Estimating Intergenerational Distribution Preferences Using Choice Modelling

Faculty of Business and Law SCHOOL OF ACCOUNTING, ECONOMICS AND FINANCE School Working Paper - Economic Series 2006 SWP 2006/26 ESTIMATING INTERGENERATIONAL DISTRIBUTION PREFERENCES USING CHOICE MODELLING Helen Scarborough & Jeff Bennett The working papers are a series of manuscripts in their draft form. Please do not quote without obtaining the author’s consent as these works are in their draft form. The views expressed in this paper are those of the author and not necessarily endorsed by the School. ESTIMATING INTERGENERATIONAL DISTRIBUTION PREFERENCES USING CHOICE MODELLING Helen Scarborough* School of Accounting Economics and Finance, Faculty of Business and Law Deakin University, Warrnambool, Vic. 3280 Australia email: scars@deakin.edu.au Jeff Bennett Crawford School of Economics and Government ANU College of Asia & the Pacific The Australian National University Canberra, ACT 0200 Australia email: Jeff.Bennett@anu.edu.au ABSTRACT Resource management decisions influence not only the output of the economy but also the distribution of utility between groups within the community. The theory of Cost Benefit Analysis provides a means of incorporating distributional changes into the decision making calculus through the application of distributional or welfare weights. This paper reports the results of research designed to estimate distributional weights suitable for inclusion in a Cost Benefit Analysis framework. The findings of a choice modelling experiment designed to estimate community preferences with respect to intergenerational utility distribution are presented. JEL classification codes: Q56, C35, D61 Keywords: Distributional weights, Cost benefit analysis, Intergenerational distribution, Choice modelling. *Corresponding author contact details Ph: +61 3 55633547, Fax: +61 3 55633320 Page 1 ESTIMATING INTERGENERATIONAL DISTRIBUTION PREFERENCES USING CHOICE MODELLING ABSTRACT Resource management decisions influence not only the output of the economy but also the distribution of utility between groups within the community. The theory of Cost Benefit Analysis provides a means of incorporating distributional changes into the decision making calculus through the application of distributional or welfare weights. This paper reports the results of research designed to estimate distributional weights suitable for inclusion in a Cost Benefit Analysis framework. The findings of a choice modelling experiment designed to estimate community preferences with respect to intergenerational utility distribution are presented. 1. INTRODUCTION Resource management policies have a range of distributional effects within the economy. It is unlikely that those benefiting from a specific policy change will be the same group as those who bear the cost. Assessment of the distributional impacts of policy alternatives can be based on criteria such as the economic status, ethnicity, age, geographical or temporal distribution of those who gain and those who lose. As awareness of the distributional impacts of policy change has heightened, debates have increasingly involved these distributional considerations (Serret and Johnstone 2006). Yet, the incorporation of distribution in policy analysis is difficult “as there is no commonly accepted definition of optimum equity: certainly nothing analogous to maximum net benefits from economic efficiency” (Sutherland 2006). Page 2 One method of including distributional considerations in policy analysis is through the application of distributional weights in a cost benefit analysis (CBA) setting. This theory is briefly outlined in section two. Johansson-Stenman (2005) illustrates a large range of cases where the application of distributional weights is (second-best) optimal thus reinforcing the argument that efficiency and distributional concerns must be analysed simultaneously. However, despite the well established welfare economic theoretical underpinnings of incorporating equity considerations in policy analysis, there has also been extensive debate in the literature regarding the efficacy of applying distributional weights. This debate was particularly active during the 1970s in the context of project appraisal by the World Bank (Dasgupta, Sen et al. 1972; Little and Mirrlees 1974; Squire and van der Tak 1975; Squire 1989). In practical terms, the approach to incorporating distributional weights in CBA varies. In most cases, the practice of the World Bank is not to apply explicit distributional weights in CBA (Little and Mirrlees 1994; World Bank 1996). In the UK, H.M. Treasury, has officially endorsed distributional weights in CBA as detailed in their Green Book (H.M.Treasury 2003). In Australia, the Commonwealth Department of Finance and Administration recommends that, as a general rule, distributional weights not be assigned “and that recommendations of cost-benefit analyses flag the need for distributional judgements to be made at the political level” (Department of Finance and Administration 2006). This raises the issue of the distinction between the application of explicit or implicit distributional weights. Adler and Possner (1999) suggest that at a national level in the US, if distributional weights are applied, the weighting appears to be made implicitly through the policy decision making process. Page 3 One of the impediments to the adoption of the application of explicit distributional weights has been the difficulty in estimating