Estimating the Benefits of Clean Air. Contingent Valuation and Hedonic Price Methods

Mohammed Belhaj Department of Economics Göteborgs University Box 640, SE 405 30 Göteborg Sweden Tel: + 46 31 773 2516 Fax:: + 46 31 773 1043 [email protected] Abstract: The higher growth rate of transportation and thereby fuel demand with higher concentration of vehicular traffic in urban areas is in general highly correlated with increased air pollution. This paper deals with the estimation of willingness to pay for a 50 per cent reduction of air pollution caused by road traffic in Rabat-Salé (Morocco) using contingent valuation and hedonic price methods. In CVM two techniques are employed. these are the iterative bidding and the dichotomous choice. Both empirical analyses are based on the same set of individuals. This strategy enables to compare the two techniques; a Generalized Tobit and Probit models are used for the estimations. Moreover parametric and non-parametric approaches are used to calculate the mean and median willingness to pay. Since data on emission concentrations is not available in Rabat-Salé we use distance to the city center such as a surrogate for environmental factors to conduct a hedonic price study and the result obtained are almost similar to the CVM estimates. 1.1

Introduction

The growth of transport in urban areas is in general highly correlated with increased air pollution. Rabat- Salé is such an area where we conduct a contingent valuation study in order to examine whether or not the inhabitants of these cities suffer from air pollution problems or not and whether they are willing to pay to reduce the emissions.

To assign an economic value to clean air, various direct and indirect methods are available. In this paper contingent valuation is used to assess the willingness to pay to reduce air pollution in RabatSalé (Morocco). Both the iterative bidding and the dichotomous choice values are estimated from the same data. This design allows for a comparison using the same respondents. For the sake of

1

comparison the hedonic price method is also used based on changes in property values as we move away from the city center where concentrations of air pollution are highest.

This paper is organized as follows: Section 1.2 discusses the CVM method and section 1.3 compares the different CVM techniques. Section 1.4 includes the estimation methodology . Section 1.5 analyzes the socio-economic determinants of WTP. Section 1.6 analyzes the hedonic price method and section 1.7 concludes. 1.2

The contingent valuation method

Different direct and indirect methods have been developed to value a good or a service that is not traded in a private market,. The contingent valuation method, CVM is a direct way of using surveys to value public goods. The CVM is a relatively new technique1 built on the idea of a hypothetical market scenario where a public good is transacted. An auctioneer represented by an interviewer presents the good to the members of the hypothetical market (i.e., the sample) to elicit willingness to pay for the good in question. This method is called contingent2 because it depends upon the hypothetical market presented to the person being interviewed.

The interviews can be conducted in various ways, e.g., by mail, by telephone or face to face3. The questionnaire generally includes three parts:

1

In the early 1960s the economist Robert K. Davis was the first to use a questionnaire to estimate the benefits of outdoor recreation in a Maine area. 2

It has been also called the survey method, the interview method, the direct questioning method, the hypothetical demand curve estimation method, the difference mapping method, and the preference elicitation method. 3

Each of these procedures possess advantages and disadvantages (see Mitchell and Carson (1989) for more detail).However, The NOAA (National Oceanic and Atmospheric Administration) panel recommended inperson interviews for their superior reliability (Hanemann, 1994).

2

i- In order to avoid misunderstanding and to increase consistency4 the respondent is supplied with detailed information about the good to be valued by way of a hypothetical market scenario ii-The respondent is asked questions to elicit his or her willingness to pay. These questions are carefully formulated so as not to bias the answers. iii- The respondent characteristics (e.g., age, income, sex...) and preferences of interest for the good valued are collected.

Before discussing the estimation framework, point (ii) above is of very special interest: To get the willingness to pay of a respondent, the valuation question can be asked in several ways which reflect the particular technique used. These techniques each possessing certain strengths and weaknesses are as follows:

a. Iterative bidding: Sometimes called the sequential bidding techniques is the oldest and most frequently used contingent valuation procedure. It is based on continuous responses. The interviewer proposes an initial starting bid5 which the respondent accepts or rejects. If the initial bid is accepted, the interviewer revises the bid upward until a final value is reached. If the initial bid is rejected the initial value is revised downward until a final value, which can be zero (or in principle even negative) is reached. An inherent weakness, however, is that the initial bid can influence the bidding process (Boyle et al (1988)).

4

The CVM literature has identified a long list of potential biases that can arise due to the hypothetical nature of the setting in which willingness to pay is elicited, and due to context effects associated with the elicitation method: In Cummings, Brookshire, and Schulze (1986), a panel that has reviewed CV methods proposes a list of "reference Operating Conditions" that must be met if CV methods are to provide reliable estimates of value. These conditions include requirements such as "subjects must understand, be familiar with the commodity to be valued" and "subjects must have had (or be allowed to obtain ) prior valuation and choice experience with respect to consumption levels of the commodity") Mcfadden et al (1992).

5

For a rigorous discussion of the starting bid, also called starting point or starting price, see Mitchell et al (1989): Assume the respondent believes that the starting price, SP, presented randomly contains some information about the value of the project or policy being offered. This assumption is usually motivated by a claim that it is reasonable for a respondent to believe that a competent researcher would not specify a SP that was "just plucked out of the air". The SP must bear some relationship to something relevant: e.g., the kinds of answers other respondents have been giving, or expert estimates of the value of what is being offered.

3

b. Payment cards: This method has been introduced by Mitchell and Carson as an alternative to iterative bidding to avoid of starting point problem. The payment cards are formulated in such a way that a sequence of prices are portrayed6 beginning at zero. These prices represent for instance estimates of what people in a specific income category paid for selected public services in the preceding years. The respondent is then asked if he is willing to pay a one of these prices. The answer is final and no bidding is involved.

c. Dichotomous choice, DC7: Based on a discrete response this technique was first used by Bishop and Heberlein (1979)8. The respondent is asked to answer yes or no to the take-it-or-leave-it offer for the object being valued. While the iterative technique and payment cards techniques analyze quantitative responses, DC analyzes qualitative answers (yes/no). These qualitative responses provide much less information about the respondents' actual values (preferences) than is utilized when continuous numerical responses are obtained with iterative bidding and/or payment cards. This method is, however, preferred by the NOAA Panel since the respondent here does not have a strategic reason to answer untruthfully9. It also reduces the incentives for strategic behavior10, since it is more difficult for the respondent to influence the mean willingness to pay which is not the case using iterative choice. Nevertheless, comparing DC and the iterative bidding, we would argue that this last technique includes a number of DC questions to which the respondent says yes or no until the final bid is accepted.

