind-
fall gains can be expected to accrue to a physical Object
or contractual right for which a market exists.

Federal agency interestixithe use of land values in
the benefit estimating process can be traced to two ex BEER
flood control project reviews made in the mid-1950's by
Wilfred Pine1 and Fred Clarenbach.2 Both Pine and Claren—
bach used current and anticipated land values to demon—
strate that the structure of the land market in the affected
areas could not support the estimated benefits claimed for
the projects. These findings, together with other evidence
Of errors in the estimating procedures originally used, led
both to recommend that a land value check be included as a

part of every project investigation.

 

 

1[Wilfred H. Pine, et a1.], "Report of Advisory
Board on Agricultural Flood Damages to Army Corps of
Engineers," April 27, 1956. (Mimeographed.)

2Clarenbach, Task Force Report, pp. 1275—1298.

 

 

 

The techn

their foundatio
stated, the the
value of land a
earnings--land

can be expected
:omparable flco

are fully infer

 

into considerat
is expected r
flood losses.
damage and 105
:ultural benef
conditions one
cultural benefi
values.

The adva1
benefit estima‘
‘38 relatively
toast to the l
Summing pro
WOVIdes a mea
lost directly

emcted earn4

7 J- W.
frisation Be?
.959), pp, 23.

The techniques used by Pine and Clarenbach have
their foundations in the theory of economic rent. Briefly
stated, the theory implies that——to the extent the market
value of land accurately reflects its expected future
earnings--land subject to periodic damage from flooding
can be expected to command a lower price than otherwise
comparable floodfree land. If land market participants
are fully informed and realistically take the flood risk
into consideration, the differential in price will reflect
the expected reduction in future earnings attributable to
flood losses. Since the prevention of expected physical
damage and loss of earnings is the major source of agri—
cultural benefits it follows that under these idealized
conditions one should be able to estimate expected agri-
cultural benefits from flood control directly from land
values.

The advantages claimed for a land value approach to
benefit estimation are several.1 For one, land value data
are relatively accessible and easily interpreted. In con—
trast to the laborious and potentially error—prone
estimating procedures currently used, the land market
provides a measure that is basedon theexperiences of those
most directly affected by the flooding hazard. In addition,

eXpected earnings as reflected in land prices have already

 

 

1J. W. Milliman, "Land Values as Measures of Primary

II‘I‘igation Benefits,” Journal of Farm Economics, A1 (May,
1959), pp. 23A—2A3.

 

 

 

 

l

:1:— \_

been discounted

term market rat
social time pre
Several o
claimed advant
of applying th
demonstrated b
exogenous chan
land (particul
allotments und
have conceptua‘
financed flood
floodplain lan
characteristic
studies, the tl
:bat land value
tent contribut:

:3 benefit-cos

The purp
Iheoretical an

lEnefit-e st ima
\

1

”JAN See, fc
Allotélené sGibE
mead

been discounted to a present value by an implicit long—
term market rate that presumably reflects the appropriate
social time preference rate for project evaluation.

Several objections may be raised to each of these
claimed advantages; nevertheless the empirical feasibility
of applying the principles of economic rent theory has been
demonstrated by a number of studies of the effects of
exogenous changes in land rent on the market value of
land (particularly changes arising from crop production
allotments under government price support programs) which
have conceptually similar effects on land values as federally
financed flood control projects.l Although the market for
floodplain land probably differs in several important
characteristics from the land markets examined in previous
studies, the theoretical and empirical evidence suggest
that land value data, properly applied, may make an impor—
tant contribution in improving present agency techniques

of benefit—cost analysis°

Objectives

The purpose of this investigation is to examine the
theoretical and practical issues involved in the use of a

benefit—estimating model based on land values. The general

 

 

1See, for example: Frank H° Maier, James L. Hedrick,

and W. L. Gibson, Jr., The Sale Value of Flue-cured Tobacco
Allotments, Technical Bulletin 148 (Blacksburg: Virginia
Agricultural Experiment Station, April, 1960).

 

 

 

 

 

 

plan of presen

develop a heme
(b) to apply t
sidered for fl
billty of adop
agency procedu

The spec

olog

lam

The thi:

land Values a,
exlooted dama
DPooess. Sin
analytical te
for field ap!

13 t0 evaluaf

 

plan of presentation followed in this report is (a) to
develop a benefit—estimating model based on land values,
(b) to apply the land price model to areas currently con—
sidered for flood protection, and (c) to judge the feasi—
bility of adopting such a model as a part of required
agency procedure.

The specific objectives of this study are:

1. To estimate the relationships between land
values and flood risks in selected areas of
the Wabash River Basin.

2. To compare flood protection benefits estimated
from land value data with benefits estimated
by conventional benefit—estimating procedures
and to evaluate the relative accuracy of the
two approaches or a combination of the two
approaches.

3. To evaluate land value data sources and method—
ological requirements for field applications of
land value estimation procedures.

The third objective is conditional upon finding that
land values are sufficiently accurate indicators of
expected damages to warrant use in the benefit estimating
process. Since the first two objectives may require
analytical techniques that may be too complex or expensive
for field applications, the purpose of the third objective

is to evaluate possible shortcuts in data collection and

 

 

 

 

 

analysis requio

the technique .
of implementin;
evaluations .

The seco

must be resolv

 

benefit-estima
nodel contribu
damage and £10
use. The ques
first is the q

adjustment to
log accuracy 1
assuming under
second aspect
model compared
eedures. This
agency viewpoi
objective is a
The diff

mates derived
benefit-estima
ObJective is t
which the accur
review of coma

necessary but

 

analysis requirements to facilitate field applications of
the technique. Also implied is an evaluation of the costs
of implementing the land value approach for specific project
evaluations.

The second objective reflects the basic issue that
must be resolved to justify federal agency adoption of a
benefit—estimating model based on land values: whether the
model contributes sufficiently to the accuracy of flood
damage and flood control benefit estimates to justify its
use. The question of accuracy has several aspects. The
first is the question of the accuracy of the land market's
adjustment to risk, or equivalently’the potential of obtain—
ing accuracy in the land value estimating technique-—
assuming underlying accuracy in the land market. The
second aspect is the relative accuracy of a land value
model compared to conventional benefit—estimating pro—
cedures. This is the relevant question from a federal
agency viewpoint and it is to this question the second
objective is addressed.

The difficulty with the direct comparison of esti—
mates derived from a land value model and from conventional
benefit-estimation procedures as proposed in the second
objective is that there are no absolute standards against
whichtflm accuracy of either approach can be judged. A
review of conventionally derived estimates will be made if

necessary but it cannot be definitive because of uncertainty

:87

 

 

 

 

pegarding the h

shodied in the
Recognizi
which either th
be measured, th
are consistent
can serve to st
data. For exam
values react in
hazards of diff
degree of faith
moor would be
to determine if
aasured. If
for example, if
agnitude is r
useful in defin
are appropriate
rould establish
degrees of risk
ihat the moneta
correctly evalL‘
raise serious c

land value app!

 

 

regarding the hydrological elements and other assumptions
embodied in the conventional technique.

Recognizing there is no ultimate standard against
which either the land market or conventional technique can
be measured, the first objective seeks to determine if there
are consistent relationships within the land market that
can serve to strengthen confidence in the use of land value
data. For example, if it can be demonstrated that land
values react in a regular, systematic fashion to flood
hazards of differing intensities, ceteris paribus, the
degree of faith one could place in land values as an esti—
mator would be increased. The first objective is intended
to determine if such relationships exist, or if they can be
measured. If relationships exist in only a limited sense,
for example, if only flooding risks greater than a certain
magnitude is recognized, this knowledge still should be
useful in defining cases where the use of land value data
are appropriate. Positive findings for this objective
would establish that land prices do discriminate among
degrees of risk although it may not necessarily establish
that the monetary value of the expected flooding damage is
correctly evaluated. Negative findings, however, should
raise serious questions about the applicability of the

land value approach.

 

 

 

 

  

At this

  
  
 
 
  
  
  
  
   
 
  

some of the mo

in this report .

Land Value Mod
To avoid
tern "land val
procedure for
:rol projects
the procedure.
lcnventional 0
Procedures
These te
land designatic
the “stage-dame
federal agencie
generally, the
attempts to re<
range 01” flood

idiot in time a

Flood dar
Plinarily in to
lllSieal damage

benefits from .

 

f————

8

Definitions
At this point it is desirable to explicitly define
some of the more frequently used terms and concepts employed

in this report.

Land Value Model

To avoid repetitious qualifications we will adopt the
term "land value model" (or "approach”) to designate any
procedure for estimating expected benefits from flood con—
trol projects that uses land values as an integral part of
the procedure.
Conventional or Agency
Procedures

These terms will be used interchangeably as a short—
hand designation of the "flood hydrograph—flood damage" or
the "stage—damage integration" methods currently used by
federal agencies to estimate flood control benefits. More
generally, the terms refer to any estimating procedure that
attempts to reconstruct in probabilistic terms the whole
range of flood risk and associated flood damages at some

point in time and space.

Benefits versus Damages

Flood damages to agricultural lands are evaluated
primarily in terms of income losses and the costs of
physical damages resulting from flooding. Conversely, the

benefits from flood control are the expected reductions in

 

 

annual flood da
nth and withou
damages and floo
detail in later
point is that t
control is comp
rarely provide
suited to esti
expected flood
qualification 3
reasonableness
also provides
its land value
limits for expe

direct estimate

 

iSuell establi
ildPOpriate; hc
Zionship betwee

rather than ex;

The steps

c"malntualiy s
3fnow the lan
118thllluish be

t
'i‘ms 01‘ a lar

‘

 

annual flood damages——the differences in expected damages
with and without the flood control structure. Both flood
damages andflood control benefits will be discussed in more
detail in later chapters; the important distinction at this
point is that the two measures are equivalent only if flood
control is complete. Since most flood control projects
rarely provide complete protection, land values are best
suited to estimating expected flood damages rather than
expected flood control benefits. This is not a serious
qualification since a land value—based judgment of the
reasonableness of conventionally estimated flood damages
also provides a basis for the judgment of expected benefits.
Thus land values should be usable in establishing upper
limits for expected benefits even if they are not usable as
direct estimators. The "benefit estimating" terminology

is well established and we will continue to use it where
appropriate; however, we are more concerned with the rela-
tionship between land values and expected flood damages

rather than expected flood control benefits per se.

Land Value Model Applications

The steps involves in applying a land value model are
conceptually simple but there are differing interpretations
0f how the land value approach should be applied. We will
distinguish between "supplemental" and "substitute" applica—

tions of a land value model.

 

 

 

expected avera

mated by conve
compared to pr
tected, and ju
can be in term
vise comparabl
of the indicat
the investment
and the capita
case is that t
with present p
indicated pric
related to pric
quality.

In this a
reliance for at
lemniQues but-

corroborating t

°f the project

 

 

 

10

The supplemental application would be an integral
part of conventional estimating procedures. After the
expected average annual damages (and benefits) are esti-
mated by conventional techniques they are capitalized,
compared to prevailing land prices in the area to be pro—
tected, and judged for reasonableness. The comparison
can be in terms of current price differentials between other-
wise comparable floodplain and floodfree land, or in terms
of the indicated price of presently unprotected land after
the investment (which will be the sum of the present price
and the capitalized benefits).1 The criteria in the first
case is that the capitalized damage estimated should agree
with present price differentials. In the second case the
indicated price after protection should be reasonably
related to prices of nearby floodfree land of comparable
quality.

In this application of the land value model primary
reliance for accuracy would be placed on conventional
techniques but the land value data would provide valuable
corroborating evidence, especially for subsequent reviewers

of the project proposal.2

 

lOr, equivalently, current or indicated price dif—
ferentials can be converted to annual equivalents and com—
pared to the conventionally estimated average annual damages
0r benefits. All of the above comparisons would be carried
out most conveniently on a per acre basis.

2The Corps of Engineers issued a well—reasoned
engineering manual in 1957 that explained the application
Of a land value check and specified that project reports

 

 

  

 

In a "su

  
  
  
  
  
  
   
   
  
  
  
  
   
 

could be relie
damages and be
advocated this
sodel based on
cheaper to app
application ma
projects but i
:o most flood
between the tw

‘ o" .
degree 04'. acca

 

”the land value

sively as the

should include
between produc
values of land
Validity of th
ccsals examine
and values in
agineers, "Ex
flutes and Fl
iogineering Ma

1
., Edward
“Economic A
some

W

2
.. A third
"“Euished, T

is“ in a limi
fnhancement n

 

 

ll

In a "substitute” application the land value model
would be relied upon exclusively in estimating expected
damages and benefits° Edward Renshaw, for one, has
advocated this approach; arguing that a benefit—estimating
model based on land prices could be more accurate and
cheaper to apply than current agency procedures.1 This
application may have promise for irrigation and drainage
projects but it would appear to have limited applicability
to most flood control projects, The primary distinction
between the two applications of the model is that a higher
degree of accuracy would be required for the latter since
the land value differentials would be relied upon exclu—

sively as the measure of expected benefits.2

 

should include a discussion of ”the extent of agreement
between productivity, damage, and benefit values with market
values of land, as a basis for indicating the essential
validity of the formera" However, none of the project pro—
posals examined for this study included any discussion of
land values in this sense, See: U.S¢ Army, Corps of
Engineers, "Examinations and Surveys: Relation of Flood
Damages and Flood Control Benefits to Market Value of Land,"
Engineering Manual EMll20—2—ll, 13 June 57, (Mimeographed.)

lEdward Fe Renshaw, Toward Responsible Government:
An Economic Appraisal of Federal Investment in Water Re—
source Programs (Chicago: Idyia Press, 1957), pp. 78-79,

 

2A third, "enhancement" application, can also be dis—
tinguishedo The enhancement application is currently being
used in a limited way by the Corps of Engineers where
"enhancement" benefits are expected to be large relative
to direct damage reduction benefits. Enhancement benefits,
generally, are the benefits realized when flood protection
makes a shift to a higher land use feasible. This is most
likely to occur in urban areas where floodplain land is
idle or in a low valued use but may also be a component of
agricultural benefits. The "enhancement" is equivalent to
the increased rent the land can command in the higher use

 

 

 

 

 

This rese

 

the Corps of E
(003), and the Ec
Jepartment of .
alternative pr
flood control
:ie study is :
eepirical anal
under the Corp
The Waba
analysis for a
siderations in
1e selection <
inaction of r
M
:53: as a prac1
:ipected markei
« a more gene]

glove but the 2
She also.

1
.. , The othv
1°“ in the Wal
Wee (SCS) .
lipe' The ur
:ie basis 0% s‘

 

12

Research Background and Area of Study

 

This research was conducted under contract between
the Corps of Engineers, United States Department of Army
(COE),and theIEconomic Research Service, United States
Department of Agriculture as a part of a larger study of
alternative procedures for the evaluation of agricultural
flood control benefits. Consequently, the orientation of
the study is toward COE techniques and procedures and the
empirical analysis is applied primarily to floodplains
under the Corps' jurisdiction,1

The Wabash River Basin was selected for empirical
analysis for administrative convenience and because of con—
siderations involved in other parts of the overall study.
The selection of specific study areas within the Basin was

a function of methodological considerations only.

 

and, as a practical matter, is estimated by estimating the
expected market value of the land with protectiono This

is a more general use of land value than the uses envisioned
above but the analysis planned will have general relevance
here also.

1The other federal agency involved in flood control
work in the Wabash River Basin is the Soil Conservation
Service (SCS) of the United States Department of Agricul—
ture. The jurisdiction ofthetwo agencies is decided on
the basis of stream drainage area with the SOS restricted to
streams or upstream segments of major watercourses with
less than 250,000 acres of drainage area; The principles
0f a land value model apply equally to the two agencies
but the typically small areas involved in 808 work may
create difficulties in applying a land value model because
0f the scarcity of small—area land value datac

 

 

 

 

 

 

 

13

The original request by the Corps of Engineers was
for a study of the "influence on agricultural land values
of improved utilization through flood protection." Hence,
the study is restricted to lands used for agricultural

purposes.

y

 

 

 

 

CHAPTER II

A THEORETICAL FRAMEWORK FOR ESTIMATING

FLOOD CONTROL BENEFITS

 

The Corps of Engineers classifies the damages from
flooding as:1
(l) Tangible flood damages

a. Physical damages, including the cost of clean—
up, damages to or loss of buildings or parts
thereof, loss of contents, including furnish—
ings, equipment, decorations, stocks of raw
materials, materials in process and completed
productso

b. Emergency costs, including those additional
expenses resulting from a flood that would
not otherwise be incurred such as evacuation
and reoccupation, flood fighting, disaster
relief, increased expense of normal operations
during a flood, increased costs of police,
fire or military patrol, and abnormal wear
and tear on alternative routes of traffic,

0. Business and financial losses, including the
various economic losses other than physical
damages and emergency costs, resulting from a
flood such as net loss of normal profit and
earnings to capital, management and labor in
the readily identifiable zone of flood influ—
ence, The estimate should exclude all losses
that may be compensated for by increased
economic activity in the area affected at a
later date (postponed sales, etc.) or in an
unaffected area at any time (alternative sales
by competitors, etc.), and also losses to

 

 

lU.S. Army, Corps of Engineers, "Survey Investigations
and Reports: General Procedures," Engineering Manual EM
1120-2—101 (includes change 16), 12 October 64 (Mimeographed),

ppo 50—50b.
IA

Intax

(2)

v. . .. _ a: AC n w“ a: .ny ‘1. Tc . i . _ .C v _ Ex .vx u ~ . .wx v.4 < w . i x w
.L .3 a v. C .2 .c e T C C .l l T. a 0 e C d e m
e ”a a c e 5 a m; to C .L ) .d 5 5 C 1o W C C h P; r C e (t P O
VA .1 .n . V» .b D. .l e C e C 5 I: a S e e u C t S p C
a: .Q S .t e .3 .Q n“ C. at :0 m :D a w. a S e D. S e a 0 e n
.u :c A: "x n my .v S :J a: A: r e n. it n .l .7. h u C d .‘i
.t a: r . w . e at S Ru 0 A» T". w . mm «C at {u 0 WI 0 a s T l
AC AC Av ”1 Hi . . 0 w” hi n; n1 5» I» we. Fl bl“ 1 Nu m; h a a). a e l e
_ . at a. _ .D . t a. . N; a C. Wu. .nu 3 l .mu 5 a a: . t Al. D; .t b V h «m. h
fi _ 5?. ‘IV hmu t
Mu my. D» S AU; h‘
2. .. . ht 3. ha «U
V

 

15

activities remote from the flooded area where
adjustments can be made during or after flood
periods to avoid or compensate for the loss.

(2) Intangible flood damages

 

Those detrimental effects of floods that cannot
be given market or monetary values, except by
assignment of arbitrary values or by assuming
them analogous or equivalent to marketable goods
or services. When given values they should be
classed as tangible damages; when not evaluated
they should be discussed objectively. No monetary
value is to be placed on loss of human life.

A further requirement is that damage estimation procedures
are to take into account prospective enhancement or

increased utilization:

 

Basic estimates of flood damages will be prepared
for the existing state of development of the area
surveyed. Forecasts will then be made of the probable
trends and nature of developments and activities in the
flood area and adjacent affected region, based on the
most probable economic use of the area both without
and with the project under construction . . . . The
prospective "normal" state of development without the
project, and the susceptibility thereof to flood
damage, over the life of the project will be the basis
for modifying the basic estimates of average annual
damages for current conditions to determine prospective
average annual damages. The modified damage data,
after correlation with flood stages and frequencies,
and adjustments for expected normal development condi—
tions without the considered project, will be used
for estimating probable flood control benefits.
Prospective deveIOpment with the project, if different
than that expected without the project, will be the
basis for estimating probable additional enhancement
of increased utilization benefits.

The Corps specified that tangible flood damages may
be evaluated by one or more of the standard approved methods;
as the cost of restoration (repair or replacement less
normal depreciation), comparative market or sales value,

or the income evaluation method. The first two methods are

 

 

 

 

 

 

01"

:v
i

NEEeSted

inc
ting
wen

the

a.

 

w.
“V
0
av
um.
r"
”C
.nu
..

‘. . ieew—
Ann—a A‘ I
. .M...» i

U

procedL

non-crog

n

 

“epnysical
I
I“ dEbI‘lS d5

16

suggested for evaluation of physical damages and emergency
costs; the income evaluation method for evaluation of agri—
cultural crop losses. The Corps also specified that when
the estimates are to be used for project evaluation (rather
than reporting on a flood of record), nonrecurring damages
and damages preventable by prudent management are to be
eliminated.

For agricultural areas this list of damage sources
can be reduced. Emergency costs, for the most part, are
borne by local (or larger) governing units. To the extent
these costs are encountered in both rural and urban areas,
there seems little point in allocating a portion of the
cost to the agricultural sector. Some emergency costs
specific to agriculture, such as farm home evacuation or
emergency harvest, flood—protection, and salvage costs may
need be considered. It is also difficult to conceive of a
significant intangible damage source stemming solely from
the agricultural sector.

The significant damage classes for the agricultural
sector, then, are tangible physical damages and business and
financial losses. In treating agricultural damages the
usual procedure is to classify physical damages into ”crop"
and "non—crop" categories. Non—crop agricultural damages
are physical damages to soils and farm structures. Damages
typically enumerated include bank cutting, erosion, sanding

and debris deposition, damages to farm buildings, fences,

 

 

 

 

 

 

 

nchinery , sto

of these damag
hy the cost of
:ethods.

Crop dart

{Crop damages
he functions
dration of f
or non-agricu
estimating t
essentially
iarmer--whic
flooding less
if needed, I
apart of th
hazard precl
The se
fellow dire.

major sourc

that flood

 

 

l7

machinery, stored crops, ditches, and livestock loss. All
of these damage sources can be straight—forwardly evaluated
by the cost of restoration or comparative market value
methods.

Crop damages are enumerated separately because they
can be best estimated by the income evaluation method which
automatically considers business and financial losses.1
(Crop damages are also enumerated separately because they
are functions of the seasonality as well as the depth and
duration of flooding to a much larger degree than non—crop
or non—agricultural damages and therefore require special
estimating techniques.) The income evaluation method
essentially counts (a) the direct income loss to the
farmer——which is the investment in the crop at the time of
flooding less harvesting costs foregone; with allowances,
if needed, for the possibility of replanting or salvaging
a part of the crop; and (b) income foregone where the flood
hazard precludes higher valued agricultural use of the land.

The sources of agricultural flood control benefits
follow directly from the sources of flood damages. The
major source is direct damage reduction. To the extent

that flood control reduces or alleviates tangible crop and

 

1We are assuming, of course, that crop production is
the major agricultural floodplain enterprise. This is true
for the Wabash Basin and would also appear to hold for
agricultural use of floodplains generally. See: Ian
Burton, Types of Agricultural Occupance of Flood Plains in
the United States, Research Paper No. 75 (Chicago: Univer-
sity of Chicago, Department of Geography, 1962).

 

 

 

 

 

 

 

and non-crop

attributable
In addition,
that protecti
level of. land
sources of ag

following sec

Conside
flooding haz
topography , .

this unit of

 

order for th
land must be
factors of pi
managerial a1
each of thes
other factor
land inputs

“1‘- Visualiz

divisible sc

l8

 

and non—crop damages, the value of the reduction is
attributable to the flood control project as a benefit.

In addition, further benefits are possible to the extent
that protection makes feasible a shift to a higher economic
level of land use. The rationale for considering these
sources of agricultural benefits should be clear from the
following section of this chapter.

Agricultural Benefits and the Role
of Land Values

 

Consider first a parcel of land without regard to
flooding hazard.l Depending on its location, climate,
topography, natural fertility, and other natural factors,
this unit of land will have an inherent productivity. In
order for this productivity to be obtained, however, the
land must be combined with some minimum amount of other
factors of production—~seed, fertilizer, labor, and
managerial and machinery services. Rather than deal with
each of these inputs separately, we can consider these
other factors of production as a composite bundle of non—
land inputs that we conveniently denote as "capital.” We
can visualize these non-land inputs as being highly

divisible so that units of capital can be combined with a

 

1The model used in this section is adapted from a
model presented in: Edward F. Renshaw, "The Relationship
Between Flood Losses and Flood—Control Benefits" in Pa ers
on Flood Problems, ed. by Gilbert F. White, Research Paper
No. 70 (Chicago: University of Chicago, Department of
Geography, l96l).

 

 

 

 

unit of land

acontinuous
The usu
characteristi
added to a fi
but, beyond 3
relationship ,
obtain a valu
as in Figure
decisions the
The value of

application

The uni
is assumed tc
:ition in fa:
6:“ all non-la
fined by its
price higher
dent 0n the 1

additional u]

(P0 P10).

19

unit of land in sufficiently small increments to approach
a continuous production function.

The usual agricultural production function has the
characteristic that as additional units of capital are
added to a fixed unit of land, total product will increase
but, beyond some point, at a decreasing rate. From this
relationship, assuming a constant product price, we can
obtain a value of marginal product (VMP) curve for capital
as in Figure 1. Over the relevant range of production
decisions the VMP curve is downward sloping to the right.
The value of product forthcoming at any level of capital
application is equivalent to the area under the VMP curve
from the origin to the point on the abscissa corresponding
to that level of capital.1

The unit cost of the non—land resources ("capital”)

is assumed to be (PO). Under conditions of perfect compe—

 

tition in factor markets (which implies perfect mobility

of all non—land resources), the price of capital is deter—
mined by its Opportunity cost——capital cannot command a
price higher than (PO) and will not be available for emp10y_
ment on the unit of land at a price less than (PO). If
additional units of capital are available at a constant
price we derive a straight line "cost of capital" function

(Po P'O).

 

1To simplify the graphical presentation we assume a
production function with a positive intercept and decreasing
returns to the variable factor over the entire function.

 

 

 

 

Figure

16nd.

 

 

 

 

P2 A
4.)
H
n
D
a
(I)
a
4.)
(I)
O
O
n
O
(D
0
q...
5.,
m
P J Q
o P'
I
I
I
|
I
I
I VMP
l
l
0 X
o
Capital/per unit of land
L Figure l.——Capital Investment and Rent to Floodfree
and.

 

 

 

 

Given t

to the final

units of capi
payment to ea
or the area (
or rent, that
Given a capit
nun price an

of land.

From th
of the maximu
iron the VMP
placed in pro
.han (P2). F
viii be left 1
751116 for prod
the cost of ca
Ellall see belc
tents to be
lent

on Compar.
I".
anal

Furtherm
.‘I
all" d1

80

Cum; 1n

4

41

a. 3
HP

unit or 1
a

 

 

21

Given the VMP and cost of capital functions, returns
to the final fixed unit of land will be maximized if (X0)
units of capital are combined with each unit of land. The
payment to capital is its price (PO) times the (X0) units,
or_the area (J Q X0 0). The residual (AQJ) is a surplus,
or rent, that accrues to land as the scarce resource.

Given a capitalization rate, this rent determines the maxi—
mum price an entrepreneur would willingly pay for the unit
of land.

From the diagram we can deduce a theoretical measure
of the maximum expected benefits from flood protection.
From the VMP curve in Figure l we know the land will not be

placed in production unless the cost of capital is less

 

than (P2). For capital costs greater than (P2) the land
will be left in its natural state (and have no market
value for productive purposes) until events either lower
the cost of capital or shift the VMP curve upward. As we
shall see below, offering flood protection may have both
effects. Therefore, it follows that the maximum annual
benefits to be acquired from providing flood control to
undeveloped agricultural land cannot exceed the annual net
rent on comparable agricultural land not subject to flood—
ing. Furthermore, if the discount rate and the time span
for discounting were the same for society and for the
entrepreneur, the present value of flood control benefits

DEF unit of land could not exceed the market price of

 

 

comparable flc
flood control
rents do accr
they would no
culties in th
perfect knowl
connon privat
concept at is
later conside
We now
Instead, assv
floodplain S‘
occurrence,
Peproduced a
labeled VMP1
(say, truck-
been Combine
such as cash
represent 1.
mstulated ;
The (‘

as the VMP
\
l
inte For
I‘Sectin
I‘SgapdlesS

22

comparable floodfree land. Generally, the benefits from
flood control will be less than in this example since some
rents do accrue to floodplain lands at present, otherwise
they would not be in production. There are other diffi—
culties in the assumptions of competitive equilibrium,
perfect knowledge, mobility of non—land resources, and a
common private and social discount rate but the net rent
concept at least establishes a theoretical benchmark for 2
later consideration.

We now drop the assumption of floodfree location.

 

Instead, assume that the unit of land is located on a
floodplain subject to periodic inundation of uncertain
occurrence, duration and magnitude. Figure 1 has been
reproduced as part of Figure 2 with the VMP curve now
labeled VMPl. The VMP curve in Figure l was drawn on the
assumption that the land was in its highest valued use
(say, truck—farming). The non—land resources might have
been combined in other ways to produce other products,
such as cash—crops represented by (VMP2), but this would
represent less—than—optimum use of the land under the
postulated floodfree conditions.1

The (VMPl) and VMP2) functions may still be viewed

as the VMP curves for truck— and cash—crops, respectively,

 

l . . .
. For SimpliCity, we draw the VMP curves as non-
lntersecting and assume a common Cost of capital function
regardless of the product produced.

 

 

 

  

Figure
and.

Price or Cost per Unit

 

 

 

 

23
P2 A
B
P D :E\\ H I 13)
l , l
I I l
| I l
| I
PO :K N Q P'
: l D\\\
i l
I
l I '
I VMP IVMP' VMP
I 3| 9
I l l
I I '
I I I
I | l
0 1 l l
X1 X2 X0
Capital/per unit of land
L Figure 2.-—Capital Investment and Rent to Floodplain
and.

 

 

 

occur. Howe

flood losses
accordingly .
especially v
be discounte
floodplain i
crop farming
it may also

inundation a
enterprise.

for the cash
reflect the

ity distribu
(VMP3) can b
gm val
enterprise i
protection I:

The v

 

from (VMPZ)
yields and
tion there
in replanti
a flood.1

 

2H

on the floodplain tract in a year when floods do not
occur. However, the entrepreneur must now take potential
flood losses into account and discount his schedules
accordingly. Truck farming may be a land use that is
especially vulnerable to flood damages and therefore must
be discounted heavily (to the point that production on the
floodplain is uneconomic). On the other hand, while cash—
crop farming is not as profitable on a floodfree location,
it may also be a use that is better able to withstand
inundation and therefore be the preferred floodplain crop
enterprise. The VMP2 schedule is not the relevant curve
for the cash—crop enterprise however, because it does not
reflect the expected damages from flooding. If a probabil—
ity distribution can be assigned to the anticipated damages,
(VMP3) can be derived. This curve must be viewed as the
expected value of marginal product curve for a cash—crop
enterprise in the floodplain. In the absence of flood
protection the (VMP3) becomes the relevant planning curve.
The value of marginal product schedule was discounted
from (VMP2) to (VMP3) to reflect the expected loss in
yields and other direct crop damage from flooding. In addi-
tion there will likely be additional capital costs incurred
in replanting, replacing leached nutrients, etc., following

a flood.1 Suppose an entrepreneur can expect to incur, on

 

1Although this exposition is in terms of crop damages,
other physical damages (debris deposition, erosion) could
be incorporated into the diagram. To include highly

 

 

 

 

 

the average,

each unit of
costs, the "
units of cap
non-land res
the rent acc
triangles (A
benefits fro

We can

benefits:

1. Si
ca:
ex
ea
di
th

ar

fl

    

localized d
could be th
could be 211

   
  

25

the average, additional costs equivalent to (Pl—PO) on

each unit of capital employed. Then, in terms of expected
costs, the "cost of capital" function is (Pl P'l) and (X1)
units of capital will be used. The expected payment to the
non—land resources is (DEXlO) and the residual (CED) in

the rent accruing to land. The difference in the rent
triangles (AQJ) minus (CED) is the expected average annual
benefits from complete flood protection.

We can further identify the sources of the expected

benefits:

1. Given the decision to invest (X1) units of
capital on the floodplain land, losses are
expected to be incurred from (a) additional
capital costs equivalent to (DEKJ) and (b)
direct crop losses equivalent to the area between
the (VMPZ) and (VMPB) curves measured by the
area (BREC). These losses will be avoided by
flood control measures. We term the prevention
of these losses direct damage reduction benefits.

2. With flood control but with a decision to con—
tinue to produce cash—crops, the relevant
schedules are (VMP2) and (PO P’O). Now, however,

(X1) is not the optimum capital input. Capital

 

localized damages (buildings, bank cutting), a "farm"
could be the fixed unit, or the damages from these sources
could be allocated OVer the acres flooded on each farm.

 

 

 

 

In pra

and, therefc
than the dii

(CED) . Howe

rent will be

     
   
    

Efficiency
Saparately
accrue imme
t0 higher 1
over time.

