World Bank Economic Review (2019).


Short-term human mobility has important health consequences, but measuring short-term movement using survey data is difficult and costly, and use of mobile phone data to study short-term movement is only possible in locations that can access the data. Combining several accessible data sources, Senegal is used as a case study to predict short-term movement within the country. The focus is on two main drivers of movement—economic and social—which explain almost 70 percent of the variation in short-term movement. Comparing real and predicted short-term movement to measure the impact of population movement on the spread of malaria in Senegal, the predictions generated by the model provide estimates for the effect that are not significantly different from the estimates using the real data. Given that the data used in this paper are often accessible in other country settings, this paper demonstrates how predictive modeling can be used by policy makers to estimate short-term mobility.

With Gary Burtless

Soc. Sec. Bull. 73 (2013): 83.


The increasing cost of employer contributions for employee health insurance reduces the share of compensation subject to the Social Security payroll tax. Rising insurance contributions can also have a more subtle effect on the Social Security tax base because they influence the distribution of money wages above and below the taxable maximum amount. This article uses the Medical Expenditure Panel Survey to analyze trends in employer health insurance contributions and the distribution of those costs up and down the wage distribution. Our analysis shows that employer health insurance contributions increased faster than overall compensation during 1996-2008, but such contributions grew only slightly faster among workers earning less than the taxable maximum than they did among those earning more. Because employer health insurance contributions represent a much higher percentage of compensation below the taxable maximum, health insurance cost trends exerted a disproportionate downward pressure on money wages below the taxable maximum.

With Barry Bosworth

Washington, Brookings (2011)

Working Papers

Implementation of a multi-centre digital trauma registry: Experience in district and central hospitals in Malawi

With Kevin Croke, Linda Chokotho, Jonna Bertfelt, Saahil Karpe, Meyhar Mohammed, and Wakisa Mulwafu

With Elisabeth zu Erbach-Schoenberg, Linus Bengtsson Erik Wetter, Andy Tatem


While human mobility has important benefits for economic growth, it can generate negative externalities. This paper studies the effect of mobility on the spread of disease in a low incidence setting when people do not internalize their risks to others. Using 15 billion mobile phone records for 9 million SIM cards, I causally quantify the relationship between travel and the spread of malaria. Estimates indicate that an infected traveler results in 1.7 additional cases reported in the health facility in the traveler's destination. I develop a simulation-based policy tool that demonstrates that using mobile phone data to inform strategic targeting of travelers based on their origin and destinations could reduce the caseload by 50\% more than the next best strategy for containment that only uses previous incidence.

Household recombination, retrospective evaluation, and educational mobility over 40 years

With Andrew Foster


Despite longstanding belief in certain circles that investment in primary health care and education can help to encourage reductions in inequality and increases in intergenerational economic mobility, evidence is scarce due to the lack of systematically collected data from developing countries that links households over multiple decades. Bangladesh would seem an especially fruitful avenue for looking at these issues given international recognition of its success in improving basic health care and expanding primary education. In this paper we use newly collected survey data connected to the Matlab Demographic Surveillance System (DSS) maintained by the International Center for Diarrheal Disease Research in Bangladesh (ICDDRB) to take a first look at this issue. A novel insight from this paper is that standard methods for correcting sampling weights in panel data do not adequately account for the process of household formation and dissolution. We develop a new approach to weighting that requires the kind of information available in the context of a DSS, and use these weights to look at long term changes in educational investment of households in the Matlab area. We show that a substantial rise in average educational investment among children 6-16 has been accompanied by high levels of economic mobility but little reduction in economic inequality.

Long-Term Effects of the Matlab Maternal and Child Health and Family Planning Program on Cohort Survival and Health

With Jane Menken, Randall Kuhn, Patrick S. Turner, Tania Barham, Abdur Razzaque, Elisabeth Dowling Root, Andrew Foster, Warren Jochem, Gisella Kagy, Nobuko Mizoguchi


Late in the 1970s, icddr,b introduced its well-known maternal and child health and family planning program incrementally in approximately half the area in which their Matlab Health and Demographic Surveillance System collected data from all households at least quadrennially. Similar services were unavailable in the Comparison area for approximately a dozen years. Effects on period fertility and infant/child mortality were immediate. But did advantage last over the lifecourse? This is the first paper examining long-term effects (to 2014) on birth cohort survival and health. Treatment area survival increased almost immediately, with differences most pronounced in cohorts born 1978-87. Area differences narrowed thereafter. Cohort fertility results reveal sizeable bias when migrants are excluded from analysis. Further analysis will utilize extensive recently collected migrant follow-up data to continue this unprecedented opportunity for long-term evaluation. Preliminarily, when migrants are included, some cohorts show treatment area advantage in survival at older ages.

Work in Progress

Bias in Measuring Population Movement Using Mobile Phone Data: Utilization of Randomly Generated "Robo-calls"