Observers have long been ambivalent about democratic representation by parties with a regionalist... more Observers have long been ambivalent about democratic representation by parties with a regionalist orientation. We estimate the causal effect of regionalist party representation on political violence in India and find regionalist parties who win power cause local violence, but not through increased conflict with the Center. Rather, successful regionalists tend to favor local ethnic majorities, causing heightened uncertainty for local minority groups. In particular, we show that the increased violence is explained entirely by electoral constituencies with significant tribal populations but no mandated political representation for tribes, and that regionalist parties further decrease local tribal persons' reported consumption of television and radio, cultural goods, and wages. Our results imply representation by local majorities further relegates local tribal minorities to the margins of society, and that they in turn respond with organized political violence.
We exploit the introduction of pedestrian countdown signals—timers that indicate when traffic lig... more We exploit the introduction of pedestrian countdown signals—timers that indicate when traffic lights will change—to evaluate a policy that improves the information of all market participants. We find that although countdown signals reduce the number of pedestrians struck by automobiles, they increase the number of collisions between automobiles. They also cause more collisions overall, implying that welfare gains can be attained by hiding the information from drivers. Whereas most empirical studies on the role of information in markets suggest that asymmetric information reduces welfare, we conclude that asymmetric information can, in fact, improve it.
We modify a simple textbook model of driver interaction, where drivers can choose to act aggressi... more We modify a simple textbook model of driver interaction, where drivers can choose to act aggressively or cautiously, to rationalize the finding that informing drivers about light changes causes more driver-driver collisions. The modification is that we allow for uncertainty in what drivers know about the time until a light change at an intersection. We show that under rather innocuous assumptions, equilibrium collision probabilities are larger when drivers know the time that remains. The result follows from two simple features of equilibrium behavior in our model. First, aggressive behavior on the part of informed drivers is more likely when less time remains. Second, drivers are most responsive to information about light changes when they learn the time that remains is less than they expected. Taken together, the features of equilibrium behavior imply that informing drivers increases the chances of a collision.
Journal of Economic Behavior & Organization, 2019
Most studies analyzing worker adjustments to discrimination focus on sorting decisions, such as o... more Most studies analyzing worker adjustments to discrimination focus on sorting decisions, such as occupations workers pursue. We instead analyze on the job adjustments, focusing on the effects of discrimination by customers and managers. Specifically, we use observational and experimental data from large-scale restaurants to investigate the effects of the symmetry of waiter facial attributes on the trade offs they make, the extent to which the trade offs are shaped by consumer preference for symmetry, and whether the trade offs affect how they are treated by managers. We find customers have a preference for symmetry and that preferred workers consequently give them less personal attention. This allows preferred workers to serve more customers, generate more revenue for the firm, and more earnings for themselves. The large ensuing productivity and earnings differentials are facilitated largely by small differences in managerial treatment of preferred workers.
Observers have long been ambivalent about democratic representation by parties with a regionalist... more Observers have long been ambivalent about democratic representation by parties with a regionalist orientation. We estimate the causal effect of regionalist party representation on political violence in India and find regionalist parties who win power cause local violence, but not through increased conflict with the Center. Rather, successful regionalists tend to favor local ethnic majorities, causing heightened uncertainty for local minority groups. In particular, we show that the increased violence is explained entirely by electoral constituencies with significant tribal populations but no mandated political representation for tribes, and that regionalist parties further decrease local tribal persons' reported consumption of television and radio, cultural goods, and wages. Our results imply representation by local majorities further relegates local tribal minorities to the margins of society, and that they in turn respond with organized political violence.
We estimate the causal impact of citizen candidates on political representation in India. To do t... more We estimate the causal impact of citizen candidates on political representation in India. To do this we exploit exogenous changes in the entry deposit candidates pay for their participation in the political process, changes that disproportionately excluded candidates with no affiliation to established political parties. A one standard deviation increase in the number of citizen candidates increases voter turnout by more than 5 percentage points, as voters who have a favored candidate choose to vote rather than stay home. The vote share of citizen candidates increases by 9 percentage points, as some existing voters switch who they vote for. These results imply citizen candidates improve the representation of voter preferences at the ballot box. We show further that citizen candidates allow winning candidates to win with less vote share, and thus decrease the probability that their constituency's elected candidate is a member of government by about 28 percentage points. This implies that there is a sizeable tradeoff between better preference representation at the ballot box and better constituency representation in government.
