Development Impact will now be on break over the next couple of weeks for the holidays, resuming in early January after the AEA annual meetings. Inspired by some of the interesting lists of favorite papers of the year (e.g. Noah Smith, Matt Notowidigdo) we thought we’d each offer three of our favorite development economics papers for the year...
Three Notables for Berk
- Hubacher et al. (2017) on a partially randomized patient preference trial among women presenting for short-acting contraceptive methods (SARCs). When we see continuation rates, reported for short- and long-acting reversible contraceptive (LARCs) methods, they usually come from observational data, following adopters of various methods and seeing what share are still using the same method after, say, one year. To get around the obvious selection issues (people who want to wait longer may adopt longer-term methods), the authors designed a study that allowed women seeking short-acting methods (i.e. pills or injectable) to either agree to a randomization between SARCs and LARCs or to choose their preferred SARC. All three cohorts were followed and continuation rates, pregnancies, and happiness with method were measured. The 12-month method continuation rates were 53% vs. 77% (p<0.001) in the SARC and LARC randomized groups, respectively, while the cumulative unintended pregnancy rates were 6.4% and 0.7% (p<0.01). These figures are closely related to typical use figures used to inform potential clients. Furthermore, while the continuation rates for SARCs in the preference group were higher than those in the random assignment group, pregnancy rates were indistinguishable in these two groups. Given that close to 80% of those randomly assigned to LARCs were satisfied with the product at 12 months, the findings are promising in being able to expand the take-up of LARCs among users of SARCs and those not using any modern methods.
- Hahn, Wang, and Yang (last revised 2016) on school autonomy differences between private and public schools in Seoul, South Korea. Taking advantage of the fact that students in Seoul are allocated to public or private schools randomly within their school district, which are funded similarly, but have different flexibility with respect to personnel decisions, this clever and clearly-written study finds that private schools outperform public schools in terms of four-year college attendance, standardized national tests in language and math, and the number of reported violent incidents. The authors speculate that the differences in personnel and remuneration characteristics likely explain the positive effect of attending private secondary schools. The paper is an R&R at JPubEc and will hopefully be forthcoming soon…
- Abadie, Athey, Imbens, and Woolridge (2017) on when you should cluster your standard errors. Need I say more? If your answer is ‘yes,’ David has written about it beautifully here.
Three of David McKenzie’s favorites
- Rotemberg and White on measuring cross-country differences in misallocation: This paper shows just how important we are on measurement choices for how we think about the world. They revisit the incredibly influential (and cited more than 2,800 times) Hsieh and Klenow (2009) paper which argued that firms in India have a lot more misallocation than those in the U.S., and show that this entire result may stem from the way the U.S. census bureau cleans firm census data – basically the census bureau cleans up its data by editing outliers, imputing data, and matching to alternative data sources, which is not done in developing countries. Measured misallocation is no higher in India than the U.S. when the same measurement methods are used in both countries.
- Rigol, Hussam and Roth on targeting high ability entrepreneurs using community information: this is Natalia’s job market paper this year, and is work I’ve seen grow from the idea stage where they pitched it as part of the SME ideas competition I ran to a fully realized and impressive paper. Much of the work on entrepreneurship shows how hard it is to predict which firm’s will succeed. You would think that predicting further who would have the highest treatment effects would be even more difficult. But this work shows that, amongst poor households in India, community members can predict which individuals will have the highest marginal returns from small grants – and that so can machine learning done on data from Sri Lanka.
- Andrabi, Das, Khwaja, Ozyurt and Singh on Upping the Ante: The Equilibrium Effects of Unconditional Grants to Private Schools. I’ll count this one, since they have been presenting it over the year, even though the working paper isn’t yet out. They take the idea of private schools as firms seriously, and investigate what happens at the market level when you give grants to only one school in a market, versus to all, and find schools respond by increasing capacity in the first case, but only increase quality in the second.
Three from Markus Goldstein
Maybe I just need a vacation, but I’m feeling particularly reflective and so I’ll pick three pieces that made me think more this past year:
- David’s paper on firm death. Because a) it’s a huge public good pulling together all of that data to give us the most definitive numbers on the topic, and b) it raised a lot more interesting questions for me than the average paper (e.g. why do female owned firms die faster)?
- Bernhardt et. al. on thinking about household investment choices in male owned vs. female owned enterprises. Because it a) raised a host of really interesting questions, b) made me go back to thinking about seperability, and c) was a nice use of existing impact evaluation data.
- Ruth Levine’s blog post on the moral case for evidence and what it means for impact evaluation. Because a) she skillfully links impact evaluation and distributive justice, b) made me really think (it’s worth reading more than once) about what I am doing and why, and c) her arguments matter more each day.
Three favorites from Dave Evans
I just came off working on the World Development Report on Education, so indulge my writing about three education papers, all of which I encountered for the first time this year. They may have existed in some form previously.
- Bold et al. on What Do Teachers Know and Do? Does It Matter? Evidence from Primary Schools in Africa: More and more recent work has shined a light on disappointing learning outcomes in low- and middle-income countries (as well as in many high-income countries). But why? Data on the proximate causes has been scarce. But the Service Delivery Indicators initiative has meant collecting standardized, detailed data from schools and health centers in a range of African nations, giving comparable data on teacher attendance and teacher ability. These data allow analysis of the relationships between teacher capacity and activity on student learning, teacher behavior across public and private schools, and much more that has yet to be mined. (There is a Journal of Economic Perspectives version of the paper, but the working paper has richer detail.) My one concern with this work is that there’s a temptation to focus all improvements on the area where we can demonstrate problems. We need more and better descriptive data on school management and governance to complement this exciting work on teachers.
- De Ree et al. on Double for Nothing? Experimental Evidence on an Unconditional Teacher Salary Increase in Indonesia: When I speak with teachers in different countries and ask about constraints to their ability to perform well, salary usually emerges at the top of the list. In this paper, the researchers use a large-scale randomized experiment across 350+ schools across Indonesia and the roll-out of a dramatic pay increase to examine the impacts. Higher pay made teachers better off: It “significantly improved teachers’ satisfaction with their income, reduced the incidence of teachers holding outside jobs, and reduced self-reported financial stress,” but “after two and three years, the increase in pay led to no improvement in student learning outcomes.” Of course, this doesn’t mean that higher pay couldn’t translate into better candidates over time. But it does provide strong evidence that higher teacher pay is no easy solution for student learning.
- Bergman on Parent-Child Information Frictions and Human Capital Investment: Evidence from a Field Experiment: Countries around the world have experimented with various ways of providing parents with better information on how the schools that service their children are performing, and the results have been decidedly mixed. But Bergman’s paper highlights that another kind of information – how individual students are performing – can be highly effective. Parents and other caregivers have a clear lever of action when they hear that their child missed an assignment (get it done!), and helping them to access it is a simple intervention with potentially big impacts. Subsequent research by Bergman and Chan show impacts on in-class exam scores, and Cunha et al. show similar results in Brazil, and Berlinski et al. do so in Chile. Providing information is no magic bullet for improving school quality, but it can help parents to support their students.
- favorite papers