On Tuesday, September 12, the U.S. Census Bureau released the 2016 income, poverty and health insurance statistics from the Current Population Survey Annual and Social and Economic Supplement, known as CPS ASEC. The 2016 data showed that the U.S. median income in 2016 was $59,039. By simply looking at median household income from previous years, one could infer that the 2016 median household income was the highest on record, surpassing the median income for 2007 ($58,149 in 2016 dollars) and 1999 ($58,665 in 2016 dollars).

However, these comparisons must be made with caution because the income questions were redesigned in the Current Population Survey in 2014 for data year 2013; therefore, some of the difference in median household income between 2016 and the previous highs is due to the question differences. To better understand potential differences from the change in questions, we used a split panel design in 2014 with the traditional and redesigned income questions. Analyzing 2013 median household income data collected from both sets of questions pointed to a 3.17 percent increase in median household income in 2013 due to the change in questions (see Description of the Split Panel Test of the CPS ASEC Income Redesign and Time Series Guidance and Appendix D of Income and Poverty in the United States:  2014).

Without adjusting for the change in the income questions, the difference between the 2016 median household income and the 1999 median household income is not statistically significant. The difference between the estimate for 2016 and the estimate for 2007 would be statistically significant, with the 2016 estimate higher than the 2007 estimate ($59,039 and $58,149, respectively). In other words, 2016 and 1999 appear to be in a statistical tie as the years with highest real median household income.

A simple method for adjusting pre-2013 income data to make them comparable to current data could be to increase them by this 3.17 percent. Doing this would increase median incomes to $60,524 for 1999 and $59,992 for 2007. If we further account for the margin of error in the 3.17 percent adjustment, the median household income for 2016 would not be statistically different than either the 1999 or 2007 medians. All three years would be in a statistical tie for highest real median household income. In fact, the 2016 median household income would not be statistically different from median income in any year between 1998 and 2008 after the 3.17 percent adjustment for the change in questions (see the full adjustments from 1967 through 2016).

* Denotes statistical difference compared to 2016, standard errors in parenthesis.

We do not make these adjustments in our official publications and table packages because they require the assumption that the 3.17 percent difference between responses to the historical questions and the redesigned questions in 2013 would have been identical in all years before 2013.

While the CPS ASEC provides income statistics, another source of income data is the American Community Survey, which provides consistent statistics back to 2005. Additionally, CPS ASEC questions on earnings were not changed in the redesign. This allows us to analyze changes in earnings over time across the redesign as we do in our blog “Median Earnings over the Last 40 Years.”