6

Other researchers have used modified payment cards (see Boyle et al (1988)).

7

Within the DC two other techniques are used. The spike model and the double bounded dichotomous choice contingent valuation model, see Kriström (1995) and Hanemann et al (1991). 8

See also Cameron and James (1987) and Kriström (1990).

9

In Hanemann (1994).

10

Loomis (1987).

4

1.3

Comparison of techniques

To compare the three techniques used in the contingent valuation methods, Boyle and Bishop(1988) valued the scenic beauty along the lower Wisconsin River where the Wisconsin Department of Natural Resources was proposing to establish a state forest. The results are presented in table 1.1. Boyle and Bishop concluded that the value estimates acquired by the three techniques are rather similar:"The iterative bidding estimate may be affected by a starting point bias (but) this bias does not appear sufficient to make a dramatic difference between the bidding estimate of value and the estimates derived with the other techniques. However, the dichotomous choice gives somewhat lower values for both the mean and the median".

Table 1.1 Estimated Willingness to Pay for Scenic Beauty (in USD) per Year. Method

Mean

Median

n

Iterative biding

29.82

15.83

176

14.56

180

11.67

176

(3.77) Payment card

29.36 (3.88)

Dichotomous choice

18.88 (3.42)

Source: Boyle and Bishop (1988). Note: Numbers in parentheses are standard errors. The estimated mean was derived by integrating one minus the estimated logit Cumulated distribution function over offers from zero to infinity.

1.4

Estimation methodology

1.4.1 Sample selection

The towns of Rabat and Salé (Rabat is the capital of Morocco) are two adjacent cities separated by the Bouregreg river. This urban area belongs to the second most polluted zone in the country where the population and number of transport vehicles is the next largest.

In July 1995 and with the help of The Moroccan statistical office in Rabat, a stratified random sample was obtained in Rabat-Salé by separating the population elements into non overlapping 5

groups, i.e., strata11. Rabat-Salé is divided into 1 433 strata or districts with 808 in Rabat and 625 in Salé. The number of households in Rabat is 127 083 and in Salé 109 250. Each strata is homogenous in the sense that it belongs to a specific district which include similar housing types and surrounding environments.

After the selection of 25 strata and before conducting the final interviews, a pilot study including 100 households was carried out. The purpose of the pilot study was twofold; First discussing the questionnaire and its formulation with the interviewee permitted us to correct misunderstandings and to include other relevant questions. Second the pilot study served to decide the stating bids which were then used in the final interviews. Using the iterative choice method, 4 randomly chosen households from each stratum were asked about their maximum willingness to pay to reduce air pollution in Rabat-Salé. From the pilot study the median willingness to pay was 50 dirhams. The starting bids which were used in the final study were calculated so that they correspond to two lower values and two upper values arrayed around the median as such:

18

34

Median = 50

66

82

Turning to the final study, the 25 strata were used to select a sample size of 400 households and a simple random sample of 16 households was chosen from each stratum. The reason for choosing the sample size of 400 households is as follows: We wanted to estimate the proportion p of households having a willingness to pay to reduce air pollution. Our population size N is equal to 236 333 and we wanted to choose a sample size n. The question is how to choose n. The natural Maximum likelihood method for estimating p is to use the relative frequency f. It is known that f is unbiased and that

11

There are three reasons stratified random sampling often results in increased information for a given cost: 1. The data should be more homogeneous within each stratum than in the population as a whole. 2. The cost of conducting the actual sampling tends to be less for stratified random sampling than for simple random sampling because of administrative convenience. 3. When stratified sampling is used, separate estimates of population parameters can be obtained for each stratum without additional sampling ( see Scheaffer et al (1979) for details).

6

f&p f(1&f) N&n n N&1

is approximately N(0,1). If N is very large as compared to n, the term N-n/N-1 becomes approximately equal to 1. Since we want to choose n so that our estimate of p became accurate in the sense that a 95% confidence interval for it is not too large, such a confidence interval has the form

p ' f ±1.96

f(1&f) n

The total length, l, of the interval is not allowed to be larger than 10%, i.e.,

l'f %1.96

f(1&f) & (f &1.96 n

f(1&f) ) n

f(1&f) # 0.10 n

' 2( 1.96

or equivalently

n'(2(1.96)2

f(1&f) 0.12

This is a function of the unknown frequency f through f(1-f). Fortunately, it is possible to obtain an upper bound for f(1-f) namely 1/4 when f = ½. If we use this case, we get

n #

(3.92)2 1/4 16(0.25 . ' 400 0.01 0.01 7

which is the sample size we used in our study. 1.4.2 The questionnaire The questionnaire12 was conducted face to face and contained 27 questions for the three parts of the CVM framework described above. A total of 382 persons, or 96 percent answered all questions, some households were never available13 and only two respondents refused to answer the questions. The questionnaire included a letter describing the hypothetical market, why the study was being conducted and what instruments were to be used to reduce air pollution. Using CVM to elicit willingness to pay to reduce air pollution is a difficult task since the degree of reduction is on the one hand complex to establish and on the other hand very hard to explain to respondents. Many studies such as Rowe et al (1980) and Shechter et al (1991) used photographs of both visibly polluted and relatively clean days before asking the respondents what they would be willing to pay to reduce air pollution. This may be possible in an industrialized area where emissions from coal fired plants produce are substantial. In the case of Rabat-Salé however with practically no industries, pollution is dominated by vehicle emissions which vary enormously from one street to another. To use photographs that give accurate descriptions of certain percentages of reduction is difficult and might be misleading. However, in order to ask respondents their willingness to pay for pollution abatement we use a 50 percent reduction (as in Shechter et al (1991)) since the concept of a "half" number is easy to understand compared to other percentages.

