Courses and Book Chapters


The Graduate Institute: Fall 2012, Fall 2013, Fall 2014, Fall 2015

 

PhD in Development  Economics: Doctoral Seminar: Development Microeconomics

Development Therapy

The Graduate Institute: Fall 2013, Spring 2014, Fall 2015

 

I have taught various versions of this course over the years in places as diverse as the Universidad Autónoma del Estado de México (UAEM) in Toluca, at Paris I Panthéon-Sorbonne, and at Renmin University of China in Beijing.  I also often teach compressed versions of the course in various executive education and tailor-made programs.

 

Impact evaluation

Syllabus

Statistically assessing the impact attributable to social programs in developing countries is increasingly, though slowly, becoming standard practice in the development community. Unfortunately, and despite the great value of such exercises in terms both of program evaluation and informed policy dialogue, much remains to be done in terms of mainstreaming impact evaluations into the project cycle. In particular, the importance of the basic underlying concept of any properly conducted impact evaluation, that of a counterfactual, is often difficult to impress on individuals involved in purely operational aspects of program/project execution.

Heavy emphasis will be placed on concrete examples where these methods have been applied, in such areas as privatization of basic services, conditional cash transfer programs, health, community-driven social programs, education, and nutrition. Supplementary technical material is also suggested, allowing participants to delve more deeply into the nuts and bolts of the methods covered (you will fi…nd this in the "Supplementary material" sections of the syllabus).

Time permitting, I will also give you examples of code in R (freely downloadable on the web at http://cran.r-project.org) that implement the methods covered in the course.  R-Studio provides a nice interface for R and you should also install it on your laptop.  It is hoped, by the end of the course, that participants will have aquired the capacity to distinguish what is from what is not a true impact evaluation, and will be able to critically assess the validity of studies of this kind prepared under the auspices of multilateral or bilateral donors.

I expect you to read the papers in the "Readings" sections before class.

1. What is an impact evaluation and the unifying framework provided by the Roy model

Introductory lecture on the concept of a counterfactual and on what does not constitute an impact evaluation.  A Beamer presentation on the heuristic reasons for which one should evaluate is available here.  A Beamer presentation on the basic assumptions underlying the five approaches to impact evaluation (dealt with in the rest of the course), with a few simple examples from my own research is available here.  I will also present the Roy model. Some class notes on the potential outcomes approach (for multiple treatment intensities) and the Roy model (for a binary treatment) are available here.

Supplementary material. Read the only paper by Heckman (2005) that you are going to …find in a sociology journal, on the scienti…c concept of causality. Three other important papers on what a causal effect is in econometrics: Holland (1986), Heckman, LaLonde, and Smith (1999), Angrist and Krueger (1999). An excellent short, though technical, summary of this material is provided in Angrist, Imbens, and Rubin (1996); we will come back to it when we deal with instrumental variables An excellent summary of much that will be covered in the course is provided by Heckman’s Nobel lecture, in Heckman (2001); this paper also furnishes you with a formal presentation of the Roy model; an even more sophisticated version is available in Heckman and Vytlacil (2005). You can also look at Heckman and Honoré (1990).

2. Randomization, RDD, and matching

Lecture on how randomization solves the evaluation problem; on OLS and on matching estimators, and on regression discontinuity design (RDD).  Notes on OLS, selection on observables, and on the (absurd) conditions under which such an assumption is tenable are available here.

Readings. For two great papers on impact evaluations based on randomizations, dissect Duflo and Chattopadhyay (2004) and Miguel and Kremer (2004). Read Angrist and Lavy (1999) for a very nice application of RDD to evaluating the impact of class size on student performance. Recent surveys of RDD methods are provided by Imbens and Lemieux (2008) and Lee and Lemieux (2010).

Examples from my own experience. A randomized HIV/AIDS intervention in Senegal: Arcand, Diallo, Sakho, and Wagner (2010). Evaluating teacher training and risky sex in the Cameroun:Arcand and Djimeu Wouabe (2010); evaluating a CDD program in Morocco.