community preferences for the distribution of utility. Revealed preferences studies such as those by Basu (1980) have estimated preferences by policy makers based on the analysis of past decisions. Yet there is a paucity of knowledge of the distributional preferences of the community. In this paper, this limitation is addressed through the application of the stated choice method of choice modelling (CM) to the estimation of distributional preferences. CM has increasingly been recognized as a method of estimating relative values for non-marketed environmental attributes (Bennett and Blamey 2001). An illustration of the application of the CM methodology to the problem of estimating distributional weights is provided in section three. Rather than taking the conventional CM focus on utility estimation, the application reported here involves utility distributions between generations being used as policy change attributes. The context used is that of environmental policy development. The research is therefore particularly relevant to the sustainability debate, where sustainable development implies some general rule about maintaining the capability of future generations to achieve the same level of well-being as the current generation (Tacconi 2000). The results of the CM experiment, which are summarised in section four, suggest that the community holds a degree of altruism towards future generations. Discussion of these results in section five indicates that the weights estimated are within the range of recent speculation regarding intergenerational distributional preferences. The findings indicating community distributional preferences favouring the utility of future generations have significant natural resource management policy implications. Furthermore, the plausibility Page 4 of the results supports the potential of CM to estimate community distributional preferences. 2. DISTRIBUTIONAL WEIGHTING AND BENEFIT COST ANALYSIS Using a Bergson-Samuelson social welfare function (SWF), social welfare (Wj) representing the social preferences of respondent, j, can be expressed as; W j = U 1 ( x1 ) + U 2 ( x 2 ) + U 3 ( x3 ) + ... + U n ( x n ) [1] where Ui is the utility of the i=1…n people in society, with xi the quantity of goods consumed by individual i. The form of the SWF depends on whose preferences are being reflected, and it is often expressed as the views of parliament or social planners [see, for example, Mäler, 1985]. In this paper, we are interested in the SWFs of individuals within the general community. Changes in social welfare resulting from policy changes can be accounted for by acknowledging that for some individuals there may be positive or negative changes in utility as a result of the introduction of a specific policy. Aggregation across individuals in this form of the welfare function assumes that the marginal utility of consumption is equal for all individuals: additions to consumption are valued equally by each individual. Recognising that this may not be the case, distributional concerns can be allowed for by assigning distributional weights to the gains and losses to various individuals: W j = α 1jU 1 ( x1 ) + α 2jU 2 ( x2 ) + α 3jU 3 ( x3 ) + ... + α njU n ( xn ) [2] Page 5 where the distributional weights, held by person j, for each individual are given by the α i j ’s. For further elaboration on the derivation of distributional weights see Johansson (1987), Dreze and Stern (1987), Mäler (1985) or Layard and Glaister (1994). The distributional weights ( α i j ’s) that can be applied to the elements of a CBA are variously referred to as the marginal social utilities of income (Johansson 1993), the welfare weights (Dreze and Stern 1987), or the marginal social utilities (Boadway and Bruce 1984). These distributional weights are the products of two components: the change in social welfare if the utility of individual i increases marginally (∂W ∂U i ) .1 j ∂U i ) and the marginal utility of consumption of individual i, ( ∂xi ∂W j ∂U i αi = ⋅ ∂U i ∂xi j [3] The first component of the weight indicates how the person, j, whose social welfare preferences are being reflected, ranks the utility of individual i in their distributional preferences. For example, in the view of person j, is social welfare enhanced or diminished if the utility of a low income person is improved relative to a high income person? Examples of characteristics that may influence these perceptions include wealth, ethnicity, race, geography or generation. 1 Strictly speaking, this is the perception, of the person whose social welfare preferences are being reflected, of the marginal utility of consumption of other members of the community. Page 6 The second component of the weight reflects the assessment, by the person whose welfare preferences are being reflected, of how the well-being of individual i changes as a result of a change in consumption, often substituted by money as a numéraire. For example, is the view held that a dollar of benefit increases the utility of a low income person more than it would increase the utility of a high income person? Although this component of the weight is often referred to in terms of income, this does not necessarily need to be the case. For example, it could also be the marginal utility of an additional unit of an environmental good for individual i. Medin, Nyborg et.al. (2001) illustrate the sensitivity of distributional weights to the choice of numéraire. The distributional weights may be different for each individual, j, in