Chapter 5

Survey Results

This chapter presents estimates of the benefits from reducing the rate of monument deterioration, where the reduction in the rate of deterioration is a result of the SO₂ emissions reductions mandated under Title IV of 1990 Clean Air Act Amendments. In particular, we focus on the approximately 100 outdoor marble monuments in Washington, DC. The benefits are for reduced injury rates to only those 100 marble monuments, and therefore underestimate the total benefits of reduced injury to marble monuments resulting from reduced SO₂ emissions nationwide.

This chapter summarizes the responses to the survey questions and the WTP modeling results. In Section 5.2, the attitudinal responses are summarized for all respondents and systematic differences among respondents are discussed. The subsequent sections present several estimates of WTP for preservation; the differences are a result of the different modeling assumptions and the use of different subsets of the data. Section 5.3 introduces the two types of valuation questions for preserving the monuments: pair-wise choices and a payment card. The pair-wise choice format is the primary elicitation approach used in the survey for valuation. The payment card and other follow-up questions are used to help assess the pair-wise results. Different statistical models were used to analyze the data from the two elicitation formats. Section 5.4 describes the pair-wise choice questions and answers, the statistical model for estimating the value of the preservation options, and the estimated WTP results. Section 5.5 covers these same topics for the payment card. Section 5.6 compares the methodologies and results from the two models.

 

5.2	Attitudinal Results

5.2.1

The Importance of the Monuments

Two findings are of particular significance: on average, people feel that the existence of the monuments is important, and this feeling appears to be unrelated to whether they have visited or plan to visit Washington, DC. Respondents often expressed strong and positive sentiments, if not reverence, for the monuments. However, reducing natural and pollution related injury to the monuments is less important to respondents than other social problems such as reducing crime, improving public schools, and maintaining economic growth.

Survey questions were designed to determine why people care about the monuments and their preservation. For example, Question 3 asked about the general importance of the monuments to the respondents. Eighty percent of the respondents gave a rating of 4 or 5. The mean rating was 4.2.

Another indication that the value of the monuments is unrelated to the likelihood of visiting the monuments is that the answers of those who have visited the monuments do not differ statistically from those who have not. The one exception is for Question 7e, which is discussed in the next section.

 

5.2.2

The Future Condition of the Monuments

Not only do people appreciate the existence of the monuments, but they are also concerned about the condition of the monuments in the future. This finding is expected given that the most important reason stated for the existence of these monuments is so they are available for future generations. After being presented with information about the expected future effects of natural and pollution-related weathering, respondents were asked in Question 4 to indicate the number of years from now when the existence of the monuments still matters to them (Figure 5-1). The percent of respondents who care whether the monuments exist in the future gradually declines with the number of years; 44% of respondents say they still care whether monuments exist more than 1,000 years from now.

Following a presentation of the effects of natural and pollution- related weathering, respondents were asked about their reactions to such changes in the condition of the monuments:

Generally, the respondents indicated that the effects of natural and pollution-related weathering diminish the appearance of the monuments. Almost 80% of respondents disagreed (rating of 1 or 2) with the statement that the changes improve the monuments by making them look old (Question 7a), and almost as many respondents agreed with an opposing statement (Question 7b). Although some respondents feel that natural weathering is all right (about 30% rated question 7d as 4 or 5), most respondents indicated that deterioration caused by pollution and natural forces is undesirable.

Consistent with answers to previous questions, few respondents indicated that they do not care about the condition of the monuments 75 and 150 years from now: only 16% of respondents gave a rating of 4 or 5 to question 7e. However, this is the only question where people who have visited the monuments provided responses that were statistically different from non-visitors. The answers of those who have visited the monuments indicated that they are more concerned than non-visitors with the condition of the monuments in the relatively near future.

The answers to Question 8 suggest that respondents feel equally strongly about the four suggested reasons why people might care about changes in the appearance of the monuments. In each case, fewer than 8% of respondents indicated that a reason was not important by giving it a rating of 1 or 2.