As some people might believe that citizens in a developing country such as Morocco are unaware of environmental problems, the questionnaire started (before the letter was read) by asking the respondents (see Appendix 2) Do you think that Rabat/Salé suffers from environmental problems? If the answer was yes, then the next question was Which are those problems in order of importance?

12

The survey was conducted in July 1995.

13

Some of these households went away on holidays and others were in Morocco only sometimes during the year (emigrants living in other countries).

8

If the answer was no, the result was reported and the introductory letter with the detailed information about the good to be valued was read. Posing such questions was fundamental since asking people about their willingness to pay when they are not aware of any environmental problems would make no sense in the CVM framework. The proportion of the population who gave a positive answer, i.e., there are environmental problems, was 93 percent. Thirty-six percent of those households considered air pollution to be the greatest problem in the area. The other problems reported are by order of importance; household waste14, noise from vehicles and water problems. 1.4.3 The valuation question

Since the iterative bidding includes a number of DC questions to which the respondent say yes or no, we used question 13 to conduct a DC study and question 16 to conduct the iterative choice study. This design allows at least for a partial comparison using the same respondents. However, it is also worth pointing out that the iterative bidding approach generates a scenario most similar to that encountered by Moroccan consumers in their usual market transactions. They do not face a "take it or leave it" situation when buying but rather a negotiation or a bargaining situation.

14

Household waste is inefficiently collected; sometimes this waste lies around the house for some days before it is collected.

9

Table 1.2 Characteristics of the sample. Mean

Std dev

Min

Max

WTP (Dirhams)

67.25

101.26

0

1 000

Age

43.2

13.26

20

80

Number of children

3.41

2.86

0

12

Household income (Dirhams)

2 759

2 152

250

15 000

Family size

5.67

2.76

1

16

Gender (male=1)

0.60

Importance of air pollution

0.36

Literacy

0.57

Environmental awareness

0.94

Marital status

0.77

Respiratory disease

0.17

Non importance of the environment

0.34

Importance of air pollution

0.21

Interest in environmental problems

0.49

Authorities’ indifference

0.56

Car ownership N= 382.

0.23

As shown in table 1.2 the mean willingness to pay per month for the iterative bidding was 67.25 Dirhams (DH). That is 8 USD (1 USD = 8.4 DH). The median WTP is 50 DH, the same amount as in the pilot study. The mean income is 2 759 DH, which sounds reasonable in Morocco where the "minimum" wage is 1 510. The family size is equal in average to 5.67. As concerns the gender variable 60 percent of the respondents were men and 40 percent were women. We believe this is a representative population since the gender distribution in the entire country is 49 percent men and 51 percent women. However, the slight overrepresentation of men in our study reflects a culture where women are not supposed to talk to strange men. We did our best to reduce this small bias by sending women enumerators to such households. Considering the question of literacy our population is very representative since literacy in urban areas in the whole country is estimated to be 57 percent. Of the interviewed people 36 percent considered the problem of air pollution in the city to be of greatest importance before the introduction letter was presented. 10

Turning to respiratory deceases, 17 percent of the respondents believed air pollution to be the cause. When discussing the authorities’ committment to taking measures for reducing emissions, 56 percent agreed that the authorities are indifferent. As concerns car ownership 23 percent of the households owned a car. 1.4.4 The DC method

Based on random utility functions used in transportations models, Hanemann (1984) provided the link between economic theory and DC methods using the random utility maximization model. Following Hanemann (1984) and Kriström (1990) we write the utility function in the following form: V ( z, m; c) = U (z, m; c) + ,

where V(.) is the individual’s utility function where prices and quantities are not considered for simplicity. z is an index of air quality, m is household income and c is a vector of household characteristics. U(.) is the observed part of the utility function and , is an error term representing the notion of random utility maximization (RUM). It is the RUM concept which provides the link between a statistical model of observed data and an economic model of utility maximization. In a RUM model it is assumed that, while the individual knows her preferences with certainty, they contain some components which are unobservable to the econometric investigator and are treated by him as random (Hanemann (1996)). These unobservables could be characteristics of the individual and/or attributes of the item. The respondent is confronted with the question whether to pay an amount A to reduce air pollution from z 0 to z 1.(z 0 being the status quo and z 1 is a situation where the air pollution is reduced by say 50 per cent). If he is willing to pay A then V ( z1, m-A; c) +,1 $ V ( z0, m; c) +,0

11

Thus, A is the maximum WTP for the change from z 0 to z 1. This formulation makes it apparent that the WTP is a random variable. To measure the mean and median of the WTP, parametric and non- parametric approaches are used. -A simplified parametric approach Following Hanemann (1995) and Kriström (1990) for any distribution function Fv and utility difference )U, the mean willingness to pay takes the form: 4

E(WTP) '

m

4

AUm f v [)U(A)] dA '

&4

where U

m

=*U/*m and m, f v, g

m

A gwtp (A) dA

&4

wtp

denotes income, densities of v and WTP respectively. The

middle part of the equation says that E(WTP) is obtained by weighting each bid A with the probability f v()U(A)) and the marginal utility of income. Hence, if the random variable v has expected value zero and the utility function is linear in income, U i= " i +$m, implying )U = "-$A, then the expected value of willingness to pay takes the form15 E(WTP) = "/$. -Non-parametric approach

To calculate the mean and the median in the case of the dichotomous choice method i.e. where the data is binary, and when the different probabilities form a monotone non-decreasing sequence of proportions, the theorem of Ayer et al (1955) is used to estimate the proportions of individuals willing to pay the preassigned amounts of money. Accordingly, the data is grouped by the yes/no answers at each given bid Ai which means that we can focus on the proportion yi of yes answers in each group. Hence, the usual maximum likelihood (M-L) estimate of the probability for a yes answer is p-i (= yi / ni ) at each Ai p-i is unbiased and has mean pi and variance p(i 1-pi ) / ni ,where ni is the number of respondents in group i. This fact can be utilized to exhibit the asymptotic properties of the proportion as an estimator of the 15

see Kriström (1990) for more details.

