Measuring Impact of Certification: Towards a Focus for Future Research ISEAL Conference, June 8, 2011 ©2009 Rainforest Alliance Demonstrating Impact: Now and Tomorrow MYTH ANECDOTE OBSERVATION COMPARISON REPLICATED RRC Documented Now Academic Practitioners Academic Future Standard Indicators Research Strategy Practitioners BUSINESS MODEL OF CERTIFICATION Consumer behavior (values) Retailers Brands Manufacturers Traders Processors Sustainability feedback: Demand for sustainable markets Business Practices in Value Chain Production Unit (Best Management Practices) Measureable Ecological, Social, Economic benefits Sustainability feedback: Evidence of Impact WHAT IS THE HYPOTHESIS TO TEST? All study designs must first start with a question or set of questions to be answered in the evaluation. For example….. • What are the monetary costs and benefits of implementing the SAN standard (including price premiums), and how do they change over the lifetime of a certificate? • How do farm productivity, crop quality and soil health compare between farms that do and do not implement best management practices we assume change these conditions? What are the independent variables (Controlled and varied by study) and dependent variables (determined by independent variable)? 4 HOW TO DEVELOP EVALUATION AND RESEARCH MODEL What is the meaningful Independent Variable? Consumer behavior (values) Retailers Brands Manufacturers Traders Processors Sustainability feedback: Demand for sustainable markets Business Practices in Value Chain What is the meaningful Independent Variable? Production Unit (Best Management Practices) Measureable Ecological, Social, Economic benefits Sustainability feedback: Evidence of Impact EVALUATION OF INTERVENTIONS NOT BUSINESS MODELS Unbundling certification to measure adoption of best management practices (BMPs). Operation 1 BMP1 BMP2 BMP3 BMP4 Operation 2 BMP1 BMP2 BMP3 BMP4 Operation 3 BMP1 Operation 4 No BMPs BMP4 Gains us a more nuanced understanding of what production and social and environmental management practices drive what economic, social or ecological returns, and under what conditions. Informs learning and refinement of standards. 6 Level of confidence Dependent variables. Determined by control and variation of independent variable Reduced encroachment Independent variable. Controlled and varied by study Indirect Adoption of RA best management practices Productivity Action Costs of measuring change Desired Impact? Time to see an impact DETERMINING HYPOTHESIS BASED ON COST, LENGTH OF STUDY AND LEVEL OF CREDIBILITY NECESSARY Hypothesis TESTING: Research Strategy Research that tests hypothesis by reaching conclusions about relationship between dependent and independent variables is Before and after implementation of BMPs/Certification attempt to discover new information Control group Independent variable Exposed to all elements of the research, except the intervention (experimental) variable are Determined by independent variable are is is Control and varied Dependent variable Intervention (experimental) group demonstrate that what we say we do is true examine the validity of a project’s hypothesis Exposed to the intervention (experimental) variable 8 Hypothesis TESTING: Research Strategy Research that tests hypothesis by reaching conclusions about relationship between BMPs and farm income variables BMP(s) Farms with no BMPs Exposed to elements such as international prices for coffee, advantageous weather conditions or improvements in processing and marketing,. Farms with BMPs are Determined by control and variation of BMPS attempt to discover new information are is is Controlled and varied by RA Real Farm Income is Before and after implementation of BMPs/Certification demonstrate that what we say we do is true examine the validity of a project’s hypothesis Exposed to all elements + BMPs 9 CONTRIBUTION VS. ATTRIBUTION DEMONSTRATING CONTRIBUTION OF RA PRACTICES • Measure before and after implementation of BMP to see changes in social, economic and environmental conditions. • The implicit assumption is that had farms not applied the BMP, their outcomes, on average, would have stayed the same. • Results will show impact of BMP but will not isolate the influence of these practices from the impact of factors unrelated to BMP implementation, e.g. international prices also influences income, advantageous weather conditions also influences productivity, a government intervention also improves working conditions. • BMP implementation must be presented as just ONE of a possible number of factors influencing desired outcomes. 10 CONTRIBUTION VS. ATTRIBUTION DEMONSTRATING ATTRIBUTION OF RA PRACTICES • Measure before and after implementation to see changes in social, economic and environmental conditions. • At the same time also measure changes in social, economic and environmental on farms that DO NOT implement BMP as the counterfactual outcomes (comparison group). • The implicit assumption is that had farms not implemented BMP , their outcomes would be the same, on average, as those of farms in the comparison group. • With a comparison group, the impact of BMP implementation can be defined as the difference between the observed intervention outcome and the counterfactual outcome. • BMPs can be attributed to desired outcomes because counterfactual outcome accounts for changes as a result of influencing factors, except BMP implementation 11 CONSTRUCTING A CREDIBLE COUNTERFACTUAL • A credible counterfactual must address selection bias, a problem that will violate assumptions when farms select themselves – or are selected by NGOs/traders – into certification. • 2 approaches to eliminate selection bias – Experimental – Quasi-experimental. • Each varies in feasibility, cost, the degree of clarity and scientific validity of results. 12 EXPERIMENTAL DESIGN (OPTIMAL APPROACH) • Involves gathering a set of farmers equally eligible and willing to participate in certification and randomly dividing them into two groups: those who receive the technical assistance (treatment group) and those from whom the intervention is withheld (control / counterfactual group). Advantages Challenges Random assignment of farms serve as a perfect counterfactual, free from selection bias (assuming sufficient sample size). Perhaps unethical and political owing to the denial of certification. Simplicity in interpreting results as true impact — difference between the means of both groups. Can be expensive and time consuming, particularly in the collection of longitudinal data points. Farms in treatment or control groups may change certain identifying characteristics during the experiment that could invalidate or contaminate results. 13 QUASI-EXPERIMENTAL DESIGN • Consists of constructing a comparison group using matching comparisons. This involves identifying non-certified farms comparable in essential characteristics to certified farms. Both groups should be matched on the basis of very similar observable characteristics that plausibly affect outcomes • Ideally matched comparison groups should be selected before project implementation, not afterwards. Advantages Can draw on existing data sources and are thus often quicker and cheaper to implement Disadvantages Reliability of the results is often reduced, as the methodology may not completely solve the problem of selection bias Matching methods can be statistically complex, thus requiring considerable expertise in the design of the evaluation and in analysis and interpretation of the results. 14 META ANALYSIS OF LITERATURE • Looks at set of specific BMPs across standard systems (Ag, Forestry, Fisheries) as list of independent variables • Categorize by dependent variables examined • Method • Statistical analysis applied • Result - Statistical significance • Geographic scope / focus • Time frame adequate - based on dependent variable (Oak tree versus Sunflower) 15 The Rainforest Alliance works to conserve biodiversity and ensure sustainable livelihoods by transforming land-use practices, business practices and consumer behavior.