It is usually best to use the tool that is most directly suited to the task at hand. Moreover when other mechanisms are also in use. one must attend to interactions that may be overlooked when focusing on a single policy instrument. Notably, many policies interact with the income tax with regard to effects on labor supply and income distribution. For these and other reasons, we should, for example, hesitate to prescribe policy for the Earned Income Tax Credit without regard to the current state of welfare programs and possible reforms in them, and we cannot assess the taxa- tion of capital income independently of other tax instruments that raise revenue from individuals in the same part of the income distribution Comparability refers to the idea that it is easier to assess choices among different types of apples than between apples and oranges(or, orse, apples and elephants). In evaluating a given policy, it is often most helpful to focus on one dimension at a time and, furthermore, to distinguish intrinsic from incidental characteristics. When shopping for an automobile, it is not very helpful to compare a white, high-end,over- sized SUv to a red, economy, subcompact car For buyers on a tight bud get to pick the suv because they hate the color red or because it comes with a free microwave oven that happens to meet a current need would be foolish. Instead, it is more illuminating to undertake a series of com- parisons of vehicles that are similar on all but one dimension, for ex ample, to investigate subcompacts in other colors if one hates red and to visit a home appliance store if a new microwave oven is desired. This commonsense notion of comparability has received too lit ttle attention in the theory of taxation. Comparability is in fact an extremely powerful idea, as much of this book will demonstrate. In particular, it turns out that in a surprisingly wide range of instances, comparability is sharpest when one completes the system-say, a luxury tax plan or a proposed expenditure on parks-by using a particular technique: an ad justment to the income tax(and transfer) system that achieves an overa result that is distribution neutral. Indeed the availability of the income ax as an instrument casts a different light upon the analysis of such pol- icies as commodity taxes, dividend and capital gains taxes, estate and gift taxes, social insurance, public goods provision, and economic regulation. It turns out that it is not generally sensible to use various indirect forms of taxation, expenditure policies, or regulations to redistribute income if an income tax is available, as it generally is in developed economies 03_Kaplow_Ch02_p011-p034 ndd 14 1016/20071121:27AM
14 chapter 2 It is usually best to use the tool that is most directly suited to the task at hand. Moreover, when other mechanisms are also in use, one must attend to interactions that may be overlooked when focusing on a single policy instrument. Notably, many policies interact with the income tax with regard to effects on labor supply and income distribution. For these and other reasons, we should, for example, hesitate to prescribe policy for the Earned Income Tax Credit without regard to the current state of welfare programs and possible reforms in them, and we cannot assess the taxation of capital income independently of other tax instruments that raise revenue from individuals in the same part of the income distribution. Comparability refers to the idea that it is easier to assess choices among different types of apples than between apples and oranges (or, worse, apples and elephants). In evaluating a given policy, it is often most helpful to focus on one dimension at a time and, furthermore, to distinguish intrinsic from incidental characteristics. When shopping for an automobile, it is not very helpful to compare a white, high-end, oversized SUV to a red, economy, subcompact car. For buyers on a tight budget to pick the SUV because they hate the color red or because it comes with a free microwave oven that happens to meet a current need would be foolish. Instead, it is more illuminating to undertake a series of comparisons of vehicles that are similar on all but one dimension, for example, to investigate subcompacts in other colors if one hates red and to visit a home appliance store if a new microwave oven is desired. This commonsense notion of comparability has received too little attention in the theory of taxation. Comparability is in fact an extremely powerful idea, as much of this book will demonstrate. In particular, it turns out that in a surprisingly wide range of instances, comparability is sharpest when one completes the system—say, a luxury tax plan or a proposed expenditure on parks—by using a particular technique: an adjustment to the income tax (and transfer) system that achieves an overall result that is distribution neutral. Indeed, the availability of the income tax as an instrument casts a different light upon the analysis of such policies as commodity taxes, dividend and capital gains taxes, estate and gift taxes, social insurance, public goods provision, and economic regulation. It turns out that it is not generally sensible to use various indirect forms of taxation, expenditure policies, or regulations to redistribute income if an income tax is available, as it generally is in developed economies. 03_Kaplow_Ch02_p011-p034.indd 14 3_Kaplow_Ch02_p011-p034.indd 14 10/16/2007 11:21:27 AM 0/16/2007 11:21:27 AM
