Narrowing the evaluation gap Session 2.A. Not just window dressing – how IPE can be used to understand effects from impact evaluations 2018 EEF Evaluators’ Conference Narrowing the evaluation gap #EEFeval18 @EducEndowFound
By re-thinking the design and use of logic models Using the IPE to better understand the effects from impact evaluations By re-thinking the design and use of logic models Dr Bronwen Maxwell
Some potential contributions of an IPE based on theory-based evaluation principles Distinguish between programme, implementation and methodological deficiency Strengthen learning by exploring variation among different levels of implementation and different contexts Understand programme success or failure Unpack the 'black box' Understand causal logic and capture causal mechanisms Address limitation of experimental designs (why and how?) Understand causation A description of how an intervention leads to change that lies between the minor but necessary working hypotheses … and the all-inclusive systematic efforts to develop a unified theory Develop middle range theory [EEF] Evaluators should develop an intervention logic model or theory of change in partnership with the delivery team to inform the evaluation [...] It is important to know not just if an intervention ‘works’ in terms of producing desired outcomes, but also if it works in the manner theorised." (Humphrey et al, 2015 p.9).
Some potential contributions of an IPE based on theory-based evaluation principles Distinguish between programme, implementation and methodological deficiency Strengthen learning by exploring variation among different levels of implementation and different contexts Understand programme success or failure Unpack the 'black box' Understand causal logic and capture causal mechanisms Address limitation of experimental designs (why and how?) Understand causation A description of how an intervention leads to change that lies between the minor but necessary working hypotheses … and the all-inclusive systematic efforts to develop a unified theory Develop middle range theory
Some potential contributions of an IPE based on theory-based evaluation principles Distinguish between programme, implementation and methodological deficiency Strengthen learning by exploring variation among different levels of implementation and different contexts Understand programme success or failure Unpack the 'black box' Understand causal logic and capture causal mechanisms Address limitation of experimental designs (why and how?) Understand causation A description of how an intervention leads to change that lies between the minor but necessary working hypotheses … and the all-inclusive systematic efforts to develop a unified theory Develop middle range theory
Potential contribution of logic models to an IPE Articulate thinking behind programme working so this can be tested Gather data on implementation and contextual variation Understand programme success of failure Articulate causal logic Provide frame work to explore causal logic and mechanisms, including independent but inter-related mechanisms Understanding causation Explore plausibility of underpinning middle range theory Develop middle range theory
Some issues with the use of logic models in EEF evaluations Issues re: explanation of causality and complexity Programme and implementation theory not distinguished Implementation theory overemphasised Causal mechanisms subsumed into arrows Causal mechanisms poorly evidenced There maybe alternative and/or recursive causal paths There may be independent but inter-related causal paths Ignores context and concurrent programmes
Some issues with the use of logic models in EEF evaluations Issues re: explanation of causality and complexity Programme and implementation theory not distinguished Implementation theory overemphasised Causal mechanisms subsumed into arrows Causal mechanisms poorly evidenced There maybe alternative and/or recursive causal paths There may be independent but inter-related causal paths Ignores context and concurrent programmes
Some issues with the use of logic models in EEF evaluations Issues re: explanation of causality and complexity Programme and implementation theory not distinguished Implementation theory overemphasised Causal mechanisms subsumed into arrows Causal mechanisms poorly evidenced There maybe alternative and/or recursive causal paths There may be independent but inter-related causal paths Ignores context and concurrent programmes
Some issues with the use of logic models in EEF evaluations Issues re: explanation of causality and complexity Programme and implementation theory not distinguished Implementation theory overemphasised Causal mechanisms subsumed into arrows Causal mechanisms poorly evidenced There maybe alternative and/or recursive causal paths There may be independent but inter-related causal paths Ignores context and concurrent programmes
Some issues with the use of logic models in EEF evaluations Issues re: context Oversimplifies context as a set of features Context is dynamic and maybe agentic Contextual factors are relational Change is immanent rather than external Contextual factors are socio-historically and culturally located
Some issues with the use of logic models in EEF evaluations Issues re: context Oversimplifies context as a set of features Context is dynamic and maybe agentic Contextual factors are relational Change is immanent rather than external Contextual factors are socio-historically and culturally located
Some issues with the use of logic models in EEF evaluations Issues re: context Oversimplifies context as a set of features Context is dynamic and maybe agentic Contextual factors are relational Change is immanent rather than external Contextual factors are socio-historically and culturally located
Some issues with the use of logic models in EEF evaluations Issues re: context Oversimplifies context as a set of features Context is dynamic and maybe agentic Contextual factors are relational Change is immanent rather than external Contextual factors are socio-historically and culturally located
Example of issues of causation, complexity and context: EEF Booktrust Summer Active Evaluation
Moving towards better understanding of effects Logic models developed using research evidence on : Evidence-informed logic model The limitations of logic models In relation to causation, complexity and context The initiative being evaluated Likely causal and implementation processes, in context and taking account of complexity Evidence relating to connections between elements in implementation paths and to the causal processes that theorise how these connections occur and lead to outcomes.
