计量经济学_metrics___experiments.ppt

Econometrics Econ 304/404/543 Instructor Ric Scarpa Lecture 15 Experiments and causal effects In psychology and medicine when researchers are interested in a casual effect they set-up randomized experiments For examples, drugs are tested in experimental trials Effectiveness in healing is measured by the differences between a group given the drug and a group given a placebo Why studying the analysis of data from randomized controlled experiments? They give you a conceptual benchmark for studying a causal relationship The outcome of experiments can be very influential, so it is important to understand the potential limitations of such studies External circumstances can cause randomization. That is, the treatment of some people can be as-if due to chance Quasi-experiments This last case takes the name of quasi-experiments This have something in common with experiments and the study of the latter can cast some light into how to handle the former The peculiarity of this subject is not given by the type of tools used, but by the special challenges posed by this type of data Program evaluations The methods developed here are most frequently used in economics to evaluate programs This is the field concerned with the estimation of the effects due to a policy, an intervention or a “program” For example, what is the effect on earning of going through a training program? What is the effect of a re-training program for home-makers with school age children on their ability to gain employment? Idealized experiments and causal effects A randomized controlled experiment: randomly draws subjects from the population of interest Randomly assigned them to a “treatment” group and to another “control” group (with no treatment) Analyzes data on the differences between groups which are causally related to the treatment Ideally…. We would like to keep everything else equal across groups except the treatment However, this is beyond our control What matters is that assignment to treatment