Supplementary material. On matching, read the basic paper by Rosenbaum and Rubin (1983) on the concept of the propensity score. Caliendo and Kopeinig (2008) provides a great summary. Heckman, Ichimura, and Todd (1997) and Heckman, Ichimura, and Todd (1998) consider a more robust version of matching that is based on exclusion restrictions, as with the instrumental variables approaches that we will consider later on. Also see the devastating critique by Deaton (2010), and the measured response by Imbens (2009).

3. Difference in differences and panel data

Lecture on the difference in differences approach.

Readings. On the impact of the privatization of water, Galiani, Gertler, and Schargrodsky (2005); on child nutrition and health, Behrman and Hoddinott (2005).

Example from my own experience. A difference in differences-based impact evaluation of a CDD program in Senegal: Arcand and Bassole (2009).

Supplementary material. This section is essentially based on standard panel data econometrics; the appropriate chapters in Wooldridge (2001) are therefore the best place to look for those who want to delve more deeply into these questions. If you want a one-paper treatment of the basic linear panel data model, the best bet is the incredibly clear paper by Hausman and Taylor (1981).  Other issues that are essential in terms of correct statistical inference are (i) clustering of standard errors when treatment occurs at a level of aggregation that is higher than that of the observations (see the standard papers by Moulton (1986) and Moulton (1990)) and (ii) autocorrelation (read Bertrand, Duflo, and Mullainathan (2004) very carefully). People often do not know what to say when they estimate a non-signi…cant treatment effect. Andrews (1989) tells you what you can say; Jacoby and Mansuri (2009) provide a very nice application in the context of sharecropping. Finally, you should also worry about the parallel trends assumption. See my paper (Arcand and Bassole (2009)) for an application.

4. Instrumental variables

Lecture on instrumental variables and what constitutes a good identi…cation strategy.  Class notes on finite sample bias in IV estimation, on the forbidden regression problem and on the unifying framework provided by k-class estimators are available here.

Readings. Miguel and Roland (2006) is a very nicely crafted paper on the impact on Vietnam of the American strategic bombing campaign.  Another great paper is Miguel, Satyanath and Sergenti 2004Levitt (1997) is a classic paper on the impact of policing on crime. Read Heckman and Navarro-Lozano (2004) for a general discussion of the relative strengths and weaknesses of IV, matching and control function approaches.

Example from my own experience. Read Arcand and Bassole (2009) for an identifi…cation strategy for the impact of completed CDD projects based on quantitative measures of political power.

Supplementary material. The best summary treatment on instrumental variables is probably that by Hausman (1983). Chapter 5 in Wooldridge (2001) (pp. 83-113) is an excellent resource. Finally, perhaps the best compact treatment of IV techniques and their application to impact evaluation is furnished by Angrist, Imbens, and Rubin (1996), mentioned earlier. When applying instrumental variables methods, apart from
having a good story concerning the orthogonality of your IVs, you should also be aware of the problems that will arise if your instruments are "weak". See the usual treatment in Stock, Wright, and Yogo (2002) and the stupidly neglected paper by Hahn and Hausman (2002).  An application of this last test is in Ai, Arcand and Ethier (2007).  In many situations, especially if you have many IVs, it is important to know how to choose them (apart from the standard tests of the overidentifying restrictions).  See Andrews (1999) for a very simple test based on the J-statistic, and a slightly more complicated test by Donald and Newey (2001).  If you absolutely want to do Arellano-Bond type GMM IV estimation, a good place to start is the excellent paper by Bond (2002).

5. Essential heterogeneity

Lecture on estimating the marginal treatment effect (MTE) using non-parametric methods, and on how one can distinguish between the average treatment effect (ATE), the effect of treatment on the treated (TT) and the effect of treatment on the untreated (TUT) using these methods.  Lecture notes on MTE, LIV estimation and on what standard linear IV estimates in the presence of essential heterogeneity are available here and here.

Reading. Take a look at Carneiro, Heckman, and Vytlacil (2011) for an application of MTE methods to estimating the returns to a college education.

