|
on Discrete Choice Models |
| By: | Susan Athey; Guido Imbens |
| Date: | 2006–01–13 |
| URL: | http://d.repec.org/n?u=RePEc:cla:levrem:122247000000001040&r=dcm |
| By: | Siebert, Ralph; von Graevenitz, Georg |
| Abstract: | This paper is a study of licensing in a patent thicket. In a patent thicket licensing allows firms to avoid hold-up. It will have different effects on firms' R&D incentives depending on whether firms license existing or future patents. Building on a model of a patent portfolio race, firms' choice between these types of licensing contracts is modelled. We find that firms' relationships in product markets and technology space jointly determine the type of licensing contract chosen. We derive several hypotheses and test these. Using data from the semiconductor industry a dynamic panel data model with unobserved heterogeneity and a lagged dependent variable is estimated. A new method suggested by Wooldridge (2005) is employed to estimate a random effects probit model using conditional ML. The hypotheses derived from the theory are confirmed. Based on our results we argue that licensing raises welfare in the patent thicket. |
| Keywords: | hold-up problem; innovation; licensing; patent race; patent thicket |
| JEL: | L13 L49 L63 |
| Date: | 2005–12 |
| URL: | http://d.repec.org/n?u=RePEc:cpr:ceprdp:5436&r=dcm |
| By: | Matilde P. Machado,; Ricardo Mora; Antonio Romero- Medina |
| Abstract: | In this paper, we propose an alternative methodology to rank hospitals based on the choices of Medical Schools graduates over training vacancies. We argue that our measure of relative hospital quality has the following desirable properties: a) robustness to manipulation from the hospital’s administrators; b) comprehensiveness in the scope of the services analyzed; c) inexpensive in terms of data requirements, and d) not subject to selection biases. Accurate measures of health provider quality are needed in order to establish incentive mechanisms, to assess the need for quality improvement, or simply to increase market transparency and competition. Public report cards in certain US states and the NHS ranking system in the UK are two attempts at constructing quality rankings of health care providers. Although the need for such rankings is widely recognized, the criticisms at these attempts reveal the difficulties involved in this task. Most criticisms alert to the inadequate risk-adjustment and the potential for perverse consequences such as patient selection. The recent literature, using sophisticated econometric models is capable of controlling for case-mix, hospital and patient selection, and measurement error. The detailed data needed for these evaluations is, however, often unavailable to researchers. In those countries, such as Spain, where there is neither public hospital rankings nor public data on hospital output measures such as mortality rates our methodology is a valid alternative. We develop this methodology for the Spanish case. In a follow-up paper we will present results using Spanish data. In Spain graduates choose hospital training vacancies in a sequential manner that depends on their average grade. Our framework relies on three assumptions. First, high quality hospitals provide high quality training. Second, graduates are well informed decision makers who are well qualified to assess hospital quality. Third, they prefer to choose a high quality vacancy rather than a low quality one ceteris paribus. If these assumptions hold, then the first physicians to choose are likely to grab the best vacancies while the ones who choose last are stuck with the worst available. Thus, it is possible to infer from physicans’ choices quality differentials amongst hospitals. We model the physician’s decision as a nested-logit a la McFadden. Unlike in standard applications of McFadden’s model, in our application the choice set is not constant across physicians but it shrinks along the sequential hospital choice process |
| Date: | 2006–01 |
| URL: | http://d.repec.org/n?u=RePEc:cte:werepe:we060201&r=dcm |
| By: | Robert L. Hicks (Department of Economics, College of William and Mary); Kurt Schnier (Department of Environmental and Natural Reseource Economics, University of Rhode Island) |
| Abstract: | This paper examines the impact of dolphin-safe eco-labeling and how it fundamentally altered the spatial distribution of fishing effort and fishermen's willingness to pay to avoid dolphins. To do this, a dynamic discrete choice econometric model is applied to the Eastern Tropical Pacific tuna fishery. This econometric approach combines a dynamic programming component with the static discrete site choice model. This estimator couples the current period projected profits associated with fishing a specific site with the value of all future location choices on the cruise, assuming choices are made optimally. The key feature of this model is that it recovers behavioral parameters and solves the dynamic programming problem recursively. The dynamic site choice model reveals a markedly higher impact on producers as compared to the commonly used static model following the labeling regime. Further, in all but a few cases the common practice in dynamic choice models of setting discount factors equal to one is rejected. |