AN INTEGRATED VIEW 15 Among the many policy instruments generally thought of in distribu tive or revenue-raising terms, the income tax has a special place, and because of this role the others need to be analyzed differently from how they often are Together, completeness, comprehensiveness, and comparability are essential aspects of an integrated view of taxation, government ex penditures, and redistribution. One can only understand each poli instrument--each piece of the puzzle-if the others are also on the table and the relationships among them are understood. Most analysis is far specialized, focusing on one particular policy, indeed, ofte en a sin gle aspect of a particular policy, and there are good reasons for this divi sion of labor. However, research is best guided and its results are most effectively employed by policy-makers if the broader, integrated frame work of which they are a part is well understood and kept clearly in view A Completeness of policy Specification The idea developed in this section, as noted previously, is that one can- not properly analyze the redistributive character of a policy unless that policy is fully articulated in certain respects. Begin by considering a gas. oline tax increase. Supposing that this tax is moderately regressive, the increase may well be opposed on distributive grounds. Next, consider improvements to public parks. Supposing that these improvements are 2 As will become apparent in chapters 4-6, it is not so much an income tax per se as a rsonalized tax system under which the taxes paid by individuals can be varied according to their level of well-being. In most developed economies, it is the income tax(combined with transfer programs)that plays this role; as subsection 9. B. 1 indicates, this role could also be filled by a personal consumption tax. Such an objection presupposes that there is too little redistribution: I bution is generally costly, due to labor supply distortion, there is an optimal degree of redis- tribution(see chapter 4). If that degree of redistribution were exceeded, then a regress hange would be desirable, not objectionable, on redistributive grounds. Where convenien however, the discussion will follow convention in referring to greater redistribution as desir able, which should be understood as crediting the benefits of further redistribution while abstracting from the costs. 03_Kaplow_Ch02_p011-p034 ndd 15 1016/20071121:27AM
an integrated view 15 Among the many policy instruments generally thought of in distributive or revenue-raising terms, the income tax has a special place, and because of this role the others need to be analyzed differently from how they often are.2 Together, completeness, comprehensiveness, and comparability are essential aspects of an integrated view of taxation, government expenditures, and redistribution. One can only understand each policy instrument—each piece of the puzzle—if the others are also on the table and the relationships among them are understood. Most analysis is far more specialized, focusing on one particular policy, indeed, often a single aspect of a particular policy, and there are good reasons for this division of labor. However, research is best guided and its results are most effectively employed by policy-makers if the broader, integrated framework of which they are a part is well understood and kept clearly in view. A. Completeness of Policy Specifi cation The idea developed in this section, as noted previously, is that one cannot properly analyze the redistributive character of a policy unless that policy is fully articulated in certain respects. Begin by considering a gasoline tax increase. Supposing that this tax is moderately regressive, the increase may well be opposed on distributive grounds.3 Next, consider improvements to public parks. Supposing that these improvements are 2 As will become apparent in chapters 4–6, it is not so much an income tax per se as a personalized tax system under which the taxes paid by individuals can be varied according to their level of well-being. In most developed economies, it is the income tax (combined with transfer programs) that plays this role; as subsection 9.B.1 indicates, this role could also be fi lled by a personal consumption tax. 3 Such an objection presupposes that there is too little redistribution: Because redistribution is generally costly, due to labor supply distortion, there is an optimal degree of redistribution (see chapter 4). If that degree of redistribution were exceeded, then a regressive change would be desirable, not objectionable, on redistributive grounds. Where convenient, however, the discussion will follow convention in referring to greater redistribution as desirable, which should be understood as crediting the benefi ts of further redistribution while abstracting from the costs. 03_Kaplow_Ch02_p011-p034.indd 15 3_Kaplow_Ch02_p011-p034.indd 15 10/16/2007 11:21:27 AM 0/16/2007 11:21:27 AM
