Environ. Sci. Technol. 2000, 34, 1456-1461
Evaluating Health and Environmental Impacts of Pesticide Use: Implications for the Design of Ecolabels and Pesticide Taxes† S U S A N A M O U R A T O , * ,‡,§ ECE OZDEMIROGLU,| AND V I V I E N F O S T E R §,⊥ T. H. Huxley School of Environment, Earth Sciences and Engineering, Imperial College of Science, Technology and Medicine, 48 Prince’s Gardens, London SW7 2PE, U.K., Economics for the Environment Consultancy Limited, 16 Percy Street, London W1P9FD, U.K., The World Bank, 1818 H Street, Washington, D.C. 20433, and Centre for Social and Economic Research on the Global Environment, University College London, Gower Street, London WC1E6BT, U.K.
This paper estimates the economic impacts of pesticide use on human health and on the environment to gather information on the structure of a possible pesticide tax and on the design of an “environmentally friendly” bread product. The relative importance of these different impacts is determined by what individuals are prepared to pay to avoid a case of human ill health and a unit of environmental damage, measured by bird species in decline. Willingness to pay is estimated using a contingent ranking approach, a variant of the standard contingent valuation method, which is capable of tackling the multidimensional effects associated with pesticide applications. The results suggest that consumers would be willing to pay substantial price markups for environmentally friendly bread loaves and, consequently, that a case could be made for a substantial pesticide tax, preferably differentiated byproduct type. It is also shown that individuals are on average only willing to tolerate between 7-8 cases of human illness to save an entire bird species.
Introduction The term “pesticides” embraces an enormous diversity of products that are used in a number of different activities. These include agriculture, amateur gardening, woodworm treatment, and public health applications. The range of damages associated with the application of these products across environmental media and different receptors is equally great, providing a particularly complex example of multidimensional environmental impacts. Loss of aquatic and terrestrial biodiversity, contamination of groundwater and agricultural produce, and poisoning of agricultural workers are among the potential consequences of pesticide use in agriculture alone. Moreover, the choice between alternative pesticide products implies a tradeoff between different types of damages †
Part of the special issue on Economic Valuation. * Corresponding author telephone: +(0)20 75949314; fax: +(0)20 75949336; e-mail:
[email protected]. ‡ Imperial College of Science, Technology and Medicine. § University College London. | Economics for the Environment Consultancy Limited. ⊥ World Bank. 1456
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 8, 2000
as agrochemical products affect the environment in disproportionate ways. While some products may be more benign in terms of certain environmental impacts, they may simultaneously inflict more damage in other areas. Thus, it is necessary to evaluate the benefits of an improvement in environmental performance along one dimension against the costs of deterioration along another. In such situations, where damages are multidimensional and tradeoffs between them are of particular interest, using the standard contingent valuation (CV) approach to value environmental impacts raises some concerns, as complex and burdensome questionnaire designs would be needed. [See ref 1 for a review of the CV method and ref 2 for an overview of the controversies surrounding the use of this technique.] As a result, valuation practitioners are increasingly developing a predilection for alternative stated-preference techniques, such as the contingent ranking method, that provide a natural way of analyzing environmental multidimensionality. Introduced by Beggs et al. in 1981 (3), the contingent ranking method has been applied to a wide range of environmental issues in recent years. These include forest landscape changes (4), water quality improvements (5), beach characteristics (6), and competing principles of environmental equity (7). This paper develops an application of the contingent ranking methodology to the economic valuation of human health and biodiversity impacts from pesticide applications to wheat production in the U.K. Disaggregated monetary estimates of the damages associated with pesticide use have a great policy application potential. Two of these applications are the focus of this paper. First, the monetary estimates of environmental damage associated with pesticides could be used as a basis to introduce price differentials in products according to the level or type of pesticides used in their production process. Bread is the particular product considered in this paper. Product labeling can be used to introduce differentiation between products purely on the basis of their production processes. For example, through labeling, consumers have been able to view “dolphin friendly” tuna differently from tuna harvested by conventional methods, even though the physical characteristics of the two products in the supermarket shelves are indistinguishable. Following similar reasoning, people may be willing to pay more for agricultural products whose production processes are environmentally less damaging. The growing market for organic produce is exemplary in this context. Second, the damage figures estimated in this paper also provide valuable information about the design of a pesticide tax. When the production activity creates costs, which are not borne by producers but by third parties (external