Census report of big jump in income is a little too good to be true

The Census Bureau recently released new statistics on the trends in poverty and household income.  In contrast to the dismal numbers the Bureau published last year, the latest statistics show robust gains up and down the income distribution. They also show a sizeable drop in the poverty rate.  The headline number in the new report was the surge in median household income, which increased $2,800 (or 5.2 percent) between 2014 and 2015. Median income fell in six out of the previous seven years, so last year’s sizeable gain was a striking contrast to a lengthy string of gloomy reports.

I suspect the reported income gain in 2015 was a little too good to be true. Based on national income statistics, it appears that average income gains were probably smaller. On the other hand, previous Census reports understated the income improvement Americans experienced earlier in the recovery. Based on comprehensive income data covering the entire economy and the full economic recovery, income gains over the recovery have been larger than implied by the household survey statistics.

The latest Census numbers show big proportional increases in the incomes of households near the bottom of the distribution. Households in the bottom one-fifth of the distribution, for example, saw their average incomes climb 6.6 percent. Not surprisingly, the improvement in bottom-end incomes produced a notable drop in the poverty rate. The percent of Americans classified as poor dropped 1.2 percentage points, falling to 13.5 percent, the lowest level in eight years.  Poverty fell in all age groups, though it dropped fastest among children. Poverty also declined in all regions of the country.

The income improvements were fueled by strong employment growth and healthy gains in worker earnings. The number of adults who reported being employed sometime during the year increased 2.1 percent in 2015 compared with 2014. That’s the biggest employment gain we’ve seen since the late 1990s. Aided by a low rate of consumer price inflation, workers also saw notable gains in the purchasing power of their earnings. In combination, growing payrolls and rising wages boosted the real incomes of the nation’s working-age families.

The good news on household income might seem a bit incongruous in a political year filled with headlines about voter anxiety and dimming middle class prospects. As most observers realize, however, the good news about income in 2015 was an exception. In recent years, it is far from the norm.

The nation’s main source of historical information about the level and distribution of family incomes is the Census Bureau’s Current Population Survey (CPS). That survey, which is the basis of the official poverty statistics and the standard measure of median income, presents a bleak picture of income gains in the 21st century. The income statistics published this week are a conspicuous exception to that pattern. They show a startling jump of 4.8 percent in average real income per person in 2015 (see Chart 1). Despite the surge in income last year, the CPS income numbers suggest average income per person in the United States was about the same in 2015 as it was back in 1999.

Real per capital growth in current recovery using two income measures

Although this pessimistic conclusion is accepted by voters and reporters alike, it does not correspond with other indicators of economic progress, including measures of personal income that are derived from sources other than household surveys. The personal income statistics published by the Commerce Department, for example, show healthier income gains in the 21st century compared with the household income changes reported in the Census Bureau’s CPS survey. According to the most recent GDP estimates, real U.S. money income per person increased almost 11 percent in the first six years of the current economic recovery—that is, from 2009 through 2015 (Chart 1). In contrast, the Census CPS survey suggests that the income gain was less than one-third as large (3.3 percent versus 10.9 percent).  To insure that the two estimates of income improvement are comparable, I have applied the same inflation adjustment used by the Census Bureau to both estimates. (The Census Bureau uses the CPI-U-RS.)

The Commerce Department’s annual income series also shows a different pattern of income gains over the course of the recovery. While the Census household survey suggests that virtually all the income improvement in the recovery occurred in 2015, the Commerce Department’s personal income statistics show sizeable income gains in four of the last five years. This gap between reported income gains in the CPS survey and in the national accounts has persisted for more than a decade.

Chart 2 shows trends in real per capita income using the two income measures. From 1973 until about 2003 the chart shows that both measures produced similar estimates of the pace of average income growth. The lines diverge after 2003. Compared with the CPS estimates, the Commerce Department’s money income series shows faster income gains in economic expansions and somewhat steeper income losses in the Great Recession. Over the entire period from 2003 to 2015, the Commerce Department’s estimate of average income gain is about 9 times larger than the meager gains captured in the CPS survey.

Trend in real per capital income under two definitions

In sum, the Census Bureau’s household survey showed exceptionally fast income growth in 2015. Between 2003 and 2014, however, the income gains in the CPS badly lagged those seen in the national accounts. The Commerce Department’s income estimates in the national accounts are derived from income tax returns, earnings records maintained by the unemployment insurance system, and administrative records provided by government transfer agencies. While the data provided by these sources are subject to error, they are almost certainly more accurate in the aggregate than estimates based on household survey responses.

Of course, the national accounts data by themselves provide us with no evidence about the prevalence of poverty, the distribution of income, or the trend in median income.  Only household survey data offer a guide to how income is distributed across households and across individuals. The fact that income gains recorded in our most important household survey are much smaller than those seen in the national accounts should be a source of concern. If the household survey is giving us an inaccurate picture of the trend in average income, it is inevitably giving a distorted picture of the trend in income at many points in the income distribution, possibly including in the middle and at the bottom of the distribution.

The implications of these survey problems depend on the types of income that are misreported and the income levels of people who understate their incomes to interviewers or fail to respond to Census surveys. The issues can and should be investigated using a combination of survey and administrative files. If the survey findings are worth reporting on the front pages of the nation’s newspapers, the data they rely on are worth checking against competing, and perhaps more reliable, sources of information.