Search

Recent discussion in the public sphere about algorithmic classification has involved tension between competing notions of what it means for a probabilistic classification to be fair to different groups. We formalize three fairness conditions that lie at the heart of these debates, and we prove that except in highly constrained special cases, there is no method that can satisfy these three conditions simultaneously. Moreover, even satisfying all three conditions approximately requires that the data lie in an approximate version of one of the constrained special cases identified by our theorem. These results suggest some of the ways in which key notions of fairness are incompatible with each other, and hence provide a framework for thinking about the trade-offs between them.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Economics, 2006., Includes bibliographical references., This dissertation presents work empirically investigating various aspects of the criminal justice system. Chapter one, coauthored with Chris Rohlfs, examines the judicial bail-setting process and the defendant decision to pay bail. Optimal bail-setting rules must balance the tradeoffs between costs to defendants and costs to society. This chapter develops a model of optimal bail that incorporates the cost of jailing the defendant, the private cost to the defendant from being incarcerated, the cost of crime, and the costs that arise when defendants abscond. The model is empirically calibrated using data from a randomized experiment. The randomized experiment allows the use of defendants' bail posting decisions to estimate their subjective values of freedom. Our estimates suggest that high-risk defendants would be willing to pay $300 to $1,000 for 90 days of freedom. We find the socially optimal level of bail to be substantially lower than levels currently set by judges. Aggregating nationally, we find that the total social benefit of reform would be on the order of $10 billion per year. Chapter two, coauthored with Marianne Bertrand and Sendhil Mullainathan, is a study of the impact of defendant race on interjudge sentencing disparity, which seeks to add to the knowledge of the role of race in the courtroom., (cont.) This chapter attempts to determine whether the legal system discriminates against minorities by addressing a related question: do judges differ in how they sentence minorities? This approach avoids the difficulty of systematic racial differences in case characteristics by exploiting the random assignment of cases to judges. We measure the between judge variation in the ratio of African-American to White defendant sentence lengths and incarceration rates. In our data set, which includes all felony cases in Cook County, Illinois from 1985-2005, we find large between-judge variation. We also find that judge characteristics, such as age, and the judge's previous work experience as a prosecutor or defender all predict their racial gap in sentencing. Chapter three presents evidence regarding the deterrent effect of incarceration. Knowing the magnitude of the deterrent effect of incarceration on crime is crucial to optimal policy setting. In this chapter I make use of sentence enhancements in gun robbery sentence lengths caused by add-on gun laws to attempt to estimate this impact. Since defendants subject to add-ons would be incarcerated in the absence of the law change, the short-term effect will be solely deterrent., (cont.) I take advantage of the temporal variation in the passage of these laws in different states to identify the causal impact of the law change. I find that add-on gun laws result in a significant reduction in gun robberies, approximately 5% within the first three years of passage, for the average add-on gun law. The results are robust to a number of tests, and do not appear to be due to a large spillover to other types of crime., by David S. Abrams., Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Economics, 2002., Includes bibliographical references., The first essay finds that the market betas of value and small stocks have decreased by about 75% in the second half of the twentieth century. The decline in beta can be related to a long-term improvement in economic conditions that made these companies less risky. The failure to account for time-series variation of beta in unconditional CAPM regressions can explain as much as 30% of the value premium. In some samples, about 80% of the value premium can be explained by assuming that investors tied their expectations of the riskiness of these stocks to the high values of beta prevailing in the early years. Moving from these findings, the second essay (co-authored with Tobias Adrian) explores in detail the relation between the 'value premium' and the decrease in value stocks' beta. We develop an equilibrium model of learning on time-varying risk factor loadings. In the model the CAPM holds from investors' ex-ante perspective. However, the econometrician can observe positive mispricing, whenever the expected beta is above the true level. Given the finding of a decreasing beta, it is likely that investors' expectation of the beta of these stocks has been above the actual level. Therefore, our model can provide an explanation for the 'value premium'. We present the results of simulations in which the