Estimated Impacts of PSA on Change in Forest Cover (1992–2005)
| Outcome | Aerial Photos (Full Sample)a | Aerial Photos (Reduced Sample) |
|---|---|---|
| Simple DIDb | 12.72*** (3.96) | 13.96*** (4.92) |
| N treated | 50 | 50 |
| N controls | 152 | 86 |
| Matching estimates | ||
| Matching DIDc | 12.09** (4.89) | 11.24*** (3.25) |
| Postmatching OLS regression,d core covariate set | 9.70*** (2.96) | 11.19*** (2.50) |
| Postmatching OLS regression,d extended covariate set | 8.48*** (2.80) | 9.60*** (3.04) |
| Matching DID with calipersc,e | 4.65*** (1.41) | 7.03*** (2.40) |
| N treated dropped by calipers | 13 | 12 |
| Γf | 2.1 | 3.1 |
| N treated | 50 | 46 |
| N available controls | 152 | 70 |
Note: DID, difference in difference; OLS, ordinary least squares; PSA, Costa Rica’s Programa de Pagos por Servicios Ambientales.
↵a Full sample includes units for which outcome or covariate values may be imputed. Reduced sample restricts sample to units for which all values are observed.
↵b A two-sided t-test of the difference in means between treated and control segments. Standard errors in parentheses.
↵c Robust standard errors in parentheses (Abadie and Imbens 2006b).
↵d Huber-White robust standard errors in parentheses. Estimates postmatching with calipers are similar and not reported here.
↵e Calipers restrict matches to units within one standard deviation of each covariate.
↵f Γ is log odds of differential assignment to treatment due to unobservables. Value reported is Γ at which the critical p-value for the estimate implies the effect is insignificantly different from zero at p = 0.10.
↵**, *** Significance at 5% and 1%, respectively.