Pilot Study: Report

National Park Service
Visibility Valuation Study:
Pilot Survey Results
Final | 18 April, 2013

prepared for:
Susan Johnson
Chief, Policy, Planning and Permit Review Branch
National Park Service
Air Resources Division

prepared by:
Industrial Economics, Incorporated
2067 Massachusetts Avenue
Cambridge, MA 02140

This page intentionally left blank.

EXECUTIVE SUMMARY

To support its mission under the Organic Act, and its consultative role on regulatory
measures to achieve Clean Air Act requirements, the National Park Service is conducting
a visibility valuation study. Following focus groups and a peer review of survey
materials a pilot study was conducted in late summer and early fall of 2012. A mail
survey was administered to a random sample of 4,000 households in the southwestern and
southeastern U.S. Response rates for the southwest and southeast surveys were 38.6 and
32.5 percent, respectively. Telephone and mail follow-up surveys of nonrespondents
were also conducted. A comparison of “benchmarking” question responses to wellestablished public opinion survey results, as well as respondent characteristics to Census
data, indicates that survey respondents are similar to, but not fully representative of, the
general populations of these regions. Analysis of valuation question responses indicates
that the magnitude of visibility improvement and the occurrence of related ecological and
human health improvements are significant determinants of program choices. Household
willingness-to-pay (WTP) for visibility improvements increases with programs that
reduce the number of lowest visibility days and increase the number of highest visibility
days over the course of a year. Models based on data weighted to reflect general
population parameters result in WTP estimates that are generally between +/- 10 percent
of unweighted estimates. Overall, the pilot study results indicate that the survey
instrument is functioning properly and is ready for full implementation with minor
revisions.

This page intentionally left blank.

INTRODUCTION
& BACKGROUND

To support its mission under the Organic Act, and its consultative role on regulatory
measures to achieve Clean Air Act requirements, the National Park Service (NPS) is
conducting a visibility valuation study. The study is designed to elicit the general
population’s willingness-to-pay (WTP) for visibility improvements in Class I areas that
would arise from reductions in haze from human sources.1 While the study is designed
specifically to estimate the benefits of improvements anticipated due to the Regional
Haze Rule (40 CFR Part 51), the results may be applied to assess the benefits of other
programs or policies that improve visibility conditions in national parks and wilderness
areas.
The study team includes the following individuals:


Robert Paterson, Principal, IEc- Project Director



Dr. Kevin Boyle, Virginia Tech- Co-Principal Investigator



Dr. Richard Carson, University of California, San Diego- Co-Principal
Investigator



Dr. Barbara Kanninen, BK Econometrics LLC



Dr. Christopher Leggett, Statistics and Economics Consulting



John Molenar, Vice-President, Air Resource Specialists

This report describes the procedures and results of a pilot study conducted in the late
summer and early fall of 2012. The study involved administration of a mail survey in
two multi-state regions, with telephone and mail follow-up surveys of nonrespondents.
The following sections discuss survey design and pre-testing procedures, pilot design and
implementation, survey responses and WTP estimates, and associated implications for the
full survey.

S U RV E Y D E S I G N &
PRE-TESTING

The survey design process comprised four phases. The first involved a comprehensive
review of existing visibility valuation literature to identify key issues and challenges. The
review included an inventory of stakeholder comments on the Chestnut and Rowe (1990)
study, the current basis for regulatory analyses involving recreational visibility. The
review identified four principal issues:

1

Specific visibility provisions were included in Sections 169A, 169B, and 110(a)(2)(j) of the Clean Air Act that directed the
Environmental Protection Agency, the states, and federal land managers to prevent any future, and remedy any existing,
human-induced visibility impairment at mandatory Federal Class I areas. Mandatory Federal Class I areas ("Class I areas") are
defined as national parks exceeding 6000 acres, wilderness areas and national memorial parks exceeding 5000 acres, and all
international parks that were in existence on August 7, 1977. There are 156 Class I areas throughout the country.

1



Collateral benefits. To avoid double counting health and ecological benefits in
regulatory analyses, estimates of WTP for visibility improvements must exclude
any health and ecological concerns.



Depiction of visibility changes. Improvements brought about by a pollution
reduction program will vary over the course of a year. Reducing changes to a
simplified measure (e.g., a change in average conditions) does not accurately
portray visibility improvements and may not be sufficient to describe changes in
visibility to survey respondents.



Isolating WTP for improvements in Class I areas. Residential and urban
visibility changes are measured separately in regulatory analyses; therefore,
development of a plausible scenario that focuses respondents’ valuation responses
on Class I areas only is necessary.



Geographic coverage. Baseline and improved visibility conditions, as well as the
number, type and characteristics of Class I areas, vary by region of the country.
Thus, multiple survey versions may be required to reflect these differences.

