.‘VMJHMIWWA‘I‘. ...L‘ltﬂ -
Sauna kw... w . .
.pﬁ}.hu.¢\ ... .
.m. z... .. W..... .

. any?

        
   

    

                         

        

 

a .

      

u... Rwy-WNW}... .
.%...ﬁwﬁéh.,.a ..
. armmuwnn-srhhanhamau4ﬂ.
u\u........._...wmvmm....amﬁ wm...ﬁ$.....$ . . .
...... . .... inﬂux-..., . .a.
1‘ W..-v.33?! ......k. A4... ,....-....3..r. . .
.123- new-rmnuwﬁvwwwu. .rmwukWuvM . .. ,... . . _
.. ... I L'ﬁwl ..."...umul ti. -. .
Jr...» #6,...”

...T , .
.{wwvz .r

  

c

     

 

We .
Eh... ......

      

v

  

   

 

. 91.1!!!
I.
‘.

 

 

. a .... habit... .111. I.
. .hahlﬂh.mmm.w.l. hummﬁﬁﬁ. mnﬁ... ......

1...
... n - . -..? ...-...... ..

.- 2-2-9. . . .. . Z . .
...»...r... . .. . . . . . .. . .. .
...4. 4 u . . .... . n . . .
x \ . - .. u . . . . . . , . u. . r . . .
.. n. .. . . .. . . f . ... . . . .... 7:9. .. ..-..t. .. . 1
. .. . .. . . . . . .. . . - . . . .. . . . ....-.. r.l.un.U...r-.Ml. . ...... . ..
u . . I - . I n .. n . . u A ’5 ..-.‘ I6V O§\l ...le . . . .
. .. .. . : . . . .. tt... 11.14:; PI It .11.. .‘lho .0! . u . .. .
. f . . . .. . . ,. ,, . . . . . “...“...ﬂwl. .. -... o .255... ... . .. .h . . ... .... .
.... . ... .. .. . .. .. . - .. 4.... ...... .. -... du 5!.

  

 

It! 99": .no‘nl 5.3.“. l f
.. . «ﬁfﬂ n I It ‘1-
“.- .......ﬁ.....u.........a...nm~um.....

, IV". .. 255.
km... 3%... g. ...».mmmw...

....

 

 

                 

 

    

\I In I ....

.. .li.

    

     

    

. . . . . ,. . . mduﬁhammmug . .3
, v.. ”9mm! hum. ......F......o.....
. .. . .. . . . . an..-” "mild-w. ....u...................m.n..w
.. .. . . .. . ... .-..va. .. ...-.... «3......aLvr... “We... . ......d. -
. . . .. ... . .amub..ﬁu-&.ﬁ.~w.l..w.ﬂ. ...»... .. .. .
tuna

      

 

       

 

.
.. . -
.. ..x. .... I” ....n . .... ..
. . ..wm ..KWEW. ......aﬂﬁnuﬁtﬁlum- \inwﬂws.
. . «we... ...... - -.rwﬂﬁﬁﬂi. F4.!............-,..-....1.\.Z ..
. . . . . ..F. . a - m” ..J...r..ki....nr..u..!...-J..$.
. ._ ......muwu .. ..-...mnn. ..

.521- :A. ... -1... v
...-Avg»... ....
...-...mmwrﬁmﬁmw-ﬁ... . , ...-......me .. .
. my1k.b.zhx.rv ..v............m.....r..n...w.....-. 41.3w...
.. . . .Mnauhxmtﬂ:
.... . inn... 5....

 

I-rJvﬁfdumun-hwnmm‘ 9.. ....A ......w .
.. .. "...-.34...-» .. 1...-.. . . ... 1.3... .. - . .. ,
lininﬁﬁvgﬁ: ......«VuVHM. tin-Km“. ...“...o... ... .. ...... , . .
. .. .. ..-...- .-.. - . ... .- .52... . , -.. ......
.....5M..:......u...\64......hl.s......... ...... .... ....wduwﬂﬁrn..: «......ﬂuhguﬂ. an»
43“. .110... I] I . .-.!1 In...» I v \It!.'§..t&l¥
. .. ......u...m. kWh-m... .. ......ﬁlnﬂ...kn.....m-d..........4...a..s...$....-yh . ... ..
6 1.. -Il!’ .. l . cial: I". .A! ......II. .
...-......m. . .. -..»...Muuuauuumnhnmhﬂwﬂﬂv-QMWﬁﬁszi ......
. .4»... .
.

        

 

      

   

 

     

..

.. v

. .. x 4

...}- ...nu-l...-. ... .... . . é ....

.-. .. u I. In . (.... A 0| 0! lb- valn..kc.p n‘ ixol ....

nu... t54?.1rus..u& Frau... ...-W..... ...vuﬁ-MWWU. . ......h .
1"! II. 9“ \

h1.§rﬁh.n..m.hwauhumlnw 5..."... ...

 

a
a

 

     

        

 

-1...
involoﬂrgvoi sin-III. 1i...
,2 u-‘f...-.. .... . .. ..- ..... . a...
. . 3.1.4 giﬁth.-:.-£.. ...-«....-kab .. ..-. . ... . .
LI). 1.... H ..-....c--. ... ...“...Jnﬁo3 ...
. .‘vuh-AI l.’ .71 8....‘Ao..7 l. nuuu ......‘x. I. 1
I. ... . . .riﬂﬂ]...;.-..P ......l‘l-.. ...»... ;
.. .. .. w. . .... .. ...)rﬂn...c...w... 3R-...-..v Id- . . ,. .
.-..vl 0. “M .I ..ftohd‘.ﬁ’ﬂ.ﬁx.'v. ”.1.“ . . .
.5... u.....- ...nw.....l -.. . .- .
(- ...-5 - , .49...- - . ...... ......
...OLFX-f . a .(r..ﬂ..-....l.

 

. ...... n. ...”. 1...... .....-wwg ... . ...
: It... 13.... .....PI... 1!.“- ..|......v... 1:4. . .
.. 16.1%....{p3-ﬁﬂ ..-.. 9.5.... --..s .. ...-.5?" xii-.....r...
. . ..I.........- No.3»... . .3

       

 

                                                 

 

       
       

     

       

    

 

       

1. - .. ....
. . . . .... ....
.... . £1..-....hm...ﬂ.-....$..-...}....-:1... ..u. ...»...293 . . a
i -. ......»- - ...-......w- ..vot! 51...... “.1 . -.QHVA... ,
. .I- 3.3.x no.3..4Jlnviu-wt.ﬂ.rn.- aka-.18.»... i1a41..u\h..§s.n... - 5...... . .. . ...-...“ .... ...-......
. huh-Luv Hmhh-h‘cﬁdf u...t'§o.il!t4....v .Oﬂlnvelrvu \ .slrg‘uu‘a o...- u‘. .193. l-xl ”...”.an A. ...-2“}. .l I.“!.’. .
...... ...... ....l...2...........-......, ...W..-......aux-04......«...-.1131.-.. {bf-...}... .. . . J... ...
119.3... 3.... 4.1-2...- .9 ....-. ..-Uz.........li.:..-......«1...... . .....- ....
z I.n.uh...n..s...m......-.w-..s...... ”kl-...... {-..-......uuwhw-s1...»...v...r....9.£2. .. ...-.9. .
L r- u-ive.....u.7....-.......3. wk... #33... ......a...u.. ...-... .....k....-.-..-....I. .24.“.- .
.V- .l ...-It... l........-.5.....ﬂ.i.-«. ......N........................i..u... ....n! ......u...-.. -...-. . . . . .
.....u-«HLJ.5......Q-... -e............mm.n. 1.3.1:”... ......f.-...¢lo ...-.... . .
... .- .de . . ‘2. hi 0. [Eh-loo}.-.l\is\t-.Qv-h. ...... .-LU.
. yr 1 HF - .M‘ ' 3 I. t '1 .’I- n. it!” 12:“ .\..lt\-- III- .
t.ﬁ‘t.i“.lll\|1\iiotv%l.' \09 iou.|.§i-C-a vi, ...r V.«)|...lv.|o. n
. ...r... $..u..4.a...ﬁ...«. . .... .n. 1 .43.}... .............c............. ...n..-....-
111. ...-.....x. ”Iii“.-. .24.... wdﬂrvm . ﬁnk-....-
P-... -..-.- “cum: 3&2»..- ... A to... ..- ....un...
.21.... .. .-.-.. . .. .. n. $1....-
. ...-(.03...
. .. .1: .
.... ...-.-
..! . .t. ....
- . A.

   

I
g:

\
‘-
v‘
3
s
3:
r
9
h

    

3
{
g.
3:.
. $2
‘ 3
i
=3
.‘
.-
M
J
i

          

4
‘3.
M.

 

0 .IllaI.I.0 .
x

. . ..

L . 2- .AI . .

.. - ol- Ov‘.tt v. .
. . utl‘.‘ ‘11.. In
u”\".\.¢. .

     

 

 

       

     

                           

      
 

   

     

        

    
      
 

       

      
   

    

1 .n- 7 14
...... aw. "H...-
.-I..\..I|-|I. .lo- «M . .
.. . .-.- . . . |., . 0.. . .
. .l!‘ v... . . I’ﬁli. 7r {U1n‘fy-AI.... .uV‘N’. . ,
lib-ﬁn.- ......H-}...u..........x..-. . -..... . . .3 - . .. ....l-oJHI- . ..lt...x.-...-:.l ...-...-..l.l.-.(-.. ...-.- 1......
I. . )l. .I..l.|..l .\. .1... t.- ...- .- ‘--I- -.. ‘ll {...}: lol- ...-“ha.“
0. Iosv [(45013 ..t-I. -vLA..Jv.I|O. . ... ....Is‘ltu..nlr. .J-pw 1:".1- OIAIuL‘Hl....vOr>.\.-.l. vtl? {‘{IP‘
....f.-....||.. UV.\(. 3.. 9-1.... N‘s-{\i ...... Io..ill)!4v-\-...bl..«."“.t11|f\!
.. tr .. , , .\ 1- .. lbs-It...u.lxn.s.IL\|“(..tll/ll‘1lvvﬁhu"Huntlﬂbri‘vs‘r‘
. . a .v. ..Utlr-OINHKLCLI..- a. ...... {I'll--11.... 9 I)“. .-
.v... 0" .- .. .....Il--...I..\..: Ill-... I .0)-
v . . -.l.i»-..I\.-.IIIII .v.\...- l...- .I‘ 1‘2"; 7?}-
. . I.” .lalu‘s..v.|ci-l I...Yt.-t.v!l-l. .13...
- ..I....v-l-..|Y..--I.TO {...-.1
.I-l‘l. 1...-.. ...) ....l.|...'.......lf-...
1.04.- t... -... .
...l-I.l-I\...l 1|..- .-0. .55....
.V.. . iokﬁuvllwl. ...». ... . 4.1.“. ..v (.10., .
in....I-.\r ...rr.v.rw\lo.h11|-\00-\.I .
..CII-v-t.-..?..l0‘¢£..p\ .--!‘v-V J
i1-..’t§...-tt:.-|)\.|-I- ...-.91
{ii-.....- .. I..- ....
- L... (...-..-..I

         

               

  

          
    

        

-..-...! - - . .- ....
- ..t-l!..-..:...l.. .
...-..- -i-v- ...--z. .
(......u...-..-......N.--.........la..t......-..... . ...-.....h...-......-.... .. --..-.
-.....nr... at In.-.lol.. .o .. . .. . .. . .’| 1|Inr-hl\LI..--.. .I.
- ..(-..-..-Y-t-..-.-r....u-. . . . . . - .. ...-4 :-...-....I...-...-u-I-
. ...-....Y... . {-..-...... 1...... x . . I .\ ......- l.--......-...l..-I.!-l ....l
n . . - - .Altihn\o|t.xll|l\l.t 5.1 .
l...lr-...-.--......!.-.. . .
---. ...... Iv... n ...!tP..-..-!.,.KI-.u.... ......L-
..I.--.-. - l. . O... - n- ..l-... ...-..9.-¥-.-(l...-......I. ...-...: .... ....- v.
- f .. -.. - . -..-..- ...... . -. t-.. -N.. .....\-.I!.
f..... ..- 3...? ..-....- . ..Kl. --.... L...-. lei-31.1.1: ......l..... -
3 t . -..; ..bh- I! It...“ i .F .. ..lc't'! v .DIIVQIUDIUb-b-Iﬁgnﬂ. in.
. .- .. .....I. 2 i.- . .... . -- -. -.. . .... . ...tEl ... .l-uiii.
. .... . . - . .9. .. .. -E... CI .0 ti“ til-9..- -.O...‘ [9... 1.....- ......- ' ...-.....ltl.).1(o.o.|lbl.\0 0!
. -. . . . . Paar. -1-vvuutsox..l|!-v.\Uu~.-..II¢i-tgv b»? . . . . 4-9....101! .I1l-I...“ ....1.‘i!..s.r.o)..lv)b.| I. ...-v
. .. I , 2 if .. . ll... - ...... .-- ...-It .... -...-1-.. -..)...3...!... ...-.-.. .... ---... ... ...i - ..|-. .
. . .. -. .- ..--.1 .....nit...-..............«..>...-...r...-.ﬁ 5-1304..- .n-uhu.l.......-.l-.- ..\l........l,..... ...:tt‘buu........-...(.IU.IL..-....b ...........\...1....!..,-....-... .... «...-......[1 ......Ir... ...-1.1....
L .....n..lln..... -- . .E.xum..nv..o-..-vh¢...llr ...-... -..... ..b...-..-L..-:.....!... . «vii-5.3.1.. - .0231..- (...-t.
.. ... - 1. ...IO-ollollio. it'll.- .llb‘lt {tilt- . .Ibil-‘ .11.." .I. In! .... ..vv‘...\. .5..\. I. -
. 11.2.31... .. I....§-....-Iﬁ.........-1...ll.1 ...-..-.z ..-.-.L.-.:.. l...........-r- ...... luv-1.0.3311...)
...l.-. - .. -..... ...-v - I - I :11--- - - - .,
. -..-..ﬂ-dt-zﬁs. lt!---M.01..l..lav...l..ﬂ§.ﬂnft<.t .....nnﬂ. Suva...
...-3.13.3.3!!! -.\J.-..-.:-.-\--o-......unﬁuf ...-....V
.--...t...-:-..-... .......-..-..I-........ ... -..-......gl
1 . ......o-.-..I..i..5~0...p!....nﬁ. ..PIDPKﬂg-uﬁnl.’

. Ik‘lrll

. I». I ..
a - E» .z rlni ..
a 1L .

 

VELRSITY

IIIIIIIIIIIIIIII’I IIIII I III’IIIIIIII

3 1293 01570 7072

 

 

 

 

 

 

\\
Iv

This is to certify that the
thesis entitled

THEORETICAL VALIDITY OF CONTINGENT VALUATION IN A
HYPOTHETICAL MARKET FOR COLLECTIVE WILDLAND FIRE
RISK REDUCTION

presented by
Gregory J. Winter
has been accepted towards fulfillment

of the requirements for

M.S. degree in Forestry

IW

Major professor

Date June 19; 1.997

0-7639 MS U is an Afﬁrmative Action/Equal Opportunity Institution

 

 

 

LIBRARY

Michigan State
University ‘

 

 

 

PLACE N RETURN BOX to romovo this chookoui from your record.
TO AVOID FINES Mum on or More duo duo.

DATE DUE DATE DUE DATE DUE

       

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU IoAn Afﬁrmative AotloNEquol OppoMniiy Institution
Wanna-0.1

THEORETICAL VALIDITY OF CONTINGENT VALUATION IN A
HYPOTHETICAL MARKET FOR COLLECTIVE WILDLAND FIRE RISK
REDUCTION

by

Gregory J. Winter

A THESIS

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

MASTER OF SCIENCE

Department of Forestry

1997

ABSTRACT
THEORETICAL VALIDITY OF CONTINGENT VALUATION IN A
HYPOTHETICAL MARKET FOR COLLECTIVE WILDLAND FIRE RISK
REDUCTION
by

Gregory J. Winter

The beneﬁt of government programs that reduce home loss in wildland-urban
interface ﬁres depends on how residents value risk reductions. We estimated the value of
actions to reduce risk for residents of a Michigan jack pine forest via contingent valuation
(CV) and assessed the construct validity of the method by testing the relationships
between household characteristics and willingness to pay (WTP) for collective risk
reduction, and comparing these to theoretical expectations. The survey responses
conﬁrmed the expectations which were based on a two-stage decision framework in
which individuals ﬁrst decide whether to participate in the hypothetical market for risk
reduction, and then decide how much a given risk reduction is worth. Homeowners’
perceptions of the risk, and objective risk assessments determined the probability of
market participation. For market participants, WTP was related to property value and

household income.

ACKNOWLEDGMENTS

This thesis would not have been possible without the support and advice from the
following organizations and individuals. I would like to'acknowledge the generous
funding from cooperative agreement #239332 with Warren Heilman and Deborah Carr of
the US. Forest Service North Central Experiment Station; data and ﬁeld assistance
provided by Don Johnson and Duane Brooks of the Michigan Department of Natural
Resources. I am grateful for the assistance provided by Frank Lupi of the Agricultural
Economics Department at MSU who held my hand through the early stages of data
analysis using LIMDEP software. Lisa Greenfeld, Demetrios Gatziolis, Georgia
Peterson, and Gem Castillo, my colleagues in the Department of Forestry GIS Lab, all
contributed, in many ways, to this document.

I would especially like to thank Jeremy Fried, my advisor, for the opportunity to
work on this project and for the countless hours he spent with me despite his many other
active projects. Daniel Stynes and Eileen van Ravenswaay, the other members of my
graduate committee provided valuable guidance and insight into difﬁcult methodological
issues.

Finally, I want to thank the residents of Crawford County for welcoming me into
their homes and sharing their sometimes heart-wrenching stories about the devastating

Stephan Bridge Road wildﬁre. Without their assistance there would be no story to tell.

iii

TABLE OF CONTENTS

ACKNOWLEDGMENTS ................................................ . ................................................. iii
LIST OF TABLES ............................................................................................................ vii
LIST OF FIGURES ............................................................................................................. x
INTRODUCTION ............................................................................................................... 1
THEORETICAL BACKGROUND .................................................................................... 5
Expected utility theory ..................................................................................................... 5
Prospective Reference Theory ......................................................................................... 7
Sequential Choice Models ............................................................................................... 8
THE CONCEPTUAL MODEL ......................................................................................... l3
SURVEY DESIGN ........................................................................................................... 15
Willingness to pay for risk reduction .............................. 17
Risk perception .............................................................................................................. 19
Hazard adjustment alternatives ...................................................................................... 21
Hazard information, beliefs, and demographic information .......................................... 22
DATA COLLECTION ...................................................................................................... 24
Risk assessment ............................................................................................................. 26
MODEL ESTIMATION ................................................................................................... 28
RESULTS .......................................................................................................................... 30
DISCUSSION ................................................................................................................... 38

iv

Attitudes and beliefs ...................................................................................................... 73

Responsibility for wildfire protection ......................................................................... 73
Property tax belief ....................................................................................................... 73
Demographic variables ................................................. 74
Residency .................................................................................................................... 74
Education .................................................................................................................... 75
Age .............................................................................................................................. 75
Household income ...................................................................................................... 75
Gender ........................................................................................................................ 75
APPENDD( C - Samples, Stratiﬁcation, and Weighting ................................................. 76
1994 Sample ................................................................................................................... 76
1996 Sample ................................................................................................................... 78
Sample stratiﬁcation .............................. 79
Poststratiﬁcation weighting ........................................................................................... 80
Relative weights ......................................................................................................... 80
Property value adjustment factor ................................................................................ 81
APPENDD( D - Data Summary ....................................................................................... 83
APPENDIX E - 1996 Survey Instrument ....................................................................... 102
APPENDIX F - Survey Handouts .................................................................................. l 18

vi

LIST OF TABLES

Table 1. Indicator Variables and Operational Deﬁnitions ................................................ 16

Table 2. Conditional (p) and Joint (1t) Probabilities of Fire Loss by Risk Reduction
Action Taken. ..................................................................................................... 27

Table 3. Weighted Sample Means and Dispersiona ......................................................... 32

Table 4. Tobit and Cragg Models for the Contingent Valuation Decision Purchase

Collective Wildland Fire Risk Reduction .......................................................... 35
Table 5. Derivation of Poststratiﬁcation Relative Weights for Sample Strata .................. 82
Table 6. Derivation of Poststratiﬁcation Adjustment Factor for Property Value .............. 82
Table 7. Respondent Age ...................................................................................... 83
Table 8. Household Income .............................................................................................. 83
Table 9. Education Level .................................................................................................. 83
Table 10. Number of Occupants ....................................................................................... 84
Table 11. Intention to Remain in Area ............................................................................. 84
Table 12. Number of Pets ................................................................................................. 84
Table 13. Employmnet Status .......................................................................................... 84
Table 14. Marital Status ................................................................................................... 85
Table 15. Gender .............................................................................................................. 85
Table 16. Evacuation Experience During Stephan Bridge Road Fire .............................. 86
Table 17. Took Steps to Protect Home ............................................................................ 86
Table 18. Steps Taken to Protect Home ........................................................................... 86

vii

Table 19.