costs), market prices do not incorporate them. An obvious example is pollution damage from pesticide use. Thus, buyers do not pay the full costs to society of using pesticides; therefore, too many pesticides are produced and sold. This inefficiency would be removed if suppliers and buyers considered the social costs of pesticide use instead of internal costs only, where social costs are the sum of internal and external costs. This is the idea upon which the use of environmental taxes is based: the tax is set equal to the value of the external costs and is added to the price of the good. This way, the sum of tax and the price of the good will equal the social cost of production at which point the market can function efficiently. Contingent ranking can be used to express the various external costs of pesticide use in terms of money in order to 10.1021/es990732v CCC: $19.00
2000 American Chemical Society Published on Web 03/22/2000
TABLE 1. Sample Contingent Ranking Question attributes
process Aa
process Bb
process C
process D
price of bread health effects on general public biodiversity effects on farmland birds
60 pence per loaf 100 cases of ill health per year 9 bird species in decline
85 pence per loaf 40 cases of ill health per year 2 bird species in decline
85 pence per loaf 40 cases of ill health per year 5 bird species in decline
£1.15 per loaf 60 cases of ill health per year 2 bird species in decline
a
Current technology for wheat cultivation.
b
Alternative environmentally friendly options for wheat cultivation
subsequently calculate the amount of a (possibly differentiated) tax that can be added to the price of pesticides.
Methods Contingent Ranking Method. The contingent ranking method (3) is one of a number of survey-based techniques originally designed by marketing practitioners to isolate the value of individual product characteristics typically supplied in combination with one another. In a contingent ranking experiment, respondents are presented with a number of alternatives, each one consisting of a number of attributes that are offered at different levels across alternatives. A price or cost variable is typically one of the attributes. Respondents are then asked to rank the alternatives according to their preferences or simply to choose their most preferred alternative. From the ordinal rankings, the monetary value associated with each attribute [i.e., the willingness to pay (WTP) measure] can be indirectly calculated. This technique therefore avoids the need for a direct and explicit elicitation of respondent’s WTP. The conceptual economic framework for contingent ranking lies in Lancaster’s (8) work which assumes that consumers’ preferences for goods can be decomposed into preferences for its composing characteristics. [A detailed overview of this and other similar statedpreference approaches involving choices between alternatives can be found in refs 9 and 10.] During the month of June 1996, a contingent ranking survey was undertaken in various locations across the U.K. It was based on face-to-face interviews, performed by a market research company, on a sample of 504 people selected to be representative of the U.K. population. The questionnaire had three sections. The first section contained a number of attitudinal and lifestyle questions that introduced the subject of pesticides. The second contained an explanation of the environmental impacts of pesticide use on wheat production, a description of each of the attributes considered, and the contingent ranking valuation exercise. The final section requested information on a variety of socioeconomic characteristics. [The implementation of the final questionnaire was preceded by a series of in-depth one-to-one pilot interviews and by two field pretests that served to test the wording and order of the questions, the information provided on the impacts of pesticide use, and the attribute descriptions and levels used in the ranking cards.] In the contingent ranking exercise, respondents were asked to rank, according to their preferences, four hypothetical bread loaves differentiated in terms of price and ecolabels that identified the environmental impacts generated by the underlying production processes and identical in all other aspects (such as appearance and taste). As noted above, the range of environmental impacts associated with pesticide use is potentially very large but can be broadly grouped into two categories: human health and biodiversity. Indicator variables were chosen to quantify each of these two dimensions in the survey and were selected to represent, as accurately as possible, the main areas of welldocumented environmental damages for the U.K. Thus, the impact on human health was measured in terms of “cases of illness as a result of field exposure to pesticides during
cultivation”, while the impact on biodiversity was measured in terms of the “number of farmland bird species in a state of serious long-term decline as a result of pesticide use in cereal cultivation”. [More precisely, the questionnaire provided the following information about the bread attributes considered: “There are 100 cases per year where members of the public suffer ill health as a result of accidental contact with pesticides being applied to neighboring farmland. The symptoms suffered include nausea, dizziness, and irritations to skin and eyes. Hospitalization is sometimes, though not always, required”. “Nine species of British farmland birds (out of a total of 40 species) have experienced a decline in numbers of about 50% during the last 10 years. Pesticide use is thought to be an important factor in this decline. The affected species are illustrated on this card (linnet, tree sparrow, bullfinch, reed bunting, spotted flycatcher, song thrush, skylark, corn bunting, and grey partridge)”.] Each bread attribute (price, ill health, and bird species in decline) was offered at three different levels: price varied between 60, 85, and 115 pence; [as of June 1999, £1 ) U.S.