This paper extends the Euler Equation (EE) representation of dynamic decision problems to a gener... more This paper extends the Euler Equation (EE) representation of dynamic decision problems to a general class of discrete choice models and shows that the advantages of this approach apply not only to the estimation of structural parameters but also to the computation of a solution and to the evaluation of counterfactual experiments. We use a choice probabilities representation of the discrete decision problem to derive marginal conditions of optimality with the same features as the standard EEs in continuous decision problems. These EEs imply a …xed point mapping in the space of conditional choice values, that we denote the Euler equation-value (EE-value) operator. We show that, in contrast to Euler equation operators in continuous decision models, this operator is a contraction. We present numerical examples that illustrate how solving the model by iterating in the EE-value mapping implies substantial computational savings relative to iterating in the Bellman equation (that requires a much larger number of iterations) or in the policy function (that involves a costly valuation step). We de…ne a sample version of the EE-value operator and use it to construct a sequence of consistent estimators of the structural parameters, and to evaluate counterfactual experiments. The computational cost of evaluating this sample-based EE-value operator increases linearly with sample size, and provides an unbiased (in …nite samples) and consistent estimator the counterfactual. As such there is no curse of dimensionality in the consistent estimation of the model and in the evaluation of counterfactual experiments. We illustrate the computational gains of our methods using several Monte Carlo experiments.
/countdown-signals-traffic-accidents/ Credit: Kat Northern Lights Man (Creative Commons BY NC) Fu... more /countdown-signals-traffic-accidents/ Credit: Kat Northern Lights Man (Creative Commons BY NC) Full transparency is not always a good thing: countdown signals at intersections make pedestrians safer, but also lead to more collisions. No matter people's major mode of transportation, they are almost certain to be a pedestrian at some stage of the journey. Having better information in busy urban areas, such as countdowns at intersections, can help to make being a pedestrian safer. Using Toronto as a case study, Sacha Kapoor and Arvind Magesan investigate the effects of introducing pedestrian countdowns on pedestrian and automobile safety. They find that while countdowns do make pedestrians safer, they also can lead to an increase in automobile accidents, as they facilitate more aggressive driving. They argue that in these sorts of contexts, full transparency of local information may not necessarily lead to the best outcomes for all.
In this paper, we derive marginal conditions of optimality (i.e., Euler equations) for a general ... more In this paper, we derive marginal conditions of optimality (i.e., Euler equations) for a general class of dynamic discrete choice structural models, including single-agent models and dynamic games. We show that these equations de…ne a policy iteration mapping in the space of players' strategies (i.e., players'choice probabilities) such that the vector of equilibrium strategies is a …xed point of this mapping. We use this Euler Equations-Policy Iteration (EE-PI) mapping to de…ne two-step (or K-step) GMM estimators of structural parameters, and estimators of counterfactual experiments using the estimated model. The Euler Equations-Policy Iteration mapping and the estimators built on it have important advantages with respect to the standard policy iteration mapping. First, the evaluation of the EE-PI mapping does not involve the computation of in…nite period forward presents values, but only expectations of payo¤s a few periods forward, i.e., N periods forward in a game with N heterogeneous players. This property results into substantial computational savings. Second, GMM estimators using a sample-based EE-PI mapping are consistent and are not subject to the curse of dimensionality in the computation of present values. These estimators do not su¤er of an asymptotic bias associated to the approximation of value functions. We present Monte Carlo experiments to illustrate the computational gains of these methods.