Table 20.

Table 21.

Table 22.

Table 23.

Table 24.

Table 25.

Table 26.

Table 27.

Table 28.

Table 29.

Table 30.

Table 31.

Table 32.

Table 33.

Table 34.

Table 35.

Table 36.

Table 37.

Table 38.

Table 39.

Table 40.

Wildﬁre Experience ......................................................................................... 87

Property Value .................................................................................................. 88
Objective Estimate of Initial Risk .................................................................... 89
Fire Insurance ................................................. 89
Type of Insurance Coverage ............................................................................. 89
Insurance Cost .................................................................................................. 89
Sources of Information About Wildﬁre Impact ............................................... 9O
Wilﬁre a Home Choice Factor ......................................................................... 90
Government Actions ........................................................................................ 90
Length of Residency ......................................................................................... 91
Seriousness of Wildﬁre Risk When Moved In ................................................. 91
Private Risk Reduction Action History ............................................................ 92
Past Risk Reduction Behavior .......................................................................... 92
Fire Protection Investment ............................................................................... 92
Government Actions Suggested ....................................................................... 93
Perception of Unconditional Risk of Wildﬁre in the Neighborhood ............... 94
Perception of Conditional Risk of Wildﬁre Destroying Home ........................ 94
Implied Joint Probability of Fire in Neighborhood and Home Destroyed ....... 95
Risk Reduction Preference ............................................................................... 9S
Hazard Salience ................................................................................................ 96
Attitude Toward Taxes ..................................................................................... 97
Responsibility for Wildfire Protection ............................................................. 97

viii

Table 41.
Table 42.
Table 43.
Table 44.
Table 45.

Table 46.

Attitude Toward Government Spending on Crime .......................................... 98
Attitude Toward Government Spending on Wildﬁre Protection ..................... 98
Policy Preference .............................................................................................. 98
Willingness to Pay for Private Risk Reduction Actions .................................. 99
Willingness to Pay to Restore Risk to Current Level ..................................... 100
Willingness to Pay for Public Risk Reduction Actions .................................. 101

ix

LIST OF FIGURES

Figure 1. Conceptual Model of a Staged Decision Process for Purchasing Collective

Wildﬁre Protection. ......................................... ' ................................................. 13
Figure 2. Study Area: Crawford County, Michigan. ........................................................ 25
Figure 3. Distribution of Property Value .......................................................................... 88
Figure 4. Length of Residency ......................................................................................... 91
Figure 5. Current Seriousness of Wildﬁre Threat ............................................................. 91
Figure 6. Perception of Unconditional Risk ..................................................................... 94
Figure 7. Perception of Conditional Risk ......................................................................... 94
Figure 8. Implied Perception of Joint Risk ....................................................................... 95
Figure 9. Willingness to Pay (in money) for 1St Private Risk Reduction Action ............. 99
Figure 10. Willingness to Pay (in time) for 1St Private Risk Reduction Action ............... 99
Figure l 1. Willingness to Pay to Restore Risk to Current Level: Initial Bid ................. 100
Figure 12. Willingness to Pay to Restore Risk to Current Level: Final Bid .................. 100
Figure 13. Willingness to Pay for Public Risk Reduction Actions ................................ lOl

INTRODUCTION

The emergence and acceleration of the Wildland-urban interface (WUI)
phenomenon, where homes are built adjacent to or within vast tracts of ﬂammable
vegetation, has greatly complicated the management of Wildland ﬁres. The traditional
tactic of constructing a fuel-free line around the ﬁre as rapidly as .possible becomes
impractical when homes and lives are immediately threatened. During a typical WUI
incident, ﬁreﬁghting resources are diverted to protect structures, thereby reducing their

effectiveness in containing the ﬁre.

Fire planning and‘efﬁciency analysis are also affected. Traditionally, Wildland
ﬁre managers have relied on efﬁciency models to minimize the sum of ﬁreﬁghting costs
and damages. The models assume that a) values at risk are distributed uniformly over
large areas (e.g. forests of some age class and stocking) and b) that prices are available to
guide the assignment of per acre loss values in the accounting of ﬁre. losses. Both

assumptions fail at the WUI.

Although state and federal ﬁre protection organizations are compelled to respond
to Wildland ﬁres which threaten residential developments, little is known about the end-
user demand for such protection. Individual homeowners can reduce ex post welfare
impacts through insurance markets; however, due to moral hazard and unreimbursed
damages, these contingent claims markets cannot fully compensate ex post losses.
Although investing in self-protection (physical modiﬁcations or changes in behavior)

reduces risk ex ante, many homeowners fail to take such precautions (Rural Fire

1

2
Protection in America Steering Committee [RFPA], 1994). As a collective remedy for
market imperfections and the misperception of risk by individuals, government risk
reduction programs can be made more efﬁcient if more is known about individual risk
reduction choices and ex ante economic values (Shrogen & Crocker, 1991). A valid
efﬁciency analysis of ﬁre protection and the initial attack system in the WUI context
requires valid estimates of WUI home protection. Such estimates are difﬁcult because of
both the dispersed nature of WUI development and the signiﬁcant non-market

components to wildﬁre damages.

The contingent valuation method (CVM) is one promising approach to obtaining
estimates of the ﬁre protection value at the WUI (Fried, Stewart, and Gilless, 1994).
Residents can be asked to state their willingness to pay (WTP) for state-sponsored
programs which reduce the risk of their homes being destroyed by a Wildland ﬁre.
Examples of such programs include improved ﬁreﬁghter training, acquisition of
specialized equipment, and ﬁre safety education programs aimed at homeowners and

tourists.

The literature contains no examples of the application of CVM in the context of
collective protection from damage by Wildland ﬁres. The novelty of this application and
the potential for bias when CVM is improperly designed or administered (Mitchell and
Carson, 1989) motivated the construction of hypotheses to test the validity of CVM
responses. The usefulness of study results to policy makers depends on demonstrating the

validity of the approach.

3
One method of testing the validity of CVM results is by determining the extent to
which the WTP responses conform to theoretical expectations. Signiﬁcant regressions of
WTP values on independent variables bearing some theoretical relationship to the
beneﬁts of the good are one form of validity conﬁrmation (Mitchell and Carson, 1989).
Variants of expected utility theory and Kunreuther‘s (1978) sequential model of choice
provide the theoretical framework for a conceptual model of the consumer decision to

purchase wildfire risk reduction.

Mitchell and Carson (1989) distinguish three types of validity tests applicable to
CV measures: content, criterion, and construct validity. Content, or face, validity is a
subjective assessment of the degree to which a measurement technique is an adequate
representation of a construct’s domain. In CVM studies, content validity tests may be
applied to the hypothetical market description. For example, in this study, a joint
probability model of destruction of homes by Wildland ﬁre was reviewed by ﬁre
management experts to assess the degree to which the description of the good adequately

described the risks presented to respondents.

Criterion validity compares one construct’s measure with another measure which
is thought to better reﬂect the underlying construct. Simulated or real market values are
sometimes used as criteria against which CVM ﬁndings are compared (e.g. Bishop,
Heberlein, and Kealy, 1983). Because CVM is typically used to measure the value of

goods for which no market exists, criterion validity tests are relatively rare.

Construct validity concerns the degree of association between measures as

predicted by theory. CVM can be evaluated using two types of construct validity:

4
convergent and theoretical validity. Convergent validity tests compare CVM values with
other valuation techniques that attempt to measure the same construct such as comparing
revealed preference measures with CVM measures for the same non-market good (e. g.
Carson, Flores, Martin, & Wright, 1996). Theoretical validity is the degree to which the
observed relationship between measures matches the relationships predicted by theory.
Internal consistency, one test of theoretical validity for surveys, compares survey
responses with theoretical constructs, or patterns of responses to different questions using

the same survey instrument (Diamond, 1996).

When CVM estimates are used to inform policy makers, they should ﬁrst satisfy
some test of construct validity (Mitchell & Carson, 1989; Whitehead, Blomquist, Ready,

& Huang, 1997). Mitchell and Carson comment on the importance of theoretical validity,

Whenever contingent valuation studies are designed with the intent of gathering
data for policy purposes, it is highly desirable that they take into account the need
to produce theoretically based regression equations or comparable evidence of
their theoretical validity. . .(p. 207)

Because this research is the ﬁrst application of CVM to wildﬁre risk reduction, testing the
theoretical validity of the ﬁndings is a necessary ﬁrst step toward developing a

meaningful beneﬁt estimation technique for wildland ﬁre managers.

This paper describes a conceptual model of the homeowner decision to purchase
collective protection and presents analyses based on this model that assess the validity of
CVM-based beneﬁt estimates for a study of ﬁre risk reduction valuation in northern

Michigan.

THEORETICAL BACKGROUND

Under the assumptions of expected utility theory, utility-maximizing individuals
uncertain about the future state of the world (i.e. whether 'or not ﬁre destroys their home)
attach value to avoiding an unpleasant outcome or reducing its likelihood. The value can

be expressed as willingness to pay for the risk reduction ex ante (Freeman, 1993).

Exgcted utility theog:

The standard model of expected utility as applied to this problem is expressed by
Equation 1, where each term represents a probability weighted utility given a future state
of the world. In the ﬁrst state, utility (expressed as an indirect utility function v) is a
function of income, y, adjusted for a loss term, L, which represents the value of assets at
risk. In the second state, income is not adjusted (because no loss occurs). Each state can
be associated with a probability of occurrence: 1: and (l-It). The amount that individuals

would be willing to pay ex ante to completely avoid a loss is the option price, op.
E(v)=7t-v(y-L)+(l-7r)-v(y)=v(y—op) (1)

For hazards such as wildland ﬁre, where 1: can be reduced but not driven to 0, the
indirect utility function can be modiﬁed and the option price, op’, representing the value

of risk reduction from It to II’, where rt’ >0, can be obtained by solving

n-v(y-L)+(1-n)~v(y)=n'~v(y-L-0p')+(1-”')°”(Y‘0P') (2)

According to this model, willingness to pay for reductions in the probability of a

6
loss increases with initial risk, It, and the value of assets at risk L. Taking the total
differential of equation (2) and setting it equal to zero shows the marginal value for

changes in It

3);: v(y)-v(y-L)
dn’ [Ir-vy_L+(l-7r)vy]

 

(3)
This result is positive, indicating a positive option price which increases with the
magnitude of the reduction in It (Freeman, 1993). In the wildland ﬁre context, forest
residents at high risk for damage (e. g., those whose property is covered with flammable
vegetation) would be willing to pay more for an increment of risk reduction (private or
collective) than residents at low risk (e.g., those with ﬁre—safe landscaping) if asset value
and marginal utility of income are held constant. This simple model assumes that
decision makers know and adopt the true probabilities of future states, that they know the
value of the potential loss, and that they accurately calculate the utility level for each
prospective state of the world. But many researchers report that the standard model of
expected utility maximization does not explain the observed behavior of individuals

facing uncertain outcomes.

The most common observations are that people (1) tend to be oversensitive to low
probability events causing them to express high bids for risk reduction that
overcompensate for the actual risk level they face, and (2) they are relatively insensitive
to high probability events (Schulze, 1993; Viscusi, 1989; Kahneman and Tversky, 1979).
Slovic (1987) proposed that people develop their own assessments of risk based on

characteristics of the hazard such as controllability, voluntariness, catastrophic potential,

7
and degree of outcome uncertainty. So one serious problem with the expected utility

approach is that it is unclear how to represent the risk term in the expected utility model.

Prosmctive Reference Theory

Rather than abandon expected utility as a model of behavior, psychologists and
behavioral economists have incorporated a model of learning into expected utility theory
to explain the apparent “irrationality” of decision makers. Viscusi (1989) proposed a
variation of expected utility theory that predicts many of the aberrant experimental
phenomena that the standard expected utility model cannot explain, especially the
apparent overweighting of low probability events, the underweighting of high probability
events, and the reluctance to purchase probabilistic insurance (Viscusi, 1989; Kahneman
and Tversky, 1979). Under this proposition, decision makers continuously update their
assessments of the probability of a hazard as they are exposed to new, relevant

information.

The prospective reference theory of risk perception formation is used to modify
the standard expected utility model by more accurately describing the probability
estimate. The updated or posterior risk perception P described below by Equation 2
replaces It in Equation 1. P is a weighted average of the prior risk perception p, and an

objective or technical risk assessment It (Viscusi, 1989).

i’(p)+‘57t

P(P)= +§ (4)

The scale factor 7 corresponds to the informational content of the individual’s prior

8

beliefs, and if represents the informational content of new, objective information. The

ratio % is the relative informational content of prior and subsequent exposure to a

hazard.

Se uential Choice Models

 

Although such reﬁnements address the representation issue with respect to risk
perception, other problems with expected utility remain. One example is a fundamental
and unrealistic underlying assumption that people’s actions are based on their explicit or
implicit multiplication of utilities by probabilities (Kunreuther, 1978). A sequential
model of choice for insurance based on the notion of “bounded rationality” represents
another approach. Kunreuther postulated that people make decisions based on relatively
simple models of the world which they construct using their limited cognitive abilities. In
the case of disasters, insurance decisions occur in a three-stage sequence: 1) is the hazard
considered a problem; 2) is the person aware of the option to purchase insurance; 3) does
the individual purchase insurance coverage. It is more likely that individuals actually rely
on simple heuristics rather than on attempts to optimize some abstract utility function.
For example, individuals may ignore the consequences of a certain hazard because they
perceive the likelihood of that event to be less than their threshold of concern
(Kleindorfer & Kunreuther, 1988). Or, the decision may be driven by their familiarity
with the hazard. Kates (1962) proposed that individuals are “prisoners of their own
experience,” and that they decide how to respond to the prospect of future disasters by

relying on their experiences with past disasters. Others term this phenomenon

9
“availability” based on the hypothesis that individuals judge the probability of future

events by the ease with which similar past events are recalled (Tversky and Kahneman,

1973).

Thus, hazard awareness and hazard experience can be thought of as components
of hazard information that inﬂuence risk perception and, ultimately, decisions regarding
hazard risk reduction. Literature addressing the WUI problem suggests that wildland
residents with different sets of hazard information have different perceptions of the ﬁre
problem and would likely provide different WTP responses when confronted with a
contingent market for risk reduction. In one California study, residents of two ﬁre-prone
areas were surveyed. Residents of one of these areas had recently experienced a series of
ﬁres; residents of the other had not. The study found that high awareness and early
awareness predicted these residents’ relatively high assessed probabilities for both
wildﬁre occurrence and personally threatening wildﬁre occurrence (Gardner, Cortner, &
Widaman, 1987). Increases in ﬁre hazard awareness over time were measured among
residents of the non-affected community. As homeowners witness ﬁres in nearby
communities and are exposed to media coverage of such events their awareness of the ﬁre

problem may increase, causing them to modify their perception of ﬁre risk.

Others have observed the relationship between risk perception and proximity to
recently burned areas (McKay, 1985). Presumably, distance to recent events can serve as
an indicator of hazard awareness. The accumulation of hazard information and the
concurrent upward adjustment of hazard awareness and risk perception conforms with

Viscusi’s model of consumers continuously updating their assessment of risks based on

10
prior beliefs and new information. The investigators proposed that direct experience with
ﬁres had a dampening effect on hazard awareness and risk perceptions of ﬁre-affected
residents who may have based their assessment “on the adage that lightning does not
strike twice in the same place” (Gardner et al., 1987). The danger in this thinking is that,
due to regrowth in burned areas, the ﬁre risk rebounds rapidly relative to the perception of
risk among those affected by previous ﬁres (Cortner and Gale, 1990; McKay, 1985).
Though the rate of ﬁre recurrence in burned areas of Michigan’s jack pine forest is
probably not as high as in those areas studied by McKay and Gardner, ’ the effect of risk
experience on willingness to take precautions may be an important determinant of the

wildland residents’ role in reducing the wildﬁre risk.

The occurrence of ﬁres in and nearby ﬁre-prone areas may erode support for
public ﬁre programs and willingness to take precautionary action because a large number
of affected people no longer see the threat as immediately serious. On the other hand, ﬁre
occurrence may bolster support for fire protection programs among residents who are
made aware of the ﬁres but are not directly affected by them. The latter hypothesis is
consistent with literature summarized in the previous section which proposes that hazard
experience tends to increase a person’s assessment of the likelihood of future hazardous

CVCIIIS.

 

' Researchers investigating another major jack pine wildﬁre near Grayling (the Mack Lake ﬁre in 1980) have
determined that within the area burned, there had been ﬁve ﬁres larger than 10,000 acres since 1820, or one
large ﬁre every 28 years (NFPA, 1991).

11

Four classes of factors that are associated with hazard adjustment decisions:
1) prior hazard experience; 2) material wealth, which can be an important determinant in
the adjustment decision as wealthier people are “more likely to experiment with a variety
of measures. . .and they have the necessary pecuniary support”; 3) personality traits (e.g.,
“a sense of inner control”); 4) perceived role of the individual in a social group (Burton,
et a1, 1993: 119). Class 4 factors may be especially relevant in the case of wildland ﬁre
protection. Individual homeowners and government land managers have opportunities to
reduce risk; therefore, the resulting mix of public and private risk reduction will depend
on the perception among homeowners of who is primarily responsible for wildﬁre
protection. The fact that individuals have several risk reduction mechanisms available to

them has important implications for the analysis of risk reduction beneﬁts.

Individuals can reduce risk through collective or self-protection and insurance
mechanisms. WUI residents can pay taxes to improve government ﬁre prevention and
suppression programs, they can alter their private landscapes making their property less
prone to wildﬁre damage, and they can purchase ﬁre insurance that would reimburse
them for ﬁre losses. Each of these actions can reduce the probability and/or severity of
wildﬁre damage. This implies that in the absence of opportunities to self-insure or self-
protect the payment for collective risk reduction would increase. In a series of controlled
experiments designed to assess the inﬂuence of alternative risk reduction mechanisms on
risk reduction valuation, Shrogen and Crocker (1989) found that self provisions of safety
are more highly valued than collective provision, and that the risk premium for collective

risk reduction can be negative if individuals are provided with opportunities for self-

12
protection or self-insurance. The existence of these alternative risk reduction

mechanisms complicates the task of valuing collective risk reduction.

THE CONCEPTUAL MODEL

These theoretical relationships and empirical ﬁndings give rise to a testable,
conceptual model of the homeowner's decision to purchase collective protection from

wildﬁre home destruction (Figure 1).

Market participation decision
... ........... f ..........

l Initial risk l-
| Risk perception l—J

Hazard information I—

Value decision

| Asset value I_
Initial risk l—L’ $

. 1
fl No I:: ,
0 '| _I

 

 

 

o
O
O
o
C
o
o
o
o
I
o
o
f

-

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Alternative 3 ° ............ ' I [Demographic l_
l protection I— : characteristic. 7
, I
Demographic I. :
f Charac‘c‘iStiCS - ——:'> Status quo selected I
I. ..................... J I

 

Figure 1. Conceptual Model of a Staged Decision Process for Purchasing Collective
Wildﬁre Protection.

Consistent with a sequential choice model, the decision occurs in stages rather
than simultaneously. Homeowners ﬁrst decide whether or not to participate in the market
for collective protection, then market participants express a value for the risk reduction

such protection provides. The conceptual model is also consistent with the variants of

13

l4
expected utility theory described above. According to this model, risk perception, hazard
information, alternative protection behavior (ﬁre insurance and self-protection), and
initial risk level determine the likelihood that homeowners will purchase collective
protection. Value of assets at risk and household income determine the amount that
homeowners are willing to pay for a given amount of protection. Demographic variables
including age, household income, education, and gender are also included as potential

explanatory variables for both stages of the decision.

SURVEY DESIGN

To test this conceptual model, evaluate the validity of the CVM approach to
obtaining estimates of risk reduction value, and gather data on the perceptions and
attitudes of WUI homeowners in support of decisions by ﬁre managers and planners, a
CVM survey was designed and implemented in Crawford County, Michigan. The CV
potion of the survey instrument was designed to elicit estimates of the value of
incremental reductions in the risk of home destruction by wildland ﬁre via a hypothetical
market in which a supplemental tax funds additional government investments in ﬁre
protection (Batts, 1993). Other questions generated independent variables to
operationalize elements of the conceptual model. Theoretical validity tests compare the
relationships between independent variables, deﬁned in Table 1, and the dependent
variables (market participation and risk reduction value) in the survey instrument with the

relationships proposed in the conceptual model (Figure 1).

