IREF - Institut de Recherches Économiques et Fiscales
Pour la liberté économique et la concurrence fiscale
George Stigler, in an influential paper on income redistribution, states an observation attributed to Aaron Director as follows : “Public expenditures are made for the primary benefit of the middle classes, and financed with taxes which are borne in considerable part by the poor and the rich.” (GEORGE STIGLER 1970, p.1) As Stigler notes, this relationship, which will henceforth be referred to as Director’s Law, was found by Director through inductive reasoning, based upon empirically observed public spending decisions. Stigler, on the other hand, sets out to lay a theoretical foundation for Director’s Law, and the core of his argument is that the middle income classes dominate political processes under democratic decision-making. Obviously, this argument of middle class domination is related to a wealth of earlier contributions, notably those concerned with the median voter approach, which point in the same direction. Also, a number of later contributions following Stigler have made more refined arguments regarding the influence of general and special interests on political outcomes, as well as regarding the impact of political institutions.
But how exactly is Director’s Law to be understood ? First of all, it is important to note that it does not focus on explicit schemes of income redistribution, such as social welfare programs, but on the distributive effects of the budget as a whole. If, for example, the tax system was very progressive and a large chunk of public funds was spent on public goods preferred by the middle classes, but less so by high income earners, then this would be a policy in line with Director’s Law. Every governmental activity has some redistributive effects, and their sum needs to be considered in evaluating whether Director’s Law holds in the real world or not. Stigler enumerates a number of policies, which he presumes to be targeted primarily at middle-income households. Among them are obvious suspects, such as tax exemptions granted for educational purposes, and the provision of schools and universities from which a middle-class offspring is significantly more likely to graduate compared to one of a lower income household. But Stigler argues that middle-class targeting also underlies less suspicious spending categories, such as police forces. The poor, in contrast to the middle classes, have not too much property to be protected, and the rich would principally be able to provide for private protection of their property themselves – in fact, this is what they usually do even with a public enforcement of property rights in place. Even a regulatory measure such as a minimum wage law can be interpreted as an attempt to protect earners of reasonable incomes from the potential competition of low-income earners.
2. Director’s Law and the dominance of the middle classes
The major underlying premise of Director’s Law is that the middle classes dominate the political process in one way or the other. There are different models in political economy within which such a conclusion can be reached. One straightforward approach is the median voter model, which assumes that voters can be ordered according to a criterion relevant for their political decision-making along a one-dimensional scale. This criterion may be their ideological position on a left-right-scale, or income, if it determines the amount of public goods demanded by the voters. If some additional assumptions hold, it can be shown in this model that the median voter’s preferences, i.e. the preferences of the voter who divides the population into two exactly equally large subgroups, will always prevail in democratic voting. With an income distribution that is not too skewed either towards high- or low-income earners, and with income being the relevant criterion according to which we sort the voters, we can then expect the median voter to be a member of the middle income classes. In other words, the preferences of the middle classes would always prevail in the political process.
Even though this result seems to be clear-cut, it can be questioned even within the framework of a median voter model, if there are publicly provided goods for which the demand rises at first with income, but declines from some relatively high threshold. An example is publicly provided university education, which may be demanded in little amounts by both the poor (who lack resources to invest into education) and the rich (who substitute public by private education). Such settings produce so-called “ends-against-the-middle” results, where the ends of the income distribution form a coalition against the middle classes. The individual with the median demand for public goods does in this case not fall naturally into the middle classes, but may well be a member of the rich or the poor.
Maybe even more importantly, there are some good reasons to be skeptical with regard to the applicability of the median voter model to questions regarding Director’s Law. Remember that the model only holds if a one-dimensional scale can be applied. If we talk about redistribution, this would for example be the case if we decided on a uniform, proportional income tax rate and if it were agreed upon in advance that the entire revenue be redistributed as identical per-capita transfers to all individuals. Obviously, real-world problems of income redistribution are not of that kind. Rather, they involve multi-dimensional tax- and transfer-schemes, with non-linear (e.g. directly progressive) income taxes and transfers that are directed only at small, clearly-defined subgroups of the population. It is therefore an interesting question if the proposed result of a dominance of the middle classes can be obtained under more general assumptions, in models allowing for multi-dimensional policies. Indeed, we discuss some models that come to this result when voting on non-linear tax schedules is introduced. However, it is important to note that the assumptions that have to be applied in order to guarantee Director’s Law-type equilibria become much more arbitrary and less general. Under more general assumptions, on the other hand, the theoretical analysis usually leads to multiple possible equilibria, and a state of the world resembling Director’s Law would only be one among many plausible alternatives.
