The choice of the personal income tax base

Following Mirrlees (1971) and Vickrey (1945), the optimal tax literature has studied the design of a personal income tax where the ideal would be to tax unobserved earnings ability but instead the only available information is actual earnings. In practice, though, taxes depend on a broader range of information about each individual than earnings. In principle, this supplementary information can help in designing a tax that not only has lower efficiency costs but also has more attractive distributional properties, by more closely approximating an ability tax. The objective of this paper is to lay out theoretically and estimate empirically how to best make use of available information about each individual in addition to earnings, solving for the optimal tax base. In contrast to most of the literature on this topic, we show how equity considerations may be quantitatively incorporated in the analysis. In accordance with current practice, we find that the optimal tax base should include capital income, at least to some degree. In contrast to current practice, though, property tax payments and mortgage interest payments should not be deductible, since these deductions are costly on equity and presumably on efficiency grounds as well. The choice of a personal income tax consists first of the choice of a tax base and then the choice of a tax rate schedule. The past literature on the optimal design of the income tax, as exemplified by Vickrey (1945) and Mirrlees (1971),1 presumes that the ideal tax base is the earnings ability (wage rate) of each individual, since this is the only characteristic that is assumed to differ across people. In practice, however, earnings ability cannot be monitored for tax purposes. A close observable proxy for earnings ability is labor income (wage rates times hours of work), so that the initial optimal tax literature presumed that labor income is the natural choice of a tax base and then derived the optimal rate schedule given this tax base. Are there any welfare gains, though, from including in the tax base not just labor income but also other observable information about individuals? Actual tax bases certainly include information beyond labor income, such as interest, dividend, and capital gains income. In addition, by taxing couples as a unit rather than taxing each spouse separately, the labor income of one’s spouse affects one’s own tax rate. Mortgage and property tax payments are allowed as deductions for those who itemize. The tax base is certainly more complicated than labor income. To what degree can these additional elements in the tax base be explained based on an optimal income tax framework, without adding subsidiary objectives or externalities? Atkinson and Stiglitz (1976) derive conditions under which the ideal tax base does not make use of information about consumption of other goods, so includes just labor income.2 One key condition is that consumption goods not differ in the degree to which they are complements to leisure. A second condition is that the amount an individual consumes is not correlated with the individual’s marginal utility of income. It is this second condition that we focus on. In doing, so we explicitly allow for heterogeneity in behavior beyond skills. Are there consumption choices, that reveal information about ability beyond what is revealed by reported earnings? Besley and Coate (1992) argue that providing low-quality in-kind rather than cash transfers help reveal who among low earners has low earnings ability, on the presumption that only those with low earnings ability are in fact willing to consume low quality goods. Similarly, Blomquist and Christiansen (2005) argue that users of excludable public goods should be charged a price different from marginal cost to the degree that demand depends on earnings ability. Kopczuk (2001) argues that tax avoidance should be facilitated if the low skilled can avoid taxes more easily than the high skilled, conditional on labor income. Closer to the choice of income tax base, Gordon (2004) argues that income from savings (dividends or interest income) should be part of the tax base to the degree to which those with high ability save more (or in different forms) than those with low ability, among those with the same labor income. The first objective of this paper is to explore empirically using PSID data to what degree the observable variables that are commonly part of the income tax base help forecast earnings ability (wage rates) among those with any given level of labor income. Are there observables that help detect high vs. low wage rates among individuals with the same labor income? The variables we focus on are interest and dividend income, the labor income of one’s spouse, and expenditures on For recent work in this tradition, see Saez (2001) and Gruber and Saez (2002) For more recent attempts to explicate this result, see Saez (2002), Laroque (2005), and Kaplow (2006).