The development of two-step econometric methods for dynamic structural models has afforded resear... more The development of two-step econometric methods for dynamic structural models has afforded researchers the ability to estimate models with large state spaces without having to compute a full solution of the model even once. However, regardless of the method used for estimation, the implementation of counterfactual experiments using the estimated model does require the full solution and thus still faces the well-known curse of dimensionality in the solution of dynamic programming models. This paper proposes an approach to compute consistent estimates of counterfactual experiments that avoids the full solution of the model and breaks the curse of dimensionality. We illustrate the computational gains associated with our model and methods using Monte Carlo experiments. Finally we illustrate our method using real data, by studying the effects of a counterfactual increase in the cost of entry in several industries in Chile.
We estimate the causal effect of independent candidates on voter turnout and election outcomes in... more We estimate the causal effect of independent candidates on voter turnout and election outcomes in India. To do this, we exploit exogenous changes in the entry deposit candidates pay for their participation in the political process, changes that disproportionately excluded candidates with no affiliation to established political parties. A one standard deviation increase in the number of independent candidates increases voter turnout by more than 6 percentage points, as some voters choose to vote rather than stay home. The vote share of independent candidates increases by more than 10 percentage points, as some existing voters switch who they vote for. Thus, independents allow winning candidates to win with less vote share, decrease the probability of electing a candidate from the governing coalition by about 31 percentage points, and ultimately increase the probability of electing an ethnic-party candidate. Altogether, the results imply that the price of participation by independents...
This paper considers the problem of an imperfectly informed regulator constrained in his choice o... more This paper considers the problem of an imperfectly informed regulator constrained in his choice of environmental regulation by the political opposition of those affected by the policy. We compare the value of two types of information to the regulator: the social cost of pollution and the profitability of firms present in the economy. We find that in environments where small increases in the losses to regulated firms greatly affect the regulator's ability to implement the policy, it is most valuable to learn the types of firms whereas it is most valuable to learn the social cost of pollution when small increases in losses are relatively ineffectual.
ABSTRACT Most studies analyzing the adjustments of workers to discrimination focus on sorting dec... more ABSTRACT Most studies analyzing the adjustments of workers to discrimination focus on sorting decisions, such as occupations workers pursue. We instead analyze on-the-job adjustments, focusing on the effects of discrimination by consumers. Specifically, using data from a couple of large-scale restaurants, we investigate the effects of an outward yet immutable physical trait - symmetry of the facial attributes of workers - on trade offs workers make, and the extent to which the trade offs are shaped by consumer preference for the trait. Large-scale restaurants are well-suited for studying these issues because, as with many jobs in the services sector, workers must trade off quality of service for the quantity of consumers they serve. Using a combination of observational data and data generated by a field experiment, we find consumers have a preference for the trait and that preferred workers deliver lower service quality. Instead they specialize in serving more consumers. The findings imply that when outward physical traits substitute for service quality in consumer preferences, preferred workers specialize in tasks having no services component because consumers punish them less for poor performance. Our evidence further implies preferred workers are better off, as they earn the same as their coworkers despite working fewer hours. We conclude that consumer discrimination shapes comparative advantage and, by doing so, generates inequality in the workplace.
This thesis collects three papers that study applied problems in economics dealing with dynamic s... more This thesis collects three papers that study applied problems in economics dealing with dynamic strategic behavior and imperfect information. In the first chapter I study the relationship between participation in United Nations Human Rights Treaties (HRT), foreign aid receipts and domestic human rights institutions. I provide empirical evidence that countries with relatively
We exploit the introduction of pedestrian countdown signals—timers that indicate when traffic lig... more We exploit the introduction of pedestrian countdown signals—timers that indicate when traffic lights will change—to evaluate a policy that improves the information of all market participants. We find that although countdown signals reduce the number of pedestrians struck by automobiles, they increase the number of collisions between automobiles. They also cause more collisions overall, implying that welfare gains can be attained by hiding the information from drivers. Whereas most empirical studies on the role of information in markets suggest that asymmetric information reduces welfare, we conclude that asymmetric information can, in fact, improve it. (JEL D82, D83, R41)
Observers have long been ambivalent about democratic representation by parties with a regionalist... more Observers have long been ambivalent about democratic representation by parties with a regionalist orientation. We estimate the causal effect of regionalist party representation on political violence in India and find regionalist parties who win power cause local violence, but not through increased conflict with the Center. Rather, successful regionalists tend to favor local ethnic majorities, causing heightened uncertainty for local minority groups. In particular, we show that the increased violence is explained entirely by electoral constituencies with significant tribal populations but no mandated political representation for tribes, and that regionalist parties further decrease local tribal persons' reported consumption of television and radio, cultural goods, and wages. Our results imply representation by local majorities further relegates local tribal minorities to the margins of society, and that they in turn respond with organized political violence.
