idsevalseminar - Centre for Development Impact

Evaluability Assessments and Choice of
Evaluation Methods
Richard Longhurst, IDS
Discussant: Sarah Mistry, BOND
Centre for Development Impact
19th February 2015
Introduction and some health warnings
• Some acknowledgements and thanks
• How this work came about: multilateral agency experience as
well as some review of literature
• Evaluability assessments (EAs) are not new, go back 25 years
• Will try to avoid getting bound up in the technical aspects ….
some of this will seem common sense …..but what matters is
trying to make explicit the basis on which decisions are
made… and how they relate to the culture of the organisation
• It is important to make judgements about choice of evaluation
methods (as this is a CDI event) and what drives choices. The
EA literature beginning to enter debate of choice of methods
• In the scope of this seminar, will not be covering every
evaluation method
Context of this work with International Programme
for the Elimination of Child labour (ILO-IPEC)
• Large technical cooperation programme (since 1992) largely funded
by US Dept. of Labor
• Causes of child labour are multi-faceted, approaches to eliminate
are equally various
• Main programme tool is Programme of Support to the national
Time Bound Programme to reduce the worst forms of child labour
• TBP involved ‘upstream’ enabling environment and ‘downstream’
action support to reduction of child labour, therefore mix of
• Also project and global interventions: at its peak IPEC carrying out
25 evaluations per year
See: Perrin and Wichmand (2011) Evaluating Complex Strategic Interventions: The
Challenge of Child Labour in Forss, Marra and Schwartz (eds), Transaction Publ.
Context: IPEC Evaluation approaches and
Information Sources
National Household Surveys
Baseline Surveys
Rapid Assessment Surveys
Child Labour Monitoring Systems and programme monitoring
Tracking and Tracer studies
One on one interviews; Focus groups
Document Analysis, Observation, Case studies
Impact and outcome evaluations, expanded final evaluations
Success case method and most significant change
Use of SPIF: strategic planning and impact framework
Context: My baseline at Commonwealth
Secretariat (1995-2002)
Starting up an expanded evaluation function
Conservative, diplomatic based organisation
An organisation with many small (<£50K) projects
About 4-5 project evaluations plus one strategic review of the
political function
• Evaluation worked with planning function and reported direct
to CEO with oversight from GB
• Many projects were hard to evaluate because of their design
• Evaluability regarded as achieved through adherence to the 2
year strategic plan
Current Use of EAs
• Use of EAs is growing:
• After their popularity in the US in the 1980s, EA guidance has
been developed by ILO, CDA, IDRC, EBRD and UNODC, with
(a German NGO).
• Encouraged by the International Financial Institutions (IFIs)
• Over half of EAs were for individual projects (balance were
country strategies, strategic plans, work plans and
Some definitions of EA from multilaterals
• OECD-DAC: ‘the feasibility of an evaluation is assessed … it
should be determined whether or not the development
intervention is adequately defined and its results verifiable,
and if evaluation is the best way to answer questions posed by
policy makers or stakeholders’. (broad)
• Evaluation Cooperation Group of the IFIs: ‘The extent to which
the value generated or the expected results of a project are
verifiable in a reliable and credible fashion’ (narrow but
• World Bank: ‘A brief preliminary study undertaken to
determine whether an evaluation would be useful and feasible
…. It may also define the purpose of the evaluation and
methods for conducting it’. (says something about methods)
Process for EAs (i)
• Common steps include (Davies):
– Identification of project boundaries
– Identification of resources available for EA
– Review of documentation
– Engage with stakeholders, then feedback findings
– Recommendations to cover: project logic and design, M&E
systems, evaluation questions of concern to stakeholders
and possible evaluation designs.
Process for EAs (ii) – Incorporating approaches
for methods
• Mapping an analysis of existing information
• Developing the theory of change to identify evaluation
questions noting linkages to changes attributable to
• Setting out priorities, key assumptions and time frames
• Choosing appropriate methods and tools
• Ensuring resources are available for implementation
• Outline reporting and communicating results of evaluation
Issues for an EA
• Review of guidance documents of international
agencies suggest EAs should address three broad
– Programme design
– Availability of information
– Institutional context (including breadth of stakeholders)
EA Tools (i)
• Checklists are normally used: ILO covers five main areas:
– Internal logic and assumptions
– Quality of indicators, Baselines, Targets and Milestones
– Means of verification, measurement and methodologies
– Human and Financial resources, and
– Partners’ Participation and use of information
(and ILO uses a rating system for this).