26

will be increased to (X2). The additional
cost is (KNX2X1), the additional return is
(RNX2X1) yielding a surplus (RNK) which we

term an efficiengy benefit.

 

3. With flood control, cash-crop farming is no
longer the best use of the flood plain land.
The entrepreneur will now shift to truck-farming
(VMPl). Capital usage will increase to (X0),
yielding a surplus (SNQ), but in addition each
unit of capital will now yield a higher return
than in its former use. The total gain will be
the area (AQNB) which we will term the higher

utilization benefit:L

 

In practice flood control is not apt to be complete
and, therefore, the expected annual benefits will be less
than the difference between the rent triangles (AQJ) and
(CED). However, the general effect of flood control invest—
ment will be to lower the expected cost of capital curve and
shift the expected value of marginal product curve to the

right. The result will be an expansion in the size of the

 

lIn COE terminology, the benefits we have termed
efficiency and higher utilization are considered jointly
as enhancement benefits. We will continue to use the COE
terminology where it is unambiguous. Our distinction seems
useful, however, because of the possibility that the
efficiency and higher utilization benefits may be realized
separately in time. Efficiency benefits are likely to
accrue immediately but, for a number of reasons, a shift
to higher land usage may be delayed or occur only gradually
over time.

 

 

I

 

 

 

rent triangl

and "new" tr

We ha
rent triang
benefits an

determines

(CED) would
teetion and
flood free 1
cost to the
protection"
or value.

If the
individual c
an investmer
equalled or

the rent tr

  
  
  
  
  
  
  
    

land would
being a ret
Therefore,
land, the p

rent to lan

  

1

The
Example , th
assessment

,

27

rent triangle and the difference in area between the "old”
and "new" triangle will measure the expected benefits.

We have mentioned that the difference between the
rent triangles (AQJ) and (CED) is a measure of the expected
benefits and also that the size of the rent triangles
determines the maximum price an entrepreneur would willingly
pay for land, given a capitalization rate. Thus, with a
common capitalization rate, the amount of the rent triangle
(CED) would determine the price of the land without pro—
tection and (AQJ) would determine the price of comparable
flood free land. Or, if protection were provided at no
cost to the entrepreneur (CED) would determine the "before

protection” price and (AQJ) the "after protection" price

 

or value.

If the decision to provide flood protection were an
individual one, the entrepreneur would be willing to make
an investment only if the annual costs of the investment
equalled or was less than the difference in the amount of
the rent triangles. In this case, the residual return to
land would still be the area (CED); the remaining "rent"
being a return to the capital invested in protection.
Therefore, unless the investment became embodied in the
land, the price of the land should not change since the

rent to land has not changed.1

 

1The investment would not become embodied if, for
example, the protected land was sold subject to an annual
assessment equal to the annual costs of the investment.

 

In the
toagricultu
oamfits tha
mreand the
was. Thes

swung the p

()
'3
53
O
'1
(D
(3
(D
'1
<‘f

borne by a s

amhuscular

\
I
“howev

al tax
as lan
of

28

In the case of flood control reservoirs, the benefits
to agricultural uses from reduced flood damages are flow
benefits that are released over the lifespan of the struc-
ture and that are specific to the protected downstream
areas. These benefits are captured by those individuals
owning the protected land. The individuals experience a
real wealth gain equivalent to the sum of the present dis-
counted values of the future income streams made possible
or more certain by the reservoir. These gains are nearly
pure windfall gains since the cost of the reservoir is
borne by a society in which the benefitting land owners are

a miniscular proportion.l Thus, in the case of Federal

 

1However, a proportion of the gains may be captured
by local taxing units in the form of higher property assess—
ments as land prices respond to the gain and the present
value of this liability should be subtracted from the
present value of the increased income stream. In the dia-
gram, the effect of higher annual tax liabilities could be
reflected by an upward shift in the cost of capital func—
tion. If assessment practices are consistent between flood—
plains and uplands, relative land prices would automatically
reflect the differences in tax liabilities, ceteris paribus.

Two general studies of the impact of reservoir con—
struction on county tax revenues indicate that a portion of
the increment in value is likely to be captured. See:
Jack L. Knetsch, ”The Influence of Reservoir Projects on
Land Values," Journal of Farm Economics, M6 (February,
1964), pp. 23l—2D3; and Claude M. Vaughn, Jr., Covariance
Analysis of Reservoir DevelOpment Effects on Property Tax
Base, Research Report No. 4 (Lexington, Kentucky: Univer—

 

 

Sity of Kentucky Water Resources Institute), 1967. The
State of Kentucky specifically requires reassessment of

the areas benefitting from flood control with the view of
capturing a part of the gain. (From conversation with
Harvey King, Basin Planning Branch, Engineering Division,
Louisville Engineer District, U. S. Army Corps of Engineers,
December 13, I967).

 

 

 

 

 

reservoir 0

 

sould eXPec
part of the
Ashe have .
rate (or th
13the same
that the pr
installatio
htheprice
mdthe pI‘ic

There

asingle par
an entreprené
This may alsc
control in SI:
excernalities
If the area 1
tion may shif
increased fac
tected land.

Ilain followiI
all users of 1

time the incre

M is vi
5‘8 ‘ Obstruc
Mateo 0

«an. n'

 

 

 

 

 

 

29

reservoir construction for flood control purposes,1 we
sould expect floodplain land prices to reflect a major

part of the benefits accruing to the agricultural sector.
As we have previously noted, if the social time preference
rate (or the long—term government investment interest rate)
is the same as the private land market rate, this implies
that the present value of benefits from a flood control
installation accruing to land cannot exceed the difference
in the price of land to be protected by the installation
and the price of comparable flood—free land.

There is an additional point that can be noted regard—
ing Figure 2. The discussion of the figure is in terms of
a single parcel of land for which the external effects of
an entrepreneur's decision can conveniently be ignored.
This may also be true for agricultural benefits of flood
control in small watersheds but for river basin work,
externalities may become important in social accounting.
If the area involved is large, the "cost of capital” func—
tion may shift upward following protection reflecting the
increased factor demand for capital on each unit of pro—
tected land. As more units of capital are invested in the
plain following hazard reduction, the cost per unit to
all users of the factors involved increase. At the same

time the increased output resulting from increased

 

1This will also be true of protection from Federal
levee construction except that here there will be come
associated on-site costs, primarily that of providing
Pight—of—way for the levee.

 

 

  

investment

 
  

the VMP sch

 
   
  

to investor

 
  

the same ti

duct prices

 
   
 
 

where to sh

 
  

that increa

 

decreased p

 
   
  

tion is les

effected.

 
 
 

If la.

 
  

of a large '

 
   
   

scale flood
expectations
Since the Co
:ethods esse
effects of b
conceivably
if net gain
likely that
eI‘fects of ti
from land va
unlikely,

An oeca
agency Droeec

DOSsible bem

fl

I 30 .
investment may depress product prices, tending to shift
the VMP schedules to the left. Then the rent accruing
to investors on the floodplain land will be reduced. At
the same time, the increased factor costs and reduced pro—
duct prices may cause operators on marginal lands else—
where to shift their land to other uses. To the extent
that increased production in the plain is offset by
decreased production elsewhere, the net gain from protec—
tion is less and only a redistribution of income is
effected.

If land owners are aware of the price consequences

of a large increase in production made possible by large

scale flood control works, they will adjust their income

 

expectations and land prices will respond accordingly.
Since the Corps of Engineers' project-by—project evaluation
methods essentially ignore the accumulative pecuniary
effects of basin and regional flood control investments,
conceivably land prices could prove a more accurate measure
of net gain than COE procedures. However, it is more
likely that individual land owners also ignore the external
effects of their action so that divergences for this reason
from land value and conventional benefit estimates is
unlikely.

An occasionally mentioned criticism of conventional

agency procedures is that the procedures fail to consider

possible beneficial effects to agricultural lands from

 

   
   
   
   
   
   
   
   
   
   
   
   
   
  

flooding:L
restores th
or enhances
water-borne
Figure 2 ar
criticism 1
be shifted
recurring b
the loss of

the benefit

 

value appro
it would be

the increme

 

be smaller t

would indica

9T

in TE
ff‘agricultt
Encoubtedly ,
that has bee
necessarily
b91118 reoeiv
Chances of
ge“ins sand
We weed 1
if Constant]
”Dortant CC
Egdiefi ther
to: official
, I betaLuSe
leptilizers,

2,,

31

flooding.1 The benefits may arise because the flooding
restores the water—table, improves the tilth of the soil,
or enhances soil fertility through the deposition of
water-borne silt. If the (VMPl) and(VMP2) curves in
Figure 2 are based on yields in floodfree years, this
criticism implies that, with protection, the curves will
be shifted to the left because the soil is derived of the
recurring benefits from flooding. If this is the case,
the loss of benefits from flooding must be subtracted from
the benefits gained from protection. In terms of the land
value approach, the rent triangle CED is not as small as
it would be in the absence of beneficial flood effects and
the increment in expected prices following protection will
be smaller than computations based on flood damages alone

would indicate.2

 

lClarenbach, Task Force Report, p. 1280

 

2In reviewing the literature no conclusive evidence
of agricultural benefits from flooding have been found.
Undoubtedly, delta regions are examples of a stock resource
that has been built up by flooding but this is not
necessarily evidence that a flow benefit (resource) is
being received on all floodplains. It would seem that the
chances of getting too much water at the wrong time, or of
getting sand instead of silt, or the chances of water—
borne weed infestation would mitigate against expectations
Of constantly recurring benefits from flooding. The most
important consideration, however, is whether landowners
believe there are positive benefits from flooding. Some
COE officials state that this opinion once was widely held
but, because of the almost universal heavy use of inorganic
fertilizers, it is now considered that the potential bene-
fits are marginal. (Conversation with Russell Whistler,
Project Planning Branch, Engineering Division, Louisville
Engineer District, U. 8. Army Crops of Engineers,
December 13, 1967.)

 

 

 

 

fioations (

were ignore
examine the
the validit
general he

1.

Elalliiina
The d

that the ma
are known to
actually ma
Pent is rec

a number of

“dared: in

 

 

32

Land Values in Practice
The diagrams in Figures l‘and 2 are highly simplified
views of reality based on assumptions of competitive equil-
ibrium, perfect knowledge, mobility of resources, and
known discount rates. In addition, certain other quali—
fications (e.g., ”comparable” land, market data sources)
were ignored. We now wish to relax these assumptions and
examine their effect on the model. The factors affecting
the validity of the assumptions can be listed under four
general headings:
l. The efficiency of the land market in evaluating
flood hazards.
2. The consistency of the land market in valuing
non—comparable lands.
3. The state of knowledge about components of land
value.

4. Sources of land value data.

Evaluating Flood Hazards

The diagram in Figure l is drawn on the assumption
that the marginal efficiency and cost of capital schedules
are known with some certainty and that entrepreneurs
actually make the adjustments indicated so that the full
rent is received and reflected in land prices. Although
a number of other value-influencing factors need to be con—

sidered,in a well—settled, homogeneous, agricultural area

 

 

 

 

knowledge

mine the e
floodplain
issue here
land marke
that must
can be up
to floodin
experience
The
ture stron
nation of
flood hazar
are underes
(area CED 1
lead to an
an more acc
over'Dr'iced
Darison wil
In effect,

indemnity c

33

we can usually expect land prices to be reasonably related
to the land's productivity.

The assumption that land—buyers also possess full
knowledge of the flooding hazard to enable them to deter-
mine the expected VMP and cost of capital curves for
floodplain locations, however, is less tenable. A basic
issue here would seem to be whether participants in the
land market-—to whom the flood hazard is only one element
that must be considered in a decision to buy or sell land——
can be expected to evaluate expected losses attributable
to flooding as accurately as can specialists trained and
experienced in the field.

The weight of the available evidence in the litera-
ture strongly suggests a human tendency toward underesti—
mation of the nature and extent of the flood hazard. If
flood hazards and, hence, expected losses from flooding
are underestimated, the expected rent accruing to land
(area CED in Figure 2) will be overestimated. This will
lead to an ”overpricing" of land relative to a price based
on more accurate expectations. If the floodplain land is
overpriced in view of the flood hazard, a land value com—
parison will underestimate the benefits of flood control.
In effect, part of the benefit of protection will serve to

indemnify current owners of floodplain land for their past

 

  

 

 

mistakes , a

 

trol structu
a "wait and

 

 

mistakes, and the increment in land values following pro—
tection will be less than otherwise would have occurred.1
A search of literature failed to reveal any studies
that have directly examined the land market's effectiveness
as a device for evaluating the hazard of floodplain occu—
pance or that have specifically evaluated the attitudes
of agricultural operators of floodplain land. However, a
number of papers sponsored by the Department of Geography
of the University of Chicago2 have general relevance to
the agricultural areas of the Wabash Basin. The Chicago
studies are primarily of urban and industrial uses of

floodplains but the behavioral and informational factors

 

1Theoretically, the opposite could occur; i.e.,
floodplain lands could be underpriced in terms of a
realistic appraisal of flood hazards. This might occur
over a period of time following the installation of a con—
trol structure if occupants of the former floodplain adopt
a "wait and see” attitude about its effectiveness but most
factors probably act in the other direction toward over—
pricing.

2Gilbert F. White, Human Adjustment to Floods: A
Geographical Approach to the Flood Problem in the United
States, Research Paper No. 29 (1945); FranCis C. Murphy,

Re ulating Flood—Plain Development, Research Paper No. 56
(1958); Gilbert F. White, et al., Changes in Urban Occu—
pance of Flood Plains in the United States, Research Paper
NO- 57 (1958); John R. Sheaffer, Flood Proofing: An Element
in a Flood Damage Reduction Program, Research Paper No. 65
(1960); Gilbert F. White, et al., Papers on Flood Problems,
Research Paper No. 70 (1961); Ian Burton, Types of Agri—
cultural Occupance of Flood Plains in the United States,
Research Paper No. 75 (1962); R. w. Kates, Hazard and

Choice Perception in Flood Plain Management, Research Paper
NO- 78 (1962); Gilbert F. White, ChOice of Adjustmentdto
Floods, Research Paper No. 93 (1955); R: W- Kates, In US-
trial Flood Losses: Damage Estimation in the Lehigh Valley,
Research Paper No. 98 (1965); W. R. Derrick Sewell, Wate:
Measssmgnt and Floods in the Fraser River BaSin, Researc
Paper No. 100 (1965).

 

 

 

 

 

involved iI
areas. We
the nature

flooding, :

(I) (I) (D
>4
U

<-r 14.,

 

 

35

involved in human occupancy probably apply also to rural
areas. We will discuss these factors under four headings——
the nature of the flood hazard, the expected frequency of
flooding, seasonality, and disaster relief.

The Nature of the Flood Hazard.——The most important

 

factor affecting the validity of the land value approach
is that floodplain occupants may have little knowledge of
the true risks involved. As White notes:1

The flood hazard is underestimated by most flood—
plain dwellers because of the infrequency of major
floods, the frailties of human memory, and the
reluctance of some people, for economic reasons
or from sheer obstinacy, to admit that past floods
may be repeated or exceeded . . . As a general rule,
the flood hazard tends to wax and wane in the public
mind in direct relation to the occurrence of high
water. Immediately after a flood, losses are
exaggerated and distorted. The almost invariable
experience of men who survey flood damages is that
early estimates appearing in newspaper headlines are
far too high, that later estimates by local organiza-
tions are also high but less exaggerated, and that
detailed surveys yield totals substantially less than
anticipated by the flood sufferers. So long as there
are no other floods, the memory of the last one grows
progressively dimmer. Its scars disappear, public
interest in preparedness or protection weakens, and
at length its ravages are forgotten. In time, indeed,
there may develop a pronounced aversion to public
mention of past flood disasters. Unscrupulous land
subdividers, false inflators of municipal reputation,
and speculative promoters of drainage enterprises
seek to eliminate the hazard simply by ignoring it.
Thus, average past or prospective losses from floods
as given in estimates of competent engineers, may
appear unrealistic to the people involved . .
Another discrepancy between concept and reality
in dealing with floods exists in the tendency of
laymen and technicians alike to assume that the
highest flood of record will never be exceeded. In

 

lWhite, Human Adjustment to Floods, pp. 51—52,

 

 

 

 

virtual
occupar
is take
that tr
probabl

In cc

~Ia1 «0 -
- Vy I.
Ahl‘e

,

\4

1:30 Subjec

y. .
‘ch Prices

. 36

virtually all flood plains of the United States,
occupance has been arranged, where any account

is taken of the flood hazard, in the tacit belief
that the largest flood of record also is the
probable maximum flood.

In connection with White's last point it is inter-
esting that a report issued by the American Insurance
Association on the feasibility of flood insurance by
Capital Stock Companies states: "(T)he investigations of
the engineers strongly indicate that neither the maximum
probable loss from floods nor the maximum probable fre—
quency of flood occurrences in any given period has yet

1 This study was

been experienced in the United States."
a report on the catastrophic floods in the Pacific Coast
and Northeastern States in 1955. Although areas in
both regions experienced record peak discharges, it was
concluded that neither had yet experienced "the most
severe combinations of meterological events that will at
some future time produce catastrophic floods in part or
in all of the area.”2

Whites' observations indicate that floodplain
occupants awareness of flood risk is partly a function of
the frequency of flooding. Thus, the flood hazard on
land subject to frequent inundation may be reflected in

land prices while land located at an elevation slightly

 

1American Insurance Association, Studies of Floods
and Flood Damage 1952—1955 (New York: American Insurance
Association, 1956), pg A.

2Ibid., p. 11.

 

 

 

above the f:
for the haze
enough to si
to land. It
higher eleva
attributable
This 1
defining the
among enginc
take the 111
synthetic "l
in many are:
discerned t1
flood of re
is that if .
year flood,
ing the flo
do not reco
ferences be
floodfree l
for flood r
the plain.
land values

\

1The
best of the
tenerally u
inter'Val in

 

37

above the frequent inundation level may reflect no discount
for the hazard although it may still be flooded frequently
enough to significantly affect the expected annual rent

to land° In this case a land value check applied to the
higher elevations will fail to reveal a price differential
attributable to flooding.

This possibility immediately creates the problem of
defining the relevant floodplain. A general practice
among engineers and other flood control specialists is to
take the limits of the historical flood of record or the
synthetic "hundred year” flood as defining the floodplain.1
In many areas of the Wabash an alluvial valley can be
discerned that may or may not coincide with the historical
flood of record. The problem for the land value approach
is that if any of these measures (flood of record, hundred
year flood, or apparent valley limits) are taken as defin—
ing the floodplain in an area where land buyers and sellers
do not recognize the risk of the larger floods, any dif—
ferences between land values on this "floodplain" and
floodfree land will under—estimate the true discounting
for flood risk that did occur on the lower elevations of
the plain. This possibility does not rule out the use of
land values, of course, since land values could still be

used as a check of damage computations over that part of

 

1The 1913 flood implicitly defines the floodplain in
most of the COE project studies for the Wabash. The $08
generally uses a synthetic flood with a 50 year recurrance
interval in PL 566 projects.

 

 

 

the frequenc
recognize, t
tion of the
Even i
cally evalue
pants cannot
eupected f1<
hdrologist
recognize an
mdland vai
bility of c:
it may be en
high water 1
maximum flo<
onthe othe:
that has he
masurable
ahust expe
Tothe exte
events that
nnl undere

White

the land ma
\—

In a
0? all reco
t° Six floo
Between Flc
white (ed.)

 

38

the frequency distribution that floodplain occupants
recognize, but this must be known a priori to the applica-
tion of the check.

Even if land buyers and sellers attempt to realisti—
cally evaluate flood hazards, individual market partici—
pants cannot be expected to have the same knowledge of
expected floodings as that possessed by the engineer or
hydrologist. Land buyers may have little ability to
recognize and discount for the hazards of the larger floods
and land values may completely fail to reflect the possi—
bility of catastrophic events.1 Lacking better information,
it may be entirely rational for an individual to take the
high water mark of the largest known flood as a measure of
maximum flood danger he faces as White suggests. Society,
on the other hand, will want to evaluate not only the worst
that has happened in the past but also that which has some
measurable probability of happening in the future, and to
adjust expectations and investment decisions accordingly.

To the extent that individuals ignore the possibility of
events that have no recorded historical precedents, they
will underestimate the risks of floodplain occupance.

White's observations also suggest the hypothesis that

the land market tends to reflect the occurrence of the rare

 

1In a 52—year period ending in 1957, more than #0%
of all recorded losses in the United States were attributed
to six floods. See: Edward F Renshaw, "The Relationship
Between Flood Losses and Flood— Control Benefits," in G. F
White (ed. ), Papers on Flood Problems, p Ml.

 

 

 

 

flood event
anajor floc
the receding
diately fol]
collapse as
nistically.
again as the
If such a pa
over- or one
the check is
the expected

Sheafi
proofing st:
evidence the
experienced
to 1959 (the
recent large
astage elex
Probable flc
Pegional flc

Sheaff

their city v
M

lShea!

A ma:
asslllllption c
those leadir

00d was de
eWaited to
graphic regj

 

 

 

39

flood event in a predictable way. As time passes without
a major flood, land values may gradually rise reflecting
the receding awareness of the flood hazard. Then, imme-
diately following a major flood, the market may tend to
collapse as the hazard is re—evaluated, perhaps pessi-
mistically. In a few years, land values may begin to rise
again as the physical evidence of the flooding disappears.
If such a pattern exists, a land value check will tend to
over— or underestimate expected benefits according to when
the check is made. It will be an accurate reflection of
the expected damages only coincidentally.

Sheaffer, in a study of the feasibility of flood
proofing structures in Bristol, Tennessee, found other
evidence that confirms White's observations.1 The city had
experienced 31 floods during a 92—year period from 1867
to 1959 (the year of Sheaffer's survey), with the most
recent large flood occurring in 1929. The 1929 flood had
a stage elevation 10 feet lower than the extimated maximum
probable flood and a.3 feet lower than the estimated
regional flood.2

Sheaffer found that many Bristol residents believed

their city was no longer subject to flooding since nearly

 

lSheaffer, Flood Proofing.

2A maximum probable flood was computed under the
assumption of occurrence of less extreme conditions than
those leading to the maximum possible flood. A regional
flood was defined as a flood which can reasonably be
expected to occur anywhere within the (Bristol) hydro—
graphic region at any time. Ibid., p. 65.

 

 

 

 

30 years he
hypothesize
flood in t]
the minds v
Many rente:
vere livin
of the haz
maximum po
then under
to their p
had taken
proofing t
cost ratic
reasons gi
fear of a
the visibl
flood pron
She:
the city ‘
actual fly
on a sepa
Stream f1
°Dt1mism
that land
Control 1

unless st

 

 

H0

30 years has expired since the last major flood. He
hypothesized that if an area has not suffered a major

flood in the past 25 years, for all practical purposes in
the minds of the inhabitants it is not subject to flooding.
Many renters and property owners were unaware that they
were living on a flood plain. The owners who were aware
of the hazard consistently regarded the 1929 flood as the
maximum possible flood that could occur in Bristol but even
then underestimated the height of that flood in relation
to their property. Few owners who recognized the hazard
had taken protective measures or were interested in flood
proofing their buildings even when shown that the benefit
cost ratios of doing so were substantial. One of the
reasons given for not undertaking flood proofing was the
fear of a negative effect on property values arising from
the visible proof of the flood hazards evidenced by the
flood proofing.

Sheaffer reported that stream flow improvements within
the city that had not been carried far enough to provide
actual flood height reduction and the construction of a dam
on a separate watershed that would have no effect on the
stream flowing through Bristol, were cited as reasons for
optimism regarding future flooding. This would indicate
that land market participants are likely to view any flood
control installation as offering complete protection

unless strong efforts are taken to inform them otherwise

 

 

 

or until ev
This findin
approach in
partial pro
damages fro
to warrant
may fail to
of (a) the
currently c
possibility
the levees.
The c‘
plain occuy
inhabitant:
with highel
expected to
than urban:
in their d:
greater kn.
it should
destructiv
is a funct
price sens
betveen do

f100d expe

 

41

or until events prove that their optimism was unfounded.
This finding also has implications for the land value
approach in areas receiving partial protection, especially
partial protection from levees. Although the residual
damages from greater—than—design floods may be great enough
to warrant further protection, the lands behind the levees
may fail to reflect the expected residual damages because
of (a) the tendency to overestimate the level of protection
currently offered and (b) the tendency to discount the
possibility of lower—frequency floods that would over—top
the levees.

The degree of flood hazard awareness by urban flood—
plain occupants is probably less than that expected of
inhabitants of agricultural lands bordering major streams
with higher incidence of flooding. Farmers should be
expected to be more aware of the unpredictability of nature
than urbanites simply because of their closeness to nature
in their day—to—day work. This may not give them any
greater knowledge of expected frequency of flooding but
it should increase their awareness of the potential
destructiveness of floods. However, if flood awareness
is a function of flood eXperiences, we may find that land
price sensitivity varies from basin to basin—~and even
between downstream and upstream reaches within a basin——as

flood experiences vary. This may mean that land price

 

 

 

 

differentia

will have 1
The__E

part of con

‘ the frequen
curve shows

maximum yea

occurrence.

occurrence
fear can be
stage is kn
by multiply
hfobability
all stages.
floods grea
talent to a
and the dam
ehheetation
The f
essumption
assumptiOn
aflood of

unaffecte d

 

M2

differentials and research findings for a particular area
will have little relevance for other areas.

The Expected Frequency of Flooding,——An integral

 

part of conventional methods of flood damage estimation is
the frequency analysis of flood flows. A flood frequency
curve shows the relationship between the magnitude of
maximum yearly flood flows and their probability of
occurrence. Given the frequency curve, the probability of
occurrence of floods of specified magnitudes in any given
year can be derived. If the damage expected at each flood
stage is known, the average expected loss can be obtained
by multiplying the expected loss at each stage by the
probability of occurrence of that stage and summing over
all stages. If land owners ignore the possibility of
floods greater than some specified magnitude this is equi—
valent to assigning a probability of zero to these floods
and the damages they would cause will not enter into
expectations.

The frequency curves and analysis are based on the
assumption that flood flows are independent events.1 This
assumption implies that the probability of occurrence of
a flood of any specified magnitude in a given year is

unaffected by the flood events of preceding years; that is,

 

1This is a commonly accepted assumption. Most trend
or cycle statistical tests of flood frequencies yield nega.
tive results; however, these tests cannot "prove" inde—
pendence. See: Roland C. Haynes, "Composition and Size
Of Flood Loss," in G. L. White (ed.), Papers on Flood
Problems, p. 11.

 

 

 

the occurre:

probability

:3 A
V“ hetwe

MB

the occurrence of a major flood will have not effect on the
probability of a major flood in the following years. There—
fore, the interpretation of flood probabilities in terms

of the time of occurrence has no validity; the relevant
probabilities for an investment decision faced by an
individual with a planning horizon of 10 years is the same
as that faced by society with a planning horizon of 50 or
100 years or more.

In Sheaffer's study, several respondents expressed
variations of the argument that a major flood would not
occur again in their lifetime. This indicates that not all
floodplain occupants subscribe to the view that floods are
independent events.1 If flood events are not independent——
or if individuals simply act on the assumption that the
events are not\independent——there will tend to be a diver—
gence between the relevant probabilities facing the individ—
ual and society because of the difference in their planning
horizons. Viewing floods as non—independent events has the
same effect as the other behavioral and informational fac—
tors previously mentioned; the individual will fail to

adequately discount expected land returns.

 

1The practice of labeling floods by their recurrence
interval may contribute to this view of non—independence
as it is easy to fall into variations of the argument that
"since the one in 100 year flood has already occurred, the
chance of it occurring again in the immediate future is
therefore lower than it would be if it were yet to come.”
Even though this argument has no statistical basis it may
lead individuals to underestimate the flood risks they face.

 

 

Tiih, the 1

'M

..: underes:

financ ial b

aveilable t
\

lRObe

Papers on F

 

ma

Seasonality.—-Another source of underestimation of
flood risks may arise from the fact that flooding in the
various river basisn of the United States tends to follow
well—defined, but not perfect, patterns of seasonality.1
Thus, in addition to a probability distribution of flood
frequencies there is also a distribution of time of
occurrence within years. If the median time of occurrence
within a basin is winter or early spring, and if the major
source of agricultural flood losses is crop damage, agri—
cultural losses will typically be low escept for the out—
of-season flood. Thus to realistically appraise the flood
risk, the individual needs to know both distributions. If
he underestimates the probability of occurrence of either,
the compounding of probabilities involved will lead to
even greater underestimation of flood risks.

Disaster Relief.--The factors discussed above all act
to lead a prospective land purchaser to discount the
expected return to land less than a realistic View of the
flood risk would dictate. Another factor that does not
directly affect the expected value of marginal product or
cost of capital schedules but which acts to reduce the
financial burden of flood losses is the disaster relief

available to floodplain occupants paid, for the most part,

 

1Robert W. Kates, "Seasonality," in G. F. White (ed),
Papers on Flood Problems, pp. llh—l3l.

 

 

 

45

by the occupants of nonflood areas. In addition to direct
relief aid offered by the Red Cross and other agencies at
the time of the flood; loans, grants, and rehabilitation
aid may be available from Federal and State agencies, and
disaster losses can be used to offset income tax liabilities.1
In addition, direct income losses resulting from a flood

may be offset by increased off—farm employment opportunities
for relief or rehabilitation work.

To the extent direct and indirect disaster relief aid
compensate for direct flood losses, an income maximizing
strategy for a floodplain farmer would be to plan his opera-
tion each year as if flooding will not occur. In the years
flooding does occur, he will be compensated for income
losses and his capital investments will be restored by
public aid. In effect, continued occupancy of the flood—
plains is encouraged by the public through these voluntary

contributions and taxation.2

Valuing Noncomparable Lands

 

A key phrase in any discussion of the use of land
value data is the term ”comparable land." The argument for

estimating flood damages or expected banefits assumes that

 

lWhite, Human Adjustment to Floods, pp. 196—199°

2Subjectively, this factor is probably not important

for rural areas except for the rare catastrophic flood; how—
ever, the proposed national flood insurance plans would have
this effect to the extent the insurees do not bear the full
cost of the insurance. See: U.S. Congress, Senate, A Bill
EQ_AQ§Q§ the Federal Flood Insurance Act of October 1§§E—EB'
Provide For a National Program of Flood Insurance, 8. 19 5
90th Cong., lst sess., 1967.

3

 

 

 

 

land (and a:
parable in 3
true compare

prices must

with
2122.: physio
for price c

One 0
aportion o
from a lev
remaining 1
iii may be
Changes in
FPOVlde som

price relat

46

land (and associated price data) can be found that is com—
parable in all respects except for flooding hazard. With
true comparability, any difference in the associated land
prices must reflect the market's evaluation of the flood
hazard. In this sense, "comparability" implies price com—
parability, i.e., a floodplain and floodfree tract of land
are comparable if they would command the same price in the
absence of flood risk. However, the expected price of
floodplain land with protection is unknown. Therefore,
the comparisons will generally have to be in terms of
lands with physical comparability under the assumption
that physical comparability is a sufficient condition
for price comparability.

One obvious exception to the above is the case where
a portion of a homogeneous floodplain receives protection
(from a levee, say) and where benefits from protecting the
remaining lands are to be computed. Here price comparabil—
ity may be a reasonable assumption. Also, the ex BREE
changes in floodplain land prices following protection may

provide some guides for evaluating, ex ante, the expected

 

price relationship in a similar area that is to receive
similar protection, but here the problems of selecting a
time period that encompasses the captialization period and

in abstracting from general price trends and nonflood—

related price factors may be severe.

 

 

 

A str
comparabili
projects w‘n
or presence
plain will
upland area
up by allut

may have 11‘

168s than .

lands Will

47

A strong case for expecting both physical and price
comparability could be made for irrigation and drainage
projects where the only physical difference was the absence
or presence of water on part of the land. However, a flood
plain will seldom be physically comparable to bordering
upland areas. On a floodplain defined by scour and built
up by alluvial fill, the fertility of the flood plain soil
may have little relationship to the fertility of soils on
the uplands. Slope, elevation, and internal drainage may
differ markedly. In mountainous regions, the valley
floor may provide the only extensive level areas available
for cultivation in the region. Even in relatively level
areas a flat floodplain may command a price premium because
of its adaptability to irrigation or continuous cropping
or, conversely, this land may suffer internal drainage and
ponding problems because of the lack of slope.1 The lack
Of physical comparability thus implies lack of price com—
parability.