15

16

Table l. Indicator Variables and Operational Deﬁnitions

 

 

Model element Variable Deﬁnition

Market WTP (Yes) 1 if homeowner bid positive amount for wildﬁre risk reduction, 0

participation otherwise

Collective WTP Bid Amount of ﬁnal wildﬁre risk reduction bid

protection value

Alternative Reducer 1 if homeowner has taken private action(s) to reduce wildﬁre

protection risk, 0 otherwise

Hazard information Survey96 1 if homeowner was interviewed in 1996, 0 if 1994

Hazard information Distance Miles (“as crow ﬂies”) to 1990 SBR ﬁre perimeter

Hazard information Damage 1 if homeowner has experienced damage to property fr'om
wildﬁre, 0 otherwise

Hazard information Numyears Number of years respondent has occupied residence for all or
part of the year

Risk perception WﬁreRank Homeowner wildﬁre hazard rank, 1 if most likely. . .4 if unlikely

Risk perception Firechance Respondent assessment of percent chance wildﬁre occurs in
neighborhood in next ten years (as decimal)

Risk perception Damchance Respondent assessment of percent chance passing wildﬁre
destroys home (as decimal)

Responsibility Responsible Responsibility for wildﬁre protection, 5 pt. Scale with 1:
primarin homeowner, 5 = primarily government

Tax attitude PropertyTax Belief about property tax spending, 1 = too much, 5 = too little

Demographic Education Education level, 1 if grade school...7 if graduate school

Demographic Age Age range, 1 if less than 25 years old, 12 if over 76 years old

Demographic Income Income range, 1 if less than $10,000/year, 7 if over $45,000 per
year

Demographic Seasonal 1 if seasonal resident, 0 if permanent resident

Demographic Gender 1 if female, 0 if male

Initial Risk InitialRisk Level of risk as percent chance that home would be destroyed in
ten years by wildﬁre (on-site assessment)

Asset value at risk PropertyValue Respondent estimate of property value in $1,000’s (or a township

assessment of property value) including land and improvements

 

l7

Willingness to pay for risk reduction

The dependent variable, willingness to pay for risk reduction, can be thought of as
a behavioral intention to take a precautionary action contingent on the hypothetical
market for collective risk reduction. Much of the debate over CVM concerns the validity
of respondent answers to WTP survey questions. When asked, “How much are you
willing to pay for a reduction in risk from It to 1t’?”, the respondent will usually provide
an answer, but the meaning of the response is often far from clear. To the extent possible,
the guidance provided to practitioners by researchers working on CVM methodology

(e.g., Mitchell and Carson, 1989; Arrow et al., 1993) has been incorporated in this study.

Four methods were considered for eliciting respondent willingness to pay for ﬁre
risk reductions: open-ended, payment card, dichotomous choice, and iterative bidding
(Batts, 1993). Over the past decade, dichotomous choice (or referendum format) has
become the most widely used method for valuing non-market environmental goods
(Arrow, et al., 1993; Mitchell and Carson, 1989). This method asks respondents whether
they would be willing to pay a certain dollar value for the good in question. A range of
speciﬁc dollar amounts are distributed randomly among the sample and respondents
simply accept or reject the offer based on the description of the good and the price. One
advantage to this method is that it resembles the situation a property owner might face
when asked to vote on a public project to reduce ﬁre risk. This is much different than

asking the same individual to estimate their maximum W'I'P for the risk reduction.

18
Recently, there has been a resurgence of interest in the open-ended CVM format.
Although the question formats for open-ended elicitation methods is less familiar to
respondents than for dichotomous choice, they elicit continuous data, provide more
information on respondents’ WTP, and are less prone to “yea-saying.” Dichotomous
choice elicitation has been found to consistently generate larger WTP estimates (Ready et
al., 1996; Boyle et al., 1996) and requires a much larger sample of respondents to achieve

the same statistical precision as open ended questions (Mitchell and Carson, 1989).

The small population sizes in this study (particularly for residents with loss
experience where N=47) and the high cost per sample element led to the selection of a

direct, open-ended WTP question format with iterative instead of dichotomous choice.

Respondents were provided with an estimate of their current contributions
towards ﬁre protection costs via property, sales, and income taxes, and queried for their
WTP for two levels of risk reduction via increases in annual property taxes: (1) from their
current, initial level of risk, It, to the next lower level in Table 2; and (2) for an additional
reduction in It to the next lower level. Households already at the lowest level of risk (1: =
0.04) were asked instead for their WTP for a reduction in It to one-half of that level (It =
0.02) and were not asked about a second risk reduction. Respondents were not told in
advance that they would be asked to value 2 risk reduction scenarios. Risk cards were

used to illustrate risk levels (see Appendix F).

The operational deﬁnition of willingness to pay for risk reduction was deﬁned by

survey responses to question 23:

19
Earlier we determined that the probability of you losing your home to ﬁre within
the next ten years was (initial risk) . Through a combination of public
prevention and suppression programs this risk can be reduced to (refer to m
show risk card 2 . Keeping in mind this action will only be taken if there is

suﬁicient demand from the public, how much would you be willing to pay each
year, in increased property taxes, for this risk reduction?

Initial risk is an objective (by domain expert) assessment of the risk (as a percent chance)
that a respondent would lose his or her home to wildﬁre in the next ten years assuming

that no individual or collective risk reduction efforts are undertaken.

Risk perception

Risk perception is the “intuitive judgment” of the risk posed by a hazard (Slovic,
1987). In this survey we sought a numerical estimate of the probabilities of occurrence of
two events which, together, comprise a joint probability of home destruction by wildland
ﬁre over a ten year period: 1) the unconditional probability, u, that a wildland ﬁre would
occur in the respondent’s neighborhood; 2) the conditional probability, p, that such a ﬁre

would destroy the respondent’s home.

Developing portrayals of wildﬁre risk and risk reduction that could be
meaningfully and reliably communicated to homeowners was challenging. Van
Ravenswaay deﬁnes risk perception as “an assessment of the chance of harm in a given
context” (as cited in Decker, 1995). Chance is determined with respect to a given
population over a specified time period. Risk assessments were performed for individual
homeowners, so the population is the household. Time period was speciﬁed as the next

ten years. Harm refers to the type of harm inﬂicted by the hazardous event. Harm was

20
speciﬁed as the destruction of the respondent’s home by wildﬁre.2 Context comprises the
conditions that result in the level of hazard exposure and that cause the reduction in that
level. Interviewers provided context by describing, in lay terms, the joint probability
model of home destruction by wildland ﬁre, reminding them of their opportunities for
collective and self-protection, and describing the speciﬁc actions and practices which
would reduce exposure to the wildland ﬁre hazard. Unconditional and conditional risk

perception were elicited with the following questions:

Which percentage best represents your estimate of the probability of a wildﬁre
moving through the block on which you live at some time in the next ten years?
[respondent has card with 0-100% range printed on it for reference] (Q 13)

If a ﬁre were to pass through your neighborhood, which percentage best

represents the probability of it destroying or severely damaging your home? (Q

14)

Another dimension of risk perception considered in this study is hazard salience.
Kunreuther and Slovic ( 1978) found that homeowners who failed to purchase earthquake
and ﬂood insurance were disinclined to worry about these low probability events. They
referred to homeowners’ limited time, energy, and attention to certain hazards as a “ﬁnite
reservoir of concern.” One method of measuring the salience or importance of a hazard is

to ask respondents to rank hazards in the order of the perceived likelihood of occurrence.

Higher ranked hazards are assumed to be more worthy of attention.

 

2The 1994 survey instrument speciﬁed harm as “. . .severely damage or destroy your home...” This was
changed slightly in the 1996 survey instrument to “. .. destroy your home. . .” To more clearly specify the

21
In this study, interviewers asked homeowners to rank 4 hazards in order of the
chance that each would occur. The events include (1) a tornado destroys your home, (2) a
car accident sends you to the hospital, (3) your home is burglarized, (4) a wildﬁre
destroys your home. Those who ranked wildﬁre home destruction high on this list
received a high salience score for the wildﬁre hazard and are, theoretically, more likely to

participate in the hypothetical market for collective protection.

Hazard adjustment Qematives

Individuals may not take precautions against a hazard because of a belief that the
community should provide the protection. This may lead to the community providing
some form of protective service for its citizens (Burton et al, 1993). Actions taken by the
government to reduce the risk of private property losses due to wildﬁre damage are
considered to be collective protection mechanisms. These include investments in ﬁre
suppression equipment, personnel training, and ﬁre prevention programs such as
homeowner education. On the other hand, individuals may believe that self-protection or

ﬁre insurance are more appropriate responses to this hazard.

22
The survey measured hazard adjustment alternatives by asking homeowners (1)
whether they carry ﬁre insurance 3 and (2) whether they had taken any self-protection

actions to reduce the risk of losing their home to wildﬁre.

The joint probability model of wildfire home destruction implies that indiVidual
homeowners have the opportunity to reduce their exposure to the joint risk, It, given the
unconditional risk of wildﬁre occurrence, it. Several speciﬁc self protective actions were
considered for reducing the conditional risk, p, of losing homes to wildﬁres: creation of a
defensible space by clearing ﬂammable vegetation, debris, and ﬂammable material such
as fuel and ﬁrewood to a safe distance from the exterior walls of homes; and retroﬁtting
homes with ﬁre-resistant building material such as metal rooﬁng. Interviewers explained
these self-protection options as well as the options for collective protection: additional
ﬁreﬁghters, upgrading ﬁre department equipment, road improvements for better access by
emergency vehicles, additional ﬁreﬁghter training, and educating homeowners about

wildland ﬁre.

Hazard information, beliefs, and demographic information
Hazard awareness and experience measures included length of residency, wildﬁre

damage experience, years since the 1990 Stephan Bridge Road (SBR) ﬁre (interview

 

3 Nearly all homeowners (97%) responded “yes” to this question; therefore, self-insurance is not included as
a variable in the model. The survey was also designed to elicit the degree of self-insurance as a variable, but
many homeowners did not know the extent of their insurance coverage resulting in high item non-response
for this question.

23
year) and proximity to the 1990 SBR ﬁre perimeter. The location of each respondent’s
home was recorded and the distance in miles from the SBR ﬁre perimeter was calculated
via geographic information system software (ArcView). Residents with homes inside the

ﬁre perimeter were assigned a distance value of O.

Homeowner assessment of property value was used as a surrogate for value at
risk. Ideally, al_l values at risk would be elicited from respondents, including monetary
estimates of items with great sentimental value (e.g. photographs) and expenses
associated with the inconvenience of losing one’s home (e. g. hotel and commuting
expense). However, those not having had the experience of losing a home would
probably ﬁnd it exceedingly difﬁcult to produce an itemized list and assign value to such

items in the course of a short interview.

Respondents were also asked about their beliefs regarding (1) responsibility of
homeowners and the government for wildﬁre protection and (2) the amount they pay in

property taxes. The latter serves as an indicator of protest bidding.

DATA COLLECTION

Following a period of iterative pre-testing and revisions, the survey instrument
was used to interview WUI residents in Crawford County, Michigan. Seventy percent of
this rural, forested county is under state or federal ownership, and jack pine (Pinus
banksiana, Lamb.), a ﬁre adapted species that is native to the area, is the most common
land cover type. Much of the public forest is managed as habitat for the endangered
Kirtland’s warbler songbird (Dendronica kirtlandii). The 1990 Stephan Bridge Road
(SBR) wildland ﬁre destroyed 76 homes, burned nearly 6,000 acres of public and private
forest (National Fire Protection Association [NFPA], undated), and left area residents
with a variety of direct and indirect wildﬁre experiences. Many who sustained ﬁre
damages were interviewed for our survey. A total of 285 residents were interviewed in
1994 and 122 in 1996. The SBR ﬁre perimeter, respondent home locations, and pine

forest coverage is shown in Figure 2.

24

A

County roads

/{/ Streams
N
i

SBR Fire

Perrmnent
Seasonal
Highways

Surveyed households

 

 

 

 

15 Miles

10

 

 

 

Figure 2. Study Area: Crawford Cormty, Michigan

26

Risk assessment

Interviews were conducted at peoples’ homes immediately following an on-site,
property risk assessment designed to estimate initial risk: the probability that a
respondent’s home would be destroyed by forest ﬁre in a ten-year period. This joint
probability estimate, 1:, was calculated as the product of u, an estimate of the ten-year
unconditional probability of wildﬁre occurring in the respondent’s neighborhood, and p,
the conditional probability that the respondent’s home would be destroyed if a ﬁre
entered the respondent’s neighborhood. Values for p, which are linked to homeowner
behavior were obtained via expert judgment of a panel of local ﬁre professionals (Table
2).4 The parameter p was estimated as 0.15 from GIS coverages of land cover and ﬁre
ignitions in the study area.5 The estimate of It obtained from this risk assessment
constituted the initial risk estimate used in the CVM interview which followed.
Interviewers also collected data on household characteristics, and indicators of the
conceptual model elements: risk perception, alternative protection measures taken,
attitudes and beliefs about government responsibilities, spending and taxation, and

willingness to pay for risk-reducing government actions.

 

‘ Post-ﬁre analysis has shown that landscaping and home construction characteristics are important predictors
of wildﬁre home losses in areas where wildﬁres occur (e.g., Foote, Martin, Gilless, 1992).
5 u=2acres burned by ﬁres igniting in jack pine between 1980 and l992/acres of jack pine in 1980 (10/ 12).

27

Table 2. Conditional (p) and Joint (It) Probabilities of Fire Loss by Risk Reduction

 

 

Action Taken.
Action taken Conditional probability (p) Joint probability (1t)*
None 0.93 0.14
Trees cleared 0.67 0.10
Grass mowed in fall 0.47 0.07
Debris free 0.27 0.04

 

*based on 11:0.15

MODEL ESTIMATION

A two-equation Cragg censored regression model (Cragg, 1971; Greene, 1995)
was speciﬁed to test the conceptual model of risk reduction purchase decisions. The
Cragg model is consistent with the underlying decision structure implied by the
conceptual model in which the latent dependent variable is the value of risk reduction to
homeowners. The value is expressed in two stages: the discrete market participation
decision and the continuous value decision. Zero values for WTP are treated as
unobserved market participation; therefore, the distribution of WTP values is censored
from below at zero. Zero values do not necessarily mean that the respondent places no
value on risk reduction. Some respondents believe they would beneﬁt from collective
protection but cannot currently afford to purchase it; others object to some feature of the
hypothetical market (the payment vehicle or the plausibility of the scenario). A
misspeciﬁcation bias arises from using a least squares approach to estimate the value
model parameters from an incomplete, or truncated, sample because the conditional mean
of the error term from the complete sample regression equation is not included as a
regressor in the equation for the incomplete sample (Heckman, 1976). The two equation

model corrects for this selection bias.

Cragg’s approach combines a probit model to estimate the market participation

function and a truncated regression model to estimate a value function which corrects for

28

29

the bias by including the conditional mean of the error term in the regression equation.6
Both models are non-linear and require maximum likelihood estimation of model
coefﬁcients. An alternative to the Cragg model is the tobit model which also accounts for
bias discussed above, but which assumes a different underlying decision structure when
applied to valuation problems. Assuming that the valuation decision is simultaneous
rather than staged, as is proposed in this study, the tobit model estimates only one
coefﬁcient for each independent variable; therefore, predictor variables affect the market
participation and value decisions simultaneously. Thus, the effect of a given variable on
the participation decision is of the same sign and magnitude as that variable’s effect on
the value decision (Greene, 1993; Blaylock & Blissard, 1992). The tobit and Cragg
models were both speciﬁed to test the hypothesis that the underlying decision structure is

staged (See Appendix A: Empirical Models for an expanded description of model

estimation and model comparisons).

 

6 The term added to the regression equation, A = (b/( l-<l>), known as the inverse Mills ratio, is a non-linear
function of the regressors, x, and the parameter estimates B (Greene, 1993).

RESULTS

The population of interest consisted of homeowners living in or adjacent to jack
pine forest in Crawford County Michigan, mostly in or near Grayling Township. Of 344
households contacted in 1994 (by phone or in person), 83% agreed to be interviewed.
Because Grayling Township Assessor’s ofﬁce records indicated that seasonal residents
were underrepresented in 1994, (23% in the sample versus 43% in the township), a
random sample stratiﬁed by type of residence (seasonal or permanent) was used in the
1996 survey to increase the proportion of seasonal residents in the combined sample. Out
of the 1996 sample of 306 households, 29 refused and 155 could not be contacted; 122
were surveyed for a response rate of 40%. Altogether, 407 interviews were conducted in

1994 and 1996.

Post-stratiﬁcation weights, developed to adjust the sample distribution by
residency (permanent or seasonal) and adjusted to correct for differential response rates
within six property value classes, were used in the data analysis. The sample frame
developed prior to the 1996 sampling was used as a reference source for population
distribution estimates (see Appendix C for the derivation of sample weights). Weighted

mean and standard deviations for the combined sample are displayed in Table 3.

Item non-response and removal of ﬁve WTP outliers, 7 reduced the useable data

 

7 Before estimating the model, ﬁve outliers were removed (WTP values greater than $500) because the
algorithm used for maximum likelihood computations failed to converge when these values were included.

30

31
set for multivariate analysis to 265 responses. Respondents in the unusable survey group
tended to be older, and reported lower household income suggesting an upward bias for

that mean risk reduction value obtained from the usable surveys.

Residents overestimated the unconditional risk of wildland ﬁre occurrence, [4,
relative to our initial risk estimate of 0.15. The mean estimate for u in the sample is 0.41
(Table 3). The mean estimate of conditional risk, p, in the sample (0.54) is remarkably
close to our estimate for the sample (0.55). The respondents’ perceived conditional risk
and our estimates based on the on-site risk assessments are moderately correlated (r=0.35,
p<0.001). Only 25% of the sample respondents were women. Median household income

was in the range $25,000 - 35,000. The weighted mean property value was $ 50,580.

Seventy ﬁve percent of the sample were willing to participate in the hypothetical
market (Table 3). The weighted mean annual WTP for collective risk reduction is $57.26
(median = $40.00). Independent sample t tests found no signiﬁcant (p = 0.1) differences
in mean WTP between 1994 and 1996 respondents and between seasonal and permanent
residents. The distribution of WTP values is characteristic of those obtained from other
open-ended CVM elicitation formats. It is bimodal with one mode at zero and another

within a bell-shaped distribution with a thick upper tail (Schulze, 1993).

32

Table 3. Weighted Sample Means and Dispersion“

 

 

Variable Mean Std. Dev.
WTP (Yes) 0.75 0.43
WTP Bid $ 57.26 $ 74.60
Reducer 0.74 0.43
Survey96 0.35 0.47
Distance 3.13 3.04
Damage 0.09 0.28
Numyears 15.75 12.40
WﬁreRank 2.75 0.97
Firechance (u) 0.41 0.24
Damchance (p) 0.54 0.30
Responsible 2.72 0.96
PropertyTax 2.40 0.95
Education 4.27 1 .5 1
Age 7.54 2.79
Income 4.96 1.94
Seasonal 0.42 0.49
Gender 0.25 0.44
InitialRisk 8.28 4.05
PropertyValue 50.58 43.44
($1,000)

N 265

 

3 Variables deﬁned in Table l.

33
Comparisons between the tobit model and Cragg model speciﬁcations for this data set
suggest that the underlying decision structure as a function of the conceptual model
elements is not simultaneous. Rather, willingness to pay for collective risk reduction is
composed of a discrete market participation decision followed by a value decision (for
market participants only). An underlying assumption of tobit is that the effect of each
explanatory variable on the decision to participate in the hypothetical market and on the
value of risk reduction is of the same sign and magnitude. The Cragg model is free of
this assumption and is more consistent with the sequential decision structure. If the
probability of a non-zero bid observation is not independent of the regression model for
non-zero bids, then the tobit model should be speciﬁed. If the tobit model is the correct
speciﬁcation, the probit and truncated log-likelihoods should sum to the tobit log-

likelihood.

The tobit model and the probit (participation) and truncated regression (value)
equations which comprise the Cragg model are displayed in Table 4. The probit equation
correctly classiﬁed 78 % of cases. A respondent is correctly classified if the predicted
probability of participating in the hypothetical market for public risk reduction is greater

than .5 for those who bid positive amounts and less than .5 for zero-bidders.

34

Using the likelihood ratio test, the null hypothesis that the tobit and Cragg models
are statistically equivalent, is rejected (x2 (“:13 = 109, p<0.1),8 suggesting that the
underlying decision process is, in fact, sequential and that the variables have different
effects at each stage in the sequential decision. For example, the tobit coefﬁcient for
property value is positive and statistically significant (Table 4); however, only the
truncated regression coefﬁcient of the Cragg model is significant for property value
indicating that this variable affects the value decision but not market participation. Initial
risk, conditional risk perception (Damchance), and property tax spending belief affect
only the participation decision. Although tobit model coefﬁcients can be decomposed to
estimate the effects of variables on the participation and value decision (McDonald and
Mofﬁt, 1981), such estimates, which suggest that all signiﬁcant variables are predictors of
both decision components, can be misleading because they are based on a single set of

coefﬁcients.

 

8 The likelihood ratio statistic 3. = -2[ln Lr(ln L,» + 1n 1:73)] where L,- is the log-likelihood for the tobit,
probit, and truncated regression models (Greene, 1993).

