Unlike the experimental or quasi-experimental methods, the structural approach is based on modeling. In its simplest form, it may simply consists in modeling participation in a treatment to separately identify the selection into the treatment and the causal effect of the treatment on the treated population. This is the Roy model. But the structural approach can also rely on a more complex modeling of the economic environment and the way it impacts and coordinates the decisions of economic agents (workers, unemployed, future pensioners, etc.).
The goal of structural econometrics is then to estimate the key parameters of the proposed model using empirical data. Once estimated, the model can then be used to evaluate the effects of the public policies integrated into the model, as well as the effects of alternative policies whose effects can simply be deduced from the model. This is what is done for example in some micro-simulation exercises.
The fact that it is based on theoretical foundations may allow the structural approach to assess the effects of public policies credibly even when it is impossible to use controlled or natural experiments. But this advantage can also be a disadvantage: the results obtained depend intrinsically on the modeling choices that were made and on the assumptions regarding the behavior and interactions of economic agents. This is not the case with non-theoretical (or “reduced form”) approaches such as controlled or natural experiments