Measuring economic development and evaluating impact Sonia Laszlo Associate Professor, Economics Associate Director, ISID Outline Motivation: why measure? Measuring economic development: • The Big Picture • The Micro picture Measuring impact of development policies • Standard toolbox • National policies • Smaller projects Conclusions What is Economic Development? "Development can be seen (...) as a process of expanding the real freedoms that people enjoy. Focusing on human freedoms contrasts with narrower views of development, such as identifying development with the growth of gross national product, or with the rise in personal incomes, or with industrialization, or with technological advance, or with social modernization. Growth of GNP or of individual incomes can, of course, be very important as means to expanding the freedoms enjoyed by the members of the society. But freedoms depend also on other determinants, such as social and economic arrangements (for example, facilities for education and health care) as well as political and civil rights (for example, the liberty to participate in public discussion and scrutiny).” From Amartya Sen, Development as Freedom, Anchor Books, 1999 (p.3) Why measure economic development? Evaluate “health” of an economy Evaluate changes in the “health” of an economy Set goals accordingly Evaluate development policies and projects To compare To better understand Why measure economic Development? Example Millennium Development Goals (MDG) • Goal 1: halve the absolute poverty rate between 19902015 • Reached in 2010 • But only thanks to China Post-2015 MDG • Now what? The Big Picture Macroeconomic indicators Gross National Product (GNP), per capita Gross Domestic Product (GDP), per capita Savings rates Human Development Index • Combines GNP, Education and longevity Happiness levels A subjective measure of wellbeing? … Socio-economic indicators Poverty rate (e.g. head count index) Educational attainment Absence of illness Life expectancy Labour force participation Inequality (of income, land, assets…) … The Big Picture Aggregate income measures (GDP, GNI): • Aggregate • Unidimensional • Measurement issues • Relation to welfare and economic wellbeing? However, they perform remarkably well in predicting other, social, outcomes: • http://hdr.undp.org/en/dat a/explorer/ The big picture Can money buy you happiness? • What does life satisfaction mean? • Is this causal? Chart by: Angus Deaton Micro measures We may want to look at individual based measures instead Maybe even focus on the poor Why? • Rawlsian Maximin & Veil of Ignorance argument • Altruism Examples: • Income, educational attainment, absence of ill health… • Ability to meet basic needs Micro measures Poverty: Total poverty gap (TPG): • Amount of money required • H= # poor, N= Population to bring all the poor up to • Poverty line? Yp the poverty line • Head Count Index = H/N • Policy relevant • Extent versus depth of poverty? Measuring impact of development policy and projects Suppose a project aims for a particular target: • Reduce poverty rate • Increase educational attainment • Improve maternal and child health • Provide access to credit via micro-finance Policy prescription Measuring impact of development policy and projects How do you go about to evaluate whether it has achieved success? Naïve approach: compare before and after Problems of identification: • Reverse causation • Self-selection bias • Omitted variable Or, lack of counter-factual Causation versus correlation “Country X invested in education and has high income” • Education affects income • Income affects education • Both determined by other factor (e.g. institutions) “Job training programs have high success rates” • Or did participants self-select into job training based on unobservable factors, such as ability Standard tool box - Quantitative methods Macro economic time series (growth, trends, cross-country) • But: identification without counterfactual? • e.g: Vietnam with and without liberalization, holding all else constant? Micro data sets • Censuses (population, agriculture, health/schools) • Household or labour force surveys (e.g. Demographic and Health Surveys, WB’s Living Standards Measurement Surveys, etc…) • Still, need to find plausible sources of exogenous variation Example 1: Education as a development policy Education is important for economic growth and wellbeing MDG2: “Achieve universal primary education” MDG3: “Promote gender equality and empower women” • Target 3.A: “Eliminate gender disparity in primary and secondary education” Canada’s foreign policy (DFATD) “Canada is participating in the global educational effort” Example 1: Education as a development policy Theory: • Education raises marginal productivity of labour (increases skills) and so increases earnings • Education serves as a signaling device for high ability, and so increases earnings • Higher education leads to better health & fewer children (but invest more in children) • Higher education associated with higher civic participation Example 1: Education as a development policy Evidence: Example 1: Education as a development policy Some policy options: • Build more schools • Make them better quality • Provide incentives