HUMAN DECISIONS AND MACHINE PREDICTIONS Kleinberg, Jon; Lakkaraju, Himabindu; Leskovec, Jure ...
The Quarterly journal of economics,
02/2018, Letnik:
133, Številka:
1
Journal Article
Recenzirano
Odprti dostop
Can machine learning improve human decision making? Bail decisions provide a good test case. Millions of times each year, judges make jail-or-release decisions that hinge on a prediction of what a ...defendant would do if released. The concreteness of the prediction task combined with the volume of data available makes this a promising machine-learning application. Yet comparing the algorithm to judges proves complicated. First, the available data are generated by prior judge decisions. We only observe crime outcomes for released defendants, not for those judges detained. This makes it hard to evaluate counterfactual decision rules based on algorithmic predictions. Second, judges may have a broader set of preferences than the variable the algorithm predicts; for instance, judges may care specifically about violent crimes or about racial inequities. We deal with these problems using different econometric strategies, such as quasi-random assignment of cases to judges. Even accounting for these concerns, our results suggest potentially large welfare gains: one policy simulation shows crime reductions up to 24.7% with no change in jailing rates, or jailing rate reductions up to 41.9% with no increase in crime rates. Moreover, all categories of crime, including violent crimes, show reductions; these gains can be achieved while simultaneously reducing racial disparities. These results suggest that while machine learning can be valuable, realizing this value requires integrating these tools into an economic framework: being clear about the link between predictions and decisions; specifying the scope of payoff functions; and constructing unbiased decision counterfactuals.
Abstract
We use large-scale federal survey data linked to administrative death records to investigate the relationship between Medicaid enrollment and mortality. Our analysis compares changes in ...mortality for near-elderly adults in states with and without Affordable Care Act Medicaid expansions. We identify adults most likely to benefit using survey information on socioeconomic status, citizenship status, and public program participation. We find that prior to the ACA expansions, mortality rates across expansion and nonexpansion states trended similarly, but beginning in the first year of the policy, there were significant reductions in mortality in states that opted to expand relative to nonexpanders. Individuals in expansion states experienced a 0.132 percentage point decline in annual mortality, a 9.4% reduction over the sample mean, as a result of the Medicaid expansions. The effect is driven by a reduction in disease-related deaths and grows over time. A variety of alternative specifications, methods of inference, placebo tests, and sample definitions confirm our main result.
We propose a new approach to studying the pass-through of credit expansion policies that focuses on frictions, such as asymmetric information, that arise in the interaction between banks and ...borrowers. We decompose the effect of changes in banks’ cost of funds on aggregate borrowing into the product of banks’ marginal propensity to lend (MPL) to borrowers and those borrowers’ marginal propensity to borrow (MPB), aggregated over all borrowers in the economy. We apply our framework by estimating heterogeneous MPBs and MPLs in the U.S. credit card market. Using panel data on 8.5 million credit cards and 743 credit limit regression discontinuities, we find that the MPB is declining in credit score, falling from 59% for consumers with FICO scores below 660 to essentially zero for consumers with FICO scores above 740. We use a simple model of optimal credit limits to show that a bank’s MPL depends on a small number of parameters that can be estimated using our credit limit discontinuities. For the lowest FICO score consumers, higher credit limits sharply reduce profits from lending, limiting banks’ optimal MPL to these consumers. The negative correlation between MPB and MPL reduces the impact of changes in banks’ cost of funds on aggregate household borrowing, and highlights the importance of frictions in bank-borrower interactions for understanding the pass-through of credit expansions.
Abstract
Using data on bank-firm relationships in Italy during the Eurozone financial crisis, we show that: (i) compared to healthy banks, under-capitalised banks cut credit to healthy but not to ...zombie firms and are more likely to prolong a credit relationship with a zombie; (ii) in area sectors with more low-capital banks, zombies are more likely to survive; (iii) bank under-capitalisation does not hurt the growth rate of healthy firms. We provide evidence that extending credit to the weakest firms during the recession mitigated the disruption of supply chains and adverse local demand externalities.