model accounts for up to 80% of the 'value premium' in the 1963-2000 sample., (cont.) The third essay analyzes the response of stock returns to earnings information. First, I test the assumption that market expectations of earnings reflect a seasonal random walk, despite the actual process being autoregressive. This hypothesis is rejected. Second, I test the opposite view that expectations are unbiased. The data rejects this possibility for small firms. On the other hand, large firms' prices provide evidence of efficiency. Finally, I show that in the case of small firms the market understates the autoregression coefficient in the earnings process, and it incorrectly assumes that this coefficient is positive, even when actual earnings are seasonal random walks., by Francesco Franzoni., Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Economics, 2002., Includes bibliographical references., The first essay finds that the market betas of value and small stocks have decreased by about 75% in the second half of the twentieth century. The decline in beta can be related to a long-term improvement in economic conditions that made these companies less risky. The failure to account for time-series variation of beta in unconditional CAPM regressions can explain as much as 30% of the value premium. In some samples, about 80% of the value premium can be explained by assuming that investors tied their expectations of the riskiness of these stocks to the high values of beta prevailing in the early years. Moving from these findings, the second essay (co-authored with Tobias Adrian) explores in detail the relation between the 'value premium' and the decrease in value stocks' beta. We develop an equilibrium model of learning on time-varying risk factor loadings. In the model the CAPM holds from investors' ex-ante perspective. However, the econometrician can observe positive mispricing, whenever the expected beta is above the true level. Given the finding of a decreasing beta, it is likely that investors' expectation of the beta of these stocks has been above the actual level. Therefore, our model can provide an explanation for the 'value premium'. We present the results of simulations in which the model accounts for up to 80% of the 'value premium' in the 1963-2000 sample., (cont.) The third essay analyzes the response of stock returns to earnings information. First, I test the assumption that market expectations of earnings reflect a seasonal random walk, despite the actual process being autoregressive. This hypothesis is rejected. Second, I test the opposite view that expectations are unbiased. The data rejects this possibility for small firms. On the other hand, large firms' prices provide evidence of efficiency. Finally, I show that in the case of small firms the market understates the autoregression coefficient in the earnings process, and it incorrectly assumes that this coefficient is positive, even when actual earnings are seasonal random walks., by Francesco Franzoni., Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Economics, 2001., Includes bibliographical references., This dissertation consists of three separate essays. In the first essay, privatization is found to improve bank performance by both better monitoring of existing firms, and improved selection of new bank clients. Data on firms borrowing from government-owned banks, and firms borrowing from a privatized bank suggest that privatization led to an improvement of 1 to 2 percent in the earnings-per-asset of firms financed by the privatized bank. Privatization influenced firms with weak outside options the most. Moreover, bank-debt became more responsive to expected profitability of firms after privatization. Finally, privatization significantly improved the ability of the bank to attract and select more profitable firms, with the newly selected firms having on average, 4 percent higher earnings-per-asset compared to other firms in their cohort. The second essay presents new empirical evidence on differences between domestic and foreign owned private banks in Africa. Domestic banks hold significantly more liquid assets, and are substantially less profitable than foreign banks. However, all of the difference in profitability can be explained by the 2.6% higher interest rate on deposits for domestic banks. Furthermore, domestic banks appear to take on riskier loans, but their return on loans is higher by 2.7%. Foreign banks on the other hand earn significantly more income through bank fees. Finally, domestic banks grow a lot faster than foreign banks with overall economic expansion in the country. A 1% growth in GDP expands domestic bank credit by 1.76% compared to 0.86% for foreign credit., (cont.) The third essay (joint with Daron Acemoglu and Michael Kremer) presents a theoretical analysis of the relative merits of markets, firms, and governments in environments where high powered incentives have both costs and benefits. Firms obscure information about workers' output, thus flattening incentives, and improving efficiency over markets. However in some cases, most notably under common shocks, firms are unable to commit to such a strategy. Under these circumstances, Governments may be able to commit to much flatter wage schedules, and improve the allocation of resources., by Atif Rehman Mian., Ph.D.

