An introduction to Impact Evaluation Markus Goldstein Poverty Reduction Group The World Bank My question is: Are we making an impact? 2 parts • Impact evaluation methods • Impact evaluation practicalities: IE and the project cycle • Use rural project examples Outline - methods • • • • Monitoring and impact evaluation Why do impact evaluation Why we need a comparison group Methods for constructing the comparison group • When to do an impact evaluation Monitoring and IE IMPACT OUTCOMES OUTPUTS INPUTS Effect on living standards - infant and child mortality, - prevalence of specific disease Access, usage and satisfaction of users - number of children vaccinated, - percentage within 5 km of health center Goods and services generated - number of nurses - availability of medicine Financial and physical resources - spending in primary health care Monitoring and IE Program impacts confounded by local, national, global effects IMPACTS OUTCOMES Users meet service delivery OUTPUTS Gov’t/program production function INPUTS difficulty of showing causality Impact evaluation • Many names (e.g. Rossi et al call this impact assessment) so need to know the concept. • Impact is the difference between outcomes with the program and without it • The goal of impact evaluation is to measure this difference in a way that can attribute the difference to the program, and only the program Why it matters • We want to know if the program had an impact and the average size of that impact – Understand if policies work • Justification for program (big $$) • Scale up or not – did it work? • Meta-analyses – learning from others – (with cost data) understand the net benefits of the program – Understand the distribution of gains and losses What we need  The difference in outcomes with the program versus without the program – for the same unit of analysis (e.g. individual) • Problem: individuals only have one existence • Hence, we have a problem of a missing counter-factual, a problem of missing data Thinking about the counterfactual • Why not compare individuals before and after (the reflexive)? – The rest of the world moves on and you are not sure what was caused by the program and what by the rest of the world • We need a control/comparison group that will allow us to attribute any change in the “treatment” group to the program (causality) comparison group issues • Two central problems: – Programs are targeted  Program areas will differ in observable and unobservable ways precisely because the program intended this – Individual participation is (usually) voluntary Participants will differ from non-participants in observable and unobservable ways • Hence, a comparison of participants and an arbitrary group of non-participants can lead to heavily biased results Example: providing fertilizer to farmers • The intervention: provide fertilizer to farmers in a poor region of a country (call it region A) – Program targets poor areas – Farmers have to enroll at the local extension office to receive the fertilizer – Starts in 2002, ends in 2004, we have data on yields for farmers in the poor region and another region (region B) for both years • We observe that the farmers we provide fertilizer to have a decrease in yields from 2002 to 2004 Did the program not work? • Further study reveals there was a national drought, and everyone’s yields went down (failure of the reflexive comparison) • We compare the farmers in the program region to those in another region. We find that our “treatment” farmers have a larger decline than those in region B. Did the program have a negative impact? – Not necessarily (program placement) • Farmers in region B have better quality soil (unobservable) • Farmers in the other region have more irrigation, which is key in this drought year (observable) OK, so let’s compare the farmers in region A • We compare “treatment” farmers with their neighbors. We think the soil is roughly the same. • Let’s say we observe that treatment farmers’ yields decline by less than comparison farmers. Did the program work? – Not necessarily. Farmers who went to register with the program may have more ability, and thus could manage the drought better than their neighbors, but the fertilizer was irrelevant. (individual unobservables) • Let’s say we observe no difference between the two groups. Did the program not work? – Not necessarily. What little rain there was caused the fertilizer to run off onto the neighbors’ fields. (spillover/contamination) The comparison group • In the end, with these naïve comparisons, we cannot tell if the program had an impact  We need a comparison group that is as identical in observable and unobservable dimensions as possible, to those receiving the program, and a comparison group that will not receive spillover benefits. How to construct a comparison group – building the counterfactual 1. 2. 3. 4. 