12

probability for acceptance. The proportion of the yes-answers converges toward the true probability of a yes-answer as the sample size increases.

A typical discrete response survey involves a number of different bids giving a sequence of proportions: p-1, p-2,...,p- k ,where p-1 corresponds to the lowest bid A1, and p-2 corresponds to A2 and so on. Intuitively, one would expect that the proportions of yes-answers would decrease with increased bids as the sample size getts larger. This might not be valid in the case of surveys with small sizes. Table 1.3 confirms this observation and reports the bids and the proportions of yes-answers16:

Table 1.3 Bids and proportion of yes-answers. Bid

Proportion of yes answers

18 34 50 66 82

35/48= 0.729 23/44= 0.522 54/95= 0.568 33/89= 0.371 62/107= 0.579

According to the Ayer et al theorem, the maximum likelihood estimates that take the order of the proportions into account are found in the following way. If p-i is smaller or equal to p- i+1 for some i (i=1,2,..,k-1) then p^i = p-i where a bar denotes as already mentioned the usual M-L estimates and p^i denotes the new M-L estimates . Otherwise the proportions p^ iand p^

i+1

are then both

replaced by : ( yi + yi+1 ) / ( ni + n i+1 ), where yi is the number of yes- answers in group i and ni is the number of respondents in group i. The procedure is repeated until the sequence is monotonic.

16

Respondents who accepted the bid 82 have an average income of 3 791 DH, while the average income for the lowest bid happened to be 1 759 DH.

13

The mean and median WTP are computed by applying this algorithm and drawing the corresponding empirical survival function. The complete data set and the survival function are given bellow.

Table 1.4 Bids and proportion of yes-answers. Bid

Survival function

18 34 50 66 82

0.729 0.545 0.545 0.475 0.475

Assuming a linear interpolation between the points described in table 1.4 and assuming further that at bid zero the survival function takes value 1 and at bid 120 (median * 2) it takes the value zero, the median . 60 Dirhams and the mean . 63 DH are computed from figure 1.1. The median is read off at p= 0.5 from the figure and the mean is calculated as the area bounded by the survival function. This is obviously quite an uncertain way of calculating the median since it depends on the assumptions about the end points of the distribution; If for example at p=0 we assume that the bid is equal to 1000 ( 1000 being the highest bid given in the iterative choice procedure) we get a mean willingness to pay equivalent to 270 Dirhams.

Figure 1.1 Mean and median WTP.

Probability of yes answers

1 .8 .6

.4 .2

20

40

60

80

100

14

120 Bids

1.4.5 The iterative bidding

In the final study and using the iterative bidding 30 percent of the representative population stated a WTP equal to zero. Sixty percent of the zero WTP was obtained from families whose monthly income is inferior to the "minimum" wage i.e., 1510 DH. These results are presumably dependent on absolute poverty and not on free riding behavior. From the remaining 40 percent of zero WTP, 13 percent have a monthly income greater than 3000 DH. Many of these respondents motivated their zero WTP by saying they thought that the authorities should solve the problem. 1.5

Socio-economic determinants of willingness to pay

Apart from the fact that we are interested in the mean and median willingness to pay for the cost benefit analysis, we are also interested in the factors that have impact on the given bid. These factors include various socio-economic characteristics of the household notably income and household composition. Also important are such factors as health, education and opinions about various issues such as the importance of environmental problems and the authorities’committment to solve them. It would have been important to also include pollution levels in order to study their effect on WTP. Unfortunately data on emissions did not exist at the time the study was conducted.

Using discrete choice and iterative bidding require the use of accurate models for estimation. In the case of DC where the dependant variable is binary, estimation with ordinary least square (OLS) would not be efficient because the dependant variable includes a large number of zeros. Therefore a probit model is used.

As concerns the iterative bidding, this method involves models with limited dependent variables i.e., the dependent variables are continuous but include a large number of zeros. Thus limiting the range of the values of the dependent variable leads to a nonzero mean of the disturbance and to biasedness and inconsistency of the OLS estimators There exist different models of estimation within the framework of limited dependent variables and the most known are Tobit, Heckman two step and the generalized Tobit models. The Tobit model and Heckman two step method are

15

applied in appendix 1 using the iterative bidding data. The generalized Tobit, called also type-two Tobit model is used here.

Turning to the variables included in the models, there is little guidance from economic theory about which characteristics such as age, gender, etc should be used for estimation. Thus, many explanatory variables, even though not significant are included for estimation. The results of the Probit and generalized Tobit model are presented in table 1.5. Significant variables are age of the respondent, household income, gender17, literacy, if the interviewed is aware of environmental problems in the area , if the respondent is married, if the authorities are considered to be indifferent toward environmental problems and if the respondent has a car.

Starting with the age of the respondent, older people have lower WTP in both models. This can be seen as a decreased interest in emission abatement as age increases. As expected, household income is positive and highly significant suggesting higher WTP as income increases. In order to check the impact of family size on WTP we replaced the variables income and family size by family size/household income in the probit model. The obtained coefficient i.e., 0.0001 is highly significant implying that larger families care about the future and are willing to pay higher amounts for cleaner air. Similarly, we replaced the variables number of children and family size by number of children/family size but this variable showed to be positive but not significant. Turning to the gender variable we run a probit model where instead of man, a woman variable was included in the model. The result showed that the marginal effect of being a woman is equal to -0.11 and highly significant. This result may be interpreted to mean that in general, women are responsible for the household’s e16conomy and are constrained by its budget. With regard to the variables literacy and environmental awareness, educated and environmentally aware respondents are willing to pay more to reduce air pollution. In the probit model, respondents suffering from respiratory disease would also contribute to decrease air pollution in Rabat-Salé. An increase in the authorities’ actions would reduce the probability of having higher WTP. Car ownership has a positive impact on higher willingness to pay, at least when the generalized Tobit model is considered. This can be seen as a willingness to pay to reduce the emissions they are generating. 17

Most of the explanatory variables used here are binary variables. Depending on the nature of the binary variables (i.e., they take only two values, say 1 for male and 0 for female), the results are not interpreted as marginal effects but rather as derivatives.

