Looking into the Black Box: Using Experimental Data to Find out How Training Works

Considers the relative effectiveness of using experimental training programmes. Uses the US experience of programme evaluation to make two main points. First, many of the key questions concerning the effects of training cannot be answered using simple estimators, even in the presence of experimental data. Second, addressing such issues in a non‐experimental setting raises additional problems, particularly the need to model the selection process by which individuals enter training.