Moving towards better understanding of effects Logic models developed using research evidence on : Evidence-informed logic model The limitations of logic models In relation to causation, complexity and context The initiative being evaluated Likely causal and implementation processes, in context and taking account of complexity Evidence relating to connections between elements in implementation paths and to the causal processes that theorise how these connections occur and lead to outcomes.
Moving towards an evidence based logic model Building the model Knowledge mobilisation literature crucial to identify three core causal mechanisms, detail on implementation processes, recognise complexity/ non-linearity of knowledge 'transfer'. Contextual factors -drawn from literature on research use Benefits Directing data gathering and analysis in ways that may not have been apparent otherwise Led to conclusion that effective advocacy-based scale-up has three key and inter-related system components - each of which different sets of enabling characteristics Iterative combination of research evidence and conversations to create underpinning hypothesis
Improving the impact of teaching assistants: EEF scale-up campaign
Moving towards an evidence based logic model Building the model: Multi-strand complexity/ causal and implementation processes Effective professional learning leading to teacher change Effective pedagogical strategies leading to pupil outcomes Professional learning and communities improving teacher retention Benefits Supported the design of tools and analysis for a pilot evaluation Provided evidence of the path from inputs to outputs- supporting the plausibility of the model Identified crucial programme components and ways of strengthening hypothesised links between inputs and outcomes.
Evaluation of the RETAIN programme : CPD for early career KS1 teachers
Key components of an evidence-informed logic model
Evaluating implementation and mechanisms Chris Bonell Professor of Public Health Sociology London School of Hygiene and Tropical Medicine
Process evaluation evaluates 2 kinds of processes Processes of implementation – actions providers and clients enact as part of the delivery e.g. train teachers to do restorative practice; teach social & emotional skills curriculum; participate in school-wide action group Mechanisms of impact – the mechanisms that interventions trigger which generate outcomes e.g. developing new skills; increasing commitment to school; developing new relationships
Processes of implementation Quantitative metrics: fidelity/quality; reach; dose delivered/received; acceptability - measured via structured observations/videos, logbooks, surveys etc. Qualitative accounts - measured via semi-structured observations, interviews, focus groups etc. Frameworks e.g. Carl May’s normalisation process theory – coherence, cognitive participation, collective action, reflexive monitoring Develop and test hypotheses about how context affects implementation e.g. baseline capacity and culture Rationale - assess whether failure of implementation or theory explains null result; refinement of implementation; assessment of potential transportability of intervention
Mechanisms of action Realist evaluation: what works for whom, how and under what conditions? Focused on ‘context-mechanism-outcome configurations’ e.g. engaging students in school to reduce risk behaviours will work better in schools with low baseline engagement and supportive SLT Hypotheses from formal or informal theory Refine with qualitative research Test with statistical analyses of mediation (what proximal outcomes explain distal outcomes?) and moderation (what baseline variables account for variation in impacts?) Rationale – further assessment of potential transportability of intervention; better theory
Mechanisms of action Realist evaluation: what works for whom, how and under what conditions? Develop, refine and test context-mechanism-outcome configurations e.g. engage students in school to reduce risk behaviours will work better in schools with low baseline engagement and supportive SLT Hypotheses from formal or informal theory Refine with qualitative research Test with statistical analyses of mediation (what proximal outcomes explain distal outcomes?) and moderation (what baseline variables account for variation in impacts?) Rationale – further assessment of potential transportability of intervention; better theory
QUESTIONS FOR SESSION 2.A ON IMPLEMENTATION AND PROCESS EVALUATION MEHODS How well do current IPE approaches support our understanding of impact results, particularly when there are null effects? How can IPE be designed to enhance impact evaluations, particularly within QED and RCT designs? What specific models or frameworks are most useful? Are there any approaches, models or frameworks that you think are underexploited? Thinking particularly from a design perspective, what elements or qualities make for good IPE? Conversely, what makes for poor quality IPE? #EEFeval18 @EducEndowFound