Category Archives: Redistribution I



V. Larcinese

Political Studies Vol. 55, (2007) pp. 568-85

 A Summary

In a Nutshell

Formal voting models have neglected to take account of the role that voter turnout plays in determining election results. This is because factors that have been shown to be determinants of abstention (education, income, sex), are also correlated with preferences over key policies such as redistribution. Thus although the extension of the voting franchise has changed the landscape of electoral competition as Meltzer Richard make clear, with regard the political demand for welfare it should be recognized that this has probably happened to a much lesser extent than most public choice models would indicate.

 Downsian Analysis of Redistribution

  • If policy space is unidimensional and preferences are single peaked then the median voter is pivotal in majority decision making. Thus parties in electoral competition converge on the policy platform preferred by the median voter.
  • With regard to tax, this means that a tax rate chosen by majority rule will be that preferred by the citizen whose income is the median in the polity. If t is the tax rate, and N is the number of citizens and µ is the mean income, then the total revenue will be tNµ and the per capita benefit will be tµ. If the t  ∈ {0,1} then any citizen above the mean income will want zero tax rate. Any citizen below will want a positive tax rate, and the preferred rate will decrease as their income increases. Anyone below the mean income will receive more than they pay; anyone above will pay more than they receive.
  • As income is always skewed to the right the median lies below the mean i.e. democratic decision making will yield a positive tax rate.
  • If the tax rate cannot be reacted to (people are voters only, and not economic agents) then the preferred tax rate is 100% for those below the mean. However, the rich in fact can react to the tax rate, and they do. Taxes reduce the incentive to produce and the optimal labour supply is an inverse function of the tax rate. Voters understand this tradeoff between size of the pie and the share they receive via redistribution. The ideal tax rate will now take this into account. The more elastic output with regard taxation, the lower the rate. Indeed, as it is possible to move assets abroad at low cost, high taxes can easily erode the taxable capital base.
  • In the Meltzer Richard framework equilibrium redistribution depends on the distance between the median voter’s income and the mean income. The voter is aware of the efficiency cost and updates his preferences accordingly. However, this implies that for a given efficiency cost of taxation, higher inequality will increase the desire for redistribution of the median voter, and hence the tax rate.
  • This is a static model. Dynamic models have been proposed by Alesina and Rodrik where the poorer the median voter’s income relative to the mean, the higher the capital tax rate, and the lower the growth rate.

 Empirical Evidence

  • The empirical evidence is less than strong.
  • It appears that high inequality actually reduces government spending.
  • The A and R model may show a link between inequality and growth, but they cannot prove the intermediate step that inequality reduces growth through redistribution.
  • The social affinity theory provides an alternative link between inequality and redistribution (Lindert). This states there will more redistribution the closer middle-income voters are to the poor and the further they are away from the rich. If those with incomes below the median get poorer the mean income reduces and hence there is less call for redistribution although the position of the median remains unaffected. This leads to a situation of social exclusion in which the poor become poorer but this does not incentivize the middle classes to support more redistribution.


  • Sources of power go beyond formal voting rights. In real life we see varying degrees of turnout. If propensity to abstain were random there would be no problem, and Downsian politics would function in the same way. But non-voters are not randomly distributed – they differ systematically in socio-economic status and therefore in their needs and preferences. E.g. Education, income, age and occupation.
  • Low voter turnout is often equated with socioeconomically biased turnout.


  • The above curve is the income distribution, they lower curve is the turnout as a function of income distribution. M* is the median participating voter in terms of income, but he represents a point to the right of the full electorate median in terms of income (M). M* is the pivotal voter. As the tax preferred is decreasing in income we know that M* prefers a lower tax rate to M.
  • When turnout is considered inequality is no longer a good predictor for redistribution. When inequality increases, the desire of the median voter to tax increases, but a concurrent effect is that it makes him less likely to vote. The overall effect is undetermined but we should not expect a rise in inequality to equate a rise in tax and spend.


  • A new regression is run the main finding of which is that the turnout variable is significant and indicates that an increase by one standard deviation in turnout raises social spending by 1.89 percent.



T. Iverson & D. Soskice

American Political Science Review Vol. 100, No.2 (May, 2006)

 A Summary

In a Nutshell

The Meltzer Richard framework where the individual with the median income  is the decisive voter claims that with typical right-skewed distributions of income the median voter will push for redistributive spending up to the point where the benefit of such spending to the median voter is outweighed by the efficiency costs of distortionary taxation. The general idea is that democratic institutions empower those who stand to benefit from redistribution.

