A Robust Estimation of the Effects of Taxation on Charitable Contributions

While many studies find that the tax-price elasticity of giving exceeds unity, several recent studies find the contrary. This is important because it can be shown that if the elasticity exceeds one, then allowing taxpayers to deduct charitable giving from their taxable income is efficient in the sense that the amount donated exceeds the loss to the treasury. Here we use Consumer Expenditure Survey data to estimate the price elasticity of all deductible contributions. Because specification tests reject the consistency of estimators such as Tobit or the two-stage Heckman we use the semiparametric method of Ahn and Powell (1993). Rather than selecting bandwidths through cross-validation we demonstrate that because high and low bandwidths lead to the standard linear model one may use visual inspection for bandwidth selection. We also do not use the covariance matrix estimator of Ahn and Powell, instead bootstrapping a confidence interval. These bootstraps are also used to remove the finite sample bias inherent in nonlinear estimators. In our results we find an elasticity estimate greater than unity for the Tobit and Heckman methods but less than one for the Ahn and Powell method. Because specification tests suggest that the likelihood assumptions ensuring the consistency of the Tobit and Heckman do not hold, our results suggest that previous high tax-price elasticities may be caused by misspecification. However, our estimate of the elasticity of contributions to just social welfare organizations exceeds unity. In this sense the deduction for those types of contributions is efficient.