16 CHAPTER 2 Income net effect Figure 2.1 Park Improvements Financed by Gasoline Tax Increase orth somewhat more(measured in dollars)to high-income individuals this plan also might be opposed on distributive grounds. If, however, these two proposals are combined into one, the distributive effect may well be favorable to the poor, as depicted in figure 2. 1. Although lower income individuals pay a somewhat higher proportion of their income in additional gasoline taxes, their payment is still a smaller absolute amount. Perhaps individuals earning $20,000 pay an additional $20 in gasoline taxes(0.1%)and those earning $100,000 pay an additional $60 in gasoline taxes(0.06%). Regarding the expenditures on parks, suppose that the dollar-equivalent benefit to those earning $20,000 is $30 and the benefit to those earning $100,000 is $50. Netting these amounts, the lower-income individuals gain by $10 per capita and the upper-income individuals lose by $10 per capita The possibility that two seemingly regressive policies, when combined, increase the extent of redistribution is readily explained by a fairly com mon although inconsistent use of baselines: proportionality for the tax In chapter 8, more will be said about the relationship between benefits measured in utility and in dollars. 03_Kaplow_Ch02_po11-p034 ndd 1016/20071121:27AM
16 chapter 2 worth somewhat more (measured in dollars) to high-income individuals, this plan also might be opposed on distributive grounds.4 If, however, these two proposals are combined into one, the distributive effect may well be favorable to the poor, as depicted in fi gure 2.1. Although lowerincome individuals pay a somewhat higher proportion of their income in additional gasoline taxes, their payment is still a smaller absolute amount. Perhaps individuals earning $20,000 pay an additional $20 in gasoline taxes (0.1%) and those earning $100,000 pay an additional $60 in gasoline taxes (0.06%). Regarding the expenditures on parks, suppose that the dollar-equivalent benefi t to those earning $20,000 is $30 and the benefi t to those earning $100,000 is $50. Netting these amounts, the lower-income individuals gain by $10 per capita and the upper-income individuals lose by $10 per capita. The possibility that two seemingly regressive policies, when combined, increase the extent of redistribution is readily explained by a fairly common although inconsistent use of baselines: proportionality for the tax Figure 2.1 Park Improvements Financed by Gasoline Tax Increase 4 In chapter 8, more will be said about the relationship between benefi ts measured in utility and in dollars. 03_Kaplow_Ch02_p011-p034.indd 16 3_Kaplow_Ch02_p011-p034.indd 16 10/16/2007 11:21:27 AM 0/16/2007 11:21:27 AM
AN INTEGRATED VIEW 17 and equal dollar benefit for the expenditure. For taxes, there is a long tradition of assessment by reference to a standard of proportionality. For government expenditures on goods and services, few would describe a project that benefits millionaires by $1000 per capita and the poor by $50 per capita as"progressive, " even when the benefit to the poor is a far higher percentage of their income than is the benefit to the rich The main lesson of the illustration is that viewing a particular change in a tax or an expenditure as increasing or decreasing redistribution is problematic. If a tax increase is involved, one must ascertain how the revenue will be spent; if taxes are cut, how the revenue shortfall will be addressed. For any given tax change, one can imagine combining a wide range of expenditure policies or other adjustments to the system. Thus, the hypothesized gasoline tax increase might fund park improvements, welfare programs, police protection, income tax reductions, or many other actions. Likewise for expenditures The problem with incomplete depictions of policies is well illus- trated by the debate over flat tax proposals Consider the question, How progressive, or redistributive, is a flat tax? The conventional answer is not at all"because, as just noted, the standard benchmark is propor tionality and a flat tax is perfectly proportional. But this response is seri- usly deficient. Take a pure flat tax: a linear income tax(a form of neg ative income tax")wherein the uniform marginal tax rate is t and the proceeds are used to finance a grant of g to each individual.