Don’t knock checklists, there is always a theory of change
embodied in them
An independent consultant is usually employed
EA tools (ii) to lead to choice of methods
• EA can be the focus for a modified design workshop that
brings together staff and participants involved in all stages of
the intervention (e.g. use of SPIF)
• Helps develop a stronger theory of change
• Can strengthen monitoring and needs for other information
• Can defuse suspicions about evaluations
• Can be very useful when a Phase I has been completed and a
Phase II has been proposed, building on an evaluation
• Allows ‘lessons learned’ from Phase I to be properly
Experience from using EAs (i)
• Generally EAs have been a good thing:
– Improved usefulness and quality of evaluations: an advance on when
evaluator arrived at the end of the project and finding no means to
– Early EAs dependent on logic models and linearity, now some signs
they are being broadened
– An opportunity for an early engagement with stakeholders, i.e. more
– Some evidence of improvements in project outcomes as well as design
– More resources applied up front helps address later problems
Experience from using EAs (ii)
• Some of the difficulties:
– Clash of work cultures between design and evaluation professionals –
working to different incentives and time scales
– Issues of how far the evaluation ‘tail’ wags the design ‘dog’, leading to
some ethical issues
– Have to be prepared for ‘cats’ put among ‘pigeons’ if there are
significant gaps in design; does it mean intervention is stopped ?
– Evaluators must not get too seduced by what EAs can achieve,
especially if original intervention design is weak
– EAs will not work everywhere and must always be light touch - there
will be a budget constraint
– Other techniques may be more appropriate (e.g. DFID approach
Linking to Evaluation Methods
• Using the starting point of Stern et al (2012) Broadening the
range of designs and methods for Impact evaluations, DFID
working Paper No 38.
– Selection of appropriate evaluation designs has to satisfy three
constraints or demands:
– Evaluation questions
– Programme attributes
– Available evaluation designs
Some criteria for choice of methods based on
the results of the EAs (criteria will interact)
• Purpose of the evaluation
• Level of credibility required: what sort of decisions will be
made on the basis of the evaluation?
• What does the agency know already, i.e. nature of existing
information and evidence
• Nature of intervention and level of complexity
• The volume of resources and nature of capacity available to
carry out the evaluation
• Governance structure of the implementing agency and
relationship with partners
Purpose of the evaluation
• This is the overarching framing question (so EA can make this
• Relates to the position of the intervention in the agency’s
planning structure and how evaluation has been initiated
• Any special role for stakeholders
• Is the evaluation being implemented for accountability,
learning or ownership purposes or for wider process
• Nature of topic: project, country, thematic, global, programme
• To set up an extension of an intervention
Level of credibility of evaluation results and
decisions to be made
• How does the decision maker need to be convinced? Independence
of the process ?
• How will the evaluation be used? What sort of evaluation
information convinces policy makers?
• What is the nature of the linkages between results and
– Attribution
– Contribution
– Plausible attribution
• If attribution is required with a need for a ‘yes/no it works/or not’
decision, then have to choose an impact evaluation
• If contribution is required, then can use contribution analysis
• If ‘plausible attribution’ is required then can use an outcome
summative method.
Other common observations on method choice
(relates to criterion of credibility)
• Experimental: demonstrates counterfactual, strong on
independence, requires both treatment and control
• Qualitative: strong on understanding, answers ‘why?’ ,
difficult to scale up findings
• Theory based and realistic evaluation: compatible with
programme planning, strong emphasis on context, requires
strong ToC
• Participatory: provides for credibility and legitimacy, enhances
relevance and use, can be time consuming
• Longitudinal tracking: tracks changes over time and can
provide reasons for change, can be resource intensive
What does the agency and its partners already
know ?
• No need to repeat evaluations if they do not add to the
agency’s ability to take decisions (value of DFID writing
approach papers)
• Role of information banks outside the agency (e.g. systematic
reviews, research studies); external validity
• Have all stakeholders been involved with information
gathering at the design stage
• How strong is the M&E, will the ‘M’ be useful for the ‘E’
• Have worthwhile decisions been made in the past on existing
information, good enough for sound design
• Is some form of comparison group required ?
Nature of the intervention and level of
• Key question on complexity is: what is the level of complexity/
reductionism at which an intervention is implemented and an
evaluation can be carried out
• Do the findings of the evaluation provide the basis for going
ahead to make a decision ?
• If complexity is addressed in design through multiple
intervention components, some where the n=1 (addressed to
governments), some where n=thousands (addressed to
children), then different evaluation methods can handle this.
• But, what do we know already that allows the evaluator to
compromise on complexity ?
Resources and capacity
• Much choice comes down to the budget line, what the
evaluation staff know and how much they are willing to take
risks on unfamiliar methods (e.g. realist evaluation) and the
time lines they work to
• There are opportunities for methods to be applied differently
based on criteria already mentioned.
• Some agency staff describe the ‘20 day’, ‘30 day’ etc.
evaluation method, defined by the resources they have
• This is why the ‘outcome summative’ method is so popular
and why efforts should be made to improve it.
Governance Structure of the Agency
• Always remains a key issue as structure often inhibit risk
taking by the evaluators
• Role of the governing body and executive varies in terms of
what evaluators can do.
Importance of strengthening the ‘outcome
summative’ evaluation
• Still remains the most common evaluation method (over 75%
of evaluations ?) but not much covered in recent literature
• Large element of evaluator’s judgement involved, familiar,
convenient, inexpensive
• But considering other factors for choice it can become the
best choice: plausible attribution, aligned closely with other
information sources, acknowledges deficiencies in addressing
complexity, borrows ideas from other more rigorous
techniques such as some form of comparison group of
retrospective baseline.
Thank you !