The second phase entailed development of a study plan that identified strategies for
addressing the above issues. The plan envisioned the use of attribute-based or choice
modeling techniques to present respondents with alternative visibility improvement
programs that would vary with respect to the extent of improvement, health and/or
ecological benefits, timing of the improvements, and cost. Baseline visibility conditions
would be described to respondents as a distribution of days over the course of the year
using multiple photographs, and improvements would be described as changes in the
distribution of days associated with each photo. Improvements would be described as
occurring within a “visibility improvement region”- geographic regions containing sets of
Class I areas that are roughly homogenous with respect to current visibility levels and
potential visibility improvements. The regions would exclude large cities in order to
minimize the potential for respondents to inadvertently include consideration of
improvements in urban visibility.
In the third phase, five sets of focus groups were conducted in different regions of the
country: Atlanta, GA; Chicago, IL; Sacramento, CA; Denver, CO; and, Boston, MA.
Four focus group sessions (two, two-hour sessions per evening on two consecutive
evenings) were held in each location, for a total of 20 groups. All respondents were
recruited at random from listed telephone numbers (i.e., not facility panel members) and
the groups were led by a professional moderator who is an economist. The target was to
have 8 to 10 participants per group.
The first groups were conducted in Atlanta and explored concepts, terminology, images
and graphics in an open-ended format. The focus group effort concluded in Boston with
participants responding to the full questionnaire.

2

In the fourth phase, all survey materials were peer reviewed by Dr. Vic Adamowicz
(Department of Rural Economy, University of Alberta) and Dr. William Schulze
(Department of Applied Economics and Management, Cornell University). Comments
from these experts were incorporated and final materials for the pilot survey were
developed.

PILOT STUDY
DESIGN &
I M P L E M E N TAT I O N

Based on the geographic distribution of current and potential improved visibility
conditions, the contiguous 48 states were divided into seven survey regions. Two regions
were selected for pilot implementation (Exhibit 1): “Four Corners” (Utah, Arizona, New
Mexico and Colorado) and “Southeast” (Delaware, Maryland, Virginia, West Virginia,
Kentucky, Tennessee, North Carolina, South Carolina, Georgia, Alabama, Mississippi
and Florida). These regions were selected because they cover a range of current and
expected future visibility conditions, and are areas where previous visibility valuation
research has been conducted (e.g., Chestnut and Rowe, 1990; Balson et al., 1990).
The survey was administered by the Center for Survey Research at Virginia Tech and
overseen by its director, Dr. Susan Willis-Walton. For each region a sample of 2,000
addresses from the U.S. Postal Service Computerized Delivery Sequence File was
acquired from Survey Sampling International. The mailing sequence was as follows:
1) A personalized pre-survey contact letter (Appendix A)
2) Main survey, including a fold-out picture set, map, cover letter, $2 bill and prepaid return envelope (Appendices B and C)
3) Reminder postcard
4) Replacement survey (if no response received)

As part of the survey administration, a nonrespondent follow-up was conducted
approximately six weeks after the initial survey mailing. All nonrespondent households
where a telephone number could be matched to the address (431 in the Four Corners
region and 576 in the Southeast region) were contacted to complete a short
nonrespondent survey (Appendix D). In addition, a sample of 600 of the remaining
nonrespondent households in each region without matched phone numbers was sent a
nonrespondent survey, via Priority Mail, with the same set of questions (Appendix E).

3

EXHIBIT 1

P I L O T S T U D Y S U RV E Y A N D V I S I B I L I T Y I M P R O V E M E N T R E G I O N S

4

M A I N S U RV E Y

The main survey contained seven sections:
1) Section A contained two background questions intended to orient the respondent
to the context of implementing and funding public programs and to gauge their
confidence in various institutions; these questions were adapted from the
National Opinion Research Center General Social Survey.
2) Section B provided information on haze and its effects on visibility.
3) Section C provided background information on national parks and wilderness
areas. Respondents were referred to an enclosed fold-out map that displayed the
visibility improvement region(s), Class I areas contained within the region(s), and
major sources of haze.
4) Section D provided information on the sources of haze affecting the region(s).
5) Section E provided information on current visibility conditions portrayed in an
accompanying picture set, and example visibility improvement programs.
6) Section F provided information on each of the choice question attributesecosystem impacts, health impacts, program timing and cost (visibility
improvements were addressed in the previous section) and the six valuation
choice questions.
7) Section G contained follow-up and standard demographic questions.

All sections of the survey contained questions that were designed to help respondents
focus on the information being presented or to collect data to be used in statistical
analyses.
Re p r e s e n t a t i o n o f B a s e l i n e a n d I m p r o v e d Vi s i b i l i t y C o n d i t i o n s

The framework for presenting baseline and improved visibility conditions within the
survey derives from the Regional Haze Rule (“Rule”), which requires states to develop
and implement plans for making “reasonable progress” toward achieving natural visibility
conditions in Class I areas by 2064.
Natural visibility conditions are defined as the distribution of visibility that would exist in
the absence of human-induced impairment. The Rule requires that states focus on
improving visibility on the haziest days of the year, defined as the "worst 20 percent
days," or all days falling below the 20th percentile of the visibility distribution. This
improvement must occur while preventing any degradation in visibility on the clearest
days of the year, defined as the "best 20 percent days," or all days falling above the 80th
percentile of the visibility distribution. For the mean of the worst 20 percent days, the
Rule requires that states consider visibility goals that would be consistent with a uniform
rate of progress (i.e., linear through time) in visibility improvement toward the mean of
the worst 20 percent days under natural visibility conditions. The Rule stipulates that
visibility goals should be expressed in "deciview" units. An increase in deciviews
corresponds to an increase in haze (and a decrease in visibility).
5