to attend school (Conditional Cash Transfers) Let’s look at some real world examples of these policies and how they can be evaluated Example 1.A.: Build more schools! (Duflo, 2001) Duflo (2001): Effect of massive investment of new schools in Indonesia Q: Infrastructure Education Q: Education Earnings Problems: • omitted variables – family background, community characteristics: third factors • Endogeneity of schooling Example 1.A.: Build more schools! (Duflo, 2001) Natural experiment – Indonesia’s “Massive school construction project” (INPRES) built approximately 2 new schools per 1,000 children in 1973-1974 to 1978-1979 • targeted kids who had low educational attainment • Primary school construction, hire and train new teachers • Financed from oil revenues. Example 1.A.: Build more schools! (Duflo, 2001) Effect of program should be: • zero for those aged 13 and above in 1974 • increasing for younger children. Example 1.A.: Build more schools! (Duflo, 2001) Results – INPRES program: • increased quantity, not quality • increased educational attainment by 0.25 to 0.40 years • increased probability of completing primary school by 12% • increased wages by approximately 3 to 5.4% Example 1.B.: Progresa/Oportunidades Conditional Cash Transfer Program in Mexico Designed as a Randomized Controlled Trial (RCT) experiment (“Pilot”: Progresa) Objective: increase educational outcomes • Poverty is a binding constraint • So, relax that constraint Provide cash transfers to poor households conditional on children’s school attendance (& visits to school nurse) • Why conditional? Example 1.B.: Progresa/Oportunidades RCT design: Treatment Control Before YTo YCo After YT1 YC1 Simple estimate of impact of program: • Difference-in-difference: (YT1-YC1)-(YT0-YC0) Measurable outcomes: • School enrolment • Test scores • Child labour • Child health • Poverty Example 1.B.: Progresa/Oportunidades IFPRI & Mexican government: • 24,000 households in 506 localities in randomly assigned PROGRESA and non-PROGRESA areas. • Formal surveys, interviews, focus groups, and workshops Their results showed huge success: • • • • Increased primary enrolment by 1.07 ppts (boys) & 1.45 ppts (girls) Increased secondary enrolment by 9.3 ppts (girls) and 5.8 ppts (boys) Reductions in child labour by up to 25% Reduced child stunting (12-36 month olds): increase of 16% mean growth per year Scaled up: Oportunidades Example 2: Sexual and reproductive health Both health and human capital implications • Excess fertility (actual > desired # of children) • Sexually transmitted infections MDG 5: Improve maternal health MDG 6: Combat HIV/AIDS, Malaria and other diseases Example 2: Excess fertility in Zambia (Ashraf et al. 2014) Excess fertility: desired number of children less than actual number of children • High rate of unwanted births • Yet, low rate of contraceptive use Outcome of bargaining process within the household over contraceptive control? Discordant preferences between men and women? Note some women hide contraceptive use from husbands Example 2: Excess fertility in Zambia Ashraf, Field and Lee (2014) The research problem: can we test for whether spousal discordance can explain excess fertility? Outcomes of interest: implications of spousal discordance in fertility decisions • Women demand more birth control than men • Women would prefer concealable forms of birth control in the presence of spousal discordance • These combined should lead to lower fertility rates Example 2: Excess fertility in Zambia Treatment - voucher for immediate and free access to contraceptives Random assignment to 2 different treatment groups: • Tindividual=1: voucher given to women alone/in private • Tcouple=1: voucher given to women in the presence of their husbands Baseline in 2007, follow-up in 2009 From Ashraf et al. 2014 Example 2: Excess fertility in Zambia Frequent shortages in contraception in local clinics... Spousal consent required by law until 2005. Nevertheless, practice still continues Problem in the study: injectable scare • Dec 2007 - March 2008 • Misinformation in the press: Injectables tested HIV positive • Will thus be difficult to say much about effect on fertility within this 2 year period Example 2: Excess fertility in Zambia (Ashraf et al. 2014) Results - Table 2: voucher take-up • give people free stuff, they take it • but effect not 100% • effect stronger in Tindividual Take up stronger in Tindividual, but only for women who do not want a child in the next 2 years Example 2: Excess fertility in Zambia (Ashraf et al 2014) Concerns: • treatment assignment to couple might have increased discussion about family planning that wouldn't have occurred in the absence of study • selection bias in the take-up of vouchers in the couples treatment (may exclude couples with extreme discordance) • involving men in family planning might not be optimal Conclusions A large set of tools in the measurement toolbox Increasingly, researchers and policy makers turning towards RCTs • But: not without some caveats • Internal vs external validity? • Need to look at theoretical concerns Other tools: • Qualitative methods • Behavioural & lab experiments…. Nevertheless, lots of innovations to measurement and impact assessment