Working Paper: NBER ID: w24903
Authors: Olivier Deschenes; Kyle C. Meng
Abstract: This paper examines the application of quasi-experimental methods in environmental economics. We begin with two observations: i) standard quasi-experimental methods, first applied in other microeconomic fields, typically assume unit-level treatments that do not spill over across units; (ii) because public goods, such as environmental attributes, exhibit externalities, treatment of one unit often affects other units. To explore the implications of applying standard quasi-experimental methods to public good problems, we extend the potential outcomes framework to explicitly distinguish between unit-level source and the resulting group-level exposure of a public good. This new framework serves as a foundation for reviewing and interpreting key papers from the recent empirical literature. We formally demonstrate that two common quasi-experimental estimators of the marginal social benefit of a public good can be biased due to externality spillovers, even when the source of the public good itself is quasi-randomly assigned. We propose an unbiased estimator for the valuation of local public goods and discuss how it can be implemented in future studies. Finally, we consider how to preserve the advantages of the quasi-experimental approach when valuing global public goods, such as climate change mitigation, for which no control units are available.
Keywords: quasiexperimental methods; environmental economics; public goods; externalities
JEL Codes: C21; H23; H41; Q50; Q51; Q52; Q53; Q54
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Standard quasiexperimental methods (C90) | biased estimates of marginal social benefit of public goods (H49) |
| ASEE (Q49) | biased estimates when externalities cause spillovers (D62) |
| AEEE (Q49) | biased due to selection effects (D91) |
| Weighted version of ASEE (C29) | unbiased estimates for local public goods (H49) |