 

5.2.3

The Relative Importance of the Condition of the Monuments

In Question 11, respondents were asked to rate the importance of 10 issues ranging from monument preservation to using energy efficient appliances to maintaining economic growth (see Appendix C). The primary purpose of including this question was to remind respondents that although they may feel that preserving monuments is important, there are other issues that may be of greater importance to them. Of the 10 issues listed, preserving monuments received the lowest average rating, a mean rating of 3.4 on a scale of 1 to 5 (where 1 is "not at all important" and 5 is "very important"). Rated slightly higher, a mean rating of 3.5, was using energy-efficient compact fluorescent light bulbs. Driving a car that gets high gas mileage and reducing taxes were next, each with a mean rating of 3.9. So although people think the existence and condition of the monuments are important, they are less important than many other social issues.

 

5.2.4

Differences among Respondents

Throughout the survey, respondents, on average, consistently indicated that the monuments are important and that pollution- related weathering negatively affects the monuments. However, respondents' attitudes differ significantly by demographic groupings. To compare the attitudes of various subgroups of respondents, the mean responses of various subgroups are compared to those of all other respondents. Only questions where the differences in the mean responses are statistically significant are discussed.¹ This analysis considers one demographic characteristic at a time and does not, therefore, account for collinearity among demographic characteristics such as with education and income. The findings should therefore be cautiously interpreted.

The most dramatic difference in attitudes is between Caucasians and non-Caucasians. Of the 272 respondents, 268 provided information about their ethnicity: 207 or 77% are Caucasian and 61 or 23% are of another ethnicity, categorized as non-Caucasian. Of the non-Caucasians, the majority (74%) are African-American.

The existence of these monuments appears to be more important to Caucasians than to non-Caucasians. Caucasians are also more bothered by the effects of pollution-related weathering. For example, the mean responses of the two groups are statistically different for each of the questions in the box below. Although no specific questions were asked that would explain this difference, a few respondents provided reasons in open-ended responses such as: "these monuments do not depict enough black history." This result is not surprising given the historic events and people memorialized by this set of monuments, most of which were erected between 1850 and 1940.

Significant differences in attitudes are also evident between respondents with lower levels of education and all other respondents. The respondents who have no more than a high school education (11% of respondents) consistently indicated that they believe the monuments are more important than did the respondents with more education. The mean responses for this group versus all other respondents statistically differ for most of the attitudinal questions.³

The respondents with some college (27.8%) or trade school (4.1%) gave answers to questions 2a, 2b, and 11g that indicate they believe the monuments are less important than all other respondents. Thus, respondents with the least education express the greatest enthusiasm for monuments; some additional education reduces this enthusiasm, yet the enthusiasm is partially restored with further education.

The female respondents, through their answers to questions 2, 3, and 11, indicated that the monuments are more important to them than the male respondents indicated.

The mean ratings of the respondents in the lowest two income groups (annual household income less than $12,000, and $12,000- $25,000) differed significantly from those for respondents with higher incomes for Questions 2b, 2f, 7a, and 7d. In each case, the lower income respondents express greater concern for the existence or condition of the monuments. There are no significant differences among income levels above $25,000.

 

5.3	Introduction to WTP Results

A primary purpose of this research is to estimate people's WTP for a reduction in the rate at which marble monuments are injured by SO₂ emissions. Recall from Section 2.5 that we did not attempt to directly value air pollution reductions, but instead developed a hypothetical preservation program to mimic the effects of reducing SO₂ emissions on marble monuments. To do this, three hypothetical preservation levels are considered: a 25% increase in the time in which the illustrated pollution-related injuries to marble monuments would occur (Option A), a 50% increase in the time in which injuries would occur (Option B), and a 100% increase, or doubling, in the time in which injuries would occur (Option C). These levels were chosen to span the potential, yet uncertain, effects of the SO₂ reductions mandated under Title IV.