35

Table 4. Tobit and Cragg Models for the Contingent Valuation Decision Purchase
Collective Wildland Fire Risk Reduction '

 

 

 

Cragg Model
Tobit Probit Truncated
Variable Model Coefﬁcienta Slope Coefﬁcient Slope
Constant *- 102.940 :"-2.098 -0.544 *-460.580 -2210.500
(42.922) (0.816) (219.320)
Reducer 9.562 *0.724 0.188 -23.685 -1 13.670
( 12.333) (0.240) (40.5 10)
Survey96 6.424 *0.786 0.204 -66.934 -321.25
(14.207) (0.287) (47.173)
Distance 2.755 0.027 0.007 9.582 45 .990
(1.830) (0.033) (6.686)
Damage -4.464 -0.35 1 ~0.091 59.468 285.41
(19.735) (0.351) (62.391)
Numyears 0.077 0.001 0.000 0.393 1.885
(0.468) (0.009) ( 1.475)
WﬁreRank *-9.559 -.090 -0.023 -6.459 -31.000
(5.548) (0.106) (18.149)
Firechance 13 .046 0.3 10 0.080 17 .459 83.795
(24.261) (0.472) (80.392)
Damchance 27 .541 *0.986 0.256 - 10.331 -49.5 85
(20.485) (0.393) (69.719)
Responsible 4. 1 12 -0.036 -0.009 32.588 156.400
(5 .712) (0.105) (22.606)
PropertyTax 5 .980 *0253 0.066 - 10.135 -48.641
(5.581) (0.109) (18.589) .
Education -0.009 -0.035 -0.009 1 .827 8.768
(4.347) (0.082) (14.130)
Age -0.558 ~0.013 -0.003 3.107 14.910
(2.214) (0.040) (7.819)
Income *9.812 I"0. 153 0.040 I"35.372 169.77
(3.798) (0.071) ( 16.250)
Seasonal -20.479 -0.373 -0. 100 - 17.001 ~81.596
(14.363) (0.265) (46.107)
Gender 7.299 I"0.606 0.157 - 19.735 -94.716
(12.603) (0.268) (44.454)
InitialRisk *4.068 I"0.098 0.001 4.532 21 .753
( 1.393) (0.029) (4.788)
PropertyValue *0.71 1 -0.002 0.000 '1' 1 .566 7.514
(0.130) (0.002) (0.352)
-Log likelihood 1203.97 1 13.90 1035.50
der-rs 109.14*

 

‘ Standard errors in parentheses.
I"signiﬁcant p<0.1

36
Several variables are signiﬁcant only in the Cragg model. Private risk reduction
action (Reducer), conditional risk perception (Damchance), property tax spending belief,
gender, and survey year affect the participation decision. Household income level is
signiﬁcant in all three models indicating that income plays a role in the market

participation and value decision.

The probit and truncated regression parameters are not interpreted like OLS
parameters. To determine the marginal effects of the independent variables on the
probability of a positive bid and on the value of the positive bid, it is necessary to
calculate the derivative of each function with respect to the vector of independent
variables at some value of the regressors. Table 4 reports these slope values computed at

the means of the independent variables.

First, consider the effect of a continuous variable on the participation decision. At
the mean of the regressors, a unit increase in a respondent’s perception, expressed as a
probability, of the conditional risk (Damchance) increases the probability of a positive bid
for risk reduction by 0.256 (for this variable the lepe value makes more sense if it is
divided by 10 and interpreted as an increase of 0.026 in the probability of a positive bid
for every 0.1 unit increase the respondent’s conditional risk perception). The marginal
effect of property tax spending belief and initial risk are similarly interpreted for the

participation decision.

The nominal variables can be interpreted in much the same way. Holding all
other regressors at their mean values, the probability of participation increases by 0.724

for homeowners who have taken private actions to reduce their risk. Women and 1996

37
respondents were more likely to participate than males and 1994 respondents. Marginal
effects for the value decision are also evaluated at the regressor means. Wildﬁre risk
reduction value increased by $7.514 per property value unit ($1,000). Income affected
participation and value. An additional unit increase in household income (measured as

ranges on a 7-point ordinal scale) increased the probability of market participation by

0.04 and increased the bid by $169.77.

The effect of property tax spending beliefs provide evidence that protest bidding
may have been more prevalent than the few cases identiﬁed by interviewers. Only three
zero bids were classified as protest bids in 1996 (protest bids were not explicitly
identiﬁed in 1994). Respondent explanations for zero bids can be divided into two
categories ( l) preference for private risk reduction mechanisms, (2) acceptably low risk
level. The researchers accepted these explanations at face value; however, an underlying
payment vehicle bias is evident for many of these bidders. Overall, 29% of respondents
returned zero bids, but 48% of the respondents who believed they pay “too much” in

property taxes (n=89) also bid zero.

Interviewer bias may explain the positive coefﬁcient for 1996 respondents
(Survey96). One of four interviewers returned surveys in 1994 with a signiﬁcantly higher
proportion of zero bidders than those returned by other interviewers whose proportion of

zero-bidding respondents was equal to that of the interviewer in 1996.

DISCUSSION

We ﬁnd that the CVM method is capable of yielding theoretically valid estimates
of collective risk reduction value given the general conformance of the empirical model
to the conceptual model introduced earlier. Consistent with a threshold model of choice,
risk perception determined the probability of market participation but not risk reduction
value; and, as predicted by expected utility theory, value of assets at risk are a signiﬁcant
predictor of risk reduction value. However, the relationships between survey variables

and the relationships between conceptual model elements was not a perfect match.

One ﬁnding that is a departure from the conceptual model concerns the effect of
initial risk on intended hypothetical market behavior. Expected utility theory predicts
that, given an equivalent percent reduction in risk, WTP values should increase with
baseline risk; however, we found that the effect of risk level on market behavior is more
consistent with a threshold decision model. Homeowners will participate in the market
for collective risk reduction only if the risk level is above a certain level; household

income and asset value at risk then drive the value decision.

Interviewers found evidence for this type of behavior by homeowners responding
to the WTP elicitation questions. Several respondents unwilling to participate in the
hypothetical market commented that the speciﬁed initial risk level was probably less than
the risk of a house ﬁre caused by arson or faulty wiring. As one respondent put it, “At
7% you could get hit by lightning.” The relationship between initial risk level, presented

to respondents as an expert risk assessment, and market participation indicates that

38

39

respondents adopted these estimates and incorporated them into the decision.

Homeowners also incorporate their prior risk perception into the decision process;
however, only their conditional risk assessments (Damchance) affected their intended
behavior. Both of these observations are consistent with a Bayesian learning model of
risk perception formation such as that embodied in prospective reference theory.
Individuals respond to hazards based on a posterior assessment of risk, the result of prior
risk perceptions updated with new information (Viscusi, 1989). Representing respondent
risk perception as the product of the conditional and unconditional risk perceptions (the
implied joint probability) in the regression equation is a more realistic portrayal of the
risk being valued; however, the model specification presented here preserves the
unconditional and conditional joint risk perception components to investigate the extent

to which each affect the decision.

Another departure from the conceptual model is the relationship between
alternative risk reduction mechanisms and hypothetical market participation. The effect
of past risk reduction behavior (Reducer) ran counter to our hypothesis that homeowners
who had already engaged in self-protective actions would be less likely to participate in
the collective protection market. Rather, evidence of past risk reduction may have acted
as an indicator of risk aversion. A market offering additional safety opportunities attracts
homeowners predisposed to reduce risk whenever possible. Other homeowners rejected

additional collective risk reduction in favor of self-protective actions.

Many zero-bidders expressed a preference for increased self-insurance or self-

protection. Respondent comments reveal several possible explanations for this

40
observation. First, some people are more apt to accept personal responsibility for future
outcomes, a concept referred to as locus of control (Schiff, 1977). Several homeowners
commented that they would rather increase their insurance coverage or that they had
“weighed” and “assumed” the risk trade-off that living in the woods entails. Second,
government actions aimed at reducing wildﬁre risk seemed futile. Remarks about the
ferocious 1990 SBR ﬁre sometimes accompanied this comment. If collective protection
seems futile, homeowners demanding additional risk reduction turn to self-protective
actions or insurance as an adjustment choice. These observations have important

implications for models of ex-ante risk reduction value.

In measuring the value of collective protection, researchers often treat risks as
involuntary or exogenous (e.g. Smith & Desvousges, 1987). This has led to ﬁndings of
increasing marginal ex ante valuations for safety. However, when self-protection
opportunities are available, endogenous risk prevails, and the directional change in ex
ante marginal valuation is ambiguous (Shrogen & Crocker, 1991). An earlier treatment
of data from this study comparing the incremental value of the ﬁrst risk reduction with
that of the second increment revealed diminishing ex-ante marginal valuations for

collective protection (Fried, Winter, & Gilless, 1996).

Commenting on the need for CVM validity tests, the NOAA Panel on Contingent
Valuation wrote, “some form of internal consistency is the least we would need to feel
some conﬁdence that the verbal answers correspond to some reality. (Arrow, 1993, p.
4604). This study shows that continuous WTP data and an econometric model that is

consistent with the consumer’s decision framework is capable of yielding theoretically

41
valid results. The two-stage model dissects the valuation decision into its component
parts - participation and valuation. This allowed us to test a theoretical model free of the
assumptions that risk is exogenous and that risk levels affect a simultaneous valuation
decision that require perfect information. Our enthusiasm. is tempered by the problems
we encountered dealing with large WTP outlier values and the bias this may have

introduced into the valuation equation.

Recommendations for Further Research

This is an important ﬁrst step in assessing the appropriateness of CVM for
wildﬁre risk valuation. Additional research should evaluate CVM in the wildﬁre risk
context along the following lines. First, our conceptual model is missing some
potentially important elements that determine risk reduction market participation. Some
individuals will reject even the most plausible hypothetical market for collective risk
reduction because it is seen as a futile attempt to control an awesomenatural force or
because of the perceived nature of the property rights to safety. Future valuation studies
should attempt to measure and control for these responses by eliciting respondents’
beliefs about the effectiveness of risk reduction mechanisms presented in the hypothetical
market. This enhances the reliability of the survey instrument and provides greater

insight into the relationship between wildland residents and their surroundings.

Second, many forest homeowners tend to reject private risk reduction actions,
especially the most effective private wildﬁre hazard adjustment: clearing vegetation a

safe distance from the home. The positive amenity values of this hazard factor limits the

42
hazard adjustment opportunities. The amenity values attached to hazardous vegetation
may be quite high and present an insurmountable obstacle to govemment-sponsored
efforts aimed at encouraging ﬁresafe landscaping due to the disutility of that activity. Put
succinctly by one resident, “People live in the woods to liVe in the woods.” This thinking
may be positively related to WTP for collective protection and has obvious implications
for wildland ﬁre policies. We have shown previously that values for private risk
reduction may far exceed the value for an equivalent amount of collective risk reduction,
but many homeowners are unwilling to engage in protective behavior that diminishes the
amenity values afforded by vegetation (Fried, Winter, & Gilless, 1996). Unfortunately,
the study design did not allow us to incorporate the value of private risk reduction into the

collective risk reduction model. This would be a worthwhile pursuit.

Third, the wildland-urban interface is a particularly difﬁcult sampling
environment. Typically, the population at risk is dispersed, difﬁcult to aggregate into a
suitable sampling frame, and extremely difﬁcult to contact. Future wildﬁre valuation

researchers should not underestimate the difﬁculty of this vexing sampling problem.

Policy implications

Understanding how people decide whether and how to respond to risk is vitally
important to the success of govemment-sponsored initiatives to improve safety. If, as we
suspect from our results, individuals use a threshold choice model in response to this
hazard, some homeowners are unlikely to support additional government expenditure on

collective protection, nor are they likely to respond to government programs to encourage

43
self-protective behavior because they view the hazard as not worth worrying about

(Kleindorfer & Kunreuther, 1988).

If homeowner decisions and economic values are based on misperceptions of the
wildﬁre risk, then policies based on these values may be misdirected (Kleindorfer &
Kunreuther, 1988). Our model and results suggest that this group of homeowners did not
consider the unconditional risk of wildfire occurrence, but did rely on their perception of
the conditional risk of home loss and on an “expert” assessment of the joint risk. In
response to both risk perception and valuation questions, homeowners remarked about
recent ﬁre experiences indicating the dynamic nature of risk perception and, perhaps,
valuation. These observations should be encouraging to public land management
agencies such as the US. Forest Service that are pursuing risk communication as a policy
response to the wildland-urban interface ﬁre problem (U. S. Department of the Interior &

U. S. Department of Agriculture, 1995).

REFERENCES

REFERENCES

Arrow, K. R, Solow, R. Portney, P. R., Learner, E. E., Radner, R. & Schuman, H.
(1993). Report of the NOAA panel on contingent valuation. Federal register,
_5_8(10), 4601-4614.

Batts, D. (1993). Ming incremental reductions in ﬁre risk at the urban-wildland
interface: Design and pretest of a contingent valuation survev instrument.
Unpublished master’s thesis, Department of Resource Development, Michigan
State University, East Lansing.

Bishop, R. C., Heberlein, M. P., & Kealy, M. J. (1986). Hypothetical bias in contingent
valuation: Results from a simulated market. Natural Resources Journal 23(3),
619-633.

 

Blaylock, J. R., & Blisard, W. N. (1992). US. cigarette consumption: The case of low
income women. Americangioumal of agricultural economics 74(3), 698-705.

 

Boyle, K. J ., Johnson, F. R., McCollum, D. W., & Desvousges, W. H. (1996). Valuing
public goods: discrete versus continuous contingent valuation responses. Land
Economics 72(3), 381-96.

 

Burton, I., Kates, R. W. & White, G. F. (1993). The Environment as Hazard, (2“1 ed.).
New York: The Guilford Press.

Carson, R. T., Flores, N. E., Martin, K. M., & Wright, J. L. (1996). Contingent valuation
and revealed preference methodologies: Comparing the estimates for quasi-public
goods. Land Economics 72(1), 80-99.

 

Cortner, H. J ., & Gale, R. D. (1990). People, ﬁre, and wildland environments. Population
and Environment 11(4), 245-257.

 

Cragg,J. G. (1971). Some statistical models for limited dependent variables with
application to the demand for durable g. Econometrica 39(5):829-45. . , eds. .

 

Decker, B. D. (1995). A review of the psychological literature on risk greeption in terms
of chance harm and context. Unpublished master’s thesis, Michigan State

University, East Lansing.

 

Diamond, P. (1996). Testing the internal consistency of contingent valuation surveys.
Journal of Environmental Economicspnd Management, 30, 337-347.

44

45

du Vair, P. & Loomis. J. (1993). Household's valuation of alternative levels of hazardous
waste risk reductions: An application of the referendum format contingent
valuation method. Journal of Environmental Management, 39, 143-155.

Eom,Y. S. (1994). Pesticide residue risk and food safety: A random utility approach.
American Journal of Agricultural Economics 76, 760-771.

 

Foote, E.I.D., Martin, R.E., & Gilless, J .K. (1992). Defensible space factor study: A
survey instrument for structure loss analysis. In P.L. Andrews & D.F. Potts, (Eds).

Proceedings of the 11th Conference on Fire and Forest Meteorology. (pp. 66-73).
Bethesda, Society of American Foresters.

Frankel, M. (1983). Sampling theory. In P.H. Rossi, J .D. Wright & A.B. Anderson (Eds),
Handbook of Survey Research. (pp. 21-67). San Diego, CA: Academic Press, Inc.

Freeman, A. M. (1993). The measurement of environmental and resource values: Theog
and methods. Washington: Resources for the Future.

Fried, J. 8., Winter, G., & Gilless, J. K. (1996). Assessing the beneﬁts of reducing ﬁre

risk in the wildland-urban interface: A contingent valuation approach.
Unpublished manuscript, Michigan State University.

Fried, IS (1993). Valuing social and economic impact of ﬁre at the urban-wildland
interface. Research proposal. May 26. East Lansing, MI: Department of Forestry,
Michigan State University.

Fried, J .S., Stewart, S.I., & Gilless, J K (1994). Objectives setting in the wildland ﬁre
system: What do the customers think? In J. Sessions & J.D. Brodie (Eds.),

Proceedings of the 1994 Symposium on Systems Analysis in Forest Resources.
Paciﬁc Grove, CA: Society of American Foresters.

Gardner, P. D., Cortner, H. J ., & Widaman, K. (1987). The risk perceptions and policy
response toward wildland fire hazards by urban home-owners. Landscax and

Urban Planning, 14, 163-172.

Great Lakes Forest Fire Compact. (1993). Fire Agency Forest Home Fire Hazard
Assessment Evaluation Form. Michigan Department of Natural Resources, Forest
Management Division, Lansing, MI.

Greene, W. H. (1995). LIMDEP Version 7.0 User’s Manual, Bellport, NY: Econometric
Software, Inc.

Greene, W. H. (1993). Econometric analysis, (2Ind ed.). New York: MacMillan.

46

Haines, P. S., Guilkey, D. K., & Popkin, B. M. (1988). Modeling food consumption
decisions as a two step process. American Journal of Agricultural Economics,
m6), 543-552.

Heckman, J. J. (1976). The common structure of statistical models of truncation, sample
selection and limited dependent variables and a simple estimator for such models.
Annals of Economic and Social Measurement 5(4), 475-492.

 

Kahneman, D. & Tversky, A. (1979). Prospect theory: An analysis of decision under risk.
Econometrica 47(2), 263-291.

 

Kates, R. W. (1962). Hazard and choice pgrception in ﬂood plain management. (Research
Paper No. 78). Chicago: University of Chicago, Department of Geography.

Kleindorfer, P. R. & Kunreuther, H. (1988). Ex ante and ex post valuation problems:
Economic and psychological considerations. In G. L. Peterson, B. L. Driver, & R.

Gregory (Eds), Amenity resource valuation: Integrating economies with other
disciplines (pp. 77-86). State College, PA: Venture Publishing.

Kunreuther, H. (1978). Disaster insurance protection: Public mlicy lessons. New York:
John Wiley and Sons.

Kunreuther, H. & Slovic, P. (1978). Economics, psychology, and protective behavior.
Psychology and economics. 68(2), 64-9.

Lee, E. S., Forthoffer, R. N., & Lorimor, R. J. (1989). Analyzing complex survey data.
Newbury Park, CA: Sage.

McDonald, J. F. Mofﬁt, R. A. ( 1981). The uses of tobit analysis. Review of Economics
and Statistics. 63. 318-321.

McKay, J. M. (1985). Community adoption of bushfire mitigation measures in the
Adelaide Hills. In D.T. Healey, F.G. Jarrett & J .M. McKay (Eds), The Economics
of Bushﬁres: The South Australian Experience. Melbourne: Oxford University
Press.

Mitchell, R. C., & Carson, RT. (1989). Using surveys to value public goods: The
contingent valuation method. Washington, DC: Resources for the Future.

National Fire Protection Association. (n.d.). Stephan Bridge Road ﬁre: A case study.
Quincy, MA: Fire Investigations Division, National Fire Protection Association.

47

Ready, R. C., Buzby, J. C. & Hu. D. (1996). Differences between continuous and discrete
contingent value estimates. Land Economics 72(3), 397-411.

 

Rural Fire Protection in America (RFPA) Steering Committee (1994). Fire protection in
rural America: A challenge for the future. Washington, DC: National Association
of State Foresters.

Schiff, M. (1977). Hazard adjustment, locus of control, and sensation seeking.
Environment and Behavior. 9(2), 233-253.

Schulze, W. D. (1993). Use of direct methods for valuing natural resource damages. In

R.J. Kopp and V.K. Smith (Eds), Valuing natural assets: The economics of natural
resource damage assessment (pp. 204-230). Washington, DC: Resources for the

Future.

Shrogen, J. F. & Crocker, T. D. (1991). Risk, self-protection, and ex ante economic value.
Journal of Environmental Economics and Management, 20, l-15.

Shrogen, J. F. & Crocker, T. D. (1989). Risk Reduction Mechanisms and Risk Valuation:
An Exmrimental Study. Ames, Iowa: Department of Economics.

Slovic, P. (1987). Perception of risk. Science 236, 280-85.

 

Smith,V. K. & Desvousges, W. H. (1987). An empirical analysis of the economic value
of risk changes. Journal of political economy, 95(10), 89-114.

Spotts, D. M. (1991). Second homes and other temporary residences in Michigan. Travel
and tourism in Michigan: a statistical proﬁle, 2nd edition. D. M. Spotts, (ed). East
Lansing, MI: Travel, Tourism and Recreation Resource Center, Michigan State
University.

 

Stynes, DJ. & Zheng, J. (1995). Seasonal homes in Michigan. East Lansing, MI:
Department of Park Recreation, and Tourism Resources, Michigan State
University.

Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and
probability. Cognitive Psychology. 5. 207-232.

U. S. Department of the Interior U. S. Department of Agriculture. (1995).Federal

Wildland Fire Management Policy and Program Review Final Report. December
18, 1995. Washington, DC.

Viscusi, W. K. (1989). Prospective reference theory: Toward an explanation of the
paradoxes. Journal of Risl_t_and Uncertainty, 2, 235-264.