Examples from my own experience. I apply these methods to estimating the impact of a social program in Senegal in Arcand and Bassole (2007).

Supplementary material. The semiparametric approach using local linear regression methods was introduced in Heckman and Vytlacil (1999) and developed further in Heckman, Urzua, and Vytlacil (2006). You should also read Heckman and Vytlacil (2001) so as to understand the concept of "policy-relevant" treatment effects.
 

The Graduate Institute: Spring 2012

 

Master in International Economics: Econometrics IIIb

Development Therapy

The Graduate Institute: Fall 2011

 

PhD in Development  Economics: Doctoral Seminar: Development Microeconomics
Master in International Economics: Econometrics II
Master in Development Studies: Introduction to Development Economics

Development Therapy

The Graduate Institute: Executive Master in Oil and Gas

 

Economic development and diversification in resource rich countries

Syllabus

It would be a very good idea if you could read the papers on the reading list before class. On each topic, I will present the basic issues in lecture format ---if you read the papers beforehand, we will be able to have an informed and intelligent discussion in class. I will do my very best to make the technical material accessible to you ---do not despair!

1. Economic development, development economics: an overview

Economists "do it with a model" as someone once put it. See why the profession does it this way according to a recent Nobel prize-winner in Economics in Krugman 1998. For a fascinating and concise view of the entire Development Economics field (inspired by another great ---Albert Hirschman), read Krugman 1994.  Development economics is a very very broad subject. Skim the survey by the former Senior VP and Chief Economist of the World Bank in Stern 1989. A puzzle solved by Lucas 1990 and the issue that (thankfully) got him thinking about economic development (and also helped him get a Nobel).

2. Solow, basic growth theory and naive growth empirics

The paper that started everything (and got him his Nobel) is Solow 1956 (only skim this), revisited with empirics added, in Mankiw, Romer and Weil 1992 ---perhaps the most cited empirical growth paper (more on this below). There are literally hundreds of papers that do growth regressions. Most of them are not particularly informative because they forgot the cardinal rule of econometrics ---spell out your identification strategy. Be this as it may, for your education you should be familiar with (at the very least): Sachs and Warner 1997.

3. Bringing natural resources into the picture

The paper that started natural resources economics: Hotelling 1931, plus two other classics: Solow 1974 and Hartwick 1977.  A nice summary is also provided by Barbier 2003.

4. Institutions versus geography

You should visit Daron Acemoglu's web site at http://econ-www.mit.edu/faculty/acemoglu/index.htm: this is what real empirical testing of economic growth is all about ---and Acemoglu is not a bona fide econometrician; he just happens to be really smart (he won the Bates-Clark medal a few years ago, and is a sure Nobel within the next 15 years). There are at least a dozen papers (published and unpublished) worth reading on everything from the Modernization Hypothesis, the effect of life expectancy, or the link between democracy and economic growth. At the very least, you should read the famous settler mortality paper Acemoglu, Johnson and Robinson 2001, as well as the reversal of fortune paper Acemoglu, Johnson and Robinson 2002.

5. Climate change and natural resources

You should read Stern 2007.  There is a very brief summary here.

6. The resource curse

A nice summary of the literature is provided in Sachs 2001. Also see Stevens 2003. For the link with educational investment, see Gylfason 2000.

7. The oil market

A very nice economic analysis is provided in Hamilton 2008.

8. Conflict and natural resources

Read two papers, one micro, the other macro.  For the micro side, see Dube and Vargas 2006.  For the macro side, see Collier and Hoeffler 2004.  If you are interested specifically in Africa, my political geography colleagues have been very active: see Buhaug and Rod 2006.

Some Chapters from my Micro-Development Textbook

 (dans la langue de Molière)

 

Chapitre 1   :  Un livre de cuisine

Chapitre 2   :  Théorie du producteur

Chapitre 3   :  Théorie du consommateur

Chapitre 4   :  Equilibre général

Chapitre 5   :  Economie de l'incertain : Les bases

Chapitre 11 :  Modèles de ménage