Deconstructing Median Income Bullshit

When I learned yesterday, 55 days before the 2016 election, that the Census Bureau and the White House had announced an historic leap in real (inflation-adjusted) median household income, my bullshit detector went into the red, went right off the scale and then ceased functioning. I'll have to get a new one 

Look at that jump! The biggest since 1967 when record keeping began.

How fucking likely is that?

My first clue appeared in the Los Angeles Times.

The clue is circled. Redesigned income questions?

And while I'm quoting the Los Angeles Times, there is this—

Obama, stumping for Clinton at a campaign event Tuesday in Philadelphia, did not pass up the moment to spotlight the census report. Obama said the uninsured rate was the lowest on record as was the pay gap between men and women.

"So, now, let's face it; the Republicans don't like to hear good news right now," Obama said. "But it's important just to understand this is a big deal. More Americans are working, more have health insurance, incomes are rising, poverty is falling, and gas is $2 a gallon. … Thanks, Obama!"

Keep those gasoline prices in mind. OK, this time I searched for "redesigned income questions" and found exactly one article at the New York Post. That's not the New York Times. No, that's the New York Post. (I added the links.)

Economic data that come out of Census, including the monthly unemployment report, should be scrutinized more carefully than they are — because they are what hurt Americans the most day to day, ruin dreams of hardworking families and, when the numbers are played with to present a better picture, erode trust in government.

Here are some details on just two points.

First, the 5.2 percent increase came during 2015 compared with the previous year. The Census Bureau doesn’t have up-to-date numbers.

But an outfit called Sentier Research, manned by ex-Census workers, does.

And while Sentier’s numbers confirm the increase in 2015, household income this year has slackened off, they told me Tuesday.

We'll look at the Sentier data below.

Gordon Green, a partner in Sentier, says median household income declined slightly in the first half of 2016. In both 2015 and the first half of 2016, incomes were boosted by a drop in gasoline prices.

“Oil prices play into this in a very, very big way,” Green said.

Follow the link to see the swan dive in gasoline prices which began in 2014 and continued into and throughout 2015. Use the 5-year chart. You'll see an insignificant seasonal rise in the summer months in 2015.

The second knock is that Census moved the goal posts.

Ah, make the data look better by doing the measurement differently. The usual move.

Starting in 2013 with a partial phase-in, which was fully implemented in 2014, Census changed the questions and the methods in calculating household income.

For example, Census, starting in 2014, began to “collect the value of assets that generate income if the respondent is unsure of the income generated.”

Also, the government started to use “income ranges” as a follow-up for “don’t know” or “refused” answers on income-amount questions.

Those are the redesigned income questions.

UPDATE: Census Bureau guidance states that the redesigned income questions increased their estimate of household median income by 3.2%. And see here, Appendix D, Table D-1.

For income, there were statistically significant differences for many key measures between the redesigned and traditional questions. For example, median household income calculated by using the redesigned questions was 3.2 percent higher than the median income found using the traditional questions.

As a result, it is difficult to assess whether an apparent change in median household income relative to previous years was a “real” change in income or an artifact of changes to survey questions.

Back to the New York Post.

There are plenty of other changes — but with just these two, income levels reported could be noticeably higher, say 5.2 percent higher, without the actual income being 5.2 percent higher.

In the fine print, Census admits the change. “The data for 2013 and beyond reflect the implementation of the redesigned income questions.”

Americans, in their guts, know the 5.2 percent gain in median household incomes isn’t true.

OK, now it was time to check out the Sentier data, which is updated monthly by ex-census employees. Doug Short is the go-to guy for this kind of thing, and I found this July, 2016 update.

Sentier Research, an organization that focuses on income and demographics, offers a more up-to-date glimpse of household incomes by accessing the Census Bureau data and publishing monthly updates. Sentier Research has now released its most recent update, data through November (available here). The numbers in their report differ from the Census Bureau's in three key respects:

1. It is a monthly rather than annual series, which gives a more granular view of trends.
2. Their numbers are more current. The Census Bureau's most recent is the 2014 annual data released in September 2015.
3. Sentier Research uses the more familiar Consumer Price Index (CPI) for the inflation adjustment. The Census Bureau uses the little-known CPI-U-RS (RS stands for "research series") as the deflator for their annual data. For more on that topic, see this commentary.

I highly recommend that you read the commentary that last link goes to. It's called Median Household Income Growth: Deflating the American Dream. The choice of the inflation-adjustment deflator makes all the difference in these median household income calculations. I will comment on that at the end.

Here's the first chart.

Sentier data deflated with the CPI (see above) shows a jump in real median household income (MHI) in 2015 which Sentier says was due in large part to low energy prices, (New York Times, August 30 2016). Given the very low inflation rate, Sentier estimated that median household income grew nearly 3.8% in 2015. If you use the core CPI, which excludes energy and food, MHI grew only 2.1% in 2015 (see the NYT report). MHI is still down from its early 2002 and 2008 peaks.

Here's the second chart, with some commentary from Doug Short. Remember, this is the latest data available on real median household income (July, 2016).

The next chart is our preferred way to show the nominal and real household income — the percent change over time. Essentially we have taken the monthly series for both the nominal and real household incomes and divided them by their respective values at the beginning of 2000. The advantage to this approach is that it clearly quantifies the changes in both series and avoids a common distraction of using dollar amounts ("How does my household stack up?").

The reality illustrated here is that the real median household income series spent most of the first nine years of the 21st century struggling slightly below its purchasing power at the turn of the century.

Real incomes (the blue line) hit an interim peak at a fractional 0.7% in early 2008, far below the nominal illusionary interim peak (as in money illusion) of 27.2% six months later and the latest at 40.3%, fractionally off the 40.4% record high in January of this year. The real median household income is now at -1.1% from its turn-of-the-century level. In essence, the real recovery from the trough has been frustratingly slow.

Summing up, real median household income (blue line just above) adjusted with the CPI is down 1.1% with respect to where it was in the year 2000. And if you follow that "Deflating The American Dream" link above, you will find out that deflating with the CPI-U-RS as the Census Bureau does puts a rosier glow on the historical data than deflating with the CPI does.

So it's all bullshit, isn't it?

And if you do this Google search, you will see all the usual suspects celebrating this nonsense. It's all over NPR this morning. Those assholes can't get enough of this happy bullshit.

I knew it was bullshit the moment I saw it, but thought it worth the time to figure out what was going on. And now we know.

Comments

[Jeff]

Thank you Dave for dissecting this so clearly. My bullshit detector also immediately blew a gasket on that news and I just assumed it was all junk. Good to see the details backing up that hunch. Appreciate your legwork here.