As shown in Appendix B and C, respondents were provided a set of five photos from a
representative Class I area. The photographs were developed by John Molenar of Air
Resource Specialists and the specific scenes (Canyonlands in the Four Corners Region
and the Great Smokey Mountains in the Southeast Region) were chosen from available
options that (1) presented a view that a visitor would actually experience from a given
vantage point, (2) provided features at varying depths within the photo, and (3) could be
reproduced with sufficient resolution for accurate and consistent presentation in a 4” by
6” format for the picture sets. The photographs were digitally manipulated and set at the
deciview mean of each quintile of days under current (baseline) conditions, from the 20
percent best days in Photo A to the 20 percent worst days in Photo E. Current conditions
were based on monitoring data from the Interagency Monitoring of Protected Visual
Environments (IMPROVE) network for the period 2000 to 2004.
For improved visibility conditions, respondents were provided bar charts that depicted
distributions with days reallocated from the lower visibility photos to higher visibility
photos. These scenarios differed between the Four Corners and Southeast surveys and
were derived from U.S. Environmental Protection Agency estimates of “natural
conditions” for these same Class I areas.
Choice Questions and Experimental Design

The valuation questions were presented as a series of binary choices comparing
current conditions to a potential visibility improvement program with varying levels
of five attributes, as described in Exhibit 2.
EXHIBIT 2

C H O I C E Q U E S T I O N AT T R I B U T E S A N D L E V E L S

ATTRIBUTE

Visibility
Improvement

Ecosystem
Impacts

DESCRIPTION

Bar chart depicting number of days in the
year associated with each of five
photographs in picture set
Particles that form haze can affect water
quality, soil, plants, and in turn, the

6 or 7 Programs Ranging
Between 5% and 100%
Progress Toward Natural
Conditions
 No Change
 A Small Reduction

growth and variety of plants and animals
Some park visitors who have respiratory

Health Impacts

LEVELS

problems may experience coughing or
shortness of breath on days with high

 No Change
 A Small Reduction

levels of human-caused haze

Timing

Cost

Number of years until specified program
improvements are realized

 10 Years
 20 Years

Recurring annual cost to household

$15, $35, $65 or $115

6

The initial ecosystem and health impact attributes were defined based on discussions with
NPS scientists, with subsequent refinement based on focus group participant feedback.
These attributes were included in the design so that these potential benefits could be
explicitly excluded from estimation of values for visibility improvements, thereby
avoiding potential double-counting in policy analyses. The timing attribute was included
to investigate preferences for the speed at which specified program improvements would
take place. The levels of the cost attribute were assigned based on data from draft choice
questions administered to focus group participants.
The visibility attribute levels were developed to support two approaches to estimating
values for visibility improvements. The first approach was to define full visibility
programs with the percentages of days associated with each of the five visibility photos,
A, B, C, D, and E. Each unique set of percentages would be represented by a different
program variable in the econometric analysis of responses to the choice questions, which
would allow for calculation of WTP for the program. A key advantage of this approach
is that it facilitates the estimation of values for specific programs, at specific points in
time, along the projected visibility improvement paths defined by the Rule.
The second approach was to define the programs as additive functions of the number of
days associated with each of the photos. The advantage of this approach is that it allows
for estimation of values for changes in the number of days represented by specific photos,
and in turn a great deal of flexibility in valuing alternative visibility distributions.
To support estimation of models using both approaches, an experimental design of the
attributes and their levels (Exhibit 2) was developed. To derive choice sets, a 24-row,
orthogonal, main-effects design matrix was drawn from a well-regarded, on-line catalog
of orthogonal matrices by Warren Kuhfeld.2 The size of this design matrix allows for
orthogonal placement of the three two-level attributes (health impacts, ecological
impacts, and time), one four-level attribute (program cost), and one six-level attribute (the
visibility programs).
To ensure adequate variation across and within choice sets for the defined programs and
photo percentages, three programs taken directly from the Rule, representing 5, 50 and
100 percent progress toward natural conditions, were specified for each region. Second,
four additional programs (three in Four Corners) were created by "perturbing" the 50percent program in the following four ways: increase (decrease) the percentage
occurrence of Photo A one-third up (down) the improvement path, and orthogonally
increase (decrease) the percentage occurrence of Photo E one-third up (down) the
improvement path. In all cases, the amount of increases and/or decreases are added and/
or subtracted from Photo C. This process resulted in a total of six to seven visibility
programs.3,4 The design contained 24 choice sets, which were divided among four
2

3

http://support.sas.com/techsup/technote/ts723_Designs.txt
Because two programs turned out to be very close for the Southeast region, only six programs are used in the final
experimental design for that region.

4

Since the design matrix only accommodates a six-level attribute, variation over the seven programs is manufactured by
mixing information from two additional two-level columns from the design matrix into the perturbation routine. 