society reflecting their distributional preferences. Assumptions regarding the first component of the distributional weight reflect varying theories of social justice. For example, in a Benthamite or utilitarian society ∂W j ∂U i = 1 for all, resulting in all changes in utility being treated equally. j Alternatively, in a Rawlsian society ∂W ∂U i = 0 for all, except the worst-off, reflecting Rawl’s view that welfare is maximised by seeking to maximise the utility of the least welloff individual. In practice, social justice preferences will most likely be in terms of groups within society who share common characteristics, rather than individuals. For this reason, the proceeding analysis is in terms of groups rather than individuals. Distributional weights are thus dependent on the impact of money, assuming this is the chosen numéraire, on the wellbeing of the group and the impact of the change in utility of a group on society’s total welfare. Page 7 There are few examples where distributional weights have been explicitly applied (Markandya 1998). In part, this may be due to the prospect of an efficiency cost arising from the incorporation of equity preferences in policy analysis. This case has been strongly argued by Harberger (1978) and Harberger and Jenkins (2002) and countered by authors such as Layard (1980), Dreze (1998) and Johansson-Stenman (2005). A further difficulty has been the question of whose social welfare preferences should be considered. A lack of knowledge of the community’s social welfare and distributional preferences and an inability to elicit and estimate distributional preferences has also contributed to the limited application of explicit distributional weights. There has been some work on estimating distributional weights with respect to income distribution (Cowell and Gardiner 1999) but even less in estimating weights that acknowledge distribution is impacted by a broader range of marginal utilities. To elicit community utility distribution preferences and hence a set of distributional weights ( α i j ), a CM experiment involving intergenerational utility redistribution arising from changing environmental policies was conducted. Environmental policies can affect the distribution of resources, both financial and environmental, between generations. Policy debates have increasingly involved generational distributional considerations. The Brundtland Commission (World Commission on Environment and Development 1987) defined sustainable development as “development that meets the needs of the present without compromising the ability of future generations to meet their own needs” and Pearce and Barbier (2000) stress the “fair treatment of future generations”. Arrow, Dasgupta, et.al. (2004) take sustainability to mean that intertemporal social welfare must not decrease over time. Page 8 These definitions highlight an anthropocentric policy focus that places emphasis on achieving a quality of life that can be maintained for future generations. It is impossible to know the exact conditions that will allow the certain existence of future generations. Consequently, intergenerational utility distribution also depends on the extent of altruism the current generations hold for future generations. Those advocating “environmental justice”, both in principle and in terms of practical political action, emphasise the distributional implications of environmental change (Agyeman, Bullard et al. 2003). They highlight the strong link between sustainability and social justice. Yet notions of social justice, both generally and with respect to sustainability and intergenerational equity, vary between individuals reflecting personal judgements regarding fairness. Broome (1995) suggests a search for a class of reasons, referred to as claims, why one group should be given priority over another. He argues that fairness is about mediating the claims of different people and requires that claims should be satisfied in proportion to their strength. The important aspect of this view of equity is the need to mediate claims as one of the paramount considerations for fairness because an aspect of justice is not just how a group fares in relation to their claims but also involves how they fare in relation to the rest of the claimants (Rescher 2002). In the context of sustainability, the claimants reflect varying generations and the question becomes one of weighing the gains and losses to different generations. Again, the application of the concepts so developed is limited because of the lack of information regarding the strength of intergenerational utility preferences. Page 9 Another key feature of the literature on fairness is that it is dependent on the actor and beliefs about what is fair are personal (Elster 1992). Elster suggests four groups of actors that can be useful for analysing distributive justice; individuals in the organisation that is charged with the allocative task, political actors, claimants and public opinion. While policy makers generally have the opportunity to express their justice principles, the public has limited forms of social choice in which to express their preferences. This is one of the strengths of using a stated preference technique to estimate community utility distribution preferences. 