Expenditure Shares, Price Measurement, and True Relative Labor Productivity Growth in Post-WWII Manufacturing: What the Aggregate Deta Suggest

Tuesday Morning Distraction: Well, I was supposed to be sitting three tables down from Aaron Edlin at the Claremont Peets this morning doing research. But I got myself distracted–convinced myself that I ought to right something about the sharp Matthew Yglesias’s (and why, Harvard Economics Department, was he not an economics major?) piece on premature deindustrialization. And then I got myself redistracted…

Let’s start with one of the standard graphs: the American share of (nonfarm) employment that is in manufacturing:

At the start of the 1930s manufacturing employment was 30% of nonfarm employment. It is now 9% of nonfarm employment. In the absence of our trade deficit in manufacturing, it would be 12% of nonfarm employment. Thus only one-sixth of the reduction in the manufacturing share of nonfarm employment can be traced to the emergence of our manufacturing trade deficit–six-sevenths of the share decline is due to extra-fast improvements in manufacturing labor productivity and to shifts in the types of goods and services that we demand.

How much of the 18%-point ex-manufacturing trade deficit decline in the manufacturing employment share can be traced to demand shifts? Answering that question requires taking a view on what an unshifted demand would look like. Another standard graph is the nominal share of manufacturing production in GDP:

On the same axes as the nominal share of manufacturing production, I have plotted a series called “real” obtained from the nominal share by (a) multiplying it by the chain price index for GDP and (b) dividing it by the chain price index for manufacturing. This is not the real share of real GDP that is real manufacturing production. It is the “real” share. “Real” shares calculated this way do not add up to totals. This is a ratio of two flawed aggregative index numbers scaled so that in 2000 the ratio matches the nominal share of manufacturing in GDP.

At the start of the 1950s manufacturing production was 27% of GDP (manufacturing labor productivity was in value terms some 5/4 of average labor productivity, including the farm sector), and it has since fallen to 11% of GDP (manufacturing labor productivity is still in value terms 5/4 of average labor productivity). But over the course of the last 70 years the measured price of manufactures has fallen relative to the measured price of GDP by 1.4%/year, so that all of the declining share of manufactures in nominal GDP can be, in some sense, accounted for as an extra-fast improvement in manufacturing productivity and thus a relative reduction in market prices.

If there were a single commodity called “GDP” and a single commodity called “manufactured goods”, then we would say that:

• The relative price of “manufactured goods” today is only 40% of its level 70 years ago…
• If the income and price elasticities of demand for “manufactured goods” were one, then we would have expected the decline in relative price to 40% of its initial value to be associated with a 2.5-fold multiplication in relative quantities, and for the nominal share of manufactures in GDP to remain the same. Thus if the “unshifted demand” has associated with it a manufactured-goods demand curve of unit price elasticity, all of the 18%-point fall could be traced to changes in our market preferences away from manufactures. But I see nothing in the world or in our culture to generate such a shift in tastes and thus in market demand away from manufactured stuff. Manufactured stuff is useful, really useful…
• If the income elasticity of demand for “manufactured goods” were one and the price elasticity of demand were zero, then we would have expected the decline in relative price to 40% of its initial value to be associated with stability in relative quantities, and for the nominal share of manufactures in GDP to fall on the same track as the price–as it has. Thus if the “unshifted demand” has associated with it a manufactured-goods demand curve of zero price elasticity, none of the 18%-point fall could be traced to changes in our market preferences away from manufactures.

For our demand for manufactured goods to have a price elasticity of zero seems to me to make little sense: Manufactured stuff is useful. When the price of manufactures drops relative to the price of other stuff, we ought to buy more manufactures–not, to be sure, enough to keep the share of our incomes we spend on manufactures constant, but somewhat.

My view is that our measurements have gone substantially awry, and that the price elasticity of demand is about 1/2. The speed with which the share of manufactures in nominal GDP has declined is, in my view, much more consistent with a manufacturing price and labor productivity growth rate differential of not 1.4%/year but more like 3%/year. That would suggest that the relative price of manufactures properly measured today is not 40% but 12% of its immediate post-World War II level, and that the right “real” share graph sees not a constant “real” share of manufactured goods in output, but rather a tripling of the share.

This has implications for the “true” rate of economic growth–an aggregate-scale underestimate of 0.3%/year from this channel alone…