$1.6 and £1 ) 100 pence] ill health cases varied between 100, 60, and 40; while bird species in decline varied between 9, 5, and 2. Currently, a standard loaf of bread (the baseline) is priced at 60 pence, and the pesticides used in its production are estimated to cause some 100 cases of ill health and to contribute to the decline of 9 bird species. The combination of three attributes at three different levels creates a total of 27 possible permutations of the hypothetical bread product. Given the prohibitive complexity of ranking 27 different bread loaves, the experimental design literature has developed procedures for selecting a subset of alternatives that still contain sufficient information to identify the primary effect of each attribute level on the ranking decision (11, 12). As a result of this process, each respondent was asked to conduct the ranking exercise with three different sets of four hypothetical loaves each. All ranking sets included the standard baseline loaf, which acted as a common reference point and, more importantly, gave respondents the opportunity to express a zero WTP as their most preferred alternative. A sample ranking question is reproduced in Table 1. A more detailed discussion of the experimental design is available in ref 13. Theoretical Framework. The random utility model (14) provides the economic theory framework for analyzing the data from the ranking exercise. According to this framework, respondents will select the bread product that maximizes their personal utility/satisfaction. Since the researcher does not observe all the determinants of individual choice, the utility function (U) for each respondent i can be decomposed into two parts: a deterministic element, which is a linear index of the attributes (X) of the j different alternative hypothetical loaves in the choice set; and a stochastic element (e), which represents unobservable influences on individual choice. This specification is shown in
Uij ) bXij + eij VOL. 34, NO. 8, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
(1) 9
1457
Thus, the probability that any particular respondent prefers option g in the choice set to any alternative option h can be expressed as the probability that the utility associated with option g exceeds the utility associated with all other options, as stated in
P[(Uig > Uih) ∀ h * g] ) P[(bXig - bXih) > (eih - eig)] (2) Under the assumption of an independently and identically distributed random error with a Weibull distribution, the probability of any particular alternative g being chosen as the most preferred can be expressed in terms of the logistic distribution (14). This specification is known as the conditional logit model:
P(Uig > Uih ∀ h * g) )
exp(bXig)
∑
(3)
exp(bXij)
j
The parameters b of the utility function can be estimated by maximum likelihood procedures. Once the parameter estimates have been obtained, consumer WTP for the human health (H) and biodiversity (B) attributes of the ecological loaves is calculated as the marginal rate of substitution between each of these attributes and the money price of bread (P), as illustrated in
Uij ) bpPij + bhHij + bbBij
(4)
WTPH )
dUij/dHij bh ) dUij/dPij bp
(5)
WTPB )
dUij/dBij bb ) dUij/dPij bp
(6)
Limitations. One of the main limitations of the contingent ranking technique is the cognitive difficulty associated with making choices between complex bundles with many attributes and levels. This burden escalates with the number of attributes used and the number of alternatives presented to each individual. Previous research in the marketing literature (15, 16) found significant differences in the preference structure implicit across ranks, with information on the most preferred alternative alone invariably being found to be more reliable than information on full rankings. This may be indicative of increasing respondent fatigue with the depth of the ranking task and consequent reliance on rules of thumb and simplifying heuristics to aid the ranking process, which may conflict with economic principles. These findings raise questions as to whether respondents are genuinely able to provide full rankings of complex alternatives. Further analysis shows that also in the context of this survey a substantial proportion of respondents found difficulty in providing coherent responses to the full ranking questions, implying that a model based on full rank information, while more efficient, may also be misspecified (13). Hence, to minimize these biases, a simpler model, the one described in eq 3, was adopted for the purposes of this paper. Note that this model does not make use of the full ranking information elicited in the survey but utilizes only information on the most preferred alternative out of each group of four that were presented to respondents. Another problem associated with ranking methods lies in the fact that the choices are dependent on the specific attributes presented in the survey and implicitly or explicitly assume that all other attributes are held constant at their current levels. In reality, it is difficult to control for the way these other missing attributes are taken into account by 1458
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 8, 2000
TABLE 2. Conditional Logit Model model coefficients price human health biodiversity log likelihood no. of observations WTP for human health WTP for biodiversity a
Standard error: 0.062.
