Greg Mankiw appears to miss the point. He writes:
Greg Mankiw's Blog: News for Emily Litella: All of us who work with macroeconomic data have learned, sometimes the hard way, that conclusions based on preliminary data are subject to change. Here is a good example. The NY Times two days ago in a front-page, above-the-fold article: "wages and salaries now make up the lowest share of the nation's gross domestic product since the government began recording the data in 1947." The NY Times online today, presumably to be reported in tomorrow's paper (on what page? [update: page C1]).... As a result, wages and salaries no longer make up their smallest share of the gross domestic product since World War II.... As Emily Litella would say, "Never mind."
By continuity, the fact that wages and salaries are almost at their lowest share of income in the government's memory is almost as interesting and alarming as the fact that wages and salaries are at their lowest share of income in the government's memory. It is simply not true that it's interesting if wages and salaries are at their lowest share and not interesting if wages and salaries are almost at their lowest share.
UPDATE: And Jason Furman writes to Greg Mankiw:
Update: My friend Jason Furman emails me his observations on the matter:
Greg,Measured properly wages and salaries in the first half of 2006 were the lowest as a share of economy since WW II.The proper denominator for factor shares is gross domestic income (GDI) or national income. Using either denominator, wages and salaries are the lowest since World War II. CEA's Economic Report of the President always shows incomes as a fraction of GDI as does NIPA Table 1.11.
You are correct that as a share of gross domestic product (GDP) wages and salaries are no longer the lowest since WW II. But this is an apples-to-oranges comparison -- which is why neither the very careful CEA nor the very careful BEA would do it. (It is, alas, a common practice to use GDP in the denominator -- one that many excellent economists and reporters, including the New York Times -- follow.)
Specifically, the statistical discrepancy between GDP and GDI shows up in the denominator but not in the numerator. In 2006-Q2, GDP was $76 billion lower than GDI. If you're using GDP in the denominator, then this $76 billion should be subtracted from some combination of wages, profits, rents, etc. Conversely, in 2005 GDP was $71 billion higher than GDI. As a result, to use GDP in the denominator you should also add this $71 billion to incomes.
Since we don't know how the statistical discrepancy breaks out between different factor incomes (that's why it's a "statistical discrepancy") we use GDI in the denominator. Or to avoid screwy results from depreciation, I generally prefer to use national income. Either way, the Times headline holds up.
P.S. I should say I'm not that interested in wage shares, compensation shares are the relevant metric for most important questions. And the upward revision in wages certainly is a good sign and helps make sense of the revenue surprises we experienced this year -- and might suggest they'll be somewhat more durable than the potentially more ephemeral capital gains and corporate profits. But since you were being picky, I thought you might as well help your Ec 10 students to learn some national accounting arcana.
I would dispute Jason's belief that compensation shares are the relevant metric for most important questions. Compensation shares are the relevant metric for some important questions. But in my view, the ratio of median compensation or the geometric mean of compensation to labor productivity is the metric for most important questions.