Most papers that employ Differences-in-Differences estimation (DD) use many years of data and focus on serially correlated outcomes but ignore the resulting standard errors are inconsistent. To illustrate severity this issue, we randomly generate placebo laws in state-level female wages from Current Population Survey. For each law, OLS to compute DD estimate its "effect" as well error estimate. These conventional severely understate deviation estimators: find an significant at 5 percent...

We study race in the labor market by sending fictitious resumes to help-wanted ads Boston and Chicago newspapers. To manipulate perceived race, are randomly assigned African-American- or White-sounding names. White names receive 50 percent more callbacks for interviews. Callbacks also responsive resume quality than African-American ones. The racial gap is uniform across occupation, industry, employer size. find little evidence that employers inferring social class from Differential treatment...

Health systems rely on commercial prediction algorithms to identify and help patients with complex health needs. We show that a widely used algorithm, typical of this industry-wide approach affecting millions patients, exhibits significant racial bias: At given risk score, Black are considerably sicker than White as evidenced by signs uncontrolled illnesses. Remedying disparity would increase the percentage receiving additional from 17.7 46.5%. The bias arises because algorithm predicts care...

Much of our understanding corporations builds on the idea that managers, when they are not closely monitored, will pursue goals in shareholders' interests. But what would managers pursue? This paper uses variation corporate governance generated by state adoption antitakeover laws to empirically map out managerial preferences. We use plant‐level data and exploit a unique feature law allows us deal with possible biases associated timing laws. find insulated from takeovers, worker wages...

The poor often behave in less capable ways, which can further perpetuate poverty. We hypothesize that poverty directly impedes cognitive function and present two studies test this hypothesis. First, we experimentally induced thoughts about finances found reduces performance among but not well-off participants. Second, examined the of farmers over planting cycle. same farmer shows diminished before harvest, when poor, as compared with after rich. This cannot be explained by differences time...

The contracting view of CEO pay assumes that is used by shareholders to solve an agency problem. Simple models the predict should not be tied luck, where luck defined as observable shocks performance beyond CEO's control. Using several measures we find in fact responds much a lucky dollar general dollar. A skimming model, has captured pay-setting process, consistent with this fact. Because some complications could also generate for test directly examining effect governance. Consistent...

Owners of business groups are often accused expropriating minority shareholders by tunneling resources from firms where they have low cash flow rights to high rights. In this paper we propose a general methodology measure the extent activities. The rests on isolating and then testing distinctive implications hypothesis for propagation earnings shocks across within group. When apply our data Indian groups, find significant amount tunneling, much it occurring via nonoperating components profit.

Machines are increasingly doing “intelligent” things. Face recognition algorithms use a large dataset of photos labeled as having face or not to estimate function that predicts the presence y from pixels x. This similarity econometrics raises questions: How do these new empirical tools fit with what we know? As economists, how can them? We present way thinking about machine learning gives it its own place in econometric toolbox. Machine only provides tools, solves different problem....

Four main messages emerge from the study of subjective survey data. First, a large experimental literature by and supports economists' skepticism questions. Second, put in an econometric framework, these findings cast serious doubts on attempts to use data as dependent variables, because measurement error appears correlate with set characteristics behaviors. Third, may be useful explanatory variables. Finally, empirical work suggests that variables are practice for explaining differences...

Poor individuals often engage in behaviors, such as excessive borrowing, that reinforce the conditions of poverty. Some explanations for these behaviors focus on personality traits poor. Others emphasize environmental factors housing or financial access. We instead consider how certain stem simply from having less. suggest scarcity changes people allocate attention: It leads them to more deeply some problems while neglecting others. Across several experiments, we show attentional shifts can...

We investigate the market for news under two assumptions: that readers hold beliefs which they like to see confirmed, and newspapers can slant stories toward these beliefs. show that, on topics where share common beliefs, one should not expect accuracy even from competitive media: competition results in lower prices, but slanting reader biases. On diverge (such as politically divisive issues), however, segment extreme positions. Yet aggregate, a with access all sources could get an unbiased...

Investment in scalable, non–price-based behavioral interventions and research may prove valuable improving energy efficiency.

Journal Article Network Effects and Welfare Cultures Get access Marianne Bertrand, Bertrand Princeton University National Bureau of Economic Research Search for other works by this author on: Oxford Academic Google Scholar Erzo F. P. Luttmer, Luttmer Chicago the World Bank Sendhil Mullainathan Massachusetts Institute Technology The Quarterly Economics, Volume 115, Issue 3, August 2000, Pages 1019–1055, https://doi.org/10.1162/003355300554971 Published: 01 2000

We examine how machine learning can be used to improve and understand human decisionmaking.In particular, we focus on a decision that has important policy consequences.Millions of times each year, judges must decide where defendants will await trial-at home or in jail.By law, this hinges the judge's prediction what defendant would do if released.This is promising application because it concrete task for which there large volume data available.Yet comparing algorithm judge proves...

Firms spend billions of dollars developing advertising content, yet there is little field evidence on how much or it affects demand. We analyze a direct mail experiment in South Africa implemented by consumer lender that randomized loan price, and offer deadlines simultaneously. find content significantly Although was difficult to predict ex ante which specific features would matter most this context, the do have large effects. Showing fewer example loans, not suggesting particular use for...

Most empirical policy work focuses on causal inference. We argue an important class of problems does not require inference but instead requires predictive Solving these “prediction problems” more than simple regression techniques, since are tuned to generating unbiased estimates coefficients rather minimizing prediction error. that new developments in the field “machine learning” particularly useful for addressing problems. use example from health illustrate large potential social welfare...

Standard theorizing about poverty falls into two camps. Social scientists regard the behaviors of economically disadvantaged either as calculated adaptations to prevailing circumstances or emanating from a unique culture poverty, rife with deviant values. The first camp presumes that people are highly rational, they hold coherent and justified beliefs pursue their goals effectively, without mistakes, no need for help. second attributes poor variety psychological attitudinal short-fallings...

We provide evidence from field experiments with three different banks that reminder messages increase commitment attainment for clients who recently opened savings accounts. Messages mention both goals and financial incentives are particularly effective, whereas other content variations such as gain versus loss framing do not have significantly effects. Nor we find receiving additional late reminders has an additive effect. These empirical results map neatly into existing models, so a simple...

Journal Article Obtaining a Driver's License in India: An Experimental Approach to Studying Corruption Get access Marianne Bertrand, Bertrand University of Chicago Graduate School Business, National Bureau Economic Research Center for and Policy Research, Institute the Study Labor Search other works by this author on: Oxford Academic Google Scholar Simeon Djankov, Djankov International Finance Corporation Rema Hanna, Hanna New York Wagner Public Service Sendhil Mullainathan Harvard The...

Can machine learning improve human decision making? Bail decisions provide a good test case. Millions of times each year, judges make jail-or-release that hinge on prediction what defendant would do if released. The concreteness the task combined with volume data available makes this promising machine-learning application. Yet comparing algorithm to proves complicated. First, are generated by prior judge decisions. We only observe crime outcomes for released defendants, not those detained....

In order to investigate the impact of limited memory on human behavior, I develop a model grounded in psychological and biological research. assume that people take their memories as accurate use them make inferences. The resulting predicts both over- underreaction but provides enough structure predict when each effect dominates. then this framework study consumption decision. results match empirical work predictability well differences marginal propensity consume from different income...