Home Address:
  239 Edwards Street, Apt. E
  New Haven, CT 06511

Telephone: (203) 624-2972 (home)
                   (203) 435-6745 (cell)

Development Economics
Applied Econometrics

Development Economics
Economics of Africa
Applied Econometrics/Microeconometrics
Labor Economics

October 2002 (Oral) Development Economics (with distinction), Econometrics
May 2002 (Written) Microeconomic and Macroeconomic Theory

Technology Adoption, Risk-Sharing and Child Labor in Sub-Saharan Africa

Professor Michael Boozer
Professor Gustav Ranis
Professor Paul Schultz
Professor Christopher Udry

May 2006

Ph.D., Economics, Yale University, Expected May 2006
M.Phil., Economics, Yale University, December 2003
M.A., Economics, Yale University, December 2003
M.A., International and Development Economics, Yale University, May 2001
M.A. (Cantab), Trinity College, Cambridge University, July 2003
B.A. (Honors), Economics, Trinity College, Cambridge University, June 1999

Research Grants
   Yale Center for International and Area Studies Dissertation Grant, 2004–2005
   Agrarian Studies Fellowship, 2004
   Lindsay Fellowship for African Studies, 2004
   Yale Center for International and Area Studies Pre-Dissertation Grant, 2003
Academic Fellowships and Honors
   Yale University Dissertation Fellowship, Spring 2006
   Ryoichi Sasakawa Fellowship, Yale University, 2003–2004
   Economic Growth Center Prize, Yale University, 2001–2005
   Yale University Doctoral Fellowship, 2001–2005
   Scholar of the Oxford and Cambridge Society of Kenya, 1996–1999
   Scholar of the Cambridge Commonwealth Trust, 1996–1999

Raymond Powell Teaching Prize, Yale University, 2005
Teaching Assistant, Graduate Level
   Econometrics and Statistics, Graduate, Professor Michael Boozer, Fall 2004 and Fall 2005
   Economics of Natural Resource Management, Professor Douglas Gollin, School of Forestry and
      Environmental Studies, Fall 2000
Teaching Assistant, Undergraduate Level
   Intermediate Macroeconomics, Professor Stefan Krieger, Spring 2001
   Poverty Under Post Industrial Capitalism, Professor Gerald Jaynes, Spring 2000

Research Assistant
   Professors Gustav Ranis and Michael Boozer, Yale University, 2000–2005
   Professor Chris Timmins, Yale University, 2000–2001
   Professor Robert Evenson, Yale University, Summer 2000
   Professor George Hall, Yale University, Summer 2000
   Professor Paul Seabright, Cambridge University, Summer 1999

North East Universities Development Consortium, October 2003, 2005
University of Chicago, Development Workshop, November 2003

The American Economic Association
The Econometric Society

English (fluent); Punjabi (fluent); Swahili (fair); Hindi (spoken, fair); French (basic)

Professor Michael Boozer
Department of Economics
Economic Growth Center
Yale University
PO Box 208269
New Haven, CT 06520-8269
Phone: (203) 432-3623
Fax: (203) 432-3898
Email: michael.boozer@yale.edu

Professor Gustav Ranis
Department of Economics
Economic Growth Center
Yale University
PO Box 208269
New Haven, CT 06520-8269
Phone: (203) 432-3630
Fax: (203) 432-3898
Email: gustav.ranis@yale.edu

My dissertation focuses on empirical microeconomic issues of development in Africa. Economies in Sub-Saharan Africa are high-risk environments, where weather shocks (such as El Nino, droughts), crop disease and human disease are prevalent. Given that their populations are predominantly rural (the World Development Indicators report 67% of Sub-Saharan Africa to be rural in 2002), these economies are largely dependent on agriculture as a source of income. All this, together with the fact that certain markets either do not exist, or if in fact they do, they often work imperfectly, means that households face extremely uncertain and variable incomes. My dissertation addresses various trade-offs that rural households face and the implications of households’ decision making for policy. The goal of this dissertation is to examine the efficiency of household behavior in Sub-Saharan Africa within the institutional boundaries and constraints they face.

I. Selection and Comparative Advantage in Technology Adoption

In the first paper, I examine technology adoption in Kenya. In particular, I am interested in the use of the main agricultural technologies in Kenya, hybrid maize and fertilizer. From experimental and other research (Gerhart (1975), KARI (1993), Hassan (1998), Duflo, Kremer and Robinson (2004)), it is clear that these technologies can potentially increase average yields of maize dramatically. However, use of such technologies is by no means universal. For example, in 1996 only 63% of my sample of households used hybrid maize and only 47% used inorganic fertilizer. This is the empirical puzzle I am interested in: despite the possibility of high returns, a significant number of households do not use these new technologies. In order to understand this puzzle, I use an economic and econometric framework that allows the effect of hybrid on yields to vary not just across the varied geography of Kenya, but also according to the inherent productivity of the farmer. Since hybrid was developed to guard against poor yields, I anticipate the gains from its use to vary by initial non-hybrid yields.