16

As concerns the starting bid, it is negative as expected. Unfortunately, this variable is significant in the probit model implying a starting bid bias. Moreover, the Rho variable in the generalized Tobit model is non significant and suggests that we did not have sample selection bias.

Comparing the results of the generalized Tobit to those of the Tobit reported in appendix 1 the the variable family size has a negative coefficient suggesting a decrease in WTP as family size increases. On the other hand, the starting bid, although not significant, positive in the Tobit model. In the case of Heckman’s two step model only income is significant.

Table 1.5 Marginal effects of the Probit and Generalized Tobit models. Probit

Generalized Tobit

Intercept

-0.21 (1.1)*

-0.77 (1.7)

Age

-0.004 (1.8)

-0.02 (2.2)

Number of children

0.01 (1.0)

0.03 (0.9)

Household income

0.00005 (3.1)

0.0001 (1.7)

Family size

-0.001 (0.1)

0.04 (1.1)

Gender (male=1)

0.10 (2.1)

0.06 (0.3)

Importance of air pollution

-0.02 (0.5)

0.02 (0.1)

Literacy

0.10 (1.9)

0.55 (2.4)

Environmental awareness

0.40 (2.8)

0.97 (2.5)

Marital status

0.07 (1.2)

0.54 (2.5)

Respiratory disease

0.10 (1.7)

0.24 (1.0)

Non importance of the environment

-0.07 (1.1)

-0.001 (0.01)

Importance of air pollution

-0.06 (1.1)

0.02 (0.1)

interest in environmental problems

0.03 (0.5)

0.36 (1.7)

Authorities indifference

-0.10 (2.5)

-0.37 (2.1)

Car ownership

0.06 (0.9)

0.67 (2.0)

Starting bid

-0.003 (2.9)

-0.006 (1.4)

Rho *) Values within parentheses are t-statistics. N=382

0.45 (1.11)

1.5.1 Mean, median and aggregated WTP 17

The DC and iterative choice methods may give different estimates of descriptive measures such as the mean and median. Referring to the results of Bishop et al in section 1.3 the DC gives somewhat lower values for the mean and median. Sellar et al (1985) present some evidence implying the DC gives higher estimates of mean WTP. Moreover, Kealy et al (1988) present results that indicate no significant difference. However, economic theory gives no explanation as to why differences may exist.

Using Probit estimates where " i.e., constant is equal to -0.21 and $ i.e., the starting price coefficient is equal to -0.003, and using the simplified parametric approach the mean and median willingness to pay are approximately equal to 70 DH i.e., 8 USD.

Table 1.6 shows the mean and median values in our study; the median in the iterative choice is lower than that of the DC method and the mean WTP is in the same range.

Table 1.6 Mean and median WTP in USD. Method

Mean

Median

Total benefits (106) Month

Year

Iterative choice

8

6

4.7

56.4

Dichotomous choice -non parametric approach

7.5

7

4.4

52.8

-parametric approach

8

8

4.9

58.8

Since CVM is a tool of cost benefit analysis its role is to provide aggregate benefits. However, expanding the sample’s values to a whole population may imply biases if proportions of nonrespondents is high. According to Loomis (1987) and Mitchell et al (1989) there are different approaches to estimating aggregate results obtained from samples with low response. One way is to assign sample average values to nonrepondents. Another more conservative approach, gives zero values to nonrespondents.

18

In this study we did not experience a high rate of nonrespondents. Thus, using the estimates of table 1.6, the monthly gross benefit for reducing atmospheric contamination by 50 percent is obtained by summing over the total representative population. We assume the respondents are representative for all adults, and children (< 18 years) have no separate WTP. We interpret the WTP elicited by individuals as their own evaluation and not the household’s18. Therefore, we aggregate over the total adult population in Rabat-Salé which was equal in 1995 to 588 46919 individuals assuming average adults by households being 2.49 persons. 1.6

Hedonic prices

There are several indirect methods ( see for instance Mitchell et al 1989) that can be used to measure the benefits of public goods and the most frequently used being the travel cost and the hedonic prices methods. The hedonic approach20 has been used to study: a) how variation in housing attributes e.g., size of the house, number of rooms, distance to central business district (CBD), can be capitalized into housing prices. b) how to impute a price for an environmental good by examining the effect which its presence has on a relevant market-priced good.

For case (a) subject to urban transport studies, many cases reveal that modernization and/or provision of new transport services from an outlying area to the CBD have led to increased value of the properties. In case (b), houses located in areas suffering from fewer environmental disamenities tend to have a higher value than identical houses situated in more polluted areas. We use the latter specification to conduct a simple hedonic price study in the case of Rabat-Salé.

Since environmental factors such as air quality concentrations are not available in our case we use distance to CBD as a surrogate factor to indirectly assess the changes in property values and to 18

Vainio (1995) tested this by asking a respondend to give his or her WTP; then the same question was directed to the spouse. It was anticipated that the spouse’s WTP would often be zero because the respondent would already have contributed for the whole household. However, this was not the case. 19

This number ( = 49 per cent of the population) is very representative of Rabat-Salé in particular and Morocco in general where the proportion of adults is estimated to be 56 percent. 20

Rosen (1974) provided a widely used model to explain the theoretical origins of the hedonic results. In that model there are demands and supplies for the various characteristics of a differentiated product. Most of the research on estimating the underlying demands and supplies for characteristics has been done in housing markets.

19

imply that the equilibrium marginal willingness to pay of the households21. This data deficiency explains why a full scale study was impossible, and why a short note to illustrate the method is presented here. 1.6.1 Data collection and characteristic of the sample

While collecting data for the CVM study, questions relevant to the hedonic study were asked. Thus the sample size is equal to that of the CVM. Since the collection of data related to house prices turned out to be a difficult task, rents are used as a proxy variable because the rents move in line with house prices.