We exploit the introduction of pedestrian countdown signals—timers that indicate when traffic lig... more We exploit the introduction of pedestrian countdown signals—timers that indicate when traffic lights will change—to evaluate a policy that improves the information of all market participants. We find that although countdown signals reduce the number of pedestrians struck by automobiles, they increase the number of collisions between automobiles. They also cause more collisions overall, implying that welfare gains can be attained by hiding the information from drivers. Whereas most empirical studies on the role of information in markets suggest that asymmetric information reduces welfare, we conclude that asymmetric information can, in fact, improve it.
We modify a simple textbook model of driver interaction, where drivers can choose to act aggressi... more We modify a simple textbook model of driver interaction, where drivers can choose to act aggressively or cautiously, to rationalize the finding that informing drivers about light changes causes more driver-driver collisions. The modification is that we allow for uncertainty in what drivers know about the time until a light change at an intersection. We show that under rather innocuous assumptions, equilibrium collision probabilities are larger when drivers know the time that remains. The result follows from two simple features of equilibrium behavior in our model. First, aggressive behavior on the part of informed drivers is more likely when less time remains. Second, drivers are most responsive to information about light changes when they learn the time that remains is less than they expected. Taken together, the features of equilibrium behavior imply that informing drivers increases the chances of a collision.
Journal of Economic Behavior & Organization, 2019
Most studies analyzing worker adjustments to discrimination focus on sorting decisions, such as o... more Most studies analyzing worker adjustments to discrimination focus on sorting decisions, such as occupations workers pursue. We instead analyze on the job adjustments, focusing on the effects of discrimination by customers and managers. Specifically, we use observational and experimental data from large-scale restaurants to investigate the effects of the symmetry of waiter facial attributes on the trade offs they make, the extent to which the trade offs are shaped by consumer preference for symmetry, and whether the trade offs affect how they are treated by managers. We find customers have a preference for symmetry and that preferred workers consequently give them less personal attention. This allows preferred workers to serve more customers, generate more revenue for the firm, and more earnings for themselves. The large ensuing productivity and earnings differentials are facilitated largely by small differences in managerial treatment of preferred workers.
Observers have long been ambivalent about democratic representation by parties with a regionalist... more Observers have long been ambivalent about democratic representation by parties with a regionalist orientation. We estimate the causal effect of regionalist party representation on political violence in India and find regionalist parties who win power cause local violence, but not through increased conflict with the Center. Rather, successful regionalists tend to favor local ethnic majorities, causing heightened uncertainty for local minority groups. In particular, we show that the increased violence is explained entirely by electoral constituencies with significant tribal populations but no mandated political representation for tribes, and that regionalist parties further decrease local tribal persons' reported consumption of television and radio, cultural goods, and wages. Our results imply representation by local majorities further relegates local tribal minorities to the margins of society, and that they in turn respond with organized political violence.
We estimate the causal impact of citizen candidates on political representation in India. To do t... more We estimate the causal impact of citizen candidates on political representation in India. To do this we exploit exogenous changes in the entry deposit candidates pay for their participation in the political process, changes that disproportionately excluded candidates with no affiliation to established political parties. A one standard deviation increase in the number of citizen candidates increases voter turnout by more than 5 percentage points, as voters who have a favored candidate choose to vote rather than stay home. The vote share of citizen candidates increases by 9 percentage points, as some existing voters switch who they vote for. These results imply citizen candidates improve the representation of voter preferences at the ballot box. We show further that citizen candidates allow winning candidates to win with less vote share, and thus decrease the probability that their constituency's elected candidate is a member of government by about 28 percentage points. This implies that there is a sizeable tradeoff between better preference representation at the ballot box and better constituency representation in government.
