Missing the Forest (Owners) for the Trees? The Potential of Matching for Estimating Program Additionality

• In Event: Lightning Session: Methods and Tools of Analysis

Abstract

Forests offer opportunities to sequester and store considerable carbon at low cost, while also providing diverse ecosystem services. Realizing the potential of forest incentive programs requires credible estimates of their effects. A central concern of the programs is additionality: the extent to which observed gains would not have occurred under the no-program counterfactual. For public forest programs, non-additional enrollment wastes scarce resources by directing payments to landowners who would have provided the environmental service even in the absence of the program (Alix-Garcia et al., 2012). For private programs, the stakes can be even higher when carbon gains are sold as credits to firms that use them to offset their own emissions: inaccurate estimates may result in too many credits issued and, in turn, increase net emissions (Calel et al., 2025).

To estimate counterfactuals and quantify environmental gains, economists have often used ex-post matching methods. Academics and carbon program developers have taken note, and methodologies that mirror ex-post matching approaches are becoming the private industry norm. However, there is little evidence on their effectiveness (Haya et al., 2032) and limited theoretical understanding of how unobservables and program eligibility shape additionality.

Our study addresses whether matching can provide reliable estimates of gains from forest incentive programs. We show in theoretical simulations that the reliability of matching in forest program evaluation hinges in large part on the role of owner unobservables—owner characteristics that affect forest outcomes and that carbon accountants can at best observe imperfectly. Because even coarse owner information is costly to acquire, new matching methodologies for forest programs assume what we call selection-on-trees: owners with similar forests today will manage them similarly tomorrow.

We empirically test some of the predictions from our simulations using data from the American Forest Foundation's Family Forest Carbon Program in five Pennsylvania counties with the highest initial enrollment. Combining Program records, tax data, and satellite imagery, we estimate treatment effects with and without matching. To assess how unobservable owner characteristics affect estimates, we compare results when matching enrolled parcels to those owned by profit-focused entities (timber companies and investors) or, separately, with owners who expressed interest in the Program but declined enrollment.

Our empirical findings align with a key theoretical prediction. The unmatched baseline shows differences in pre-program harvesting and a null effect of the Program, consistent with our eligibility rule simulations. In contrast, the matched baseline shows no harvesting differences before the Program and a 25 percent reduction in timber harvesting over its first five years, a more encouraging finding than prominent studies of other forest carbon programs. However, matching enrolled parcels with those of profit-focused owners doubles the estimated Program effect, with differences emerging most strongly during post-pandemic timber price increases. Because the matching-on-trees approach relies on nontrivial assumptions about owners, it may not provide the accuracy sought by high-integrity carbon credit programs, especially in populations with diverse owner types.

Author

• Jeremy G Weber, University of Pittsburgh
  Presenting Author