Matthew Olckers


My favorite papers at the ASSA Meeting 2018

  • picks

Last year, I trawled through the list of papers which were presented at the ASSA Annual Meeting and picked a few of my favorites. It was interesting to look back on what caught my eye. I decided to do the same this year. Below are a selection of papers I find interesting from the hundreds of papers that will be presented in January.

“Behavior Revealed in Mobile Phone Usage Predicts Loan Repayment” by Daniel Björkegren and Darrell Grissen presented at the session titled Economic Applications of Machine Learning.

Many households in developing countries lack formal financial histories, making it difficult for banks to allocate capital and for potential borrowers to obtain loans. However, many unbanked households have mobile phones and even prepaid phones generate rich data about their behavior. This project shows that behavioral signatures in mobile phone data predict default with accuracy approaching that of credit scoring methods that rely on financial histories. The method is demonstrated using call records matched to loan outcomes for a sample of borrowers in a Caribbean country. Individuals in the highest quartile of risk by our measure are 5.5 times more likely to default than those in the lowest quartile. We obtain this performance despite the fact that our sample is poor and uses phones infrequently. We outline several ways our method could be practically implemented.

“New Evidence on the Choice of Retirement Income Strategies: Annuities vs. Other Options” by Jeffrey Brown, James Poterba and, David Richardson presented at the session titled Annuity Markets and Retirement Income Security.

This paper exploits rich longitudinal data on the contribution behavior and subsequent distribution decisions of the participants in a large, multi-employer, defined contribution pension plan (TIAA). The data set tracks participants over more than a decade and it provides new evidence on the heterogeneity of participants’ draw-down strategies. Only a minority of participants follow the strategy of accumulating until they retire and then either purchasing an annuity or rolling their plan balance to an IRA. Many participants defer distributions until many years after their last contribution. Among those who annuitize, annuitizing part of the plan accumulation is more common than annuitizing the entire accumulation. The paper concludes with a discussion of which of these findings are consistent with standard lifecycle models and which are difficult to explain in this framework.

“Measuring Poverty and Vulnerability in Real-time” by Joshua Blumenstock, Michael Callen, Tarek Ghani, Niall Keleher and Jacob Shapiro presented at the session titled New Methods for Measuring Poverty and Welfare

In wealthy nations, novel sources of data from the internet and social media are enabling new approaches to social science research and creating new opportunities for public policy. In developing countries, by contrast, fewer sources of such data exist and researchers and policymakers often rely on data that are unreliable or out of date. Here, we develop a new approach for measuring the dynamic welfare of individuals remotely and in near real-time, through analyzing their patterns of mobile phone use. To benchmark these methods, we conducted high-frequency panel surveys with 1,200 Afghan citizens and with the respondent’s consent, matched each individual’s responses to his or her entire history of mobile phone-based communication, which we obtained from Afghanistan’s largest mobile operator. We show that mobile phone data can be used to accurately estimate the social and economic welfare of respondents and that machine learning models can be used to infer the onset and magnitude of positive and negative shocks. These results have the potential to transform current practices of policy monitoring and impact evaluation.

“Optimal Paternalistic Savings Policies” by Christian Moser and Pedro Olea presented at the session titled Optimal Policies in a Behavioral World

We study optimal savings policies when there is a dual concern about under-saving for retirement and income inequality. In our model, agents differ in time preferences and earnings ability, both unobservable to a planner with paternalistic and redistributive motives. We characterize the solution to this two-dimensional screening problem and provide a decentralization using realistic policy instruments: forced savings at low incomes—similar to Social Security—but a choice between savings accounts with different subsidies and caps at high incomes—like 401(k) and IRA accounts in the US. Offering more choice in savings for high-income individuals acts as a screening device that facilitates redistribution. We calibrate our model to microdata on income and wealth accumulation and find that the current US savings and tax system is off the Pareto frontier, with large welfare gains available from simple reforms.

“Discovering Heterogeneous Effects Using Generic Machine Learning Tools” by Victor Chernozhukov and Esther Duflo presented at the session titled Machine Learning for Policy Research

We propose inference methods and tools for characterizing heterogenous treatment effects in randomized experiments and observational studies, which are applicable in conjunction with any high-quality modern prediction method from machine learning. We provide point and interval estimators, where the latter quantify the uncertainty associated with point estimates. We demonstrate the utility of the approach in a variety of empirical examples.