Thesis (Ph.D.)--Massachusetts Institute of Technology, Dept. of Economics, 2002., Includes bibliographical references., This dissertation consists of three empirical essays related to information and investment. The first essay asks whether investments in advertising bias the information consumers receive about advertisers' products. Using ratings from two U.S. wine publications, only one of which accepts advertising, I find that advertisers earn differentially higher ratings than non-advertisers and that these higher ratings appear to arise through selective retastings of advertisers' wines before ratings are published. However, the small fraction of wines tasted more than once limits this bias and suggests that reputational considerations may induce publications largely to insulate reviewers from advertisers. The second essay asks whether investments in research by mutual fund families increase the returns of their actively managed funds. Using detailed expense data from 1996-1999, I find evidence that funds recover a fraction of their management fees and brokerage commissions through higher before-expense returns, but that these results are driven by positive relationships between expenses and returns in 1999. I then find more robust positive relationships between aggregate measures of research spending within the family and fund-level returns. Finally, I find that pairwise correlations of fund returns are significantly higher, on average, when both funds belong to the same family, and that these higher correlations are not simply the result of management team overlap. The final essay (joint with Eric Zitzewitz) asks whether the recent rise and fall of U.S. financial markets had real economic consequences in online auctions for fine wine., (cont.) Using data from 63 auctions covering the 30 months from July 1999 through December 2001, we document a strong positive relationship between the number of bottles sold within an auction and fluctuations of the S&P 500 index during that auction. While average prices tend to decline as the number of bottles sold increases, we also present evidence that price changes of the rarest and most famous wines are positively correlated with stock market returns. Overall, our results imply that investors may respond more quickly to short-run changes in stock market wealth than existing theoretical and empirical work has suggested., by Jonathan Michael Reuter., Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Economics, 2003., Includes bibliographical references., This dissertation brings together three essays on the relationship between education, health, and family structure in developing countries. The first essay studies the impact of the AIDS epidemic on children's schooling in Kenya. I draw on the relationship, established in previous literature, between the lack of male circumcision and HIV prevalence. The Luo ethnic group, who does not generally practice male circumcision, had a much larger increase in the HIV prevalence rate between 1993 and 1998. I show that there was a corresponding increase in orphan rates and a decrease in educational achievement among the children in this group. This does not seem to be accounted for by alternative explanations, such as changes in the political clout of the Luo or mean reversion. The second essay examines the impact of sibling sex composition on educational outcomes of children in Tanzania. The estimates suggest that 14-year-old children with three sisters are 24 percentage points less likely to complete primary school (7th grade) after completing 6th grade than children with no sisters, and 8.4 to 9 percentage points less likely to complete primary school overall. Having two or more older sisters, however, can benefit children in completing 4th, 5th, or 6th grade of primary school. These results are robust to the inclusion of parental background characteristics and an index measuring household assets. The third essay, co-authored with Professor Esther Duflo, takes advantage of a school construction program that took place in Indonesia between 1973 and 1978 to estimate the effect of education on fertility and child mortality. Time and region varying exposure to the school construction program generates instrumental variables for the average education in the household, and the difference in education between husband and wife., (cont.) We show that female education is a stronger determinant of age at marriage and early fertility than male education. However, female and male education seem equally important factors in reducing child mortality. We suggest that the OLS estimate of the differential effect of women' s and men's education may be biased by failure to take into account assortative matching., by Lucia Breierova., Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Economics, 2003., Includes bibliographical references., This thesis consists of three distinct essays in public finance economics. In the first I use the presence of minority candidates on the ballot to test two implications of the spatial theory of voting. I find, contrary to theoretical predictions, that the magnitude and types of voters who come out to the polls is responsive to the race of the candidate on the ballot. While both black and white voters turn out in greater numbers when there is a black on the ballot, the whites who are propelled to the polls are more often Republican than not. I use this shift in the electorate caused by the racial mix of the candidates as a test of candidate responsiveness, a second implication of the spatial theory. I find evidence in support of this prediction. In Chapter 2 I seek to understand why thirty-five to forty-five percent of low-income American households do not possess a bank account I demonstrate that the low-income household's banking decision responds to the price of savings accounts, particularly their minimum balance requirements. Despite this, I show that government banking regulation to this point has been ineffective in connecting households to transaction accounts. On the other hand, regulation of the fringe banking market has proved more successful. One approach to covering the uninsured that is frequently advocated by policy makers is subsidizing the employee portion of employer-provided health insurance premiums. In Chapter 3, joint with Jonathan Gruber, we study an example of such subsidies: the introduction of pre-tax premiums for postal employees in 1994, and then for the remaining federal employees in 2000. We find that there is a very small elasticity of insurance takeup with respect to its after-tax price, and a modest elasticity of plan choice. Our results suggest that the federal government did little to improve insurance