(To simplify the present discussion, ignore both incentive effects and the need to . This observation has also been made by Steuerle(2003) There exists a substantial, often technical literature measuring the progressivity and redistributiveness of taxes by reference to proportionality, most of which does not attend to the problems discussed in the text. See, for example, the surveys and analysis in Lambert (1999, 2001). Such measures are examined in section 15.A. Progressive, proportional, and regressive taxes are ordinarily defined as ones v average rates rise, are constant, or fall with income. Occasionally, these terms are associated with marginal rates, but that usage will not be followed here. A motivation for focusing or average rates is that"progressive "taxes are often associated with redistributive taxes. In any event, the concept of progressivity is of limited importance because policies should be as- sessed by using the pertinent social welfare function, not by measuring how they fare under some index of progressivity(see section 15.A This tax is formally analyzed in subsection 4.B.1 03_Kaplow_Ch02_po11-p034 ndd 1016/20071121:28AN
an integrated view 17 and equal dollar benefi t for the expenditure.5 For taxes, there is a long tradition of assessment by reference to a standard of proportionality.6 For government expenditures on goods and services, few would describe a project that benefi ts millionaires by $1000 per capita and the poor by $50 per capita as “progressive,” even when the benefi t to the poor is a far higher percentage of their income than is the benefi t to the rich. The main lesson of the illustration is that viewing a particular change in a tax or an expenditure as increasing or decreasing redistribution is problematic. If a tax increase is involved, one must ascertain how the revenue will be spent; if taxes are cut, how the revenue shortfall will be addressed. For any given tax change, one can imagine combining a wide range of expenditure policies or other adjustments to the system. Thus, the hypothesized gasoline tax increase might fund park improvements, welfare programs, police protection, income tax reductions, or many other actions. Likewise for expenditures. The problem with incomplete depictions of policies is well illustrated by the debate over fl at tax proposals. Consider the question, How progressive, or redistributive, is a fl at tax?7 The conventional answer is “not at all” because, as just noted, the standard benchmark is proportionality and a fl at tax is perfectly proportional. But this response is seriously defi cient. Take a pure fl at tax: a linear income tax (a form of “negative income tax”) wherein the uniform marginal tax rate is t and the proceeds are used to fi nance a grant of g to each individual.8 (To simplify the present discussion, ignore both incentive effects and the need to 5 This observation has also been made by Steuerle (2003). 6 There exists a substantial, often technical literature measuring the progressivity and redistributiveness of taxes by reference to proportionality, most of which does not attend to the problems discussed in the text. See, for example, the surveys and analysis in Lambert (1999, 2001). Such measures are examined in section 15.A. 7 Progressive, proportional, and regressive taxes are ordinarily defi ned as ones whose average rates rise, are constant, or fall with income. Occasionally, these terms are associated with marginal rates, but that usage will not be followed here. A motivation for focusing on average rates is that “progressive” taxes are often associated with redistributive taxes. In any event, the concept of progressivity is of limited importance because policies should be assessed by using the pertinent social welfare function, not by measuring how they fare under some index of progressivity (see section 15.A). 8 This tax is formally analyzed in subsection 4.B.1. 03_Kaplow_Ch02_p011-p034.indd 17 3_Kaplow_Ch02_p011-p034.indd 17 10/16/2007 11:21:28 AM 0/16/2007 11:21:28 AM
18 CHAPTER 2 finance government expenditures on goods and services. If t is set at 0%, in which case g must be $O, no redistribution takes place, consistent ith the standard view of a proportional tax. If t is set at 100%, in which case g would equal mean income, everyone's after-tax(and transfer )in come is fully equalized. In other words, in this hypothetical world with no incentive effects, both an entirely nonredistributive tax and the most redistributive possible tax are flat taxes. For levels of t between 0%and 100%, intermediate degrees of redistribution are possible This example, by illustrating the importance of attending to the ex nditure of tax revenue, illustrates the more general notion that an ted view can be highly misleading. Actually, the problem with incompleteness is worse. In the example, it was specified that all revenue financed a particular sort of expenditure, namely, a uni- form grant to each individual. But for any given flat tax rate t, one could instead assume that the revenue finances public goods with benefits proportional to income, in which case there would be no redistribution The specific lesson that the flat tax cannot be characterized with regard to the degree of redistribution involved