b
t ratios
-3.52 -9.44 -2.39 -9.70 -1.84 -8.25 -621.94 501 0.68 pence per loafa 5.25 pence per loafb Standard error: 0.574.
respondents and even more difficult to predict how responses would have changed had they been included in the choice sets. This is especially problematic if attributes that respondents feel strongly about are left out of the ranking set. In the context of the pesticides survey, this problem was addressed by informing respondents that the two pesticide impacts presented in the cards were the major pesticide problems in the U.K. as documented by the existing evidence. Finally, the contingent ranking method does not allow respondents to be indifferent between the alternatives presented in the ranking sets, i.e., no ties are allowed between options. This may introduce bias in respondents stated choices and is more likely to occur when ranking least preferred alternatives, as respondents may find it difficult to identify which option they dislike the least. Hence, using only information on respondents most preferred alternative rather than full ranking information is also thought to minimize the occurrence of this bias.
Results and Discussion Table 2 presents the results of estimating the conditional logit model on the pesticides data. The model coefficients all have the expected negative signs (cost, ill health, and bird species loss all have negative impacts upon consumer welfare) and are highly statistically significant. Consumers were found to be prepared to accept an average markup of 0.7 pence per loaf of bread to avoid a case of human illness and 5.3 pence per loaf to protect one species of farmland bird. Hence, on average, respondents are willing to tolerate the decline of one farmland bird species to prevent 7-8 additional cases of minor human illness, which is indicative of the importance attached to the preservation of human health. These results were shown to perform well in terms of standard validity tests. In other words, WTP was found to vary systematically with the socioeconomic characteristics of respondents in ways that might be predicted by economic theory, to be sensitive to the scope of the damage, and to be broadly consistent with estimates obtained in other studies (13). As noted earlier, the economic estimates of the social costs of pesticide applications can be put to use in a range of different applications. Perhaps the most obvious of these is the design of a bread product which could be marketed under an ecolabel. Here the model can be used to inform as to the balance of price, human health, and biodiversity impacts that is most likely to maximize the market share of such a loaf. But the relative weighting of health and biodiversity impacts obtained from the model also has wider uses. In particular, in providing information about the choice between alternative agrochemical products that differ in their range of environmental damages. Indeed, such weights could ultimately be used in the design of a differentiated pesticide tax. Both of these issues will be considered. Design of a “Green” Loaf. The model estimated above predicts that, on average, 80% of consumers will choose one of the alternative environmentally friendly loaves of bread
TABLE 3. Implicit Reservation Prices for a Range of Alternative Green Loaves no. of cases of human ill health per year 100 80 60 40 20 0
no. of farmland bird species in decline 9 7 5 2 0 a
TABLE 4. Price of Standard and Organic Bread Loaves (January 2000)a
naa £0.70 £0.81 £0.97 £1.07
£0.74 £0.84 £0.95 £1.10 £1.21
£0.87 £0.98 £1.08 £1.24 £1.34
£1.01 £1.11 £1.22 £1.37 £1.48
£1.14 £1.25 £1.35 £1.51 £1.61
£1.28 £1.38 £1.50 £1.65 £1.75
na, not applicable.