48

Viscusi, W. K. & W. A. Magat. (1987). Learning about risk: Consumer and worker
responses to hazard information. Cambridge, MA: Harvard University Press.

Whitehead, J. C., Blomquist, G. C., Ready, R. C., & Huang, J. (1997). Construct validity
and polychotomous choice contingent valuation questions. (Working Paper
#ECU9612). Greenville, NC: Department of Economics, East Carolina
University.

APPENDICES

APPENDIX A

EXPANDED LITERATURE REVIEW

APPENDIX A -- Expanded literature review

The main objectives of this study are to develop beneﬁt estimates of WTP for
wildﬁre risk reduction that could be incorporated into a beneﬁt-cost ﬁre management
planning model, and to test the validity of CVM responses based on theoretical
expectations. Several models of consumer choice were consulted to develop a conceptual
model of the homeowner decision. The theories discussed below differ primarily in their
assumptions about the cognitive abilities of the decision maker and how information

about the hazard is incorporated into the decision process.

Exgted utility theog

Traditionally, those analyzing decisions involving uncertain outcomes have
appealed to expected utility theory which assumes that risk averse, utility maximizing
agents possess perfect information regarding the probability of future states of nature and
their associated economic gains and losses. Within this framework the homeowner
decision to engage in protective behavior (e. g. purchase risk reduction) is a function of
asset value at risk due to ﬁre damage and the risk of ﬁre damage expressed as a
probability of occurrence. Utility-maximizing individuals uncertain about the future state
of the world (ﬁre destroys home or does not) attach value to avoiding the unpleasant
outcome or to reducing its likelihood. The value can be expressed as willingness to pay

for the risk reduction ex ante (Freeman, 1993).

49

50
The standard model of expected utility is expressed by Equation 1, where each
term represents a future state of the world. In the ﬁrst state utility, expressed as an
indirect utility function v, is a function of income y minus a loss L (value of home
destroyed by ﬁre); in the second state there is no loss of income. In addition, each state
carries an associated probability of occurrence: It and (I -n). The amount that individuals

would be willing to pay ex ante to completely avoid a loss is the option price, op

(Freeman, 1993).
E(v)=7r-v(y-L)+(l-7r)°v(y)=v(y-0P) (5)

For many hazards such as wildﬁre, the risk cannot be completely eliminated;
therefore, the option price for risk reduction represents the value of probabilistic
insurance (Kahneman& Tversky, 1979). For these cases the indirect utility function is

modiﬁed and the option price, op’ for risk reduction from It to 1t’ is the solution to
7t-v(y- L)+(1—n)-v(y) = 7t'-v(y—L-op')+(1—7r')-v(y—op') (6)

According to this model, willingness to pay for reductions in the probability of a
loss increases with the magnitude of the risk reduction tt-tt’. In the wildﬁre damage
context, forest residents exposed to high risks of damage (those with ﬁre-prone
landscaping) would be willing to pay more for given percentage of risk reduction (private
or collective) than residents at low risk (those with ﬁre-safe landscaping) holdingasset

value and marginal utility of income constant.

This simple model assumes that decision makers know and adopt the true

51
probabilities of future states, that they know the value of the potential loss, and that they
accurately calculate the utility level for each prospective state of the world. But many
researchers report that the standard model of expected utility maximization does not
explain the observed behavior of individuals facing uncertain outcomes. The most
common observation is that people tend to be oversensitive to low probability events
causing them to express high bids for risk reduction that overcompensate for the actual
risk level they face, and they are relatively insensitive to high probability events (Schulze,
1993; Viscusi, 1989; Kahneman and Tversky, 1979). Slovic (1987) proposed that people
develop their own assessments of risk based on characteristics of the hazard such as
controllability, voluntariness, catastrophic potential, and degree of outcome uncertainty.
So one serious problem with the expected utility approach is that it is unclear how to

represent the risk term in the expected utility model.

Prospgctive reference theog

Rather than abandon expected utility as a model of behavior, psychologists and
economists have incorporated a model of learning into expected utility theory to explain
the apparent “irrationality” of decision makers. Viscusi (1989) proposed a variation of
expected utility theory that predicts many of the aberrant experimental phenomena that
the standard expected utility model cannot explain, especially the apparent overweighting
of low probability events, the underweighting of high probability events and the
reluctance to purchase probabilistic insurance (Viscusi; Kahneman and Tversky, 1979).

There seems to be growing agreement among researchers that individual decision makers

52
facing uncertainty act as if they are solving a Bayesian estimation problem for which they
estimate the probability of a loss based on imperfect information. Under this proposition,
decision makers continuously update their assessments of the probability of a hazard as
they are exposed to new, relevant information. Viscusi’s prospective reference theory of
risk perception formation is used to modify the standard expected utility model by more
accurately describing the probability estimate, 2: , in Equation 1. The updated or posterior
risk perception P , described by equation 2, replaces It in the ﬁrst equation and is a
weighted average of the prior risk perception p, and an objective or technical risk

assessment It (Viscusi, 1989).

= 700+?!

P(p) 7+5 (7)

The scale factor 7 corresponds to the informational content of the
individual’s prior beliefs, and 5 represents the informational content of new, objective

information. The ratio % is the relative informational content of prior and subsequent
exposure to a hazard. Viscusi and Magat (1987) found it feasible to measure y in an
7

experiment that found chemical hazard warning labels with strong informational content

had high % values.

The links between hazard information, risk perception and precautionary
action are well-developed in the morbidity risk and economics literature, but few studies
have explored these relationships in the wildfire context. Extensive studies of consumer

and worker behavior (Viscusi and Magat, 1987) show that risk perception is a key

53
variable in explaining individuals’ willingness to engage in precautionary behavior.
Likewise, Born (1994) found risk perception to be an important component of behavior
among consumers choosing between produce with and without pesticide residue. In a
contingent valuation study of hazardous waste risk reductions, respondent risk perception
(belief regarding the likelihood of coming in contact with hazardous waste) was a
signiﬁcant predictor of risk reduction value (duVair and Loomis, 1993). From her study
of ﬁre-affected communities in South Australia McKay ( 1985) concludes that the level of
risk perception distinguished those homeowners that undertook costly ﬁre hazard

precautionary measures from those who undertook low-cost measures.

Smith and Desvousges (1990) applied the prospective reference model of risk
perception in a study of radon risk. Their experiment measured prior radon risk
perception of 2,300 homeowners, then measured radon levels at each of their homes.
Homeowners were later sent the testing results along with information about the radon
hazard, the meaning of risk levels, and suggested mitigation measures. Finally, risk
perceptions were measured again after homeowners had received the objective risk
information. The investigators manipulated the informational content of the material sent
to homeowners to measure its effect on posterior risk perceptions. Their study used a
simpler form of Viscusi’s prospective reference model to explain their observation of

systematic responses to risk information.
PA=aBPB+aMPM (8)

Perceived risk after receiving information PA is a weighted average of perceived

54

risk measured prior to receipt of information PB and risk inferred from the new
message PM . The a, ‘s are the relative informational content associated with each risk

and the subscripts designate variables associated with base (prior) information B , and
objective risk inferred from the message M . The P; ‘s may not be numerical per se; for
example, an individual may not think in terms of numerical probabilities, but assigns
some degree of seriousness to the threat (the measurement approach employed by Smith
and Desvousges). The assessed seriousness carries with it relative informational content

(a, ‘s) which acts as a weight or scale factor. Likewise, the objective risk message may

not be conveyed in terms of numerical probability. Rather, it could consist of a brochure
that allows homeowners to rate their own risk as high, moderate, or low based on
objective criteria. Observations from the Smith and Desvousges experiment were
consistent with the proposed Bayesian updating of subjective risk lending empirical

support to the prospective reference model.

Seguential choice models

Even if individuals’ risk perceptions can be accurately reﬂected through these
reﬁnements, expected utility may still be a poor descriptive model of choice because it
assumes that people act on the basis of accurate multiplication of utilities by probabilities
Kunreuther (1978) proposed a sequential model of choice for insurance based on the
notion of “bounded rationality.” Within this model, people make decisions from a very
simple model of the world they construct in response to their limited cognitive abilities.

Kunreuther conceptualized this model in the context of natural disaster insurance

55
purchases. Individuals decide whether or not to insure against earthquake losses, for
example, in a three-stage sequence. First, the hazard must be considered a problem;
second, the person must be aware of the option to purchase insurance; the third and ﬁnal

stage is the decision to purchase or not purchase insurance Coverage.

In deciding whether or not to purchase insurance, individuals rely on simple
heuristics rather than maximization of some objective function. For example, under a
threshold model of choice, individuals ignore the consequences of a certain hazard if the
perceived likelihood of the hazardous event is below a given level (Kleindorfer &
Kunreuther, 1988). Familiarity with the hazard is another basis for a simple decision rule.
Kates (1962) proposed that individuals are “prisoners of their own experience,” and that
they decide how to respond to the prospect of future disaster by relying on their
experiences with past disasters. Tversky and Kahneman (1973) refer to this phenomenon
as “availability” whereby individuals judge the probability of some future event by the
ease with which a similar past event is recalled. Burton, Kates, and White (1993)
discovered experience-based risk perceptions and response decisions in cross cultural

surveys of residents in disaster prone areas around the world.

If such heuristics are used in the context of decisions to pay for wildﬁre risk
reductions, we would expect that individuals who have experienced recent wildﬁres
would be more likely to assign high probabilities to future ﬁre events. Prospective
reference theory may accommodate this expectation if experience with wildﬁres is
considered informational content which acts as a weight on an individual’s subjective

probability estimate of wildﬁre occurrence.

56

Burton, et al (1993), refer to prior hazard experience as one of four classes of
factors that are associated with hazard adjustment decisions. First, prior experience with
the hazard - which includes the length of time one is exposed to the consequences of
hazardous events and the degree of its severity — is relatedto how people adjust to natural
hazards. Second, material wealth can be an important determinant in the adjustment
decision as wealthier people are “more likely to experiment with a variety of
measures. . .and they have the necessary pecuniary support.” Third, personality traits
such as “a sense of inner control” are related to hazard adjustment decisions. Fourth, the
perceived role of the individual in a social group can inﬂuence whether adjustments are
made. This last class of factors may be especially relevant to the case of wildﬁre
protection. Because individual homeowners and government land managers have
opportunities to reduce risk, the resulting mix of public and private risk reduction will
depend to some degree on the perception among homeowners of who is primarily
responsible for wildﬁre protection. The fact that individuals have seVeral risk reduction
mechanisms available to them has important implications for the analysis of risk

reduction beneﬁts.

Individuals can reduce risk through collective or self-protection, and self-
insurance mechanisms. WUI residents can pay taxes to improve government ﬁre
prevention and suppression programs, they can alter their private landscapes making their
property less prone to wildﬁre damage, and they can purchase ﬁre insurance that would
reimburse them for items destroyed by ﬁre. Each of these actions can reduce the

probability and severity of wildﬁre damage. This implies that in the absence of

57
opportunities to self-insure or self-protect the payment for collective risk reduction would
increase. In a series of controlled experiments designed to assess the inﬂuence of
alternative risk reduction mechanisms on risk reduction valuation, Shrogen and Crocker
(1989) found that self provisions of safety are more highly valued than collective
provision and that the risk premium for collective risk reduction can be negative if
individuals are provided with opportunities for self-protection or self-insurance. The
existence of these alternative risk reduction mechanisms complicates the task of valuing
collective risk reduction. This phenomenon was accounted for in the research design of
the present study and it provides an additional variable in the conceptual model of

willingness to pay for government risk reduction.

Empirical Models

An objective of this study is to find out how much forest residents value risk
reduction in a hypothetical market and the relationship between value and selected
independent variables. A feature of open-ended valuation responses complicates the data
analysis: the willingness to pay distribution is typically bimodal, containing a high
proportion of zero bids, and a second mode within a bell-shaped curve (Schulze, 1993).
Therefore the respondents can be divided into two groups, (1) those who bid positive
amounts for risk reduction for which there are observations on the independent variables
and on the dependent variable: WTP for risk reduction, and (2) those who chose not to
participate in the hypothetical market (zero bidders), for which there are only

observations on the independent variables. Under these conditions, the range of the

58
dependent variable is limited to non-negative values. One option for data analysis is to
estimate willingness to pay for market participants only. But for ordinary least squares
regression, the resulting parameter estimates using this approach will be biased and
inconsistent because if only market participants are considered, it can not be assumed that
the expected mean of the error term will be zero (Gujarti, 1995). Fortunately, several

econometric techniques are available to analyze limited dependent variable data.

The tobit model is one example of a regression model widely used to estimate
demand for consumer goods. In the context of this research, the tobit model takes the

form

Yi =[3X,-+u,- if BXi+ui >0
Y,- = 0, otherwise (9)

where Y is the payment for risk reduction, X is a vector of independent variables, [3 is a
vector of unknown coefﬁcients, and u is an independently distributed error term with

mean of zero and constant variance (McDonald & Mofﬁtt, 1981).

An assumption underlying this analysis is that a zero bid represents a comer
solution so that if values for explanatory variables changed, say an increase in income,
zero bids would turn positive (Blaylock and Blisard, 1992). But in this study, such a
change may not affect willingness to pay. Consider those residents who strictly prefer
self-protection over collective protection, or those whose level of initial risk is below
their threshold of concern. Some residents will not participate in the hypothetical market
for public risk reduction; however, tobit estimation restricts parameter estimates of

participation and bid value to the same sign and magnitude. This can be seen by

59
examining the mathematical relationship of parameters in the tobit analysis to the
dichotomous participation decision and continuous value decision. The probability of

observing a limit observation (Y,- = 0) is

P(Y,- = 0) = <I>(-X,-13/ 0) (10)

where (I) is the standard normal distribution function. The likelihood function for

1,518

L=Ho(—X,-B/a)“"i’-1/o¢([Y,--X,I3]/a)l’ (11)

where 1,- =0 for limit observations and 1,- =1 otherwise, and (I) is the standard normal

density function. Estimates of B and o are obtained by maximizing the likelihood
function. Because there is only one coefﬁcient, B, in (l 1) that determines both the
probability that Y is a non-limit observation and the mean of Y for positive values of Y
(Haines, Guilkey, & Popkin, 1988), explanatory variables necessarily have the same
effect on both stages of the consumer decision (participation and value). Thus the tobit
model can dissect the WTP observations into the participation and value decisions, but
because the parameter estimates for a given regressor must be of the same sign and
magnitude for both decisions, it is inconsistent with a conceptual model that proposes
differential effects of independent variables on the market participation and value

decisions.

A generalization of the tobit model developed by Cragg permits separation of the

consumer decision into two stages. The probability of a limit observation is

60

P(Y.-=0)=<D(-Xu31) (12)

The density function of Yi conditional on being a positive amount is normal with

mean X :13 and variance 0'2:

f(Y,'/Yi>0)={1/O'¢([Yi‘xil32]la)}/¢(XIIBZIO) (13)

The likelihood function for the observed sample is
l.
L: r1¢(—x,p, /a)“""-{<I>(x.a)[1/a¢([n — X,[5]/o)]/¢(X,-B2 10)} (14)

The function can be maximized for B. and B2; therefore, the explanatory variables
X; can have different effects in determining the probability of a non-zero value for Y,- and
in determining the mean of Y,~. The tobit estimator is a special case of the Cragg estimator
where B; = leo. Generating Cragg model estimators B; and B2 consists of estimating a
probit model for the probability that Y,- >0 and a truncated regression model for the non-
limit observations (Greene, 1993). If tobit is the correct speciﬁcation, the tobit estimators
are equal to the sum of the Cragg estimators (probit coefﬁcients plus truncated regression
coefﬁcients). A likelihood ratio test is used to test this null hypothesis. The likelihood
ratio test statistic, -21nA is chi-squared, with degrees of freedom equal to the number of
variables speciﬁed. For this case the test statistic is given by,

2 LM +Lm
x =-2|n —-L-—- (15).

"I

If the null hypothesis is rejected, the estimators for the alternative models are not

equal and it is assumed that the independent variables do not have equal effects on (1) the

61
probability that Y,- is a non-limit observation and (2) the variance of nonlimit

observations (Greene, 1993).

APPENDIX B

EXPANDED SURVEY DESIGN SECTION

APPENDIX B - Expanded Survey Design Section

This section presents conceptual and operational deﬁnitions of the key variables
that were measured using the contingent valuation survey instrument, and proposes
hypotheses based on expectations from the theoretical and empirical literature described
in the literature review. Each variable acts as an indicator of a conceptual element; several
conceptual elements have multiple indicators. Unless otherwise stated, the operational
deﬁnition of key variables consists of responses to survey questions designed to elicit
indicators of the underlying concepts. Survey questions or procedures designed to
measure each variable appear in italics at the end of the conceptual deﬁnitions along with

the associated question number on the 1996 survey form (Appendix E).

Initial risk

Initial risk is the expert assessment of the risk that a respondent would lose his or
her home to wildﬁre in the next ten years assuming no private or public risk reductions
are implemented. Each respondent’s initial risk level is expressed as a percent chance of
home destruction in a ten year period given the current private property conditions and
current level of government ﬁre prevention and suppression capabilities. The initial risk
is the product of the ten-year unconditional risk of wildﬁre occurrence in a respondent’s
neighborhood and the conditional risk that a passing wildﬁre would burn the respondent’s
home. We refer to this conception of initial risk as the joint probability model of wildﬁre

home destruction.

62

63

Hazard adjustment alternatives

Collective protection

Individuals sometimes fail to take precautions against a hazard because they
believe that the community should provide protection. This behavior may lead to the
community providing some form of protective service for its citizens (Burton et a1, 1993).
Government efforts to reduce the risk of wildﬁre damage to private property are an
example of collective protection. Investment in ﬁre suppression equipment, personnel
training, and ﬁre prevention programs such as homeowner education all belong in this
category. However, individuals may believe that private action is a more appropriate
response to a hazard. Private actions fall into two categories: self-protection and self-

insurance.

Self-protection

The joint probability model of wildﬁre home destruction implies that individual
homeowners have the opportunity to reduce their exposure to the risk given the
unconditional risk of wildﬁre occurrence. In the context of this study, speciﬁc self-
protective actions to reduce the risk of losing homes to wildﬁres were explained to
respondents: 1) creation of a defensible space by clearing ﬂammable vegetation, debris,
and ﬂammable material such as fuel and ﬁrewood to a safe distance from the exterior
walls of homes; and 2) retroﬁtting homes with ﬁre-resistant building material such as
metal rooﬁng. Interviewers also outlined options for collective protection (government

risk reduction actions). To assess whether respondents had undertaken self-protection,

we asked:

20) Have you done anything to reduce the risk of losing your home to wildﬁres?

Sel -insurance

In addition to self-protection, homeowners also have the opportunity to reduce the

risk of wildfire losses through homeowners’ insurance. They were asked:

35) Do you carry ﬁre insurance?

Risk p_e_rception

Risk perception is the “intuitive judgment” of the risk posed by a hazard (Slovic,
1987). Risk perception is conceived here as the assigning of a numerical probability to the
future occurrence of two events comprising the joint risk of wildﬁre home destruction
over a ten year period. Because the main objective is to determine the value of risk
reductions, it is vitally important that the researcher and the respondent have similar
conceptions of risk, of the amount of risk reduction being valued, and of the conditions
which result in the reduction in risk. As stated earlier, there is overwhelming evidence
that scientists’ risk assessments do not coincide with the public perceptions of risk;
therefore, one of the challenging methodological aspects to this research was to develop
concepts of wildﬁre risk and risk reduction that could be meaningfully and reliably
communicated to homeowners. Van Ravenswaay’s deﬁnition of risk perception provides
guidance for designing and evaluating risk perception measurement and risk
communication in the context of valuation surveys (Decker, 1995). Van Raveswaay

deﬁnes risk perception as “an assessment of the chance of harm in a given context.” The

65
concepts chance, harm, and context must each be clearly deﬁned when measuring risk

perception and risk reduction valuation across individuals.

Chance is determined with respect to a given population over a speciﬁed time
period. Because risk assessments were performed on an individual homeowner basis, the
population is the individual homeowner and the time period was speciﬁed as ten years.
Harm refers to the type of harm inﬂicted by the hazardous event. Harm was speciﬁed as
the destruction of the respondent’s home by wildﬁre.9 Context comprises the conditions
that result in the level of hazard exposure and that cause the reduction in that level. Here,
context was provided by describing, in lay terms, the joint probability model of wildﬁre
home destruction, reminding them of their opportunities for collective and self-protection,
and describing the speciﬁc mechanisms which would result in a reduced exposure to the

wildﬁre hazard.

The responses to the following questions were used to calculate the homeowner’s

implied probability assessment that his or her property will be damaged by ﬁre:

13) Which percentage best represents your estimate of the chance of a wildﬁre
moving through the block on which you live at some time in the next ten years?
[respondent has card with 0-100% range printed on it for reference]

14) If a ﬁre were to pass through your neighborhood, which percentage best
represents the probability of it destroying or severely damaging your home?
Hazard salience is another dimension of risk perception relevant to this study.