| Keywords: | location choice, dynamic random utility modeling, dolphin-safe eco-labeling |
| JEL: | C35 Q20 Q58 |
| Date: | 2006–01–12 |
| URL: | http://d.repec.org/n?u=RePEc:cwm:wpaper:25&r=dcm |
| By: | Celestino Suárez-Burgnet; Inmaculada Martínez-Zarzoso; Laura Márquez-Ramos |
| Abstract: | This paper provides new empirical evidence on the determinants of regional integration agreements (RIAs) in a discrete choice modelling framework. The research has two main aims: first, to empirically analyse the determinants of different levels of integration, re-examining the evidence presented by Baier and Bergstrand (2004) in the JIE 64 (1); and second, to analyse the importance of additional factors, in particular socio-political factors. The results show that geographical factors alone are the most important explanatory factors for the probability of regional integration agreement formation or enhancement, thus supporting the theories on "natural" trading partners. The socio-political factors considered, democracy dummy, the level of economic freedom and the common language dummy are all statistically significant, although their relative importance in explaining RIA formation is low. |
| URL: | http://d.repec.org/n?u=RePEc:fda:fdadef:05-10&r=dcm |
| By: | Skogsvik, Kenth (Dept. of Business Administration, Stockholm School of Economics) |
| Abstract: | Probabilistic business failure prediction models are commonly estimated from non-random samples of companies. The proportion of failure companies in such samples is often much larger than the proportion of failure companies in most real-world decision contexts. This so-called “choice-based sample bias” implies that calculated failure probabilities will be (more or less) biased. The purpose of the paper is to analyse this bias and its consequences for standard applications of probabilistic failure prediction models (for example probit/logit analysis) and in particular to investigate whether the bias can be eliminated without having to re-estimate the underlying statistical model. It is shown that there is a straightforward linkage between sample-based probabilities of failure and the corresponding population-based probabilities. Knowing this linkage, sample-based probabilities can be adjusted for the “choice-based sample bias”, provided that sufficiently large samples of randomly selected failure companies and randomly selected survival companies have been used in the estimation of the underlying statistical model. Empirical observations in previous research are in line with the theoretical results of the paper. |
| Keywords: | Business Failure Prediction; Choice-Based Sample Bias; Financial Analysis; Probabilistic Prediction Model; Probit/Logit Analysis |
| Date: | 2005–12–01 |
| URL: | http://d.repec.org/n?u=RePEc:hhb:hastba:2005_013&r=dcm |
| By: | Carlsson, Fredrik (Department of Economics, School of Business, Economics and Law, Göteborg University); Martinsson, Peter (Department of Economics, School of Business, Economics and Law, Göteborg University) |
| Abstract: | In a split sample design, we examine how the number of choice sets, design of the first choice set (starting point), and the choice of attribute levels in the cost attribute affect the precision in the elicited preferences in otherwise completely identical choice experiment surveys. These issues are investigated for Swedish households’ marginal willingness to pay to reduce power outages. Our results indicate that neither the number´of choice sets nor the starting point choice set has a significant impact on estimated marginal willingness to pay, while the effect was significant for theadditive scaling of the cost vector. At the end of the paper we discuss the implications of our results on future developments and applications of choice experiments. <p> |
| Keywords: | Attribute levels; choice experiment; complexity; length; power outages; starting point |
| JEL: | C25 C93 D12 Q41 |
| Date: | 2006–01–19 |
| URL: | http://d.repec.org/n?u=RePEc:hhs:gunwpe:0191&r=dcm |
| By: | Robert W. Fairlie (University of California, Santa Cruz, National Poverty Center and IZA Bonn) |
| Abstract: | The Blinder-Oaxaca decomposition technique is widely used to identify and quantify the separate contributions of group differences in measurable characteristics, such as education, experience, marital status, and geographical differences to racial and gender gaps in outcomes. The technique cannot be used directly, however, if the outcome is binary and the coefficients are from a logit or probit model. I describe a relatively simple method of performing a decomposition that uses estimates from a logit or probit model. Expanding on the original application of the technique in Fairlie (1999), I provide a more thorough discussion of how to apply the technique, an analysis of the sensitivity of the decomposition estimates to different parameters, and the calculation of standard errors. I also compare the estimates to Blinder-Oaxaca decomposition estimates and discuss an example of when the Blinder-Oaxaca technique may be problematic. |