If the advantages of floodplain occupance stemming
from the positive factors such as fertility and slope are
less than the risks of flood damage, then the nonflood

lands will command the higher price per acre, but the

 

lIn parts of the Upper Wabash Basin, drainage problems
on the floodplains are cited as outweighing flooding DTOb'
lems. See: J. Charles Headley, "Agriculture and Wabash
Basin Development,” in Regional Development and the Wabash
Basin, ed. by Ronald R. Boyce (Urbana: University of
Illinois Press, 1964). p. 51»

 

Price diffe
was the onl
not he adeq
marhct priC
advantages

the floodpl
ms “59’

even higher
was: phi

Either expe

~ :1 y
i

v

be knot
how the 133
zealous 131
themes: ii
agriculturi
such as 803
one approal
farm accow
terms of 01

non rate,

indicated '

 

48

price differential will be less than if flood risk alone
was the only noncomparable element. Therefore, it will
not be adequate to take only the difference in current
market prices as a measure of control benefits. If the
advantages of floodplain location outweigh the risks,

the floodplain lands will command the price premium. In
this case, flood control would be expected to result in
even higher relative prices for the floodplain land, but
current price data will provide few guides for estimating
either expected land prices or benefits.

If physical differences between floodplains and flood—
free land are great, it will be necessary to have some
means of adjusting one of the sets of land prices for com-
parability. The difficulty, however, is that generally the
relationship between physical and price comparability will
not be known; i.e., we do not have complete knowledge of
how the land market evaluates the relative values of
Various land attributes. A reasonable assumption is that
the most important value determinants of land purchased for
agricultural use will be those related to productivity
such as soil fertility, slope, and drainage. Therefore,
one approach is to determine from experimental data or
farm accounting records what these differences mean in
terms of crop yields. Given price data and a capitaliza—
tion rate, fertility differences can be converted into

indicated value differences, but there is no assurance the

 

 

 

narket cons
may lack th
of the fert
be correlat
ferent type
:ay be attr

‘l'v‘ term A
.uln, me

It?“

49

market consistently imputes value in this manner. Buyers
may lack this detailed knowledge of the income implications
of the fertility differences. Or, because soil types may
be correlated with floodable and nonfloodable lands, dif—
ferent types of buyers with different risk preferences

may be attracted to land on and off the flood plain. This,
in turn, may affect the relevant capitalization rates.

If a number of adjustments must be made in this man—
ner, the process can rapidly become very complicated or
yield unrealistic results because of the number of assump—
tions required. A particular problem associated with this
approach (or related approaches based on enterprise or
whole farm budgeting) is the scarcity of specific data
about flood plain yields, production practices, or costs.
Practically no regularly published agricultural data are
systematically collected below the county level, and none
is specific to floodplains.

The question of land comparability is largely an
empirical one that must be answered on a case—by—case
basis, depending on the degree of departure from compar—
ability and the availability of data necessary to
adjust market values. In some cases, skilled land
appraisers, extension agents, or knowledgeable farmers may
be able to estimate the value of noncomparable elements
Within the bounds of required precision. For greater pre—

Cision, more involved adjustment procedures may be

 

 

—.._- _- ._L

required.
eliminate

procedures

Components
One

Figures 1

Unfc
constitute
hath/91y,

PSlat iVe Q

 

 

50

required. As a result, cost (and time) considerations may
eliminate the use of land values as a check on conventional

procedures.

Components of Land Value

 

One of the assumptions underlying the diagrams in ' .
Figures 1 and 2 was the assumption that land is the only
fixed factor and,therefore, the residual claimant after
all other factors of production have been paid their
opportunity cost. If the return to a tract of land is
imputed in this manner and if the price at which the
present owner would sell and a prospective buyer would buy
is known, the capitalization rate can be determined. With
a well-functioning capital market, there would tend to be
one capitalization rate which would be closely tied to
the opportunity cost of capital——the difference, if any
exists, being attributable to the difference in the risk
attached to holding land and the risk of alternative
capital investment. Knowledge of the capitalization rate
is essential to the land value approach in order to con—
vert land price differentials (a stock value) to annual
equivalents( and also to correct for the noncomparable
elements if necessary).

Unfortunately, there is little agreement about what
constitutes the opportunity cost of capital; or alter—.
natively, there are no good methods of measuring the

relative degree of risks of alternative capital investments.

   

 

argument 5
marshalled
prime more

priate cap

equally pe

would be

rs} n— 4
wsalons

stern an

51

Arguments of varying degrees of persuasiveness can be
marshalled for the choice of a particular rate such as the
prime mortgage rate or landlord's return as the appro-
priate capitalization rate, but for each choice there are
equally persuasive objections. Wendt's appraisall that
the choice of a capitalization rate ”is essentially one
of subjective estimation” seems valid. The choice may be
of little consequence in applications involving relative
comparisons but can be critical to the land value approach.
A possible resolution for the capitalization rate
problem for the small area applications considered here
would be to find the rate implied by net land return/price
relationships. This would require the use of upland
return and price data since to compute net return to flood—
plain land would involve the use of data we are seeking to
check. However, the use of upland data requires an assump-
tion that the capitalization rate for both areas is the
same. Renshaw suggests that individuals with a propensity

to gamble are likely to be lured into the floodplain.2

 

1Paul F. Wendt, Real Estate Appraisal (New York:
Henry Holt and Company, Inc., 1956), quoted in Raleigh
Barlowe, Land Resource Economics (Englewood Cliffs,
New Jersey: Prentice—Hall, Inc., 1958), p. 193.

2Edward F. Renshaw, ”The Relationship Between Flood
Losses and Flood—Control Benefits," in G. F. White (ed,)
Papers on Flood Problems, p. 31.

3

 

 

 

A pronounc
up 5100de

1
pay.

the tender
farmers wz‘
They have

more to g:

expected
\

given yea

Capital 0
Cation.

as his p1
T£301” fl:
emnded
:0 averag

Ennual r5

52

A pronounced risk taker (i.e., a "gambler") is apt to bid
up floodplain prices over the price a risk averter would
pay‘1

If Renshaw is correct, it may also be likely that
the tendency to gamble on flood occurrences is higher among
farmers with land located entirely on the floodplains.
They have more to lose from flooding but they also have
more to gain if they plan their operation as though flood—
ing will not occur, and it does not. This can occur
because the (VMP3) and(Pl P‘l) curves in Figure 2 are
expected curves, given the associated flood risk. In any
given year, however, only a flood of a certain magnitude
(or no flood at all) occurs and, EE post, some level of
capital other than (X1) may have been the optimum appli~
cation. Thus, the entrepreneur who views (VMPl) and(PO P'O)
as his planning curve and who hits a run of years free of
major floods will receive the rent (AQJ). Over any
extended period of time, however, this strategy will lead
to average capital losses of (EIJXb X1) and an average

annual rent (CED). Thus, to the risk adverter, the

 

lTolley suggests that familiarity with the floodplain
farming may convey comparative entrepreneurial advantage to
these operators. If the floodplain land market is "made"
by a select group of buyers and sellers with prior experi—
ence in managing floodplain lands, this may result in two
separate agricultural land markets in a region. Thus, an
apparent risk preference on the part of the floodplain
Operators.may actually reflect greater knowledge of the
true risks involved (or of special risk—minimizing strate-
Sies). See: G. S. Tolley, ”Analytical Techniques in Rela-
tion to Watershed Development," Journal of Farm Economics,
60 (August, 1968), pp. 653—665.

 

 

 

 

 

expected 1
while the
average re

A m
that land
lncreasinl
may also '
both land
returns 1“

PEEUI‘HS t

 

 

 

53

expected rent (CED) will determine his bid price for land
while the risk taker will gamble on receiving a greater
average return and bid for land accordingly.l

A more general problem with the model, however, is

that land may not be the only residual claimant. There is

 

increasing awareness that in many cases, farm operators I K
may also be immobile or fixed in an economic sense. Then,
both land and labor become residual claimants to the
returns from agricultural production and, by definition,
returns to land and labor cannot be allocated separately.
It is possible to impute a return to labor and treat land
as the residual in order to deduce the capitalization rate,
but if labor is fixed, there is no assurance that the
imputed labor return is the correct one. In particular,

a case can be made that under certain circumstances, farm
operators will be led to capitalize part of their labor
earnings into land values.2 This will result in higher

land values than would otherwise prevail if farm operators

 

had economic mobility. If an annual return to this land

 

1A special field of economics (Portfolio Analysis)
has been developed that deals with decision—making in
similar rate—of—return/risk tradeoff situations but data
sources appear too scarce to make applications to this
study.

 

2The circumstances will not be discussed here, but
for a good theoretical discussion and an attempt at empiri—
cal verification, see: Roger W. Strohbehn, Resource Pro—
ductivity and Income Distribution With Implications for
Farm Tenure Adjustment, Bulletin 720 (Urbana: University
Of Illinois Agricultural Experiment Station, August, 1966).

 

 

 

 

is compute
capitaliza
rate), the

In a
ti" labor I
there are
expectatic
land that
"warrant e<

returns a1

technolog;
Values or

reservoir

should no
:0 detem
related t
County or
My Specj
affect tt
Specify t

COPDS of

54

is computed on the basis of some standard proxy for a
capitalization rate (such as the average mortgage finance
rate), the expected return to land will be overstated.

In addition to the possibility of the capitalization
of labor returns into land values under certain conditions,
there are other factors associated with landowner's
expectations or with the supply and demand schedules for
land that may result in land prices above or below the
”warranted value" of the land based on current realized
returns and a nominal capitalization rate. For example,
anticipations of price inflation or accelerating rates of
technological changes may become capitalized into land
values or landowners displaced by land purchases within a
reservoir site in one area may greatly expand the demand
for land in other surrounding areas. Generally, if all
farmers within a county or region are capitalizing their
labor earnings or expectations at about the same rate,

relative land prices between the uplands and floodplains

 

should not be affected but it may be extremely difficult
to determine that portion of land prices that are directly
related to land productivity. Again, land prices in each
county or region will have to be evaluated on the basis of
any special considerations that are known or believed to
affect the price structure of that area. In particular, to
Specify the use of a single capitalization rate as the

Corps of Engineers does (the average long—term mortgage

 

 

 

 

financing

 

y. ‘w

..‘\. "ell.

\I-A h‘

.utCtIY,
-,

“m.

...; " V‘

I ya I”

 

 

55

financing rate)1 may impose serious restraints on the

applicability of the land value approach.

Sources of Land Value Data

 

Implicit in the preceeding discussion of the land
value model is the presumption that a single—valued price
for uplands and for floodplains exist or can easily be
determined. However, "land" is a very heterogeneous
product even in its natural state. The market for land is
not well organized, specific tracts are traded infre—
quently, land sales may include sunken capital investments
that cannot be valued separately, and land prices may
refelct the influence of a number of factors (location,
grantor-grantee relationships, etc.) that have no direct
bearing on land productivity. Skilled land appraisers and
regular market participants may be of aid, but frequently,
it will be necessary to rely on recorded deeds of sale as
the primary data source.

There are a number of drawbacks to reliance on
registered deeds as a source of farm sales data. Between
196“ and 1966, the average turnover rate for all farm
sales in Illinois and Indiana was “4.3 per thousand; 27.4
Of these were voluntary with the remaining 16.9 per

thousand representing foreclosures, estate settlements,

 

 

1U. S. Army, Corps of Engineers, "Survey Investiga-
tions and Reports," p. 59-

 

 

 

 

and other
farms repc
counties c

Wabash Bas

COinlete t
With the C

Varies wit

\
l

.. U ,
"“aI‘ket .
w

1
fl) Vol e
N‘nqiana H
KIOVQ-r, 3

56

and other involuntary sales.1 In terms of the number of
farms reported in the 196” Census of Agriculture for 39
counties of Indiana and Illinois that are within the
Wabash Basin,2 these turnover rates are equivalent to an
average of 31 voluntary sales and 19 involuntary sales per
country, per year for all farms; and 23 voluntary sales
and 1M involuntary sales for commercial farms.

These sales estimates, of course, are based on all
sales in the county. Sales of floodplain land will be
smaller; perhaps in proportion to the amount of floodplain
in the county. However, personnel in the Louisville COE
Office report that much of the desirable floodplain lands
are in large, closely—held farms that seldom reach the
market except as estate settlements.

The increasing use of land contracts as a means of
transferring ownership also decreases the reliance that
can be placed in deed records. A land contract is a form
of conditional sales contract for land in which the seller
retains title of the land as security against failure to
complete the contract. The contracts can be registered
With the County Recorder, but this practice probably

varies with local custom or according to the advice of

 

 

1U. S. Department of Agriculture, Farm Real Estate
Market Developments (Compiled from various issues, 1965-66).

2 S. Bureau of the Census, Census of Agriculture,

U.
%2§fl, Vol. I, Statistics for the State and Counties, pt. 11,
Indiana," and pt. 12, "Illinois“ (Washington, D.C.:
Government Printing Office, 1967).

 

 

 

 

 

 

the lawye
that some
that all

The
that the
reduced.
payment a

mortgages

.. s
(‘D
' r3
"3
(1)
U)
(D
:3
(- c

R
physicals
aspubli.

57

the lawyer involved in the transfer.1 Thus, the fact
that some land contracts are recorded is no assurance
that all contracts are.

The first effect of the use of land contracts is
that the number of current observations of farm sales is
reduced. In addition, one of the primary attractions of
the land contract is its greater flexibility in the down—
payment and repayment terms compared to conventional
mortgages. To the extent that this flexibility is used

to finance the larger sales, deed records will not be

 

representative of all farm sales in the county.

The use of "informed persons" as the primary data
source may be adequate for some applications of land values
as a check on conventional procedures. For more detailed
land value data, the regional COE Offices might find it
feasible to regularly collect data for anticipated needs
as a part of other routine work (through the Real Estate
Divisions, for example). However, for specific studies,
it may be necessary to collect the data directly from the

county offices and this may involve appreciable costs.

 

Land Values in Practice: Summary
The possible sources of divergence from the theo-

retical land market are several and, taken together, may

m

lRegistering the contract is not universal because
physical possession of the land by the buyer (which serves
as public notice) and the seller's promise to convey a
marketable title are the buyer's primary defenses.

   

 

 

generate
must be r
are also
0: assump
expect ab
van: ques

model co.

( n
,L
I“)
.3
,4.

(11

Cr

(1

r!

I

, 4.
(D
“S
{/1

Ipequemy

generate pessimism about its applicability. However, it
must be remembered that conventional agency techniques

are also error—prone and subject to numerous uncertainties
of assumptions and data. Thus, it is unreasonable to
expect absolute accuracy from either approach. The rele—
vant question, generally, will be whether the land value
model contributes significantly to the accuracy of conven—
tional procedures to justify its use.

The sources of possible divergence from the theoreti-
cal model have been presented without attempting to assess
the relative effects on the applicability of the model.

In most cases the possible divergences require empirical
data for evaluation. Probably the most important considera—
tion is the likelihood of failure of land buyers and
sellers to fully recognize the flood hazard. This is-
likely to lead to overpriced floodplain lands in absence
of recent or frequent flooding but may also result in
temporarily depressed floodplain values after severe
flooding. The land market is least likely to reflect the
expected damages from flooding in areas where the flood
risk is slight or from catastrophic floods. In the latter
case, however, flood frequency data as used in conven—
tional techniques are also suspect because of the lack

of observations needed to establish the upper end of the

frequency curve.

 

 

The
of deterr
poses.
areas whe
considers
Land pric
price car

Minor prc

 

 

 

59

The next major consideration is probably the problem
of determining comparable lands for price comparison pur—
poses. This consideration may restrict the number of
areas where the model will be applicable. A mitigating
consideration here, however, is that most micro—studies of
land prices indicate that most variation in farm sale
price can be explained in terms of a few major variables.l
Minor productivity differences can be overcome by fertili—
zation; improved transportation facilities reduce the impor—
tance of farm location as a function of distance from
trading centers; and so forth. Therefore, price compar-
ability may not be a problem within fairly broad limits.

An inherent difficulty in working with land values

 

is that land prices may reflect a number of value—
influencing components not directly related to its pro—
ductivity. Agricultural lands may be purchased as a
hedge against inflation, in anticipation of future agri—
cultural technological advances, or to provide employment

for otherwise unused resources, particularly family labor.

 

1For instance a study of the farmland market in the
Mississippi River Delta cotton region found that up to 97%
of the variation in farm sale price could be explained in
terms of differences in acreage of cotton allotment, wood-
land and openland. Thirteen other variables for factors
such as land productivity or type of road were occasionally
statistically significant but their effect was far over—
shadowed by the influence of the three major variables.
See: J. B. Penn, Bill Bolton, and Willard F. Woolf, The
Farm Land Market in the Mississippi River Delta Cotton“.
Region, 1964—1965, D.A.E. Research Report No. 372 (Univer-
sity Station: Louisiana State University Agricultural
EXperiment Station, April, 1968), p. 61

 

To the e:
an "appai
current a
return re

these fa<

the recur
tripartit
and the c

two factc

CT Pegion
can DI‘Oba
On approp
the proba
DTiCes an

cl a cam

 

60

To the extent that these considerations affect land prices,
an "apparent” rate of return calculated on the basis of
current agricultural yields understates the true rate of
return realized from land ownership. The importance of
these factors is difficult to evaluate, especially in a
small—area land market.

Since the purpose of a land value model is to cal—
culate a return to floodplain land that is comparable to
the return estimated by conventional procedures, the
tripartite relationship between net returns, land values
and the capitalization rate requires that both the latter
two factors be known. In this review of theoretical
issues more emphasis has been placed on land value rela—
tionships than on the determination of capitalization
rates because the theoretical issues of relative land
price relationships are more numerous. Again, this is an
empirical question but, in practice, available data from
enterprise or representative farm studies and from state
or regional data on rates of return to land investments
can probably serve to establish relatively narrow limits
On appropriate capitalization rates. Beyond this point,
the probably range of error associated with both land

prices and estimated land returns make precise estimation

Of a capitalization rate improbable.

 

 

A
advocatv
evaluat:
in the :
applica1
mentions
reviews
value st

SDeCific

prOJ'ects
CePhed H
W the .
chiefly

\

lc

CHAPTER III

PREVIOUS APPLICATIONS

Although numerous researchers and authors have
advocated the application of a land value check in project
evaluation, only a few actual applications have been found
in the search of the literature. Two of the earliest
applications found were the studies by Clarenbach and Pine
mentioned previously. Both of these studies were ex p252
reviews of completed project investigations. Two land
value studies completed since then were independent of

specific project evaluations.

Verdigris River, Kansas

In his investigation of Federal agency procedures
for the Task Force Report on Water Resources and Power,
Fred Clarenbach examined four projects; one each in
Oklahoma and Kansas, and two in Texas.1 Two were COE
projects and two were SCS projects. His inquiry was con-
cerned "primarily with the reasonableness and reliability
of the results" reported by the COE and 808. He looked

chiefly at estimated agricultural flood damages (mainly,

 

lClarenbach, Task Force Report.

61

 

 

 

 

62

to crops) and the estimated benefits from land enhancement.
His analysis of the claimed benefits for the proposed
Toronto dam on the Verdigris River in southeastern Kansas
illustrate his techniques and are typical of his findings
for all four projects.

Clarenbach chose the immediate downstream reach from
the dam site for analysis. This reach contained 10,430
cultivated floodplain acres. The COE claimed annual bene—

fits from four categories:1

1. Flood losses prevented ($183,552) based on
"experienced" floods. This was equivalent
to assuming flood losses prevented were equal
to flood damages and apparently assumed com—
plete protection against recurrence of past
floods of record.

2. Project flood benefits ($1,334) from additional
possible benefits from protection against a
"design" flood larger than any of record.

3. Increased land utilization ($124,627) stemming
from increases in net income from agricultural
production, after protection, resulting from a
higher land use and improved timing of farm
operations.

4. Future development ($45,888) calculated at 25
percent of "experienced” flood losses prevented,
as an allowance for damage prevention to
”future development under present flood condi—
tions.”

Categories 1 and 2 can be considered estimates of
direct damages to crops and are equivalent to $9.09 per
acre annually, based on 1953 prices. However, in an

earlier draft report using 1949 price data the Corps had

 

1Ibid., p. 1280. (The description of the benefit
Categories has been paraphrased.)

 

n

a a _ «J a. \u/
wl n: u: n . a:

 
 
 
 

 

63

estimated that the "net earning power" per acre of culti—
vated land in flood—free years was $8.28. Converting the
$9.09 direct damage estimate to the same 1949 base yielded
an estimated annual crop damage of $6.87 per cultivated
acre and a residual return to cultivated land of $1.41
under existing flood conditions. At a 5 per cent capital—
ization rate this residual return would justify a land
value of only $28 per acre. The Corps'estimate of the
market value of this land was $94 per acre. The Corps
also claimed other (non-crop) direct damages from several
sources that totaled to 97 cents an acre leaving a
residual of 44 cents or an indicated value of land of about
$9 an acre.

The Corps of Engineers also claimed benefits from
increased land utilization of $6.29 per acre (categories
3 and 4, above). The total benefits claimed ($6.29 +
$7.84) would support an increase in capitalized value of
$280 per acre following protection. The Corps estimate
of the value of land after protection was $99-77.l
Clarenbach concluded that the inconsistencies between COE
estimates and apparent land values were sufficiently great
to completely refute the benefits claimed.

Perhaps the most significant factor of Clarenbach's

analysis is that it required a minimum of data about the

 

 

1Ibid., p. 1285.

 

-D'

 

a: is nu. .«u . s h. \v .I» Tu fit. a» I
.. l .. . "\u . .3 \x. h.” n . A: A» it a a» (A .. . .\,.
. . .x i 2 . v .. . : . .. . r0. nu VA» \\ a». «\s x .

 

 

 

64

level of land prices in the area. The obvious conclusion
from his analysis is that a simple exercise such as the
one he followed could have revealed these inconsistencies
at the time the project analysis was prepared. A second

conclusion affecting the applicability of the land value

 

approach is that gaggg estimating errors would be

revealed by applications of the principles of land value
determination and do not depend importantly on assumptions
about a capitalization rate or require exact knowledge of

land price levels or differentials.l

Grand River Reach, Missouri

 

Another pp ppgp review conducted at about the same
time took the land value approach further than Clarenbach
did in that land price differentials were used to estimate
experienced flood damages. This report was prepared by an
”Advisory Board" consisting of a farm management special—
ist, a banker, and a retired College Dean. The Board was
formed at the request of the Kansas City District COE

office and was assisted by economists William Green of the

 

lRenshaw used a procedure similar to Clarenbach's
in examining the U. S. Department of Agriculture's (SCS)
procedures for PL566 watershed projects. In the Sny
watershed in eastern Illinois he found that capitalized
benefit estimates ranged from a fraction of a per cent
0f average Census land and building values to 2370.8
per cent and averaged 92.2 per cent for the entire water-
gged. (Renshaw, Toward Responsible Government, pp. 73-

 

 

 

: .l .1 1C A: x.» s _ e . n:
n n... n . .r.) G» a... A: : _ v .. r“ A .. . . .l s. . . s it
. V o a» Li. Vt. r. . .. :u 1. J. 5.. 3 . . .. as In ..\x .. »\. ~ .

 

 

55

United States Department of Agriculture and Wilfred Pine
of the (then) Kansas State College.1

The board was requested to evaluate crop and non—
crop agricultural flood damages along the main stem of
the Mississippi River. They selected the Platte River
reach in the Omaha COE District and the Grand River
reach in the Kansas City COE District for analysis but
later found that the Platte River reach had not experienced
significant flooding in the 10 year period prior to their
investigation. The Board took this lack of flooding
experience as prima facie evidence that the flood hazard
in this reach was negligible.

In the Grand River reach the major conclusion of the
Board was that the estimates of average annual agricultural
flood damages contained in the Corps' report on the flood
hazard were in no sense indicative of the existing flood
hazard. The Board was convinced that the Corps' estimate
of agricultural damages was about fourfold greater than
actual monetary damage over a representative period of
years.2

The Corps of Engineers estimated average annual crop
and pasture damage in the Grand River reach to be $25.50
per acre after allowing for the effect of existing pro-

tective works. The Board estimated average annual

 

lPine, Report of The Advisory Board.

2Ibid., p. 6.

 

 

 

 

66

agricultural damages of between $5 and $8 per acre. The
Board used several approaches to derive this estimate; two
of which involved the use of land value data. First, an
analysis of land prices covering a period of 10 years
indicated a price difference of about $140 per acre

between average land in the Grand River reach and flood-
free land of comparable quality. At average capitalization
rate in the area (considered to be 5 per cent) this dif—
ferential indicated that average annual net returns on
flood—free land were from $6 to $8 per acre higher than

on floodable land in the reach. Interviews with 70 flood-
plain farmers in the reach indicated that, on the basis of
average conditions, flood protection would be worth about
$110 per acre, indicating an annual net return of about
$5.50.

Further collaborating evidence for these estimates
was found by carefully reviewing the assumptions and data
used by the COE in their conventional hydrographic analysis.
Serious biases in the choice of assumptions or data were
found at nearly every stage of the Corps' analysis and the
Board concluded that:

After groping at great length with the many complex
and related problems of agricultural science, econ—
omics, and hydrology involved in damage appraisal,

the utility of land price analysis seems even more
apparent and the need even more urgent.1

 

 

lIbid., Supplement IV, p. 1.

 

 

 

 

67

The land value analysis was based on a total of 48
sales, 19 of which were protected by Federal levees. The
land sold was considered comparable except for the presence
or absence of levee protection; adjustments in sale prices
for time of sale were made by a land price index series
for the State of Missouri; and the analysis was essentially
restricted to comparison of average prices per acre with
some effort to estimate reasonable ranges of error in the
price data.

The Board recognized several potential shortcomings
of the land value approach. They pointed out that the
technique may be inappropriate for areas of infrequent
flooding, that obtaining an adequate number of sale observa—
tions may be a problem, and that locating "comparable”
land data may be difficult. Therefore, they urged that a
need for flexibility in application of land price data be
recognized and recommended that a land classification
system be developed to facilitate separate analyses of
areas differing materially in physical characteristics,
land use and flood hazards. However, the Board strongly
recommended that a land sales analysis of some type be
used as a check upon the accuracy of all flood damage com—

putations on agricultural land.

 

 

 

 

 

mm.
now
,
"m
u._.

 
 

 
 

 

 

68

Ray, Clay and Platte Counties, Missouri

 

A recent study of two areas along the Missouri river
slated to receive levee protection employed professional
land appraisal techniques.1 The study was made by the Real
Estate Research Corporation (BER) for the Kansas City COE
District. The purpose of the appraisal was to evaluate
approximately 41,000 acres of Missouri River bottom land
situated in two proposed levee districts known as Project
345—330-L and 409—L along the Missouri River. The appraisal
was to include the market value of the farm areas consider—
ing appropriate highest and best use and assuming (a) the
absence of any levee protection, (b) the present state of
protection and (c) protection by the proposed levees.2 The
Corps‘ reason for sponsoring the appraisals was not stated
and, if an eg peep check was intended, the RER apparently
was not furnished with the Corps‘ estimates. Since the
findings for the two areas to be protected are similar,
only the empirical data for Project 345-330—L will be
reported.

The areas to be protected by the levee system lie
between the Missouri River and the bluffs of the River.

Presently protected lands are primarily agricultural,

 

1U. S. Army Engineer District, Kansas City, Corps
Of Engineers, "Appraisal Analysis on Agricultural Levees
345—330—L and 408—L, Ray, Clay, and Platte County,
Missouri," prepared by the Real Estate Research Corporation
St. Louis, Missouri, July, 1967. (Mimeographed.)

)

2Ibid., p. 1.

 

cop
the

 

w l n; w M .n: w u . ab 7... . . A .1 rd. 3. 3»
WM G» AL V“.- 1. . fink Tn ":4 1‘; Al. F u C e 5. x \\ “A as!) “H
1 U nun 1H - 1 \V I v. I. . .vlu V . u . an.» W\ \L r u A.- x «NH a» w..\

 

 

69

consisting of highly developed farms. The RER reported
that it was not difficult to find sales of land in areas
protected by private levees. The quality of these levees
varied from good to very poor. It was extremely difficult
to find sales of comparable land as if protected by a
Corps—built agricultural levee as most protected areas
were too remote from the studied areas to be considered
comparable. It was also difficult to find sales of land
that could be considered completely unprotected because
of the numerous private levees in existence.

The BBB began the study by contracting for the pre—
paration of soil maps and productivity ratings of soils
in the study area. A productivity rating of from 1 to 10
was assigned to the soils but, "because of statistical
limitations,"1 the ratings could not be related to land
values. Instead, the soils were grouped into calcareous
and non—calcareous soil groups. Average values per acre
of productive calcareous land was found to be 103 per cent
Of the average of all sales compared to 93 per cent for
non—calcareous lands. In area 345—330—L 40 sales of farm-
land covering a period from 1960 to 1967 were abstracted
from county records. AdJUSted for time (by moving
averages), the average price per productive acre for the
entire sample was $566. There were 23 sales that included

land substantially flooded in 1965 and therefore could be

M

1Ibid., p. 32.

 

 

 

70

considered subject to flooding. These tracts sold for an
average of $536 per crop acre. Sixteen sales involved
land that had not flooded since 1951 and these sales com—
manded an average of $609 per acre. The breakdown of

sales by soil type and flooding conditions was:1

 

Unflooded Flooding
Levee Project 345—330—L Cropland Cropland Differential

 

Entire sample $609 $535 $73
Calcareous soil 627 552 75
Non—calcareous soils 566 498 68

 

The RER also estimated present values for farm
buildings and waste or idle land to derive an aggregate
estimate of market value for the entire area (28,974 acres)
of $18,025,362 under present conditions. The RER then
used the unflooded cropland values, by type of soil, to
estimate aggregate land values after protection. The
estimate included allowances for higher building value
(increased 25 per cent) and some shifts of land from non—

productive to productive use behind the levee (some land

 

lIbid., p. 37. The calcareous soils were primarily
recent alluvium in first bottom positions while the non—
calcareous soils were of older alluvium in high bottom
positions. It seems possible that the price differentials
attributed to soil differences also represented some flood
risk differential.

 

 

 

 

 

 

71

that would be outside the proposed levee was valued down—
ward). The total market value after protection was esti—
mated to be $18,748,024.

The difference between the present and projected
market values of $722,662 apparently can be considered the
present value (at the land market rate of discount) of the
COE levee; i.e., in the absence of annual operating costs
the land market data would apparently justify a one—time
capital investment in levee construction of up to $722,622.
Unfortunately, neither COE cost nor benefit estimates were
reported so that it is difficult to evaluate this figure
in terms of conventional benefit estimating procedures.2

The RER also estimated values under the three condi—
tions using an income capitalization approach. The compo—
site net return per productive area (no flood condition)
was budgeted as $35.14. This indicated a capitalization
rate of 5.8 per cent for the floodfree cropland (at $609
per acre). The annual expense of flood occurrence for
each crop grown was estimated on the basis of clean—up
and replanting costs for the predominantly spring floods.
The private levees had been overtopped twice in the past
15 years. Therefore, the return interval was assumed to
be once every 7.5 years, yielding a probability of

occurrence of 0.13 in any year. Land outside levees was

 

 

2The RER also estimated the aggregate market value

Of land in the absence of all protection ($16,843,041) but
there were errors (discussed later) in the per—acre values
used for this calculation.

 

 

 

estima

IY|

FPesent

1w...
“WTCI

72

estimated to flood once every 2.3 years, or with a prob—
ability of 0.43. The expected annual cost of flooding was.
then calculated by multiplying the appropriate probability
by the annual cost of flooding for each crop grown on the
floodplain and computing a weighed composite loss per acre
across all crops.

The estimated composite risk of loss was $2.39 per
acre under existing conditions and $7.89 in the absence of
all protection. These correspond to capitalized values
(at 5.8 per cent) of $41 and $136 per acre, respectively.
The indicated land prices, therefore, would be $473 with
no protection, $568 behind present levees, and $609 with
the COE levee.

From market value data RER had previously estimated
values of $536 for land without protection and $566 for the
present level of protection but this apparently is a mis—
interpretation of what the Corps intended. The $566 value
is a weighed average of $536 for land subject to flooding
(with or without private levee protection) and $609 for
floodfree land. Technically, this represents ”land values
under the present level of protection” but, logically, what
should have been calculated was the value of land presently
protected by private levees. Similarly the $536 figure
includes some land with private levee protection and is pg:
the.value of completely unprotected land. Therefore, the
market value data presented in the RER report is not

adequate for comparison with the income capitalization

 

 

approal
exist 1
by prix

value c

 

 

73

approach or to determine if differentials in market value
exist between completely unprotected land and land protected
by private levees. The method used to calculate aggregate
value differences between the present state of protection
and protection by a COE—built levee is apparently valid but
the estimated difference of $722,662 cannot be evaluated
in the absence of comparable Corps figures.
Methodologically, the procedures used by the appraisal
firm were similar to those used in the Advisory Board
report discussed earlier in that relatively simple time
adjustments and average values per acre were used. The
presence of two non—comparable elements (soil type and
flood risk) complicate an analysis based solely on average
values per acre but this problem was handled by working
with aggregate data to derive final estimates. The limita—
tion to this approach is the possible lack of class observa—
tions if a number of non—comparable elements are present.
Although sample data for farm sales were used, no attempts
were made to evaluate the possible range of errors in the

data.1

 

 

1The price data were presented in the report
enabling a test of difference to be made for the estimates
of $536 and $609 per acre values under present and poten_
tial protection levels, respectively. The difference was
not quite significant at the 0.05 level using a one—tailed
E test and assuming equal variances.

