Saturday, February 21, 2015

How US income distribution changed between 2007 and 2013

[The previous version had a couple of mistakes due to an error I made with the CPI index. The graph likewise has been corrected. My apologies.]

The Great Recession has brought an incredible (in the real sense of the word) increase of interest in income and wealth distribution in the United States. This has of course been a welcome development not only to many people (like myself) who have been working for years, in a kind of a Platonic cave, on an underground topic. It is a welcome news, shaking its complacency, to the US political and economic establishment which has largely ignored changes in income and wealth inequality. And it is a very good news for the academia, which similarly --whether because of ideology or due to more prosaic concerns (like economic incentives)-- has also tended to ignore it.

As one would expect, this new interest in the matters of distribution has proven to be politically very contentious. And since people have strong political opinions and since income and wealth inequality have become the topic of the day, many people who otherwise never dabbled in income distribution have had their field day. This is best seen in the proliferation of income, consumption and wealth measures. I have written a bit on it here, and I do not want to go into all details of definitions in this short post. But some people have acted as if no standards existed on how income and income distributions are measured. More than half-a-century of work on the topic was ignored (or more likely, those who wrote about it did not even know it existed), Thus all kinds of bizarre measures have been proposed as if the entire corpus of knowledge had to be reinvented, or as if America were an island which needs to have its own measures of income and inequality unrelated to what is done in the rest of the world. One could, I guess, as well start inventing American concept of Gross Domestic Product.

So, in this note I will use the international  definition of disposable income (income after transfers and direct taxes) to show what happened to US income distribution between 2007 and 2013. Disposable income includes all annual flows of income from work (like wages and fringe benefits), from property (dividends, interests, profits of the self-employed etc.), and from social transfers (social security benefits, unemployment benefits, and near-cash transfers like food stamps). It deducts direct taxes paid. This is economists’ favorite concept used in all international  income comparisons. The data come from US Census Bureau’s Current Population Survey (CPS), from its recently revised and improved historical data series. CPS has since 1962 been the prime instrument for monitoring poverty and income inequality in the United States. The CPS data are “lissified” by Luxembourg Income Study which means that the definitions of income aggregates are made comparable between the countries, so that one can easily compare (say) income or wage levels in the US with those in Canada, France or Brazil. LIS data are available here (but you have to be registered to use them).

In this post, however, we are concerned only with US income distribution. One thing which is important to note is that income concept used does not include capital gains and losses. According to the Canberra groups handbook definitions (a “bible” for economists working with household surveys, see 2.3.6, page 16) and UN system of National Accounts capital gains and losses are considered changes in the wealth position, not income.

There are several substantive and practical reason to leave capital gains and losses out of income. They are volatile and would often make havoc of income distribution statistics with otherwise rich people, who in some bad years register large capital losses, being treated as the poorest people in the country (with negative incomes). This clearly does not make sense. But dropping capital losses and using only capital gains (which would also increase volatility of incomes) would be inconsistent. So you have to drop both. This does not mean that capital gains and losses do  not matter: it simply means that they do not affect income, but that they do increase or decrease wealth.

Figure below shows the change in real disposable per capita income across US income percentiles for 2007-2010 and 2010-2013.  The percentiles run from the poorest (number 1 on the left, on the horizontal axis) to the richest (number 100, or the infamous top 1%, on the right). US Consumer price index is used as the deflator.

Consider the first period (blue line). It is remarkable that real income of all groups declined. But the hardest hit were the rich, with percentage losses increasing as we move toward to right portion of the graph, and the very poor.  I am not an expert on US welfare system, but it seems to me that the system failed to protect the poorest people from substantial income losses between 2007 and 2010. But for the bulk of the population, the years of the Great Recession meant a modest real income decline. The median person’s real income went down by a little over 3 percent. The upper middle class (the people between the 80th and 90th percentiles) did not see much change in their real income. But the top 10% clearly lost out: notice how the blue line starts decreasing ever more steeply as you move toward the top 1%. The Gini coefficient decreased by less than 1 point.

Now, look at the red line which shows the real change in the second period. It is almost a mirror-image of what happened in the first. The growth was zero or positive along the entire distribution, the strongest among the very poor (around the lowest 5th percentile) and among the rich (the top 10%). Median inflation-adjusted per capita income decreased by just under 1%. For the two top percentiles, which  got clobbered by the recession, real income growth was in excess of 10%.

Very little changed in the shape of the distribution: the rich got back almost all  they lost in the first three years, and so did most of the poor, except the bottom two percentiles which seem to be the outliers (the bottom percentile always is). The poor also seemed to have been better protected by the social safety net in the second period. (People more diligent than I can use LIS or CPS data to figure out the exact income composition of both the poor and the rich, and for the former to check whether it was the social safety net or greater labor participation or something else that did the trick in 2010-13.)

So, the conclusions are, I think, three:

1.  As expected in the financial crises, the first negative income shock affected the rich. But they fully recovered as the economy improved.

2. The vast bulk of the population, from around the median to the 90th percentile, went along with the economy: when the economy “tanked”, they lost; when the economy improved a bit, their real incomes were flat. On balance, they are not better off in 2013 than they were in 2007.

3. Little changed in the income distribution except for the slight increase in the share of the richest decile: they control larger share of  total income in 2013 than in 2007. The Gini index increased from 40.7 in 2007 to 41.2 in 2013. So, if people think that something serious has to be done to improve US income distribution, that job is not done: it remains in the future.  If people think that nothing needs to be done, then everything is just fine.

Nota bene: There is a caveat. Income of the top 1% in 2013 is a preliminary estimate. A “final”  number will be soon available, but it is unlikely to change any of the trends or our conclusions.

If you wish to check how the CPS data match National account aggregates, here is a comparison of CPS mean income growth rates with the growth rate of real GDP per capita and real per capita household consumption.

Period real growth (cumulative).

CPS real mean per capita income
Real  per capita GDP
Real per capita personal consumption

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.