This paper extends the Euler Equation (EE) representation of dynamic decision problems to a gener... more This paper extends the Euler Equation (EE) representation of dynamic decision problems to a general class of discrete choice models and shows that the advantages of this approach apply not only to the estimation of structural parameters but also to the computation of a solution and to the evaluation of counterfactual experiments. We use a choice probabilities representation of the discrete decision problem to derive marginal conditions of optimality with the same features as the standard EEs in continuous decision problems. These EEs imply a …xed point mapping in the space of conditional choice values, that we denote the Euler equation-value (EE-value) operator. We show that, in contrast to Euler equation operators in continuous decision models, this operator is a contraction. We present numerical examples that illustrate how solving the model by iterating in the EE-value mapping implies substantial computational savings relative to iterating in the Bellman equation (that requires a much larger number of iterations) or in the policy function (that involves a costly valuation step). We de…ne a sample version of the EE-value operator and use it to construct a sequence of consistent estimators of the structural parameters, and to evaluate counterfactual experiments. The computational cost of evaluating this sample-based EE-value operator increases linearly with sample size, and provides an unbiased (in …nite samples) and consistent estimator the counterfactual. As such there is no curse of dimensionality in the consistent estimation of the model and in the evaluation of counterfactual experiments. We illustrate the computational gains of our methods using several Monte Carlo experiments.
/countdown-signals-traffic-accidents/ Credit: Kat Northern Lights Man (Creative Commons BY NC) Fu... more /countdown-signals-traffic-accidents/ Credit: Kat Northern Lights Man (Creative Commons BY NC) Full transparency is not always a good thing: countdown signals at intersections make pedestrians safer, but also lead to more collisions. No matter people's major mode of transportation, they are almost certain to be a pedestrian at some stage of the journey. Having better information in busy urban areas, such as countdowns at intersections, can help to make being a pedestrian safer. Using Toronto as a case study, Sacha Kapoor and Arvind Magesan investigate the effects of introducing pedestrian countdowns on pedestrian and automobile safety. They find that while countdowns do make pedestrians safer, they also can lead to an increase in automobile accidents, as they facilitate more aggressive driving. They argue that in these sorts of contexts, full transparency of local information may not necessarily lead to the best outcomes for all.
In this paper, we derive marginal conditions of optimality (i.e., Euler equations) for a general ... more In this paper, we derive marginal conditions of optimality (i.e., Euler equations) for a general class of dynamic discrete choice structural models, including single-agent models and dynamic games. We show that these equations de…ne a policy iteration mapping in the space of players' strategies (i.e., players'choice probabilities) such that the vector of equilibrium strategies is a …xed point of this mapping. We use this Euler Equations-Policy Iteration (EE-PI) mapping to de…ne two-step (or K-step) GMM estimators of structural parameters, and estimators of counterfactual experiments using the estimated model. The Euler Equations-Policy Iteration mapping and the estimators built on it have important advantages with respect to the standard policy iteration mapping. First, the evaluation of the EE-PI mapping does not involve the computation of in…nite period forward presents values, but only expectations of payo¤s a few periods forward, i.e., N periods forward in a game with N heterogeneous players. This property results into substantial computational savings. Second, GMM estimators using a sample-based EE-PI mapping are consistent and are not subject to the curse of dimensionality in the computation of present values. These estimators do not su¤er of an asymptotic bias associated to the approximation of value functions. We present Monte Carlo experiments to illustrate the computational gains of these methods.