 

 

A J
.rovides
obtained
flood cox
issue the

the Okla}

”939d we
"qw ~
“V“P‘vhous

Cultural

74

Washita River Basin, Oklahoma

 

A recent study of six upstream watersheds in Oklahoma1

provides some indication of the results that might be
obtained from a rigorous application of land value data of
flood control evaluation problems and illustrates the major
issue that must be resolved in judging its feasibility. In
the Oklahoma study, three of the six watersheds had been
essentially developed for flood protection by the Soil
Conservation Service for three or more years prior to the
study; these were paired with three unprotected watersheds
that were similar in general location and type of agricul—
ture.

A two stage analysis was performed, First, a regres—
sion analysis was used to obtain empirical estimates of the

effects of watershed development on land prices in the pro—

 

tected watersheds. Sales data were obtained by searching
courthouse records for bona fide sales of tracts of agri—
cultural land containing some floodplains in the water—
sheds. Sales data from 1947 through 1962 were collected.
There were 95 sales in the three protected watersheds and
89 sales in the unprotected watersheds. Independent vari—
ables in the regression analyses included acres of flood—

plain land, acres of upland cropland, assessed value of

 

1John E. Waldrop and Daniel D. Badger, Effects of
Upstream Watershed Development on Prices and Values of
Eggmland in the Upper Washita River Basin, Processed
Series P—529 (Stillwater: Oklahoma State University
Experiment Station, March, 1966).

 

arm impr
nizicam:
prote' I:

priate va

"V A: Ab W e e .7. S .1 3C r e ..,\ Z
at .1 urn T. v... .1“ a n y 13 4d e .1 e I
Ru a .t w . a l 1|. Cu 6 Mn at m; C a» h .l. S 1
r1. Mn 2. e . -. e .r t n: at 9.. n1 .nL «b a... a 1
AG 7.. .7. CO A: w. . at ‘C a: «L. e . .
.i .3 . . .‘_ A: P.“ H. 7C a» nu . _ wt. hm .1. h. e
. l . . n . A: C n; C W n xv D. at h.
nu .n... C. 2. : n at n . . . .l. v Q n . a: n .9.
.l. . . . . «C A . 1» . . 3. «u r» .. _ . 51 C :1 v»

 

 

75

farm improvements, mineral rights transferred and appro—
priate variables for time of sales and location. A sig—
nificant increase in land values attributable to flood
protection was found in two of the watersheds. No increase
in land values was found for the third watershed, but in
this watershed there was reason to believe that the capi—
talization occurred prior to the study as a result of
earlier watershed development activities.

The second stage of the analysis involved the use
of a linear programming model to estimate directly the
changes in the productivity value of farm real estate due
to watershed development. The major implication of this
programming analysis was that by far the major portion of
direct agricultural benefits resulting from flood protec—
tion accrued to the land protected from flooding.

The 808 estimates were also available. The increases
in land values indicated by the regression analysis were
compared to the productivity values of protection derived
from the linear program and the 808 direct estimates (both
Capitalized at five per cent). The resulting estimated
increase in value per acre attributable to protection

were 2 1

1Ibid., p. 35.

 

 

 

 

 

76
SCS
Watershed Regression Programming Estimates
Barnitz Creek $121.28 $391.00 $MH7.HO
Calvery Greek 0.00 275.80 338.60
Saddle Mountain 50.76 102.80 97.80

 

Both the regression and the programming analysis
indicate that a part of the direct agricultural benefits of
flood protection are imputed as returns to land. However,
from the larger viewpoint of the relative accuracy of the
land value approach, these results are inconclusive. In
the two watersheds yielding positive results, the SOS
estimates indicate (at a five per cent rate) two or three—
fold greater benefits compared to the land value data. The
difficulty is that the analysis provides no basis for
deciding which are the correct (or more nearly correct)
estimates. The programming results more nearly agree with
the SCS estimates, but this may only reflect the fact that
"without protection” yields in the model were derived from
SCS crop damage factors.

Waldrop and Badger did not attempt to reconcile
these figures beyond noting some possible sources of the
discrepancies. However, the results are in accord with
the implications drawn in the preceding chapter. This

indicates that the most important question to be answered

 

 

18 NC

I

CCU

The ;
pcrat
1 .~
« .1‘
Board

 

 

77

is not whether land values reflect flooding risks, but how
accurately.

Conclusions
The Clarenbach study did not deal with relative land

values per s: but it is a good example of how the princi—

 

ples of value determination may be applied in discovering
gross inconsistencies in the benefit—estimation process.
The procedures are very simple and could easily be incor-
porated into project evaluation methodology. The diffi—
culty, however, is its limited applicability, assuming
some improvement in COE procedures since 1955. Ideally,
land values should provide point estimates to which some

range of confidence can be attached and this range may be

 

critical where small changes in per acre estimates of
damages or benefits determines the difference between a
favorable and unfavorable benefit—cost ratio.

The other three studies at least establish that the
land market apparently does recognize and discount for
floodrisk. The Real Estate Research Corporation report
does not present enough data to evaluate the land market's
assessment of expected damages or benefits. The Advisory
Board report would indicate close agreement between the
land market and the Board's estimate based on revisions
Of COE techniques within a range of $5 to $8. However,
the Board reported that it was not able to completely

evaluate all of the assumptions and data required by the

 

floo<

CHI‘EI

COHCI

78

flood hydrograph—flood damage integration method and a
careful application of the conventional approach could
conceivably reduce the apparent agreement between the
approaches used.

The Waldrop—Badger study most clearly illustrates
the dilemma likely to be encountered in attempting to
reconcile estimates of flood benefits derived from differ—
ent sources. Thus, the question of relative accuracy of

the land value approach becomes paramount.

 

 

CHAPTER IV

SELECTION OF STUDY AREAS AND

DATA COLLECTION

The Wabash River Basin encompasses an area of
33,100 square miles.1 The Basin includes a small portion
of the state of Ohio, the major part of central and south—
ern Indiana, and the southeastern section of Illinois
(Figure 3). The length of the Basin is about 285 miles
and its maximum width is about 190 miles. The Wabash
River, a major tributary of the Ohio River, originates in
Mercer County, Ohio and flows in a northwesterly direc—
tion to the vicinity of Huntington, Indiana and thence
generally southwest to its confluence with the Ohio River
between Mount Vernon, Indiana and Shawneetown, Illinois.
The gradient of the Wabash River averages approximately
1 1/2 feet per mile, ranging from approximately A feet per

mile in the headwaters to about 0.6 feet per mile near

 

1The discussion in this section is based primarily
on the following sources: U.S. Department of Agriculture,
"Wabash River Basin Comprehensive Basin Study: Appendix H,
Agriculture," WAC Review Draft, September, 1968 (Mimeo—
graphed); U. S. Army Engineer District, Louisville, Corps
Of Engineers and Cooperating Agencies, Thg Wabash River
Basin: Water Resources Planning (32 page brochure, no
date, number nor publisher given); and Ronald R. Boyce,
ed., Regional Development and the Wabash Basin (Urbana:
University of Illinois Press, 196A3.

79

 

 

 

SOIL CONSERVATION SERVICE

“BASH

'IIPIUN

\
\ijun
‘1

IADISDN

FAYETTE

SHELBY

necnurfi

“0le
«\/

IASHINGTON

f

/ {-2,
JASPER

DUBOIS Figure 3

WABASH RIVER BASIN

ILLINOIS, INDIANA AND OHIO

U. S. DEPARTMENT OF AGRICULTURE
ECONOMIC :ESEARCHI SE
SERV In. I-Is-u
SOIL CONSERVATION ISERVICE 5.3'15JIIC

 

 

 

 

a_1

81

the mouth. The principal tributaries of the Wabash include
the Salamonie, Mississinewa, White, Embarass, Little
Wabash and Patoka rivers.

The terrain of the Basin varies from nearly flat to
gently rolling plains and lowlands in the north and central
portions of the Basin to the relatively strong reliefs of
the Crawford and Norman Uplands in southern Indiana (Figure
A). Natural drainage is frequently poor in the level
areas. Coal, interbedded with Pennsylvanian shales and
sandstone, is found in the southwestern part of the Basin
in the Wabash Lowlands, Mount Vernon Hill Country, and
Springfield Plain areas. It is currently mined (primarily
by stripping) in the eastern portion of this area. Petro—
leum and natural gas are also produced from a number of
pools located principally in the Mount Vernon Hill Country,
Springfield Plain and the southwestern part of the Wabash
Lowland physiographic units.

The climate of the Basin is classed as humid conti—
nental. Rainfall is fairly well distributed throughout the
year. Average annual precipitation varies from 36 inches
in the northern part of the Basin to A5 inches in the
southern part. The average precipitation during the grow—
ing season (April through November) is 26 inches. The
average growing season varies between 1A5 to 200 days over

the Basin.

 

 

 

IROQUOIS

ILLINOIS
\ INDIANA

KANKAKEE OUTWASH

TIPTON

ELIIIRI

IIPPECAMDE

Iflflllnltlv
FIHIIAII

III-ILLIUI

IAIIIEI

zvnsvl LLE

 

SOIL CONSERVATION SERVICE

I

Illlllfll

 

illeIl

REGIONAL
SLOPE

 

 

LOWLAND

’-d~ WISCONSIN GLACIAL BOUNDARY
.00... ILLINOIAN GLACIAL BOUNDARY

SCALE 25 50 MILES

Fig ureA
WABASH RIVER BASIN
PHYSIOGRAPHIC MAP

ILLINOIS, INDIANA AND OHIO

U. 5. DEPARTMENT OF AGRICULTURE
EC CONOMIC RESEA ARCHI SE VI
RES ST SER
SOIL CONSERVATION ISERVICE

II I l- Is-n
5, S 15 I12

83

The largest single economic activity in the Wabash
Basin is farming. The agricultural sector is highly pro—
ductive and the Basin's location conveys additional ad—
vantages from its favorable marketing position. Manu—
facturing is the next significant major economic activity
and is concentrated in east—central Indiana.

The Basin's farms produce a wide variety of crops
ranging from extensive acreages of corn, soybeans, wheat
and oats to the intensive production of commercial fruits
and vegetables. Feed grains, hay and pasture support a
large livestock population. The major livestock enter—
prises are the production of pork and beef. The major

types of farming regions are delineated in Figure 5.

FloodingrProblems

 

The floodplains of the Wabash River and its tribu—
taries comprise over a million acres but about one-half
of this acreage receives levee protection.1 The major
flood damages occur on the exposed floodplains located
primarily in the southern one-third of the Basin. There
have been 15 major floods on the Wabash since 1875. Flood
damages from the 1913 flood were $30 million; from the 19A3
flood, $21 million; and from the 1950 flood, $6 million,

all in 1953 dollar terms.2

 

lBoyce, Regional Development, p. A2.

 

21bid.

 

 

 

 

STARKE

SOIL CONSERVATION SERVICE

IIAYNE ‘

FAVETIE

Figure 5

WABASH RIVER BASIN
LAND RESOURCE REGIONS
AND
MAJOR LAND RESOURCE AREAS
ILLINOIS, INDIANA AND OHIO

U. 5 DEPARTMENT OF AGRICULTURE
ECONOMIC RESEARCH SERVICE
FOREST SERVICE

SOIL CONSERVATION SERVICE '

5

IS-II

II. I.
.S-25.I15

 

85

The "Water Resources Planning" brochure states
that ". . . the average annual damages in the basin, as
reduced by the limited flood control works built to date,

presently amount to about $35,000,000.”l

This estimate
seems high considering that the damages from the 1913
flood (generally the flood of record in the Basin) was
estimated at only $30 million in 1953 dollar values. An
estimate of current average annual damages for the main—
stem of the Wabash River (reaches W—l through W—ll) places
the average annual damages at $3.9 million after allowing
for protective works now in place.2 Nearly half (A9.A
per cent) of this amount is attributed to agricultural crop
damage and another A2.2 per cent to non—crop agricultural
damages. The remaining damages are distributed about
equally between damages to transportation facilities,
levees and urban areas.3

Floods in the Wabash Basin are primarily late
winter—early spring floods resulting from a combination
of rainfall on frozen or saturated ground and snow melt.
Late spring or summer floods are usually the result of

convection—type storms of limited aerial extent. Kates,

 

lU.S. Army Engineer District, Louisville, The Wabash
River Basin: Water Resources Planning, p. 5.

2Data supplied by the Louisville District Office,
U.S. Army Corps of Engineers.

3The COE data for urban damages are residual to all
local protection in place or authorized.

 

 

 

86

in his study of seasonality of flooding in the Ohio River
Basin,1 found that flooding in the Wabash and White water—
sheds was less seasonal (less concentrated) than in the
other major watersheds of the Ohio Basin. The cumulative
concentration of flood events by months did not exceed 70

per cent until the month of May, and the mode of occurrence

 

was March—April.2 Thus, flooding in the Basin encroaches
on the crop—planting and early growing season to a much
greater extent than in other watersheds of the Ohio Basin.

In 1967, the COE reported that there were 28

 

separate Congressional Resolutions outstanding which

request study of portions or all of the Wabash River Basin.3
The earliest comprehensive report of survey scope dealing
with water resources of the Wabash Basin was completed in
1932.“ This survey found that improvements of the Wabash
River by the Federal Government were not advisable at that

time. A subsequent survey report in 19AA concluded that

flood control by levees would be the most attractive

 

1Robert W Kates, "Seasonality,” in G. F. White
(ed ), ape rs on Flood Problems, pp. 115—128.

 

2Modes were March for the northeast and north—central
portions of the Ohio basin and along the main stem of the
Ohio River, and January—February for the southern tribu—
taries of the Ohio.

3 U. S. Army Engineer District, Louisville, Corps of
Engineers, "Wabash River Basin Comprehensive Study,"
Interim Report No. 3, Vol. III, Appendix C, p. 2. (Here—
after referred to as Interim Report No.3).

“U.S. Congress, House, Wabash River,,0hio, Indiana
and Illinois, H. Doc. 100, 73rd Cong., lst sess., 1932.

N

 

iIlI. DRUG IIIZDI: s V F .I. I. .KL

 

87

improvement from a financial viewpoint.l Since the 19AA
report, which lead to the authorization of several
federally—financed levee projects, emphasis has switched
to construction of multi—purpose reservoirs as the primary
flood control medium. The reservoirs are operated pri—
marily for flood control purposes but also fulfill some
low-flow augmentation and water-based sport and recreation
functions.

Presently there are six federally-financed reservoirs
in the basin: Monroe, Cagles Mill, Mansfield, Mississinewa,
Salamonie, and Huntington. Construction of five additional
reservoirs was authorized by Congress in the Flood Control
Act of 1965.2 In addition, the third Interim Report found
favorably for five additional reservoirs.3

Prior to federal assumption of the financial respon—

sibility for flood control work, reliance for flood

 

lU.S. Congress, House, Wabash River and Tributaries,
Indiana and Illinois, H. Doc. 197, 80th Cong., lst sess.,
19AA.

 

 

2See: U.S. Congress, Senate, Lafayette and Big Pine
Reservoirs, Wabash River Basin, Indiana, S. Doc. 29,
89th Cong., lst sess., 1965, and U.S. Congress, House,
Lincoln, Clifty Creek, and Patoka Reservoirs, Wabash River
Basin, Indiana and Illinois, H. Doc. 202, 89th Cong., lst
sess., 1965.

3Louisville, Helm, Big Walnut, Big Blue, and Downey—
Ville. The Interim Reports are so called because they are
part of a Wabash River Basin Comprehensive Study scheduled
for completion in 1969. Interim Reports No. 1 and No. 2
recommended construction of the reservoirs authorized by
S. Doc. 29 and H. Doc. 202 (footnote supra).

 

 

 

 

88

protection was placed primarily in levee construction,
either through individual efforts or through local levee
districts. The Corps has incorporated many of these
privately—owned levees and build additional levees into
an extensive system, particularly along the lower reaches
of the Wabash. From examination of topographic maps it
appears that almost all areas in the Basin where the
ratio of area protected to levee miles required is favor—
able are receiving some type of levee protection. Many
of the private levees are inadequately designed and main-
tained but even a small levee can be effective in prevent—
ing inundation from the frequent small floods.l

In the process of this study essentially all the
floodplains in the Indiana portion of the basin that are
accessable by automobile were examined. The general
impression one receives is that the flood risk is recog—
nized by the majority of the floodplain inhabitants and
adjustments have been taken to minimize risk. Where
topography permits, farmsteads have been located off the
floodplain or to take advantage of elevations existing on
the floodplain. In the areas of extensive floodplains
(Knox and Sullivan counties), the type of agriculture is

such that on—farm residency is not required and the rate

 

 

1According to Corps personnel in the Louisville
District office, there are 1A5 named levees in the Wabash

 

Basin, with the majority constructed by private interests.
In addition, there are other smaller levees that the COE
has not enumerated.

 

 

 

 

 

 

89

of abandonment of existing structures is high. In this
area, there are some low-grade summer homes or fishing
shacks located behind leveed areas but most substantial
housing is located so that the main inconvenience from
all but the largest floods would be isolation. The farm—
ing areas behind levees are nearly all cleared and inten—
sively cultivated, except for sloughs in former riverbeds.
Corn for grain, soybeans, some wheat and an occasional
field of alfalfa were observed in the summers of 1967 and
1968. In unprotected areas crop production appeared
restricted primarily to corn production and the amount of
idle or uncleared land was greater. There is little
fencing, or need for it, in the floodplains.

It was also noticed that the major stream valleys
tend to be "U—shaped" in cross-section, particularly in
the Upper Wabash area and along some of the unprotected
areas of Knox and Sullivan Counties. As a result, moderate—
sized floods (say, those associated with a 5 to 10 year
recurrance interval) tend to inundate essentially all
the valley floor. Larger floods result in greater depths
of inundation but add relatively little to the areal
extent of flooding. From examination of topographic maps
this characteristic appears typical of the Basin. For
this reason the Wabash Basin did not provide opportunities
for fully testing the relationship between land values and

flood risk on infrequently flooded land.

 

 

 

 

 

 

90

Selection of Areas for Investigation

 

Because of the nature of the Wabash Basin and the
study objectives, the selection of counties and areas
within counties for investigation was made on a systematic
basis. The counties selected for this study are not
necessarily representative of all agricultural areas in
the BaSin and the results probably cannot be aggregated
to represent the Basin or portions of the Basin affected
by a specific project. However, conclusions about tech-
niques and principles should have relevance to similar
agricultural areas both within and outside the Wabash
Basin.

The selection procedure involved, first, elimination
of certain counties that, a priori, appeared unfavorable
for a land value check and then selecting from the remain—

ing areas counties with flood risk and farming conditions

 

consistent with the objectives of this study but which could

also be considered "typical" of a range of conditions
likely to be encountered in applying a land value check.
The consideration of study areas was restricted to
counties located downstream from existing or proposed
flood control reservoirs to insure that COE damage esti—
mates were available for comparison purposes. Areas with

large urban centers were excluded because urban site

 

 

91

values may unduly affect the price for agricultural land,
especially along streams which may have residential site
value with protection, and because the large number of
urban lot transactions make deed searching time—consuming.
Counties excluded from consideration for this reason were
Marion (Indianapolis), Vigo (Terra Haute) and Tippecanoe
(Lafayette).

A tier of counties along the East Fork of the White
River in the Crawford and Norman Uplands and the Mitchell
Plain was not considered favorable for a land value check
because these counties contain large holdings of federal
and state forest lands that would complicate land value
determinations. The White River is deeply entrenched and
what farming is found on the narrow floodplains appears to

be of a subsistance nature.

 

Personnel at the Louisville District Office have
characterized farming in the extreme lower reaches of the
Wabash (roughly, from Mt. Carmel, Indiana to the Ohio River)
as being highly speculative. This area includes Gibson
and Posey Counties, Indiana, and White and Gallatin
Counties, Illinois. Parts of Posey and Gallatin are sub—
ject to backwater flooding from the Ohio and other areas——
such as land in Gibson County between the White and Patoka
Rivers——are subject to overland flows. Because of the
flooding and the slow drainage, many large tracts are

uncultivated. The Rochester and McClearys Bluff Federal

 

92

Levees afford protection to part of the area, and other
levees are under construction or authorized which may
further speculation in land values in anticipation of
future protection. Therefore, these counties were ranked
low as potential study areas.

The analysis used in this study depends importantly
on being able to locate tracts of land on topographical
maps to facilitate determination of the flood risk. The
United States Geological Survey has published topographic
maps (”quadrangles") covering the entire Indiana portion
of the Basin at a scale of l:2A,000 (1 inch = 2,000 feet).
These maps cover a quadrangle of 7 1/2 minutes of latitude
and longitude, and generally show 10 feet contour inter—
vals (and 5 feet supplemental contours in some cases).

Mapping of the 7 1/2 minute quadrangles is much less
advanced in Illinois, and only 15 minute maps are avail—
able for most areas of the Illinois portion of the Basin
(the exceptions are primarily the area bordering the Wabash
River in the lower part of the state). The scale of the
15 minute maps is l:62,500 (1 inch = approximately 1 mile)
and contour intervals are 10 or 20 feet. This scale is
much less satisfactory for the intended use in this study.

The greater scale of the 7 1/2 maps favored the
choice of counties in Indiana. The areas of Illinois that
are covered by the 7 1/2 minute quadrangles border the

Wabash and reflect essentially the same topographical and

 

93

flooding conditions as found on the Indiana side of the

river. Since working on only one state would facilitate
some data collection activities, it was decided to con-

fine the study to counties in Indiana.

From the remaining counties, three sample areas con—
taining eight counties were designated for the analysis
of land market transactions. The choice of each area
reflected different land market or flood risk character—
istics that were desired to have represented in the analy—
sis. The counties and areas chosen were:

1. Knox and Sullivan Counties in the Lower

Wabash area;

2. Carroll, Cass, Miami, and Wabash Counties

along the upper reaches of the Wabash; and

3. Bartholomew and Jackson Counties in the

White Subarea.

Some general agricultural characteristics of the
three areas are presented in Table 1, together with the
corresponding figures for the State of Indiana. The river
reaches and acres of rural floodplain in the counties are

indicated in Table 2.

The Lower Wabash Area

 

Knox and Sullivan Counties were chosen because they
include reaches of the Wabash and White Rivers with the
largest rural floodplains in the Basin. The unprotected

floodplains of the two counties are subject to considerable

 

 

 

 

 

 

   

 

 

 

 

 

 

 

 

 

 

9A

 

TABLE 1.——Farms and Farm Characteristics by Sample Areas, Wabash Land Value
Study, 1956 and 1965

 

 

 

Upper Lower

Item Unit State Wabash Wabash Subarea
All Farms 196A Nr. 108,082 5,163 2,355 2,388
All Farms 1956 Nr. 128,160 6,195 3,019 2,635
Change 1956—6A % —15.6 —16.6 —22IO —9.A
Proportion of Land in Farms % 77.A 89.0 80.A 69.8
Average Size 196A Ac. 165.9 175.7 212.6 172.7
Average Size 1959 Ac. 1A5.2 151.0 17A.3 150.0
Value of Land and Bulidings

Per Farm 196A 3: 51,616 61,854 58,051 A8,1A2

Per Acre 196A $ 309.8A 3A9.99 270.A9 287.2A

Per Acre 1956 $ 265.00 316.77 217.90 228.60

Change 1956—6A % +16.9 +10.5 +2A.l +25.6
Value of Product Sold

per Farm $ 10,227 13,390 13,288 9,359
Commercial Farms 196A

Proportion of All Farms Z 59-6 80-5 76.0 63.8

Average Size Ac. 212.3 207.6 266.9 278.9

Average Value Acre $ 312.55 3A9.08 266.1 283.3u

 

Source: U.S. Bureau of the Census, Census of Agriculture, 196A, Vol. I,
Statistics for the State and Counties, pt. 11, Indiana (Washington,
D.C.: Government Printing Office, 1967)-

 

 

 

AB
EU
3
O‘.’

 

'11

 

95

TABLE 2.—-Study Areas, Reaches, and Floodplain Acres, Wabash
Land Value Study.

Floodplain
Study Area Stream Reach R1616 2:322” acres per
p stream mile

 

 

Upper Wabash Wabash W—8* 6,000 561

—do— w—9 19,000 062

—do— w—10 11,500 590

—do— w—11* 3,000 199

Lower Wabash Wabash W—2* 11H,000 2,092

—do— W—3 99,700 3,357

—do- w—u* 154,000 1,700

White WH—l 24,200 069

—dO— WW—1* 120,000 1,367

@2129 EeF. White EW—3 15,600 382
Subarea

“———- _do— Ew—u 71,600 1,311

~do— EW-5 3,u00 596

Clifty Creek CC—l 3,500 185

Driftwood DR—1 8,600 573

Flatrock FR—l 9,000 379

 

*
Only part of reach included in the study area.

Source: Compiled from various project justification
reports, Wabash River Basin.

 

at .1~ at at \ Wu K
n. w . 1. . O. E: TL

 

 

 

96

flood risk. It was believed that Knox County, especially,
would yield a high number of observations because it is
bounded on three sides by the Wabash, the White, and the
West Fork of the White. Sullivan County has a long
frontage on the Wabash.

The second factor in the choice of these two counties
was that both contain extensive levee systems. The combina—
tion of terrain and levee systems thus provided three types
of land——unprotected flood plains, protected flood plains,
and uplands——for which sale data could be obtained. Nearly
all the levees on the Wabash River are part of the COE—
built systems (Brevoort, Niblack, and Gill Township). Two
levee systems are authorized for the White River within
Knox County but the existing levees were privately built.
Parts of these levees have been repaired and upgraded by
the COE but the system is not as complete as the Wabash
levee systems.

Both Knox and Sullivan Counties are in the corn,
wheat and truck type—of—farm area.l Truck farming is
restricted primarily to a ”sand hill" area bordering the
river valley. The floodplains are intensively cultivated
for cash grain production. The townships in the north—

central part of Knox and the eastern part of Sullivan

 

lLy nn Robertson, et a1. Types of Farming in Indiana,
Bulletiny 628 (Lafayette: Purdue University Agricultural
Experiment Station, December, 1955), p. 23.

 

 

 

97

appear less productive and it is in this area that strip—
mining is found. Of the three areas, Knox and Sullivan
Counties have the largest farms and highest proportion of
commercial farms, but the lowest per—acre values of land

and buildings (Table l).

The Upper Wabash Area

 

A tier of four counties from Carroll to Wabash along
the upper reaches of the Wabash River was chosen to repre—
sent an area that has undergone a significant recent
change in flood risk.1 One purpose of choosing this area
was to determine if land values on the floodplain have
responded to this change, as an indication of the respon-
siveness of land values to varying flood risk.2 These
counties, which include parts of reaches W-8, W—9, W—lO
and W—ll, may also be representative of a number of other
reaches of the Wabash and its tributaries with moderately
narrow floodplains (around 500 acres per stream mile) with
respect to the problems likely to be encountered in working
with land values on floodplains of this size. Because the
floodplains are relatively narrow, four counties were

selected to increase the number of observations of farm

 

1Construction of the Mississinewa, Salamonie, and
Huntington Reservoirs is expected to essentially eliminate
flooding down river to Logansport.

2This objective is developed further in the anlysis
sections of this study.

 

 

 

 

98

sales on the floodplains. Only sales of land in townships
bordering the river were taken.

Because of the narrowness of the floodplains the
majority of the farms in the area ”overlap“ or are only
partly embraced by the floodplains. The farmsteads are
generally located on the upland portion of the farm or on
the fringe of the floodplain.

The parts of the four counties for which sales data

were taken are in the Central Grain and Livestock area of

 

Indiana.l Robertson describes this region as a glaciated
area which contains more fairly good farm land and pro—
duces a greater total value of agricultural products than
any other area in the state. No other area is higher in
average productivity of all land.2 Carroll County is the
most agricultural of the four counties and appears to have
many characteristics similar to the area between Terra
Haute and Lafayette with respect to the size of floodplains
and agricultural development. Census land values in all
four counties are the highest of the counties selected
(Table 1),

Cass, Miami and Wabash Counties each contain a city
of over 10,000 population, with Logansport in Cass County
being the largest. These cities are only slowly expanding

their populations and do not appear to exert much influence

 

lRobertson, Types of Farming in Indiana, p. 22.

2Ibid.

 

 

 

 

99

on land values outside their immediate vicinity. The
north bank of the Wabash between Logansport and the Wabash
County line is a historic travel route and land ownership
patterns are cut—up by old rights—of—way for a canal and
an interurban line, and by an existing highway (U.S. 24)

and a rail 11ne.l

The White Subarea

 

Bartholomew and Jackson counties were chosen for the
third study area because, being much less homogeneous, the
area provides a wide contrast in topography, type of farm—
ing, and nature of flood risk——both within the area and in
contrast to the other study areas. The "average” flood—
plain size is intermediate to the floodplains in the Upper
and Lower Wabash areas but this average is made up of one
reach with 1,311 acres per stream mile and other reaches
with very narrow floodplains on the tributaries of the
East Fork of the White River (see Table 2). Most of
Bartholomew County and the floodplains of Jackson County
are in a corn—wheat—hog farming area with the remainder
in a general farming area. The two counties are within
the area expected to receive protection from one reservoir
proposed in the Second Interim Report, and two reservoirs

proposed in the Third Interim Report. Therefore, current

 

1The analysis of land values indicated no difference
in value for land lying north or south of the Wabash that
Might be attributed to this factor. The abandoned right—
Of—ways are cultivated where this is physically possible.

 

100

examples of conventional COE estimating procedures are

available for these reaches.

Data Sources

The general sources of data for this study included
public records at the county level, U.S. Geological Survey
topographical maps; USDA Soil Conservation Service soil
maps and reports; county plat maps; aerial photographs
obtained through the Agricultural Stabilization and Con—
servation Service, USDA; COE working material, and miscel—
laneous primary and secondary data on file with the North
Central Resource Group, Economic Research Service. The
public records at the county level include warranty deeds
and contracts of sale registered with the County Registrar
of Deeds and farm appraisal data from the Office of the

County Tax Commissioner.

ProEerty Records
The property records on file with the County Regis—

trar of Deeds include deeds of sales, contracts for sale,
and mortgage instruments. Information available from the

instruments includes:

Type of instrument Acres transferred

Date of recording Sub—surface rights reserved

Date of sale Special convenants and
easements

Grantor—grantee
Price paid

Legal description

Federal Revenue Tax paid

 

 

 

 

101

Some of this information was directly usable, the remainder
was useful in determining if the property sold could
reasonably be expected to be agricultural land1 and if the
sale was a bona fide transaction between a willing buyer
and seller. The data for analytic purposes included the
'date of sale, price paid, legal description, acres trans—

ferred, and mineral or other rights reserved.

Assessment Records

The State of Indiana is currently on a seven—year
real property appraisal schedule. The last statewide
appraisal for tax purposes was made in 1961. Another
appraisal was scheduled for 1968, but the records available
for this study reflected the 1961 appraisals updated to
reflect physical changes since 1961.2 Real property is
assessed at one—third its appraised value.

The County Assessor maintains an office file of the
actual worksheets used by the appraisers. In addition to
the usual assessment data, these cards provide considerable

information on the physical features of each tract,

 

lThe presumption was that, in the absence of other
indications to the contrary, sales of sizeable tracts in
Open agricultural areas were for farming purposes. For
convenience, we may occasionally refer to these tracts as
"farms." However, since many tracts are bought for farm
expansion purposes, the tract itself may not have all the
characteristics (such as a set of farm buildings) usually
associated with the concept of a farm as an entity.

2The counties maintain only the last proceeding
appraisal in an inactive file. Thus, the existence of ade-
quate appraisal data is likely to be the major obstacle to
Ehe collection of sales data over any extended period of
me.

 

 

 

102

including acreage of land by use and grade, topographical
features, and building grade factors. The use of appraisal
data always raises questions of accuracy and consistency;
however, appraisals are the only practical source of
building value data. The information on physical features
were checked against other sources where possible.1
Generally, an appraisal card is maintained for each
tract of land even where one person owns several tracts,
including contiguous parcels of land operated as a unit but
with different ownership histories, so that few problems
of separating appraisal data for combined units were

encountered.