7

different survey versions with six questions per survey and randomly assigned to
respondents.
To verify that the experimental design would identify all parameters, simulations were
run on the Four Corners experimental design with 1,000 replications. Each replication
assumed a sample size of 400 (100 responses to each of the four survey versions) and
utilized representative utility parameters from focus group data. Results indicated that all
parameters could be estimated precisely.
MAIN & FOLLOW-UP
S U RV E Y R E S P O N S E S

Overall response rates for the Four Corners and Southeast surveys were 38.6 and 32.5
percent, respectively.5 Exhibits 3 and 4 present the geographic distribution of
respondents to the main survey by zip code for each region. Complete response
frequencies and summary statistics are embedded in the questionnaires in Appendices B
and C. A summary of responses to the open-ended question (#25) is also included at the
end of each survey.
The response rates for the phone follow-up survey for the Four Corners and Southeast
regions were 17 and 13.3 percent, respectively (Appendix D), while the response rates for
the mail follow-up survey were 7.5 and 3.2 percent (Appendix E).
The phone and mail follow-up surveys were conducted to investigate if there was a
systematic difference between those who did or did not respond to the main surveys. As
noted, questions were also included in the main survey that could be compared to results
from existing public opinion surveys. The following sections provide summaries of
responses to selected questions in the main surveys for each region and responses to
comparable questions from the General Social Survey and American Community
Survey.6
G E N E R A L S O C I A L S U RV E Y B E N C H M A R K I N G Q U E S T I O N S

The General Social Survey (GSS) is a highly regarded in-person survey conducted
annually or biennially since 1972 by the National Opinion Research Center at the
University of Chicago. Recent GSS survey questionnaires were reviewed by the study
team and three questions were selected from the 2010 questionnaire for replication in the
main surveys (as Questions 1, 2, and 28) for comparison purposes. The results of these
comparisons are presented in Exhibits 5-7 below. The GSS Southeast region is identical
to the Southeast pilot survey region. In addition to UT, AZ, CO and NM, the GSS
Mountain region also includes MT, ID, WY and NV.

5

Calculated as Response Rate = Complete / (Complete + No Response + Refused + Incomplete)

6

http://www3.norc.org/gss+website/; http://www.census.gov/acs/www/

8

EXHIBIT 3

GEOGRAPHIC DISTRIBUTION OF RESPONSES- FOUR CORNERS REGION

9

EXHIBIT 4

GEOGRAPHIC DISTRIBIBUTION OF RESPONSES- SOUTHEAST REGION

10

EXHIBIT 5

GOVERNMENT SPENDING
“ We a r e f a c e d w i t h m a n y p r o b l e m s i n t h i s c o u n t r y, n o n e o f w h i c h c a n b e s o l v e d
e a s i l y o r i n e x p e n s i v e l y. L i s t e d b e l o w a r e s o m e o f t h e s e p r o b l e m s . F o r e a c h o n e
c i r c l e w h e t h e r y o u t h i n k w e ’ r e s p e n d i n g t o o m u c h m o n e y o n i t , t o o l i t t l e m o n e y,
or about the right amount.”

ABOUT THE

N

TOO LITTLE

641

51.3%

34.2%

14.5%

73

58.9

29.3

11.8

548

58.6

33.4

8.0

287

61.9

26.7

11.4

639

24.1%

42.4%

33.5%

71

20.1

56.0

23.9

547

21.6

40.8

37.7

275

17.2

40.9

41.9

639

69.6%

21.4%

8.9%

74

79.4

14.9

5.7

543

70.6

22.1

7.3

286

80.2

16.8

3.1

635

44.6%

33.4%

22.0%

RIGHT AMOUNT

TOO MUCH

THE ENVIRONMENT

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast
SPACE EXPLORATION

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast
EDUCATION

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast
HEALTH

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast

74

61.3

12.1

26.6

543

56.2

28.9

14.9

286

62.1

15.0

22.9

2.5%

17.0%

80.5%

ASSISTANCE TO OTHER COUNTRIES

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast

636
73

8.3

11.1

80.6

550

2.0

16.2

81.8

284

4.3

29.8

65.9

11

EXHIBIT 6

CONFIDENCE IN INSTITUTIONS
“ L i s t e d b e l o w a r e s o m e i n s t i t u t i o n s i n t h i s c o u n t r y. A s f a r a s t h e p e o p l e r u n n i n g
these institutions are concerned, would you say you have a great deal of
confidence, only some confidence, or hardly any confidence at all in them?”

N

A GREAT
DEAL

ONLY SOME

HARDLY ANY

BANKS & FINANCIAL INSTITUTIONS

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast

643

7.2%

47.4%

45.4%

98

3.9

41.2

54.9

553

7.8

51.5

40.7

381

12.6

44.5

42.9

643

1.9%

33.4%

64.7%

95

4.2

48.9

46.9

553

3.8

29.5

66.7

377

9.0

48.4

42.6

643

41.8%

49.9%

8.2%

92

52.7

44.1

3.2

544

32.4

55.3

12.3

365

42.1

52.3

5.6

CONGRESS

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast
SCIENTIFIC COMMUNITY

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast

EXECUTIVE BRANCH OF THE FEDERAL GOVERNMENT

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast

644

9.9%

39.6%

50.5%

96

8.4

55.1

36.6

551

9.8

39.6

50.6

377

17.0

44.7

38.2

642

9.3%

56.1%

34.6%

95

10.1

59.6

30.3

553

8.1

57.5

34.4

373

14.9

60.0

25.1

641

17.5%

42.7%

39.8%

94

18.1

54.1

27.8

552

20.8

50.0

29.2

366

21.6

53.8

24.6

MAJOR COMPANIES

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast
ORGANIZED RELIGION

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast

12

EXHIBIT 7

I N C O M E TA X E S
“Do you think the amount of federal income tax you have to pay is too high,
about right, or too low?”