in preference to the conventional baseline loaf. This result can be compared with attitudinal questions asked at the outset of the questionnaire enquiring how regularly respondents purchased environmentally friendly products. In answer to this question, 26% of the sample claimed to purchase such products regularly, while 51% answered that they only sometimes did so. Overall, this comes to a total of 77%, which is very similar to the predictions made by the model. However, the comparison suggests that the predicted probabilities are more likely to be indicative of occasional rather than habitual purchase of the green loaf. The predicted probability of choosing any particular type of green loaf varies between 9% and 63% depending on the characteristics of the loaf, thereby indicating the importance of obtaining the right combination of price and environmental benefits. To clarify the nature of these relationships, it is useful to calculate the average consumer’s “reservation prices” for different types of green loaves. This is done by fixing the levels of the human health and biodiversity attributes and then raising the price variable until the linear utility score implied by the model (eq 4) is equal to that associated with the standard baseline loaf. This point of indifference indicates the threshold price above which the average consumer would no longer choose to purchase the green loaf. Table 3 presents the implicit reservation prices for 35 possible permutations of the green loaf. The table can be used as a ready-reckoner to aid the design of the ideal green loaf by bringing to bear information about the cost of securing such environmental improvements. Green loaves whose cost exceeds the associated reservation price are clearly infeasible. Among those which are feasible, the best (or utility maximizing) loaf can be identified as the one with the highest reservation price. The main finding is that consumers would on average be willing to pay up to £1.75 for a loaf that did absolutely no damage to human health and biodiversity. However, this estimate should be treated with caution since it involves linearly extrapolating values beyond the range of attributes considered in the survey, which are likely to be nonlinear. For the most environmentally friendly loaf actually presented in the survey (40 cases of human ill health per year and 2 bird species in decline), the reservation price is found to be £1.37. Even this is more than twice the price of the conventional baseline loaf. One way of testing the validity of the contingent ranking WTP predictions is by matching them up with values of real payments made in comparable circumstances (criterion validity). In the context of this application, a formal validity test is obviously not possible as it would imply observing the real prices of nonexistent hypothetical loaves, as described in the questionnaire. However, an informal comparison can be made between the estimated prices in Table 3 and the prices of organic bread loaves found in the U.K. market. Organic bread loaves are produced without artificial fertilizers, chemicals, pesticides, or genetic modification and can be considered a close comparator to the hypothetical green
price (£) product (800 g loaves) sunflower wholemeal granary rye white
standard loaves nab 0.38-0.66 0.59-0.75 na 0.17-0.75
organic loaves 0.85-1.45 0.59-1.45 1.45 1.09-1.65 0.59-0.85
a Sources: Organic Delivery, Tesco, Sainsbury, Safeway, Asda, Iceland. b na, not applicable.
loaves described in the survey. Table 4 presents the current prices of standard and organic 800-g bread loaves, as found in a number of U.K. supermarkets. Although large variations in price exist between stores, makes of bread, and bread types, most standard bread loaves currently cost about 60 pence, as in the time of the survey. In contrast, organic bread loaves of similar size are found to be significantly more expensive on average, although large variations of price also exist. Most stores explain the price differential as a premium paid to account for increased production costs due to more labor-intensive practices and possibly lower yields. A comparison between the figures in Tables 3 and 4 reveals that the implicit reservation prices estimated in the survey are close to the higher bracket of organic bread prices found in the market. However, the hypothetical loaves do not possess as many desirable attributes as organic loaves (e.g., nothing was mentioned about the lack of genetically modified components), which suggests that the estimated figures could be upward biased when compared with what people actually pay in practice. The implication is that the monetary values obtained in this simple survey would probably need to be somehow adjusted before being applied to answer policy questions. Illustrative Design of a Pesticide Tax. An interesting application of monetary measures of pesticide damage is to aid in the calculation of a pesticide tax. The level of a uniform pesticide tax could be calculated as the aggregate WTP to avoid pesticide damage (i.e., the external cost of pesticides) over a given time period, say 1 year, divided by the total volume of pesticide applications to cereal crops over the same period. To obtain aggregate WTP, the marginal WTP in terms of pence per loaf per unit of damage, calculated in a manner similar to this study, would be multiplied by the total environmental damage figures and by the aggregate number of loaves purchased every year, adjusted by the price elasticity of demand for bread. It should be noted that the contingent ranking survey summarized here was