Maps and Photographs

 

A complete set of USGS topographic maps were assembled
for each area. These were taken to Louisville where they
were matched to COE floodplain charts. The inundation

limits of natural or synthetic floods for a range of

 

lAcreage figures from both deed and assessment
records were spot—checked against local agricultural sta—
bilization and Conservation (ASC) office records in several
counties. The deed and assessment records were more
detailed and considered more reliable than the ASC records
because the ASC filing system is designed for a specialized
use and was not well suited to matching with other sales
records. The ASC also maintains records of the feed—grain
base on each farm but this was not considered a significant
determinant of value. Studies have indicated that the
profitability of participation in the feed—grain program
is quite sensitive to grain prices and therefore capital—
ization of a feed—grain base was not expected. See:
James Vermeer, Profitability of Participation in the 1962
Feed Grain Program in the Corn Belt, ERS—362, U. S. Depart—
ment of Agriculture (Washington, D. C.: Government Printing
Office, n.d.).

 

 

103

exceedence frequencies was transferred to the maps and pro—
vided the basis for flood risk determination for flood—
plain sales. In the editing process the boundaries of

each floodplain tract sold was located on the topographic
map. All upland sales were spotted on the maps to check
for any unusual topographic conditions that might affect

the sale (quarries, frontage on small streams that might
present local flooding hazards, etc.).

Ownership plat maps were obtained locally or from
commercial sources.1 The plat maps, together with the
legal descriptions of the tracts sold, were used to check
locations on the topographic maps and aerial photographs.
A11 plat maps were current (1963 to 1966) except for

Miami County for which only a 1957 map was available.

 

Index sheets (a photograph of all aerial photographs
of a section of a county laid out in proper spatial rela—
tionships) were obtained for each county with the intention
of ordering enlarged photographs of the relevant areas of
the county. However, the index sheets, with magnification,
were found usable and advantageous for easier location of
each tract. Every tract for which a data schedule was
Obtained was located and examined on the photograph. All

Counties had been photographed since 1960.

 

lRockford Map Publishers, A525 Forest View Avenue,
Rockford, Illinois 61108.

 

 

 

 

 

104

Soil maps were available for Miami, Cass, Carroll,
Knox and Bartholomew Counties. Wabash, Sullivan and
Jackson County are being surveyed but only generalized
soil maps were available for this study. The use of the

soil maps is discussed in a later sectionxof this report.

Data Collection

Farm sales data were collected from county records in
the eight study counties by enumerators hired from the
Statistical Reporting Service, U.S.D.A. Because recording
practices are not uniform from county to county it was
necessary to give each enumerator individual instructions
at his assigned county courthouse. The first stage of data
collection involved a search of deed record books to record
all eligible sales between the dates assigned for each
county. In the Lower Wabash and White Subarea, the dates
were from January 1, 1964 to December 31, 1966. The dates
were chosen on the basis of a pre—test in Knox County which
indicated that about three years of sales records would
yield an adequate number of floodplain sales for statisti—
cal reliability. The dates initially chosen for the Upper
Wabash were the calendar years 1955-56, 1960-61, and 1965-
66. In these counties it was desired to have data on farm
sales over a span of time in order to investigate the
effect of flood risk reduction on floodplain land values.
Inadequacies in earlier assessment records determined the

Starting date of 1955.

 

 

 

105

The enumerators were instructed to record all trans—
actions of more than 10 acres1 that were outside corporate
limits, that carried tax stamps or a statement of price
paid, that were general or special warranty deeds or land
contracts, and that were not otherwise ineligible. Trans-
actions were not recorded if:

1. The special warranty deed was given to settle

a tax claim or mortgage foreclosure.

2. The transaction was for the purpose of granting

an easement or quit—claim.

3. The transaction was made as a gift, a life

 

estate was retained, or the sale was for
reconveyance.
A. Personal property that could not be valued

separately was included.

 

5. Only timber or subsurface rights were transferred.

6. An undivided partial interest was granted. (How-
ever, the transaction was recorded if undivided
partial interests were conveyed, as in the case
of a buyer assembling partial interests from
several heirs, and all interests could be

accounted for).

 

1The minimum acreage was deliberately set low to
determine if these small acreage sales could be used to
Supplement the number of usable observations. However, it
was found that sales of tracts containing between 10 and 19
acres contained a disproportionately high percentage of
Purchases for residential purposes. Consequently, the
final sample contains only sales of 20 acres or more.

 

 

106

7. The transaction involved a corporate deed or
a business firm not clearly identified with
farming purposes.

The criteria were designed to ensure that the sales
recorded were primarily bona—fide market transactions of
agricultural land. Further office checks were made to
locate kinship sales and sales with locational attributes
that might affect sale price.

The schedules from the first stage of data collection
were mailed to the East Lansing office as they were com—
pleted. Here they were checked for completeness and
accuracy and edited. In addition, grantor—grantee informa—
tion was transferred to a second schedule to facilitate
the collection of farm assessment data for each sale. For
the second stage of data collection, three of the enumera—
tors (one for each area) were instructed to search the
county assessment records for assessment data. In addition,
the enumerators checked questionable schedules collected
in the first stage and, in the Upper and Lower Wabash
areas, made additional searches for sales in years not
originally assigned.

A search for additional sales in these two areas
was decided upon after the initial returns of first data
collection stage indicated that fewer floodplains sales
than expected had occurred in the Upper Wabash area. The

number of floodplain sales in the Lower Wabash was about

 

 

 

107

as expected but relatively few involved sales of unpro—
tected lands. Rather than resume a search of deed records
for all sales in the years not originally assigned, a
listing was made of all floodplain landowners as deter—
mined from the ownership plat maps. The enumerators were
instructed to check this list against records in the
County Auditor's office to determine if any ownership
transfers had occurred over the period and, if so, to
record the information from the deed and assessment
records. This procedure required less enumerator time
than would have been required to search the deed books
but the time saved was offset by the clerical requirements
of compiling and alphabetizing the ownership list. The
procedure was not used in the White Subarea because the
initial returns indicated a sufficient number of flood—
plain observations was being obtained.

In the office, the schedules were again checked and
edited. Each eligible sale was located on a plat map,
soil map (where available), aerial photograph and topo—
graphic sheet in order to identify locational attributes,
soil characteristics, and topographic and flooding con-
ditions. This information was then coded and transferred
to punch cards for processing and analysis.

In the Upper Wabash, sales were collected for the
townships bordering the Wabash River and covered a period

from 1952 through 1966. After final editing, there were

 

 

108

A06 sales recorded, of which 60 contained some floodplain
land. This included 15 sales that were subject to flooding
from minor tributary streams. Sales on these streams were
not deliberately sought but were picked—up as part of the
collection of upland sales.

In the Lower Wabash, sales in all of Knox county
were collected. In Sullivan County all sales of land
lying west of Range 8W were selected. This eliminated
three civil townships and parts of two other; that con-
tained no floodplains. The final count yielded 33A sales
covering the period 1962 through 1966. There were 7A sales
containing some floodplain land, of which four were along
minor streams.

In the White Subarea all of Bartholomew and Jackson
counties were originally sampled but sales from Harrison,
Ohio, and Jackson Townships in Bartholomew and from Salt
Creek, Pershing, and Owen Townships in Jackson were sub—
sequently eliminated because they lay in the Norman Upland
physiographic region. This region is heavily forested,
farms are small and differ considerably from farms in the
lowland regions of the two counties. There were a total
of 199 observations after final editing; of these 52 were
floodplain sales. The sales covered the period of 19A6-66

only.

 

 

109

Floodplain and Upland Comparisons

Tables 3 through 6 present a comparison of the upland
and floodplain sales in the three study areas. In the pre—
liminary regression analysis sales in each county were
analyzed separately and in combinations. The results of
the separate analysis for the four Upper Wabash counties
were quite similar and therefore all observations were com—
bined for subsequent analysis. Several tests of the regres-
sions based on the combined county observations revealed no
significant difference in sale price attributable to county
location.

The separate analyses for Knox and Sullivan counties
were also in agreement although not to the degree found in
the Upper Wabash counties. Part of the difference appeared
attributable to differences in the importance of the farm
buildings in the two counties. When the regression
analysis for the combined counties was adjusted for this
factor, there was no significant difference in sale price
attributable to county location. Therefore, sales in these
two counties were analyzed jointly in the final regressions.

Bartholomew and Jackson counties differed substan-
tially in several important respects and it was not pos—
sible to aggregate the sales data for these counties. The
primary difficulty was that the average price per acre for
upland tracts in Jackson county was much lower than for

upland tracts in Bartholomew or floodplain tracts in

either county.

 

 

 

 

110

Upper Wabash Area

In the Upper Wabash, upland farms commanded a price
premium of slightly over $100 per acre relative to flood—
plain sales (Table 3). However, these prices are unad—
justed for differences in the composition of the average
upland and floodplain farm. For example, although the
floodplain sales are approximately A0 acres larger, on
the average, than upland sales, the major part of the
acreage difference is made up of pasture and woodlands.

The initial regression analysis indicated that these lands
are very low valued.

The average upland and floodplain tract had about
the same percentage of cropland graded A or B but the
floodplain sales contained more grade D and E cropland.
(There was little land in either case graded A or E and
therefore these acreages were combined with grades B and
D.) There were no strong relationships between floodplain
land graded D or E and the degree of flood risk on the
land. Apparently these grades represent primarily quality
differences rather than a downgrading of A, B or C land
because of the flood hazard.

The upland sales cluster around 1955—56, 1960—61
and 1965—66 because the original deed search was confined
to those years. The upland sales for the remaining years
were picked—up in the process of recording additional

floodplain sales in the years not originally searched.

 

 

 

111

TABLE 3.——Selected Characteristics of A06 Farm Transfers,
» Upper Wabash Area, 1952—1966.

 

 

 

 

 

 

 

 

 

 

Upland Sales Floodplain Sales
Item
Unit % Unit %
Number of Observation 3A6 60
Price per Farm $ 26,929 $25,572
Average Price/Acre $325.39 $221.03
Cropland Acres/Farm:
Grade AB 17.0 25.2 22.A 27.5
Grade C 32.6 A8.2 31.2 38.A
Grade DE 18.0 26.6 27.7 3A.l
TOTAL 67.6 77.5 81.3 63.8
Total Pasture Acres 8.7 10.0 2A.5 l9 2
Total Woodland Acres 8.5 9.7 17.A 13.6
Other Land Acres 2.A 2.8 A.3 3 A
TOTAL Land in 87 2 100.0 127 5 100.0
Farm
Value of Farm
Improvements $5,238 $A,300
Purchases for
Expansion 115 32.2 12 20.0
Land Contract Sales 51 lA.7 5 8.3
Special Warranty Sales 57 16.5 7 11.7
Year of Sale:
1952—5A 5 l.A 6 10.0
1955 56 16.2 5 8.3
1956 A9 lA.2 9 15.0
1957—59 2 0.6 A 6.7
1960 AA 12.7 2 3.3
1961 50 1A.5 3 5.0
1962—6A 7 2.0 13 21.7
1965 60 17.3 10 16.7
1966 73 21.1 8 13.3

 

 

   

 

 

 

112

Lower Wabash Area

 

Floodplain tracts sold in the Lower Wabash Area
averaged twice the size of the upland tracts, and in
addition contained a higher percentage of cropland (Table
A). However, the cropland grade distribution was nearly
the same for the twosales groups. Very little pasture was
found on the average floodplain farm. The low percentage
of pasture land is primarily a reflection of the fact
that only a few farms contained any pasture at all.
Similarly, the lower average appraised value of farm
improvements on the floodplain is partially a reflection

that many tracts sold were without buildings.

White Subarea

Separate tables are presented for Bartholomew and
Jackson counties in order to point out the difference in
the average upland and floodplain tracts in the two
counties. Several regression runs were made with the two
counties combined but the results were unsatisfactory and
indicative of conditions in neither county. Therefore,
these two counties in the White Subarea were analyzed
separately.

Bartholomew county is similar to the counties in
the other study areas in that the floodplain sales were,
on the average, larger than the upland farms sold (Table
5). However, 65 per cent of the floodplain cropland was

graded A or B, in contrast to the grading in the other

 

 

 

 

 

 

 

113

TABLE A.-—Selected Characteristics of 33A Farm Transfers,
Lower Wabash Area, 1962—1966.

 

 

 

 

 

 

 

Item Upland Sales Floodplain Sales
Unit Per Cent Unit Per Cent

Number of Observations 260 7A
Price per Farm $2A,l96 $39,752
Average Price $303.33 $255.65

per Acre
Cropland Acres per

Farm:
Grade ABC 15.0 26.8 31.0 22.5
Grade D 30.1 53.9 76.0 55.0
Grade E 10.8 19.3 31.1 22.5

TOTAL 55.9 66.8 138.1 82.5
Total Pasture Acres 12.8 15.3 A.3 2.6
Total Woodland Acres 9.2 11.0 1A.3 8.5
Other Land Acres 5.8 6.9 10.7 6.A

TOTAL Land in Farm 33.7 100.0 167.A 100.0
Value of Farm Improve-

ments $3,276 $2,676
Purchases for

Expansion 123 A7.3 A6 62.2
Land Contract Sales A8 18.5 5 6.8
Special Warranty Sales 36 13.8 11 1A.9
Sand Hill Location 63 2A.2 20 27.0
Year of Sale:

1962—63 10 3.8 ll 1A.9

196A 80 30.8 2A 32.A

1965 87 33.5 26 35.1

1966 83 21.9 13 17.6

 

 

 

 

11A

TABLE 5.——Selected Characteristics of 81 Farm Transfers,
Bartholomew County, White Subarea, 196A—1966.

 

Upland Sales Floodplain Sales

 

 

 

 

 

 

 

 

 

Item
Unit Per Cent Unit Per Cent

Number of Observations 65 16 ‘
Price per Farm $23,2A8 $35,815
Average Price per

Acre $3A8.83 $335-29
Cropland Acres per

Farm:
Grade AB 21.8 A1.5 53.8 65.0
Grade C 1A.2 27.1 16.0 19.3
Grade DE 16.5 31.A l3 0 15.7

TOTAL 52.5 72.A 82.8 73.1
Total Pasture Acres 6.2 8.6 5.A A.8
Total Woodland Acres 11.5 15.8 16.5 lA.6
Other Land Acres 2.3 3 P 8.5 7 5

TOTAL Land in Farm 72.5 100.0 113.2 100.0
Value of Farm

Improvements $3,569 $7.356
Purchase for Expansion 29 AA.6 7 “3.8
Land Contract Sales 2 3.1 1 6.2
Special Warranty Sales 7 10.8 3 18,8
Sandhill Location 1 1.5 2 12.5
Year of Sale:

1964 20 30.8 6 37.5

1965 28 A3.l 3 18.7

1966 17 26.2 7 A3.8

k

 

 

115

areas or on the upland farms of this county. The high
percentage of land graded A or B plus the fact that flood—
plain land sold at an average price per acre very close

to the price for the upland tracts indicate that the
floodplains are desirable farmlands despite the flood
risk.

The situation is even more pronounced in Jackson
County (Table 6). Here the floodplain lands commanded a
$110 price premium. The low price per acre for the
upland sales is primarily a reflection of the poor quality
of these tracts which contained relatively little of the
higher grades of cropland. The upland and floodplain
tracts both averaged 80 acres in size but 6A per cent of
the average floodplain tract Was in cropland compared to
A8 per cent for the uplands.

In Jackson county the upland farms were primarily
located either on the edge of the Norman Uplands which is
a dissected plateau of strong relief, or in an area of
Very poor drainage along the Muscatatuck River. By con~
trast, the floodplain consists primarily of a large

(71,600 acres including some land in Bartholomew County),

level area lying between the East Fork of the White River

and the Little White Creek. Most of this land is cleared

and intensively cultivated.
There were no land contracts recorded in Jackson

county and only three in Bartholomew during the period

 

 

116

ABLE 6.——Selected Characteristics of 118 Farm Transfers,
Jackson County, White Subarea, 196A—1966.

 

Upland Sales Floodplain Sales

 

 

 

 

Item

Unit Per Cent Unit Per Cent
imber of Observations 82 36
rice per Farm $17,583 $22,A79
verage Price per Acre $227.5A $338.32
ropland Acres per
Fram:
rade AB 1.7 3.6 11.7 18.A
rade C 13.2 27.8 33.5 52.8
a... DE _3_2_Ǥ_ .5010 is; 2.8.28
TOTAL A7.5 59.A 63.5 78.6
otal Pasture Acres 11.2 lA.0 6.8 8.A
otal Woodland Acres 17.3 21.6 6.1 7.5
ther Land Acres A.0 5.0 __£;fl __5;5
TOTAL Land in Farm 80.0 100.0 80.8 100.0
a188r8§e§868s $3,A72 $2,666
Archase for Expansion A1 50.0 21 58.3
oecial Warranty Sales 5 6.1 8 22.2
and Hill Location 7 8.3 8 22 2
ear of Sale:
196A 3A Al.5 1A 38.9
1965 20 211.11 15 41.7
1966 28 . ~ 2A.l 7 19.A

 

M..—

117

196A—66. Presumably, this is a reflection of local custom
regarding the recording of contracts rather than an
indication of the prevalence of the use of contracts in

transferring land.

 

 

CHAPTER V

REGRESSION ANALYSIS: GENERAL

Land Price Models

The analytical technique selected for the analysis
of the land value data was multiple regression analysis.
Regression analysis provides a means of determining how
one factor (the dependent variable) varies with changes
in other factors (the independent variables) that are
believed to significantly affect the dependent variable
under the circumstances being considered. In this case,
the dependent variable would be some measure of land prices
and the independent variables would be those quantifiable
factors——including measures of flood risk——that are believed
to affect market value of land. Regression techniques pro-
vide a means for ”holding constant” other factors that may
affect the market value of land while measuring the effect

of variables related to flood risk. Estimates of the

regression coefficients for variables included in the model
also provide a means for adjusting the sale price of ”non—

comparable" tracts of land for comparison purposes to the

extent the non—comparable elements are quantifiable and

included in the regression model.

118

 

 

119

Basically, two models of the land market were con—
structed. Both models postulate a given structure of the
land market but differ in the way in which the variables
were entered into the model. The general form of both

models is:

.X >

Y = f (xl,x2,.. x x.,. x /xm,... n

., i’ J .., k

Model I can be defined as a "total sale price"
model. Y, the dependent variable, is the total sale price
of the land and buildings composing a tract of land (a

"farm"). Total sale price of a floodplain farm is postu—

 

lated to be a function of (l) certain factors that are
common to all farm sales such as acres of cropland, acres
of woodland, or value of buildings, designated by the

(2) certain variables unique to flood-

variables X ,X.'

1,... 1’
plain sales such as acres of land inundated at selected
exceedence intervals, designated by the variables Xj"°‘xk’

and (3) certain other factors, variables X ...,Xn, repre—

m’
senting those influences on land values that have not been
quantified and therefore must be ”controlled" by either
holding the sample constant with respect to their influence
(i.e., by editing out sales known to be divergent with
respect to some unquantified factor) or by assuming that

the influence of the unquantified factors is randomly

distributed.

 

 

120

Model II——the "average sale price per acre" model——
postulates the same basic structure as Model I but in this
model the dependent variable Y is entered as average price
per acre for each farm (i.e., total sale price divided by
total acres sold). With Y defined on a per acre basis it is
also necessary to define the independent variables in a
consistent manner. This can be done by expressing the con—
tinuous variables as percentages. Thus, for example, Model
II would read as: Average sale price per acre is a function
of per cent of cropland on the tract, per cent of land
inundated at the 5-year interval, and so forth.

The multiple regression equation consistent with

either model is of the form:

 

Y = bO + lel + b2X2 + + biXi + + kak + E

The dependent variable, Y is either total sale price or
average sale price per acre. The Xi are the independent
variables that were quantifiable on at least a nominal
scale and that were believed to significantly affect land
values. Linear relationships between the dependent and the
independent variables were assumed (with one exception dis—
cussed in Chapter VI).

The error term, t, represents those factors (vari—
ables Xm,...,Xn) that could not be specified or quantified,
including some elements of "randomness" of human behavior

in a bargaining situation. We assume that these disturbance

 

 

121

terms have a normal distribution and that the observations
for this study are random samples from this population of
disturbance terms.

Ordinarily, one would expect Model I to explain a
very high percentage of the variation in sale price (have
a high R2) because the sale price of a tract of agricul-
tural land is basically a linear function of a price per
acre for crOpland times the acres of cropland, a price per
acre for other land use categories times the corresponding
acres plus the value of farm buildings, plus other adjust-
ments for location, soil fertility, and amenities. These
latter factors, to the extent they are thought to be impor—
tant value determinants, can be incorporated into the re—

gression equation by the use of discontinuous variables.

 

The difficulty with this model, however, is that the betas
for the discontinuous variables must be interpreted at
mean total sale price and it may be unrealistic to draw
inferences about the price effect of the discontinuous
factors for the very small or very large farms.l

This difficulty can be avoided by using average sale

price per acre for the dependent variable as in Model 11.

With this model, the beta for a discontinuous variable can

 

1For example, if a discontinuous variable for a
certain soil type were included, the beta would be inter—
preted as indicating that a farm with that soil type would
command a total sale price differential equivalent to the
beta value-—regardless of whether the farm contained 20
acres or 200 acres. Logically, however, the effect of
soil quality on total farm sale price should vary also
with size of farm.

 

 

122

be interpreted as the per—acre price differential for
sales with the attribute. The initial regressions were
run with Model I but Model 11 was adOpted for the final
analysis because it yielded more consistent estimates for
the discontinuous variables (which include some vari-
ables measuring flood risk). Model II yields lower R2
(approximately 0.60 vs. 0.80) but the relative standard
errors of the estimates were at least equivalent or

smaller than those for Model 1.

Variables

 

Listed below are the variables or data sources avail-
able for the analysis. All were used in the regressions
at one time or another but only the statistically sig—
nificant variables1 were included in the final estimating

equations.

Sale Price

 

Sale price, the dependent variable, is the total
sale price of the tract including farm buildings and
associated structures. For Model 11, sale price was con—
verted to an average price of land and buildings per acre.
If the price was stated in the deed or contract this price
was used. Otherwise sale price was computed from the

Federal Revenue Tax stamps on the deed. Since the tax is

 

1The criterion was to include in the equations only
those variables that reduced mean square error.

 

 

 

 

 

 

 

123

imposed at the rate of 55 cents per 500 dollars of sale
price, the price represented by the last 55 cents was
estimated by comparing the stated sale price with the
price computed from the tax stamps for those deeds con—
taining both. The average value of the last 55 cents for
these sales was $397. This was rounded to $400 and used
to adjust the calculated price of the remaining sales.
Sales subject to mortgages were included if the outstand—

ing balance of the mortgage could be determined.1

Date of Sale
The date the sale occurred was recorded and used as

a time or trend variable in the regression equations.

Total Acres Sold

The total acres sold was either stated in the deed
(most cases) or estimated from the legal description.
Checks on total acres in farm were available from plat

maps and assessment records.

Acres of Cropland by Grade

The best source of data on cropland on the tract
was found to be the farm appraisal worksheets used by
the county or township assessors. In Indiana, farm land

is appraised for tax purposes by determining the categories

 

1The amount of the outstanding balance of the mort-
gage was required because this amount is not subject to
the Federal transfer tax used in computing sale price.

 

 

124

of land use on the farm, assigning a letter grade from

A to E to the land by use category, and then assigning

a standard price by grade and category. It was possible
to interview several of the assessors during the data
collection stage and they indicated that they attempted
to evaluate as many factors—-topography, drainage, loca-
tion, and soil characteristics——as possible in assigning
grade categories. In particular, they evaluated land
according to its use-potential rather than its actual use.
Thus pasture capable of being cropped was appraised as
cropland and the pasture category was reserved for land
with limitations preventing higher use.

It is impossible to determine precisely all the
factors evaluated by the individual assessors or to judge
the accuracy of their work either relatively or against
some objective standard. However, the best judgment that
can be offered of the appraisal work is that the price
relationships between grades as determined by the regres—
sion analysis were consistent in all counties except
Jackson. In Jackson, part of the difficulties appear
attributable to the wide heterogeneity of land quality
in that county.

The acreage by land use categories (but not by
grade categories) was spot checked against Agricultural
Stabilization and Conservation records in several counties
and all sales were located on aerial photographs as a

rough check.

 

125

Acres of Pasture and Acres of
Woodland

These acreages were also determined from appraisal
worksheets. The grades for these land use categories
were not significant in any regressions and only total
pasture or wood acreage was included. The appraisers
rarely assigned a grade of C or better and tracts with
these higher grades were usually found to have residential
or commercial site potential and were eliminated from the

sample.

Acres of Other Land
On tracts sold for agricultural purposes acreage in
this category was small and included land in ditches,

roads, farmsteads and waste.

‘Type of Instrument

Sales were coded according to whether the sale was
by general warranty deed, special or commissioner's
warranty deed, or by land contract. Special warranty
deeds were primarily deeds given by court appointed
executors for estate settlement sales° Information from
land contracts was taken if the contract was recorded

but it could not be determined if the practice of record-

 

ing contracts was universal.

 

 

126

Expansion

A tract was presumed to be purchased for farm
expansion purposes if the grantee owned other land nearby,

as determined from plat maps and assessment records.

Per Cent of Mineral and
Oil Rights Sold

The per cent of mineral and oil rights transferred

 

to the grantee was determined and coded separately. Knox
and Sullivan Counties were the only counties in which an
appreciable number of transactions involved transfers of
less than 100 per cent of mineral and oil rights. Most

of the mineral (primarily coal) rights were alienated from
surface rights around the turn of the century when under—
ground coal mining first began. Since present production
is based on strip mining in limited areas of the counties,
the absence of mineral rights to the land should not be
expected to appreciably affect sale price and this vari—
able was not found significant in any regressions.

There are several producing oil pools in the two
counties but most transfers of agricultural land included
all oil and gas rights. In most cases where oil rights
were retained by the grantor the rights were to revenue
from producing wells. No significant difference in sale

price attributable to the oil rights was found.

 

 

 

127

Value of Farm Improvements

 

The appraised value of the farm residence and of
other farm buildings was taken from farm appraisal work-
sheets in county offices. The appraisals were made in
1961 but the sheets are updated for any material changes
in the buildings between_appraisals. Only values for the
buildings on the farm at the time the sale occurred were
recorded. The appraisal sheets contained considerable
information about the farm dwelling (type of construction,
number of rooms, etc.) and listed all farm buildings by
type. The appraiser also judged the relative quality of
all buildings (intermediate, low average, poor, etc.).

The following was recorded on the data collection schedule:

 

appraised value of the house, a letter equivalent of the
house grade, total value of all other farm buildings
existing at the time of sale, and a brief description of
the types of buildings.1

The average appraised value of the farm buildings
is relatively low (Tables 3 to 6, Chapter IV) but the
prices are consistent with the general quality of the
buildings, both as indicated by the appraisers‘ grading
and descriptions and from visual observation of farm—

steads in the study area. The farm dwellings, for the

 

lThe enumerators were instructed to describe in

detail any buildings on a tract that were not normally
associated with farming as a check in the editing process,

 

128

most part, are old and outbuildings are minimal. However,
one would expect some tendency for the appraisals to
understate actual building value (at least in the Lower
Wabash and White study areas) because of the time difference
between appraisal date and sale date. A number of investi—
gations were made of the relationship between sale price

of farm and appraised value of improvements in the initial
runs using Model I. Here the working hypothesis was that
one dollar of appraised value of buildings added a dollar
to sale price of the farm; i.e., that the beta for the
Value of Improvement variable was equal to one ($1.00).

The betas for Upper Wabash, Bartholomew and Jackson sales
generally were not significantly different from $1.00.
Betas for Sullivan County ranged between $2.00 and $2.AO,
indicating either that buildings in this county were under—
appraised or that buyers valued the buildings more highly
than in the other counties. The beta values for Knox

were about $1.00 for farms purchased by grantees not owning
other land in the vicinity but only $0.53 for farms pre—

sumably purchased for expansion purposes.

House Grade

House Grade was also included as an independent
variable for some runs of Model I. House Grade was gen—
erally positively related to sale value but the relation—
Ships were not strong enough to warrant inclusion of this

variable in the Model II runs.

 

 

 

129

Type of Road

The sales were coded according to whether they were
located on primary, improved (secondary), or unimproved
roads as a possible measure of some locational and access
values. The difficulty was that all the counties had
good road systems with very few unimproved roads remaining
and location on primary or arterial roads did not appear
to offer sufficient advantages over secondary roads to be
reflected in price. Consequently this variable was deleted
early in the investigation. A variable for distance from
town was considered but not developed because the spatial
distribution of the towns and villages in all counties was
good and no area could be considered isolated from the

major trading centers.

TOpography

A general classification of topography (river bottom
or level, rolling, or hilly) was taken from the appraisal
sheets and checked against topographical maps. However,
the effects of tOpography are also reflected in land use
and land grade categories and this variable was not sig—
nificant in the initial regression runs.

Some appraisers made notations on the appraisal
sheets about farm drainage condition ("wet," ”tiled,"
etc.). It appears that the notations were made in order
to justify quality grades assigned to the land and there—

fore a separate variable for drainage was not developed.

 

 

130

Other than notations on the worksheets, there were no

data available on drainage installations.

Township

A variable for township was included, primarily as
an alternative (to type of road and distance from town)
measure of location. Only in Knox and Sullivan were any
township variables found to be significant and ex BREE
explanations are generally available for the differences

in each case.

Land CapabilityfiUnit Groups

 

Soil maps of varying currency were available for
Miami (published in 1927), Cass (1939), Carroll (19AO),
Knox (193A) and Bartholomew (1936) Counties. It was not
feasible to develop variables for individual soil types
on each farm because of the large number of soils found
on a typical farm. Instead, the soils were grouped in
land resource area/land capability units (LRA/LCU) which
represent soils with similar management requirements and
productive characteristics. However, the LRA/LCU system
requires fairly recent soil maps based on current soil
definitions. Only Carroll county had a sufficiently
current soil map, but by keying soil descriptions in
Miami and Cass to the Carroll County maps, LRA/LCU group—
ings were developed for all three counties. This was not

possible for Knox or Bartholomew Counties however, and

 

131

sale tracts in these counties could not be related to LCU
groups.

Even with the LRA/LCU groupings, several LCU's were
usually found on a typical farm. It was not possible to
accurately measure the acreages in each group or precisely
delineate cropland areas, therefore the predominant LCU
was estimated by using a calibrated overlay on the soil
map after locating cropland boundaries with the aid of
aerial photographs.

There were seven LRA/LCU's suitable for cropland in
the Upper Wabash counties and each observation was coded
according to the predominant LCU on that tract. However,
the LRA/LCU variables were not significant in any run of
Model I or II. The probable explanation is that with one
exception (an LCU with an estimated corn yield of 80
bushels), the estimated corn yields under high level man—
agement practices were within 1A bushels (103 to 117
bushels per acre) over the LRA/LCU's. This indicates that
the cropland soils of the Upper Wabash are quite homo—

geneous with respect to crop yields.

Type of Farm

Since LCU groupings could not be developed for Knox
or Bartholomew because of the outdated soil maps, land in
these counties was grouped into "type of farm" categories
based primarily on soil characteristics. Generalized

soil maps were used for Sullivan and Jackson Counties to

 

 

 

 

 

132

develop similar categories. The type of farm designations
were, for Sullivan, Knox and Jackson: River Valley, Sand
Hills, and Uplands; and for Bartholomew: River Valley;
Sand Hill; Gravel Benchlands; Upland Corn, Wheat and
Livestock; and Upland General Farming. The Sand Hill area
is a pronounced region between the Wabash River and the
Upland areas of Knox and Sullivan Counties. The soils are
especially adapted for truck cropping and orchard produc—
tion (although the number of orchards is declining and no
sales of orchard land were included in the sample). A

similar area is found in Bartholomew and Jackson but of

 

much small extent.

Flood Risk

The following information was coded for each observa—
tion with land on the floodplains: The stream designation;
whether the tract had stream frontage; levee protection if
present, coded according to whether the levee was privately
constructed, COE repaired, or COE constructed; whether any
buildings were subject to flooding and the frequency; and
the acres of open land and acres of all land located in up
to five frequency zones.

Each tract sold was located on an aerial photograph
and on a topographic map on which the limits of floods of
selected exceedence frequencies had been superimposed.

Acreages were estimated by using calibrated overlays to

 

first estimate the proportion of open land and other land

 

 

l33

in each flood zone. These proportions were then converted
to acreages on the basis of known acreages, by land use
classes, in the farm. Open land (cropland plus pasture)
was measured instead of cropland because the two classes
could not be sufficiently distinguished in the aerial
photographs.

The limits of flooding along the major rivers and
streams were obtained from flood charts and/or flood pro-
files obtained from the Louisville COE office. Some upland
farms were located in small watersheds and subject to some
flooding from these streams. The extent of flooding on
these farms was estimated from data and maps in several PL

1 The PL 566 information was gen—

566 project work plans.
erally sketchy and usually presented only the limits of a
synthetic 50 year flood. Information for these streams
was coded if available but in some cases observations
were deleted for the lack of adequate information.