The development of two-step econometric methods for dynamic structural models has afforded resear... more The development of two-step econometric methods for dynamic structural models has afforded researchers the ability to estimate models with large state spaces without having to compute a full solution of the model even once. However, regardless of the method used for estimation, the implementation of counterfactual experiments using the estimated model does require the full solution and thus still faces the well-known curse of dimensionality in the solution of dynamic programming models. This paper proposes an approach to compute consistent estimates of counterfactual experiments that avoids the full solution of the model and breaks the curse of dimensionality. We illustrate the computational gains associated with our model and methods using Monte Carlo experiments. Finally we illustrate our method using real data, by studying the effects of a counterfactual increase in the cost of entry in several industries in Chile.
We estimate the causal effect of independent candidates on voter turnout and election outcomes in... more We estimate the causal effect of independent candidates on voter turnout and election outcomes in India. To do this, we exploit exogenous changes in the entry deposit candidates pay for their participation in the political process, changes that disproportionately excluded candidates with no affiliation to established political parties. A one standard deviation increase in the number of independent candidates increases voter turnout by more than 6 percentage points, as some voters choose to vote rather than stay home. The vote share of independent candidates increases by more than 10 percentage points, as some existing voters switch who they vote for. Thus, independents allow winning candidates to win with less vote share, decrease the probability of electing a candidate from the governing coalition by about 31 percentage points, and ultimately increase the probability of electing an ethnic-party candidate. Altogether, the results imply that the price of participation by independents...
This paper considers the problem of an imperfectly informed regulator constrained in his choice o... more This paper considers the problem of an imperfectly informed regulator constrained in his choice of environmental regulation by the political opposition of those affected by the policy. We compare the value of two types of information to the regulator: the social cost of pollution and the profitability of firms present in the economy. We find that in environments where small increases in the losses to regulated firms greatly affect the regulator's ability to implement the policy, it is most valuable to learn the types of firms whereas it is most valuable to learn the social cost of pollution when small increases in losses are relatively ineffectual.
ABSTRACT Most studies analyzing the adjustments of workers to discrimination focus on sorting dec... more ABSTRACT Most studies analyzing the adjustments of workers to discrimination focus on sorting decisions, such as occupations workers pursue. We instead analyze on-the-job adjustments, focusing on the effects of discrimination by consumers. Specifically, using data from a couple of large-scale restaurants, we investigate the effects of an outward yet immutable physical trait - symmetry of the facial attributes of workers - on trade offs workers make, and the extent to which the trade offs are shaped by consumer preference for the trait. Large-scale restaurants are well-suited for studying these issues because, as with many jobs in the services sector, workers must trade off quality of service for the quantity of consumers they serve. Using a combination of observational data and data generated by a field experiment, we find consumers have a preference for the trait and that preferred workers deliver lower service quality. Instead they specialize in serving more consumers. The findings imply that when outward physical traits substitute for service quality in consumer preferences, preferred workers specialize in tasks having no services component because consumers punish them less for poor performance. Our evidence further implies preferred workers are better off, as they earn the same as their coworkers despite working fewer hours. We conclude that consumer discrimination shapes comparative advantage and, by doing so, generates inequality in the workplace.
This thesis collects three papers that study applied problems in economics dealing with dynamic s... more This thesis collects three papers that study applied problems in economics dealing with dynamic strategic behavior and imperfect information. In the first chapter I study the relationship between participation in United Nations Human Rights Treaties (HRT), foreign aid receipts and domestic human rights institutions. I provide empirical evidence that countries with relatively
We exploit the introduction of pedestrian countdown signals—timers that indicate when traffic lig... more We exploit the introduction of pedestrian countdown signals—timers that indicate when traffic lights will change—to evaluate a policy that improves the information of all market participants. We find that although countdown signals reduce the number of pedestrians struck by automobiles, they increase the number of collisions between automobiles. They also cause more collisions overall, implying that welfare gains can be attained by hiding the information from drivers. Whereas most empirical studies on the role of information in markets suggest that asymmetric information reduces welfare, we conclude that asymmetric information can, in fact, improve it. (JEL D82, D83, R41)
Uploads
Papers by Arvind Magesan