This research note aims to enrich our understanding regarding the market valuation implications of financial reporting under an Integrated Reporting (IR) approach. In order to do so, we focus on the ...Johannesburg Stock Exchange (JSE) and we examine whether the value relevance of summary accounting information (i.e., book value of equity and earnings) of firms listed on the JSE has enhanced after the mandatory adoption of an IR approach under the King III Report. Our study can be seen as a response to the recent calls for a closer investigation of the usefulness of the new reporting trend for investors. More specifically, our study can be seen as a response to the stance taken by the International Integrated Reporting Council (IIRC) Framework that the adoption of an IR approach improves the usefulness of financial reporting for investors. For our empirical tests we utilize a sample of 954firm-year observations and employ a linear price-level model which associates a firm’s market value of equity with its book value of equity and earnings. In line with the IIRC Framework’s expectations, we find strong evidence of a sharp increase of the earnings’ valuation coefficient. However, contrary to the Framework’s stance, our results indicate a decline in the value relevance of net assets. Such a decline may be imputed to risks and/or unbooked liabilities that are revealed or measured more reliably after the introduction of an IR approach on the JSE. It should be noted, however, that despite its cause, the decline in the value relevance of net assets can be seen as a further argument in favor of the IIRC stance to assign equal importance to a wide range of “capitals,” such as human, social and natural capital. We believe that our findings are of particular interest to a wide range of regulators, standards setters, practitioners, and academics but first and foremost to the JSE and IIRC.
We document basic facts about prices in online markets in the United States and Canada, which is a rapidly growing segment of the retail sector. Relative to prices in regular stores, prices in online ...markets are more flexible and exhibit stronger pass-through (60-75 percent) and faster convergence (half-life less than two months) in response to movements of the nominal exchange rate. Multiple margins of adjustment are active in the process of responding to nominal exchange rate shocks. Properties of goods, sellers, and markets are systematically related to pass-through and the speed of price adjustment for international price differentials.
We examine the consequences of survey underreporting of transfer programs for prototypical analyses of low-income populations. We link administrative data for four transfer programs to the CPS to ...correct its severe understatement of transfer dollars received. Using survey data sharply understates the income of poor households, distorts our understanding of program targeting, and greatly understates the effects of anti-poverty programs. Using the combined data, the poverty-reducing effect of all programs together is nearly doubled. The effect of housing assistance is tripled. Correcting survey error often reduces the share of single mothers falling through the safety net by one-half or more.
We evaluate explanations for the absence of disinflation during the Great Recession and find popular explanations to be insufficient. We propose a new explanation for this puzzle within the context ...of a standard Phillips curve. If firms'inflation expectations track those of households, then the missing disinflation can be explained by the rise in their inflation expectations between 2009 and 2011. We present new econometric and survey evidence consistent with firms having similar expectations as households. The rise in household inflation expectations from 2009 to 2011 can be explained by the increase in oil prices over this time period. (JEL D84, E24, E32, E52, E58, Q35)
House prices exhibit substantially more momentum, positive autocorrelation in price changes, than existing theories can explain. I introduce an amplification mechanism to reconcile this discrepancy. ...Sellers do not set a unilaterally high or low list price because they face a concave demand curve: increasing the price of an above-average-priced house rapidly reduces its sale probability, but cutting the price of a below-average-priced house only slightly improves its sale probability. The resulting strategic complementarity amplifies frictions because sellers gradually adjust their price to stay near average. I provide empirical evidence for concave demand using a quantitative search model that amplifies momentum two- to threefold.
Abstract
This paper studies whether firms trade political contributions for public procurement contracts. Combining data on Lithuanian government tenders, corporate donors, and firm characteristics, ...I examine how a ban on corporate contributions affects the awarding of procurement contracts to companies that donated in the past. Consistent with political favoritism, donors’ probability of winning falls by five percentage points as compared to that of nondonor firms after the ban. Evidence on bidding and victory margins suggests that corporate donors may receive auction-relevant information affecting procurement outcomes in their favor.