3. Beyond pure self-interest and deterministic voting : How do individuals decide on income redistribution ?
The arguments we have considered so far have in common that they rely on purely self-interested and perfectly informed individual decision-makers in the political sphere. Moreover, they assume that voters decide purely by looking at the issues, and not due to personal sympathies or party loyalties. These assumptions, however, both appear to be rather questionable, given the peculiar incentives for individual deliberation in collective decision-making. Looking for the motives that drive redistributive politics discussed in the economics literature, one stumbles across voluntary redistribution as one driving force. Relatively rich individuals may suffer from negative externalities when they are surrounded by relatively poor individuals. These externalities may be of a moral or of a more practical kind, e.g. if crime rates are strongly correlated with poverty. Redistribution that results from such a motive – or even from pure altruism – is likely to be targeted precisely towards relatively poor income groups. It would clearly contradict Director’s Law. However, we argue that these measures make only for a relatively small fraction of overall government activity in OECD countries. And given that Director’s Law focuses on the overall budget incidence, the existence of this motive does not exclude that the middle classes nevertheless are the overall beneficiaries of aggregate government activity.
The impact of party loyalty and related issues is somewhat more interesting. The literature dealing with these issues builds on the assumption of “probabilistic voting”, which simply means that individual voters react towards changed electoral programs not with certainty, but only with some strictly positive probability. One important model of this type makes the assumption that left-wing voters tend to be relatively poor, while right-wing voters tend to be more affluent. The economic interests of the rich and the poor are thus closely entangled with their ideologies. The level of private consumption of the poor depends to a large degree on them receiving transfers, and the level of private consumption of the rich depends to a large degree on a policy that is associated with small efficiency losses. Both rich and poor are therefore, on average, very reluctant to vote for a party that takes an ideological stance opposing their private interests. The middle classes are politically much more mobile between the left and the right party, and therefore will benefit from electoral platforms being framed in their interests. A plausible equilibrium is thus one with a uniform income tax and spending programs that are largely to the benefit of the middle classes – which is in accordance with Stigler’s phrasing of Director’s Law.
But again, some cautioning remarks are in order. Probabilistic voting models that focus on the revenue side of the budget come to rather different conclusions regarding the tax system and predict a much more fragmented income tax schedule, complete with a pork barrel of targeted tax deductions to the benefit of small, politically mobile groups of voters. As so often, the result that corroborates Director’s Law relies on a set of plausible, but by no means necessary assumptions, and models with other plausible assumptions yield different predictions. This leads us to the question whether some more unambiguous theoretical results can be reached if the institutional framework of decisions on income redistribution is more explicitly included in the theoretical analysis.
4. Political institutions and decisions on income redistribution
A first distinction that can be made is that between purely representative and direct-democratic frameworks. The availability of a referendum seems to make it more difficult for representatives to deviate from median voter preferences, since any such decision may be overridden in a referendum. On first sight, this should increase the power of the middle classes in the political process. Matters are more complicated, however. As we have seen above, the median voter model may also yield “ends-against-the-middle” results. On the other hand, Director’s Law-type results may require the formation of coalitions between the middle classes and other income groups. In both cases, the impact of direct-democratic mechanisms threatens the position of the middle classes – in the first instance because the median voter is not a member of a middle income group, in the second instance because the results of political bargaining may be invalidated by a popular referendum. This “issue unbundling” effect makes it very difficult for representatives to reach actually binding coalition agreements.
Moreover, some empirical results in the literature appear to suggest that the funds available for redistributive politics are smaller under direct than under representative democracy. The increased pressure to finance state activity through user charges pushes fiscal policy into the direction of an adherence of the benefit principle, which by definition precludes redistribution. Analysis of Swiss data also suggests that direct-democratic control only reduces the financial means that are invested into the process of income redistribution, while the degree of income equalization remains unchanged. In other words, direct democracy appears to lead to a more efficient process of income redistribution, without having a significant negative impact on income equality in the secondary income distribution. But if income redistribution becomes more targeted, then this can be counted as indirect evidence against Director’s Law in systems with strong direct-democratic control.