3. INTERGENERATIONAL DISTRIBUTION CM EXPERIMENT A CM experiment has been undertaken where, rather than estimating utility as a function of the attributes of goods consumed as in conventional applications, social welfare is the dependent variable and the utility levels of different groups, in this case generations, are the attributes that are varied. This application of CM addresses the question of the distributional effects of policies and the consequent social welfare outcomes of policy alternatives. Rather than applying the model to the estimation of individual well being and value in a dollar measure, the emphasis is on the estimation of the relative distributional preferences of respondents. It is the respondent’s conception of social welfare rather than utility that is being maximised in the choices being made. Choices between the distribution associated with the status quo and changes in policy resulting in distributional changes were presented to respondents. The attributes of the policy options that were varied were the levels of utility or well-being of particular groups Page 10 within society and the measure of interest is the willingness of respondents to trade-off a change in the utility of one group for a change in the utility of another group. Arrow (1963) suggests that people have two distinct personalities: their self-interested selves essentially disjoint from aspects of their ethical selves. Self-interested preferences guide day-to-day participation in the market economy while their ethical ones apply to participation in collective decision making. Nyborg (2000) formalises this distinction between “Homo economicus”, the individual maximising personal well-being, and “Homo politicus”, the individual expressing their social justice preferences. Focus on the behaviour of “homo politicus” allows for a sense of social justice that Musgrave and Musgrave (1989) argue is essential for the definition of a good society and the functioning of a democratic society. Broome (1995) describes this as a notion of communal good that is separate from the good of individuals. Hence, this CM experiment is aimed at eliciting respondents’ distributional preferences reflecting their social justice preferences. The ability of respondents to view policy in this manner is supported by a study of the equity considerations of the burden of meeting the costs of environmental policy by Atkinson, Machado et.al. (2000) who found the proposition that respondents significantly allowed their own position to influence their ranking of different options was not strongly supported. A degree of interpersonally comparable cardinal utility must be assumed so that respondents are able to make judgements about the well-being of other groups in society. In this application, respondents use their knowledge of the well-being of groups within society under the status quo policy. Therefore, decision-making is seen in a broader context of welfare maximization within a social structure rather than individuals maximising their utilities. Hence, each individual Page 11 has a personal view of social justice based on their distributional preferences. Respondents were encouraged to adopt this approach by the following introduction to the survey instrument: “Many environmental policies result in a transfer of both income and resources between generations. For example, some environmental policies are paid for by current taxpayers with the aim of improving the environment for future generations. We are interested in finding out what you think about the way these policies lead to gains for some generations and costs for other generations.” Hypothetical policies with generic labels (A, B, C etc) were used as the sources of distributional change for the CM choice sets in an attempt to ensure that values other than distribution preferences were not reflected in the respondent’s choices. This also encouraged the respondent to focus on their social justice preferences. This does not mean that respondents did not bring preconceived beliefs to the decision making process, rather that these beliefs are part of ethical preferences regarding social welfare. The attributes in this experiment were described in terms of the impact on the utilities of individuals from different generations resulting from the three hypothetical and generic policy options. Individuals with specific generational characteristics were used as proxies for the group described. Following Mackay (1997), a time span of 25 years was taken as a generation. The attributes and levels are described in Table 1. [Insert Table 1 about here] Page 12 The chosen design limited the choices to generations currently living to avoid time and discounting complications, acknowledging the trade-offs required when considering the cognitive demands placed on respondents. The total time period of the analysis could have been extended by increasing the number of attributes, however, this also would have increased the cognitive burden for respondents and there is likely to be a trade-off between the number of attributes and valid responses (Louviere, Hensher et al. 2000). The levels of the attributes were described in dollar terms. The dollar terms reflected the change in utility to the individual with the specific characteristic described by the attribute. Dollars were adopted as a metric with which respondents could associate. The main advantage with this numéraire is that dollars are a common and well understood metric to respondents. However, respondents were advised in the following way that the dollar values represented the general utility of the individuals, and should not be interpreted as financial wealth alone: “In this survey, dollars have been used to