 

9The 1994 survey instrument speciﬁed harm as “. . .severely damage or destroy your home.. This was
changed slightly in the 1996 survey instrument to destroy your home. . To more clearly specify the
harm.

66
Kunreuther and Slovic (1978) found that homeowners who opted not to purchase
earthquake and ﬂood insurance were disinclined to worry about these low probability
events. They referred to homeowner’s limited time, energy, and attention to certain
hazards as a “ﬁnite reservoir of concern.” One method of measuring the salience or
importance of a hazard is to ask respondents to rank hazards in the order of the perceived
likelihood of occurrence. Higher ranked hazards are assumed to be more worthy of

attention.

Wildﬁre hazard salience is high for respondents who assign a low number to that

event when presented with the following exercise

15) Each of the following four events poses some risk to you. Please rank , from 1
to 4, these events by your estimate of the chance they will occur. Use ( I ) for most
likely and (4 ) for least likely.
The events are:
I) a tornado destroys your home _

2) a car accident which sends you to the hospital _

3) being a victim of a burglary or robbery crime __

4) a wildﬁre destroys your home _

Hazard Information

A focus of this study is the effect of information on homeowners’ risk perceptions .
and, in turn, on their willingness to take precautionary action by paying for public ﬁre
protection programs or undertaking privately ﬁnanced structural or landscape safety
improvements. Of particular interest is the nature of information in the wildﬁre context.
How do people learn about the wildﬁre hazard, and how does this learning process affect

their perceptions of the problem and their behavioral response? We propose that

67
information associated with current risk perceptions of WUI homeowners is composed of
hazard awareness and hazard experience. These information components shape the
perceptions of wildland residents and help explain the variation in WTP for risk

reduction.

Gardner, Cortner, & Widaman ( 1987) studied wildﬁre risk perceptions among
wildland homeowners in several southern California communities. They deﬁned hazard
information as a combination of hazard awareness and hazard experience and studied two
ﬁre-prone areas: one which had recently experienced a series of ﬁres, and one that had
not. Fire hazard awareness among residents of the non-affected community grew over
time. High awareness and early awareness (respondents were asked to recall their ﬁre
hazard awareness at the time of home purchase) was correlated with high assessed
probability of both wildﬁre occurrence and personally threatening wildﬁre occurrence.
They proposed that, over time, as homeowners witness or are exposed to media coverage
of ﬁres in nearby communities, their awareness of the ﬁre problem increases, and this
shapes their perceptions of ﬁre risk. Along the same lines, McKay (1985) observed
increased perceived risk among those living closer to burned areas. Presumably, distance
to recent events can act as an indicator of hazard awareness. The accumulation of this
hazard information and the concurrent upward adjustment of hazard awareness conforms
with Viscusi’s model of consumers and workers continuously updating their assessment

of risks based on prior beliefs and new information.

Of particular interest to this study is the ﬁnding that present ﬁre awareness

(measured on a scale of respondents belief of the seriousness of the present wildﬁre

68
situation) and assessed probability of wildfire occurrence were both much lower in the
ﬁre-affected community (Gardner et al., 1987). The investigators proposed that direct
experience with fires had a dampening effect on hazard awareness and risk perceptions of
ﬁre-affected residents who may have based their assessment “...on the adage that
lightning does not strike twice in the same place.” The danger in this thinking is that, due
to regrowth in burned areas, the ﬁre risk rebounds rapidly relative to the perception of
risk among those affected by previous ﬁres (Cortner and Gale, 1990; Gardner, et al.;
McKay, 1985). Though the rate of ﬁre recurrence in burned areas of Michigan’s jack
pine forest is probably not as high as in those areas studied by McKay and Gardner, '0 the
effect of risk experience on willingness to take precautions may be an important
determinant of the wildland residents’ role in reducing the wildﬁre risk. The occurrence
of ﬁres in and nearby ﬁre-prone areas may erode support for public ﬁre programs and
willingness to take precautionary action because a large number of affected people no
longer see the threat as immediately serious. On the other hand, ﬁre Occurrence may
bolster support for ﬁre protection programs among residents who are made aware of the
ﬁres but are not directly affected by them. This hypothesis is consistent with literature
summarized in the previous section which proposes that hazard experience tends to

increase a persons assessment of the likelihood of future hazardous events.

 

'0 Researchers of another major jack pine wildﬁre near Grayling (the Mack Lake ﬁre in 1980) have
determined that within the area burned, there had been ﬁve ﬁres larger than 10,000 acres since 1820, or one
large ﬁre every 28 years (NFPA, 1991).

69

Year moved in to home

Length of residence was calculated as the difference between reported year moved

into home:
I ) What year did you move into this home?

and the year in which respondent was interviewed (1994 or 1996). The longer residence
time increases the likelihood that the homeowner has been exposed to information about

the region’s ﬁre hazard.

Wildﬁre experiences

A “bum-out” list of residents who experienced a ﬁre-related loss during the 1990
Stephan Bridge Road was produced by the Michigan Department of Natural Resources.
Attempts were made to interview a member of every household on the bum-out list. In
1996, respondents were directly asked whether a forest ﬁre had ever caused damage to
their property. Several additional respondents, not on the original list of 1990 ﬁre victims
were coded as having experienced wildﬁre damage as a result of this question. An
additional indicator of wildﬁre experience is whether respondents had ever had to

evacuate because of wildﬁre. This question was asked explicitly in 1994 and in 1996.

7) Now I 'd like to ask you about your experiences with forest ﬁres. Please refer to
Item ( 7) on the response card. This is a list of experiences you may have had with
forest ﬁres in this region (the region near your vacation home). Please tell me
which of these experiences you have had.

1) Forest ﬁre has caused damage to my property Go to Q 8

2) Have had to evacuate due to threat of forest ﬁre in the region

70
Proximity to past ﬁre
The location of each respondent’s home was digitized (to within 1%: mile of true
location) and the distance in miles from the Stephan Bridge ﬁre perimeter was calculated
using a geographic information system (ArcView 3.0 and 1.0). Respondents living inside

the ﬁre perimeter were assigned a distance value of 0.

Interview year

Considering McKay’s ﬁnding that precautionary behavior in the wake of
damaging wildﬁres diminishes with the passage of time, it is important to take this
variable into consideration. This is especially important to the design of this study
because the interviews were conducted in 1994 and 1996. If there is a measurable
temporal effect, it is expected to cause a decrease in risk perception and in the willingness

to pay for risk reductions.

Willingness to pay for risk reduction

Willingness to pay for risk reduction is a behavioral intention to take a
precautionary action contingent on the hypothetical market for collective risk reduction.
This is the dependent variable of the study. Much of the debate over CVM concerns the
validity of respondent answers to WTP survey questions. When asked, “How much are
you willing to pay for X risk reduction?”, the respondent will give an answer, but what
does the answer mean? Many investigators working on this problem have provided
guidance to practitioners and, to the extent possible, these have been incorporated into the

study (e.g., Mitchel and Carson, 1989; Arrow et al., 1993).

71

Early design plans for the survey instrument outlined four different methods to
elicit respondent willingness to pay for ﬁre risk reductions: open-ended, payment card,
dichotomous choice, and iterative bidding (Batts, 1993). Dichotomous choice (or
referendum format), which has become the most widely used approach to valuing non-
market environmental goods (Arrow, et al., 1993; Mitchell and Carson, 1989), entails
asking respondents whether they would be willing to pay a certain dollar value for the
good in question. A range of speciﬁc speciﬁc dollar amounts are distributed randomly
among the sample and respondents simply accept or reject the offer based on the
description of the good and the price. One advantage is its resemblance to the situation a
property owner might face when asked to vote on a public project to reduce ﬁre risk.
However, this is much different than asking the same individual to estimate their
maximum willingness to pay for the risk reduction. Unfortunately, the dichotomous
choice (DC) method requires a large sample of respondents to achieve the same statistical
precision as open ended questions (Mitchell and Carson, 1989). This was the main
reason for rejecting the method in this study. Due to the small population and the high
cost of data per sample element, the more direct open-ended WTP question format with

iterative bidding was chosen to elicit the value of risk reduction.

Recently, the open-ended CVM format has come back into favor. The main
concern with open-ended elicitation methods is that the question format is less familiar
than the dichotomous choice method which is similar to voting, a familiar exercise for
most people (Ready et al., 1996; Arrow et al., 1993). However, OE methods which elicit

continuous data provide more information on respondents’ WTP and are less prone to

72
“yea-saying” than are dichotomous choice methods. Ready et a1 (1996) and Boyle et al
(1996) in their own research and reviews of previous studies found that DC elicitation

consistently generates larger WTP estimates than OE methods.

The operational deﬁnition of willingness to pay for risk reduction is the survey

response to question 23:

23 ) Earlier we determined that the probability of you losing your home to ﬁre
within the next ten years was (x) . Through a combination of public
prevention and suppression programs this risk can be reduced to (refer to
risk card) . Keeping in mind this action will only be taken if there is
suﬁ‘icient demand from the public, how much would you be willing to pay
each year, in increased property taxes, for this risk reduction?

Value of loss

Homeowner property value as listed in the Grayling Township Assessor’s
database was used as a surrogate for value at risk. Ideally, all values at risk would be
elicited from respondents (including monetary estimates of such goods as items with
great sentimental value -- i.e. photographs -- and expenses associated with the
inconvenience of losing one’s home -- i.e. hotel and commuting expense). However,
those not having had the experience of losing a home would probably ﬁnd it exceedingly
difﬁcult to produce an itemized list and assign value to such item in the course of a short

interview. Instead, we asked directly:

48) What is the value of your house and property?

73

Attitudes and beliefs

Responsibility for wildﬁre protection

The perceived role of the individual in a social group can inﬂuence whether
adjustments are made (Burton, et al., 1993). Because individual homeowners and.
government land managers have opportunities to reduce risk, the resulting mix of public
and private risk reduction will depend to some degree on the perception among
homeowners of who is primarily responsible for wildﬁre protection. The fact that
individuals have several risk reduction mechanisms available to them has important

implications for the analysis of risk reduction beneﬁts.

18) Who do you feel is ultimately responsible for protection from wildﬁres in
Crawford County? Individual homeowners, local and state government, or a
combination of homeowners and governments?

1) individual homeowners

2) share, more homeowners

3) share equally

4) share, more government

5) government

Property tax belief

Negative reaction to hypothetical market payment vehicles can inﬂuence CVM
WTP estimates (Mitchell & Carson, 1989)., A question was included in the survey
instrument to measure respondents’ beliefs about their property taxes. This serves as an
indicator of protest bidding and is speciﬁed in the two-stage model with the hypothesis
that respondents who believe that they pay too much property tax will be less likely to

participate in the hypothetical risk reduction market.

74

41) For each of the following kinds of taxes, do you believe that you pay too
much, just about the right amount, too little, or something in between? Pick the
number that
best matches your belief:
Property tax ?

I) too much

2) between too much and the right amount

3) just about the right amount

4) between too little and the right amount
5) too little

Demographic variables

Residency

There are reasons to believe that forest ﬁre risk perception and willingness to take
precautionary action is related to whether or not the wildland residence is the primary
home or a seasonal home. Seasonal homes in Michigan have relatively low occupancy
rates during the spring (Stynes & Zheng, 1995) when most forest ﬁres occur; therefore,
out-of-town homeowners are less likely to be exposed to wildﬁres and ﬁre protection
agency warnings than are permanent residents. Furthermore, Burton, et al (1993) found
that urban residents face extreme hazards that are more likely than natural hazards and are

therefore likely to discount natural hazards accordingly.

For sample selection purposes, a seasonal home is a property with a dwelling unit
(trailer, cabin, house) that is owned by someone whose mailing address on the Grayling
Township property tax database is not Grayling, MI. However, for the purpose of data

analysis a second, operational deﬁnition was used:

75
2) Is this your primary residence?

Standard demographic questions were also asked:

Education

42) On the back page of your answer sheet, please read me the letter that
represents the highest level of schooling you have completed?

I grade school (0-8)

2 some high school (9-11)

3 high school graduate (12)

4 specialized trade or art school

5 some college or junior college

6 college graduate (4-5 year degree)

7 post graduate work or degree

Age

43) From the back of your answer sheet, please read me the letter which best

represents your age group.
AGE (indicate letter)

Household income

44) Please tell me which letter best represents your total 1993 household income.

You only need to tell me the letter.
HOUSEHOLD INCOME RANGE: (indicate letter)

Gender

Recorded by interviewer.

APPENDIX C

SAMPLES, STRATIFICATION, AND WEIGHTING

APPENDIX C — Samples, Stratiﬁcation, and Weighting

1994 Sample

The ﬁrst step in identifying households was to obtain a GIS coverage of roads,
streams, and vegetation for the area from the 1980 Michigan Resource Information
System (MIRIS) to generate a map of the study area. Unfortunately, this coverage
reﬂected only a fraction of the paved roads in the study area, and did not include road
names. Commercial maps of the study area were incomplete with respect to the road
network. Further confusion with respect to the road network in the study area resulted

from the ubiquitous local practice of using multiple names for the same road.

Telephone listings, tax records, and a “911” database summary were consulted to
identify the households living in the study area. A complete and correct list of names,
addresses, and telephone numbers for households living in the study area could not be
developed due to deﬁciencies in these data sources, as noted below. Nevertheless, a
majority of the households interviewed were successfully contacted by mail before

interviews were conducted.

Telephone listings did not reﬂect unlisted numbers or households without
telephones, making identiﬁcation of seasonal residents particularly difﬁcult. Commercial
CD-ROM telephone listings included information on less than half of the households in
the study area, so the more extensive listings in the 1993 Grayling telephone book had to

be manually entered into a computer database.

76

77

Addresses obtained from telephone listings were of limited utility for contacting
households by mail, because the US Postal Service recognizes only rural route and box
numbers for a substantial fraction of the study area. An alternate address source, the
Grayling Township property tax assessor’s database, contained only legal property
descriptions (i.e., referencing township, range and section rather than street address) and
the mailing address to which property tax bills are sent. Over forty percent of these tax
bill addresses were outside the county, reﬂecting the high proportion of seasonal residents
in the study area. Even “local” tax addresses were not necessarily useful inlocating
properties, because many individuals own more than one property, and the records did not
distinguish between properties with and without structures. Ultimately, address ranges
along named roads had to be determined in the ﬁeld utilizing maps and the telephone

book database.

Starting from a list of households (with telephones) living on named roads in the
study area, attempts to schedule interviews by telephone were problematic due to low at-
home rates except in the evening hours - the same time period preferred time interviews.
Interviewers supplemented this sample by making door-to-door inquiries without prior
telephone contact. Records were kept of all households who declined to participate
regardless of contact method to allow calculation of an overall “decline—to-participate”

rate, and to ensure against further contact.

78

1996 Sample

The population and sampling method for the 1996 survey differed from those of
the 1994 survey for several reasons. First, seasonal residents make up almost half of the
property owners in Crawford County, but relatively few seasonal residents were present
in the 1994 sample. Second, an objective of the research is to investigate the effect of
past wildﬁre experience on risk perception and willingness to pay for risk reduction;
however, the 1994 sample contained very few residents burned out by the 1990 ﬁre. The
goals of the 1996 sampling protocol were to develop a sampling frame of jack pine
residents from which a random sample could be drawn, stratify the population by tenure
(permanent or seasonal residency), include all locatable residents who were burned out by
the 1990 ﬁre, exclude those included in the 1994 sample, and minimize data collection

COSt.

The MIRIS land cover/land use theme and the legal property descriptions in the
Grayling Township property tax database were combined in a GIS to identify
homeowners in areas where jack pine is the predominate cover type. The development of
this sample frame necessitated several intermediate steps to eliminate non-target property

owners.

Identifying Homeowners. Strictly speaking, the population sampling unit is
homes, not homeowner; however, the unit of analysis is the homeowner associated with
the properties in the database. Jack pine homeowners were selected in a series of steps
designed to eliminate vacant property, property outside the jack pine coverage area, and

non-residential property. A separate sample frame was used to identify residents burned

79
out by recent ﬁres. The list of residents who suffered property damage during the 1990

Stephan Bridge ﬁre was supplied by the Michigan DNR ofﬁce in Grayling.

Eliminating vacant land. A frequency distribution of 1994 survey respondents
showed that 97% of property values were greater than $10,000 State equalized value
(SEV). It was assumed that property values below this threshold indicate vacant land.
The Grayling Township Supervisor agreed that this was a reasonable assumption.
Eliminating properties assessed less than $10,000 eliminated 52.2% of the records from

the full database reducing the number of records from 5,818 to 2,780.

Eliminating property outside jack pine coverage. Each property in the database
carries a unique property code that includes Township range and section information.
Properties in sections containing less than 50% jack pine coverage were eliminated from
the database. Also, properties located in sections containing greater than 50% jack pine
coverage but having no private property in the jack pine area were eliminated from the
database. This step eliminated 60.9% of the records from the remaining 2,780 records for

a ﬁnal target population of 1,104 property owners.

Sample stratiﬁcation

Two reasons to use a stratiﬁed sample design are (1) include sufﬁcient number of
elements from small key groups and (2) to allow for different sample designs among
different groups (Frankel, 1983). The 1996 survey study exploits stratiﬁcation primarily

to include sufﬁcient numbers of seasonal and bumed-out residents.

80

The overall sample frame comprises seasonal and permanent residents and
homeowners burned out by recent ﬁres. The latter group is equally divided among the
seasonal and permanent resident subgroups (National Fire Protection Association, n.d.).
Because of the relatively small number of bumed-out residents in the population, and in
anticipation of the difﬁculty in locating some of them, we attempted to sample 100% of
this subgroup. The separate sample frame used for bumed-out residents is a subset of the
population sample frame; therefore, the 1996 sample is stratiﬁed by type of residency
(permanent/seasonal) and whether or not a loss was experienced in the SBR ﬁre
(bumout/non-bumout) to produce four strata: permanent non-bumed-out, seasonal non-

bumed-out, permanent bumed-out, and seasonal bumed-out.

Poststratiﬁcation weighting

Sample weights were developed and used in the data analysis to adjust the sample
distribution by the four strata described above. These weights were then adjusted to
correct for differential response rates within six property value classes. The sample frame
developed prior to the 1996 sampling was used as a reference source for population
distribution estimates. The following sections describe the method of deriving the

adjusted weights used in the data analysis.

Relative weights
Relative weights were developed to adjust the sample distribution by the four
strata. Relative weights are developed in two steps. First, the expansion weight is

calculated as the reciprocal of the selection probability 1/ f for each stratum. The pooled

81
sample of 1994 and 1996 households was split between four strata for selection
probabilities, f1(=166/570), f2(=52/414), f3(=28/60), and f4(=19/60). The expansion

weights w,=1/f,- (Table 6) are then adjusted to produce a relative weight, M., which is the

expansion weight divided by the mean of the expansion weights, 5 = Zwiln (Lee,

Forthoffer, & Lorimor, 1989). A relative weighted estimator for the population,

A _
therefore, is Y = W2 rwi yi.

Property value adjustment factor

An additional poststratiﬁcation adjustment was made to make the sample
composition similar to the population composition with respect to property value class —
an important predictor of willingness to pay. Property value class cut points were chosen
by dividing the sample frame distribution of property values into six equal groups. Table
7 shows how the relative weight adjustment factors were derived. The adjustment factor
is calculated for each property value class by dividing the population distribution by the
sample distribution. The adjusted weight used to analyze the data is the product of the

relative weight and the adjustment factor (Lee et al., 1989).

82

Table 5. Derivation of Poststratiﬁcation Relative Weights for Sample Strata

 

 

Population Sample Expansion Relative Weight
Sample Stratum Distribution Distribution , Weight (w,) (rw,)
Permanent
resident, non- 570 166 3.433 0.824
burnout
Seasonal resident,
non-humout 414 52 7.962 1 .91 1
Permanent
resident, burnout 60 28 2.143 0.514
Seasonal resident,
burnout 60 19 3.158 0.758
Total 1,104 265

 

Table 6. Derivation of Poststratiﬁcation Adjustment Factor for Property Value

 

 

Property Value Groups
(state equalized value in Population Distribution Sample distribution Adjustment
$1,000) factor
< 12.5 .167 .113 1.478
12.5-15.6 .167 .091 1.835
15.7-19.8 .167 .117 1.427
19.9-25 . 167 .21 1 0.791
25.1-33.8 .167 .155 1.077
>338 .167 .313 0.534
Total 1.002 1.000

 

APPENDIX D

DATA SUMMARY

APPENDIX D - Data Summary

Table 7. Respondent age
0: From the back of your answer sheet, please read me the letter which best represents your age group.

Age range Count °/o

 

<25 4 1
26-30 10 2
31 -35 20 5
36-40 26 6
41 -45 28 7
46°50 43 1 1
51 -55 38 9

56-60 42 10
61-65 52 13
66-70 51 13
71-75 41 10

>75 47 12
Total 402 100

 

Table 9. Household Income
0: Please tell me which letter best represents your total 1993 (1995) household income.
You only need to tell me the letter.