Pilot Four Corners
GSS Mountain
Pilot Southeast
GSS Southeast

N

TOO HIGH

ABOUT RIGHT

TOO LOW

662

46.5%

48.0%

5.4%

90

32.0

63.3

4.6

564

51.6

45.2

3.2

353

56.8

42.4

0.8

As shown, the pilot results are generally similar to those from the 2010 GSS. It is
important to note that minor differences would be expected, given the difference in
survey years (2012 versus 2010), target populations (the GSS Mountain region includes
more states than the Four Corners region), and survey modes (mail versus in-person).
D E M O G R A P H I C C O M PA R I S O N S

Demographic questions in the main survey and in the mail and phone follow-up surveys
were designed to facilitate direct comparison to American Community Survey (ACS)
data. The ACS is an ongoing statistical survey implemented by the U.S. Census Bureau.
It is viewed as the definitive source for data on the characteristics of U.S. households.
The ACS summary statistics in Exhibits 8-11 below are from the 2011 survey. Note that
the ACS data are based on samples of adults, while the pilot and follow-up survey data
are based on samples of heads of household.

EXHIBIT 8

AGE

FOUR CORNERS

SOUTHEAST

MAIN

NR
MAIL

NR
PHONE

ACS

MAIN

NR
MAIL

NR
PHONE

ACS

18-29

6.8%

2.4%

3.9%

23.4%

6.6%

17.7%

8.3%

21.3%

30-39

14.9

14.3

9.6

18.0

12.8

11.8

10.0

16.5

40-49

16.4

19.1

13.5

17.4

19.6

23.5

18.3

18.4

50-59

21.4

31.0

15.4

17.2

24.7

17.7

26.7

17.8

60-69

22.4

11.9

21.2

12.7

18.3

29.4

11.7

13.6

70-79

13.7

11.9

15.4

7.0

10.4

0.0

11.7

7.7

80+

4.4

9.5

21.2

4.2

7.5

0.0

13.3

4.8

630

42

52

546

17

60

N

13

EXHIBIT 9

ETHNICITY AND RACE

FOUR CORNERS

SOUTHEAST

MAIN

NR
MAIL

NR
PHONE

ACS

MAIN

NR
MAIL

NR
PHONE

ACS

8.5%

19.5%

11.8%

23.1%

5.7%

11.8%

5.1%

9.5%

N

622

41

51

542

17

59

American Indian or Alaskan
Native

1.3%

0.0%

2.1%

3.6%

0.9%

0.0%

1.8%

0.4%

Asian

1.8

5.1

0.0

2.6

1.1

11.8

3.5

2.8

Black or African American

2.1

2.6

2.1

3.5

10.7

11.8

10.5

22.0

Native Hawaiian or Other
Pacific Islander

0.7

2.6

0.0

0.3

0.0

0.0

0.0

0.0

White

92.2

89.7

93.8

86.7

84.7

76.5

82.5

72.7

Two or More Races

2.0

0.0

2.1

3.2

2.6

0.0

1.8

2.2

612

39

48

542

17

57

Hispanic, Latino, Spanish
Origin

N

Note: ACS covers all individuals while pilot survey covers individuals 18 and over