independently of the level of the tax remains true, although in less xtreme form, even if comparisons are limited to graduated and flat rate taxes, with revenue held constant and transfers through the tax system ruled out (or held constant). The reason is that adjusting the exemption level under a flat tax-essentially using the revenue raised by a higher tax rate to finance a more generous exemption-allows the degree of redistribution to vary substantially. Thus, for a given graduated income tax, an equal-revenue flat tax with the same level of exemption will be unambiguously less redistributive (Those with income elow the exemption pay nothing under both schemes, those with low incomes above the exemption have a higher average tax rate under the flat tax, and those with very high incomes have a lower average tax rate under the flat tax. ) An equal-revenue flat tax with a sufficiently high exemption(and correspondingly higher tax rate, in particular one at least equal to the top graduated tax rate), however, would be unambiguously more redistributive (Those with tax; those with somewhat higher incomes would pay something under the flat tax but still less than under the graduated tax, and those with the highest incomes would pay more under the flat tax ) For intermediate levels of the exemption, middle-income taxpayers would pay more under the flat tax, while the rich and the near-poor(those with income high enough to pay ome tax) would pay less. Whether such a state of affairs under the flat tax is viewed as more or less redistributive depends on whether greater importance is associated with the poor pay- ng less or with the rich getting off easier. For more technical characterizations, see Davies and Hoy(2002) 03_Kaplow_Ch02_po11-p034 ndd 1016/20071121:28AN
18 chapter 2 fi nance government expenditures on goods and services.) If t is set at 0%, in which case g must be $0, no redistribution takes place, consistent with the standard view of a proportional tax. If t is set at 100%, in which case g would equal mean income, everyone’s after-tax (and transfer) income is fully equalized. In other words, in this hypothetical world with no incentive effects, both an entirely nonredistributive tax and the most redistributive possible tax are fl at taxes. For levels of t between 0% and 100%, intermediate degrees of redistribution are possible. This example, by illustrating the importance of attending to the expenditure of tax revenue, illustrates the more general notion that an incomplete, unintegrated view can be highly misleading.9 Actually, the problem with incompleteness is worse. In the example, it was specifi ed that all revenue fi nanced a particular sort of expenditure, namely, a uniform grant to each individual. But for any given fl at tax rate t, one could instead assume that the revenue fi nances public goods with benefi ts proportional to income, in which case there would be no redistribution 9 The specifi c lesson that the fl at tax cannot be characterized with regard to the degree of redistribution involved independently of the level of the tax remains true, although in less extreme form, even if comparisons are limited to graduated and fl at rate taxes, with revenue held constant and transfers through the tax system ruled out (or held constant). The reason is that adjusting the exemption level under a fl at tax—essentially using the revenue raised by a higher tax rate to fi nance a more generous exemption—allows the degree of redistribution to vary substantially. Thus, for a given graduated income tax, an equal-revenue fl at tax with the same level of exemption will be unambiguously less redistributive. (Those with income below the exemption pay nothing under both schemes, those with low incomes above the exemption have a higher average tax rate under the fl at tax, and those with very high incomes have a lower average tax rate under the fl at tax.) An equal-revenue fl at tax with a suffi ciently high exemption (and correspondingly higher tax rate, in particular one at least equal to the top graduated tax rate), however, would be unambiguously more redistributive. (Those with incomes moderately above the graduated tax’s exemption would pay nothing under the fl at tax; those with somewhat higher incomes would pay something under the fl at tax but still less than under the graduated tax, and those with the highest incomes would pay more under the fl at tax.) For intermediate levels of the exemption, middle-income taxpayers would pay more under the fl at tax, while the rich and the near-poor (those with income high enough to pay some tax) would pay less. Whether such a state of affairs under the fl at tax is viewed as more or less redistributive depends on whether greater importance is associated with the poor paying less or with the rich getting off easier. For more technical characterizations, see Davies and Hoy (2002). 03_Kaplow_Ch02_p011-p034.indd 18 3_Kaplow_Ch02_p011-p034.indd 18 10/16/2007 11:21:28 AM 0/16/2007 11:21:28 AM