| Keywords: | decomposition, logit, probit, Blinder-Oaxaca decomposition, race, gender |
| JEL: | C6 J15 J16 |
| Date: | 2006–01 |
| URL: | http://d.repec.org/n?u=RePEc:iza:izadps:dp1917&r=dcm |
| By: | BOSSERT, Walter; SUZUMURA, Kotaro |
| Abstract: | The rationalizability of a choice function on an arbitrary domain under various coherence properties has received a considerable amount of attention both in the long-established and in the recent literature. Because domain closedness conditions play an important role in much of rational choice theory, we examine the consequences of these requirements on the logical relationships among different versions of rationalizability. It turns out that closedness under intersection does not lead to any results differing from those obtained on arbitrary domains. In contrast, closedness under union allows us to prove an additional implication. |
| Keywords: | Rational Choice, Domain Closedness, Coherence Prorties |
| JEL: | D11 |
| Date: | 2005 |
| URL: | http://d.repec.org/n?u=RePEc:mtl:montde:2005-21&r=dcm |
| By: | Sendhil Mullainathan; Ebonya Washington |
| Abstract: | In traditional models, votes are an expression of preferences and beliefs. Psychological theories of cognitive dissonance suggest, however, that behavior may shape preferences. In this view, the very act of voting may influence political attitudes. A vote for a candidate may lead to more favorable interpretations of his actions in the future. We test the empirical relevance of cognitive dissonance in US Presidential elections. The key problem in such a test is the endogeneity of voter choice which leads to a mechanical relationship between voting and preferences. We use the voting age restrictions to help surmount this difficulty. We examine the Presidential opinion ratings of nineteen and twenty year olds two years after the President's election. Consistent with cognitive dissonance, we find that twenty year olds (who were eligible to vote in the election) show greater polarization of opinions than comparable nineteen year olds (who were ineligible to vote). We rule out that aging drives these results in two ways. First, we find no polarization differences in years in which twenty and nineteen year olds would not have differed in their eligibility to vote in the prior Presidential election. Second, we show a similar effect when we compare polarization (for all age groups) in opinions of Senators elected during high turnout Presidential campaign years with Senators elected during low turnout non-Presidential campaign years. Thus we find empirical support for the relevance of cognitive dissonance to voting behavior. This finding has at least three implications for the dynamics of voting behavior. First, it offers a new rationale for the incumbency advantage. Second, it suggests that there is an efficiency argument for term limits. And finally, our results demonstrate that efficiency may not be increasing in turnout level. |
| JEL: | D0 H0 |
| Date: | 2006–01 |
| URL: | http://d.repec.org/n?u=RePEc:nbr:nberwo:11910&r=dcm |
| By: | W. Kip Viscusi; Joel Huber |
| Abstract: | This article examines revealed rates of time preference for public goods, using environmental quality as the case study. A nationally representative panel-based sample of 2,914 respondents considered a series of 5 conjoint policy choices, yielding 14,570 decisions. Both the conditional fixed effect logit estimates of the random utility model and mixed logit estimates implied that the rate of time preference is very high for immediate improvements and drops off substantially thereafter, which is inconsistent with exponential discounting but consistent with hyperbolic discounting. The implied marginal rate of time preference declines and then rises. Estimates of the quasi-hyperbolic discounting parameter range from 0.48 to 0.61. People who are older are especially likely to have a high disutility from delays in improving water quality. |
| JEL: | Q20 D90 H40 |
| Date: | 2006–01 |
| URL: | http://d.repec.org/n?u=RePEc:nbr:nberwo:11935&r=dcm |
| By: | Kevin F. Hallock; Craig Olson |