 This implies that redistribution will be greater in democracies relative to authoritarian regimes. This is partly borne out by empirical studies. Yet it also implies that systems where inequality is high will also redistribute more as the distance between the median income and the mean income will be greater indicating the critical voter (the median income individual) will prefer more redistribution than the same individual in societies where income is more evenly distributed (and hence there is less distance between him and the mean income). However, many studies now show that this relationship does not hold, and is in fact the reverse.

 So the reason that some democracies distribute more than others does not seem to be because of inequality levels. One possible alternative is that the power of the working class varies across democracies. If redistribution is a function of policy and policy reflects the preferences of those that govern, then differences in government partisanship could explain the variation. In particular left governments may redistribute more than right governments. But that begs the question: why are some democracies dominated by left governments and some by right?

 It is here argued that the difference is driven by electoral systems. Empirically majoritarian systems are centre right (75%), and PR systems are centre left (75%). The electoral system affects coalition behavior and leads to systematic differences in government partisanship, and this difference has distributive outcomes. In a two party majoritarian system a centre-right party is more likely to win power, and in a PR system it is the centre-left. This outcome is largely related to a departure from the stand Meltzer Richard model such that taxes and transfers can vary across classes, and also that the middle classes fear taxation by the poor even though they face an incentive to ally with them in order to expropriate from the rich.

 The Model

  • Individuals are a member of one of three classes, L, M, H of equal size.
  • The poor always gain from redistribution, the rich always lose, and the middle always does less well than the poor. Thus redistribution (unlike in the MR model) is progressive. There is no longer a simple median voter result – redistributive politics can only be understood as the result of class coalitions. This assumption reflects the fact that governments are constrained by some notion of fairness (perhaps enforced by free press etc. – meaning it may not apply in less developed democracies).
  • Thus the optimal policies are:
    • L: tax H and M and keep all the receipts
    • M: tax H and share the receipts with L
    • H: No tax for no redistribution.
    • Thus the middle class can either ally with the poor to tax the rich, or it can ally with the rich to avoid being soaked by the poor. Which motive dominates, and hence which coalition emerges depends upon the electoral system.


  • Majoritarian systems can only sustain two parties in equilibrium so there is either a centre-left LM party or a centre right MH party. There is thus bargaining between LM or MH as to policy platforms. Because L always will vote L, and H will always vote H, winning is equivalent to getting the M vote. Thus both L and H parties have an incentive to moderate their policy platforms.
  • However the L and H parties cannot write binding agreements with voters that they will adhere to their platforms. So there is some chance that L in LM will deviate left, and H in MH will deviate right i.e. enacting their true preferred policy. There is some probability that the L, or H elements will defect, and this probability is learned by the voters in the campaign. Voters then vote strategically.
  • The upshot of this is that M has more to fear from an L defection than an H defection. If the probability of defection is the same for both, then they will prefer an MH party to a LM party. As M’s vote determines the election, this motive determines the ruling coalition.
  • If L of LM deviates left the middle class will be taxed in favour of redistribution to the poor. If the H of MH deviates there will be no tax. This means that the middle class will not benefit from redistribution, but it will gain itself from not being taxed. Thus an MH defection is better than a LM defection and hence why given equal probabilities of defection, the middle classes will vote centre right.
  • Middle class voters do not like right wing governments but they have less to fear from them. This does not mean that MH will always win, because in any given election leader will be perceived to have different likelihood of defecting.


  • Three class parties represent L, M and H. If no majority there must be a coalition. In such a situation one party forms the coalition (requesting the other party to join). Parties will split the redistribution pie subject to the non-regressivity constraint.
  • Policies are set by a majority of two parties that bargain.
  • Here party M is the coalition former. M prefers to ally with L. by contrast with the majoritarian case, M has no fear that it will defect, as it represents its own interests in coalition negotiations. Hence M will vote M, and then ally with L, so they can share the taxable pie.
  • There is no credible threat of switching.


  • Empirical analysis of democracies since WWII.
  • Inequality is negatively correlated with redistribution, but this changes when the political institutional variables are included.
  • 2/3 of governments under PR are left of legislative median, and 2/3 to the right under majoritarian.
  • They also show that partisanship has a significant effect on redistribution.


  • The model may not work in less established democracies where the poor have little or no collective action capabilities or the rich can threaten coups. If the poor can be ignored when not in government, then MH becomes more likely as the middle classes do not have to share with L.