ABSTRACT
This is the final chapter of Part 1. In it we compare three approaches that have been used in empirical approaches to understanding development policy and which have been introduced in the previous chapters. The first uses modelling with survey data, the second is an approach to policy analysis using experiments (as introduced in the last chapter) and third is a calibration method which we introduced in Chapter 1 (although we did not give it that name). Researchers have shown great ingenuity in all three methods and part of our task in this chapter is to show how these different approaches are related. As we discussed in the last chapter, an experiment can, by design, avoid some of the endogeneity issues that are always an aspect of any analysis with survey data. However, the modelling approach and that of programme evaluation are more closely linked than may at first be apparent. We begin Section 13.2 by revisiting two methods of estimation – multivariate regression and panel estimators – which can be reinterpreted through the lens of the programme evaluation approach. In Section 13.3 we provide an account of one of the earliest and most influential interventions: the PROGRESA programme in Mexico, which was set up as an RCT.