The survey included two types of WTP questions: pair-wise choices and a payment card. The pair-wise choices are the primary valuation questions; the payment card results provide a basis for comparison. Because of this, the pair-wise questions address all three levels of preservation (Options A, B, and C), while the payment card addresses only Option C. These two elicitation methods require different economic models and estimation techniques for inferring WTP estimates. The following sections present an overview of the analyses and results for both valuation approaches. A complete and more technical description of the models and results is included in Appendix D.

 

5.4	Pair-Wise Choice Model and Results

5.4.1

Pair-Wise Choice Elicitation Format

The effects of the three preservation options were presented as permanent reductions in the rate at which natural and pollution- related weathering degrades the marble. A fourth level, no preservation, was also included in the pair-wise choices so that the benefits associated with changes from no preservation to each of the three reduced injury levels could be measured. Each respondent was presented with 10 pairs of preservation alternatives, where an alternative consists of a preservation option and a dollar amount. The dollar amount, or price, is a one-time payment for a permanent change in the rate of injury. The prices for the options were $0.25, $1, $3, $7, $10, $15, $25, and $50; and the price for no preservation, the status quo, was always $0. For each of 10 pairs, respondents were asked to select the alternative they preferred, understanding that the dollar amount would be the price their household would pay for that preservation option to be undertaken. We constructed 20 sets of 10 pairs to cover all feasible combinations of options and prices and each respondent was presented with one of the 20 sets.⁴

 

5.4.2

Modeling WTP from Pair-Wise Choice Data

We assume that individuals are able to rank preservation alternatives in terms of their price and level of preservation, and that individuals will choose the alternative they prefer from any given pair of alternatives. These choices, or rankings, may vary across individuals as a function of personal characteristics such as income, gender, age, and ethnicity. By observing how a large number of different individuals choose among different pairs of alternatives, it is possible to estimate how a representative individual would rank any given pair of alternatives. Based on this ranking, a representative individual's WTP for different alternatives can be inferred.

For example, if an individual chooses Option C for $10 over the status quo ($0), we know that the individual's WTP for Option C is at least $10. If that same individual also chooses the status quo over Option C at $20, then we know that her WTP for Option C is at least $10, but less than $20. If Option C at $15 is chosen over Option A at $10, then we know that the individual's WTP for increasing the level of preservation from A to C is at least $5. The combination of all such inferences for many individuals with different characteristics allows us to estimate the WTP for varying quantities of preservation, depending on those individual characteristics.

Each respondent was asked to make 10 choices by choosing the alternative they prefer from each of 10 pairs of alternatives. Thus, for 272 respondents, a total of 2,720 choices could be made. Our final data set of choices is smaller than 2,720 because a few respondents did not complete all 10 pairs or did not provide complete demographic information. The number of choice observations eliminated because of missing data is 152, leaving 2,568 choice observations from 259 respondents.

We used the 2,568 choice observations to estimate a statistical model of choice (see Appendix D for the details of the model). The model chosen and presented here is the one that best predicts the actual choices made. The model correctly predicts 75% of the actual choices made. The factors that influence choice are price, level of preservation, income, age, gender, and ethnicity. Because the model predicts a significant proportion of the choices made, it provides a sound basis from which to derive WTP estimates.

 

5.4.3

Pair-Wise Choice WTP Results

Estimated WTP for preservation is a function of the price and level of preservation, and the following individual characteristics: gender, annual household income, age, and ethnicity. Other socioeconomic characteristics were not found to be significant determinants of WTP. This section describes the extent to which a representative household's WTP for preservation varies with the level of preservation and the household's demographic characteristics.⁵

For the three preservation levels included in the survey, the WTP values for the study sample are reported in Table 5-1. Shown are the predicted minimum, maximum, mean, and median of household WTP for all individuals in the sample, male and female, derived from the estimated pair-wise choice model. Recall that this is a one-time payment for a permanent reduction in the rate at which the monuments are affected by pollution. For each of the three levels of preservation, the mean values are significantly different from zero.