measure the gains and losses to different generations. The dollar amounts represent gains and losses from changes to access to environmental resources such as air, water, forests and beaches as well as monetary wealth.” It is recognised that a disadvantage associated with this choice of numéraire is the difficulty for respondents to think in terms of general well-being or welfare and not just monetary income. The distribution of preferences may be sensitive to the choice of numéraire and it is possible that if a different numéraire was applied, the distributional preferences may vary. Page 13 Theoretically, a possible solution to this difficulty would be to describe the attributes in terms of an “index of well-being”. This has been used in making a theoretical case in an example by Broome (1995) but not in an empirical exercise. While an index of well-being would encourage respondents to think in terms of utility being broader than money and therefore more in line with the notion of utility in the literature (Sen 1982; Sen 2000), the difficulty and subjectivity in developing an index, determining the values for components of the index and descriptors of the index make it impractical. Even if these issues were resolved, the cognitive difficulty for respondents of making complex decisions in an unfamiliar metric would remain a concern. For these reasons money was selected as the numéraire. The levels of the attributes involve the manipulation of attribute differences, not absolute values of the attributes. The hypothetical dollar values represent a one-off loss or gain to the individual representing the group described by the specific characteristic determining the attribute. In this case, there are five levels for each attribute with each level varying well-being to the value of A$500. Feedback from focus groups suggested this degree of variation was large enough to be significant to respondents in determining a choice, and not unrealistic in representing a once-off gain or loss. A fractional factorial design taken from Lazari and Anderson (1994) was used to create 25 choice sets, an example of which is presented in Figure 1. The 25 sets were blocked into groups of five so that each respondent was presented with five choice sets in a survey. Respondents were provided with a reference key such as that in Figure 2 when asked to complete the choice sets. Page 14 [Insert Figures 1 and 2 about here] A survey was conducted in July 2005 across a random sample of households in Warrnambool, a regional city in South West Victoria, Australia. A personal drop off and pick up form of distribution and collection was used. Acknowledging one of the strategies for data collection suggested by Dillman (2000), respondents were provided with the additional motivation to respond through the opportunity to participate in the draw for a A$150 shopping voucher at a major retail chain if they completed the questionnaire. A total of 431 questionnaires were distributed. Of the 337 that were collected or returned by mail, 295 were usable giving a response rate of 68.5%. Each of the 295 usable responses included five completed choice sets giving a total of 1475 completed choice sets. Each respondent also completed socio-demographic questions and two qualitative questions; one regarding specific strategies they had employed in answering the choice set questions and one regarding general comments they wished to make about the survey. Comparison of the survey sample’s socio-demographics with the Australian Bureau of Statistics (2001) census data indicates a slightly higher representation of females and younger people completing the survey than in the general population. Table 2 provides a comparison of the age profile of the sample with that of the 2001 census as this variable is particularly relevant to the analysis. [Insert Table 2 about here] Standard choice experiment procedures were applied with the distinction that an indirect welfare function rather than an indirect utility function has been assumed. It is assumed W z j = wzj + e zj ; where wzj is the deterministic component for respondent j and choice z, Page 15 j j and can be decomposed in wzj = [ X z , S j ] where X z is a vector of the attributes of alternative z and S j is a vector of the characteristics of respondent j. The stochastic component of welfare is e zj . Hence, j j wzj = ASC z + ∑ β z X z + ∑ γ n (S j ∗ ASC z ) [4] where ASCz is an alternative specific constant associated with the change options, β z are j the coefficients associated with each attribute and γ n is the vector of the coefficients associated with the socio-demographic characteristics intersected with the ASC to avoid singularities. Assuming that the random component of welfare is distributed as IID and with an extreme value (Gumbel) distribution, then the probability of an option being chosen can be expressed as the multinomial logit (MNL) or conditional logit (McFadden 1974). P(Wmj > W z j , ∀z ≠ m) = j exp( μwm ) ∑ exp μwzj z [5] where μ is a scale parameter which is inversely proportional to the standard deviation of the j error term and wm and wzj are conditional indirect welfare functions for choice options m and z, which are assumed to be linear in parameters. The key outputs of the welfare based choice model are the social marginal rates of substitution (SMRS). Given that the attributes of the choice model are the changes in utility accruing to particular groups then the SMRS are estimated by the ratios of the marginal Page 16 welfare changes (βs). Focussing on the ratios of the welfare parameters also overcomes the problem of confounding presented by the scale parameter in the choice model. For example, assuming a specific policy, m, the SMRS by respondent j, between those aged 50 and those aged 25 is: SMRS j m Aged 50 Aged 25 j j δβ mAged 50 β mAged 50 = = j j δβ mAged 25 β mAged 25 [6] In effect, the SMRS reflects a willingness to accept distributional change, which can be represented graphically by the slope of the SWF. This reflects the respondent’s notion of social justice. For example, in Figure 3, a movement from R to S indicates a willingness to trade a decrease in the utility of those in the aged 50 generation for an increase in the utility of those in the aged 25 generation. 