EXHIBIT 10

E D U C AT I O N A L AT TA I N M E N T

FOUR CORNERS

MAIN

NR
MAIL

NR
PHONE

No schooling

0.2%

0.0%

Some schooling < grade 12

1.9

High school graduate

15.5

Some college
Associate’s degree

SOUTHEAST

ACS

MAIN

NR
MAIL

NR
PHONE

ACS

1.9%

1.0%

0.0%

0.0%

0.0%

1.2%

16.7

5.7

11.4

4.7

11.1

1.7

13.6

16.7

18.9

24.0

20.8

11.1

33.9

29.8

23.7

21.4

34.0

25.0

24.8

38.9

17.0

20.9

7.2

14.3

7.6

8.3

8.0

11.1

3.4

7.7

Bachelor’s degree

26.2

26.2

20.8

19.2

21.6

27.8

28.8

16.7

Master’s degree

16.0

4.8

7.6

7.9

12.7

0.0

11.9

7.1

Professional degree
beyond bachelor’s

4.7

0.0

1.9

1.7

4.0

0.0

0.0

1.8

Doctoral degree

4.7

0.0

1.9

1.3

3.4

0.0

3.4

1.2

638

42

53

552

18

59

N

Note: ACS data for education are for individuals 25 and over

14

EXHIBIT 11

HOUSEHOLD INCOME

FOUR CORNERS

SOUTHEAST

MAIN

NR
MAIL

NR
PHONE

ACS

MAIN

NR
MAIL

NR
PHONE

ACS

$10,000 or less

3.8%

5.0%

11.1%

7.8%

6.8%

5.6%

4.0%

8.7%

$10,001 to $20,000

7.8

15.0

6.7

10.7

10.5

5.6

12.0

12.4

$20,001 to $30,000

7.3

17.5

6.7

11.1

10.3

11.1

26.0

11.9

$30,001 to $40,000

9.5

7.5

8.9

10.5

13.7

16.7

0.0

10.7

$40,001 to $50,000

11.7

10.0

13.3

9.5

8.0

22.2

10.0

9.2

$50,001 to $60,000

9.6

15.0

6.7

8.4

8.0

16.7

8.0

8.1

$60,001 to $75,000

12.9

7.5

6.7

10.5

9.3

0.0

14.0

9.7

$75,001 to $100,000

14.3

2.5

20.0

11.9

10.5

22.2

18.0

11.0

$100,001 to $125,000

8.5

7.5

8.9

7.6

8.8

0.0

6.0

6.9

$125,001 to $150,000

5.2

7.5

2.2

4.3

4.4

0.0

0.0

3.9

$150,001 or more

9.5

5.0

8.9

7.8

9.9

0.0

2.0

7.5

614

40

45

526

18

50

N

The percent of male respondents in the Four Corners and Southeast regions was 62.7 and
51.4 percent, respectively, compared to 49.5 and 48.1 percent of adults in the ACS. As
shown above, there are relatively modest differences in age (older), ethnicity (white),
education (higher) and income (higher) between pilot respondents and the general
population in both regions. In some cases there are larger differences between main
survey and phone and mail follow-up respondents; however, these statistics are based on
a relatively small number of responses. We investigate the impact of differences between
main survey respondent characteristics and ACS statistics on WTP estimates in the next
section.

15

A N A LY S I S O F
CHOICE QUESTION
RESPONSES & WTP
E S T I M AT I O N

Summaries of all choice question responses are provided in Appendix F. Formal analysis
of responses is based on the random utility framework (e.g., see Haab and McConnell,
2002). Under this approach, individual i's utility associated with a particular visibility
program j, which is defined by a set of K attributes, can be expressed as:

(1)

∑

ε

where yi is individual i's money income, Cj is the cost of visibility program j, and Xjk is
the level of attribute k that is offered in visibility program j.
The βk's are the marginal utilities for each of the K visibility attributes and βy is the
marginal utility of income. Under the random utility specification, and given individuals'
stated responses to binary choice questions comparing program j to no program, these
parameters are estimated using a conditional logit model. Once estimated, the marginal
value of any particular attribute k can be computed as:

(2)

As discussed earlier, the experimental design was tailored to allow estimation of two
principal types of models: one in which visibility programs were identified individually
by separate binary variables, and one in which the number of days associated with
various photographs were included as continuous variables. For comparison with
previous research, a third model was estimated using mean annual visibility. Mean
annual visibility is calculated as a weighted average, where the weights are equal to the
percentage of days associated with each photo. The specific equations estimated for each
region were as follows:

(3)

α

ε

(4)

∑

ε

(5)

α

ε

_

16

where Adays and Edays are the number of days in photos A and E, respectively, in
program j; PROG are binary variables identifying the seven programs in the Four Corners
region and six programs in Southeast region; MEAN is the weighted deciview average for
program j; HEALTH is the binary health attribute; ECO is the binary ecological attribute;
and, COST is the annual household cost of program j.
For reference, Exhibits 12 and 13 below provide examples of the numbered improvement
programs in each region.

EXHIBIT 12

FOUR CORNERS VISIBILITY IMPROVEMENT PROGRAMS

17

EXHIBIT 13

SOUTHEAST VISIBILITY IMPROVEMENT PROGRAMS

Logit models were estimated using STATA v.12. The “cluster” option that estimates
standard errors accounting for correlation among choices made by the same respondent
was utilized. Estimation results are reported in Exhibit 14.
Across all Four Corners models, program cost is negative and significant at the onepercent level. The health and ecological attributes are positive and also highly significant
in all models. Consistent with expectations the number of photo A days is positive and
significant and the number of photo E days is negative and significant. A variety of
models containing different combinations of photo days were also estimated. Across
these specifications the variables representing the number of days associated with photos
A and E were generally significant, while other photos and combinations of photos in the
distribution generally were not. Lack of sensitivity to changes in the interior of the

18

distribution may be an artifact of the experimental design, which will be re-evaluated
prior to full implementation. Five of the seven program binary variables in model (2) are
positive and significant; program 7 is negative and significant. A likelihood ratio test
indicates that equivalence of coefficients for programs 2, 3, 5 and 6 cannot be rejected (χ2
= .02, df = 3). Finally, the coefficient on mean visibility in model (3) is negative as
expected, since higher deciviews mean lower visibility, and significant at the one-percent
level.
EXHIBIT 14

C H O I C E M O D E L E S T I M AT I O N R E S U LT S

FOUR CORNERS
(1)

Photo A
Photo E

(2)

Program 3
Program 4
Program 5
Program 6
Program 7

Cost
Constant

N

(5)

(6)

.015***
(.002)
-.020*
(.010)
1.239***
(.141)
.536***
(.134)
.792***
(.143)
.586***
(.119)
.622***
(.131)
-.313***
(.119)

.244***
(.057)
.171***
(.083)
.023
(.057)
-.014***
(.001)
.055
(.186)

.155***
(.063)
.220***
(.081)
-.067
(.065)
-.013***
(.001)

-.407***
(.055)
.206***
(.055)
.172**
(.083)
.039
(.055)
-.014***
(.001)
2.889***
(.410)

3902

3902

3902

Mean

Time

(4)

.638***
(.153)
.317***
(.115)
.306**
(.145)
-.056
(.138)
.323**
(.135)
.319**
(.155)
-.489***
(.110)

Program 2

Ecological

(3)

.0082***
(.003)
-.038***
(.008)

Program 1

Health

SOUTHEAST

.218***
(.067)
.020
(.089)
-.150**
(.067)
-.013***
(.001)
-.205
(.183)