Table 1.7 shows the characteristics of the sample. The mean rent payed in Rabat-Salé is equal to 1 023.67 Dirhams, the mean distance to CBD is 3.25 km and the mean living area in m2 is 7.46. Households living in apartments are 16 percent and those living in Moroccan houses22 are 72 percent of this representative population.

Table 1.7 Characteristic of the sample. Mean

Minimum

Maximum

Rent (DH)

1 023.67

75.00

4 000

Distance to CBD (km)

3.25

0

6.00

Area (m2)

76.46

4

400

Apartment

0.16

Moroccan house

0.72

Villa (detached house)

0.01

Hut

0.11

21

For more details see Johansson, P-O. (1987).

22

A typical Moroccan house is a kind of chain house without garden having maximum of three floors.

20

1.6.2 Estimation results

In order to identify the relationship between the value of the house and its attributes two functional forms are used: The linear (Model 1) and the expanded23 (model 2). The log linear model could not be used since most of the explanatory variables include zero in the range of values. The results are presented in table 1.8. All variables are highly significant. The variable number of rooms is not introduced in the models in order to avoid multicolinearity problems since this variable is highly correlated with the house’s size in m2. Moreover, the variable income is not considered since the inclusion of this variable in hedonic price studies is questionable; income is not an attribute of the house, but of the occupants24.

In both models the variable distance to CBD is positive and highly significant and we use it here as a surrogate for the environmental factor, since consumers would prefer to live far away from the CBD where air pollution and noise are less. Of course there may be other factors affecting preferences of location but typically people prefer to live in the center to avoid transport and to be close to culture etc. Thus our method should imply an underestimating of the real WTP to reduce air pollution.

For the dummies (apartments and Moroccan houses25), a general interpretation would be that an increase in the demand for these kinds of housing would lead to higher rents (or property values). For the surface area, the result is quite intuitive as bigger houses have a higher value.

Since our main interest when using hedonic prices is to assess how property values would change as environmental factors are modifiedand since we want to compare the results to those obtained using CVM, we double the variable distance to CBD by 50% and keep the other dependent

23 24

In this model the dependent variable is in log form while the independents are not. See Ball, M.J. (1973) for more details.

25

Villas (detached houses) and huts are not included in the estimation since the villas constitute only 1 percent of houses and huts are not permanent housing.

21

variables constant in the models. This would be an approximation of the mean willingness to pay to reduce air pollution in Rabat-Salé by 50% ( if pollution is linear in distance to CBD).

Using model 1 and doubling the distance to CBD in the regression gives a marginal effect equal to 130 DH. If we take the difference between this value and the marginal effect when distance to CBD is not modified the result equals 65 DH, which would be the mean willingness to pay to reduce air pollution by 50%. This result is in the same range as those estimated in the CVM study. Accordingly, the results for model 2 suggest a mean willingness to pay equal to 82 DH26.

Since using distance to CBD as a surrogate factor to indirectly assess the changes in property value, to our knowledge, has never been used in hedonic price studies these results should be interpreted with caution.

Table 1.8 Estimation results.

Intercept

Distance to CBD (km)

Area (m2)

Apartment

Moroccan house

Adj. R2

1.7

Model 1

Model 2

-2.24

5.30

(-2.64)

(51.91)

65.19

0.16

(4.53)

(4.74)

9.61

0.009

(18.10)

(13.97)

566

0.81

(6.68)

(7.96)

294

0.48

(4.43)

(5.95)

0.53

0.45

Conclusion

This chapter presents several pieces of evidence showing that interest exists for reducing air pollution problems in Morocco. Variables such as being young, having a higher income, being a 26

While the functional form used for the estimation is: ln rent = "+$ distance to CBD +...+,, the marginal effect is defined in the case of the distance to CBD as: $R where R is the mean rent.

22

man, having some degree of education and being aware of environmental problems have a higher impact on WTP.

Despite the fact that in the iterative bidding method 70 percent of the respondents have a positive WTP while in the dichotomous choice the proportion of Yes-answers is only 56 percent, the means and medians of the two methods are similar. The willingness to pay per month to reduce air pollution using CVM in Rabat-Salé is on average equal to USD 8. The yearly total benefit of reducing emissions range between USD 57 million using the iterative choice method and USD 59 million when using the simplified parametric approach. That is roughly a yearly total benefit of USD 60 million to reduce emissions by 50 percent in Rabat-Salé.

As concerning the case of hedonic prices where distance to CBD is used as a proxy for environmental factors, the mean willingness to pay to reduce air pollution by 50% using two different functional forms is again quite similar to that obtained using CVM.

However, the estimated values should be regarded as approximations because they are not only contingent upon the hypothetical market scenario presented to the respondents in the case of CVM and upon the approximations in the hedonic price´s case, but also upon the 50% reduction which is a simplified number we used du to lack of emissions data in Rabat-Salé.

23

Appendix 1 Estimation of Tobit and Heckit

Variable

Tobit

T static

Heckit

T static

Intercept

-64.91

(-1.93)

-21.60

(-0.19)

Age

-0.88

(-1.77)

-0.59

(-0.56)

Number of children

3.44

(1.49)

4.07

(0.98)

Household income

0.01

(5.04)

0.01

(4.17)

Family size

-1.06

(-0.58)

-3.85

(-1.21)

Gender (man=1)

17.43

(1.95)

25.01

(1.68)

Importance of air pollution

7.52

(0.88)

14.87

(1.27)

Literacy

24.93

(2.29)

22.97

(0.83)

Environmental awarness

50.42

(2.33)

28.81

(0.45)

Marital status

18.7

(1.79)

8.28

(0.32)

Respiratory disease

6.3

(0.58)

3.69

(0.20)

Non importance of environment

-1.84

(-0.17)

-3.29

(-0.20)

Importance of air pollution

-6.85

(-0.67)

-14.47

(-0.90)

Iterest in environmental problems

9.39

(0.94)

-0.41

(-0.02)

Authorities’ indifference

18.5

(-2.24)

-20.88

(-1.16)

Car ownership

7.64

(0.71)

0.39

(1.14)

Starting bid

0.1

(0.52)

0.39

(0.01)

46.97

(0.56)

Lambda N=382

24

Appendix 2 ( The original questionnaire is in french) I'm a researcher working at the University of Göteborg in Sweden and I'm interested in environmental problems. You have been chosen, through random sampling, as one of the persons to participate in a survey regarding the environmental contamination in Rabat-Salé. Your answers are voluntary and will be kept strictly confidential.

