A treatment effect is the causal effect of some "treatment" or policy intervention on an outcome variable. Such effects can be estimated with data from randomized or quasi experiments, and clinical trials or with observational data and methods for causal inference.
A treatment effect is the causal effect of some "treatment" or policy intervention on an outcome variable. Typical examples are the effect of participation in a job market program or the effect of a particular drug. The usual difficulty is to control for selection bias which arises if treated units are different from non-treated units due to reasons which are unrelated to the treatment itself. This can be achieved by utilizing data from randomized or quasi experiments, and clinical trials or with observational data and methods for causal inference (e.g. instrumental variables or matching).
Treatments can have continuous intensity but in the standard potential outcomes framework they are assumed to be binary. The two most commonly used measures are the average treatment effect (ATE) and the average treatment effect on the treated (ATT): $$\begin{align} ATE &= E[y_1 - y_0] \newline ATT &= E[y_1 - y_0|D = 1] \end{align}$$ where the dummy $D$ denotes treatment status ($1 =$ treated, $0$ otherwise), whilst $y_1$ and $y_0$ denoted the potential outcomes in the two states. ATE is the expected treatment effect on a randomly extracted unit from the population. ATT is the expected treatment effect on a randomly extracted unit from the sub-population that has been exposed to the treatment.
