It is critical to use real data to make a valid comparison because prices for many goods such as computers have fallen sharply. Since GDP data are calculated in money rather than volume terms, failing to take account of this fact would give a distorted picture of what is going on. For example, suppose output of some product rose by 10 percent in terms of units, while falling 10 percent in price, due to higher productivity. Using the nominal data would make it appear as if there had been no increase in output. Using real data captures the increase.
Mr. Bartlett has likely made a significant, though common mistake here in his manipulation of the BEA's real data.
While he doesn't cite exactly where he got his BEA 'real data', I assume because he calls it 'real' and describes the price vs volume advantages of real data, I suspect he is using the BEA's Real Gross Domestic Product by Industry in Chained (1996) Dollars.
If so, his mistake is that aggregates in a chained dollar series are not additive, which means that the components don't add up to the total GDP, which means that each component is not represented in its proper proportion or share of the total GDP, which ultimately means Mr. Bartlett can not compute manufacturing's percent of 2001 GDP as:
16.2 ~ 16.17 = 1,490.3 (from col 2001 line 12) / 9,214.5 (from col 2001 line 1)That math, normally valid, is invalid with chained data. That's why the BEA provides tables with GDP share computed such as Gross Domestic Product by Industry in Current Dollars As a Percentage of Gross Domestic Product
Therein, are the only correct percentages that manufacturing has fallen from 17.4 percent [in 1991] of GDP to 14.1 percent [in 2001] and as also quoted from Paul Craig Roberts critique of Bruce Bartlett Trade nothink.
Lest anyone be concerned that the BEA's computations of GDP share don't adjust for inflation, they do. Inflation is the same percentage for GDP as it is for a component of GDP and so computing manufacturing share of GDP as
MFG share of GDP = (MFG component * inflation pct) / (GDP * inflation pct)is correct because the inflation percentages (whatever they are) cancel out anyway.
Bottom line - Manufacturing's share of GDP has fallen consistently each year from 19.2 percent in 1988 to 14.1 percent in 2001, a decline of 27 percent over the 14 year period is correct as Roberts stated.