5. Randomization Matching Difference-in-Difference Instrumental variables Regression discontinuity 1. Randomization • Individuals/communities/firms are randomly assigned into participation • Counterfactual: randomized-out group • Advantages: – Often addressed to as the “gold standard”: by design: selection bias is zero on average and mean impact is revealed – Perceived as a fair process of allocation with limited resources • Disadvantages: – Ethical issues, political constraints – Internal validity (exogeneity): people might not comply with the assignment (selective non-compliance) – Unable to estimate entry effect – External validity (generalizability): usually run controlled experiment on a pilot, small scale. Difficult to extrapolate the results to a larger population. Randomization in our example… • Simple answer: randomize farmers within a community to receive fertilizer... • Potential problems? – Run-off (contamination) so control for this – Take-up (what question are we answering) 2. Matching • Match participants with non-participants from a larger survey • Counterfactual: matched comparison group • Each program participant is paired with one or more nonparticipant that are similar based on observable characteristics • Assumes that, conditional on the set of observables, there is no selection bias based on unobserved heterogeneity • When the set of variables to match is large, often match on a summary statistics: the probability of participation as a function of the observables (the propensity score) 2. Matching • Advantages: – Does not require randomization, nor baseline (preintervention data) • Disadvantages: – Strong identification assumptions – Requires very good quality data: need to control for all factors that influence program placement – Requires significantly large sample size to generate comparison group Matching in our example… • Using statistical techniques, we match a group of non-participants with participants using variables like gender, household size, education, experience, land size (rainfall to control for drought), irrigation (as many observable charachteristics not affected by fertilizer) Matching in our example… 2 scenarios – Scenario 1: We show up afterwards, we can only match (within region) those who got fertilizer with those who did not. Problem? • Problem: select on expected gains and/or ability (unobservable) – Scenario 2: The program is allocated based on historical crop choice and land size. We show up afterwards and match those eligible in region A with those in region B. Problem? • Problems: same issues of individual unobservables, but lessened because we compare eligible to potential eligible • now unobservables across regions An extension of matching: pipeline comparisons • Idea: compare those just about to get an intervention with those getting it now • Assumption: the stopping point of the intervention does not separate two fundamentally different populations • example: extending irrigation networks 3. Difference-in-difference • Observations over time: compare observed changes in the outcomes for a sample of participants and non-participants • Identification assumption: the selection bias is timeinvariant (‘parallel trends’ in the absence of the program) • Counter-factual: changes over time for the nonparticipants Constraint: Requires at least two cross-sections of data, preprogram and post-program on participants and nonparticipants – Need to think about the evaluation ex-ante, before the program • Can be in principle combined with matching to adjust for pre-treatment differences that affect the growth rate Implementing differences in differences in our example… • Some arbitrary comparison group • Matched diff in diff • Randomized diff in diff • These are in order of more problems  less problems, think about this as we look at this graphically As long as the bias is additive and timeinvariant, diff-in-diff will work …. Y1 Impact Y1* Y0 t=0 t=1 time What if the observed changes over time are affected? Y1 Impact? Y1* Y0 t=0 t=1 time 4. Instrumental Variables • Identify variables that affects participation in the program, but not outcomes conditional on participation (exclusion restriction) • Counterfactual: The causal effect is identified out of the exogenous variation of the instrument • Advantages: – Does not require the exogeneity assumption of matching • Disadvantages: – The estimated effect is local: IV identifies the effect of the program only for the sub-population of those induced to take-up the program by the instrument – Therefore different instruments identify different parameters. End up with different magnitudes of the estimated effects – Validity of the instrument can be questioned, cannot be tested. IV in our example • It turns out that outreach was done randomly…so the time/intake of farmers into the program is essentially random. • We can use this as an instrument • Problems? – Is it really random? (roads, etc) 5.Regression discontinuity design • Exploit the rule generating assignment into a program given to individuals only above a given threshold – Assume that discontinuity in participation but not in counterfactual outcomes • Counterfactual: individuals just below the cut-off who did not participate • Advantages: – Identification built in the program design – Delivers marginal gains from the program around the eligibility cut-off point. Important for program expansion • Disadvantages: – Threshold has to be applied in practice, and individuals should not be able manipulate the score used in the program to become eligible. Figure 1: Kernel Densities of Discriminant Scores and Threshold points by region 3.9e-06 Density .002918 Density .00329 Density .003412 2.8e-06 759 Discriminant Score 0 753 Discriminant Score Region 3 751 Discriminant Score Region 4 5.5e-06 Density .003639 Density .004625 Density .004142 Region 5 8.0e-06 752 Discriminant Score Region 6 4.5e-06 571 Discriminant Score Region 12 691 Discriminant Score Region 27 Density .002937 .000015 757 Discriminant Score Region 28 Example from Buddelmeyer and