.214***
(.068)
.011
(.090)
-.154**
(.067)
-.013***
(.001)

-.211***
(.020)
.216***
(.068)
.020
(.089)
-.152**
(.066)
-.013***
(.001)
4.063***
(.371)

3351

3351

3351

Standard errors in parentheses
*** Significant at 1%, ** 5%, * 10%

19

Program cost is negative and significant in all Southeast models. The health attribute is
positive and significant; however the ecological attribute is not significant in any of the
three models. Program timing is significant at the five-percent level in each model and
negative. The number of photo A days is positive and significant and the number of photo
E days is negative and significant (though at the ten-percent level). All program variables
are significant at the one-percent level in model (5), as is mean visibility in model (6).
E X A M P L E W T P E S T I M AT E S

For illustrative purposes, we calculate annual, per-household WTP estimates for the
seven example improvement programs using the above results. The values are calculated
as follows (a superscript of “0” represents baseline conditions and a superscript of “1”
represents improved conditions):
Photos A and E Models:

(6)

1

∗

Program Models:
1

(7)

∗

Mean Visibility Models:

(8)

1

∗

_

_

_

20

Exhibit 15 presents mean WTP estimates and 95-percent confidence intervals by program
and model.

EXHIBIT 15

ANNUAL HOUSEHOLD WTP BY PROGRAM AND MODEL

PROGRAM

1

2

3

4

5

6

7

$88
(69,109)
49
(28, 67)
79
(61, 100)

$66
(50, 83)
24
(8, 39)
50
(38, 62)

$51
(40, 64)
23
(2, 43)
44
(33, 55)

$56
(40, 73)
(4)
(-28, 16)
37
(28, 47)

$54
(36, 74)
25
(5, 43)
32
(24, 40)

$40
(28, 54)
24
(2, 47)
26
(20, 33)

$6
(1, 11)
(37)
(-58, -20)
(1)
(-1, -1)

$118
(93, 149)
92
(72, 113)
118
(94, 147)

$78
(55, 105)
40
(21, 57)
86
(69, 108)

$71
(53, 92)
59
(38, 80)
80
(64, 100)

$62
(36, 89)
44
(27, 60)
68
(55, 86)

$54
(36, 74)
46
(29, 64)
62
(50, 78)

$8
(-2, 17)
(23)
(-43, -5)
7
(6, 9)

Four Corners
Model 1 (A & E)
Model 2 (Program)
Model 3 (Mean)
Southeast
Model 4 (A & E)
Model 5 (Program)
Model 6 (Mean)

Confidence intervals estimated using the Krinsky-Robb method, 5,000 iterations

WEIGHTED MODELS

As noted above, there were modest differences between the demographic characteristics
of survey respondents and the demographic characteristics of the adult populations in
each region. As a result, weights were developed such that for each region, the weighted
sample matched the population with respect to age, gender, Hispanic/Latino ethnicity,
race, education and income. Models using weighted data were estimated for each
characteristic independently and resultant WTP estimates were compared to the
unweighted results. Of these, Hispanic/Latino ethnicity, education, and age resulted in
the largest average changes in WTP estimates (in the range of $2 to $6).
Models using weighted data combining these characteristics were then estimated.
Specifically, the weights were defined as: (1) Hispanic/Latino ethnicity, (2) percentage of
respondents with bachelor’s degree or higher, and (3) percentage of respondents age 40 or
older. For each region, eight mutually exclusive and exhaustive categories were
developed based on the binary ethnicity/education/age classifications (8 = 2 x 2 x 2).
Next, an iterative procedure (i.e., raking) was used to identify a single weight for each of
the eight categories such that the weighted percentage of respondents in each category
equaled the population percentage. The final weights are shown in Exhibit 16.

21

EXHIBIT 16

P O S T - R A K I N G S U RV E Y W E I G H T S

ETHNICITY

EDUCATION

< Bachelor’s
Hispanic/Latino
Bachelor’s +
Four Corners
Not
Hispanic/Latino

< Bachelor’s
Bachelor’s +
< Bachelor’s

Hispanic/Latino
Bachelor’s +
Southeast
Not
Hispanic/Latino

< Bachelor’s
Bachelor’s +

AGE

WEIGHT

<40

5.51

40+

2.30

<40

2.50

40+

1.04

<40

2.19

40+

0.92

<40

0.99

40+

0.41

<40

3.62

40+

1.35

<40

1.64

40+

0.61

<40

2.60

40+

0.97

<40

1.18

40+

0.44

Exhibit 17 presents estimation results for the weighted models. Patterns of sign,
significance and magnitude of coefficients are similar to the unweighted models.
However, several program variables and the ecological attribute in models (1) and (3) are
now insignificant in the Four Corners models.