Before going into the explanation of the aim of the study I would like to ask you the following questions:

1.

Do you think that Rabat-Salé suffer from some environmental problems?

Yes

No

If the answer is yes, what are then these problems is order of priority?

a.

b.

c.

d.

Before going into the questionnaire we will give you a brief introduction about the health-contamination relationship, and about the purpose of this study.

Thank you for your cooperation.

25

INTRODUCTION

Rabat-Salé are one of the most polluted cities in Morocco. In the center of the cities high concentrations of toxic gases such as carbon monoxide, sulfur oxide, nitrogen oxides, hydrocarbons, lead, suspended particles etc. can be observed. These gases are emitted mainly from two sources; public and private transportation (approximately 90%). The high concentration of these gases radically affect the ecosystem, the environment and most importantly, human health, causing effects such as: bronchitis, allergies, cardiac and cerebral damage, reduction in the pulmonary function, fatigue, headache etc.

PURPOSE OF THE SURVEY

The demand for consumer goods is generally regulated through the price of the good. Public goods such as recreational parks and the air we breathe, are however, goods whose benefits cannot exclude anyone. For this reason setting a price on these goods is more complex.

The deterioration of environmental public goods demands the application of measures to repair this deterioration. The application of these measures, inevitably, gives rise to a cost, which directly or indirectly has to be paid by all of us. The atmospheric contamination existing today in our capital which indirectly excludes many persons from the essential right to breathe clean air constitutes an example. In order to compare the social costs of air contamination plus the required costs to improve air quality with the benefits that clean air gives us, it is necessary to know the willingness to pay for this good. This survey intends to evaluate the willingness to pay and its relation to the total cost provided by the air contamination. You, by answering this questionnaire can make it possible.

QUESTIONNAIRE

The questionnaire is divided into two sections, A and B. Questions belonging to section A are directly related to the problem of atmospheric contamination. Section B contains questions related to personal data such as profession, interests. etc.

26

SECTION A

2.

What degree of importance do you place on the problem of air pollution compared to other problems such as delinquency, unemployment, inflation and water contamination?

3.

1.

Critical

2.

Very serious

3.

Serious

4.

Less serious

5.

Not important

Rank each one in order of priority (1 being the highest priority), the environmental problem that you find of most concern. 1 2 3 4

4.

5.

1.

The extinction of marine species

2.

Water contamination

3.

Air pollution

4.

Urban noise

Do you consider yourself as a person:

1.

Very interested in the environment?

2.

Somewhat interested in the environment?

3.

Less interested in the environment?

4.

Indifferent?

Do you think that leaving a better environment to future generations is something:

1.

Very important?

2.

Important?

3.

Rather important?

4.

Not important at all?

27

6.

Do you, or any other member of your family work actively in any environmental organization ?

1.

Yes -----> Which?_______________________________ 2.

7.

8.

9.

No

Do you think that atmospheric contamination is an issue that concerns:

1.

The authorities?

2.

The decontamination commission?

3.

Every citizen?

Concerning the atmospheric contamination in Rabat/Salé, do you find that the authorities have given:

1.

A lot of attention to the problem?

2.

Some attention to the problem?

3.

Not too much attention to the problem?

4.

No attention at all?

In other large cities around the world with similar atmospheric contamination problems, several measures have been applied in order to solve the problem. Which of the following measures do you think should be applied in Rabat/Salé? (1 being the most important).

1.

A fuel tax

2.

Hard regulation of traffic

3.

The creation of non-traffic areas

4.

The installation of catalytic converters on all cars

5.

The use of green buses (gas combustion buses)

6.

An improvement in road infrastructure

7.

A road toll for entering downtown Rabat-Salé

8.

The building of lanes for bicycles

9.

Other, which?

10.

What form of transportation do you usually use?

28

11.

12.

1.

Bus

2.

Taxi

3.

Own car

4.

Other, which?

How do you find the public transportation system in Rabat/Salé?

1.

Very efficient

2.

Efficient

3.

Inefficient

If the public transportation system in Rabat/Salé worked efficiently, would you be willing to change your way of traveling?

13.

1.

Yes, I would use public transportation only

2.

Yes, I would decrease the use of my own car and increase the use of public transportation.

3.

No, I would continue using my own car

There are several measures that could be taken in order to improve the level of air quality in Rabat/Salé. Among possible measures are: the installation of catalytic converters on all gasoline cars built 1990 and later, the creation of non-traffic areas, the elaboration of gasoline without lead, the use of green buses, improved road infrastructure etc. As mentioned before, the applications of these measures cause a cost which directly or indirectly will be paid through all of us. This payment could be through: more expensive cars, increased fuel (gasoline and diesel) prices and public transport fares. Suppose the authorities presented a program which would decrease the level of atmospheric contamination by 50%. Would you be willing to contribute (x) Dirhams per month in order to cover in part the cost of the program?

14.

1.

Yes ---> go to 14a

2.

No ---> go to 14b

Would you be willing to contribute with a sum of:

a.

1½x=

Dirhams

29

15.

16.

1.

Yes ---> go to 15a

2.

No ---> go to 16

b

¾x=

1.

Yes ---> go to 16

2.

No ---> go to 15b

Dirhams

Would you be willing to contribute a sum of: a.

2x=

Dirhams

1.

Yes ---> go to 16

2.

No ---> go to 16

b.

x/2=

1.