2 [Insert Figure 3 about here] The SRMS also yields distributional weights applicable to a CBA setting. For example, the distributional weight, associated with Policy m, between those aged 50 and those aged 25 can be estimated by the SMRS. SMRS j Aged 50 m = Aged 25 ∂U mAged 50 ∂Wmj ⋅ ∂U mAged 50 ∂x mAged 50 ∂U mAged 25 ∂Wmj ⋅ ∂U mAged 25 ∂x mAged 25 [7] 2 Figure 3 assumes a well-behaved utilitarian SWF. Page 17 This reflects the two components of the distributional weight as indicated in Section two and equation [3]. The hypotheses to be tested using the model are that the distributional weights for each age group are not equal to one. For example, if there is altruism towards the younger generations by the community then when responses are aggregated: β aged 25 β newborn β aged 50 ≥ 1 and β aged 50 ≥ 1 [8] 4. RESULTS The MNL was estimated, using the Stata software program. The IIA (independence of irrelevant alternatives) property was tested using the test suggested by Hausman and McFadden (1984) and compliance was confirmed. Each of the variables used in the model is specified in Table 3. [Insert Table 3 about here] Model results are summarised in Table 4. Each choice set attribute parameter is significant at the 1% level and signed as expected a priori indicating that the utility of each age group contributes positively to the social welfare function. [Insert Table 4 about here] Of the demographic characteristics, the age, income and parental status variables are significant at the five percent level. The interpretation of the signs for the demographic Page 18 characteristics is complicated by the changes being both positive and negative in the design of the choice experiment. Table 5 summarises the 95% confidence intervals for the mean social marginal rates of substitution. These results indicate a distributional preference towards the younger generations with the ratio of the welfare parameters being greater than 1 for both the aged 25 generation and newborns relative to the aged 50 generation. For the Aged25/Aged50, the SMRS suggests a relative distributional weight of 1.70 and for the Newborn/Aged50 a weighting of 2.35. [Insert Table 5 about here] The preference towards the utility of younger generations evident in the quantitative analysis is consistent with comments made by respondents to the qualitative questions regarding the strategy they had used to make choices. (One hundred and fourteen of the 295 respondents chose to explain the strategy they had used in answering the choice questions.) Examples of these comments include: “Help younger generation and early workforce people.” “Picked ones that were most likely beneficial to the younger generation.” “Thinking about effect on future generations.” 5. DISCUSSION AND CONCLUSION The findings of this research are particularly relevant to the sustainability debate and in the context of the trade-off between consumption today and consumption in the future. The Page 19 magnitude of the weighting towards the utility of future generations supports the contention of Arrow, Dasgupta et.al. (2004) that individuals “derive a positive externality (outside of the marketplace) from the welfare of future generations”. The findings also support estimates made by Johansson-Stenman, Carlsson et.al (2002) in an experiment involving students’ preferences with respect to “imaginary grandchildren” and future income distribution, where they found that respondents were willing to trade-off “non-negligible” amounts of money for increasing their grandchild’s relative standing in society. Arrow, Dasgupta, et.al (2004) have approached the question of intergenerational equity by estimating the elasticity of marginal (social) utility. Theoretically, for a consumption path in a market economy to be socially optimal, the market rate of return on investment, i, must be equal to the social rate of interest on consumption, denoted by r(Arrow, Dasgupta et al. 2004). If i exceeds r, markets are biased toward insufficient saving and excessive current consumption. Based on the estimation of an intertemporal social welfare function, the social rate of interest on consumption r, is given by the relation r = δ + ηg , where δ is the social rate of pure time preference, η is the elasticity of marginal (social) utility and g is the rate of growth in aggregate consumption. The choice of δ and η are value judgements which are likely to vary between individuals. The term η, which is of relevance to the findings reported here, is interpreted by Arrow, Dasgupta et.al. as “a social preference for equality of consumption among generations”. They speculate that the value of η is linked to the intertemporal elasticity of consumption and based on Hall’s (1988) time series