It seems the Incredibly Incurious Journalist Jonathan Chait has decided to double down. Picking up the story from where we make our appearance in Jonathan's cloistered world:

Likewise, the blog Political Calculations — whose work Pethokoukis has cited as refuting the “myth” of income inequality — is unhappy as well. My post pointed out that its supposed refutation of rising inequality is erroneous, because it relies on census data. The Census Department does not collect detailed information about rich people’s income, which is why inequality researchers look elsewhere when they want to study changing income among the very rich. Lane Kenworthy, an actual expert in this topic, helpfully explained the folly of the Political Calculations chart. I thought his explanation was too detailed to be of interest to readers here, but since they’re complaining, I’ll reprint a longer excerpt of his e-mail to me:

You know, it *still* hasn't occurred to the Incredibly Incurious Journalist Jonathan Chait that perhaps, just perhaps, questions about our work should be directed to us, as he has *still* made no apparent effort to contact us with any such questions! (We've even checked our e-mail spam filter!)

Although he isn't doing anything more than regurgitating the contents of an e-mail he received from an individual he has proclaimed to be an "actual expert", let's see what we can learn from it. Better yet, let's also see if we can do what appears to be beyond Jonathan Chait: ask questions about the information he is accepting at face value! From the e-mail of Chait's "actual expert" Lane Kenworthy:

Over the period since 1994 the Census Bureau's standard Ginis for households and for families, which are the ones used in the Political Calculations chart, suggest little or no change in income inequality. So too do Ginis for earnings inequality among full-time year-round employed individuals (see IE-2 at http://www.census.gov/hhes/www/income/data/historical/inequality/index.html).

We're mostly off on a good footing so far - Chait's "actual expert" appears to have checked our source data to confirm that our chart is accurately reflecting the Census' data for families and households, and he's added some new information: the earnings inequality data for full-time, year-round employed individuals also shows little-to-no change. Although we used the data for "All Persons", rather than "Full-Time, Year-Round Employed Individuals", we would expect these different data sets to be similar.

But it's well-known that the Census Bureau data miss what's going on at the top, because they "top code" very high earnings and incomes. What the Census data tell us, and what the Political Calculations post in effect simply reiterates, is that within the bottom 99% there has been little change in income inequality since 1994, whether we're looking at households, families, or individuals.

This is where it really begins to go a bit south for Chait's "actual expert" Lane Kenworthy. Now, we were pretty up front in our previous post on the topic that yes, the Census' reported data is "top-coded", which is "Census-speak" for the situation where all those with total money income above a certain threshold are lumped together into a single grouping that doesn't identify specifically what each makes, so yes, we're pretty sure we're aware of the practice.

But there are two key things that Chait's "actual expert" Lane Kenworthy has missed here: the minimum income threshold to qualify as being in the Top 1% for Families, Households and Individuals and also the level of income at which the Census actually "top-codes" its data for each!

The omission might lead one to think that Chait's "actual expert" actually thinks that the Census' income data is only valid for the "bottom 99%". That somehow, the possibility that the Census' income data might significantly overlap the entire annual incomes earned by a large number of individuals, families and households in the "Top 1%" for each category has escaped him. As does the possibility that these numbers might in fact be large enough to constitute a large enough statistical sample of the "Top 1%" that we can indeed make valid assessments of the overall actual trends in income inequality using just the Census' data.

A curious journalist might ask:

"How many of the Top 1%'s total money income is fully contained within the Census' data before it hits the actual income level where it has been 'top-coded'?"

Alas, the Incredibly Incurious Jonathan Chait is not such a journalist....

We can ask "Why begin in 1994?"

Hey, a curious journalist might have asked us the same question. We even know the answer!

But the bigger problem is ignoring what happened at the top of the distribution.

During this particular period, income inequality increased due to the top pulling away from the rest. According to the Census Bureau's calculations for households, there was essentially no change in inequality from 1994 (Gini .456) to 2007 (Gini .463). (Again, these are the data for households used in the Political Calculations chart.)

That's strange - we would describe a change in the Gini value from 0.456 to 0.463 as an "increase". In fact, we did, but more on that later....

Check this out now - our chart runs from 1994 through 2010 (as it happens, these are all the years for which the Census has made the data for families, households *and* individuals easily accessible on its own web site in a web-friendly format that anyone can access to confirm that the data we presented was correct - see, we really did know the answer to the question of "Why begin in 1994?"!).

But for some strange reason, Chait's "actual expert" decided to stop the income inequality clock at 2007. A curious journalist might ask:

"Did something happen after 2007 that might affect the income inequality numbers that the 'expert' doesn't seem to want to talk about?"

or...