“The Ostrich in Us: Selective Attention to Financial Accounts, Income, Spending and Liquidity” by Arna Olafsson and Michaela Pagel presented at the session titled Departures from Rationality in Finance

A number of theoretical research papers across multiple fields in economics analyze attention but direct empirical evidence on attention remains scarce. This paper investigates the determinants of attention to financial accounts using panel data from a financial management software provider containing daily logins, income, spending, balances and credit limits. We first explore whether individuals pay attention in response to the arrival of income payments. Here, we utilize that weekends and holidays generate exogenous variation in regular payment arrival using a fixed-effects approach. We find that individuals are five times more likely to log in because they get paid, even though the new information associated with regular income payments should be very limited. Moreover, we estimate a comparable marginal propensity to log in using plausibly exogenous income payments. Beyond looking at the causal effect of income on attention, we examine how attention depends on spending and individual financial standing, such as cash holdings, savings and liquidity. We find that attention is decreasing in individual spending and overdrafts and increasing in cash holdings, savings and liquidity. Finally, attention jumps discretely when balances change from negative to positive. All of these results are consistent with Ostrich effects and anticipatory utility as the first-order motivation for checking financial accounts. To rationalize our findings, we set up a model assuming individuals experience utility over news, or changes in expectations about consumption, as proposed by Koszegi and Rabin (2009). Because agents dislike bad news more than they like good news, paying attention to financial account is considered unpleasant, especially when remaining cash holdings are low.

“Workplace Signaling and Financial Commitment: Evidence From a Field Experiment” by Emily Breza, Martin Kanz and Leora Klapper presented at the session titled Financial Inclusion Through Savings: Commitment Devices, Mobile Money and the Role of Trust

This paper provides evidence of signaling motivations in financial decisions. We conduct a field experiment with workers at a large manufacturing firm in Bangladesh, who are offered the opportunity to sign up for a basic commitment savings product linked to their payroll accounts. We vary whether a manager provides an endorsement of the product and whether information about the worker’s take-up decision is communicated back to the employer. We find evidence consistent with workplace signaling motivations in the decision to open commitment savings accounts. Workers respond to the endorsement only when take-up information is communicated back to the employer. In contrast, the employer endorsement alone has no effect on the decision to open an account and employer feedback decreases take-up when it is not paired with an explicit employer endorsement. The effects are driven by promotion-seeking individuals with a longer intended duration of employment, who can credibly signal their commitment to the firm by entering into a financial contract with a longer time horizon than their peers. Using a separate experiment, in which managers evaluate employees based on work performance and financial information, we show that workers’ decisions are rational: managers are more likely to invest in employees who signal commitment to the firm by signing up for a savings contract with longer maturity. These findings indicate the presence of signaling motivations and significant strategic behavior when financial decisions are observable to the employer. Our results also point to an important rationale for tying financial decisions to the workplace – the reduction of information asymmetries in the employer-employee relationship.

“Targeting High Ability Entrepreneurs Using Community Information: Mechanism Design In The Field” by Reshmaan Hussam, Natalia Rigol and Benjamin Roth presented at the session titled Mechanism Design Meets Development

The impacts of cash grants and access to credit are known to vary widely, but progress on targeting these services to high-ability, reliable entrepreneurs is so far limited. This paper reports on a field experiment in Maharashtra, India that assesses (1) whether community members have information about one another that can be used to identify high-ability microentrepreneurs, (2) whether organic incentives for community members to misreport their information obscure its value and (3) whether simple techniques from mechanism design can be used to realign incentives for truthful reporting. We asked 1,380 respondents to rank their entrepreneur peers on various metrics of business profitability and growth and entrepreneur characteristics. We also randomly distributed cash grants of about $100 to measure their marginal return to capital. We find that the information provided by community members is predictive of many key business and household characteristics including marginal return to capital. While on average the marginal return to capital is modest, preliminary estimates suggest that entrepreneurs given a community rank one standard deviation above the mean enjoy an 8.8% monthly marginal return to capital and those ranked two standard deviations above the mean enjoy a 13.9% monthly return. When respondents are told their reports influence the distribution of grants, we find a considerable degree of misreporting in favor of family members and close friends, which substantially diminishes the value of reports.