estimates of this suggest that “plausible values for η might lie in the range of 24”. Although these authors have approached the question of intergenerational distribution from a different perspective, the distributional weights between those aged 50 and newborns estimated in this study falls within this expected range. Page 20 In conclusion, the results of this research demonstrate distributional weights that are not equal to one and positively favour the younger generations. The positive distributional preferences towards future generations may be due to a combination of factors including altruism toward future generations and diminishing marginal utility as raised by Arrow, Dasgupta et. al. (2004). The implication of the results is that environmental policies that favour intergenerational transfers of utility from current to future generations will be more favourably treated in CBA and hence more likely to be accepted as preferred options by the community. Furthermore, the findings of this research show that choice modelling is a useful method for eliciting the utility distributional preferences of the community. This has implications for the incorporation of the distributional impacts of environmental policies into CBA and hence decision making. An advantage of this approach is that it provides the policy maker with information regarding the community’s preferences across generations. Policy interventions have to be sensitive to the gainers and losers, not only because that matters from a social justice point of view, but also because the political acceptability and effectiveness of the measures will depend on the distribution of costs and benefits. • We appreciate the comments of Assoc. Prof. Michael Burton and participants of the 2006 conference of the Australian Agricultural and Economics Society on an earlier draft of this paper. REFERENCES Adler, M. D. and Posner, E. A. (1999). "Rethinking Cost-Benefit Analysis." The Yale Law Journal 109(2): 164-247. Page 21 Agyeman, J., Bullard, R. and Evans, B., Eds. (2003). Just Sustainabilities: Development in an Unequal World. Earthscan.London. Arrow, K. J. (1963). Social Choice and Individual Values. 2nd Wiley, New York. Arrow, K. J., Dasgupta, P., Goulder, L. H., Daily, G. C., Ehrlich, P., Heal, G., Levin, S., Mäler, K.-G., Schneider, S., Starrett, D. and Walker, B. (2004). "Are We Consuming Too Much?" Journal of Economic Perspectives 18(3): 147-172. Atkinson, G., Machado, F. and Mourato, S. (2000). "Balancing competing principles of environmental equity." Environment and Planning 32: 1791-1806. Australian Bureau of Statistics (2001). Census of Population and Housing.Commonwealth of Australia, Canberra. Basu, K. (1980). Revealed Preference of Government. Cambridge University Press, Cambridge. Bennett, J. and Blamey, R. (2001). The Choice Modelling Approach to Environmental Valuation. Edward Elgar, Cheltenham, UK. Boadway, R. and Bruce, N. (1984). Welfare Economics. Basil Blackwell, Oxford, U.K. Broome, J. (1995). Weighing Goods. Basil Blackwell, Oxford, U.K. Cowell, F. A. and Gardiner, K. (1999). Welfare Weights.Report to the UK Office of Fair Trading, United Kingdom. Dasgupta, P., Sen, A. and Marglin, S. (1972). Guidelines for Project Evaluation. United Nations, New York. Department of Finance and Administration (2006). Introduction to Cost-Benefit Analysis and Alternative Evaluation Methodologies. Australian Government, Canberra. Dillman, D. A. (2000). Mail and Internet Surveys: The Tailored Design Method. John Wiley & Sons, New York. Dreze, J. (1998). "Distribution matters in cost-benefit analysis: Comment on K.A.Brekke." Journal of Public Economics 70: 485-488. Dreze, J. and Stern, N. (1987). The Theory of Cost- Benefit Analysis in Handbook of Public Economics. A. Auerbach and Feldstein, M. II: 909-989. Elsevier Science. Amsterdam Elster, J. (1992). Local Justice: How Institutions Allocate Scarce Goods and Necessary Burdens. Cambridge University Press, Cambridge. H.M.Treasury (2003). The Green Book: Appraisal and Evaluation in Central Government. London. Hall, R. E. (1988). "Intertemporal Substitution in Consumption." The Journal of Political Economy 96(2): 339-357. Harberger, A. C. (1978). "On the Use of Distributional Weights in Social Cost-Benefit Analysis." Journal of Political Economy 86(Supplement): 87-120. Harberger, A. C. and Jenkins, G. P., Eds. (2002). Cost-Benefit Analysis. The International Library of Critical Writings in Economics. Edward Elgar.Cheltenham, UK. Hausman, J. and McFadden, D. (1984). "Specification Tests for the Multinomial Logit Model." Econometrica 52(5): 1219-1240. Johansson, P.-O. (1987). The Economic Theory and Measurement of Environmental Benefits. Cambridge University Press, Cambridge. Johansson, P.