Basic Information

Position: Assistant Professor, School of Information
Director, Data-Intensive Development Lab
University of California, Berkeley
Contact: jblumenstock[@]berkeley[.]edu
p: (510) 642-1464 / f: (510) 642-5814
for encrypted correspondence: my public key
My work focuses on developing new methods for using massive, spatiotemporal network data to better understand the economic lives of the poor. Most of this work is based in developing and conflict-affected countries.

News and updates

Recently Published

Blumenstock, JE (2016). Fighting Poverty with Data, Science, 353(6301), 753-754
Policy-makers in the world's poorest countries are often forced to make decisions based on limited data. Consider Angola, which recently conducted its first postcolonial census. In the 44 years that elapsed between the prior census and the recent one, the country's population grew from 5.6 million to 24.3 million, and the country experienced a protracted civil war that displaced millions of citizens. In situations where reliable survey data are missing or out of date, a novel line of research combines big data and machine learning to offer promising alternatives.
Blumenstock, JE, Cadamuro, G, On, R (2015). Predicting Poverty and Wealth from Mobile Phone Metadata, Science, 350(6264), 1073-1076
Accurate and timely estimates of population characteristics are a critical input to social and economic research and policy. We show that an individual's past history of phone use can be used to infer his or her socioeconomic status, and that the predicted attributes of millions of individuals can in turn be used to accurately reconstruct the distribution of wealth of an entire nation, or to infer the asset distribution of micro-regions comprised of just a few households. In resource-constrained environments where censuses and household surveys are rare, this creates an option for gathering localized and timely information at a fraction of the cost of traditional methods.
Blumenstock, JE, Eagle, N, Fafchamps, M (2016). Airtime Transfers and Mobile Communications: Evidence in the Aftermath of Natural Disasters, Journal of Development Economics, 120: 157-181
We provide empirical evidence that an early form of "mobile money" is used to share risk. Our analysis is based on the entire universe of mobile phone-based communications over a four-year period in Rwanda, including millions of interpersonal transfers sent over the mobile phone network. Exploiting the quasi-random timing and location of natural disasters, we show that people make transfers to individuals affected by economic shocks. Unlike other documented forms of risk sharing, the mobile-phone based transfers are sent over large geographic distances and in response to covariate shocks. Transfers are more likely to be sent to wealthy individuals, and are sent predominantly between pairs of individuals with a strong history of reciprocal exchange. [View Video]

In Progress

Through a field experiment in Afghanistan, we show that default enrollment in payroll deductions increases rates of savings by 40 percentage points, and that this increase is driven by present-biased preferences. Working with Afghanistan's primary mobile phone operator, we designed and deployed a new mobile phone-based automatic payroll deduction system. Each of 967 employees at the country's largest rm was randomly assigned a default contribution rate (either 0% or 5%) as well as a matching incentive rate (0%, 25%, or 50%). We find that employees initially assigned a default contribution rate of 5% are 40 percentage points more likely to contribute to the account 6 months later than individuals assigned to a default contribution rate of zero; to achieve this effect through financial incentives alone would require a 50% match from the employer. We also find evidence of habit formation: default enrollment increases the likelihood that employees continue to save after the trial ended, and increases employees' self-reported interest in saving and sense of financial security.
Social Networks and Migration - joint with Xu Tan (University of Washington)
How does the structure of an individual's social network affect his decision to migrate? We study the migration decisions of roughly one million individuals in Rwanda over a period of several years, using novel data from the monopoly mobile phone operator to reconstruct the complete social network of each individual in the months prior to migration. We use these data to directly validate several classic theories of migration that have historically been difficult to test, for instance that individuals with closely-knit networks in destination communities are more likely to migrate. Our analysis also uncovers several empirical results that have not been documented in the prior literature, and which are not consistent with common theories of how individuals derive value from their social networks. We propose a simple model of strategic cooperation to reconcile these results.
Why do referrals work? Selection and peer influence in a million-person network experiment - joint with Greg Fischer (LSE), Dean Karlan (Yale), and Adnan Khan (LSE)
In contexts ranging from health and agriculture to product marketing and job search, individuals referred through social networks are often more likely to take an action than individuals acting in isolation. We conduct an experiment on Pakistan's largest mobile phone network to understand why referrals are effective. The experiment allows us to separate the effect of "selection," whereby individuals referred through social networks have characteristics that make them more likely to act (Beaman & Magruder), from "influence," or the effect of the referral itself, including endorsement (BenYishay & Mobarak), enforcement (Bryan, Karlan, Zinman), and learning (Conley & Udry).
We develop a predictive model of Mobile Money adoption that uses billions of mobile phone communications records to understand the behavioral determinants of adoption. We describe a novel approach to feature engineering that uses a Deterministic Finite Automaton to construct thousands of behavioral metrics of phone use from a concise set of recursive rules. These features provide the foundation for a predictive model that is tested on mobile phone operators logs from Ghana, Pakistan, and Zambia, three very dierent developing-country contexts. The results highlight the key correlates of Mobile Money use in each country, as well as the potential for such methods to predict and drive adoption. More generally, our analysis provides insight into the extent to which homogenized supervised learning methods can generalize across geographic contexts. We find that without careful tuning, a model that performs very well in one country frequently does not generalize to another.
Violence and Financial Decisions: Experimental Evidence from Mobile Money in Afghanistan - joint with Michael Callen (UCLA) and Tarek Ghani (UC Berkeley)
We provide evidence that violence affects how people make financial decisions. Exploiting the quasi-random timing of several thousand violent incidents in Afghanistan, we show that individuals who are exposed to violence are less likely to adopt and use mobile money, a new financial technology, and are more likely to retain cash on hand. This effect is corroborated using data from three independent sources: (i) the entire universe of 5 years of mobile money transactions in Afghanistan; (ii) high-frequency data from a randomized experiment designed to increase mobile money adoption; and (iii) a behavioral lab-in-the-field experiment with experienced mobile money users. Collectively, the evidence highlights an economic cost of violence that operates through individual beliefs, which is large enough to impede the development of formal financial systems in conflict settings.
We exploit a novel source of data to model the impact of migration and urbanization on segregation in Estonia. Analyzing the complete mobile phone records of hundreds of thousands of Estonians, we find that the ethnic composition of an individual's geographic neighborhood heavily influences the structure of the individual's phone-based network. We further find that patterns of segregation are significantly different for migrants than for the at-large population: migrants are more likely to interact with coethnics than non-migrants, but are less sensitive to the ethnic composition of their immediate neighborhood than non-migrants.