coverage, but much to increase health care expenditures, through this policy change., by Ebonya Lia Washington., Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Economics, 2001., Includes bibliographical references., When a contract is signed between two economic agents, it is likely to produce some effect on non-contracting, third parties and provide new information to the contracting parties. This thesis examines how such third party externality and newly generated information should affect the initial form of the contract. In the first essay, a headquarters of a firm designs a mechanism with which it extracts the division managers' superior information about the external market opportunities. The information allows the headquarters to provide optimal investment incentive to the managers and make efficient trading decisions. The essay provides a justification for a real-world headquarters, who delegates all the operating decisions to the division managers but maintains the ultimate authority within the firm. The second essay examines how a client would contract with her lawyer to provide best incentive to the lawyer and maximize her return from litigation. By providing different contingent shares based on settlement and judgment, she is able to provide better incentive without diluting her return from litigation. At the same time, when she has relatively poor bargaining leverage against the counter party, the essay shows that delegating the settlement authority to the lawyer and leaving him a large rent would be more beneficial for the client. The third essay analyzes the salary contracting problem faced by the owner of a firm who is aware of a potential opportunity to sell her firm in the future. The essay demonstrates that when the owner grants a large severance payment to the employee, she would be able to defer the compensation burden to the potential buyer and increase her net return from the firm., by Albert H. Choi., Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Economics, 2003., Includes bibliographical references., This dissertation brings together three essays on the relationship between education, health, and family structure in developing countries. The first essay studies the impact of the AIDS epidemic on children's schooling in Kenya. I draw on the relationship, established in previous literature, between the lack of male circumcision and HIV prevalence. The Luo ethnic group, who does not generally practice male circumcision, had a much larger increase in the HIV prevalence rate between 1993 and 1998. I show that there was a corresponding increase in orphan rates and a decrease in educational achievement among the children in this group. This does not seem to be accounted for by alternative explanations, such as changes in the political clout of the Luo or mean reversion. The second essay examines the impact of sibling sex composition on educational outcomes of children in Tanzania. The estimates suggest that 14-year-old children with three sisters are 24 percentage points less likely to complete primary school (7th grade) after completing 6th grade than children with no sisters, and 8.4 to 9 percentage points less likely to complete primary school overall. Having two or more older sisters, however, can benefit children in completing 4th, 5th, or 6th grade of primary school. These results are robust to the inclusion of parental background characteristics and an index measuring household assets. The third essay, co-authored with Professor Esther Duflo, takes advantage of a school construction program that took place in Indonesia between 1973 and 1978 to estimate the effect of education on fertility and child mortality. Time and region varying exposure to the school construction program generates instrumental variables for the average education in the household, and the difference in education between husband and wife., (cont.) We show that female education is a stronger determinant of age at marriage and early fertility than male education. However, female and male education seem equally important factors in reducing child mortality. We suggest that the OLS estimate of the differential effect of women' s and men's education may be biased by failure to take into account assortative matching., by Lucia Breierova., Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Economics, 2002., Includes bibliographical references., The first chapter of this dissertation examines the behavior of long- and short-term capital gains distributions after the passage of the Tax Reform Act of 1997, which lowered the maximum tax rate on long-term gains. Using a panel of mutual fund data, I examine the ratio of long-term to total gains distributions around the time of the Act, and find that fund managers appear to tilttheir distributions towards the long-term after 1997. This behavior is consistent with the hypothesis that managers are tax-sensitive, and the estimates are robust to the inclusion of fund-level fixed effects and other controls. I also examine fund capital gains patterns in a difference-in-differences framework, comparing actively managed to index funds; this technique gives a lower-bound estimate of the increase in the fraction of gains paid-out as long-term after the Act of five percentage points. The second chapter examines equity mutual fund dividend yields. Dividends are highly-taxed for individual investors, but tax-favored for corporate investors. Consequently, I hypothesize that corporate investors may prefer to hold higher-dividend yield funds than non-corporate investors. I use institutional funds as a proxy for corporate, trust, or non-profit shareholders, and find that these funds do have systematically higher dividend yields than their retail counterparts. These results are consistent with the tax clientele hypothesis, and are robust to the inclusion of a number of fund- and industry-level controls., (cont.) Chapter three, co-authored with James Poterba, documents the increasing use of redemption fees throughout the mutual fund industry. These fees are levied against investors who liquidate their positions before a specified time, and are paid back into the fund to compensate existing investors for the negative externalities of redemptions. We find that foreign and sector funds are more likely to impose redemption fees, and that institutional funds are less likely to have fees. Using data from the SEC, we find that the largest funds with redemption fees collected nearly $58 million in fees in 2000. We also find that funds with redemption fees appear to have lower turnover and higher returns than their no-fee counterparts, controlling for time, fund objective, and other characteristics., by Stephanie L. Plancich., Ph.D.