-O. (1993). Cost-benefit analysis of environmental change. Cambridge University Press, Cambridge. Johansson-Stenman, O. (2005). "Distributional Weights in Cost-Benefit Analysis-Should We Forget about Them?" Land Economics 81(3): 337-352. Johansson-Stenman, O., Carlsson, F. and Daruvala, D. (2002). "Measuring Future Grandparents' Preferences for Equality and Relative Standing." The Economic Journal 112: 362-383. Page 22 Krinsky, I. and Robb, A. L. (1986). "On approximating the statistical properties of elasticities." The Review of Economics and Statistics 68(4): 715-719. Layard, R. (1980). "On the Use of Distributional Weights in Social Cost-Benefit Analysis." Journal of Political Economy 88(5): 1041-1049. Layard, R. and Glaister, S. (1994). Cost-Benefit Analysis. Second Cambridge University Press, Cambridge. Lazari, A. G. and Anderson, D. A. (1994). "Designs of Discrete Choice Set Experiments for Establishing Both Attribute and Availability Cross Effects." Journal of Marketing Research XXX1: 375-383. Little, I. M. D. and Mirrlees, J. A. (1974). Project Appraisal and Planning for Developing Countries. Heinemann, London. Little, I. M. D. and Mirrlees, J. A. (1994). The costs and benefits of analysis: Project appraisal and planning twenty years on in Cost-Benefit Analysis. R. Layard and Glaister, S.: 199-234. Cambridge University Press. Cambridge Louviere, J. J., Hensher, D. A. and Swait, J. D. (2000). Stated Choice Methods. Cambridge University Press, Cambridge U.K. Mackay, H. (1997). Generations: Baby Boomers, their parents & their children. Macmillan, Sydney. Mäler, K.-G. (1985). Welfare Economics and the Environment in Handbook of Natural Resource and Energy Economics Volume 1. A. V. Kneese and Sweeney, J. L.: 3-60. Elsevier Science. Amsterdam Markandya, A. (1998). "Poverty, Income Distribution and Policy Making." Environmental and Resource Economics 11(3-4): 459-72. McFadden, D. (1974). On Conditional logit model of qualitative choice behaviour in Frontiers in Econometrics. P. Zarembka. Academic Press. New York Medin, H., Nyborg, K. and Bateman, I. (2001). "The assumption of equal marginal utility of income: how much does it matter?" Ecological Economics 36: 397-411. Musgrave, R. A. and Musgrave, P. B. (1989). Public Finance in Theory and Practice. Fifth McGraw-Hill, New York. Nyborg, K. (2000). "Homo Economicus and Homo Politicus:Interpretation and aggregation of environmental values." Journal of Economic Behaviour and Organization 42: 305-322. Pearce, D. and Barbier, E. (2000). Blueprint for a Sustainable Economy. Earthscan, London. Rescher, N. (2002). Fairness. Transaction Publishers, New Brunswick (USA). Sen, A. (1982). Choice, Welfare and Measurement. Basil Blackwell, Oxford. Sen, A., K (2000). Social Justice and the Distribution of Income in Handbook of Income Distribution. A. B. Atkinson and Bourguignon, F. 60-85. North Holland. Amsterdam Serret, Y. and Johnstone, N. (2006). The Distributional Effects of Environmental Policy. Edward Elgar, Cheltenham, UK. Squire, L. (1989). Project Evaluation in Theory and Practice in Handbook of Development Economics. H. Chenery and Srinivasan, T. 2: 1093-1137. Elsevier Science. Squire, L. and van der Tak, H. G. (1975). Economic Analysis of Projects. Johns Hopkins University Press, Baltimore. Sutherland, R. J. (2006). The distributive effects of direct regulation: a case study of energy efficiency appliance standards in The Distributional Effects of Environmental Policy. Y. Serret and Johnstone, N.: 171-196. Edward Elgar. Cheltenham Tacconi, L. (2000). Biodiversity and Ecological Economics: Participatory Approaches to Resources Management. Earthscan, London. Page 23 World Bank (1996). "Handbook on economic analysis and investment operations." World Bank. World Commission on Environment and Development (1987). Our Common Future: The Brundtland Report. Oxford University Press, New York. Page 24 Figure 1 Example of an intergenerational utility distribution choice set Figure 2: Reference key for choice set in Figure 1 Page 25 Figure 3 Example of social welfare function Utility of the Aged 50 generation R S SWF Utility of the Aged 25 generation Table 1 Attributes and levels in intergenerational distribution choice experiment Attribute Utility change Person Aged 50 Utility change Person Aged 25 Utility change Newborn Levels ($A) -$1,000 -$1,000 -$1,000 -$500 -$500 -$500 +$500 +$500 +$500 +$1,000 +$1,000 +$1,000 +$1,500 +$1,500 +$1,500 Page 26 Table 2 Age profile of Respondents Age group 18-24 25-34 35-44 45-54 55-64 Over 65 No response Number 34 49 56 70 46 36 4 % of sample 11.5 16.6 19.0 23.7 15.6 12.2 1.4 % of Census 2001* 13.8 18.0 20.1 17.7 11.6 18.8 Total 295 100 100 *Taken as % of census population aged 18 and over (Australian Bureau of Statistics 2001). Table 3 Variables used in the CM application Aged50 Aged25 Newborn Age Income Parent Gparent Noschild Gender Education Change in the well-being of person representing those aged 50 Change in the well-being of person representing those aged 25 Change in the well-being of person representing those newborn Age of respondent in years Income of respondent in last year in thousands of Australian dollars Parental status of respondent Grandparental status of respondent Number of children of respondent Gender of respondent Education level of respondent; pre secondary, secondary, tertiary Page 27 Table 4 Intergenerational utility distribution MNL model results Variable Coefficient Std error z P>/z/ ASC Aged50 Aged25 Newborn age income parent noschild gparent gender edu Model Statistics Log L Adj Rho-square -1.1362 0.0003 0.0005 0.0006 0.0153 0.0094 -0.5253 -0.0273 0.2128 -0.1694 0.0225 .2581 .0001 .0000 .0001 .0063 .0034 .2357 .0641 .2080 .1359 .0846 -4.40 5.47 9.06 10.21 2.44 2.75 -2.23 -0.43 1.02 -1.25 0.27 0.000 0.000 0.000 0.000 0.015* 0.006* 0.026* 0.670 0.306 0.213 0.790 -1133.29 0.0911 *Significant at 5% level Page 28 Table 5 Social marginal rates of substitution Aged 25/Aged 50 Mean 95% CI* Newborn/Aged 50 Mean 95% CI* Newborn/Aged 25 Mean 95% CI* Model excluding SDC 1.50 (0.97, 2.37) 2.28 (1.47, 3.74) 1.54 (1.12, 2.10) Model including SDC 1.70 (1.03, 2.88) 2.35 (1.43, 4.28) 1.39 (1.00, 1.98) *95% confidence intervals estimated with the Krinsky-Robb(1986) method using 1000 replications. Page 29