Friday, June 6, 2014

Did you hear the one about how income inequality isn't increasing?

Analysis at the "Political Calculations" blog suggesting that income inequality has been flat for decades has been making the rounds, including with Mark Perry and Don Boudreaux. I'm no inequality expert, but I wanted to make few points that I think are getting lost in sharing this around.

1. First, the basic analysis is right that household formation patterns are very important for understanding different measures of inequality. This is widely acknowledged.

2. But, what I think people are missing is that there are a lot of ways of measuring inequality besides the Gini coefficient. The Gini coefficient aggregates changes across the entire income distribution, which is a very important property! But it doesn't necessarily give you a good measure of what's going on at different points in the distribution. Most people who talk about inequality have been referring to what's going on at the upper end of the income distribution. Gini coefficients are fine, but don't use a Gini coefficient to try to refute a completely different claim.

3. I am taking the result at face value for the time being because I haven't crunched the numbers myself and they come right from the Census Bureau! But, we should be careful. They come from the CPS, which top-codes income, and of course we know that the biggest changes in inequality have come at the top of the income distribution. As recent Economic Policy Institute analysis has shown, this is even true of wage income. They use Social Security data which I assume does not top-code like the CPS data. So would the Gini coefficient hold as steady if better data were used? Maybe, maybe not. You just have to know what you're dealing with when you pass this stuff around.


  1. Social Security data may well top-coded (it's been a fairly long time since I looked at it), because taxable income for Social Security and Medicare is capped (at different levels, just to make things more confusing).

    1. Thanks. I suppose it depends on what the data ultimately comes from. Obviously it goes to the IRS with all the data. Whether that is in the Social Security records is a different question, though.

  2. It would be interesting to see if the Census Bureau income series across time increases at the same rate as GDP income increases across time. That would catch some problems are GDP income is quite accurately measured (as far as I understand).

  3. The Census Bureau top codes the data it publishes for both very high and very low income earners (such as those who report negative income), which it does to keep the identities of those whose incomes are such that they might otherwise be able to be identified from the information they provide as part of the CPS income survey. That said, the Gini coefficient that is calculated and published by the Census is based on the full sample of data they collect - top coding is not a factor in its determination.

    The biggest weakness of the Census data in representing income distribution is that it excludes income from capital gains, which for most Americans, really represents a once-in-a-lifetime event (such as a doctor or lawyer selling their practices when they retire). There are those whose investments generate capital gains on a more regular basis, but that source of income tends to be highly volatile from year to year. In practical terms however, if that source of income were included, the impact on the calculated Gini coefficients for individuals, families and households would be to shift them upward to slightly higher values - it wouldn't do much, if anything, to alter the trends they indicate over time.

    The CPS data also underreports the effective incomes of those at the lowest end of the income spectrum, where it omits the value of welfare benefits (cash, food, housing, medical care, etc.) received by people with low incomes. If that data were included, it would lower the reported Gini coefficients. I'm afraid I don't know to what extent that source of income for low income earners would wash out the effect of excluding the capital gains earns by those at the upper end of the income spectrum.


All anonymous comments will be deleted. Consistent pseudonyms are fine.