1. Introduction The examination of the taxpayers' behavioral response to changes in marginal tax rates is essential in estimating the impact of different tax policies so as to minimize the individual's bias and avoid erroneous policy recommendations. The effectiveness of government intervention is affected by changing behavior, as taxpayers' reporting decisions are subject to the prevailing tax schedule. The lack of government revenues can also be partially explained by the potential responsiveness of taxpayers. Given that Greece is considered a country with a high rate of tax avoidance and evasion, the estimation of the reported income elasticity could prove useful, especially for policy makers and taxpayer advocates, for the evaluation of alternative tax policies and the prediction of tax revenue effects. Initially, labor supply elasticities were used to design appropriate tax and fiscal policies, though these are likely to underestimate taxpayers' response to tax rate changes, measuring only how taxpayers alter their work schedule. Recent studies have used elasticities of (taxable) income, accounting explicitly for tax avoidance and implicitly for exclusions and deductions (e.g. Lindsey, 1987; Feldstein, 1995; Sammartino and Weiner, 1997; Auten and Carroll, 1999; Gruber and Saez, 2002). The obtained results vary though considerably, depending on the method of estimation used, the particular tax reform examined and the country under consideration. Two reasons may explain the conflicting results. First, it is often problematic to compare reported income before and after a tax reform, as changes in the definition of taxable income are introduced apart from tax rate changes. Second, most studies attribute the widening of income inequality to tax reforms, though evidence has shown that other factors may have increased inequality. Nevertheless, results show that income heterogeneity should be considered when estimating the taxpayers' reporting decision, as the responsiveness of taxable income to taxes may be higher in higher income classes, for which a larger share of income is likely to come in forms that are easier to hide from tax authorities (Thalassinos and Liapis, 2013). A suitable approach for this line of empirical analysis was recently employed by Alm and Wallace (2007 and 2010); namely quantile regression. Quantile regression was developed by Koenker and Bassett (1978) as a robust alternative estimation technique compared to conditional mean regression against outliers, and a useful approach in cases of heteroskedasticity. The magnitude of differential responses across income classes can be further examined, since regression quantiles allow analyzing the responsiveness of a wide range of reporting behavior to marginal tax rates and the responses of individuals at different points of the income distribution; a task that is not investigated thoroughly. Both empirical studies estimate though taxpayers' reporting decision using arbitrarily 'typical' quantiles such as 0.2, 0.4, 0.6, 0.8, which are very unlikely to always correspond to income classes that are taxed differently so that the reported estimations may lead to a possible bias of the real magnitude of the differential responses across income classes. In addition, using quintiles, that is a 'truncation on the dependent variable' that segments the sample into subsets based on its unconditional distribution, and doing least squares fitting on these subsets yields to inconsistent estimates (Koenker and Hallock, 2001). Such strategies are condemned to failure for all the reasons so carefully laid out in Heckman's (1979) work on sample selection, implying that the reported OLS quintiles estimations should not be directly compared to the respective quantile regression estimations. In this framework, this paper contributes to the examination of the responsiveness of a wide range of reporting behavior to marginal tax rates and the responses of individuals at the different points of the income distribution that correspond to specific tax brackets. …
Read full abstract