Partisan Inequality?

Dani Rodrik recently posted a picture from Larry Bartels’ latest book – in which Bartels shows that in the US, inequality rises under Republicans, and falls under Democrats. Here’s the key graph:

What’s curious about this is that it seems to be a finding restricted to the US federal system. In this paper, I look at US states, and find that while the effects are statistically significant, the magnitudes are very small. And using top incomes data, I find no systematic pattern across countries. Here are the top income results, from a chapter that I have forthcoming in the OUP Handbook of Economic Inequality. Basic result – on average, top incomes inequality rises a smidgin faster under right-wing governments (though in Australia, the pattern goes the other way). But in no case is the difference discernable from statistical noise.

Note: Top income shares from Leigh (2007) and partisan coding from Armingeon (2006). The party coding refers to whether right-wing or left-wing parties hold the largest share of cabinet posts. For simplicity, I include centre parties with left-wing parties. The mean difference refers to the mean of the column (i.e. the mean difference for countries that have data on changes in top income shares under both right-wing and left-wing parties). The p-value is from a t-test of equality between right-wing and left-wing governments, with each run of right-wing or left-wing governments treated as a separate observation (e.g. in the United Kingdom case, the period of Conservative rule from 1979-1997 would be treated as a single observation for the purpose of this t-test). For the Netherlands, there is only one run of right-wing governments during this period, so the p-value cannot be estimated. In the last row, the p-value is from a t-test that combines data from all countries.

Update, 8 Apr: On Rodrik’s blog, Bartels responds to some of his critics.

Advertisements
This entry was posted in Uncategorized. Bookmark the permalink.

7 Responses to Partisan Inequality?

  1. Andrew,

    According to Rodrik…

    … Democratic presidents generate higher income gains for all income groups (although the difference is statistically significant only for lower income groups)

    In that light would be interesting to rebuild your table for income share of bottom quintile.

    Nathan.

  2. Andrew Leigh says:

    Nathan, I agree. But Australian income surveys only started in 1968, are only really comparable from 1981, and even then are done on a quiquennial basis…

  3. Peter Whiteford says:

    Andrew

    Your response to Nathan raised the question in my mind of how one deals with lagged effects, particularly when for some countries you don’t have annual income data. For example, as you point out the first truly comparable time series on income distribution starts in 1981 with the next survey in 1985 and then 1989 or 1990, as I recall (unless you were using tax data). I recall that income inequality increased between the early and mid-1980s and even by the end of the 1980s my not have gone down to where it was at the beginning of the decade.

    I think that there are also policy effects that carry over from governments of opposing poltical philosophies. For example, one of the first things that the Labour government did in the UK when it was elected in 1997 was to cut benefits for lone parents (in order to meet the previous government’s spending plans). In contrast, one of the first things that the Liberal government did in 1996 in Australia was to increase benefits for lone parents through one of the new family tax benefits. (Subsequently, both increased benefits for lone parents and low income families in real terms). but the initial policy response reflected the starting point – in the case of Australia, 13 prior years of a Labor Government and in the UK 17 prior years of a Conservative government.

  4. reason says:

    I’m not sure that this:

    The p-value is from a t-test of equality between right-wing and left-wing governments, with each run of right-wing or left-wing governments treated as a separate observation (e.g. in the United Kingdom case, the period of Conservative rule from 1979-1997 would be treated as a single observation for the purpose of this t-test).

    makes sense since policies may change in each electoral cycle, and policy affects may be cumulative.

  5. Pingback: Income Inequality: A Function of Parties or Study? : Tree of Knowledge

  6. Richard Green says:

    I wonder if there is some effect in US federal governments that isn’t present in state or other national governments due to the vastly larger electorate. Since the standard deviation from the median political opinion is presumably larger accross 300 million people in a vast area than within a single state or within smaller countries (even electorates in Germany and Britain are a fraction of the US) maybe US federal elections encourage more polarisation when a president plays to the base, and policy subsequently causes these outcomes.

    Afterall, the left and right is really relevant to the internal political culture of the electorate, and a Democrat in Virginia is closer to the Republicans in Virginia than a Democrat in California, and vice versa.

  7. Molesworth says:

    Could it be that one reason why the posited link seems to be more apparent at a national level is that particular areas of policy that are primarily the responsibility of the US federal government affect income inequality more than areas within the control of the states?

    Could another be that it’s really only at a federal level that we can talk about Democrat=left and Republican=right over the time span you look at in your paper? (Given that most of the southern Democratic Governors before the 1970s were right wingers even in regional terms while there were a number of northeastern Republican Governors who were clearly on the left.)

Comments are closed.