Table 7 presents the flood frequency data available
by streams in the sample. Land lying outside levees (i.e.,
between the levee and river) on tracts receiving levee pro-

tection was coded according to actual risk. Land inside

levees was considered subject to flooding from levee

 

1In order to simplify terminology, we will refer to
these small upland streams as "SCS streams," indicating

that they are the type of streams over which the SOS
normally has jurisdiction for development work. This
terminology is not meant to imply that all the small
watershed streams in this sample are currently under con—
sideration for PL 566 projects.

 

 

134

TABLE 7.——Flood Frequency Zones by Streams, Wabash Land
Value Study.

 

Recurrance Interval (years)

 

 

 

Stream

Zone 1 Zone 2 Zone 3 Zone A Zone 5
Wabash l 2 5 25 100
White ' 1 2 5 25 100
West Fork White 1 - 5 25 100
East Fork White 1 — 5 10—25 100
Flatrock — 1.7 7.7 — 100+
Driftwood 0.5 — — 20 100+
Clifty Creeka l — 5 10—25 100
SCS Streams - — - 50 —

 

aThe only sales on Clifty Creek were in the backwater
area of the East Fork of the White. The recurrance inter-
vals are for the White.

Source: From information supplied by the Louisville
District, U.S. Army Corps of Engineers.

 

135

overtopping from floods with a recurrance interval of 20

to 25 years and coded in zones A and 5.

Presentation and Interpretation
of Regression Results

 

 

In the following three chapters, the results of the
regression analysis of farm sale data in each of the three
study areas are presented. In Chapter IX the results of
the land value analysis are compared with the Corps of
Engineers estimates of expected average annual damage.

In reporting the results of the regression analysis
in each of the study areas, the estimating equation used
can be read from the appropriate table accompanying the
discussion (see, for example, Table 8, Chapter VI). The
dependent variable in each equation is average sale price
per acre (Y). The independent variables (the Xi's) can be
read from the left—hand column of each table (variable
names are assigned to each variable instead of variable
numbers to facilitate reading the tables).

All of the regression coefficients may be read
directly as dollars. The variables for cropland by grade
are continuous variables and were entered into the regres—
sion as decimal fractions of all land in the tract. "Log
Acres" and improvements by size of farm are also contin—
uous variables. The remaining variables are discontinuous
variables that take the value of one if the observation

possesses the attribute and take the value of zero

 

 

 

136

otherwise. The regression coefficients for the discon-
tinuous variables may be interpreted as the difference
in average sale price per acre for observations possess—
ing the attribute relative to the average price per acre
for observations without the attribute. For example, a
variable for ”Year of Sale, 1966" would take on a value
of one if the observation was transferred in 1966 and the
remaining year—of-sale variables would assume values of
zero. The coefficients for year—of sale should be inter-
preted as the difference in average price per acre for
sales in the year indicated over the average price per
acre in the base year (which varies in each study area).
The interpretation of the other discontinuous variables
should be self—explanatory from their names except that it
might be noted that the coefficients for "Contract” and
"Special" variables are relative to sales transferred by
general warranty deeds.

The variable ”Log Acres" is defined as the total
acres of land sold for each observation and is entered
in logarithms to the base 10.

The use of average price per acre as the dependent
variable in Model II creates special problems in generat—
ing consistent variables for value of buildings. The
building variable cannot be entered simply as average
value per acre because a poor set of buildings on a

small farm would average out to the same per acre value

 

 

 

137

as a good set of buildings on a large farm. After experi-
menting with several attempts to subtract out building
value from sale price, the observations for each study
area were divided into three approximately equal groups
on the basis of total acres in farm. The size categories
chosen were less than 50 acres, 50 to 120 acres, and more
than 120 acres. The value of buildings, per acre, was
then calculated with each observation falling into only
one of the three categories. This procedure yields inde-
pendent variables that are consistent with the dependent
variable while minimizing the effect of wide variance in
total acres on building value per acre. When defined in
this manner the independent variables for building value
are expressed in the same units (dollars per acre) as the
dependent variable. Therefore, the regression coefficients
for these variables should be interpreted as the relative
contribution of an additional dollar of appraised value
of buildings, per acre, to total sale price of the land
and buildings, per acre. Thus a coefficient of 1.0 would
indicate that $1.00 of appraised value of buildings per
acre contributed $1.00 to sale price. Similarly a co-
efficient of less than (greater than) 1.0 indicates that
$1.00 of per—acre appraised value added less than (more
than) $1.00 to per—acre sale price.

A three—step procedure was used in the analysis of

sale data for each of the three study areas. First, a

 

 

 

138

general regression analysis was made of all observations
(both upland and floodplain) for each study area in order
to determine the general nature of the value relationships
and to test if the average sale price of the floodplain
tracts differed significantly from the upland tracts. An
analysis of specific flood risk—land value relationships
is limited in this type of analysis, however, because it is
possible that some of the influence of flood risk may be
picked up by other independent variables (such as propor-
tion of cropland by grade) which are more strongly cor—
related with sale value in the combined regression analysis.
Therefore, in the second step of the analysis, the

upland sales were analyzed separately. The purpose of

 

this step was to estimate the influence on sale value of
those variables common to both upland and floodplain
tracts of land from observations that were free of flood
risk. The regression coefficients obtained therefore
should be free of any discounting for flood risk.

In the third stage of the regression analysis in
each study area, the sale price of each floodplain obser—
vation was estimated on the basis of the regression coef—
ficients estimated from the upland observations in the
second step. The assumption in this process is that in
the absence of flood risk those factors common to both
upland and floodplain sales would influence price in the

same way. The residual differences between the actual

 

 

L.__A.‘

 

 

 

 

139

per—acre sale price of the floodplain observations and
the estimated price, given the above assumption, there—
fore represents the component of floodplain land values
attributable to flood risk. These residual differences
thus provided the basis for additional investigations

into flood risk—land value relationships.

 

 

 

 

 

CHAPTER VI

REGRESSION ANALYSIS: UPPER WABASH AREA

 

General
The final regression analysis for the four counties in
the Upper Wabash area was based on A06 observations of farm
sales for the years 1952 through 1966. There were 60 sales
of farm tracts containing floodplain land but 15 of these
were subject to flooding from SCS streams. The remaining

A5 observations represent essentially a complete accounting.

 

of all registered transactions of floodplain land that can
be considered bona—fide sales of agricultural lands and for
which adequate information was available for anlaysis.
Conceivably, the number of observations could be increased
if the circumstances surrounding some sales between related ,

persons could be determined but most of these transactions

obviously were made at less than full market value.

Floodplain Sales 1952-1966

 

As a part of the search for eligible sales, a listing
was made of all ownership tracts in the four counties that
included land on the floodplain. The listing was compiled
from county plat maps and included all tracts of 20 acres

or more. Tracts in Peru township, Miami County and Eel

1A0

 

 

1A1

township, Cass County as well as tracts in the immediate
vicinity of Wabash in Wabash County and Delphi in Carroll
County were excluded as urban property. An effort was
made to consolidate all tracts held by the same person.
There were 307 identifiable ownership units. This list
was checked against real property transfer records in the
auditors office of each county beginning with the records
for 1951-52. Of the 307 total ownership units, 186 were
held in the same name the entire period. There were 62
transfers that, prima facie, represented bona-fide sales
(on the basis of information in the auditors office) but 17
of these were found to be subject to extra—market influences.
There were 73 transfers involving tracts that passed by
descent, gift, divorce settlements, or litigation and tracts
sub-divided or commercially developed.l

Although this enumeration may be subject to some errors
(in defining floodplain tracts, consolidating ownerships
units held in one name, etc.) it is indicative of approxi—

mate turn—over rates. It also indicates that a major dif—

ficulty with the land value approach in areas of narrow
floodplains may be that of obtaining an adequate number of

representative farm sales for analytical purposes.

 

lThese categories sum to 321 because there were some
multiple transfers over the period and some tract lelSlonS
or consolidations.

 

 

1A2

Combined Analysis, All Sales

The results of the regression analysis of all sale
.bservations in the Upper Wabash area are presented in
Table 8. The mean sale price per acre for all observations
Lincluding floodplain sales) was $309.97 with a standard
leviation of $151.A2. The estimating equation had an R2
3f 0.67 and the standard error of estimate was $89.A2.

"Expansion" was a discontinuous variable that took on
the value of one if the farm was purchased for farm expan—
sion purposes. These sales commanded an average of $18 per
acre over sales to grantees not owning other land in the
immediate vicinity. The positive coefficient for expansion
is in accord with economic logic.

"Log Acres" is a measure of the effect of total size
(of the tract on the average price per acre. The negative
coefficient indicates that as total size increases, price
er acre decreases. This occurs in part because building
value is included in the total price per acre and one would
expect value of buildings to exert a greater influence on
average price per acre for small farms than large farms.
When sale prices were adjusted for buildings the magnitude
of this coefficient was greatly reduced but still negative
$n sign. The variable was entered in logarithmic form after
Xperimentation with various first and second order forms of
:he variable indicated the function was approximately of

1 logarithmic form.

 

 

 

 

1A3

CABLE 8.——Estimated Regression Coefficients, A06 Observa-
tions Upper Wabash Area, 1952—1966.

 

Item Regression Standard Significance
Coefficient Error Level

 

Number of Observations A06

 

 

 

32 0.68
31' 309.97 89.43‘31
Constant 286.93 55.00
Independent Variables:
Expansion 18.32 9.98 0.06A
Log Acres —1AA.95 2A.3A 0.0005
Proportion Crop AB 275.23 23.63 0.0005
Proportion Crop C 181.15 2A.71 0.0005
Proportion Crop DE 9A.AA 27.82 0.001
Contract ' 90.15 13.68 0.0005
Special 30.87 12.80 0.016
Improvements:
Small Farm 0.91 0.09 0.0005
Medium Farm 0.99 0.12 0.0005
Large Farm 1.A9 0.2A 0.0005
Year of Sale:
1955 22.62 30.75 O.A69
1956 30.91 30.62 0 31A
1957—59 -5.35 A5-75 0.873
1960 AA.72 31.51 0.153
1961 26.50 31 0A 0.398
1962—6A £15.20 33 7A 0 178
1965 112 97 30 63 o 0005
1966 15A 82 30 0A 0 0005
Floodrisk:
Frontage ~15.l3 18.A8 0.Al9
No Frontage —15.08 29.73 0.618

 

 

aStandard error of estimate.

 

1AA

The variables for proportion of cropland by grades AB,
C and DE were highly significant and of the correct rela—
tive magnitudes, indicating that the grading system was
consistent with the land market's evaluation of tract
value. Pasture and woodland coefficients were not signifi—
cantly different from zero.1

Several considerations suggest that farms sold by land
contracts should command a price premium over farms sold
by general warranty deeds with conventional financing.

Sales by special warranty, on the other hand, are sold

 

under conditions approaching a public auction and would be
expected to command about the same price as general warranty
sales, or perhaps at slightly lower prices because of the
element of timeliness involved in these involuntary sales.
The positive sign and magnitude of the coefficient for
special sales was not expected;2 nor was the magnitude of
the coefficient for the contract sales. The magnitude of
these coefficients may indicate, then, that some of the
general warranty sales were made at less than full market

value. In view of this possibility, the editing procedure

 

1The coefficients for the cropland grade variables
can be interpreted as percentages by adjusting the decimal
point; e.g., a 10 per cent increase in the per cent of
Cropland grade A or B in the farm would, on the average,
increase average price per acre by $27.52, ceteris paribus.

 

2However, the positive coefficient may indicate only
that the commissioners handling the sales were especially
diligent in obtaining the best price.

 

 

 

1A5

ior all observations was reviewed but there were no obvious
examples of sales that should be excluded.1
The coefficients for value of farm improvements by
size of farm indicate that one dollar of appraised value
of building per acre on the large farms add more to average
sale price per acre ($1.A9) than one dollar of appraised
value per acre on the small farms ($0.91). Several hypo-
theses can be offered in explanation for this difference.

For example, to the extent the smaller farms are more

likely to be purchased for expansion purposes, the farm

 

buildings on the tract may be redundant and therefore not
valued highly by the purchaser. Similarly, on the large
farms the purchase price may tend to be high relative to
the appraised value of both land and buildings. However,
explanations for the differences in coefficients by size of
farm were not pursued because (1) both the coefficients
for small and large farms are within one standard error
confidence limits of the coefficient for medium—sized farms
and (2) the interpretation of these coefficients are not
critical to the purposes of this study.

The coefficients for year of sale measure the differ-

ence in average price per acre attributable to year of sale

 

relative to the base years 1952-54- The COEffiCientS

 

1There were three sales (two by land contract and
one by special warranty) that were made at relatively high
prices per acre ($750—$900). When these were eliminated
the coefficients were $77.51 and $29.29 for contracts and
Special warranty sales, respectively.

 

 

1A6

indicate a tendency for prices to increase over time but
the significant increases occurred only in 1965 and 1966.
There were only six sales recorded between 1957-59 and
this estimate is particularly poor. The year of sale
variables are discussed further below.

In this particular equation the variables for flood
risk are for observations on the Wabash only. There were
3A observations with frontage and 11 observations subject
to flooding but without direct Wabash frontage. The coef—
ficients have a negative sign but are not statistically

significant at the 10 per cent level. Other variables for

 

flood risk tried (with slightly different sets of other

 

independent variables) included variables for flooding from
the Wabash River and from SCS streams. In no equation was
the SOS variable significantly different from zero. The
variable for Wabash flooding was significant and negative
(around $—21 and at the 0.03 significance level). The
relationship between flood risk and land prices is examined

in more detail below.

Upland and Estimated Floodplain Sales
The analysis of the combined observations includes
60 floodplain sales and therefore reflects, in part, the
influence of flood risk on the sale price of the floodplain
farms. Depending on how the flood risk is defined, the
analysis indicates that there is a negative land price

component attributable to flood risk but the information

1A7

provided by a flood risk variable in the over—all analysis
is limited. Therefore, the sample of Upper Wabash sales
was broken into two classes: upland and floodplain sales.
The 3A6 upland observations represent sales of land
free of flood risk and sold for an average price of $325
per acre. These sales were analyzed separately and the
regression coefficients obtained (Table 9) therefore repre—
sent the influence ofifle independent variables on the price

of land free of any floodrisk.

Year of Sale Adjustments

 

The precision of the estimates of the influence of
year of sale on the average price per acre for the upland
observations alone was improved considerably relative to
the analysis of all sales combined, although the standard
error still exceeded the estimated beta value for the years
1957—59. However, this estimating equation yielded only
single estimates for the years 1957—59 and 1962-6A. Since
the number of floodplains sales that occurred during these
years were relatively important1 it was considered desir—
able to develop individual estimates for each of these
years.

The individual yearly estimates were developed by
converting the beta values obtained from the upland regres_

sion for each of the years 1955—56, 1960-61, and 1965—66

 

1There were two floodplain sales each in 1957 and
959, four in 1962, one in 1963, and eight in 196A.

 

 

 

 

 

1A8

TABLE 9.——Estimated Upland Regression Coefficients and Mean Values for Upland and
Floodplain Observations, Upper Wabash Area. 1952-1966.

 

Upland Regression Means
Item _____________.___

 

Standard Significance
Coefficient Er e1

Unit Upland Floodplain
ror

 

No. of Observa—

 

tions 3A6 60
R2 . 0.66
Y 325.39 91.99a $ 325.39 221.03
Constant 287.07 71.10 0.005 '
Independent
——V§?IEBI€§:
Expansion 16.01 10.90 0.129 % 33.2 20.0
Log Acres ~160.56 28.66 0.0005 — 1.87 1.99
Proportion Crop AB 273.A3 26.28 0.0005 % 20.2 15.1
Proportion Crop C 178.30 27.12 0.0005 % 37.A 2A.A
Proportion Crop DE 85.AA 30.80 0.006 % 20.0 22.2
Contract 95.93 1A.88 0.0005 % 1A.7 8.3
Special A7.8A 1A.07 0.001 % 16.5 11.7
Improvements:
Small Farm 0.87 0.09 0.0005 $ 2A.85 10.82
Medium Farm 0.99 0.13 0.0005 $ 31.Al 21.08
Large Farm 1.60 0.27 0.0005 $ 12.30 13.98

Year of Sale

1955 52.65 03.77 0.228 56 5
1956 50.66 00.27 0.252 09 9
1957-59 39.70 78.73 0.620 2 0
1960 70.15 00.11 0.090 00 2
1961 57.02 03.91 0.192 50 3
1962—60 97.09 50.73 0.072 7 12
1965 100.12 03.88 0.002 60 10
1966 180.36 03.03 0.000; 73 8

 

a
Standard error of estimate.

1A9

to index equivalents and plotting these computed indices
against the actual land price indices for the State of
Indiana for the same years (Figure 6). The indices for
the remaining years (1957, 1959, 1962, 1963 and 196A) were
then estimated by straight—line interpolation and converted
back to estimated price increments equivalent to the values
derived from the regression.1 This procedure yielded an
estimate of the influence of year—of—sale on sale price for
each year of the period 1955—66. These yearly estimates
were then used in the analysis of floodplain sales (below).
Data to estimate the trend in value for the six observations
sold between 1952 and 1955 were inadequate for estimation
purposes.

It may also be noted from Figure 6 that the agreement
between the state—wide index and the computed index for the
Upper Wabash sales is very close for the beginning and
ending years of the period examined. Real estate sales in
the four county areas during 1960 and 1961 apparently com—
manded lower relative prices than the state averages but
the computed indices did reflect the state—wide drop in
prices for 1961. It was concluded therefore that the esti-

mated price increments attributable to year of sale in the

 

1The estimates were: $55.39 for 1957, $69.72 for
$333, $79.27 for 1962, $98.38 for 1963, and $117.09 for

 

15O

 

 

ommaammma

n
mmOHUQH Umpwermm one nwcmach .

nwms< swans: some:

 

 

 
   
 
  

mm
OHUCH mafihm pcmq Hm3p0<ll.m madman
mmma mm .
no mm mm H@ ow mm mm mm mm mmm
_ H

. . 0. x om
T
E
U
0..
.uow .d
a
I.
O
wwn< swarm: some: to a
x09; emcmsflmm(¥\ m.
\\ 4.. OOH m
« msmflpsH flu
xoUQH Hmdpo< %
m
.22 W
8

..oma

 

 

 

 

151

regression analysis were reasonably related to average

state—wide price changes over the study period.1

Price and Flood Risk Adjustments

The average price per acre of the 60 floodplain
sales was $221.03. However, part of this difference in
price per acre represents quality differences in the
average upland and floodplain tract. For example, only
61 per cent of the floodplain tract was in cropland, com-
pared to 77 per cent for the upland sales, and relatively 1
less of this cropland was graded A, B or C. Similarly,
building values on the small and medium farms were much
lower. If corrections are made for these differences (by
assigning estimated upland values to the means for flood—
plain sales) the estimated price per acre is $231 (Table
10).2 The assumption implicit in this procedure is that
in the absence of flood risk, the factors common to both
upland and floodplain sales would influence sale price in
the same way. The difference in this estimated price
($231) and the actual sale price ($221) therefore represents
the component of floodplain land values attributable to

flood risk (see part A, Table 10).

 

1A similar check was made for the year—of—sale vari-
ables for the Lower Wabash area analysis. Good agreement
between the state index and the estimated area index was
found.

2This figure can be derived by converting the variable
means for the floodplain observations in Table 9 to decimal
equivalents (except the means for value of improvements) and
mgltiplying by the corresponding coefficients for each vari-
a le.

 

 

 

lf52

PABLE 10.—-Average Land Prices and Flood Risk Calculations, Upper Wabash Area.

 

1. Average Price per Acre Analysis

 

L. Price per acre of average farm, 3A6 upland sales $325.39
2. Price per acre of average farm, 60 floodplain sales 221.03
3. Unadjusted difference -10A.36
A. Estimated price of floodplain sales at upland valuesa 231.13
5. Difference attributable to flood risk (A—2) _ -10.lO

 

3. Per Acre Residual Analysis

 

 

 

Cate or Mean Standard t
g -y Residual Error Value Significance
50 Floodplain sales after 1950 —l6.0A 10.23 1.57 0.10
15 SCS streams +11.92 20.63 < 1 - N.S.
39 Wabash River observations —26.79 11.9“ 2.3A 0.025
a .
Estimated as:
= = A A w . . . A v - - - + A 7
Y 80 + 8101 + + BiAi + Bn‘ n

A

(here I = estimated average price per acre, all floodplain obsgrvations. The
3i were obtained from the upland regression analysis and the X, were set equal
:0 the means of the floodplain observations. ‘

bA residual (R1) was calculated for each floodplain obse-vation as:
Ri = 11 — Y1
= Yi — (so + 81x11 + - ' ' + Bnhni)

’here Y1 = average price per acre, ith floodplain observation.

 

 

 

 

 

 

 

153

The $10 difference in actual floodplain sales and the
estimated price in the absence of flood risk is based on
average price per acre for all floodplain sales. In order
to facilitate further analysis, the ”flood-free" price was
estimated for each floodplain observation and subtracted
from the actual price in order to estimate the flood risk
component for each sale. The analysis based on these resi-
duals is indicated in part B of Table 10. Year of sale
data were not available for the six sales occurring before
1955 and these sales were eliminated from the analysis.

The variable for flood-risk for tracts located on SCS
streams was not significant in any analysis of combined
upland and floodplain sales. In this residual analysis the
residual for the 15 SOS streams was found to average a
positive $12. This value was not significantly different
from zero and indicates that these farms were valued by
farm buyers about the same as surrounding upland farms.

After deleting the 6 sales that occurred before 1955
and the 15 sales on SCS streams, there were 39 sales repre—
senting tracts subject to flooding for the Wabash River.

The sales commanded an average of $26.80 less than their
estimated value under flood-free conditions. This differ—
ence was statistically smaller than zero at better than the
three per cent level.

The difference of $26.80 represents the average dis-

count across all flood risk zones and includes some

 

 

 

 

 

15A

non—floodplain land, since a number of the farms overlapped
the floodplain and uplands (55 per cent of all open land
on the average floodplain farm was on the floodplain; of
this 3A per cent was subject to flooding once in five years
or more frequently). In addition, the sales cover a span
of 11 years during which there has been a substantial
reduction in flood risk due to reservoir construction.
Therefore, additional investigations were made using the
residual values. The results may be indicative of general
relationships but the statistical reliability is very low
because of the large standard errors associated with the
residuals used as the dependent variable.

The best estimating equation from the residual
analysis was:

FP Residual = 09.16 — 68.00 (HG RISK) — 00.81 (OTHR FP)
(36.23) (A3.78)

- 3.68 (YEAR)
(2.77)

This equation postulates that the residual component
is a function of per cent of open land in the high risk
zones (recurrance interval of l to 5 years), the per cent
of open land on the floodplain but outside the high risk
zones, and year of sale. The coefficient for HG RISK is
significant at the 0.067 level, the remaining coefficients
2

are not significant at the 0.10 level. The R for this

equation was 0.1A and the standard error was $69.06.

 

 

 

155

The signs and the relative magnitude of the two
‘lood risk variables are consistent and indicate, for
xample, that a 10 per cent increase in land in the high
isk zone would increase (make larger negatively) the
‘loodplain location residual by about $7.

One of the reasons for choosing the Upper Wabash area
’or study was to determine if the effect of reservoir con-
truction in the area was reflected in land values. During
he period examined the Salamonie Reservoir was completed,
he Mississinewa Reservoir reached operational state and
onstruction on the Huntington Reservoir was begun. Expec—
ations were that this construction activity, and resulting
ncrease in protection levels, would be reflected in sale
'alue. However, the negative coefficient for year of sale
ndicates that this has not occurred.

The lack of significance for the year of sale variable
ay indicate either than there are too few observations in
ny year (and over all years) for a strong trend line to
e developed, that capitalization has not yet occurred, or
hat capitalization occurred before 1955 in anticipation of

rotection.l The latter is at least a possibility because

 

1An additional possible explanation is that farmers

n Wabash and Miami Counties who were displaced by site
cquisition for the Salamonie and Mississinewa Reservoirs

n the early sixties bid—up the price of other farmland in

he study area in seeking new farmsteads. This might explain
be greater rise in the computed land price index for the
rea, relative to the State of Indiana, between 1960—61 and
965—66. If this in fact occurred, any rise in floodplain
and values attributable to the reduction in flood risk may

 

 

 

 

156

authorization for investigation of flood control for the
area was approved in 19A9; the report recommending the
reservoir construction was submitted to the Committee of
Public Works in June, 1956;1 and the Corps was actively
buying land in Wabash and Miami County in the early years
of this study.

In this study relatively little importance is attached
to the fact that the per-acre sale value of farms bordering
the upland streams was not significantly different from
other upland farms. The study was not designed to investi—
gate the effects of flooding from these streams and the
available data are sketchy. Examination of topographic
maps and aerial photographs indicate that nearly all the
land immediately adjacent to the streams (and therefore
presumably in the high risk zones) have been left in wood-
lands. To the extent conditions in the Upper Wabash are
typical of other upland streams this does indicate, however,
that the land value check would have little applicability
under these conditions.

The implications of the analysis of floodplain land
values are discussed further in Chapter IX; where the
results of the analysis of all three study areas are related

to Corps damage estimates.

 

be obscurred by the general rise in land values in the area
which was also related (indirectly) to the reduction in
flood risk.

lU.S. Congress, House, Committee on Public Works, Head-
water Reservoirsi Wabash River, Indiana [Issued as H. Doc.
535], 8Ath Cong. 2nd sess., 1956.

 

 

 

 

CHAPTER VII

REGRESSION ANALYSIS: LOWER

WABASH AREA

Basically the same analysis was performed for Knox
and Sullivan County as in the four county Upper Wabash
area. There were 33A observations covering farm sales
between the years 1962 and 1966. Of these, 7A were flood—
plain sales. The results of the combined analysis are
presented in Table 11. Variable definitions are the same
except that the value of improvements, by size of farm,
was entered separately for Knox and Sullivan since the pre—
liminary analysis indicated that buildings had a relatively
more important influence on total farm sale price in Sulli—
van County. The coefficient for value of building on large
farms in Knox County was not significant in any equations.
"Sand Hill" is a location variable for farms located on
the sandy soil areas of the county and had a positive value,
as expected. The township variables are additional mea-
sures of locational attributes. The signs for these vari—

ables are in accord with expected relative price differ—

ences over the two counties.

 

 

 

 

 

 

 

TABLE ll.—-Estimated Regression Coefficients,

158

33A Observa~

tions, Lower Wabash Area, 1962—1966.

 

Regression Standard Significance

 

 

 

Item Coefficient Error Level
lumber of Observations 33A
{2 0.61
.7 292.76 97.8381
lonstant 37.99 32.78
gggependent Variables:
Proportion Crop ABC 321.99 27.93 0.0005
Proportion Crop D 223.90 26.23 0.0005
Proportion Crop E 201.52 31.18 0.0005
Contract 32.A9 15.A5 0.03A
Special 53.68 16.17 0.0001
Improvements, Sullivan:
Small Farm 1.02 0.16 0.0005
Medium Farm 0.85 0-30 0.005
Large Farm 1.99 0.60 0.001
Improvements, Knox:
Small Farm l.Al 0.16 0.0005
Medium Farm 0.89 0.25 0.001
and Hill 38.02 1A.63 0.010
ownships:
Fairbanks —120.97 21.03 0.0005
Gill —116.38 23.A9 0.0005
Hadden — 59.83 22.86 0.009
Hamilton ~ 8A.15 27.01 0.002
Turman —100.A8 25.93 0.0005
Busseron —l62.87 26.30 0.0005
Steen — A2.6A 32.17 0.183
Vigo — 99.00 28.76 0.001
Washington — 51.03 20.99 0.015
Widner — 6A.82 28.00 0.020
ear of Sale:
196A 63.52 2A.66 0.010
1965 89.70 20.66 0.0005
1966 138.88 2A.83 0.0005
- 7A.A3 20.07 0.0005

loodplain, no levee

 

 

aStandard error of

estimate.

 

 

 

159

Again, the coefficients for cropland by grade are
consistent in relative magnitude. The coefficient for loca-
tion on unprotected floodplain land indicates a significant
negative value component associated with these sales. The
coefficient for location on protected floodplains, on the

other hand, did not differ significantly from zero.

Upland and Estimated Floodplain Sales

As in the Upper Wabash analysis, a separate analysis
of the 260 upland sales in Knox and Sullivan Counties was
made in order to derive estimated coefficients free of the
influence of flood risk for those variables common to both
upland and floodplain sales. The regression results and
average values for the variables on upland and floodplain
farms are presented in Table 12. Again, the coefficients
for cropland are consistent, although the value difference
between cropland graded D and E is small. The coefficients
for township are relatively large, and negative, because
the township eliminated from the regression to avoid matrix
singularity was Curry Township in Sullivan County. Land
values in Curry Township were, on the average, the highest
for any township in Sullivan County and were about the

1

same as in five of the 10 townships in Knox County. The

coefficients for the township variables indicate the

 

1In Table 12, Fairbanks through Turman Townships are.
in Sullivan; the remainder are in Knox. There wered6.Sulli-
van townships in the study (Curry plus those llste in
Table 12) and 10 Knox County townships.

 

16(1

FABLE 12.—-Estimated Upland Regression Coefficients and Mean Values for Upland and
Floodplain Observations, Lower Wabash Area, 1962—1966.

 

 

 

Mean
Item St d d Si ifi —————_——__—_—‘—_‘—
Coefficient :3 ar gneveiance Unit Upland Floodplain
lumber of Observations 260 7A
:2 0.58
7 7 303.33 101.32a_ $ 303.33 255.65
Bonstant 38.20 A2.80
Independent Variables:
Proportion Crop ABC 31A.88 33.20 0.0005 % 17.0 21.3
Proportion Crop D 207.A7 29.85 0.0005 % 36.8 AA A
Proportion Crop E 199.20 35.90 0.0005 % 15.A
Contract 38.07 17.A8 0.030 % 18.5 6.8
Special 5A.99 19.52 0.005 % .13.8 10.9
Improvements, Sullivan:
Small Farm 1 0A 0.17 0.0005 $ 11.99 --
Medium Farm 0.90 0.32 0.006 $ 7.87 0.86
Large Farm 1.98 0.68 0.00A $ 2.51 2.69
mprovements, Knox:
Small F 1.23 0.19 0.0005 $ 9.A8 A.A9
Medium Fzgm' 0.86 0.29 0.003 $ 10.73 3.81
and Hill 01.96 17.32 0.016 % 20.2 27.0
ear of Sale:
.8 32.0
1 A 66.03 35.17 0.060 % 30
1385 35.61 35 17 0.016 % 33.5 35.1
1955 100.76 35.05 0.0005 % 31.9 17.6
awnships:
0 6
F -l20.5A 20.63 0 0005 3
Giiibanks —120.99 29.90 0 0005 12 1;
Hadden — 63.57 25.88 0 015 $9 0
Hamilton — 82.30 28.98 0.005 20 2
Turman -105.96 28 9A 0.0085 12 9
Busseron -l30-90 35-61 0‘00 5 7 A
Steen — A8.22 A; 81 g 582 6 12
V1 0 _ 52. 0 . .
Waghington — 83-16 22‘96 0'06; 15 i
Widner - 08.12 30.50 0-11

 

 

aStandard error of estimate. ‘

 

 

161

difference, ceteris paribus, in average price per acre
relative to average prices in Curry Township (and in five
of the Knox County townships which were not significantly
different from Curry).

The difference in the simple average sale price per
acre for the upland and floodplain farms was $A7.68
($303.33—$255.65). However, the floodplain sales had more
cropland per tract and therefore had higher percentages of

cropland in each grade. The upland sales, on the other

 

hand, had much higher building values per acre and more
sales in 1966. If the floodplain sales are adjusted for
these differences, the estimated sale price would have
been $285.50 per acre. The difference between this price
and actual price provides an estimate of an average dis—
counting for flood risk of $29.85 per acre (Table 13).

