SUMMARY
In the article “How much can we generalize from impact evaluations?” The author has described that Evaluation is considered an instrument to understand systematically how an intervention affects results. In the context of developmental economics, the role and the significance of impact evaluation could be immense, as it provides information regarding the variance in the results. It is a known fact that an economic model can only consider a few factors that strongly impact the results of similar projects implemented in different settings. The factors that are not considered also impact the results and they cause the variance in them. When we study some identical projects, being implemented at different locations (different contexts), we learn that these projects are producing slightly different results. These variations in results may facilitate identifying factors that bring the variance in results.
The article asserts that meta-analysis can be emphasized to extrapolate from the results of various projects. However, how accurate these inferences would depend largely on the credibility of the basic measures of heterogeneity. The article also points out that it is imperative to have a clear understanding about the estimation of models, as the accuracy of the outcomes related to the variance in results because of the designed interventions, depend upon the estimation of models (Vivalt).
According to author, there are the two types of models, which the article suggests, are 1) Fixed-effect Model and 2) Random-effect Model. The fixed Effect model assumes that there is only one type of effect of a particular type of intervention (project) and the variances that occur between different studies (about intervention). In contrast, the random-effect model does not make such an assumption, and it does not consider variation in different studies as an error term. The article suggests that for these estimations, the random-effect model is more appropriate. Also, the article also reveals that the application of the random-effect model can be broadened if an explanatory variable is added to the model. The explanatory model will transform the random-effect model into a mixed model (in the paper random and mixed models are considered because of their broader applicability and objective of the study).
Impact Evaluation provides a unique opportunity for policymakers to learn about the effectiveness of a program without truly implanting it. Generally, interventions are estimated after their implementation; however, Impact Evaluation that develops models to estimate the results of similar projects will provide essential information, which could improve the intervention.
The magnitude of the variation between different projects/interventions is essential to estimate. However, these estimations will not be valid, if they lack context. Also, the purpose of the project also directly affects the impact evaluation. For instance, some of the projects have a great margin of error, in comparison to the others. The room for error, in a development project, relies solely on the objective of that project. The study suggests that a statistical relationship can be established between measures that to be taken and the results of similar projects that are implemented in different settings.
The author has mentioned that in recent years, emphasis on impact evaluation has increased as it allows more rational and optimal allocation of the scarce resource. International organizations and governments are criticized for having a casual attitude regarding major projects. Also, large projects tend to produce such results that are slightly different from the projected results. Therefore, impact evaluation is required as an instrument to evaluate the results. However, policymakers must be cautious, as impact evaluation has not been refined as an instrument (Vivalt).
Work Cited
Vivalt, Eva. “How much can we generalize from impact evaluations?” Austrial National University (2017): 1-67.
+447462439809