Household
income range Count %

<$10K 32 9
$10,001-15,000 34 9
$1 5001 -20,000 43 1 1
”0.001-25,000 45 12
525,001-35,000 67 18
35.001 -45.000 45 12
>545,000 109 29

Total 375 100

 

Table 9. Education level
0: Please read me the letter that represents the highest level of schooling you have completed?

 

Respondent

education Count %
Grade school 14 4
Some high school 32 8
High school grad. 96 24
Trade school 48 12
Some college 113 28
College grad. 36 9
Pea graduate 58 15

Total 397 100

 

83

84

Table 10. Number of occupants (permanent residents only)
0: How many people live here?

Number Count %

1 41 14
2 164 56
3 35 12
4 19 6
5 25 8
6 7 2
>6 4 1

Total 295 100

 

Table 11. Intention to remain In area
0: Do you plan to move or sell this home within the next ten years?

 

Count %
Yes 44 12
No 326 88

Total 370 100

 

Table 12. Number of pets'
Q: How many pets live in this house?

Number of pets Count %

0 128 43
1 79 27
2 47 1 6
3 1 7 6
4 5 2
5 9 3
>5 1 1 3

Total 296 100
' asked only in 1994

 

Table 13. Employment status'
0: What is your occupation?

 

Response Count %
Employed 129 43
Retired 170 57

Total 299 100
' asked only in 1994

 

Table 14. Marital status
0: Are you married?

 

 

 

Count %
Yes 319 81
No 74 19
Total 393 100
Table 15. Gender
Count %
Male 287 72
Female 113 28
Total 400 100

 

85

86

Table 16. Evacuation experience during Stephan Bridge ﬂre‘
0: On May 8th. 1990. the forests near Grayling were under siege by a large fire. Did you have to
evacuate your residence during the fire?

 

Count %

Yes 65 31
No 146 69
Total 21 1 100

 

'asked only in 1994

Table 17. Took steps to protect home
0: As the fire approached your home. were you able to take any steps to further protect your

 

residence?
Count %
Yes 20 16
No 109 84
Total 129 100

 

Table 18. Steps taken to protect home
0: What were the activities that you performed to protect your residence from the fire?

 

Steps Count %
Cleared vegetation/other fuel from the house 30
Watered down the root 1 5
Removed fuel from the roof 4 20
Closed all windows and doors 1 5
Other 8 40

Total 20 100

 

87

Table 19. Wildfire experience'
Now l'd like to ask you about your experiences with forest fires. Please refer to item (7) on the response card.
This is a list of experiences you may have had with forest fires in this region .
Please tell me which of these experiences you have had.

 

Experience' Count Percent
Have suffered fire loss
Yes 39 32
No 83 68
Total 122 100
Have you been evacuated
Yes 34 28
No 88 72
Total 122 100
Friends. relatives. neighbors affected
Yes 93 76
No 29 24
Total 122 100
Other iorest fire experience
Yes 61 50
No 61 50
Total 122 100
No experiences with forest fires
Yes 5 4
No 1 17 96
Total 122 100
Experiences with SBR fire?
Yes 73 60
No 49 40
Total 122 100
Experience with other fires?
Yes 63 52
No 59 48
Total 122 100

 

‘amdoniyiniees

88

Table 20. Property value
Q: What is the value of your property?

 

 

 

 

Mean $ 53,165
Median 35,500
SD 51,879
Minimum 3,000
Maximum 450,000
N 374
Property value

     

%ofsampie
8 8 ‘5’ 82

_a
0'

40

O
N
V

88§§§8§

thousands of S

200
>200

 

 

 

Figure 3. Distribution of Property Value

89

Table 21. Objective Estimate' of initial Risk

 

Count %

4% 154 40
7% 102 27
10% 31 8
14% 95 25
Total 382 1 00

 

' risk estimate by interviewer using an on-site risk assessment form and joint probability model

Table 22. Fire insurance
0: Do you carry fire insurance?

 

Count %

Yes 393 98
No 9 2
Total 402 100

 

Table 23. Type oi insurance coverage
0: What type of coverage do you carry?

 

Type Count %
None 2 1
Less than full replcement 32 11
Full replacement 259 88

Total' 293 100

 

' 28% of respondents were unsure

Table 24. Insurance cost
0: What is your annual fire insurance cost?

 

Mean $ 369
Median 300
SD 251
Minimum 20
Maximum 2.500
N' 313

 

' 23% of respondents were unsure

90

Table 25. Sources of information about wildfire impact‘
O: From which of the following information sources have you learned the most about the
impact of forest ﬁres on people in Crawford County?

 

lrmact information source Count %
Personal experience 48 39
Neighbors. friends. relatives 35 29
Media 29 24
Government 5 4
Toured SBR fire 2 2
Other 3 2

Total 122 100

 

'eskedoriyhim

Table 26. Wildfire a home choice factor
Q: Did you consider the risk of wildfire when you moved into this home?

 

Considered risk Count %
Yes 97 24
No 309 76

Total 406 1 00

 

Table 27. Government actions
D: What actions could your state and local fire protection authorities take to make fire protection and prevention
more effective in this area?

 

One or more suggested Count %
Yes 312 85
No 55 15

Total 367 100

 

91

Table 28. Length of residency
0: When did you move into this home?
(years of residence computed by subtracting response from survey year)

 

Nurnbar of years
Mean 16
Median 13
SD 13
Minimum 0
Maximum 70
N 407

 

 

Length of residency

 

 

<5 1 0 1 5 20 25 30 35 40 >40
years

 

 

 

Figure 4. Length of Residency

Table 29. Seriousness of wildfire risk when first moved in'
0: On a scale from 1 to iowith i being not serious and 10 being very serious. which nurnberbest represents how serious
you thought the risk of Mldfire was back when you purchased or built this home?

 

1 2 3 4 5 6 7 8 9 10
not very
serloiis serious

"7.... moved in

 

 

Mean 4.77

Median 5.00

SD 2.35

Minimum 1

Maximum 10

N 1 13
‘ asked only in me

 

Current seriousness of wildfire threat

 

 

 

 

 

 

Figure 5. Current Seriousness of Wildfire Threat

92

Table 30. Private risk reduction action history
Q: What have you done to reduce the risk of losing your house to fire and in what year did you do so?

 

Action When Hrs/yr‘ s/yr' Total hrs” Total 3"
Cleared brush and trees
Mean 1982 102 1.114 311 1,135
Median 1986 35 150 100 500
SD 12 294 1.737 504 1.221
Minimum 1924 1 30 4 100
Maximum 1994 2.400 5.000 2.000 4.200
N 173 113 9 56 13
Mow in fall
Mean 1979 11 745 130 15
Median 1984 4 12 4
SD 15 29 1.876 212
Minimum 1928 0 6 1 15
Maximum 1 994 280 5.000 500 1 5
N 142 117 7 9 1
Cleared yard of debris
Mean 1981 22 73 96 25
Median 1986 10 65 29
SD 13 29 57 168
Minimum 1929 1 12 0 25
Maximum 1994 200 150 500 25
N 119 93 4 8 1
Other action
Mean 1981 45 11 255
Median 1982 50 5 200
SD 13 33 14 315
Minimum 1 936 5 2 20
Maximum 1994 85 27 800
N 23 5 - - 3 5

 

" for actions that are repeated annually
" for one-time actions or actions repeated infrequently

Table 31. Past risk reduction behavior
Q: Have you done anything to reduce the risk of losing your home to wildfires?

 

N °/o

Yes 1 12 28
No 292 72
Total 404 1 00

 

Table 32. Fire protection investment
0: in your opinion. have your next-door neighbors invested more. less. or about
the same amount as your family in personal fire prevention efforts?

 

Response N %
more 81 36
about the same 115 51
less 30 13

Total 226 1 00

 

93

Table 33. Government actions suggested'|
Q: What actions could your state and local fire protection authorities take to make
fire protection and prevention more effective in this area?

 

Government action Count %
Educate the public about wildfire risks
Yes 48 51
No 47 . 49
Total 95 100
Better enforcement of fire safety laws
Yes 31 33
No 62 67
Total 93 100
Upgrade equipment or water system
Yes 14 15
No 81 85
Total 95 100
Manage public land for fire safety
Yes 1 1 12
No 82 88
Total 93 100
increase the number of firefighters
Yes 8 8
No 87 92
Total 95 100
Restrict incendiary activities of the National Guard
Yes 7 8
No 86 92
Total 93 100
Remove or plant less jack pine
Yes 6 6
No 87 94
- Total 93 100
Other government actions
Yes 68 71
No 28 29
Total 96 100

 

‘ asked only in 1996

94

Table 34. Perception of unconditional risk of wildfire In the neighborhood
0: Which percentage best represents your estimate of the chance of a wildfire moving through the block
on which you live at some time In the next ten years?

 

For sure No chance
100%-80%-60%-40%-20%-0%

Chance wildfire in the neighborhood

Mean 39
Median 40
SD 25
Minimum 0
Maximum 100
N 382

 

 

Perception of chance of fire In neighborhood

 

 

70 80 90 1M

<10 20 30 40 50 80
16 chance In ten years

Figure 6. Perception of Unconditional Risk

 

 

 

Table 35. Perception of conditional risk of wildfire destroying home
0: If a fire were to pass through your neighborhood. which percentage best represents your estimate of the c
of it destroying or severely damaging your home?
For sure No chance
100% - 80% - 60% . 40% - 20% - 0%

Chance wildfire destgys home

 

Mean 52
Median 50
SD 30
Minimum 0
Maximum 100
N 392

 

 

Perception of chance fire will destroy home

 

 

<10 20 so 40 59. chm? 70 80 90 100

 

 

 

Figure 7. Perception of Conditional Risk

95

Table 36. Implied Joint probability of fire In neighborhood and home destroyed
(product of conditional and unconditional chance estimates)

Imglled |oint grobability

 

 

 

 

 

Mean 22
Median 1 6
SD 20
Minimum 0
Maximum 100
N 373
Implied chance of wildfire in neighborhood and destroyed home
40
as
so
525
E
3 20
S
g is
10
5
0 ,
<10 20 so so so so 70 so so 100
93 chance in ten years

 

 

 

Figure 8. implied Perception of Joint Risk

Table 37. Risk reduction preference
0: Imagine that you own two different houses In two different locations. One house Is exposed to a risk

of being destroyed by a fire of 2 in 1,000. while the other house has a risk of 20 in 1.000. You cannot
avoid either of these risks. but you can choose to have one of them reduced. Which would you prefen

a) the rlskoi2 In 1.000 reduced to 1 in 1.000, or
b) the risk of 20 in 1.000 reduced to 15 in 1.000?

 

Count %
a) 77 21
291 79

Total 368 100

 

96

Table 38. Hazard Salience
Q: Each of the following four events poses some risk to you. Please rank. from 1 to 4. these events
by your estimate of the chance that they will occur. Use (1) for most likely and (4) for least likely.

W:

- a tornado destroys your home

- a car accident which sends you to the hospital
- being a victim of a burglary or robbery crime

- a wildfire destroys your home

 

 

Tornado Car accident Burglary Wildfire
Rank N % N % N % N %
1 42 10 220 55 93 23 49 12
2 78 19 87 22 151 38 87 22
3 1 11 28 56 14 84 21 152 38
4 172 43 39 1o 74 18 1 15 29

Total 403 100 402 100 402 100 403 100

 

Table 39.

97

 

 

Attitude toward taxes
0: Do you believe that you pay too much. just about the right amount. too little, or something in between?
Sales tax State income tax Property tax
Response N % N % N %
Too much 78 20 80 22 100 26
Between too much and the right amount 46 12 51 14 66 17
Just about the right amount 255 66 231 63 208 53
Between too little and the right amount 9 2 3 1 11 3
Too little 1 0 1 0 5 1
Total 389 100 366 100 390 100

Table 40.

 

Responsibility for wildfire protection

 

home-owners. local and state government. or a combination of homeowners and governments?
Response N %
Homeowners 45 11
Share. more homeowners 69 17
Equally shared 210 53
Share. more government 53 13
Government 23 6
Total 400 100

 

: Who do you feel is ultimately responsible for protection from wildfires in Crawford County? individual

98

Table 41. Attitude toward government spending on crime
O: Considering your estimate of the likelihood of a burglary or robbery occurring and the loss which you
would face do you believe government is spending too much. just about the right amount. too little or
something in between?

 

Response Count %
Too much 19 6
Between too much and the right amount 15 4
just about the right amount 161 47
Between too little and the right amount 72 21
Too little 77 22

Total 344 100

 

Table 42. Attitude toward government spending on wildfire protection
O: Considering your estimate of the likelihood of a forest fire damaging your home and the loss which you
would face. do you believe government is spending too much. just about the right amount. too little or
something in between?

 

Response Count %
Too much 5 2
Between too much and the right amount 4 1
Just about the right amount 178 58
Between too little and the right amount 61 20
Too little 61 20

Total 309 100

 

Table 43. Policy preference'
0: Would you be in favor of a program that is subsidized by the tax-dollars of all citizens which
provides low-interest loans for individual homeowners to invest in fire protection projects that reduce the
risk of their homes burning in a wildfire?

 

Moms Count %
Yes 146 58

No 107 42

Total 253 100

 

' asked only in 1994

99

Table 44. Willingness to pay for private risk reduction actions'
O: While evaluating your property for ﬁre risk. I noticed (state actions not taken e.g.. not having 30'
perimeter cleared of trees and brush). These are actions you can take to reduce the risk that forest
fire poses to your home. Are these things you would do yourself, or pay to have done?

1) Respondent indicates they would do the work themselves (ask further questions in terms of time)
2) Respondent indicates they would contract the work out to others (ask further questions in terms of money)

Keeping in mind that (state ﬁrst action to be taken) will reduce the risk of your home being
destroyed or damaged by forest fire from the current risk of (refer to risk card) to (refer to risk
card). How much timelmoney would you be willing to spend on this project to achieve this risk
reduction?

Non-zero bids‘
Final bid for ist private action Final bid for 2nd private action
SM Hrs/yr Slyr H

 

 

rs/yr
Mean 1 .325 162 335 6
Median 500 31 225 6.5
SD 2.916 442 369 4
Minimum 24 1 10 1
Maximum 15.000 3.000 1.000 10
N 29 50 6 4

' 3% ol m wrﬁ opportunity to Ruse risk Sta zero

 

WTP (In money) for 1st private risk reduction action

 

    
    

<250 500 750 1.000
dollars per year

Figure 9. Willingness to Pay (in money) for ist Private Risk Reduction Action

 

 

WTP (in time) for 1st private risk reduction action

 

<25 50 100

300 4‘” >400

 

 

M233: year
Figure 10. Willingness to Pay (in time) for 151 Private Risk Reduction Action
' This question was involved only on the 1994 survey lorrn.

 

100

Table 45. Willingness to pay to restore risk to current level'

For a moment. suppose that due to the growing scarcity of volunteer firefighters (partially due to
increased training requirements). the city and township governments are unable to provide the current
level of forest fire prevention and suppression. This increases the risks of a fire destroying your home
from the current level of (z) to (k) . (USE ALTERNATE SCENARIO OF 'DNR BUDGET CUTS
ELIMINATE ONE FIRE ENGINE' IF NECESSARY; IF ALTERNATE IS USED. NOTE HERE

) How much would you be willing to contribute in the form of increased taxes for fire
prevention and protection programs which would reduce this risk back to the original risk of (z) ? (iF
RESPONDENT HESITATES FOR MORE THAN 10 SECONDS. PROVIDE AMOUNTS BEGINNING
WITH $25.00 AND WORKING UP IN $50.00 lNTERVALS UNTIL RESPONDENT ACKNOWLEDGES A

 

 

GIVEN AMOUNT)

WTP to restore risk

Initial bid Final bid
Mean 60 144
Median 30 30
SD 137 1.030
Minimum 0 0
Maximum 1 .300 1 5.000
N 208 214

 

 

WTP to reduce risk to current level

 

Initial bid

    

°issss§e

dollersperyear

 

§

>200

 

Figure 11. Willingness to Pay to Restore Risk to Current Level: initial Bid

 

WTP to reduce risk to current level

 

Final bid

    

 

4O
35
30
E25
20
315
$10
5
0
° *5 8 i s is .8.
dollarsperyear

§ §

 

Figure 12. Willingness to Pay to Restore Risk to Current Level: Final Bid

‘askedonly in 1996

 

 

101

Table 46. Willingness to pay for public risk reduction actions
0:

Earlier we determined that the probability of you losing your home to fire within the next ten years

was (x)__ Thmunh n

radioed to _(reier to risk card)_. Keeping' in mind this action will onlyh bei taken if there Is sufficient
demandfromthepublic. howmuchwo wouldyoubewlllingtopayeachyea

for this risk reduction? (IF RESPONDENT HESITATES FOR MORE THANh 10 SECONDS, PROVIDE
AMOUNTS BEGINNING WITH $25 00 AND WORKING UP IN :50 00 INTERVALS UNTIL

RESPONDENT ACKNOWLEDGES A GIVEN AMOUNT)

WW:
Assume that these ﬁrst projects are adoptwed. reducing the risk of"" forest fire damaging or destroying

your home to __(a)

additionaf pibiic programs. your risk can fwurther be reduced from""

_(final WTP)_
_(ﬂ)_II'J-._(bl_Koo1:Ing In mind

thisactionwillonfybetakenifthereissufficientdemandfromthapubllc.howmuchwouldyoubewllling
to pay in a further increase In annual property taxes for this risk reduction?

1st ublic action 2nd Mic action

Initial bid Final bid Initial bid Final bid
Mean 43 74 15 16
Median 25 40 0 0
SD 66 127 37 39
erilrrium - - 0 0
Maximum 500 1.000 300 300
N 357 360 269 270

 

 

WTP for 1st public action: lnitlal bid

 

 

 

'" Ki
5 — '-
doiErs peryear A

em: a

..
0°

WTP for 2nd public action: initial bid

 

 

  

a} ..

'7 9
dolﬁrsperyeu

 

>200

 

WTP for 1st public action: final bid

 

o «*g'Wa a

°iss8nsi§§

dollirs per’year A

 

 

 

oa§°mga

WTP for 2nd public action: final bid

 

 

 

 

 

Figure 13. Willingness to Pay for Public Risk Reduction Actions

APPENDIX E

1996 SURVEY INSTRUMENT

102
FIRE RISK VALUATION SURVEY

Form SB 0396 Stephan Bridge/Luzerne Fires

Before entering house, complete the MDNFi risk rating form (skipping the double pane
windows assessment), assess the probability of house loss (and note appropriate
starting risk card below). and complete name/address/survey code sheet.

Risk rating from form (0-100):

Public Risk Card used (a-d, bb or dd):
Approximate parcel size (acres):
Survey code (also write on “handout”):
Start time:

 

 

 

 

 

Hello MrJMrs/Miss (respondent's name: memorize before intervigwj . I’m

Greg Winter from Michigan State University’s Department of Forestry. We’re talking to
people in Crawford County about forest fires. The survey is funded by the US. Forest
Service with cooperation from the Michigan Department of Natural Resources. We are
interested in addressing the wildfire problem in Crawford County, but it’s hard to know
where to begin without a better understanding of how important this issue is to you. It is
critical that we know your concerns and values before considering changes in policies.
We’re interviewing you and other homeowners in your neighborhood to find out about
your perceptions of ﬁre risk and the value you place on reducing that risk.

Most of this survey concerns your opinions and attitudes. so there are no right or wrong
answers. Your responses will be kept entirely confidential by us and your name will not
be associated with your answers in any reports, nor will your name appear in any report.
The survey will take approximately 30 to 40 minutes and I think you will find it a good
ieaming experience.

Section I: BackgroundI perception, and context guestions
First, l'd like to ask you a few questions about your property and your home.

1) What year did you move into this home?

1) _—
-2) CAN'T REMEMBER (probe for estimate)
-3) REFUSED

1.1 (If they moved into their home in 1990, get month (1-12) _)

2) Do you consider this to be your primary residence?
1) yes
2) no
-3) REFUSED

103
3) How many people live here?

1 NUMBER:
-2 REFUSED

 

4) Did you consider the risk of wildfire when you moved into this home?

1) yes
2) no
-3) DK
-4) Refused

4b) Were you aware of the wildfire risk in this region at the time you moved here?

1) yes
2) no
-3) DK
-4) Refused

4c) On a scale from 1 to 10 with 1 being not serious and 10 being very serious, which
number best represents how serious you thought the risk of wildfire was back when
you purchased or built this home? [refer to card for scale]

Number (-2) DK -3) Refused)

5) Do you plan to move or sell this home within the next ten years?

1) Yes
2) No
3) DK
4) Refused

 

 

6) is the risk of fire a...?

