Our approach

Three complementary dimensions of evaluation.

Estimate causal effect

Estimating the actual effects of public interventions on the affected populations.

The scientific evaluation of the effects of public interventions on the socioeconomic, educational, health, and environmental conditions of populations contributes to the improvement of individual and collective well-being. In particular, it aims to establish a cause-and-effect relationship between public interventions and the changes observed across these various dimensions.

Analyze heterogeneous effects

Estimating differences in actual effects across intervention beneficiary groups.

Estimating, wherever possible, the heterogeneity of the effects of public policies and programs across program beneficiary groups. For example, a program with an unfavorable average effect may nonetheless produce beneficial effects for certain participants. An evaluation based solely on the average effect therefore risks masking significant differences between beneficiary groups.

Inform public policy decisions

Synthesizing scientific findings to support the improvement of policies and programs.

This approach takes place in a context where governments set their directions and priorities according to the political commitments that brought them to power. Scientific research findings can therefore provide them with empirical insights into the expected or observed effects of public policies and help guide their adjustment in light of the evidence.

Evaluation methods

A diverse range of approaches to address different questions.

Econometric Methods

  • OLS and fixed effects models
  • Instrumental variables
  • Difference-in-differences
  • Regression discontinuity design
  • Matching propensity scores
  • Others

Statistics and Machine Learning

Statistical and algorithmic tools to analyze complex relationships and differences in effects across population groups :

  • Causal Random Forests

Robustness of Causal Estimates

Assess the sensitivity of results to omitted variable bias :

  • Oster approach
  • Rosenbaum test
  • Robustness Checks
Scroll to Top