designed to uncover people’s tradeoffs between human health and environmental damage but was not explicitly designed for the purposes of a uniform tax calculation. That would have required larger sample sizes, more attributes reflecting other pesticide impacts, more than three levels of each type of environmental damage considered, a full physical damage assessment of pesticide impacts, and more information on the exact quantity of pesticides used in bread production in the U.K.. Unfortunately, none of this was possible within the constraints of this study. But despite these caveats, it is still of interest to illustrate how similar results could have been used to inform the design of a pesticide tax. For the purposes of this illustration, two simplifying assumptions are made: (a) that the total yield of cereals in the U.K. is used for the production of bread and (b) that the total damage per loaf of bread is assumed to be one. The first assumption ignores other uses and the import/export of VOL. 34, NO. 8, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1459
wheat. The second assumption ignores that there can be more or less than one unit of damage per loaf of bread and avoids dealing with the scientific uncertainty surrounding estimates of environmental damage. Given these simplifying assumptions, the WTP estimates reported above can be used to illustrate the calculation of the level of a pesticide tax to be applied uniformly per unit weight to all agrochemical products. First, the overall WTP per loaf for the protection of human health and biodiversity (£0.007 + £0.053 ) £0.06) would be aggregated over the 160 loaves purchased on average each year by the U.K.’s 20 million households. [The estimated number of loaves consumed under the new price was calculated using a price elasticity of demand of -0.09 (17).] This “aggregated marginal WTP” would then be divided by the total 15 million kg of pesticides used in cereal crops in the U.K. to obtain a uniform tax value of £12.59/kg of pesticide. Compared with a current average price of pesticides of the order of £20/kg, this would represent a tax rate of over 60%. Note that these calculations can only be regarded as illustrative in view of the limitations pointed out above. Moreover, it was noted that the damages associated with pesticide use are not confined to a single dimension but are rather multifarious in nature. Hence, a single damage figure is unlikely to be satisfactory as the basis for setting a uniform pesticide tax. Indeed, at first sight, it is quite surprising that, where pesticide taxes and charges have been introduced (e.g., Sweden and Denmark), the price has tended to affect all products equally with no attempt at differentiation. The problem with a uniform tax is that it does not provide any incentives for farmers to substitute toward less environmentally damaging agrochemicals and therefore does not necessarily guarantee environmental improvements. To differentiate a tax, it is necessary to have some means of determining the overall environmental damage associated with a particular product taking into account its full range of impacts. A number of studies have tried to rank pesticides according to the environmental and health risks posed. Typically, this is done simply by summing up the scientific damage scores for each dimension without any attempt to reflect the relative importance of different types of damages (18, 19). A preliminary attempt to apply weights to these scores was made by Higley and Wintersteen (20); however, a significant drawback of this study was the way in which it combined attitudinal rankings of the relative importance of each dimension with a very generalized contingent valuation question addressing WTP to avoid “low, moderate, or high levels of risk” from pesticide use. The present study, which was explicitly designed to make respondents consider the relative importance of different types of damages, provides precisely the kind of information that is required (i.e., a means of weighting the various dimensions of environmental impact) to show how such a uniform tax or charge could be differentiated to better reflect the relative environmental damages associated with different kinds of agrochemical products given administrative constraints. As an illustration, Table 5 provides health and biodiversity scores for a range of hypothetical agrochemical products. The aggregate damage scores obtained by taking a simple average of the two impacts are contrasted with those that result when the two impacts are weighted in accordance with the results of the contingent ranking survey. The illustrative figures show that weighting makes a material difference to the outcome by reversing the relative rankings of products B and C. This information can be used to design a differentiated pesticide tax. Normalizing the weighted score for each product against that for the most highly ranked product indicates the extent to which the tax level for that product would need to be marked up relative to that baseline. If the tax is set so that the mean level is equal to the value 1460
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 8, 2000
TABLE 5. Illustration of a Differentiated Pesticide Tax agrochemical A agrochemical B agrochemical C human health score biodiversity score simple score simple rank weighted scorea weighted ranka relative tax illustrative tax
100 2 51 third 20 third 1.20 £13.69/kg
50 9 30 first 18 second 1.11 £12.58/kg
85 7 46 second 16 first 1.00 £11.38/kg
a Based on model coefficients, these are 0.12 per unit health and 0.88 per unit biodiversity.