If we look at the differences between presidential and parliamentary regimes, we see that presidential regimes are subject to additional checks and balances due to the lacking threat with a vote of confidence vis-à-vis parliament. Moreover, the cohabitation in times of divided government requires broader political compromises and makes it more difficult to target redistributive activities only to a small constituency. On the other hand, divided government in a presidential regime may also lead to fewer spending compared to a parliamentary regime, if the majority in parliament denies the executive branch its agreement to spending proposals. The empirical evidence, however, is mixed and suggests that presidential regimes can be expected to be associated with lower spending, but do not have a significant impact on its composition.
Regarding the difference between plurality rule and proportional representation, it is at first striking that under the former, a support of just over 25% of the population may be sufficient to become elected as prime minister : One’s party needs to win half the seats in parliament, so that convincing 50% of the voters in 50% of the districts would suffice. Furthermore, Duverger’s Law predicts plurality rule to converge to a two party system. Under proportional representation, on the other hand, a government needs to test on a far broader majority of 50% of the population. Proportional representation will thus be associated with broader transfer schemes. This prediction is well supported by the empirical literature. But what does this mean in detail with regard to Director’s Law ? It seems to be straightforward that the targeting of expenditures under plurality rule is to a large degree a geographical matter. Those pivotal electoral districts that can be won over with some targeted spending, e.g. for local infrastructure, receive an extra portion of public goods or transfers, while those districts that are hopelessly lost for the incumbent government receive nothing or are even fiscally exploited if such discrimination is possible. However, it is important to note that if the electoral districts are sufficiently small, there will also be differences regarding the average income levels of voters between districts, given that there is some degree of regional sorting of individuals according to their income levels. In that case, targeting expenditures geographically and targeting them along voters’ incomes are two sides of the same coin. Plurality rule does then accommodate a fiscal policy along the lines of Director’s Law, because it enables a government to rest upon a relatively narrow base of predominantly middle income districts, while paying disregard to predominantly poor or rich electoral districts. If there is pronounced regional sorting according to income, we can thus expect to observe Director’s Law in a more strictly executed fashion under plurality rule than under proportional representation.
Finally, there is also a distinction to be made between unitary and federally organized jurisdictions. A very fragmented federal system would allow individuals to sort geographically according to their incomes. Taxation would converge towards taxation according to the benefits principle, and redistribution would collapse due to out-migration of wealthier individuals whenever they are taxed for redistributive purposes. At best, some voluntary redistribution would survive, but this would hardly resemble a Director’s Law-type state of the world. One might nevertheless think of a mechanism that at least induces one characteristic of Director’s Law, namely a regressive budget incidence at the threshold from low income to medium income earners. Jurisdictions have hardly any incentive to use their fiscal policy to attract low-income earners, who are often hampered by a small stock of human capital and exposed to a high risk of becoming unemployed and thus net transfer recipients. If formal barriers to immigration are not feasible, which is usually the case within a political union, sub-central governments might find it worthwhile to deter low-income earners by setting high marginal tax rates, providing only a modest welfare state and a low supply of other public goods that are typically preferred by this group of individuals.
5. The empirical analysis
In this paper, we do not look at the different spending or taxation instruments employed in the process of redistribution, but we focus on the actual redistribution achieved by overall governmental activity. We proceed in three steps in order to test the comparative impact of different constitutional environments on income redistribution. First, a cross section of 60 plus countries at the end of the 1990s is used to find out whether there are differences in the distribution of final (disposable) income that can be attributed to the constitutional differences outlined in Section 4. However, this cross-country data set does not contain any measure of the primary (market) income distribution. It is therefore not possible to infer with certainty whether constitutional differences also affect income redistribution.
This is therefore done in a second step of the analysis. For a small sample of 13 OECD countries between 1981 and 1998, a yearly panel data set is constructed on the basis of data provided by the Luxembourg Income Study, which allows analyzing the impact of institutional factors on the primary and final income distributions as well as on fiscal redistribution. This data set has another drawback, however, as the 13 countries contain too small numbers of presidential systems or systems with plurality rule such that it cannot be analyzed how they affect income redistribution. Nevertheless, these data allow us to focus on differences in fiscal decentralization in addition to the study of fiscal redistribution.
The third step consists in an analysis of income redistribution in Switzerland. We discuss the main results from the paper by FELD, FISCHER and KIRCHGÄSSNER with respect to its implications for Director’s Law. Switzerland is a particularly interesting case as the cantons have strong fiscal competencies including spending and taxation. This leads to intensive tax competition between the cantons and at the local level. Still, there are differences between the cantons as to the intensity of tax competition they have to cope with. Moreover, cantonal differences obtain with respect to the extent of direct democratic decision-making that is employed in fiscally relevant decision-making. Finally, there is not a presidential, but a directorial system at the Swiss cantonal level, since the members of the cantonal governments are directly elected by the citizens. In addition, plurality rule does almost not play any role for the election of parliament. Only three cantons have majoritarian systems which consequently do not have any measurable impact on spending or revenue, which leads us to also neglect it in this paper.