Thesis (Ph.D.)--Massachusetts Institute of Technology, Dept. of Economics, 2002., Includes bibliographical references (p. 120-124)., This dissertation consists of three essays on the local regulation of housing supply and its influence on house price volatility. I begin by describing the local supply-side dynamics of housing price and stock in each of 47 U.S. metropolitan area housing markets using a unique, market-level panel dataset. The data are analyzed with a conditional vector-autoregression, which characterizes the dynamic responses of price and stock to an increase in housing demand caused by a shift in employment. These response time-paths are used to create measures of short-, medium- and long-run supply elasticities. Both the time-paths and the implied elasticities vary widely. I use several area characteristics to explain the variation in the supply elasticity measures across metropolitan areas. The results suggest that governments with a greater incentive to regulate housing have a slower house-price response. This directly contradicts the predictions of the existing land-use literature. In order to explain the result, William C. Wheaton and I formalize a general equilibrium theory of a housing market whose local governments determine house price by regulating their supplies of land. The regulatory decisions are made by current residents who already own housing, but the impact of these decisions on prices is determined by new entrants who must purchase housing. The choice of how much to regulate is shown to vary by town size, existing town density, and the amount of open land currently available for development. This is broadly consistent with the results of recent empirical research., (cont.) I also find that the predictions of this theory are directly borne out in data on land-use restrictions across much of Massachusetts. Two types of residential land-use regulation are considered: regulation that restricts the share of land on which housing can be built, and regulation that restricts housing density on a given plot of land. I find evidence suggesting that the two types of regulation may enter the optimization problem of Massachusetts towns very differently, and that towns may regulate new housing more strictly than they did old housing., by Bengte Evenson., Ph.D.

Thesis (Ph.D.)--Massachusetts Institute of Technology, Dept. of Economics, 2002., Includes bibliographical references., This dissertation consists of three empirical essays related to information and investment. The first essay asks whether investments in advertising bias the information consumers receive about advertisers' products. Using ratings from two U.S. wine publications, only one of which accepts advertising, I find that advertisers earn differentially higher ratings than non-advertisers and that these higher ratings appear to arise through selective retastings of advertisers' wines before ratings are published. However, the small fraction of wines tasted more than once limits this bias and suggests that reputational considerations may induce publications largely to insulate reviewers from advertisers. The second essay asks whether investments in research by mutual fund families increase the returns of their actively managed funds. Using detailed expense data from 1996-1999, I find evidence that funds recover a fraction of their management fees and brokerage commissions through higher before-expense returns, but that these results are driven by positive relationships between expenses and returns in 1999. I then find more robust positive relationships between aggregate measures of research spending within the family and fund-level returns. Finally, I find that pairwise correlations of fund returns are significantly higher, on average, when both funds belong to the same family, and that these higher correlations are not simply the result of management team overlap. The final essay (joint with Eric Zitzewitz) asks whether the recent rise and fall of U.S. financial markets had real economic consequences in online auctions for fine wine., (cont.) Using data from 63 auctions covering the 30 months from July 1999 through December 2001, we document a strong positive relationship between the number of bottles sold within an auction and fluctuations of the S&P 500 index during that auction. While average prices tend to decline as the number of bottles sold increases, we also present evidence that price changes of the rarest and most famous wines are positively correlated with stock market returns. Overall, our results imply that investors may respond more quickly to short-run changes in stock market wealth than existing theoretical and empirical work has suggested., by Jonathan Michael Reuter., Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Economics, 2003., Includes bibliographical references., This thesis consists of three distinct essays in public finance economics. In the first I use the presence of minority candidates on the ballot to test two implications of the spatial theory of voting. I find, contrary to theoretical predictions, that the magnitude and types of voters who come out to the polls is responsive to the race of the candidate on the ballot. While both black and white voters turn out in greater numbers when there is a black on the ballot, the whites who are propelled to the polls are more often Republican than not. I use this shift in the electorate caused by the racial mix of the candidates as a test of candidate responsiveness, a second implication of the spatial theory. I find evidence in support of this prediction. In Chapter 2 I seek to understand why thirty-five to forty-five percent of low-income American households do not possess a bank account I demonstrate that the low-income household's banking decision responds to the price of savings accounts, particularly their minimum balance requirements. Despite this, I show that government banking regulation to this point has been ineffective in connecting households to transaction accounts. On the other hand, regulation of the fringe banking market has proved more successful. One approach to covering the uninsured that is frequently advocated by policy makers is subsidizing the employee portion of employer-provided health insurance premiums. In Chapter 3, joint with Jonathan Gruber, we study an example of such subsidies: the introduction of pre-tax premiums for postal employees in 1994, and then for the remaining federal employees in 2000. We find that there is a very small elasticity of insurance takeup with respect to its after-tax price, and a modest elasticity of plan choice. Our results suggest that the federal government did little to improve insurance coverage, but much to increase health care expenditures, through this policy change., by Ebonya Lia Washington., Ph.D.