Skoufias, 2005 RDD in our example… • Back to the eligibility criteria: land size and crop history • We use those right below the cut-off and compare them with those right above… • Problems: – How well enforced was the rule? – Can the rule be manipulated? – Local effect Discussion example: building a control group for irrigation • Scenario: we have a project to extend existing reaches and build some new canal • An initial analysis shows that farmers who are newly irrigated have increased yield…was the project a success? • What is the evaluation question? • What is a logical comparison group and method? Investment operation vs adjustment/budget support • Project – Maybe evaluate all, but unlikely • Pick subcomponents • Adjustment/budget support – Build a strong M&E unit • Impact evaluation designed by govt – Evaluate policy reform pilots – e.g. health insurance pilot, P4P, tariff changes – Anything economy wide ≠ impact evaluation Prioritizing for Impact Evaluation • It is not cheap – relative to monitoring • Possible prioritization criteria: – Don’t know if policy is effective • e.g. conditional cash transfers – Politics • e.g. Argentina workfare program – It’s a lot of money • Note that 2 & 3 are variants of not “knowing” – in this context, etc. Summing up: Methods • No clear “gold standard” in reality – do what works best in the context • Watch for unobservables, but don’t forget observables • Be flexible, be creative – use the context • IE requires good monitoring and monitoring will help you understand the effect size Impact Evaluation and the Project Cycle Objective of this part of the presentation • Walk you through what it takes to do an impact evaluation for your project from Identification to ICR • Persuade you that impact evaluation will add value to your project We will talk about… • General Principles • In the context of 3 project periods: – Evaluation activities – the core issues for evaluation design and implementation, and – Housekeeping activities—procedural, administrative and financial management issues • Where to go for assistance Some general principles • Government ownership as whole—what matters is institutional buy-in so that the results get used • Relevance and applicability—asking the right questions • Flexibility and adaptability • Horizon matters Ownership • IE can provide one avenue to build institutional capacity and a culture of managing-by-results – so the IE should be as widely owned within gov’t as possible • Agree on a dissemination plan to maximize use of results for policy development. • Identify entry points in project and policy cycles – midpoint and closing, for project; – sector reporting, CGs, MTEF, budget, for WB – Budget cycles, policy reviews for gov’t • Use partnerships with local academics to build local capacity for impact evaluation. Relevance and Applicability • For an evaluation to be relevant, it must be designed to respond to the policy questions that are of importance. • Clarifying early what it is that will be learned and designing the evaluation to that end will go some way to ensure that the recommendations of the evaluation will feed into policy making. • Make sure to to think about unintended consequences (e.g. export crop promotion shifts the intrahousehold allocation of power or S. Africa pensions) – qualitative and interdisciplinary perspectives are key here Flexibility and adaptability • The evaluation must be tailored to the specific project and adapted to the specific institutional context. • The project design must be flexible to secure our ability to learn in a structured manner, feed evaluation results back into the project and change the project mid-course to improve project end results. • Can be broad project redesign or push in new directions e.g. feed into nutritional targeting design This is an important point: In the past projects have been penalized for affecting mid-course changes in project design. Now we want to make change part of the project design. Horizon matters • The time it takes to achieve results is an important consideration for timing the evaluation. Conversely, the timing of the evaluation will determine what outcomes should be focused on. – Early evaluations should focus on outcomes that are quick to show change – For long-term outcomes, evaluations may need to span beyond project cycle. e.g. Indonesia school building project • Think through how things are expected to change over time and focus on what is within the time horizon for the evaluation Do not confuse the importance of an outcome with the time it takes for it to change—some important outcomes are obtained instantaneously ! But don’t be afraid to look at intermediate outcomes either Stage 1: Identification to PCN Get an Early Start How do you get started? • Get help and access to resources: contact person in your region or sector responsible for impact evaluation and/or Thematic Group on Impact Evaluation • Define the timing for the various steps of the evaluation to ensure you have enough lead time for preparatory activities (e.g. baseline goes to the field before program activities start) • The evaluation will require support from a range of policy-makers: start building and maintaining constituents, dialogue with relevant actors in government, build a broad base of support, include stakeholders Build the Team • Select impact evaluation team and define responsibilities of: – – – – – program managers (government), WB project team, and