Yes ---> go to 16

2.

No ---> go to 16

Dirhams

How much would you maximally be willing to contribute per month? __________ Dirhams

SECTION B

17.

Sex

1.

Male

2.

Female

18.

Age

___________

19.

What degree of education do you have?

30

1. Primary incomplete 2. Primary completed 3. Secondary incomplete 4. Secondary completed 5. Technical professional incomplete 6. Technical professional completed 7. University incomplete 8. University completed 9. Post graduate 19*

Social Science or natural science?................................................

20.

Profession or activity ___________________

21.

Marital status

1.

Married

2.

Divorced

3.

Separated

4.

Widowed

5.

Single

22.

Number of children

23.

Number of household members

24.

Do you or any other member of your family suffer from one or more of the following diseases?

1.

Pollen allergy

2.

Bronchitis

3.

Asthma

4.

Some other disease related to the atmospheric contamination, Which? ______________

25.

Monthly income (approximate) in Dirhams_____________

31

26.

27.

How often do you and your family visit recreation areas such as parks, forests, etc.?

1.

Several times a week

2.

Once a week

3.

Once a month

4.

Once or twice a year

5.

Never

If you and your family visit recreational areas, which type of transportation do you use on this occasion?

1.

Bus

2.

Own car

3.

Taxi

4.

Other, which? ___________________

32

References:

Amemiya,T.,(1973) "Regression Analysis When the Dependent Variable Is Truncated Normal", Econometrica No 41 Ayer, M , Brunk, H.D., Ewing, G.M. & Silverman, E. (1955) "An Empirical Distribution Function for Sampling with Incomplete Information." Annals of Mathematical Statistics, 2 6 , p.641-647. Ball, M. J., (1973) "Recent empirical work on the determinants of relative house prices", Urban Studies, 10, 210-233. Barllow, R.E., Bartholomew, D.J., Bremner, J.M., Brunk. H.D. (1972) " Statistical Inference Under Order Restrictions" John Wiley & Sons. Bishop, R.C. & Heberlein, T.A. (1979) "Measuring values of extra-market goods. Are indirect methods biased?" American Journal of Agricultural Economics, 61. Bourbonnais, R. (1993) "Économétrie" Dunod, Paris 1993. Boyle, K.J. & Bishop, R.C. (1988) "Welfare measurements using contingent valuation: A comparison of techniques." American Journal of Agricultural Economics, No.1. Cameron, T.A. & James, M.D. (1987) "Efficient Estimation Methods For "Closed-Ended" Contingent Valuation Surveys". The Review of Economics and Statistics, No. 2. Diamond, P.A. & Hausman, J.A. (1994) "Contingent valuation: Is some number better than no number?", Journal of economic perspectives, Vol. 8, No. 4. Hanemann, W.M. (1984) "Welfare evaluations in contingent valuation experiments with discrete responses". American Journal of Agricultural Economics, No. 66. Hanemann, W.M. (1994) "Contingent Valuation and Economics". Department of Agricultural and Resource Economics and Policy. Working Paper No. 697. Hanemann, W. M (1994) "Valuing the environment through the contingent valuation". Journal of economic perspectives, Vol. 8, No. 4. Hanemann, W. M. & Kanninen, B., (1996) "The statistical analysis of discrete-response CV data" Forthcoming in Bateman et al "Valuing environmental preferences: Theory and practice of contingent valuation methods in the US, EC and developing countries", Oxford University Press.

33

Johansson, P.-O. (1987) "The Economic Theory and Measurement of Environmental Benefits". Cambridge University Press, Cambridge. Kealy, M.J., Dovidio, J.F. & Rochel, M.L. (1988) "Accuracy in valuation is a matter of degree" Land Economics. Kmenta, J. (1986) "Elements of Econometrics" Second edition. Macmilan Publishing Company. Kriström, B.(1990) "Valuing Environmental Benefits using the Contingent Valuation Methods An Econometric Analysis". Ph.D. thesis, Umeå Economic Studies, No 219, University of Umeå. Kriström,B. (1995) "Spike Models in Contingent Valuation: Theory and Illustrations". Working report 210, 1995. The Swedish University of Agriculture. Kriström, B & Riera, P (1995) "Is the income elasticity of environmental improvements less than one?", Environmental and Resource Economics. Loomis, J.B. (1987) "Expanding contingent value sample estimates to aggregate benefit estimates", Land Economics. McFadden, D.L. & Leonad, G.K. (1992) " Issues in the Contingent Valuation of Environmental Goods: Methodologies for Data Collection and Analysis". Paper presented at the Symposium "Contingent Valuation: A Critical Assessment". Cambridge Economics, Inc. Washington , D.C. April 2 and 3, 1992. Mitchell, R.C. & Carson, R.T. (1989) " Using Surveys to Value Public Goods: The Contingent Valuation Method". Washington D.C. Resources for the Future. Randall, A. & Stoll, J. R., (1983) "Existence value in a total valuation framework". In R. D. Rowe, R.D & L.G.Chestnut (eds.) (1983), "managing Air Quality and Scenic Resources at National Parks and Wilderness", Westview Press, Boulder, Colorado. Rowe, R. D, D’Arge, R. C.& Brookshire, D. S, (1980) "An Experiment on the Economic Value of Visibility", Journal of Environmental Economics and Management, No 7 Sellar, C., Stoll, JR & Chavas, J.-P. (1985) "Validation of empirical measudres of welfare change: A comparision of nonmarket techniques." Land Economics. Scheaffer, R.L et al. (1979) "Elementary Survey Sampling". Duxbury Press. North Scituate, Massachusetts. Shechter, M. & Kim, M. (1991) "Valuation of Pollution Abatement Benefits: Direct and Indirect Measurement". Journal of Urban Economics, No. 30.

34

Vainio, M., (1995) "Traffic noise and air pollution, Valuation of externalities with hedonic price and contingent valuation methods" Helsinki school of economics and business administration Varian, H. R. (1992) "Microeconomic analysis" Norton International Student Edition.

35