Estimated WTP increases with the level of preservation, but at a decreasing rate. That is, as the level of preservation is increased, people are willing to continue to pay for additional increments of preservation, but not as much as they were for earlier increments. For example, Option C (a 100% increase in time) is four times as effective as Option A (a 25% increase in time), yet the mean WTP for Option C is just over twice that for Option A. The graph in Figure 5-2 shows how the mean WTP varies with the level of preservation, from a level of 0% (no reduction in pollution) to 100% (Option C), a decrease in pollution that corresponds with a 100% increase, or doubling, in the amount of time at which specific injuries occur.

Figure 5-3 shows the frequency distribution of the predicted household WTP amounts and the mean and median WTP amounts for Option C of the 259 respondents.⁷ That the median WTP amount is less than the mean WTP amount indicates that a majority of the respondents are willing to pay less than the mean amount.

---

The characteristics of the individuals with the minimum and maximum WTP amounts suggest the ways in which the individual characteristics influence WTP. The maximum WTP amount, $155.04 for Option C, is that of a 77-year-old, Caucasian woman with a household income between $12,000 and $25,000. The minimum WTP amount, $11.59 for Option C, is that of a 22-year-old, non-Caucasian man with a household income of less than $12,000. The effects of each of these individual characteristics - income, age, ethnicity, and gender - are described below.

Women are willing to pay more than are men. This finding is supported by the attitudinal data showing that female respondents believe that the monuments are more important than do male respondents and, thus, are willing to pay more for preserving them. As an example, the estimated WTP amounts in Table 5-2 are for a representative male and female, where both are 45 year old (the mean age of all respondents) Caucasians with income of $12,000 or more.

The differences are dramatic. The ratio of the WTP of males to females depends on the individual's income level. In general terms, the WTP of males with income of $12,000 or more is 53% that of females with similar income, regardless of age, ethnicity, and the level of preservation. The WTP of males with income less than $12,000 is 72% that of females with similar income.

That women have a higher WTP than men complicates the interpretation of these amounts as a household's WTP. Respondents were instructed to give WTP amounts for their household and, since many households have both a male and a female decision maker,⁸ the predicted WTP amount for that household depends on the extent to which the male and female are involved in the decision making process. We have no information on this to help in weighting the male and female WTP estimates for the average household. Therefore, we will report the most conservative estimates by calculating WTP as though all respondents are male, i.e., setting gender equal to male for all respondents. This is a less than ideal resolution of this issue, but it keeps the direction of the error unambiguous - we clearly understate rather than overstate WTP. Table 5-3 shows the minimum, maximum, median, and mean estimated WTP values and the 95% confidence interval for the mean WTP from the pair-wise choice model.

An individual's pair-wise choices, and thus WTP, depend upon whether or not annual household income is less than $12,000: households with low income (less than $12,000) are willing to pay less than households with higher income. This finding was not suggested by the attitudinal data, which indicated that respondents with lower incomes believe that the monuments are more important than do other respondents. It is likely that respondents with lower incomes may care more about the monuments, but cannot afford to pay for preservation; that is, their WTP is critically dependent on their ability to pay. Perhaps the unexpected finding is that the only significant difference among income levels is between respondents with incomes below $12,000 and all other respondents: there were no differences among individuals with higher income levels. One possible explanation for this is that the price levels considered for the preservation options, ranging from $0.25 to $50, only become 'unaffordable' at very low incomes. Table 5-4 reports the WTP amounts for representative individuals with income less than $12,000 and with income of $12,000 or more; both are 45 years old, Caucasian, and male.

More generally, the ratio of WTP amounts for respondents with low-income to other respondents depends on the respondent's gender. The WTP of low-income males is 60% that of males with higher incomes, regardless of age, ethnicity, and the level of preservation. The WTP of low-income females is 44% that of females with higher incomes. The combined effect of income and gender on WTP for Option C is shown in Table 5-5.

