Using statistical learning methods for impact evaluation: the effect of European Union membership on economic growth

  • Title: Using statistical learning methods for impact evaluation: the effect of European Union membership on economic growth
    Year: 2015
    Keyword(s): impact evaluation, statistical learning
    Description: Studies known as impact evaluation or treatment effect evaluation are traditionally based on regression models that include categorical covariates. Another method, known as "synthetic counterfactual analysis" tries to determine how some variables would have evolved in the absence of the event of interest. This paper proposes "statistical learning" as an alternative method of impact evaluation. Statistical learning fits an econometric model to a subset of the data (the training set) and tests its predictions on another subset of the data (the test set). The parameters of the econometric model are determined by evaluating the model’s performance in predicting the variable of interest in the test data subset. Using prediction methods in impact evaluation problems is a novelty. The method is exemplified on the effect of EU membership on a member country's GDP. The results, however, are not yet satisfactory, probably since the current statistical learning methods are not suitable for panel data.
    Notes: Presented on May 30, 2015 at the Canadian Economics Association conference held at Ryerson University in Toronto, Ontario.
    Peer Reviewed: No
    Type of Item: Conference Materials
    Publication Information: Colonescu, C. (2015, May). Using statistical learning methods for impact evaluation: The effect of European Union membership on economic growth. Paper presented at the conference of the Canadian Economics Association, Toronto, ON. Retrieved from https://economics.ca
    Language: English

72 Views