| Abstract: | This study empirically investigates the value employees place on stock options using information from the option exercise behavior of individuals. Employees hold options for another period if the value from holding them and reserving the right to exercise them later is higher than the value of exercising them immediately and collecting a profit equal to the stock price minus the exercise price. This simple model implies the hazard describing employee exercise behavior reveals information about the value to employees of holding options another time period. We show the parameters of this model are identified with data on multiple option grants per employee and we apply this model to the disposition of options received in the 1990s by a sample of over 2000 middle-level managers from a large, established firm outside of manufacturing. Exercise behavior is modeled using a random effects probit model of monthly exercise behavior that is estimated using simulated maximum likelihood estimation methods. Our estimates show there is substantial heterogeneity (observed and unobserved) among employees in the value they place on their options. Our estimates show most employees value their options at a value greater than the option's Black-Scholes value. |
| JEL: | J3 G3 |
| Date: | 2006–01 |
| URL: | http://d.repec.org/n?u=RePEc:nbr:nberwo:11950&r=dcm |
| By: | Yusuke Sakata (Departmen of Economics, Kinki University); Junyi Shen (Osaka School of Interna ional Public Policy, Osaka University); Yoshizo Hashimoto (Osaka School of Interna ional Public Policy, Osaka University) |
| Abstract: | Transport modal choice is considered to be influenced by natural environmental change and transport network improvement. This paper reveals how these impacts affect individualsf decisions on selecting transport mode under an extension plan of Osaka Monorail Loop-line. To estimate these impacts, a Stated Choice (SC) experiment is carried out for collecting the neighborhood data around the monorailfs extended area. Our model is estimated by the Heteroscedastic Extreme Value (HEV) specification in order to avoid Independence from Irrelevant Alternatives (IIA) assumption in the Multinomial Logit (MNL) model. Both the results of full-sample and sub-sample data imply that residents prefer public transport modes (monorail or bus) to private car when either natural environment becomes worse or transport network is improved. |
| Keywords: | Environmental Deterioration; Network improvement; Network externality; Choice Model (CM); Heteroscedastic Extreme Value (HEV) Model |
| JEL: | C35 D12 Q51 R41 |
| Date: | 2006–01 |
| URL: | http://d.repec.org/n?u=RePEc:osk:wpaper:0604&r=dcm |
| By: | Junyi Shen (Osaka School of Interna ional Public Policy, Osaka University); Yusuke Sakata (School of Economics, Kinki University); Yoshizo Hashimoto (Osaka School of Interna ional Public Policy, Osaka University) |
| Abstract: | This paper applies two recent stated choice survey datasets of Japan to investigate the difference between the latent class model (LCM) and the mixed logit model (MLM) for transport mode choice. A detailed comparison is carried out, focusing on comparing values of time savings, direct choice elasticities, predicted choice probabilities and prediction success indices. Furthermore, a test on non-nested model is also utilized to help determine which model is superior to another one. The results suggest that the LCM performs better than the MLM in both datasets. |
| Keywords: | latent class model (LCM), mixed logit model (MLM), transport mode choice, predicted choice probability, prediction success index |
| JEL: | C35 D12 R41 |
| Date: | 2006–01 |
| URL: | http://d.repec.org/n?u=RePEc:osk:wpaper:0605&r=dcm |
| By: | Donata Favaro (University of Padua); Carlofilippo Frateschi (University of Padua) |
| Abstract: | In this paper we present an empirical analysis of the 'patterns of cultural choice' in the musical domain in Italy. By employing micro data from the Italian Survey on Households. Citizens and Leisure, we investigate the way in which musical tastes appear to be differentiated in the Italian society. The main goal of the paper is to verify whether musical tastes could be characterised by the traditionally accepted dichotomy between an elite of 'cultural snobs' and a mass of consumers of 'lowbrow' genres or whether they are diversified, with the presence of a new group of "cultural omnivores". By applying the discrete choice analysis, we implement a multivariate logit model in which each equation represents the choice of one of the mutually exclusive alternatives. Our dependent variable is constructed in a way to exactly identify the set of the alternatives and to include all possible individual choices. Our estimates point to the existence of different 'degrees of omnivorouness' and capture the differences between individuals listening to (or attending concerts of) only one musical genre, more than one genre or all genres. |
| JEL: | D12 C25 Z1 |
| Date: | 2005–12 |
| URL: | http://d.repec.org/n?u=RePEc:pad:wpaper:0010&r=dcm |
| By: | Gábor Békés (Institute of Economics - Hungarian Academy of Sciences and Central European University) |