The main results support the view that the way income is redistributed in different countries depends on the respective constitutional framework, which is associated with a particular set of incentives. The cross-country analysis reveals that countries with presidential systems have a significantly more unequal distribution of disposable incomes than parliamentary systems. This result is robust to the inclusion of additional explanatory variables and could thus be seen as a ceteris paribus result. Strictly speaking this does not tell us directly how income redistribution is going on in presidential systems. Together with the result that presidential systems have significantly lower welfare spending it could be concluded however that presidential systems obviously allow for less broad-based redistribution of income and end up with higher inequality. With respect to Director’s Law, these results imply that presidential systems put restraints on redistribution towards the middle income classes aimed at by the median voter.
In contrast to presidential systems, plurality rule does not significantly affect the distribution of disposable incomes. Majoritarian systems do thus not come up with a higher inequality than systems with proportional representation despite the fact that welfare spending is lower under plurality rule. More research is thus needed to find out how mechanisms of income redistribution are working under plurality rule. With respect to Director’s Law, it must be concluded that plurality rule is not an effective restraint on attempts to redistribute towards the middle incomes.
In this cross section analysis, we do also not observe any significant effect of federalism on the distribution of disposable incomes. This is not surprising given the way federalism is captured. The dummy variable used is not able to distinguish between federations with more or less fiscal competition at the sub-federal levels. Moreover, fiscal competition might also occur in unitary states if local governments have high fiscal autonomy. We thus have a closer look at the extent of tax autonomy in 13 OECD countries and its impact on income distributions. It turns out in the analysis that higher tax autonomy is associated with a significantly more equal primary income distribution, a significantly more unequal distribution of disposable incomes and consequently also less fiscal redistribution. The latter result also holds when the primary income distribution is controlled for. Interestingly, trade openness is associated with more equal primary and final income distributions and with significantly higher fiscal redistribution. While a competitive federalism and trade openness apparently increase the chances of individuals to raise their incomes overall, only competitive federalism in the end puts a restraint on the possibilities of governments to redistribute incomes along the lines of Director’s Law.
Finally, the analysis by FELD, FISCHER and KIRCHGÄSSNER for the Swiss cantons reveals that direct democracy – ceteris paribus – induces significantly lower overall spending and revenue, but in particular lower welfare spending and tax revenue. This is also associated with a significantly more equal primary income distribution, a more equal final income distribution, and significantly less fiscal redistribution in direct democracies. The latter result however becomes insignificant when the primary income distribution is controlled for, i.e. when the gap between the rich and the poor left by the market is widening. In direct democracies, the ability to follow an income distribution consistent with Director’s Law is thus surprisingly restricted. The analysis conducted by FFK does not reveal a robust pattern of redistribution along the income classes. Direct democracy significantly reduces the tax payments in each income quantile and thus leaves all citizens more money in their pockets. Depending on the specification, the higher, the medium or the lower income classes may gain more from this effect. No final conclusion could thus be drawn. It should also be noted that tax competition between the cantons is associated with more unequal income distributions, but more redistribution at the cantonal level. This latter result is however not robust to the inclusion of the primary income distribution. Neither tax competition nor spending decentralization has a robust impact in the analysis reported by FFK.
Summing up, our analysis suggests that the occurrence of Director’s Law heavily depends on the constitutional and political system prevailing in the different countries. Redistribution from the rich to the middle income classes and perhaps also to the poor becomes more probable in representative democracies which are either organized as unitary states, or federations allowing for low levels of sub-central fiscal autonomy, constituting their executive in the way of parliamentary systems. Given the empirical ambiguity of the effect of plurality rule, Director’s Law is thus probably not only a problem for continental European countries. It might as well occur in the UK, as in Germany, the Netherlands or Italy.
It should finally be noted that the empirical results reported in this paper are based on income distributions and redistributions which only consider monetary instruments. The spending components which are at the center of STIGLER’s analysis of Director’s Law are even not taken into account. Calculating the redistributive effects of education or cultural spending might easily enforce or exacerbate the results reported in this paper. In any case, it cannot be expected that income redistribution through public budgets is primarily aimed at increasing the welfare of the poor, as normative approaches to income redistribution would demand.