1) major motivation

2) minor motivation

3) not a motivation

-4) DON'T KNOW (go to question 5)
-5) REFUSED (go to question 5)

 

 

 

104
6) From which of the following information sources have you ieamed the most

about the impact of fires on people in Crawford County?
1) Neighbors. friends, relatives
2) Media sources: newspaper. television, radio
3) Federal, state, or local gov’t (brochures. ads, agency officials)
4) Personal experience with fires '
5) Other sources:
6) DK
7) Refused

 

7) Now l’d like to ask you about your experiences with forest fires. Please refer to
Item (7) on the response card. This is a list of experiences you may have had with
forest fires in this region (the region near your vacation home). Please tell me which of
these experiences you have had.

 

1) Forest fire has caused damage to my property _’ Go to O 8

2) Have had to evacuate due to threat of forest fire in the region

 

 

 

3) Have friends. relatives or neighbors affected by forest fire
4) Other experience —’ Skip to O 13

5) No experiences with forest fires in the region

 

 

8) Were these experiences with the 1990 fire at Stephen Bridge Road or another fire?
1) Stephan Bridge fire
2) Other fire When

3) DK / Can’t remember

105

Questions for those having suffered damage from the 1990 Stephan
Bridge fire, or the 1992 Luzerne fire

 

 

9) As the tire approached your home. were you able to take any steps to
further protect your residence? (iF RESPONDENT IS UNSURE OF
WHAT CONSTITUTES A FIRE PROTECTION ACTIVITY, READ LIST IN
W1)

W

2) No
-3) Can’t remember/ DK -> Skip to O 11
-4) REFUSED

 

 

 

 

 

 

 

 

 

 

10) What were the activities that you performed to protect your residence
from the fire? (if respondent hesitates, read the list below) (circle
appropriate numbers)

10.3) Cleared away brush, vegetation, and other fuel from the
house

10.4) Watered down the roof

10.5) Removed fuel from the roof

10.6) Closed all windows and doors

10.7) Other (list all):

 

 

 

11) Describe and estimate the dollar value of fire-related losses and any
associated expenses you experienced?

 

1 ) Amount

 

2) No
-3) CAN'T REMEMBER/DON'T KNOW
-4) REFUSED

12) Which of these were not reimbursed by insurance and what was the
total amount of your unreimbursed losses?

 

1) Amount
2) No
~3) Can’t remember/ DK
-4) Refused

 

 

 

106

13) Which percentage best represents your estimate of the chance of a wildfire
moving through the block on which you live at some time in the next ten years.
[Refer R to card]

Stated Risk
-2) DK (probe)
-3) Refused

 

14) if a fire were to pass through your neighborhood, which percentage best
represents your estimate of the chance of it destroying your home?

Stated Risk
-2) DK (probe)
-3) Refused

 

15) Each of the following four events poses some risk to you. Please rank , from 1 to 4.
these events by your estimate of the chance they will occur. Use (1) for most likely
and (4) for least likely.

The events are:

1) a tornado destroys your home _

2) a car accident which sends you to the hospital _
3) being a victim of a burglary or robbery crime __
4) a wildfire destroys your home __

You ranked the risk of burglary/robbery (give rankings from Q8) and fire
. Think about how you feel about the current level of government

spending in these areas. For 2 of these areas. crime & fire. i will be asking for your
Opinion about changes in the level of spending.

16) Considering your estimate of the chance of a burglary or robbery occurring and the
loss which you would experience if a burglary occurred, do you believe
government is spending too much, just about the right amount. too little or
something in between?

Burglary/Robbery
1) Too much
2)
3) Just about the right amount
4)
5) too little
-6) DK
-7) Refused

107

17) Considering your estimate of the chance of a forest fire damaging your home and
the loss you would experience if one did. do you believe government is spending
too much, just about the right amount. too little or something in between?

Fire
1) Too much
2)
3) Just about the right amount
4)
5) too little
-6) DK
-7) Refused

18) Who do you feel is ultimately responsible for protection from wildfires in Crawford
County? individual homeowners, local and state government, or a combination of
homeowners and governments?

1) individual homeowners

2) share. more homeowners
3) equally

4) share, more government
5) government

-6) DK

-7) Refused

19) What actions could your state and local fire protection authorities take to make fire
protection and prevention more effective in this area? (IF THE RESPONDENT
HESITATES READ THE LIST BELOW)

1 one or more actions suggested
2 no actions suggested

-8 DON’T KNOW

-9 REFUSED

19.0 prompt. Enter Y if prompting was needed

 

 

19.1) Increase the number of firefighters

19.2) Upgrade equipment and expand the reliability of water systems
19.3) Widen or realign streets to improve firefighter access

19.4) Underground relocation of electrical distribution lines

19.5) Provide free training in fire evacuation and prevention techniques

19.6) Educational programs
19.7) Other:

 

108

20) Have you done anything to reduce the risk of losing your home to wildfires? (if
respondent unsure of what constitutes fire protection activities, read list below)

 

 

1) Yes

 

 

2) No ,
-3) DK -> Define risk and hand card
-4) Refused

 

 

 

Circle all that apply:

 

20.1) Cleared brush and trees away from house

20.2) Mow in the fail

20.3) Cleared yard of flammable debris

20.4) Screened vents to prevent entry of burning embers and
sparks

20.5) Installed external sprinklers

20.6) Other actions

 

 

 

 

21) How much have you spent on such measures since the 1990
Stephan Bridge (Luzerne) fire?

1 L
2 DK (PROBE)
3 Refused

DEFINE THEIR CURRENT RISK, AND HAND RISK CARD

According to our calculations, the probability of a fire passing through your

neighborhood and destroying or damaging your home within the next ten years is
percent.

There are two ways of reducing the risk of losing your home to wildfire.

One is through private actions that you take as an individual. For example, you as a
homeowner can clear trees and brush around your house, mow the grass, and keep
flammable materials away from the house. A second is through actions financed by the
public. This would include additional spending on fire stations, staff, equipment and
other infrastructure like water systems.

I will ask you about public financed approaches to reducing the risk of forest fire. This
risk card represents the likelihood of your house being destroyed by a wildfire within the
next 10 years. The risk is represented in three different ways: as a percentage, a ratio,
and as a probability grid. You face an % chance of losing your home, which is
equivalent to 1 in , or the chance that your home is one of the black squares on the
grid, where the black squares represent homes which are destroyed or damaged by
wildfire and the white squares represent homes which are not affected.

 

109
21) Do you understand how private and public spending affects the different fire risks?

1 ) Yes

 

2) No ——’ Try to clarify the misunderstanding

 

 

 

The USFS, MichiganDepartment of Natural Resources, and other fire protection
organizations are considering changes that would improve ﬁre protection. Through a
package of public actions (e.g., increasing the number of firefighters, upgrading
equipment, etc.,) forest fire suppression and prevention will increase, making it less
likely that that a ﬁre will pass through your neighborhood and therefore less likely that
your home will be damaged or destroyed by forest fire.

emphasis:
These actions will only be taken if residents are willing to pay for them.

SECTION Ilb. VALUATION A_ND WTP FOR RISK REDUCTIQN / PUBLIC
EXPENDITURES

 

t = (AMOUNT OF THE RESPONDENT'S PROPERTY TAX THAT
ROTECTION AS DETERMINED by .002‘assessed value of property)
House value (3): How determined (TWP/ASKIESTIOTH):

110

According to our calculations, approximately dollars per year of your property
taxes is spent on fire protection and prevention by the city and township of Grayling and
by the Michigan DNR.

23) Earlier we determined that the probability of you losing your home to fire within the
next ten years was (x) . Through a combination of public prevention and
suppression programs this risk can be reduced to . (refer to risk card) .
Keeping in mind this action will only be taken if there is sufficient demand from the
public, how much would you be willing to pay each year, in increased property
taxes, for this risk reduction? (IF RESPONDENT HESITA TES FOR MORE THAN
10 SECONDS, PROVIDE AMOUNTS BEGINNING WITH $25.00 AND WORKING
UP IN $50.00 INTERVALS UNTIL RESPONDENT ACKNOWLEDGES A GIVEN
AMOUN'D

23.1) Beginning Risk 23.2) Risk with change

 

23.3) Initially stated WTP 23.4) WTP after iterative bidding

 

 

 

 

 

    

 

0) Zero or “nothing” 1
-2) DK Skip to Q 32
-3) Refused

 

 

24) Would it be worth anything to you (your household) if the risk
reduction achieved through these programs were larger, say in
your case, the probability of losing your home to fire in the next

ten years would be reduced from (x) to (a) ? (use
entire risk increment)

 

 

V

 

 

 

 

 

 

 

 

 

1) Yes ’ initial amount
2) No £— After bidding
-3) DK Skip to G 35

4) Refused

 

 

 

 

If “zero” on 023 and “No” on this
L—> question, skip to next question. All others
skip to 026.

 

 

 

 

111
ASK ONLY IF ANSWERED “ZERO” TO 0 23 AND “NO” TO 0 24, ASK:

 

25) People have different reasons for saying zero dollars or nothing. For some
people that is all that wildfire risk reduction is worth to them. They don’t want
to pay anything for it. Other people give different reasons for saying zero or
nothing. Did you say zero dollars because that is what wildfire risk reduction is
worth to you (your household) or because of other reasons?

1) That’s what it’s worth to me —’ Skip to O 35

 

 

2) Other reason Go to G 26
-4) DK /7 Skip to O 35
-5) Refused

 

 

 

 

 

 

IF “DON’T KNOW” OR “REFUSED” TO 023, AND “DON’T KNOW" OR “REFUSED” TO
0 24 :

People have different reasons for saying they don't know or can’t answer these
questions. I’m going to read you some reasons. Please tell me whether or not
they represent your feelings about this question.

26) Did you give this answer because you think the government should be able to
reduce the risk with the money they have or because you think the government
. wastes too much money?

 

1) Gov’t should be able to keep risk low with money they have.

 

 

 

 

2) Gov't wastes too much money. ,__, Go to O 27
3) No
4) DK —’ Skip to Q 28

5) Refused

 

 

112
IF “NO”, “DON’T KNOW”, OR “REFUSED” ON 0 26, ASK:

27) It is very important to us to learn the value you place on reducing the risk of wildfire
damage to your property when you are given the chance to make the choice
yourself. This value is the highest amount you would be willing to pay for an
efficient and effective program to achieve each of the risk reductions i am asking
about. Would you be willing to answer these questions if I noted here that the
amounts you give are based on the assumption that the fire protection programs
be efficient, effective, and well managed?

1) Yes ———> Go back to Q 23

 

2) No
-3) UK ‘—> Skip to O 35
-4) Refused

 

 

 

28) Did you give this answer because you are (your household is) paying too much in
taxes already and don’t want to spend more?

1) Yes _, Skip to G 35

 

2) No
-3) DK —> Go to O 29

-4) Refused

 

 

 

IF “NO”, “DON'T KNOW”, OR “REFUSED” ON 0 28, ASK:

29) Did you give this answer because you think it is not possible for improvements in
public fire protection programs to achieve these risk reduction levels?

1) Yes —> Skip to Q 35

 

2) No
-3) UK —> Go to O 30

-4) Refused

 

 

 

113
IF “NO”, “DON’T KNOW", OR “REFUSED” ON 0 29, ASK:

30) Did you give this answer because you would rather take responsibility yourself for
the protection of your property from wildfire damage.

 

 

 

1) Yes ‘—’ Skip to 0 35
2) No

-3) DK ,_, Go to 0 31
-4) Refused

 

 

 

IF “NO”, “DON’T KNOW", OR “REFUSED” ON Q 30, ASK:

31) is there a reason why you gave this answer (Answer to 034 and Q 35) other than
the ones I’ve asked about?

 

 

 

 

 

1) Yes
Reason
2) No L» Evaluate and skip to 0 33 or 0 35
-3) DK
-4) Refused —’ Go to 0 35

 

 

 

32) Assume that these first projects are adopted, reducing the risk of forest fire
damaging or destroying your home to (a) , while increasing your annual
property taxes by (final WTP) . Through additional public programs, your
risk can be reduced further, from _(a)_ to _(b)_. Keeping in mind this action
will only be taken if there is sufficient willingness to pay for it on the part of the
public, how much would you be willing to pay in additional annual property taxes for
LIE additional risk reduction?

 

 

Beginning Risk Risk with change
initially stated WTP WTP after iterative bidding
-2) DK

-3) Refused

114
35) Do you carry fire insurance?

1 Yes

 

 

2No

-3 DK ‘—’ Skip to 0 37
-4 Refused

 

 

 

 

V

 

36) What type of coverage do you carry (replacement cost? living
expenses? for how long?) and what is your annual fire insurance
cost?

1) Coverage:
2) Cost:
-2 DON'T KNOW
-3 REFUSED

37) Imagine that you own two different houses in two different locations. One house is
exposed to a risk of being destroyed by a fire of 2 in 1,000, while the other house
has a risk of 20 in 1,000. You cannot avoid either of these risks, but you can
choose to have one of them reduced. Which would you prefer:

 

 

 

 

 

a) the risk of 2 in 1,000 reduced to 1 in 1,000, or
b) the risk of 20 in 1,000 reduced to 15 in 1,000?

1) a

2) b

-3) DK (Probe)
-4) Refused

38) On a scale from 1 to 10 with 1 being not serious and 10 being very serious, which
number best represents your description of the seriousness of the wildfire risk to you
and your property? [refer to card for scale]
Number
39) For each of the following kinds of taxes, do you believe that you pay too much,
just about the right amount, too little, or something in between? Pick the number that
best matches your belief:
State income tax?
1 too much
2 between too much and the right amount
3 just about the right amount
4 between too little and the right amount
5 too little
-6 DON'T KNOW
-7 REFUSED

115

40) Sales tax?
1 too much
2 between too much and the right amount
3 just about the right amount
4 between too little and the right amount
5 too little
-6 DON'T KNOW
-7 REFUSED

41) Property tax ?
1 too much
2 between too much and the right amount
3 just about the right amount
4 between too little and the right amount
5 too little
-6 DON'T KNOW
-7 REFUSED

SECTION III DEMOGRAPHICS

42) On the back page of your answer sheet, please read me the letter that represents
the highest level of schooling you have completed?

grade school (0-8)

some high school (9-11)

high school graduate (12)
specialized trade or art school
some college or junior college
college graduate (4-5 year degree)
post graduate work or degree

-8 DON'T KNOW

-9 REFUSED

mehQN-A

43) From the back of your answer sheet, please read me the letter which best
represents your age group.

1. AGE (indicate letter)
-2. REFUSED

44) Please tell me which letter best represents your total 1993 household income. You
only need to tell me the letter.

1 HOUSEHOLD INCOME RANGE: (indicate letter)
-2 REFUSED

116

45) Finally, think back to the questions concerning what you would pay to reduce your
wildfire risk, your understanding of those questions, and the responses that you
gave. Are you very sure, somewhat sure, somewhat unsure or very unsure of your
responses.

very sure somewhat sure somewhat unsure very unsure

46) interviewer believes (answer BEFORE asking 34)

very sure somewhat sure somewhat unsure very unsure

47) Receive research summary? yes no

48) What is the value of your house and property?
DK Refused

49) Are you married?

 

1 yes
2 no
i. interviewer should try to note if there is a live in partner without asking
directly.
yes=1 no=0 uncertainz. (circle one)
-3 REFUSED
50) ASK ONLY IF NOT OBVIOUS. How would you describe your ethnic or racial
background?

1 European-American

2 Hispanic (includes Chicano and Mexican)
3 African-American

4 Asian

5 North American Indian

6 Mixed

7 Other:

-8 DON'T KNOW

-9 REFUSED

 

51) GENDER

1 Male
2 Female

117

53) Focus group invitation: As a follow-up to our survey research, we hope to
interview you and others who participated in this survey in a group setting at a
convenient location in Grayling. Would you be willing to participate in a group
interview later this summer? Yes No

54) [Respondents who lack a defensible space due to proximity and density of trees]
The trees growing near your house increase your fire risk, why haven’t you removed
them? ‘

55) Additional notes:

52) End time:

 

APPENDIX F

SURVEY HANDOUTS

118

MICHIGANSTATE
UNIVERSITY

 

Department of Forestry
and

USDA Forest Service
North Central Experiment Station, E. Lansing

in cooperation with

Michigan Department of Natural Resources
Forest Management Division

FIRE RISK VALUATION SURVEY WORKSHEET
To accompany Form SBB, Stephen Bridge Fire

Survey code:

 

119

4c) On a scale from 1 to 10 with 1 being not serious and 10 being very
serious, which number best represents how serious you thought the
risk of wildfire was back when you purchased or built this home?

 

 

1 2 3 4 5 6 7 8 9 10

 

Not serious Very serious

 

6) From which of the following information sources have you learned the
most about the impact of fires on people in Crawford County?

1) Neighbors, friends, relatives

2) Media sources: newspaper, television, radio

3) Federal, state, or local gov’t (brochures, ads, agency ofﬁcials)
4) Personal experience with ﬁres

5) Other sources:

 

7) Please tell me which of these experiences you have had.
1) Forest ﬁre has caused damage to my property
2) Have had to evacuate due to threat of forest ﬁre in the region
3) Have friends, relatives or neighbors affected by forest ﬁre

4) Other experience

 

5) No experiences with forest ﬁres in the region

120

13) Which percentage best represents your estimate of the chance of a
wildﬁre moving through the block on which you live at some time
in the next ten years.

 

For _ . . , ' I I No}
rsure”. i 7 Chance

100% . 80% . 60% . 40% . 20% . 0%

14) If a ﬁre were to pass through your neighborhood, which of the
following best represents the chance of it destroying or severely
damaging your home?

 

iisure - '   . ‘ '1 ' .'Chance;:

100(70 O 800/0 0 600/0 0 ' 400/0 ‘0 20(70 0 00/0

121

15) Each of the following four events poses some risk to you. Please
rank , from 1 to 4, these events by your estimate of the chance they will
occur. Use (1) for most likely and (4) for least likely.

The events are:

__ A) a tornado destroys your home
__ B) a car accident which sends you to the hospital
_ C) being a victim of a burglary or robbery crime

D) a wildﬁre destroys your home

For following questions please select a number from 1 to 5.

16) Considering your estimate of the likelihood of a burglary or robbery
occurring and the loss which you would face do you believe

government is spending too muchI just about the right amount or too
little ?

 

 

122

17) Considering your estimate of the likelihood of a forest fire damaging
your home and the loss which you would face, do you believe

government is spending too much, just about the right amount or too
little ?

 

 

 

 

 

lust abbu t e
the right
A 1, JM ,— J.,,Lamount , ,g .. .
1:1 to 2 ‘ 3 "p4,;

 

 

18) Who do you feel is ultimately responsible for protection from
wildﬁres in Crawford County? individual homeowners, local and
state government, or a combination of both homeowners and
government?

 

   

 

 

, home eqdallym h local .
2 owners ,i;§}_;_f,_fﬁjj responsible governmen

PLEASE STOP HERE

123

 

 

 

0.53?»
3on 98:38an 3335 o
wamﬁab Hoiwﬁoum who—2 o
mmoooe
notes now $63 255:: o
33:35”;
Event—«mop ohm £63me a
mumanwmoum v.85 334 o

 

233. 335

 

 

wcmooa anemic“
team :33 moon mum—mom o
min—mu oEmEEmE 258mm o
=3 3: E 252 o

:33 ER womb RED o
2839,54 355.5

 

 

 

 

124

37) Imagine that you own two different houses in two different
locations. One house is exposed to a risk of being destroyed by a ﬁre
of 2 in 1,000, while the other house has a risk of 20 in 1,000. You
cannot avoid either of these risks, but you can choose to have one of
them reduced. Which would you prefer:

a) the risk of 2 in 1,000 reduced to 1 in 1,000,
OR

b) the risk of 20 in 1,000 reduced to 15 in 1,000?

38) On a scale from 1 to 10 with 1 being not serious and 10 being very
serious, which number best represents your desCription of the
seriousness of the wildﬁre risk to you and your property?

 

 

1 2 3 4 5 6 7 8 9 10

 

Not serious Very serious

 

125

 

For each of the following kinds of taxes, do you believe that you pay too much, just
about the right amount, too little, or something in between? Pick the number that
best matches your belief:

 

39) State income tax?

 

3‘ WV ? R» __ .3...) )3“;sz '.‘

nght
Amount

 

   

 

40) Sales tax? -

 

 

 

 

' H-xS-l'arﬂw ~W’7ﬁ’: A .. {71‘1"11 1. ““.’7'7‘.‘. -14 17.-4227‘; .. :n1-'-,--"{,".’.l mﬁW.ZWr--
.

‘1.- ,. — nght . ._ g, -- Too»,
EvarMudl '1 l ' Amount Littlel

 

agate-a- .~ '1 are :‘ﬁ .

‘1 2‘ 3 -4 ..5

 

126

For the following, you need only circle the appropriate letter
Which of the following categories best represents you?

 

127

45.Finally, think back to the questions concerning what you would pay to
reduce your wildﬁre risk, your understanding of those questions, and
the responses that you gave. Are you very sure, somewhat sure,
somewhat unsure or very unsure of your responses (please circle one).

very sure somewhat sure somewhat very unsure
unsure

47.Would you like to receive a report summarizing the ﬁndings of this
research (please circle one)? yes no

"11111111111111ES