| Abstract: | In a new economic geography framework with input-output linkages, this study analyses decisions made by foreign firms about their location within Hungary. These firm-to-firm contacts are modelled by creating several corporate customer and supplier access measures for all new foreign corporations. In order to see the impact of these variables other forces of agglomeration such as distance to Western European markets and dispersion forces such as high wages are taken into account. Investigation is carried out on a small-to-medium sized European economy that has just gone through economic transition involving almost unprecedented rapid market liberalisation. A rich dataset of corporate tax returns of Hungarian firms between 1992 and 2002 as well as annual labor surveys are used to get location, sales and wage data. Various econometric specifications of both discrete choice and count data models are applied to provide robustness of results that may be crucial when working with firm level data. |
| Keywords: | economic geography, industrial location, FDI, regional policy, discrete choice models |
| JEL: | F23 R3 R12 C35 |
| Date: | 2005 |
| URL: | http://d.repec.org/n?u=RePEc:mnb:wpaper:2005/07&r=dcm |
| By: | Newell, Richard (Resources For the Future); Anderson, Soren |
| Abstract: | We show that after a simple normalization of explanatory variables so that they equal zero at some desired reference point, marginal effects for continuous variables in probit and logit models simplify dramatically, becoming a function of only the estimated constant term. We present similar simplifications for computation of the asymptotic variance of marginal effects, as well as for the effects of dummy variables on predicted probabilities. We provide a simple table, which in combination with raw probit or logit estimates, is all one needs to compute the desired effects. |
| Keywords: | logit, probit, discrete choice, binary choice, marginal effect, data normalization |
| JEL: | C25 C51 C81 |
| URL: | http://d.repec.org/n?u=RePEc:rff:dpaper:dp-03-38&r=dcm |
| By: | Stavins, Robert; Plantinga, Andrew; Lubowski, Ruben |
| Abstract: | Changes in the use of land in the United States produce significant economic and environmental effects with important implications for a wide variety of policy issues, including protection of wildlife habitat, management of urban growth, and mitigation of global climate change. In contrast to previous descriptive and qualitative analyses of the trends in national land use, this paper uses an econometric approach to isolate the importance of historical changes in land-use profits and key government policies in determining national land-use changes from 1982 to 1997. The policies we examine are the Conservation Reserve Program (CRP) and total government payments to crop producers. We estimate a national-level discrete choice model of changes among the major land-use categories (crops, pasture, forest, urban, range, and CRP) with parcel-level observations of land use and land quality from the U.S.D.A. National Resources Inventory NRI) and measures of countylevel land-use net returns from a variety of sources. We then use fitted values from the econometric model to simulate land-use change from 1982 to 1997 under a series of factual and counterfactual scenarios that isolate the effects of different economic and policy factors. The simulations suggest how changes in economic returns and government policies have driven land-use changes in the past and will continue to affect nationwide land-use changes in the future. For example, we find that the introduction of the CRP and the decline in crop profits were the most significant explanatory factors driving the decline in cropland. Our results highlight some “unintended consequences” of government policies and the importance of net returns to a range of alternative land uses as determinants of land area change for each particular use. |
| Keywords: | land use; econometric model; counterfactual simulation; Conservation Reserve Program (CRP) |
| JEL: | C53 Q1 Q24 R14 R15 |
| URL: | http://d.repec.org/n?u=RePEc:rff:dpaper:dp-03-47&r=dcm |
| By: | Smith, V. Kerry; Banzhaf, H. Spencer (Resources For the Future) |
| Abstract: | We document the sensitivity of welfare estimates derived from discrete choice models to assumptions about the choice set. Such assumptions can affect welfare estimates through both the estimated parameters of the model and, conditional on the parameters, the substitution among alternatives. Our analysis involves estimates of the benefits of air quality improvements in Los Angeles based on discrete choices of neighborhood and housing. We further illustrate the use of meta analysis to document and summarize voluminous information derived from repeated sensitivity analyses. |
| Keywords: | Meta analysis, random utility model, choice set, air quality, housing |
| JEL: | C15 Q25 R21 |
| URL: | http://d.repec.org/n?u=RePEc:rff:dpaper:dp-03-61&r=dcm |
| By: | Pizer, William (Resources For the Future); Newell, Richard (Resources For the Future) |
