Impact evaluation of infrastructure investments: the experience of the Millennium Challenge Corporation

In many developing countries, aging or inadequate infrastructure is a binding constraint to economic growth. The Millennium Challenge Corporation (MCC), a US government agency providing development assistance, has committed more than $4 billion to upgrade or rehabilitate roads, ports, electricity, water, sanitation and major irrigation systems in 16 countries between 2004 and 2010. In at least eight of these countries, the MCC has developed evaluations that will assess the causal impacts of these investments on a variety of outcomes, including household incomes and consumption. These evaluations primarily rely on difference-in-differences estimation, complemented by random assignment, propensity score matching, geographic information systems (GIS) models, and regression discontinuity designs. The relatively large number of evaluations (13 in all) and the diversity in their approaches offer a unique opportunity to compare these evaluations in terms of the techniques used, their ability to control for selection bias, and their flexibility under changing implementation plans. This paper studies the conditions that led to the design of each evaluation, including differing mechanisms for selecting infrastructure to be upgraded. It compares the propensity score matching approaches used in many of these evaluations, noting key observable characteristics used to match treatment and control communities. It also studies the GIS modelling approaches used in four of the roads evaluations. Finally, it reviews the flexibility of each evaluation design in response to changes in the project implementation plans that arise when there are cost over-runs and/or poor policy performance, there are delays in construction, or there are changes to the roll-out strategy. Several of these evaluations will provide the first rigorous evidence on the impacts of highway or secondary road improvement in developing country contexts. Similarly, a number of evaluations will offer important evidence on the extent to which water and sanitation improvements can raise the income level of households. By incorporating multiple methods, a number of these evaluations will also illustrate whether these methods produce different impact estimates, another notable contribution to the literature.