Older respondents are willing to pay more for preservation than are younger respondents. This finding is not suggested by the attitudinal data: there were no systematic differences by age to the attitudinal questions. Generally, for any 10-year difference in age, WTP differs by no more than 7%. As an example, the WTP amounts for higher-income (i.e., income above $12,000), Caucasian males of various ages are shown in Table 5-6.

A substantial difference is observed in WTP between Caucasians and non-Caucasians. Recall from Section 5.2 that this result is strongly supported by the attitudinal data. Across almost all attitudinal questions, the answers of non-Caucasians indicate that the existence or condition of the monuments is less important to them. Depending on the age of the individual, a representative non- Caucasian's WTP ranges from 34% to 60% of that of a Caucasian. The difference is most dramatic for young, non-Caucasian respondents. Table 5-7 shows the differences in WTP amounts for 45-year-old, higher-income males.

Each of the four individual characteristics - gender, income, age, and ethnicity - significantly influences an individual's WTP. The answers to the attitudinal questions suggest that education levels might also be a determinant: recall that the responses of respondents with lower levels of education differed from those of respondents with higher levels of education. In particular, those with no more than a high school diploma were more enthusiastic about the existence and condition of the monuments than all others, and those with some college or trade school were less enthusiastic than all others. However, no difference among education levels is evident in the pair-wise choices and, thus, the WTP amounts.

Consistent with the attitudinal data, whether or not a respondent has visited the Washington, DC monuments does not appear to be a determinant of one's WTP for preservation of the monuments. This further supports the use of survey methodologies that account for values not reflected by one's visitation behavior.

The estimated WTP amounts are one-time payments for a permanent stream of benefits: the preservation options were presented as permanent reductions in the rate of degradation. Thus, WTP amounts can be interpreted as the present value of a stream of future benefits. Likewise, these one-time payments can be represented as a stream of payments into the future. It is uncertain what discount rate should be used in converting these data from a present value to a stream of annual payments, but it is useful to make this calculation over a range of plausible discount rates to show what an equivalent annual WTP might be. Many WTP studies estimate annual rather than one-time payment values for various public goods. Table 5-8 gives a range of perpetual annuities equivalent to the mean and median WTP amounts for males from the pair-wise choice model.

 

5.5	Payment Card Models and Results

5.5.1

Payment Card Elicitation Format

In the payment card WTP question, Question 14, respondents were asked to indicate the maximum one-time payment their household would be willing to pay for Option C.⁹ Respondents were provided with the following schedule of amounts from which to choose:

Question 16 asked respondents the extent to which they were thinking about each of 12 valuation-related issues (e.g., free riding) while answering the payment card and pair-wise choice questions. One purpose of Question 16 was to help the respondents identify reasons why they might have misstated their true maximum willingness to pay when circling the payment card amount.

In Question 17, they were asked to consider whether they might have over- or under-stated their WTP when answering the payment card question, and were given the opportunity to select a new payment card amount. About 60% percent of respondents said that their answer was correct, 30% said that they were uncertain whether their answer was correct, and 10% (23 respondents) said that the amount they selected was either too high or too low. Of those 23 respondents, 16 gave revised payment card amounts: 5 respondents gave higher amounts than those given in Question 14 and 11 gave lower amounts. The remainder of the analysis uses the original amounts given in Question 14 unless a revised amount was given in Question 17, in which case the revised amount is used.

 

5.5.2

Payment Card Data

The payment card data can be analyzed as interval data by assuming that the respondent's true maximum WTP is at least as high as the amount circled on the payment card, but less than the next highest amount listed on the card. This interpretation presumes that a respondent will not circle any amount that exceeds her true maximum WTP, and therefore circles the highest amount on the card that is less than or equal to her maximum WTP.¹⁰ This interpretation means, for example, an individual's response is in the X to Y interval if they circled X, where Y is the next largest amount on the payment card. So if an individual circled $25, then his WTP is assumed to be between $25 and $35.