22

EXHIBIT 17

W E I G H T E D C H O I C E M O D E L E S T I M AT I O N R E S U LT S

FOUR CORNERS
(1)

Photo A
Photo E

(2)

Program 3
Program 4
Program 5
Program 6
Program 7

Cost
Constant

N

(5)

(6)

.013***
(.003)
-.032**
(.013)
1.361***
(.172)
.599***
(.152)
.804***
(.176)
.780***
(.147)
.700***
(.168)
-.274*
(.150)

.296***
(.086)
.169
(.106)
.064
(.086)
-.014***
(.002)
-.010
(.273)

.214**
(.089)
.218**
(.103)
-.053
(.092)
-.013***
(.002)

-.382***
(.075)
.262***
(.084)
.170
(.105)
.073
(.083)
-.014***
(.002)
2.674***
(.555)

3561

3561

3561

Mean

Time

(4)

.672***
(.205)
.233
(.176)
.212
(.189)
-.121
(.182)
.323*
(.189)
.338*
(.202)
-.422***
(.156)

Program 2

Ecological

(3)

.008*
(.005)
-.034***
(.012)

Program 1

Health

SOUTHEAST

.305***
(.086)
.033
(.115)
-.195**
(.087)
-.015***
(.001)
.012
(.230)

.297***
(.087)
.027
(.116)
-.195**
(.086)
-.015***
(.001)

-.219***
(.024)
.305***
(.086)
.030
(.114)
-.195
(.086)
-.015***
(.001)
4.293***
(.439)

3056

3056

3056

Standard errors in parentheses
*** Significant at 1%, ** 5%, * 10%

23

Exhibit 18 presents a comparison of WTP estimates for the example programs from the
weighted and unweighted models for the A and E photo models.

EXHIBIT 18

W E I G H T E D V S . U N W E I G H T E D M O D E L W T P E S T I M AT E S - A & E P H O TO M O D E L S

PROGRAM

1

2

3

4

5

6

7

$81
(55, 111)
88
(69,109)

$60
(39, 83)
66
(50, 83)

$47
(32, 65)
51
(40, 64)

$50
(30, 73)
56
(40, 73)

$49
(24, 75)
54
(36, 74)

$36
(20, 54)
40
(28, 54)

$5
(-2, 12)
6
(1, 11)

(8%)

(9%)

(8%)

(11%)

(9%)

(10%)

(17%)

$114
(85, 149)
118
(93, 149)

$82
(55, 114)
78
(55, 105)

$71
(51, 95)
71
(53, 92)

$69
(41, 101)
62
(36, 89)

$58
(38, 82)
54
(36, 74)

$12
(1, 24)
8
(-2, 17)

(3%)

5%

-

11%

7%

50%

Four Corners
Weighted
Unweighted
Difference
Southeast
Weighted
Unweighted
Difference

Confidence intervals estimated using the Krinsky-Robb method, 5,000 iterations

CONCLUSIONS &
I M P L I C AT I O N S F O R
F U L L S U RV E Y

This pilot study was designed to test a survey of the public’s WTP for reductions in
human-caused haze and resultant visibility improvements in designated Class I national
parks and wilderness areas. The survey was fielded by mail in two regions and telephone
and mail follow-ups were conducted with nonrespondents.
A number of important insights arise from the empirical analysis of the pilot survey data:


The survey response rates for the Four Corners and Southeast administrations of
the surveys (39 and 32 percent, respectively) are similar to those observed for
surveys conducted for other environmental applications. However, these
response rates are not sufficient to exclude the potential for survey nonresponse
to affect WTP estimates.



Comparisons with data from the mail and telephone follow-up surveys, and
comparisons with national probability survey results, indicate that characteristics
of people who responded to the pilot survey are not fully representative of the
population that the pilot samples were drawn from. These results suggest that
data analyses should consider weighted models that bring the sample into
consistency with known population parameters.

24



The estimated valuation question response equations differ between the Southeast
and Four Corners regions. This indicates that it is important to implement final
surveys in different regions of the country with baseline visibility and visibility
improvements calibrated to each survey region.



Health and ecological considerations are significant in explaining WTP in the
Four Corners regions while only health considerations are significant in the
Southeast region. These results imply that it is important to control for these
effects to avoid double counting these effects in computing aggregate benefits of
visibility improvements.



The statistical results suggest that people are most concerned with reducing the
number of lowest visibility days and increasing the number of highest visibility
days. Prior to full implementation, the experimental design may be modified to
increase variation in changes in the interior of the distribution across programs.



Weighting data to account for sample nonresponse decreased estimated WTP in
the Four Corners region and generally increased estimates in the Southeast.
These results indicate that it will be important to provide the opportunity to
weight survey data to representative population characteristics for each
implementation region in the administration of the final survey.

Overall the pilot survey performed very well and the qualitative findings are largely
consistent with similar environmental studies in the peer-reviewed literature. With minor
editing to customize the survey to each of the remaining five survey regions and possible
revision of the experimental design, the survey is ready for full implementation.

25

REFERENCES

Balson, W.E., R.T. Carson, M.B. Conaway, B. Fischoff, W.M. Hanemann, A. Hulse, R.J.
Kopp, K.M. Martin, R.C. Mitchell, J. Molenar, S. Presser, and P.A. Ruud. 1990.
“Development and Design of a Contingent Value Survey for Measuring the Public’s
Value for Visibility Improvements at the Grand Canyon National Park.” Revised
Draft Report. Prepared by Decision Focus Incorporated for the Salt River Project.
Chestnut, L.G. and R. D. Rowe. 1990. Preservation Values for Visibility Protection at
the National Parks. Prepared for U.S. Environmental Protection Agency, Office of
Air Quality Planning and Standards, and National Park Service, Air Quality
Management Division.
Haab, T.C. and K.E. McConnell. 2002. Valuing Environmental and Natural Resources:
The Econometrics of Non-Market Valuation, Edward Elgar.

26