Good governance has become a core theme of development policy in the past decades. Each year, billions of dollars were spent to support governance-related programmes. Yet, despite the efforts of donors, very little is still understood about the impact of these programmes.
Rigorous impact evaluation measures the extent to which observed outcomes can be attributed to the intervention using experiments or quasi-experiments. Unlike in traditional sectors such as health, labor and education, rigorous evaluations in governance remain rather limited. There are several reasons for this: (i) governance outcomes are difficult to quantify, (ii) governance is characterized by complex interventions, (iii) lack of incentive to do evaluations due to the politically sensitive nature of governance topics, (iv) lack of reliable baseline and longitudinal data, (v) small sample size, among others.
Despite these challenges, evidence suggests that some aspects of governance programmes have been subjected to rigorous impact evaluation. For example, randomized control trials have been employed to evaluate governance issues like corruption, political systems, elections, and community development. Quasi-experimental designs have been utilized to evaluate interventions concerning public administration, political systems, elections, crime, participation, and government spending behaviour.
The discussion regarding the challenges of conducting rigorous impact evaluation in governance has practical implications for program design and implementation. To boost the conduct of rigorous impact evaluations, it is important to:
- Employ an evaluator in the early phases of programme design
- Cooperate with the relevant government agencies: they are important sources of information on the nature of the target beneficiaries and can help establish the likelihood of programme take-up
- Identify from the outset components of the programme that permit quantitative analysis
- Collect data for the control group. It is important to set aside a budget to fund control groups.
Furthermore, lessons from other sectors can be helpful in overcoming technical problems. For example:
- Use context-specific information or perception surveys to capture hard-to-measure outcomes
- The lack of baseline data can be compensated for by baseline reconstruction or using single difference like propensity score matching
- Higher levels of government can be evaluated using cluster randomisation
- Complex governance interventions can be investigated by unpacking the intervention into smaller dimensions
- Programmes covering the entire population can be rolled out in stages to enable impacts to be measured.