Figure 5-4 shows the frequency distribution of these selected WTP intervals. In all, 237 respondents provided numerical answers to the payment card question (i.e., they did not choose "Don't Know" or skip the question) and information about their demographic characteristics.

Figure 5-5 is the frequency distribution of the WTP midpoints corresponding to the WTP intervals. The mean of the midpoint data is $57.23¹¹; the median is $30.00.

No observations were omitted from the analysis unless the respondent did not answer the questions, including demographic characteristics. Respondents choosing the extreme payment card amounts, zero and "more than $200," were reviewed individually. In each case, the respondent's answers to the attitudinal questions do not conflict with the payment card amount chosen.

Of the 18 respondents selecting $0, 14 gave comments indicating that they did not care about preserving these monuments, that these programs were not worth anything to them, or that they could not afford to pay anything for these programs. Of the remaining four with ambiguous responses, two indicated that the government should pay for this on current budgets, but indicated in attitudinal questions that monument preservation was not very important to them. The remaining two are somewhat questionable zero values because they expressed some interest in monument preservation in the attitudinal responses but said they would not pay for these programs because either the government should pay or they had concerns about the specific proposed program.

All but 2 of the 29 respondents who selected $200 or more than $200 on the payment card gave explanatory comments indicating that it was important to them that the monuments be protected, that they could afford to pay the amount selected, or that they thought this was a fair amount to pay for these programs, as well as attitudinal responses indicating a concern for the preservation of these monuments. The remaining two respondents gave no explanatory comments, but indicated a concern for monumental preservation in their attitudinal responses. Two of the respondents giving high responses offered some ambiguous evidence of the accuracy of these responses by giving a high agreement rating in Question 16e to the statement, "I chose amounts larger than I would actually be willing to pay because I won't have to pay."

 

5.5.3

Modeling WTP from Payment Card Data

We use the circled amount to construct a WTP interval for each respondent. We interpret the amount circled as the lower bound on this interval and the next highest amount as the upper bound. For example, if a respondent circled $35.00, we take this to mean that he would be willing to pay at least $35.00 for Option C, but would not be willing to pay the next highest amount listed, $45.00. This procedure is appropriate for most of the amounts circled on the payment card, but does not apply to the highest amount, "More than $200," and the next highest amount, $200.

For those who circled "More than $200," their lower bound is assumed to be $200 and their upper bound is assumed to be $300. The choice of $300 is somewhat arbitrary yet irrelevant: the results of the interval models in this section are insensitive to this choice. The respondents who circled $200 are assumed to have a maximum willingness to pay equal to $200. Since they circled this amount, we assume that it is their lower bound, and because they did not circle "More than $200," it is also their upper bound.

Special consideration is also given to respondents who circled $0. These respondents were not given the opportunity to choose negative WTP amounts. A respondent's WTP for increasing preservation is negative if she prefers less to more preservation; this is a distinct possibility for anyone who dislikes the monuments. However, all 18 individuals who circled $0 indicated in their responses to the attitudinal questions that they think it is at least a little desirable that the monuments exist and that pollution-related weathering is undesirable.

The WTP amounts estimated for this model are only for the level of preservation associated with Option C, a 100% increase, or doubling, in the amount of time in which specific injuries occur to the monuments. WTP is assumed to be a function of an individual's characteristics and a random component that causes WTP to vary across individuals, even if they have the same characteristics. The estimated parameters in the WTP function are those that best explain the WTP interval chosen by the respondents. The function used to generate an estimate of each household's WTP for Option C is dependent on the household's demographic characteristics.

Estimation of an interval model requires specification of the distribution of the random component in the WTP function, that is, the distribution of WTP across all individuals with the same characteristics. Two distributional assumptions were investigated: normal and log-normal. The normal distribution allows WTP to take both positive and negative values, and is a symmetric distribution. The log-normal assumption restricts WTP values to be positive, and the distribution is skewed to the right. The data suggest consideration of a log-normal distribution: no WTP for increased preservation is negative and more than 10% of the sample indicated a WTP of $200 or more suggesting a long tail to the right.