of £12.59 calculated above, this results in differentiated tax rates that range from £11.38 to £13.69/kg. The results of the valuation study reported here have been used to inform the design of a differential pesticide tax in the U.K., which is currently going through a consultation process (21). The tax is based on five hazard bands, differentiated on the basis of the ecological and human health effects of pesticides with higher rates for more hazardous products. A “composite” indicator was used to identify the five hazard bands for the pesticides covered. It included the following: human/mammalian health impacts, risk to aquatic species, risk to soil beneficials, risk to bees, risk to birds, and hazard posed to groundwater. The results of the contingent ranking exercise were used to weight the human health and bird indicators alongside other weighting schemes. The use of the monetary weights changed the scores somewhat but did not alter the bands into which pesticides fall by a significant amount.
Acknowledgments The research underlying this paper results from a study funded by the U.S. Environmental Protection Agency. We would like to thank David Pearce and Robert Tinch for their contributions to that study.
Literature Cited (1) Mitchell, R.; Carson, R. Using Surveys to Value Public Goods: The Contingent Valuation Method; John Hopkins Press: Baltimore, 1989. (2) Portney, P. J. Econ. Perspect. 1994, 8, 3-17. (3) Beggs, S.; Cardell, S.; Hausman, J. J. Econometrics 1981, 16, 1-19. (4) Garrod, G.; Willis, K. Ecol. Econ. 1997, 21, 45-61. (5) Smith, V.; Desvousges, W. Measuring Water Quality Benefits; Kluwer-Nijhoff: Boston, 1986. (6) Machado, F.; Mourato, S. CSERGE Working Paper GEC 99-09; University College London, University of East Anglia: Norwich, 1999. (7) Atkinson, G.; Machado, F.; Mourato, S. London School of Economics, mimeo. (8) Lancaster, K. J. Polit. Econ. 1966, 84, 132-157. (9) Ben-Akiva, M.; Lerman, S. R. Discrete Choice Analysis: Theory and Application to Travel Demand; MIT Press: Cambridge, MA, 1985. (10) Boxall, P. C.; Adamowicz, W. L.; Swait, J.; Williams, M.; Louviere, J. Ecol. Econ. 1994, 18 (3), 243-253. (11) McClean, R.; Anderson, V. Applied Factorial and Fractional Designs; Marcel Dekker: New York, 1984. (12) Louviere, J. J. Analyzing Decision-Making: Metric Conjoint Analysis; Quantitative Applications in the Social Sciences Series; Sage University Paper 67; Sage Publications: Newbury Park, CA, 1988. (13) Foster, V.; Mourato, S. J. Agric. Econ. 2000, 51 (1), 1-21. (14) McFadden, D. In Frontiers in Econometrics; Zarembka, P., Ed.; Academic Press: New York, 1973. (15) Ben-Akiva, M.; Morikawa, T.; Shiroishi, F. J. Bus. Res. 1991, 23, 253-268.
(16) Hausman, J. A.; Ruud, P. A. J. Econometrics 1987, 34, 83-104. (17) Household Food Consumption and Expenditure: Annual Report of the National Food Survey Committee; Ministry of Agriculture, Fisheries and Food; Her Majesty’s Stationery Office: London, 1989. (18) Reus, J. A.; Pak, G. A. Med. Faculty Landbouww 1993, 58 (2a), 249-255. (19) Penrose, L. J.; Thwaite, W. G.; Bower, C. C. Crop Protect. 1994, 13, 146-152. (20) Higley, L. G.; Wintersteen, W. K. Am. Entomol. 1992, 38, 34-39.
(21) Design of Tax/Charge Scheme: Final Report to the Department of the Environment, Transport and the Regions; ECOTEC, University of Hertfordshire, Central Science Laboratory, EFTEC, University of Newcastle upon Tyne; DETR: Birmingham, 1999.
Received for review June 30, 1999. Revised manuscript received January 31, 2000. Accepted February 2, 2000. ES990732V
VOL. 34, NO. 8, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1461