other donors, lead evaluator (impact evaluation specialist), local research/evaluation team, and data collection agency or firm Selection of lead evaluator is critical for ensuring quality of product, and so is the capacity of the data collection agency • Partner with local researchers and research institutes to build local capacity Shift Paradigm • From a project design based on “we know what’s best” • To project design based on the notion that “we can learn what’s best in this context, and adapt to new knowledge as needed” Work iteratively: – Discuss what the team knows and what it needs to learn–the questions for the evaluation—to deliver on project objectives – Discuss translating this into a feasible project design – Figure out what questions can feasibly be addressed – Housekeeping: Include these first thoughts in a paragraph in the PCN • e.g. ARV evaluation – funding constraints shifted radically, quickly – design changed, and changed again Stage 2: Preparation through appraisal Define project development objectives and results framework • This activity – clarifies the results chain (logic of impacts) for the project, – identifies the outcomes of interest and the indicators best suited to measure changes in those outcomes, and – the expected time horizon for changes in those outcomes. • This will provide the lead evaluator with the project specific variables that must be included in the survey questionnaire and a notion of timing for scheduling data collection. Work out project design features that will affect evaluation design • Target population and rules of selection – This provides the evaluator with the universe for the treatment and comparison sample • Roll out plan – This provide the evaluation with a framework for timing data collection and, possibly, an opportunity to define a comparison group • Think about non-objective undermining changes that will enhance the evaluation (and this will likely be iterative) Narrow down the questions for the evaluation • Questions aimed at measuring the impact of the project on a set of outcomes, and • Questions aimed at measuring the relative effectiveness of different features of the project Questions aimed at measuring the impact of the project are relatively straightforward • What is your hypothesis? (Results framework) – By expanding water supply, the use of clean water will increase, water borne disease decline, and health status will improve • What is the evaluation question? – Does improved water supply result in better health outcomes? • How can do you test the hypothesis? – The government might randomly assign areas for expansion in water supply during the first and second phase of the program • What will you measure? – Measure the change in health outcomes in phase I areas relative to the change in outcomes in phase II areas. Outcomes will include use of safe water (S-T), incidence of diarrhea (S/M-T), and health status (L-T, depending on when phase II occurs). Add other outcomes. • What will you do with the results? – If the hypothesis proves true go to phase II; if false, modify policy. Questions aimed at measuring the relative effectiveness of different project features require identifying the tough design choices on the table… • What is the issue? – What is the best package of products or services? • Where do you start from (what is the counterfactual)? – What package is the government delivering now? • Which changes do you or the government think could be made to improve effectiveness? • How do you test it? – The government might agree to provide a package to a randomly selected group of households and another package to another group of households to see how the two package perform • What will you measure? – The average change in relevant outcomes for households receiving one package versus the same for households receiving the other package • e.g. extension vs fertilizer+extension vs fertilizer+extension+seeds • What will you do with the results? – The package that is most effective in delivering desirable outcomes becomes the one adopted by the project from the evaluation onwards Application, features that should be tested early on • Early testing of project features (say 6 months to 1 year) can provide the team with the information needed to adjust the project early on in the direction most likely to deliver success. • Features might include: – alternative modes of delivery (e.g. use seed merchants vs. extension agents), – alternative packages of outputs, or – different pricing schemes (e.g. alternative subsidy levels). Develop identification strategy (to identify the impact of the project separately from changes due to other causes ) • One the questions are defined, the lead evaluator selects one or more comparison groups against which to measure results in the treatment group. • The “rigor” with which the comparison group is selected will determine the reliability of the impact estimates. • Rigor? – More-same observables and unobservables (experimental), – Less-same observables (non-experimental) Explore Existing Data • Explore what data exists that might be relevant for use in the evaluation. – Discuss with the agencies of the national statistical system and universities to identify existing data sources and future data collection plans. – Check DECDG website • Record data periodicity, quality, variables covered and sampling frame and sample size, for – – – – Censuses Surveys (household, firms, facility, etc) Administrative