Consider first, the simple case where the expected value of WTP is the same for all households, independent of an individual's characteristics. Assuming a normal distribution, the interval model gives an expected value of WTP of $55.68. Alternatively, assuming WTP has a log-normal distribution, the expected value of a household's WTP is $99.04. Although there is a large estimated variation in WTP across household WTP for both distributional assumptions, in both cases mean household WTP is significantly positive. For the normal distribution, the 95% confidence interval on average household WTP is $47.66 to $63.70. For the log-normal distribution, the 95% confidence interval is $49.25 to $148.86.

Assuming household WTP has a log-normal distribution dramatically increases the estimated expected value of household WTP. This happens for two reasons. First, there is the large variation in the payment card responses that is unexplained.¹² Second, with the log-normal assumption the expected value of WTP is a function of the expected value of the log of WTP and the estimated variance of the log of WTP, increasing as this estimated variance increases.¹³ These two factors cause the higher expected value of WTP when assuming a log-normal distribution.

The remainder of this section describes models for each distributional assumption. Section 5.6 compares the results of all three models: the pair-wise choice model, the payment card model assuming a normal distribution and the payment card model assuming a log-normal distribution.

The WTP estimates reported here are generated from an interval model assuming the random components are distributed normally (payment card model 1). The estimated WTP and standard deviation vary with individual characteristics. The WTP estimates are a function of income and ethnicity only; gender was not a significant determinant of WTP. WTP increased with income: annual household income less than $12,000, income from $12,000 to $24,999, and income greater than $25,000. Caucasians have a higher WTP than non-Caucasians.

The standard deviation on household WTP varies significantly with income and ethnicity; the variation in household WTP is larger for Caucasians than non-Caucasians and increases with income. The differences by income are most likely because higher income households have more flexibility in the amount that they are able to spend on preservation. That WTP varies less among non-Caucasians than Caucasians suggests that the WTP of non-Caucasians is more closely related to their ethnicity. This is supported by some of the open-ended comments of non-Caucasians saying that the monuments do not represent their history.

Table 5-9 includes the minimum, maximum, median, and mean estimated household WTP and the 95% confidence interval of the mean for the 237 individuals who answered the payment card question and provided demographic information. The maximum WTP of the sample is the estimated WTP for more than half of the respondents (all respondents who are Caucasian and have incomes of $25,000 or more), thus the maximum is equal to the median WTP for the sample.

The characteristics of the individuals with the minimum and maximum WTP amounts follow the same pattern as those in the pair-wise choice model. The minimum WTP amount, $14.32, is that of non-Caucasians with household incomes less than $12,000. The maximum WTP amount, $68.06, is that of Caucasians with incomes of $25,000 or more. Table 5-10 shows the six possible estimated values of WTP by ethnicity and income categories.

The WTP estimates reported here are generated from an interval model assuming that the random components are distributed log-normally (payment card model 2). The WTP estimates are a function of age, gender, income and ethnicity; no other socioeconomic variables were found to be significant. WTP increases with income: annual household income less than $12,000, and income greater than $12,000. WTP was not sensitive to income distribution above $12,000 in this model. Caucasians have a higher WTP than non-Caucasians. Older respondents, those over 50, are willing to pay more than are younger respondents. And, females are willing to pay more than males.

Table 5-11 includes the minimum, maximum, median, and mean estimated household WTP and the 95% confidence interval of the mean for the 237 individuals.

The minimum WTP amount, $13.71, is that of non-Caucasian males under 50 with household income less than $12,000. The maximum WTP amount, $190.58, is that of Caucasian females over 50 with income greater than $12,000.

Similar to the reporting of the pair-wise choice model results, Table 5-12 includes the minimum, maximum, median, and mean estimated household WTP and the 95% confidence interval of the mean for the 237 individuals, all as males.