The 7A floodplain observations include A tracts sub—
ject to flooding by upland streams. When these four
observations are removed, the residual difference between
the estimated and actual sale price for the floodplain
sales was $31.23. This was significantly smaller than
zero at about the one per cent level. However, the 70
observations were composed of 38 observations receiving
protection from levees and 32 observations outside or with-
out levee protection. When residuals were calculated
separately for these groups, the observations with protec—

tion had a positive residual of about $A.00. This residual

 

 

1
_
v.
.1

162

.oH wanes nmopocfioom comm

 

 

 

 

 

 

 

oa.o wo.a :s.ma sa.mm| owmpsoam 02 NH
mooo.o ma.s mm.oa :m.saal owmpcosm ma
mooo.o mm.: Hm.:H mw.msl moanm oo>oa 02 mm
.m.z H v Hm.mH mw.m + moanm pouoopoad oo>oq mm
Ho.o om.m Ho.mH mm.am| Amom mmofiv moamm Qflwadpooam on
cosmoflaflsmflm oSHm> soshm awmsoflmom msowopwo
p esmecmpm one:
mflmsawc< Hospflmom oboe mom .m
mw.mm- Amuev ewes eooae on oaoeesoanooa oooonoeoao .m
om.mwm nmosHm> oceans pm moamm Qflmadpooam wo conga Umpwsflpmm .:
ww.s:| oooonoecfle soundness: .m
mm.mmm moamm ofldeUooam as «sand omwso>m do when mod monsm .m
mm.momw moadm oceans 0mm .Eswe omwso>m do meow sod ooanm .H
mflmzfimc< oso< sod moasm owmao>< .<
.nop<

amend: sosoq .mdonpwHSOHmo xmflm pooam one moonsm pang ommso><ll.ma mqm<a

 

163

was not significantly different from zero which indicates
that, with protection, the floodplains are valued com-
parably with upland farms.l

Correspondingly, the residual for the 32 observa-
tions without protection was $—72.86. This group was
further broken into two categories——those with stream
frontage (15 observations) and those without (17 observa—
tions). The corresponding residuals were $—ll7.8A which was
highly significant (smaller than zero) and $—33.17 which
was significant at the 10 per cent level. The difference
in the residuals for the frontage and no—frontage sales is
partly a reflection of differing flood risk and partly the
effect of averaging some upland values in the residual for
the no frontage observations. For the sales with frontage,
an average of 75 per cent of all open land on the farm
was in the high risk zone, and 19 per cent was located on
the floodplain between the limits of the 5 and 100 year
floods. Thus, essentially all of the open land (9A per
cent) on these farms was subject to risk. In the case of

the tracts without frontage, 62 per cent of the open land

 

1The average price per acre for the 38 protected sales
was $321.26. Theoretically, there may still be some flood
damage expectations built into these protected floodplain
values. However, the major levees of the area (Gill Town—
ship and Brevoort) have not been over—topped since they
were completed in 19A7 and the upstream reservoir construc—
tion since then has probably further strengthened local
confidence in the systems. The land value analysis cannot
prove that no residual flood risk discounting remains but
it seems unlikely that the discount would be very large.

 

 

160

was in the high risk zone and 8 per cent was between the
limits of the 5 and 100 year floods.

A regression analysis of the residuals for the Lower
Wabash area similar to the analysis in the four Upper
Wabash counties was attempted but was plagued by intercor-
relation between the variable for frontage and the variable
for high risk. When the frontage variable was deleted,
the remaining independent variables (per cent of land in
high risk zone, per cent in low risk zone) had essentially
no explanatory power. However, the estimated residual for
the 15 no—levee sales with frontage should be comparable
to COE estimates (Chapter IX) because the distribution of
floodplain land among the flood risk zones for these obser—
vations is about the same as for all floodplain land in the

1
reaches.

 

1According to data supplied by the Louisville District
foice, 7A per cent of all floodplain land unprotected by
levees is in the high risk zone.

 

CHAPTER VIII

REGRESSION ANALYSIS:

WHITE SUBAREA

Although Bartholomew and Jackson Counties are con—
tiguous, the land market in the two counties differs suf—

ficiently so that they could not be analyzed jointly.

 

There were 81 observations (16 floodplains) in Bartholomew
and 118 observations (36 floodplains) in Jackson for the
period 196A—1966.l

Table 1A presents the separate estimating equations
for all observations in each county. The variables are
defined in the same manner as in previous equations except
for "Farm Improvements.” This is a discontinuous variable
that takes the value of one if the tract contained build—
ings and takes the value of zero otherwise. This variable
was not significant in any previous equations. The nega—

tive coefficient in the Jackson County regression indicates

 

lNo effort was made to extend the sample to earlier
years as was done in the other areas because the initial
sample——which included all sales over 10 acres and in all
townships——indicated a very large number of observations
would be obtained. However, many of these were sales of
less than 20 acres, which were unusable, or were located
in several (upland) townships that were later omitted
because of comparability problems. Also the county records
were much poorer than in the other areas and the culling
rate was consequently higher.

165

 

.mudEHme .HO mLOLhm ULNUCMumd

 

 

08.0 3.2 3.3 $3
H:H.O :m.wm $0.0: mwma
Noamm 00 snow
msa.0 wm.ma 00.:H . mamm 00 snow
:3H.0 m0.mm. 0m.H: coapmooq Cadadpooam
500.0 sa.mm mw.sw Hmauodm
mso.o mm.0 mm.H Esme omens
mooo.o m:.o w©.H mooo.o om.o mo.H Ehmm ESHUwS
mooo.o Hm.o mo.m mooo.o ma.o OM.H Ehmm HHmEm

”mesoEo>0LQEH

 

 

 

 

 

m00.0 mm.:o Hm.:ma 0wH.0 sm.ms mm.sm manummm COHBLOQosm
m000.0 mm.0m mw.mom mma.o sa.mw mm.:wa mo dose coapLOQOLd
m000.0 0m.:m . mm.mom H00.0 mm.mw 0m.m:m 0 pogo coapLOQOLm
mH0.0 mm.ms 0m.:wa m000.0 HN.H0 mm.:mm m< ooso coachedosm
mmme $4; 3.3: moses. moo
MW mmo.0 mm.mm mo.wm| mpCme>0hQEH Ehmm
11 "moanmflsm> ucopcm oUCH
mn.w>a mm.mam| m:.m0a 30.00a ucmpmcoo
mom.moa :m.awm mem.mm wa.w:m m
00.0 00.0 mm
mHH Hm mCOHum>ammno 00 hopes:
Ho>oq Lonam psoflofiwmwoo Ho>mq Loshm ucmfioflmmooo
mesmoHMHcmHm Upmpcmpm scammmhmmm 00:00Hmflcmfim ULNUCMpm coammoswom EopH
:omxomh Boanozphmm
.momalzmma

“mosmndm mpflnz wmfiucsoo cowxomh new BmEOHogpnmm .moawm HH< .mpcwflOflmmwoo coamwosmmm pomeHummll.:H mqm<a

 

167
that tracts with buildings tended to command lower prices
per acre than tracts without buildings. Jackson County is
also unique in that it was the only county in the study
with a positive coefficient for flooplain location. The
coefficient for floodplain location in Bartholomew was not
significantly different from zero. Both of these results
are consistent with the relative average price per acre for
upland and floodplain sales in the two counties.

Proportion of total land in pasture came into the
estimating equation for Bartholomew but was of very low
statistical significance. Pasture was significant, and
highly valued, in Jackson County. It may also be noted
that the value estimates for cropland by grade in Jackson
County are not consistent with the assumed quality relation-
ships implied in the cropland grade categories.

Upland and Floodplain Values,
Bartholomew Counpy

Table 15 presents the results of the analysis of
upland farm sales in Bartholomew County. The average
price per acre for upland ($349) and flOOdplain ($335) is
quite close. However, as the discussion of Table 5:

Chapter IV indicated, the composition of the average upland
and floodplain farm differed considerably. Sixty—five per
cent for the cropland on the floodplain farms was graded

A or B compared to A2 per cent for the upland farms. This
is also reflected in the relative percentage Of cropland

grade A or B of all land in farms in Table 15. When the

 

 

.opmsflpmo do posse osmpCMpmm

 

 

 

 

 

m.wm a Hoo.o ss.om mo.mHH some

H.me a ema.o mo.mm mm.we mama
"onm 00 snow

sm.am a _ mooo.o mm.o sH.H Esme eased:

mm.am a mooo.o sa.o mm.a stem Haesm
"meowEo>0LQEH

N.mm a oaa.o mm.ms mm.maa we done coaotodosd

0.0H a . Hoo.o sm.ao Hm.:mm 0 dose coaotooood

m.:m a mooo.o 0m.mm :m.mmm ma 0000 eoaoLOQOLd
"moHQMHLm> pcoocodoUCH
0m.:: 00.0s peopmcoo
mm.mmm mm.m:m a eem.om mm.wem m
me 0 mm
0H mm mm mQOHpm>Lomno mo sonssz

Ho>oq Lossm
cfleadeooae nomads one: oozeOAcnemnm eeeeeoom pdoaoaceooo

sooH

memo: COflmmoswom woman:

 

.mwmalzwma .zpcsoo zosofiogpsmm .m:0fi0m>hmwno camadpooam

one ocean: you mosam> coo: pcm mpCoHOHMMooo COHmmoswmr osmAQD oopmsapmflll.na 040:3

169

average price of the floodplain sales is corrected for
these differences the estimated average price is $370.68
per acre (Table 16). The difference between this estimated
price and average sale value per acre for the floodplain
sales is $—35.39, or in terms of the average residual,
$—35.90.l This difference provides an estimate of the
discounting for flood risk on these sales, but the residual
is not significantly different from zero. Only 56 per cent
of the open land on the average floodplain farm was located
on the floodplain (19 per cent in the high risk zone) so we
would expect the discount for land entirely in the flood—
plain to be greater. However, the number of observations
and the statistical reliability is too low to enable
further analysis in this county.

From the viewpoint of the land value approach, however,
the problem with the analysis is at least partially attrib—
utable to the lack of observations (which might have been
remedied by extending the data collection to earlier years).
The techniques used do indicate that differences in land
values attributable to flood risk exist and can be identi-
fied. On the other hand, even with more observations the
standard errors would probably be relatively large because
the floodplains narrow considerably in Bartholomew County

and most of the farms overlap the floodplains.

 

1The two measures differ because the first is the dif—
ference in averages and the latter is the average of the
differences and some difference in weighing is involved.

 

 

 

 

 

 

170

 

 

 

.0H manna “moUOQ0000 comm

 

 

 

.m.z s0.H mm.mm om.mm1 moHMm QHmHQUOOHm ma
sossm Hospflmom

6888203 dawns. esoeodpm a one: Someone
wflmzams< Hoopflmom oso< mom .m
mm.mm1 Amlzv Xmas pooam on oanmpsoflsppm cocmsodmfio .m
00.0sm cmosam> ocean: pm moamm cfinadpoon 00 moasd popmeflumm .:
:m.mHI cocosommflp poemsnpmgb .m
INMuWWWI moamm cwmfldeooae 0H .esmm omeso>m mo when too conga .m
mm.&:mw moamm oceans m0 «Spam ommso>m do when sod moflsm .H

 

mwmzame< oso¢ sod deflsm ommsm>< .<

 

.mpssoo zosoaonusmm «moonumHSOHmo xmflm eooam one moofism pend ommso><ll.ma mqm¢s

 

171

Upland and Floodplain Values,_
Jackson County

As pointed out previously, upland farms in Jackson

 

ounty sold for an average price per acre of $110 lgsp
han floodplain land. The upland observations in the
ample exclude sales from some of the poorest farming
ownships in the county but there are still very large
ifferences in the upland and floodplain farms. This is
llustrated by the fact that 60 per cent of all land was
n grade A and B cropland on the floodplain, compared to
ess than 2 per cent on the upland farms (Table 17).

A regression analysis was performed (Table 17) for
he upland sales alone but the results were unsatisfactory
or adjusting floodplain sales for comparability. There
as not enough A or B grade cropland on the upland farms
0 derive a reliable estimate of value and the relative
alues for the different grades of cropland were inconsis—
ent. The explanation for the high value assigned to
asture also is not clear. Basically, however, the prob-
ems seem to lie in the heterogeneity of both the area
ad the land market in the county. Regression runs were
lso made on various combination of sales in townships
ith similar characteristics both in Jackson alone and in
artholomew—Jackson combinations but were not successful.

With the contrast between the upland and floodplain
ales and the limitations of available data, there is no

eans of estimating the value of the higher grades of

 

 

 

 

 

 

 

 

.oumEHumm mo posse osmocmumm

 

 

 

 

m m 00H.0 0m.mz 00.0» wamm Hofioodw

mm.0 mm.oa e mao.o 00.0 mm.a scam owned

w0.m m0.wa m m000.0 w:.0 Hm.w Show Eswooz

Ho.mm om.ea a mooo.o mm.o sa.m stem Hanan
”mucoEm>0LdEH

0.: 0.2 a mooo.o mode smimmm gamma coaooooofi

H.m N.m: a mooo.o mm.s: mm.aam we done coaunonoti

:.mm m.:H R H00.0 mm.mm 0:.smfi 0 doso coaunooosm

:.mm 0.H a Hma.o ms.mom oa.mmm me oooo coaododosd

smifi Hw.H I :wm.0 00.0m mm.mm woes/s. woq

91 0.0m m.m0 N 000.0 H:.mm 0m.sm| mucoso>osdEH Esme

7
ll “moanmflbw> pcwoco oocH
mm.mm H:.@>I acoumcoo
mm.mmm :m.smm a emm.mm :m.swm 0
mm 0 mm
mm mm mm mcofipm>comoo do soossz
pHCD Ho>oq aospm
Camaduooam ocean: mucosamficwflm osoocmpm pdoaoaemooo
socH
mcmoz coammwnwom ocmadp

 

 

 

173

crOpland under no flood risk conditions or of estimating
the price this land would command in the absence of flood—
ing.l This does not mean that no discounting for flood
risk occurs on the floodplain but simply indicates that

in the absence of flood risks, floodplain land prices
would probably be even higher.2 The difficulty, however,
is that there is no way to estimate the potential incre—
ment from available land price data. If the Corps of

Engineers damage estimates are accepted the increment

 

should be substantial since 92 per cent of all Open land
on the average floodplain farm was located in the flood—

plain.

 

 

1An attempt was made to estimate flood-free land
values in Jackson County on the basis of upland values in
Bartholomew County but this still understated eXpected
values. A regression analysis of the 36 floodplain sales
was also made to determine if there were differences in
the floodplain prices relatable to degree of flood risk.
This was also unsuccessful, presumably because of cor-
relations among crOpland acres and the various measures
of flood risk.

2Better estimates from the analysis of Bartholomew
County sales would have been useful in estimating probable
price increments in the absence of flooding.

 

 

 

 

 

CHAPTER IX

COMPARISONS OF LAND VALUE

AND FLOOD RISK

In this section we will attempt to relate the dif-

erentials in land value attributable to flood risk to

 

he estimates of flood damage developed by the Corps of
ngineers for the same areas.1 The COE estimates for

11 reaches in this study were based on 1960 or 1963 value
ata. As a part of this study, the three Interim Reports
ere reviewed and procedures used by the COE in deriving
gricultural damage estimates were checked for accuracy

nd reasonableness. On the basis of the review of Corps

   
  
  
  
 
   
  
 

ocedures, certain adjustments to the estimates of

 

~erage annual damages for the reaches in the White sub-
ea and for the West Fork of the White River in the Lower
bash area were made, as eXplained below in the discus-

°on of these areas. Otherwise, the damage estimates

 

1The primary source of data was computer printouts

d other calculations of estimated average annual damages
der natural and modified conditions for each reach in the
udy supplied by the Louisville Office. These data were
ecked against other published data in the Interim

ports and House Document A35, 0p. cit. In COE termin—
ogy, "natural” conditions include allowances for levees

d other protective structures already in place.

170

 

 

 

175

prepared by the Corps appeared reasonable and were accepted

as objective measures of eXpected flood damages.l

Capitalization Rates

 

The capitalization rate selected for illustrative
purposes was 5.0 per cent. Rate of return to land data
for the State of Indiana are fragmentary and nearly non—
existent at the county or smaller unit level. However,
5.0-6.0 per cent seems a reasonable estimate of prevail-

ing rates of return, whether calculated by the landlord

 

approach or as a residual payment to capital.

The landlord approach uses actual net cash rent
received by the landlord from rented land as an estimate
of the income stream from all farm real estate.2 Data on
the landlord's returnenxaavailable only on a state basis.
For Indiana the rate of gross rent to value for whole

farms ranged between 6.8 and 7.0 per cent between 196A and

 

1966. After deducting landlord expenses, this corresponds

3

to a net yield of about 5.0 per cent a year.

 

1This statement must be qualified since it was not
possible to completely review all aspects of COE techniques
because of the specialized nature of the hydrological and
statistical skills required for flood damage estimation.
Therefore, the evaluation of estimated agricultural dam-
ages is based on the premise that the Corps data for flood
frequencies and areas inundated by flood stage are correct.

2Harry Bruce Huff, ”Land Values and Valuation: A
Landlord Approach" (unpublished M.S. dissertation, Michigan
State University, 1967).

3U.S. Department of Agriculture, Farm Real Estate
Market Developments, CD—68 (July, 1966), p. 22, and CD—7l
(December, 1968), p. 17.

 

 

176

Purdue University maintains a farm account project
which is operated primarily as a research and system devel—
opment vehicle rather than as a service to farmers.1
Therefore the number of farm cooperators is limited and
generally restricted to less than 10 accounts in any one
county. Since the cooperators represent a selected group,
statistical inference to a population of Indiana farmers
is doubtful but these records are the only practical source
of data for estimating a capitalization rate from return-
on—investment data. A drawback of this data source is
that returns are calculated on total farm investment;
however, a major part of this capital is for land. There
is probably also a tendency for farm cooperators to be
better—than—average farmers. If so, the rate of return
earned by the average farmer is overstated by account data.
7 The rate earned on investment varies widely from
year—to-year, depending on weather and price fluctuations.
The year 1965 was particularly favorable with some types
of farms earning returns of up to 15 per cent. By con—
trast, 1967 was unfavorable and some account farms had
negative returns on investment after imputing returns to
operator labor and management. Nevertheless, for all farms
in the project, the average rate of return was A.3 per

cent for 1958—62. Over the last five years (1963—67)

 

1Data from the Purdue project was supplied by George

3. Irwin, Department of Agricultural Economics, by personal
correspondence.

 

 

 

 

 

177

returns averaged 5.7 per cent if 1967 is included and 6.6
per cent, if 1967 is excluded (Table 18).

Farms in the northern part of the State (which
include the Upper Wabash area) yielded lower rates of
return than farms in Central or Southern Indiana. Farms
that included hog production as an enterprise best repre-
sent the farms in the study areas. The only data on gen—
eral farms (found in Bartholomew and Jackson Counties)

indicate returns of A.2 and A.A per cent in 1965 and 1966

 

respectively.

Upper Wabash Area

 

The four counties in the Upper Wabash area include
parts of reaches W—8 and W—ll and all of W—9 and W—lO.
Estimates of crop and non—crop agricultural damages for
each reach were obtained from computer printouts and sup—
plemental data supplied by the Louisville District office.
The damage estimates, for this report, are reported on a
per acre basis and represent a weighted average of
expected damages over all reaches or proportion of reaches

within the Upper Wabash study area.

 

Averaged over all reaches, agricultural crop damages
were estimated to average $A.83 per acre, based on 1960
price and yield data, under natural conditions (no flood

protection). With completion of the three reservoirs, the

 

residual crop damages, under the same price assumptions,

would be $1.03. If the losses attributable to non-crop

 

 

 

178

TABLE 18.—-Average Rate Earned on Investment, Indiana Farm

Account Cooperators, Selected Years.

 

 

 

Category 1958-1962 1963—1966 1963—1967
All Farms A.3 6.6 5.7
By Area
North A.O A.9 A.l
Central A.A 6.7 5.7
South A 6 8.A 7.6
By Type
Crop-Hog A.8 7.A 6.3
Hog—Feeder Cattle—Crop A.9 5.6 A 6
Crop—Feeder Cattle A.0 A.A 3.5
Crop 5.0 5.2 5.3
Source: George D. Irwin, Department of Agricultural

 

Economics, Purdue University, by personal cor—
respondence.

 

 

 

 

I."

 

 

agricultural damages are assumed distributed uniformly
across the floodplain, total agricultural damages would
average $9.72 before protection and $2.01 after protection.

If annual earnings are capitalized into land value
at 5.0 per cent, the Corps estimates of average annual
agricultural damages from flooding indicate that before
the reservoir protection was offered, uplands should have
commanded a price premium of $l9A over comparable flood—
plain land. With completion of the three reservoirs, the
residual discount should be reduced to an average of $AO
per acre over the reaches.

The implication of reservoir construction in the
Upper Wabash area is that the construction should be accom—
panied by an increase in floodplain land values. There
is first the question of when the increases would be
expected to occur. If price adjustments were discrete
phenomena that followed completion of construction, the
effects would not be fully reflected in the years sampled
because only two of the three reservoirs were fully opera—
tional by January, 1967. If the increments in land value
accompanied construction starts, land prices should have
begun to respond in the early 1960's. Finally, if the
capitalization of expected benefits occurs in anticipation
of protection, the capitalization process may have been
underway at the beginning of the sample period since the

reservoir investigations had already occurred by 1955.

 

 

 

 

—--

 

180

Conceptually, one would expect at least some capi—
talization to occur in anticipation of protection but with
full capitalization perhaps delayed until sometime after
completion of construction because of uncertainty about the
effectiveness of the protection offered. The analysis of
land price data is at least not inconSistent with this
explanation but the data on the trend in floodplain land
prices over time is inadequate for determining actual land
price response to the reservoir construction.

The mean residual difference for the 39 sales of
property subject to flooding from the Wabash River was
$-26.79. This amount was statistically different from
zero at the 0.025 probability level but it probably biased
downward because most of the sales overlapped the flood-
plain and therefore contained some upland. The best esti-
mate of the relationship between floodplain location and

the size of the residual can be obtained from the equation:

FP Residual = A9.160 — 68.A00 (HG RISK)

- 55.809 (OTHR FP) — 3.680 (YEAR)

The mean value of HG RISK (per cent of open land
flooded once in 5 years or more frequently) was 0.336.
The value for OTHR FP was 0.207 and mean YEAR was equivalent
to 1961. Maintaining the same relationship between HG RISK
and OTHR FP but assuming that all land in the average tract

was located on the floodplain yields values of 0.619 and

 

 

181

0.381 per cent, respectively, and an estimated residual of
$-55.77.

The regression based on the analysis of residual
differences is very low in explanatory power and therefore
relatively small confidence can be placed in the estimated
$56 residual. However, the most difficult problem with
the analysis is in determining how this residual relates
to the Corps' estimate of expected flood damages. The
coefficient for year of sale is not significantly different
from zero and at best indicates no strong trend in flood—
plain land prices after adjusting for the general rise in
land prices over the period. However, the negative sign

for the coefficient is inconsistent with the expected posi-

 

tive sign under the assumption that changes in land value
are associated to some degree with construction or actual,
effective reductions in risk.

The mean year of sale for the sample was 1961 which
preceded any effective protection in the area but, as
previously noted, the determing factor in the capitaliza—
tion process may be the time at which land owners first
begin to anticipate protection rather than the time con—
struction starts or protection becomes effective. If some
capitalization was already occurring during the early
years of the sample then the estimated discount of $56 is

not inconsistent with the capitalized COE estimate of $A0

per acre after protection.

 

 

182

In anticipated ex pppp applications of the land value
check (Ex ppgp to authorization) this problem would not
exist because there would be only one conventionally—
derived estimate and presumably the investigation would
sufficiently precede authorization so that anticipatory
capitalization would not be a problem. For this antici-
pated use, the present analysis indicates (a) a land price
differential can be discerned, even on relatively narrow
floodplains and (b) prices of land on floodplains apparently
do reflect differentials relating to flood risk (as
indicated by the negative and significant coefficient for
HG RISK in the preceding equation and the smaller, non—
significant but still negative coefficient for OTHR FP).
However, the standard errors associated with these esti—
mates are large and it required a search of 11 years of
sales records over four counties to obtain the 39 flood-
plain observations on which the analysis is based. It is
probably impractical to extend the sample further in time
because the quality of the records maintained at the county
level (especially assessment records) deteriorates rapidly1
and because of the possibility of large structural changes
in the land market over time. Extending the sample in
space would also create problems from heterogeneity of the

geographical area to be investigated. Finally, the

 

lGenerally, only the records for the assessment pre-
ceding current assessments are maintained in the counties
selected for this study.

 

 

 

 

183

analysis of turnover rates reported above (in Chapter VI)
indicates the possibility of significantly increasing the
number of observations through Grantor—Grantee interviews
to determine the status of questionable transactions
eliminated from this sample is low. Therefore, the desir-
ability of applying a land value check to an area with the
characteristics of the Upper Wabash area may require a
careful weighing of the costs of obtaining the data against
the expected reliability of the land price data obtained.

A possible cost reduction procedure is suggested in the

conclusion of this report.

Lower Wabash Area

Knox County, in the Lower Wabash area, is bordered
on three sides by the Wabash, the White and the West Fork
of the White River. Sullivan County has a long frontage
on the Wabash. Together river frontage in the two
counties includes all or the major portion of reaches W—3,
W—A, WH—l and WW—l. The Wabash reaches receive protection
from the three upstream reservoirs protecting the Upper
Wabash area plus the Mansfield reservoir.1 The White River
reaches receive protection from the Cagles Mill and Monroe

Reservoirs. Damage estimates based on the existing

 

1Since the downstream effect of the Huntington
reservoir alone is small, it was assumed that protection
from this reservoir was effective in order to simplify the
analysis of COE-data.

 

 

 

 

 

 

L___ _

18A

modifications were used. The Wabash reach data are based
on 1960 prices; the White reach data on 1963 prices.

The average annual crop damage estimated by the
Corps was $3.59 per acre (weighed over all reaches). Non—
crop damage was estimated to average $3.97 per acre, for
total estimated average annual agricultural flood damages

1 At 5.0 per cent an average land price

of $7.56 an acre.
differential of $151 per acre is indicated.

The best indication of the land market‘s evaluation
of the flood risk is the 15 sales of tracts of land with
river frontage. These sales were almost completely

embraced by the floodplain (9A per cent of all open land

on the farms was within the limits of the 100 year flood).

 

The mean residual for these sales was $—117.8A per acre,
with a standard error of $17. The standard error indicates
that with a confidence interval of slightly over 95 per
cent we could not say the COE and the land market esti—
mates are significantly different. With a confidence
limit of one standard error (roughly a 66 per cent confi—
dence interval) the range of expected damages, in annual
equivalents at 5.0 per cent, is $5.06 to $6.72 per acre,

compared to the COE estimate of $7.56.

 

lTotal damage estimates for WH—l and WW—l are about
twice the size of the estimates for the Wabash reaches.
Part of the difference appears attributable to assump—
tions made by the Corps for expected damages in the high
risk flood zones. This assumption is discussed in the
following section on the Bartholomew—Jackson area.

 

 

185

On the basis of the closeness of agreement between
the land value and COE analysis there is no strong basis
to question the accuracy of the Corps' estimate. How—
ever, in attempting to reconcile the wide discrepency
found in the Jackson County analysis, a procedural error
was noted that could lead to an overestimation of
expected damages from frequent or out—of—season flooding.1
The same procedure appears to have led to an overstate—
ment of damage estimates for reaches WH—l and WW-l of the
White River. If the computed damages for floods with
modified frequencies larger than once a year are ignored,
the weighed average agricultural damage for the Knox—
Sullivan area would be $6.81 instead of $7.56 per acre.
This estimate is very close to being within one standard
error, at a 5.0 per cent capitalization rate, of the best
land market estimate.2

The point damage estimates derived from the land
market analysis are $5.89 or $7.07 per acre at 5.0 and
6.0 per cent, respectively. These estimates embrace the
corrected COE estimate of $6.81. Beyond this point the
anayisis cannot specify which estimate, if any, is the

"best" estimate. The decision to accept any of the three

 

1See page 190.

2The agreement is closer if a 6.0 instead of a 5.0
per cent discount rate is used. At 6.0 per cent, the
$~ll7 differential is equivalent to an annual loss of
$7.07 ($6.08 to $8.06 at one standard error limits).

 

 

186

would depend partially on how critically the choice
affected the benefit-cost ratio. If_the choice were
critical, a full—scale review of the entire COE estimat—

ing procedure would probably be the logical next step.

White Subarea

 

Floodplain sales in the White subarea were located
on reaches EW—3, EW—A, and EW—5 of the East Fork of the
White River and along the first reach each of the Drift—
wood and Flatrock Rivers. Bartholomew and Jackson Counties
were sufficiently different to prevent a joint analysis
and consequently there were only 16 floodplain sales in

the analysis for Bartholomew County. The unadjusted dif—

 

ference in the mean price per acre for the upland and
floodplain sales in this county was only $-l3.5A ($308.53
versus $335.29). However, the analysis indicated that,

at upland values, the floodplain lands would command a
price of $370.68, or $35 over the current per acre price.
The indicated residual of $—35 was DOC statistically
significant. This residual probably underestimates the
expected price of floodplain land with protection, how—
ever, because most of the farms overlapped the floodplain
and only about half the open land on the average tract
was subject to flooding. Therefore, the expected flood-
free price of land located entirely on the floodplain
might rise to $AOO—$A10 an acre ($335 plus twice $35) but

no statistical reliability can be attached to this estimate.

187

Jackson County presents even more of a problem
because the floodplains are apparently highly desirable
farming lands and already command a $110 premium over the
average Jackson County upland farm and are comparaby
priced relative to upland farms in Bartholomew County.
Presumably, land prices on the floodplain would be even
higher in the absence of flood risk but sale price data
are not available for estimating the expected price.

Jackson and Bartholomew Counties are also unique
because the COE estimated annual average flood damages are
the highest of the three areas. The weighted agricultural
crop damages were estimated at $16.89 per acre. In addi—
tion, non—crop agricultural damages were estimated to be
about $A.27 per acre.

Considering only the estimated crop damages, a per
acre return of $16.89 would support an average value per
acre of land of $338 at 5.0 per cent. This is equal to
the current price of floodplain in Bartholomew and Jack—
son Counties and implies that in the absence of flood
damages land prices would double. In order to support
the current price of $338 per acre while simultaneously
incurring an average yearly crop loss of $16.89, the
floodplains would have to yield a net return of $33.78

per acre.1 It can be noted that the floodplain lands in

 

1The available soil productivity data for alluvial
soils indicate that the productivity of floodplain soils
in the White subarea is no greater than the productivity of
typical alluvial soils found in the other study areas.

 

 

 

 

188

the two counties already support a higher price per acre
than the uplands or adjusted floodplain prices in either
of the other two areas.

A possible explanation of the current high flood—
plain land prices, assuming that flood damages are as high
as the COE estimates, is that some farm operator immobility
may exist in the area. Since the uplands are relatively
unattractive for farming purposes (at least in Jackson
County), farmers may be willing to accept lower returns
to their capital and labor than in the other areas and thus
they have bid up the price of floodplain land despite the
presence of risk. But even if landowners accepted only a
2.0 per cent yield on their current land investment, the
floodplains would have to be capable of yielding a net
return of nearly $2A a year to cover the estimated flood
damages and a 2.0 per cent residual return to the operator.
This required return still exceeds the indicated returns
in the other study areas and indicates that the explanation
of the discrepancy apparently must lie in errors in COE
data or procedures.

Corps data for reach EW—A of the East Fork of the
White River were selected for review.1 Based on revised
data supplied by the Louisville District Office, the

average annual crop damages in this reach was estimated

 

1Interim Report No. 3, Vol. III, Appendix C, Plate

A.

 

 

189

at $1.26 million, or $17.63 per acre over the 71,600
acres in the reach. Non—crop agricultural damages were
an additional $A.9A per acre. In reviewing the Corps
data, errors were found (1) in transferring crop value
data and (2) in estimating the expected damages on fre—
quently flooded land. These two error sources apparently
explain a major part of the high crop damage estimates.

In this Interim Report the Louisville Office adopted
the use of the Water Resource Council's long—run normalized
prices in their computations which would tend to yield
slightly higher damage estimates than the 1963 prices used
in the Lower Wabash area or in previous investigations of

EW—A (for Interim Report No. 2). However, after correctly

 

computing the value—at—risk (current crop yields times
price per unit) on the basis of normalized prices, con-
siderably higher values for several crops were used in
computation of the stage—damage curves. For example, the
value—at—risk for corn should have been $88 per acre.
Instead a value of $100 was used in the first stage of
crop damage computations and carried throughout the suc—
ceeding computations. Correcting for this error, the crop
damage estimate would be $1.13 million for the reach or

$15.8A per acre.1

 

1Since this discrepancy was fairly obvious it may
have been corrected in further reviews of the Interim
Report; however, a "re—computation" of the data as of
January, 1967 (supplied by the Louisville Office) apparently
still contained the error.

 

190

The second procedural error apparently arose because
a high proportion of the floodplains in this reach is
flooded more frequently than once a year. The COE cal—
culates that crop damages begin at stage 11.0 which
occurs, on the average, A.22 times per year (Table 19).
At stages 16.0 and 17.0 flooding would still occur with a
natural frequency of 1.A7 and 0.865 times per year,
respectively, and would inundate 28,800 and 38,A00 acres
of the total 71,600 acre floodplain. Assuming that the

one year flood occurs at approximately the 16.5 foot stage,

 

nearly half (A9 per cent) of all annual crop damages for
the reach are assumed to occur on land flooded more fre—
quently than once a year.

