Friday, July 27, 2012

Inaccurate Consequences

Greg Mankiw has a link to Edward Conard's new book, Unintended Consequences. Appreciating Mankiw's recommendation, and remembering that I had watched Conard on the Daily Show a little while back, I read the available-online introduction and first chapter.

Mankiw, in particular, pointed to a little chart on page 22 of Conard's book, which I've reproduced below within my understanding of rules of fair use.

I haven't read the entire book, but I want to briefly comment on this chart in particular. Conard uses it to argue that there has been no median-income stagnation in the United States, and that what's really happened is that since 1970 we've brought into the labor force large numbers of low-income minorities. Conard alleges that the consequence is that even though median real incomes within each demographic have risen, overall median real income has stagnated.

I admit, it is an intellectually appealing idea. I almost wrote this blog post about it as a follow-up to Wednesday's post, which explained how the emergence of a global market for labor in the early 1970s released the force of factor-price equalization, and as a result, real incomes haven't kept up with productivity.

But then, when I began to look at the data myself, I realized that Conard is wrong. First, I believe his data is inaccurate. Second, his conclusions are testable and do not hold up to my analysis.

Let's talk about the data inaccuracy first. No blogger, especially me, ever wants to have to accuse someone of fudging the numbers. Having to debate the facts distracts from considering the spectrum of valid opinion facts ought to produce.

But here I am, looking up the Census data, and I can't see where Conard's numbers are coming from. Let me be very specific, in hopes of getting him to respond or having someone explain to me where I've gone wrong. Tables P-2 and P-4 of the Historical Income Tables of the Census -- the "P" stands for people, as opposed to "H" for households or "F" for families -- do not reflect, even by rough approximation, any of the numbers in the above chart. (These data include part-time workers, as Conard does.)

Although Conard uses 2005 constant dollars and the new data uses 2010 dollars, we can cross-check his numbers with the 2005 current dollar income on Table P-2. None of them match, and it's not even close. Here's just one example: Conard's chart says that white men had a median income of $35,200 in 2005; the actual Census data says $31,275.

Nor are Conard's 1980 numbers any good, and what that means is that his percentage changes are also inaccurate. The Census data in P-4 says that median real personal income, inclusive of all demographic groups, rose 35.7 percent between 1980 and 2005. Conard alleges it increased 3 percent.

Conard's conclusion is that median income stagnation, holding demography constant, is a myth. In fact, he's got it totally backwards.When you don't hold demography constant, median income has risen, although not nearly as quickly as mean income or real GDP per capita. Stagnation in median incomes appears only in particular demographic groups -- namely men, and in particular, white non-Hispanic men. For these groups, there has been no increase in real incomes since the early 70s, and that conclusion is unavoidable.

By extension, median income has risen in the last 40 years because female and some minority median incomes have increased substantially. These increases has moved this large fraction of the population away from being clustered at the low-end of the income distribution and towards the all-population median. The statistical consequence is that these demographics stopped holding down the median value of real income and thus median real personal income across all demographics rose.

3 comments:

  1. "Here's just one example: Conard's chart says that white men had a median income of $35,200 in 2005; the actual Census data says $31,275".

    The $31,275 figure you cite is for ALL males.

    If you go to "White Alone, Not Hispanic" the number for 2005 is $35,345.

    Switching that number to 2010 money gets you $39,475. Compare that to "White, Not Hispanic" circa 1980 and you get $34,466 (2010).

    That's a 14.53% increase, which is in the ballpark of Conard's chart of 15%.

    ReplyDelete
  2. The chart is confusing, and I cannot get it to foot exactly (though I can somewhat recreate it). I'm not convinced you two are really in disagreement, though.

    You say, "The Census data in P-4 says that median real personal income, inclusive of all demographic groups, rose 35.7 percent between 1980 and 2005. Conard alleges it increased 3 percent".

    Conard writes, right before the chart, "Median wages have increased 30%, on average, across all demographics".

    He does not allege that it is 3%. To get the 3% - near as I can tell - he takes 2005 income and weights it using 1980's demographics (obviously he's not saying this is actually what happened between 1980 and 2005).

    So, had there been no demographic change and everything magically remained the same in those terms, overall Median Income would be up 3% (holding demographic income equal).

    Adding in the actual change in demographics gets the number to 30% (or above).

    That's what I hear Conard saying.

    ReplyDelete
  3. I find the description of the tables to be highly confusing. I mean, wth does this mean:

    Race and Hispanic Origin of People by Median Income and Sex

    @Evan Soltas - where do you get older numbers for non-hispanic whites? From what I can see, the data for non-hispanic whites starts in 1993.

    ReplyDelete