 

5.6	Comparison of Pair-Wise Choice and Payment Card Models and Results

5.6.1

Comparison of Results

The estimated household WTP amounts for Option C from the three models, the pair-wise choices, the payment card assuming a normal distribution (payment card model 1), and the payment card assuming a log-normal distribution (payment card model 2), are shown in Table 5-13. These are the WTP amounts associated with all respondents as male, a restriction that affects the pair-wise choice results and the payment card model 2 results; the payment card model 1 estimates are not a function of gender.

Figure 5-6 shows the 95% confidence intervals for these mean values. Payment card model 1 and the pair-wise choice model generate effectively the same estimates of household WTP for preservation Option C. The confidence interval for payment card model 2 is much wider and spans the intervals of the other two models.

In both models, it is estimated that poorer households and non- Caucasians are willing to pay less. The main difference in the explanatory variables across the models is that age and gender are significant explanatory variables in the pair-wise model and payment card model 2, but not in payment card model 1.

An additional comparison of the pair-wise choices and the payment card responses was made by comparing choices made between Option C and no program to the dollar amount selected on the payment card. The expectation was that if these responses were consistent a respondent would not choose Option C with a higher price than the amount selected in the payment card. Conversely, a respondent would not choose no program over Option C at a price less than the amount selected in the payment card. Eight-five percent of respondents gave answers that were fully consistent when compared in this way.

Eleven percent of respondents indicated a higher WTP on the pair-wise questions than on the payment card question. For example, in a pair-wise choice format a respondent might choose Option C at $25 over the status quo, indicating that she places a value of at least $25 on Option C. Yet, in the payment card format, she might say that her maximum willingness to pay for Option C is $16 (putting her maximum WTP within the interval of $16 to $20).

Alternatively, 4% of respondents indicated a lower WTP on the pair-wise questions than on the payment card question. For example, in the pair-wise choice format a respondent might choose the status quo over Option C at $25, indicating that she would rather have no additional preservation than pay $25 and get Option C. Yet, in the payment card format, she might say that her maximum willingness to pay for Option C is $35.

 

5.6.6

Differences in Elicitation Formats

The theory underlying both the pair-wise choice and payment card models assumes that WTP has a random component that varies from choice occasion to choice occasion and so is consistent with variation in answers both over time and across question types. However, there may be systematic differences in how people answer pair-wise choice questions versus payment card questions, even when the mean and median estimates generated by the two models are very similar. Before listing some reasons why answers might differ depending on the elicitation format, it is important to note that for most individuals in the sample (85%), the responses to the pair- wise choices and payment card question are quite consistent.

One explanation of any differences might be that the pair-wise choice and payment card elicitation formats require respondents to make different decisions. For the pair-wise choices, respondents are asked to choose between two alternatives. This is a rather common decision-making process, not unlike choosing cereal at the grocery store or an entree at a restaurant. In contrast, the payment card question asks respondents to decide the maximum amount they are willing to pay for a good, in this case Option C. This is a somewhat different task and may yield somewhat different responses.

Responses may also differ because there are different potential biases associated with the two types of questions. For example, when asked a payment card question in the focus groups, some individuals had a tendency to calculate the total revenue raised for such a program if all households paid the amount they were willing to pay. Doing this appeared to lead some respondents to state lower amounts than their maximum WTP. This concern and reaction was revealed through debriefing comments. There appeared to be less tendency to "cost calculate" with the pair-wise choice questions. (This issue is also discussed in Section 3.1.)

Responses might also vary because preferences vary. And, even if preferences are constant, an individual's preferences regarding monument preservation are likely not to be at her "cognitive fingertips" and so require time and cognition to recover. If this is the case, questions asked later in a survey - allowing more time to recover one's preferences - can lead to more accurate answers.

13. The mean of a log-normal distribution is , where is the standard deviation of the .

Contact me via email | Document Index | Home Page

Last Update: 1-9-98