| Abstract: | We estimate a carbon mitigation cost curve for the U.S. commercial sector based on econometric estimation of the responsiveness of fuel demand and equipment choices to energy price changes. The model econometrically estimates fuel demand conditional on fuel choice, which is characterized by a multinomial logit model. Separate estimation of end uses (e.g., heating, cooking) using the 1995 Commercial Buildings Energy Consumption Survey allows for exceptionally detailed estimation of price responsiveness disaggregated by end use and fuel type. We then construct aggregate long-run elasticities, by fuel type, through a series of simulations; own-price elasticities range from –0.9 for district heat services to –2.9 for fuel oil. The simulations form the basis of a marginal cost curve for carbon mitigation, which suggests that a price of $20 per ton of carbon would result in an 8% reduction in commercial carbon emissions, and a price of $100 per ton would result in a 28% reduction. |
| Keywords: | commercial energy demand, carbon policy, climate change, discrete choice |
| JEL: | Q28 Q48 Q41 C35 C15 |
| URL: | http://d.repec.org/n?u=RePEc:rff:dpaper:dp-05-13&r=dcm |
| By: | Smith, V. Kerry; Mansfield, Carol |
| Abstract: | This paper provides the results of a field test of contingent valuation estimates within a willingness to accept framework. Using dichotomous choice questions in telephone-mail-telephone interviews, we compare responses to real and hypothetical offers to survey respondents for the opportunity to spend time in a second set of interviews on an undisclosed topic. Five hundred and forty people were randomly split between the real and hypothetical treatments. Our findings indicate no significant differences between people's choices with real and hypothetical offers. Choice models indicate the size of the offer and income were significant determinants of respondents' decisions, and these models were <u>not</u> significantly different between real and hypothetical offers. |
| URL: | http://d.repec.org/n?u=RePEc:rff:dpaper:dp-97-09&r=dcm |
| By: | Smith, V. Kerry; Huang, Ju-Chin |
| Abstract: | This paper argues that the widespread belief that discrete contingent valuation (CV) questions yield substantially larger estimates of the mean (and the median) willingness to pay (WTP) for nonmarket environmental resources in comparison to estimates from open-ended CV questions is unfounded. A set of Monte Carlo experiments estimate the factors influencing the performance of WTP estimates based on discrete response models. Most of the error in the WTP estimates arises from the specification errors that are common in most of the empirical models used in the literature. These experiments suggest models based on choices where WTP is dominated by non use (or passive use) values are likely to have smaller errors than where large use values influence these decisions. |
| URL: | http://d.repec.org/n?u=RePEc:rff:dpaper:dp-97-23&r=dcm |
| By: | Banzhaf, H. Spencer (Resources For the Future) |
| Abstract: | This paper illustrates how public goods may be incorporated into a cost-of-living index. When public goods are weak complements to a market good, quality-adjusted prices for the market good capture all the welfare information required. They are also consistent with a Laspeyres index that maintains the bound on a true cost-of-living index. The paper recovers this information from a discrete-choice model, using a simulation routine to solve for the appropriate price adjustments. These concepts are applied to the case of housing, education, crime, and air quality in Los Angeles for 1989 to 1994. Over a period of time when they are improving, incorporating pubic goods into the index lowers the estimated change in the cost of living by 0.5 to 2.6 percentage points. In other years, when public goods diverge, the estimated annual adjustment differs by model, with a range of -0.2 to +1.3 percentage points. |
| Keywords: | air quality, discrete choice models, green accounting, nonmarket valuation, price inde |
| JEL: | C51 D12 D60 E31 H40 R10 |
| URL: | http://d.repec.org/n?u=RePEc:rff:dpaper:dp-02-10-&r=dcm |
| By: | Macauley, Molly (Resources For the Future); Walls, Margaret (Resources For the Future); Anderson, Soren |
| Abstract: | We study determinants of market organization of local public services by an empiricalexamination of one of the most visible municipal services, residential waste management. Usinga multinomial logit model and data for 1,000 U.S. communities, we explore the effect of politicalinfluence, voter ideology, environmental constraints, production costs (i.e., “economies ofdensity”), and contracting transaction costs on a community’s choice of market arrangement forwaste collection and recycling. We find that cost factors are a significant determinant of servicedelivery method. In contrast, few of the political variables are statistically significant. Theseresults hold for our models of both waste and recycling, lending further evidence to theconclusion that local governments emphasize costs when choosing between private and publicprovision. |
| Keywords: | Market organization, solid waste management, state and local government |
| JEL: | Q20 H70 |
| URL: | http://d.repec.org/n?u=RePEc:rff:dpaper:dp-02-35-rev&r=dcm |