I use an extensive panel dataset covering the period from 1996 to 2004 and representative of maize growing Kenya. I helped design and collect the 2004 round of this panel survey. I find that aggregate adoption over the period is stable across time, even when I look across regions or asset/wealth quintiles. The striking feature of my data is therefore the immense cross sectional heterogeneity, accompanied by a lack of time variation, in aggregate adoption patterns. At the household level, however, at least 30% of my households switch both into and out of use of hybrid seed from year to year.

My approach to these questions of technology adoption is very different from the standard literature. I examine whether the yield returns to adopting hybrid maize may vary across a given sample of farmers, such that high average returns to the technologies are accompanied by low returns for the marginal farmer. I therefore analyze adoption decisions within a regression model that allows for household-specific heterogeneity in returns and hence controls for selection via both absolute and comparative advantage differences across households. To estimate these differences in returns across my sample, I generalize Chamberlain’s approach to the fixed effects model to accommodate both heterogeneous slopes and intercepts. This allows me to isolate the sample of households for whom returns would be extremely high. I find that these households, in fact, do not use hybrid maize, although they are a small fraction of all farmers. The lack of adoption for this small set of households appears to stem from supply constraints, such as distance to seed and fertilizer distributors. On the other side of the coin, I also find a comparatively larger group of households for whom the returns are extremely low, almost zero, yet they universally adopt hybrid maize.

My results point to the need for focused interventions for policy to be cost effective. For the constrained farmers, alleviating their constraints by improving infrastructure and distribution would improve yields dramatically. For the unconstrained farmers, however, there is a critical need to develop new hybrid strains. This is extremely important at a time when domestic food security is at the forefront of the policy agenda in Kenya as yields of agricultural staples have, in fact, fallen on average over the past two decades (in contrast to maize in Mexico and wheat and rice in India).

II. Spillovers in Village Consumption: Testing the Extent of Partial Insurance

The second paper uses the same survey data as in my first to contribute to the literature on risk-sharing. I develop a test to capture the extent of risk-sharing across space. This test is in the spirit of Townsend’s (1994) original idea in that it is a general equilibrium omnibus test of Pareto efficient allocations of risk. As per the earlier literature, I assume that the relevant risk pooling group is the village, since the group itself cannot be identified from cross sectional data (as described in Manski (1993, 2000)). In work in progress with Michael Boozer, we are able to relax this assumption and use extensive panel data together with the strong implications of perfect risk-sharing to identify the relevant group.

Pareto efficiency has two implications: i) household consumption co-moves one for one with average village consumption and, ii) individual incomes play no role in determining household consumption. This places strong restrictions on how households respond to income shocks: they should be able to smooth away all idiosyncratic shocks, but not aggregate village shocks. I propose a contrast estimator that allows me to both test the null of Pareto efficiency and to also estimate the extent of risk-sharing if I reject the null. The contrast estimator compares how households respond to shocks within villages to the between village effect of shocks, the latter providing essentially a counterfactual. This approach makes two important contributions to the literature on risk-sharing and, more generally, social interactions. It solves an econometric issue that has pervaded previous research (Townsend (1994) and Deaton (1990)) by extending the experimental estimator derived by Boozer and Cacciola (2001) to the case of non-experimental data. In addition, it tests jointly the two implications of Pareto efficiency by directly comparing the responses of household consumption to idiosyncratic and aggregate shocks. In this way, it identifies the extent of risk-sharing within a village as a spillover at the village level, i.e. the extra benefit a household gets as part of being in a village with respect to being able to smooth consumption.

I study maize and food consumption in rural Kenya and find that households have the ability to smooth their consumption extremely well, though I reject complete risk sharing in a few cases. Not only do I examine how well households within villages pool risk, but I also show how well villages are able to pool risk within districts. These findings have strong policy implications for the level of aggregation at which governments should target safety net policies such as food subsidies, food security and other insurance programs. Since households are able to smooth staple consumption well, but villages less so, it points to a strong role for government policies when shocks aggregate to districts are experienced.

III. Child Labor and Schooling Decisions in Ghana (co-authored with Michael Boozer)

This paper investigates the choices involved in the trade-off between child labor (outside of the household) and schooling hours in Ghana in the late 1980’s. Households were surveyed once over an eleven month period in both the Northern and Southern regions of Ghana, which have distinct long-run rainfall patterns over the year. The differences in survey dates and rainfall patterns allow us to use the synthetic panel approaches of Deaton (1985) to identify exogenous variation in child labor due to rainfall shocks. We examine the different behavioral responses to short run and long run rainfall, and hence, to a degree, income patterns. We find evidence of strong substitution effects: an hour of labor reduces school attendance by approximately 0.38 hours. In addition, income or poverty in Ghana appears not to affect the degree of this substitution, and so it may not be as important a factor in formulating child labor policies as other research highlights, as there is no evidence of heterogeneity in these substitution effects across income groups.