Damages of this magnitude for land in the one—year
flood zone are unreasonable and indicate either that the
frequency data are wrong or that most flooding occurs in
the winter when damages should be small. Since flooding
at these frequencies should be easily verifiable, the
latter explanation appears most reasonable.

If the floods that occur one to four times a year
are primarily winter or early spring floods, crop damages
should be very low or non—existent. This is recognized by
the COE in that very low damage factors are assigned to
winter floods. However, the COE also makes an adjustment
for seasonality that has the effect of averaging over all

floods regardless of seasons, yielding a much higher

 

 

191

‘ABLE 19.——Flood Stages, Acres Inundated, Natural Frequen—
ies, and Estimated Average Annual Crop Damages; Reach
EW—A, Jackson County, Indiana.

 

 

 

 

 

Average
‘t 1 Acres Natural Crop Damage Annual
3 age Inundated Frequency by Stagel Damages
(dollars) by Stage3
(dollars)
10.0 0 A.75 0
5,177
11.0 1,600 A.22 19,535
13,73A
12.0 3,A00 3.77 A1,507
28,905
13.0 5,700 3.25 69,821
50,100
1A.0 9,200 2.70 112,50A
94,713
15.0 19,100 2.15 231,905
197,028
16.0 28,800 1.07 308,867
206,183
17.0 38,000 0.865 060,961
233,600
18.0 07,600 0.015 573,059
‘ 166,898
19.0 56,000 0.106 667,019
72,661
20.0 60,000 0 000 757,307
20.795
21.0 71,600 0.013 802,302
1,130,318
1Interim Report No. 3, 111, Appendix C, Plate A
(The "crop damage by stage" column has been corrected for
computational errors).

 

2Supplied by COE, Louisville District, January, 1967
data.

3Recomputed from COE data.

 

 

192

average damage factor.1 If this latter damage factor is
applied to the floods which occur primarily in winter,
average annual damages will be greatly overestimated.

It was not possible to adjust the Corps estimate of
average annual crop damages completely for this apparent
source of overestimation; however, if we assume that all
the floods of stage 16.5 or less are winter or spring
floods causing no crop damages then the damages from the
remaining floods (over a 16.5 foot stage) would total
$7.82-per acre, after correcting for the errors in crop
values—at—risk.

Although this estimate ($7.82) is not precisely com-
puted, it appears to be a more reasonable estimate of
expected damages. It may still be biased upward because
the damage factors for stages over 16.5 feet were also
averaged over all seasons. However, data were not available
for checking this possibility.2

The non—crop agricultural damage estimate of $A.90

per acre also seems quite high but there were no data from

—__‘

1For example, the damage factor for corn land flooded
between January 1 and March 31 is only 0.7 per cent,
regardless of depth or duration. But the average damage
for corn flooded to one foot over all floods of record is
equivalent to 17 per cent. See: Interim Report No. 3
III. Appendix 0, page C—71 and Plate 3.

5

 

2It is noted that the corrected crop damage estimate
is still about twice the estimate for the Lower Wabash
erea. Because the floodplains in Jackson County are
intensively farmed, damages from a seasonal flood might be
EXpected to be somewhat higher (but not necessarily doubled).

 

 

 

 

 

 

193

the land value survey that could be used to check, or
refute, the estimate derived by the Corps.

Because of the possibility of further bias in COE
procedures, the estimates of $7.82 and $12.72 (total
agricultural damages) probably should be viewed as upper
limits. Assuming for the purpose of illustration that the
total agricultural damages are in the neighborhood of
$10 an acre, a potential price increment of $200 an acre

in the absence of flooding is indicated. Whether this is

 

reasonable or not can only be evaluated subjectively.
Thus, in areas with characteristics similar to those of
Jackson County, the utility of a land value approach rests
ultimately on its function of indicating what is £22
reasonable rather than what is. In such a situation land
values alone can not yield objective estimates of eXpected

flood damages.

 

 

 

 

CHAPTER X

CONCLUSIONS AND RECOMMENDATIONS

The three areas of the Wabash Basin selected for
investigation each provided a different set of conditions
for testing the applicability of the land value approach.
The results of the investigation in each area, as they
relate to the second objective, are summarized briefly
below. These summaries are followed by a discussion of

the results of the study as they pertain to the first

 

objective-—the estimation of the relationships between
land values and flood risks in the study areas. Finally,
an analysis of the cost of this study and some suggestions
for reducing the costs of applications of a land value
check are presented in Appendix A.

Although the details of the investigation have not
been presented in this report, the procedures and data
used by the Corps of Engineers in deriving estimates of
expected average annual damages over the river reaches
included in this study were reviewed for each area and
corrections were made in the Corps' estimates for the
Lower Wabash and White subarea. Although the review can—
not be considered definitive because of uncertainties

regarding the hydrological and statistical aspects of the

19A

 

 

195

COE procedures, it is believed that the corrected COE
damage estimates are reasonably close to the actual flood
risk in each area, and therefore are relevant for compari—
sons with the estimates derived from the land value model.

In the Upper Wabash area the investigation of land
price relationships was complicated by the narrowness of
the floodplains in this area which yielded relatively few
floodplain sales in any year. In addition most of the
farms "overlapped” the floodplains and the uplands. This
latter consideration complicates the determination of the
flood risk discounting that may occur because only part of
the land in a typical sale is subject to flood risk. The
investigation in this area did indicate, however, that it
is at least feasible to apply the land value approach to a
narrow floodplain area and that there is some evidence of
regularity in land price response to flood risk.

The indicated price differentials for floodplain
land could not be related to a conventionally derived
estimate of expected flood damages because of the change
in flood risk over time. In this respect the sub—objective
of the investigation in this area—-to determine if the
effects of flood risk reduction would be observed in flood—
plain land price data——was not achieved, presumably because
the processes associated with the risk reduction (project
investigations, local hearings, projection authorization,

and construction) were spread over the entire study period

 

 

 

 

 

196

and there is no theoretical justification for expecting
any one of the processes alone to trigger the capitaliza—
tion process. In the anticipated applications of the
land value check by federal agencies this problem should
not exist because the land value investigation could
sufficiently precede authorization so that anticipatory
capitalization would not be a problem.

The analysis of land price differentials in the
Lower Wabash area produced close agreement with the COE

damage estimates. Although there were only 32 observa—

 

tions without levee protection (and only 15 of these
observations had frontage) the relationship between flood
risk and land price differentials was strong. Uncertainty
about the appropriate capitalization rate becomes a major
limitation in relating damage estimates. Allowing for the
possible over—estimation of damages by the COE in the
White River reaches, the differences in the COE and the
land value—derived damage estimates are not statistically
significant.

The fact that the levee—protected sales failed to
reflect any residual discounting for flood risk was also
consistent with a priori expectations. The fact that the
major COE-built levees have not been overtopped since
their completion in the 1940's would be expected to

generate considerable confidence in the efficacy of the

 

 

 

R

 

 

197

levees in the minds of the floodplain landowners and the
results of this analysis do not contradict this expectation.
The analysis for the White subarea yielded unusable
results with respect to the objective of estimating flood
damages from land value data. In Bartholomew County the
problem was basically a lack of observations. This
occurred because Bartholomew and Jackson Counties could
not be analyzed jointly and too few years of sale data were
taken to provide adequate statistical estimates for Bartho—
lomew County alone. This is not a reflection on the
analytical techniques used——except to the extent that it
indicates problems likely to be encountered from the large
standard errors associated with sales data from overlapping
farms——since the analysis does indicate that some price
adjustment to flood risk has occurred in the county.
Jackson County is the extreme illustration of the com—
parability problem in this study. The inferences that can
be drawn directly from land value data in this situation
are limited by the fact that the contrast between the up—
lands and floodplains (and the associated land markets)
is so great that adjustments for comparability were not
possible. The value of a land value check in such a
region1 depends primarily on its utility in detecting only

the large errors in conventional agency techniques or data.

 

1We would consider that mountaineous or heavily for—
ested upland regions generally would be typical of areas
where the comparability problem would be severe.

 

 

 

 

198

The statistical limitations of the data for the
three study areas did not allow a complete investigation
of the relationships between land values and flood risk
as proposed in the first objective but the findings of this
study are at least indicative of underlying regularities
in the land market. Considering the results from the three
areas jointly, it is noted that the analysis did not yield
results in any area that are inconsistent with a priori
expectations based on considerations of land rent theory.
The results, particularly between the Upper and Lower
Wabash areas and between leveed and nonleveed areas, are
consistent in relative magnitude and in relation to degree
of flood risk. These considerations are indicative of
underlying regularities in the land market and, subjec—
tively, are encouraging for regular applications of a land
value check. However, it must also be recognized that land
value data is not a precise estimating vehicle and that the
standard errors associated with a point estimate of flood
risk discounting are apt to be large, particularly in areas
of overlapping farms and/or areas with relatively low
flooding hazards. The standard errors are not necessarily
an indication of wide variance in flood risk discounting
but stem from variance in all the factors determining land
values——many of which cannot be adequately measured or

included in an analytical technique.

 

 

 

 

199

The "average price per acre" regression model was
chosen for the final estimating equations in each study
area because this model gave more consistent estimates of
the effect of the discontinuous independent variables on
sale price. In addition, the final estimating equations
were determined by including only those independent vari-
ables that reduced mean square error of the regressions.
It was believed that these criteria were more relevant
than the frequently used criterion of relative size of the
coefficient of determination (R2). For this reason it is
difficult to compare the results of the equations used in
this study with other land value models. However, Appendix

B may be consulted for some indication of the effect of the

 

choice of variables (specifically, cropland grade variables)
on the relative precision of the per—acre price estimates.
On the basis of the results in the Wabash River Basin
it appears that further applications and tests of the land
value approach in other areas to determine if the same

relationships prevail are warranted.1 As suggested in the

 

1In selecting areas for additional study, an analysis
of land value—flood risk relationships would be facilitated
if an area could be found with extensive acreage of flood-
plain land subject to only occasional inundation (say,
between the 5 and 100 year flood zones) in order to deter—
mine the nature of the price relationships in these flood
risk zones. In the Wabash Basin these areas occurred
naturally to some extent in Knox and Sullivan Counties but
are now largely protected by levees. In the remaining
areas investigated in this study, the stream valleys are
approximately "U—shaped" in cross section and as a result
the total acreage inundated by the 100 year flood is not
significantly greater than the area inundated by a flood

 

200

specific recommendations made below, considerations should
be given to the possibilities of reducing the cost of
applying a land value check. Ultimately, if a check is
adopted as a part of agency procedure, it should be pos-
sible to integrate the land value analysis into the project
justification studies as they are made. The goal of such
integration would be to bring all available data together
to obtain a single estimate of expected flood damages and
benefits that draws on the strengths of both the land value

and conventional techniques.

 

Based on the results of this analysis the following
recommendations are offered:

1. The application of capitalization principles
should be made at the appropriate steps of each project
justification study. The results should accompany the
project report and any wide divergence justified. This
recommendation is intended primarily as a procedural
check to catch errors of logic or assumptions at each stage
of the benefit estimation procedure before they become
embodied (and perhaps hidden) in the next stage. This pro—
cedure would not require precise land value data but the
values used should be consistent with prevailing land
prices. It is noted that estimates of total capital values

in the floodplain are now routinely reported in project

 

with approximately a 5 year recurrence interval. For this
reason the Wabash Basin was not ideal for testing hypotheses
about land price relationships on infrequently flooded land.

   

 

201

investigation reports. These estimates serve no apparent
purpose and the time devoted to their preparation could
as well be spent in deriving per—acre land value
estimates.1

2. Further studies of the relationship between land
values and flood risk should be made in other areas pre—
senting some variety in topographic, farming and flood
risk situations. Additional studies are recommended pri-
marily to determine the transferability of relationships
found in the Wabash and to accumulate a store of knowledge
of land market—flood risk interrelationships. Further
experiences would also be helpful in developing procedural
guides.

3. If additional investigations are made in other
areas emphasis should be placed on reducing the costs of
applying a land value check, especially through systema—
tizing land data collection.2 For example, assuming that
each District Office knows approximately which reservoir
or levee projects are likely to be considered over the

next decade or so, sample counties could be selected and

 

lAn Engineering Manual prepared in 1957 required that
the procedure recommended above be adopted. However, per—
sonnel in the Louisville office reported that the manual had
never been distributed to them and were not aware of the
land value check requirement. See: U.S. Army, Corps of
Engineers, "Examinations and Surveys: Relation of Flood
Damages and Flood Control Benefits to Market Value of Land"
Engineering Manual EM 1120—2—1-1, 13 June 57 (mimeographed).

2A note on the costs of this study may be found in
Appendix A.

 

 

 

 

202

abstractors or the County Recorder in each county could

be contracted to supply the district office with farm sales
data relevant to each agricultural sale occurring in a
specified area on and bordering the floodplain. By virtue
of his knowledge of the conditions surrounding each sale,
the abstractor could also provide an editing and informa—
tional service. The district office could thus maintain

a continuing, current file of data useful for general land
price index purposes (see 1, above) and at the time for a

land value analysis have available a stock of basic data,

 

including Grantor—Grantee names, on which to base addi—
tional data collection. The cost and practicality of
this, or a similar plan, should be investigated.

A. In selecting areas for investigation some con—
sideration should be given to improving land value esti—
mates in narrow floodplain areas. Although these areas
are basically unfavorable for a land value application,
topographical considerations usually dictate reservoir
location in such areas. Thus, damage reduction estimates
in the areas immediately downstream from the reservoir may
contribute in large measure to total benefits stemming
from the project and should be subject to an independent
check. Since the opportunity of increasing the number of
observations in such an area is limited, consideration
should be given to reducing the standard errors associated

with the land value data through supplemental interviews

   

 

 

_rr=7 , :‘_._A, ,1 .,,

 

 

203

or professional appraisals. This would require a careful
balance of the costs against the likely improvement in
precision of estimates and is probably justified only where
the damage estimates in the narrow floodplains significantly
affect the benefit—cost ratio. However, some investigation

of this question appears warranted.

 

 

 

BIBLIOGRAPHY

 

 

20L}

 

BIBLIOGRAPHY

American Insurance Association. Studies of Floods and

Flood Damage 1952—1955. New York: American Insur—
ance Association, 1956.

ASCE Subcommittee on Floods. "Review of Flood Frequency
Methods," Transactions of American Society of Civil
Engineers, 118 (1953), pp. 1220—1230.

Barlowe, Raleigh. Land Resource Economics. Englewood
Cliffs, N.J.: Prentice—Hall, Inc., 1958.

Boyce, Ronald R., ed. Regional Development and the
Wabash Basin. Urbana: University of Illinois
Press, 1955.

Burton, Ian. Types of Agricultural Occupance of Flood
Plains in the United States. Research Paper No. 75,
Department of Geography, University of Chicago,
1962.

Commission on Organization of the Executive Branch of
the Government, Task Force Report on Water Resources
and Power, Vol. III. Washington D.C.: Government
Printing Office, 195“.

Huff, Harry Bruce. "Land Values and Valuation: A Land—
lord Approach." Unpublished M.S. dissertation,
Michigan State University, 1967.

Kates, R. W. Hazard and Choice Perception in Flood Plain
Management. Research Paper No. 78, Department of
Geography, University of Chicago, 1962.

. Industrial Flood Losses: Damage Estimation in the
Lehigh Valley, Research Paper No. 98, Department
of Geography, University of Chicago, 1965.

 

Knetsch, Jack L. "The Influence of Reservoir Projects
on Land Values,” Journal of Farm Economics, 46
(February, 1964), pp. 231—2A3.

205

 

 

 

206

Maier, Frank R., James L. Hedrick, and W. L. Gibson, Jr.
The Sale Value of Flue—cured Tobacco Allotments.
Technical Bulletin 198, Virginia Agricultural
Experiment Station, Virginia Polytechnic Institute,
Blacksburg, Virginia, April, 1960.

Milliman, J. W. ”Land Values as Measures of Primary
Irrigation Benefits," Journal of Farm Economics,
”1 (May, 1959), pp. 234-243-

 

Murphy, Francis C. Regulating Flood-Plain Development.
Research Paper No. 56, Department of Geography,
University of Chicago, 1958.

 

Penn, J. B., Bill Bolton, and Willard F. Woolf. The Farm
Land Market in the Mississippi River Delta Cotton
Region 196A—1965. D.A.E. Research Report 372,
Louisiana State University Agricultural Experiment
Station, Louisiana State University, University
Station, Louisianna, April, 1968.

 

[Pine, Wilfred H., et al.] "Report of Advisory Board on
Agricultural Flood Damages to Army Corps of Engineers,"
April 27, 1956. (Mimeographed.)

 

Renshaw, Edward F. Toward Responsible Government: An
Economic Appraisal of Federal Investment in Water
Resource Programs. Chicago: Idyia Press, 1957.

 

 

Robertson, Lynn, et al. Types of Farming in Indiana.
Bulletin 628, Purdue University Agricultural Experi—
ment Station, Purdue University, Lafayette, Indiana,
December, 1955.

 

Sewell, W. R. Derrick. Water Management and Floods in the
Fraser River Basin. Research Paper No. 100, Depart—
ment of Geography, University of Chicago, 1965.

 

 

Sheaffer, John R. Flood Proofing: An Element in a Flood
Damage Reduction Program. Research Paper No. 65,
Department of Geography, University of Chicago, 1960.

 

Strohbehn, Roger W. Resource Productivity and Income
Distribution With Implications for Farm Tenure
Adjustments. Bulletin 720, University of Illinois
Agricultural Experiment Station, University of
Illinois, Urbana, August, 1966.

 

Tolley, G. S. "Analytical Techniques in Relation to Water—
shed Development," Journal of Farm Economics, 60
(August, 1968), pp. 653—665.

 

207

Army, Corps of Engineers. "Examinations and Surveys:
Relation of Flood Damages and Flood Control Benefits
to Market Value of Land." Engineering Manual EM
ll20—2—lll, 13 June 57. (Mimeographed.)

. "Survey Investigations and Reports: General Pro—
cedures.” Engineering Manual EM 1120—2—101 (includes
change l6), 12 October 65. (Mimeographed.)

Army Engineer District, Kansas City, Corps of
Engineers. "Appraisal Analysis on Agricultural
Levees 3A5—330-L and MOB—L Ray, Clay, and Platte
County, Missouri," prepared by the Real Estate
Research Corporation, St. Louis, Missouri, July,
1967. (Mimeographed.)

Army Engineer District, Louisville, Corps of Engineers.
"Wabash River Basin Comprehensive Study." Vols.
I—III of Interim Reports No. 1 (March, 1963);
No. 2 (January, 1964), and No. 3 (March, 1967).
(Mimeographed.)

, and Cooperating Agencies. The Wabash River Basin:
Water Resources Planning. (32 page brochure, no date,
number, nor publisher given.)

Bureau of the Census. Census of Agriculture, 196A.
Vol. 1, Statistics for the State and Counties, pt.
11, "Indiana, and pt. 12, "Illinois." Washington,
D.C.: Government Printing Office, 1967.

Congress, House. Wabash River, Ohio, Indiana, and
Illinois. H. Doc. lOO, 73rd Cong., lst sess., 1932.

. Wabash River and Tributaries, Indiana and Illinois.
H. Doc. 197, 80th Cong., lst sess., 196“.

, Committee on Public Works. Headwater Reservoirs,
Wabash River, Indiana. [Issued as H. Doc. 435],
84th Cong., 2nd sess., 1956.

. Lincoln, Clifty Creek, and Patoka Reservoirs,
Wabash River Basin, Indiana and Illinois. H. Doc.
202, 89th Cong., lst sess., 1965.

, Task Force on Federal Flood Control Policy. A
Unified National Program for Managing Flood Losses.
H. Doc. A65, 89th Cong., 2nd sess., 1966.

U.S. Congress, Senate. Lafayette and Big Pine Reservoirs,

Wabash River Basin, Indiana. S. Doc. 29, 89th Cong.,
lst sess., 1965.

 

 

 

208

. A Bill to Amend the Federal Flood Insurance Act
of October 1956 to Provide for a National Program
of Flood Insurance. S. 1985, 90th Cong., lst sess.,
19W.

U.S. Department of Agriculture. Crop Production: 1968
Annual Summary, Acreage, Yield, Production. S.R.S.
Crop Reporting Board CR—PR 2—1 (68), December, 1968.

 

. Farm Real Estate Market Developments. CD—66,
October, 196A.

. Farm Real Estate Market Developments. CD—67,
August, 1965.

. Farm Real Estate Market Developments. CD—68,
July, 1966.

. Farm Real Estate Market Developments. CD—70,
April, 1968.

. Farm Real Estate Market Developments. CD—7l,
December, 1968. ‘

”Wabash River Basin Comprehensive Basin Study:
Appendix H, Agriculture." WAC Review Draft,
September, 1968. (Mimeographed.)

Vaughn, Claude M., Jr. Covariance Analysis of Reservoir
Development Effects on Property Tax Base. Research
Report No. 4, Water Resources Institute, University
of Kentucky, Lexington, Kentucky, 1967.

Vermeer, James. Profitability of Participation in the 1962
Feed Grain Program in the Corn Belt. ERS—362, U.S.
Department of Agriculture, Washington, D.C.:
Government Printing Office, [n.d.].

Waldrop, John E. and Daniel D. Badger. Effects of Upstream
Watershed Development on Prices and Values of Farm—
land in the Upper Washita River Basin. Processed
Series P—529, Oklahoma State University Experiment
Station, Oklahoma State University, Stillwater,
Oklahoma, March, 1966.

White, Gilbert F. Human Adjustment to Floods: A Geographi—
cal Approach to the Flood Problem in the United States.
Research Paper No. 29, Department of Geography,
University of Chicago, 19A5.

 

 

 

209

, et al. Changes in Urban Occupance of Flood Plains
in the United States. Research Paper No. 57,
Department of Geography, University of Chicago,
1958.

, ed. Papers on Flood Problemg. Research Paper
No. 70, Department of Geography, University of
Chicago, 1961.

. Choice of Adjustment to Floods. Research Paper
No. 93, Department of Geography, University of
Chicago, 1964.

 

APPENDICES

 

210

 

. -_..

 

 

 

APPENDIX A

COST CONSIDERATIONS

Work on this study began during the early summer of
1967 and was essentially completed by January 1, 1969.
Although there was considerable overlap of the phases of
the study, an approximate breakdown of the professional
time expended was one quarter year for general planning
and search of data sources, one—half quarter year for
specific familiarization with the Wabash Basin, two
quarters for various aspects of data collection and pro—
cessing, and two quarters for data analysis. Report writing
efforts were mainly interspersed with data collection and
analysis work.
In chronological time requirements, nearly a year
was devoted to data collection and processing. Some of the
factors involved here were:
l° The selection of three study areas (eight counties)
required individual contact, training and super—
vision of the eight enumerators.
2. The geographical separation of the three study
areas from each other and from the East Lansing
office created large demands on travel and data

transmission time.
211

 

3. Because of the inexperience of the enumerators,
wide tolerances were established for the field
determination of eligible sales and the editing
burdens were passed to the East Lansing office.
A two—stage data collection plan was also
adopted to reduce the work load of the enumera—
tors and this introduced time lags in the field
work.

A. Because of the one—time nature of the study and
the turnover rates of part—time clerical help,
it was not feasible to train an assistant to
assume major responsibility of schedule editing
and data processing throughout the study.

Many of these time—lags and requirements on profes—
sional time could be reduced or eliminated in routine
applications of a land value check, assuming that respon—
sibility for the check would be delegated to the district
offices. First, the time spent in general orientation,
collection of data from COE sources, and selecting sample
areas would be eliminated. Better geographical location
with respect to study areas would be probable, and possibly
only one area would be investigated at a time.1 Finally,
routines for data editing and processing could be set up

to reduce professional time requirements.

 

lIn balance, the time required for data collection in
three separate areas outweighed any economies of time
realized in the data analysis stage.

 

 

213

The enumerators were hired from the Statistical
Reporting Service of the USDA. They were well trained and
experienced in general data collection work and required
only about a half day of instruction to begin the first
stage of data collection (deed searches). The work
required for the second stage of data collection was
fairly involved since the enumerators had to correlate
records in three county offices but even this task was
handled capably. The SRS enumerators proved to be a prac-
tical medium for primary data collection and, with more
training, could have assumed a larger share of the editing
functions.

These professional enumerators are one potential
medium for systematic land value data collection.1 If
future study sites could be anticipated, the enumerators
could be hired for a one or two week period each year to
search county data sources for land transaction data for
the preceding year. This scheme, by spreading the data
collection process over several years, might prove to be
more efficient than a one—time survey.

The collection of schedule I data required a total
of 793 hours, including travel time, for the eight counties.
Three enumerators expended an additional 679 hours in col—

lecting the schedule II‘s. The hourly rates for the

 

1An alternative scheme would be to contract with
county land title abstractors or county recorders for the
needed data on an annual basis.

 

 

 

 

 

 

21A

enumerators varied from $2.05 to $2.85 an hour, for a
straight—time labor bill of $2,120 for the first stage

. and $1,670 for the second stage.1 One enumerator lived
in the county to which he was assigned, the others were
up to 50 miles from their work. The total payment for
travel was $1,230.

About 700 hours of clerical time was required for
the study, although not all of this time is directly
attributable to data processing needs. Some clerical
time was spent on mapping farm location but the work mostly
involved schedule checking and collating, simple arithmetic
checks, and data coding.

Keypunching and verifying required 37 hours and cost
$154. Computer charges were slightly under $500 for about
one and a half hours of computer time. A standard least
squares program was used but about $400 of professional pro—
gramming time was required for input preparation.

Assuming clerical time at $2.00 an hour, $1,200 total
expenditure on computer services, and data collection costs
of $5,000; the total nonprofessional expenditures were

around $7,600 (excluding secretarial services).2 The time

 

1The actual expenditures were somewhat higher because

overtime was authorized for the enumerators living at
distance from their assigned courthouses.

2Professional travel directly related to the study
was about $865, not including an allowance for government
vehicle use.

 

 

215

requirements for each area were about equal, indicating a
cost of about $2,500 for a study of any one of the areas
(or $3,000, allowing for some economies of time). In total,
these costs are small relative to the approximately one and
a half professional man years expended.

It is difficult to estimate what the professional time
requirements would be if a land value check were adopted
as a part of COE procedure since a number of time and cost
efficiencies that were not available for this study could
be utilized. Potential efficiencies are greatest in data
collection and processing, especially if potential study

areas can be specified sufficiently in advance of the project

 

justification study. Once data collection is initiated,
professional time requirements should be low and could be
handled on a part—time basis at the district office level
or by one full—time professional over several offices.

The marginal cost of adding a land value check to
each project justification study would appear to be
relatively low although this can be evaluated only by the
COE. However, any comparison of relative costs should be
based on the costs required to make a thorough application
of presently used techniques. In the Louisville office,
at least, much of the area—inundated data is based on out—
dated floodplain charts, agricultural production and cost
data for the floodplains are not collected in any systematic

way, and "standard" damage coefficients or factors (e.g.

 

216

for crop damage by season, stage and duration) are
applied across the Basin without regard to changing
geographical or type of farm conditions. If a land
value check is to have validity, it must be based on
actual conditions on the land to be protected, as this
study illustrates. Therefore the need, and expense, of

individual studies for each project must be recognized.

 

 

 

APPENDIX B

SENSITIVITY OF REGRESSION ANALYSIS TO

LAND USE AND GRADE DATA AVAILABILITY

If a land value model is to have the widest possible
applicability it should be based on data that is available
at reasonable cost. The primary data for this study were
obtained from public sources and were typical of the types
of information routinely available at the county level.
Thus, the type of analysis used in this report should be
easily repeatable in other states and river basins.

The only data used in this study that might not be
generally available are the land use and grade data (spe—
cifically, cropland grade data) obtained from the County
Assessors' Offices. It is not known how many states
follow Indiana's practice of appraising property by grade
but if this information were not available, and grade
data were critical to the analysis, it would probably be
necessary to use some form of on—site appraisal techniques
to obtain equivalent information. To test the importance
of the cropland grade information, regression analyses for
the Upper and Lower Wabash areas were made under the
assumption that only gross categories of land use (cropland,

pasture, woodland) were available. At least this
217

 

 

 

 

218

information should normally be available in the Assessor‘s
Office but, if not, it could be obtained by planimetering
aerial photographs.

When only the variable for percentage of cropland
was included in the regression equations for the Upper and
Lower Wabash (percentage of pasture was non—significant in
both areas) the R2 were reduced and the standard errors
were increased. In the Upper Wabash the R2 for the upland
observations analysis was reduced from 0.66 to 0.56 and the
standard error of the regression raised from 92 dollars to
103 dollars. In the Lower Wabash the corresponding R2,
for the upland analysis were 0.58 and 0.54 and the standard
errors were 101 and 105 dollars, respectively. Thus, the
effectiveness of the analyses in estimating per acre farm
sale value was reduced (more so in the Upper Wabash than
in the Lower Wabash area).

The effect of the lower precision in the estimating
equations was to reduce the effectiveness of the procedure
used to adjust floodplain sales for comparability and thus
to increase the residual differences attributable to flood
risk1 (Tables B—1 and B—2). The residual differences by
categories of flood risk (part B of each table) were also
larger for most categories and the standard errors were

increased.

 

1This is to be expected because if no other information
were available only the differences in average price per acre
for floodplain and upland sales ($104 in the Upper Wabash and
$47 in the Lower Wabash) would be available as an estimate of
the flood risk discounting.

 

 

 

 

 

219

 

.mposso pampcwpm ppm mcmcapscmmm CH mm m
9: as

 

 

 

 

 

III/HI!
Amo.mav ANN.HHV
H.mm- .am-
w mQOpr>pmmno mw>flm nmwpmz mm
nom.m_._uv Amw.omv
sm.m + mm.HH+ w
Emcppm mow m
H
Akuanv Amm.OHv
Ho mm] no man :mma tough moamm camfiooooam mm
II[Nillllllllllllllllillll
Hmflm std. Hddvfimom mes/w Mom .m
wm.mfil oa.oal Mmflh Uooam on cannuSQthpw cocohompfio
mo.amm mo.amm mmamm eflmaoeooau om .moaad mwmhm><
om mmmw ma Hmmm mozfim> pcmaas
pm mcamm camadpooam po cease pwpmsflpmm
mfimzamc< cpo< pom coasm mmmsm>< .4
momma
gopo oz coopm Qopo EcpH

 

“mmh< swans: soda: .mpmm compo pstQopo
psogpfiz pew nus: mQHQmQOHpmHmm osam> posh ess xmflm cooam popmsflpmMIl.Hnm mqm<e

 

 

 

220

 

.wposao pthQMpm who mcmcspCmst CH mopswfi
m

 

IIIIII‘III$

 

 

H
A:N.QHV 7IlIIllIIllIllIIIIIllllllllllllllllllllllll

 

 

“$.03 .
Fm.wzl NH mm! mwmpCOM
. h 02 NH
gems 39%”-
mm HmHI aw. mwwucotflm ma
mm.mm| mm mm! woadw mc>oa 02 mm
AmH.mHV Aamnaav
w©.N. + mm m + mmfiwm UmpOGUOhQ ®m>mnfl mm
Am:.mav AHo.mHV
ww.mml mm.aml Amow mmmav moamw Camampooam ow
I; a
mm.o:l mw.mml xmfiu pooam op wapmpsnwpppm cocmsmMMHo
mm.mmm qmlmMI modem £33003 i. 63.3 emote:
s:.mmmw om.mmmw mosam> pcmaas
pm moadm cfimHQUooaw mo mafipg ooumsflpmm
mflmzamc< chug hog moanm owmhm>< .<
macaw compo mono EmpH
goso oz

 

.mos< gmwnmz mozoq .mpmm psmadopo

peoepaz sew ess: .mdaemcoapmamm msflm> sewn ess swam sooam empmsapmmnu.mlm mamas

 

 

 

 

221

It appears that the cropland grade data made an impor—
tant contribution to the analysis of farm sale value in this
study. The larger question is whether the contribution was
significant enough to warrant the cost of collecting the
data if it had not been available. There is no simple
answer to this question since the costs that would be
incurred or the importance of the data in an analysis of
another area are not known. However, this does illustrate
the desirability of integrating a land value analysis into
project justification studies at the time they are made.

The evaluation could then be made in the context of (a)

how the additional data Would contribute to the accuracy

of the land value estimates and (b) how greater accuracy in
the land value estimate would contribute to the accuracy of
the final estimate of flood damages or expected flood

control benefits.

 

 

 

 

 

NICHIGQN STQTE UNIV. LIBRQRIES
1111111111111111111l111111111111111
31293010616815