data Data from the project monitoring system New Data • Start identifying additional data collection needs. – Data for impact evaluation must be representative of treatment and comparison group – Questionnaires must include outcomes of interest (consumption, income, assets etc), questions about the program in question and questions about other programs, as well as control variables – The data might be at household, community, firm, facility, or farm levels and might be combined with specialty data such as those from water or land quality tests. • Investigate synergies with other projects to combine data collection efforts and/or explore existing data collection efforts on which the new data collection could piggy back • Develop a data strategy for the impact evaluation including: – – – – The timing for data collection The variables needed The sample (including size) Plans to integrate data from other sources (e.g project monitoring data) Prepare for collecting data • Identify data collection agency • Lead evaluator or team will work with the data collection agency to design sample, and train enumerators • Lead evaluator or team will prepare survey questionnaire or questionnaire module as needed • Pre-testing survey instrument may take place at this stage to finalize instruments • If financed with outside funds, baseline can now go to the field. If financed by project funds, baseline will go to the field just after effectiveness but before implementation starts Develop a Financial Plan • Costs: – – – – Lead evaluator and research/evaluation team, Data collection, Supervision and Dissemination • Finances: – – – – – BB, Trust fund, Research grants, Project funds, or Other donor funds Housekeeping • Initiate an IE activity. The IE code in SAP is a way of formalizing evaluation activities. The IE code recognizes the evaluation as a separate AAA product. – Prepare concept note – Identify peer reviewers –impact evaluation and sector specialist – Carry out review process • Appraisal documents – Include in the project description plans to modify project overtime to incorporate results – Work the impact evaluation into the M&E section of the PAD and Annex 3 • Include the impact evaluation in the Quality Enhancement Review (TTL). Stage 3: Negotiations to Completion Ensure timely implementation • Ensure timely procurement of evaluation services especially contracting the data collection, and • Supervise timely implementation of the evaluation including – Data collection – Data analysis – Dissemination and feedback Data collection agency/firm • Data collection agency or firm must have technical knowledge and sufficient logistical capacity relative to the scale of data collection required • The same agency or firm should be expected to do baseline and follow up data collection (and use the same survey instrument) Baseline data collection and analysis • Baseline data collection should be carried out before program implementation begins; optimally even before program is announced • Analysis of baseline data will provide program management with additional information that might help finalize program design Follow-up data collection and analysis • The timing of follow-up data collection must reflect the learning strategy adopted • Early data collection will help modifying programs mid course to maximize longerterm effectiveness • Later data collection will confirm achievement of longer-term outcomes and justify continued flows of fiscal resources into the program Watch implementation closely from an evaluation point of view • Watch (monitor) what is actually being implemented: – Will help understand results of evaluation – Will help with timing of evaluation activities • Watch for contamination in the control group • Watch for violation of eligibility criteria • Watch for other programs for the same beneficiaries • Look for unintended impacts • Look for unexploited evaluation opportunities  Good evaluation team communication is key here Dissemination • Implement plan for dissemination of evaluation results ensuring that the timing is aligned with government’s decision making cycle. • Ensure that results are used to inform project management and that available entry points are exploited to provide additional feedback to the government • Ensure that wider dissemination takes place only after the client has had a chance to preview and discuss the results • Nurture collaboration with local researchers throughout the process Housekeeping • Put in place arrangements to procure the impact evaluation work and fund it on time • Use early results to inform mid-term review • Use later results to inform the ICR, CAS and future operations Summing up: Practicalities • Making evaluation work for you requires a change in the culture of project design and implementation, one that maximizes the use of learning to change course when necessary and improve the chances for success • Impact evaluation is more than a tool – it is an organizing analytical framework for doing this – it is not about measuring success or failure so much as it is about learning… Where to go for assistance / more information • Clinics – Brochure here, PREM • TG resources – – – – Searchable database of evaluations Searchable roster of consultants Doing IE series – general and sector notes Website (http://impactevaluation) • Courses – workshop on IE, WBI training, PAL course • South Asia resources: Jishnu Das (12/06) Thank you