"Should I really trust an 'actual expert' who's clearly using a cherry-picked set of data to make their 'expert' claim?"

Alas, the Incredibly Incurious Jonathan Chait is not such a journalist....

The best source of income data for households is the CBO, which merges the Census household survey data with IRS tax record data. According to the report the CBO put out a few weeks ago, household income inequality including the top 1% actually increased sharply between 1994 and 2007. The CBO report is at http://www.cbo.gov/doc.cfm?index=12485; see Figure 11 on p. 20; the closest CBO counterpart to the Census income measure is "market income plus transfers". Eyeballing the CBO chart, it looks to me like the Gini goes from .45 in 1994 to about .52 in 2007. That's a large increase.

It's all fine and dandy to compare the results of the equivalent households from the CBO's study to the households of the Census to get a sense of how different they might be from one another with one adding in data not present in the other. In this case, much of that additional data covers an especially volatile component of income that can represent a large portion of the annual earnings of high income earners.

For our purposes, what Chait's "actual expert" Lane Kenworthy has accomplished here is to confirm that the Census data for households shows an increase in the measured income inequality of households from 1994 through 2010 and that the CBO's data, which includes that additional volatile income component that is not fully captured by the Census data, shows a larger magnitude increase, which is really what anyone should expect. Seeing as we made no claim to the contrary with respect to *household* data, our official position is "Yes, and...?"

A curious journalist, or even an "actual expert", might then ask:

"What results did the CBO obtain for individuals and how did they compare to the CBO's results for households?"

Or for that matter:

"What results did the CBO obtain for individuals and how did they compare to the Census' results for individuals?"

After all, the whole point of Political Calculations' original analysis is to compare the respective income inequality trends over time for Families, Households *and* Individuals. Because the Census' data for each was collected simultaneously, through identical methods and handled by the Census throughout its analysis using consistent methods, top-coded or not, the trends for each over time may be directly compared to one another, with valid conclusions being able to be drawn from an analysis of those trends.

Doing that, we observed that the Gini coefficient for both Families and Households increased from 1994 through 2010, but that this measure of income inequality for Individuals was flat, or even slightly fell, over this time. That observation indicates that things other than economic factors are responsible for the observed increases in the measured income inequality of both Families and Households, because if economic factors were involved, the same kind of increase in measured income inequality would have been observed in the data for Individuals.

Other than almost confirming that we accurately represented the data recorded by the Census for Individuals (remember, he looked at "full-time, year-round income earners rather than "all persons"), Chait's "actual expert" has avoided any discussion of that particular data, obsessing instead over the data for Households.

That's useful for us up to a point, given that Chait's "actual expert" has confirmed that the Census data is indeed picking up on the same general increasing trend that exists when the CBO's additional volatile income component is factored into a separate analysis. At least with Census household data and what the "expert" has identified as being the nearest equivalent measure to that in the CBO's analysis.

But Chait's "actual expert" is awfully silent regarding the trends for income earned by Individuals - the foundational component of both Families and Households. It is as if Chait's "actual expert" lacks the actual expertise to offer any comment in this area.

The only question remaining then, which might be asked by a curious journalist or even by an "actual expert", is:

"If the data to produce such an analysis for the trend in income inequality for Individuals exists, why hasn't it been produced?

It's not like that analysis can't be done. You would think that someone who has been proclaimed by a journalist to be an "actual expert" on income inequality would be capable of doing that themselves. After all, the CBO used the Census' income data as a foundation upon which it grafted IRS data to produce its "equivalent-to-Census households" results. It would be odd that they wouldn't be able to do the same thing with the data for Individuals, even if other "actual experts" cannot bring themselves to do it.

Pethokoukis triumphantly concludes, "Chait's theory of the case has come apart." In fact, "Chait's theory of the case" also happens to be the consensus view of every expert who actually studied the issue and knows which data to examine.

Seeing as the "experts" upon whom the Incredibly Incurious Journalist Jonathan Chait relies for his theories apparently don't have their act together in this area, otherwise we would already have the answers to each of the questions above, Chait's proclaimed "consensus" among them perhaps isn't all it's cracked up to be. Remember, these are the same people to whom it apparently never occurred to compare the income inequality trends over time for Families, Households *and* Individuals.

Maybe if somebody were actually curious enough to start asking them the right questions....

Alas, the Incredibly Incurious Jonathan Chait is not such a journalist!

Labels: income inequality, quality

- posted by Ironman at 5:57 AM | Permalink | Home |