In his email to you he discussed Manufacturing's share of the U.S. economy, as measured by real Gross Domestic Product (GDP*)
That I'm aware of, there is only the table I've cited that reports Real GDP for manufacturing as a component of total GDP. That narrows the tables he can rely on for discussing manufacturing's contributions to GDP to the two tables I've already cited (and you noted).
These numbers of Reynold's did span only periods close to the base.
No they don't or there wouldn't be differences of 27% decline since 1988 versus 5% growth since 1992. And even using numbers close to the base is still invalid - the error is just smaller.
He disputes that "Manufacturing's share of real gross domestic product... has dropped to between 16 and 17 percent and disputes the validity of a 27% decline since 1988 (alluded to in his reference to Similar statistical exercises recently led to an interesting debate between my old friends Bruce Bartlett and Paul Craig Roberts.).
To support his view, Reynolds would have to go back to 1988 as Bartlett and Roberts did. Further, he (or whomever) had to go back 1992 to support a claim of it actually grew by 5.3 percent a year from 1992 through 2000. Unless, of course, he was just dismissing someone elses math without checking it against his own.
Additionally, he wasn't adding or doing ratios vertically,
He (or whomever) is relying on the implicit assumption that the column components vertically total to 100% at the top - else computing a lower line item's percentage of the total at the column top has no meaning.
However, without Reynolds showing his math, your guess is no better than mine as to what he (or whomever) did. We only have his allusions to Manufacturing's share of Real GDP. Regardless, computing 1988, 1992, or 2000 percentages (to either affirm or refute a viewpoint) from Manufacturing's Real GDP chained dollar data as did Bartlett (whom Reynolds references and holds similar views) is invalid.
I’ve finally got a handle on this. Current Dollar (nominal) and Chained weighted (“real”) GDP estimates are measuring two distinctly different things. Current Dollar is measuring the economy’s components over time by their value/price, and Chained is measuring them by their units of serviced or units constructed (loosely speaking)
The wild variance over such a short time that you site between the two (-27% verses 5%) isn’t a product of inaccuracy in the chained weighted method as you say. (It would be utterly worthless if so.) The inaccuracy in chained weighting is very small over that 8 year time period, something like 1%. It’s broken out by the BEA in the GDP table we’ve been discussing under “Not Allocated by Industry”
|
1992 |
1993 |
1994 |
1995 |
1996 |
1997 |
1998 |
1999 |
2000 |
GDP |
6880.0 |
7062.6 |
7347.7 |
7543.8 |
7813.2 |
8159.5 |
8508.9 |
8859.0 |
9191.4 |
Not Allocated by Industry |
-59.3 |
-28.3 |
-2.2 |
-2.2 |
0.0 |
-33.3 |
-48.9 |
-97.1 |
-159.1 |
|
-0.9% |
-0.4% |
0.0% |
0.0% |
0.0% |
-0.4% |
-0.6% |
-1.1% |
-1.7% |
Current Dollar simply measures the output price of the components/industries of the GDP each year (adjusting for inflation). What each industry actually produced is irrelevant. Manufacturing could have doubled the number of widgets produced last year, but if they generated the same income as the previous year no change is reported.
Chained Weighted is a kind of cross between Current Dollar and an earlier method called “Fixed Weighted”. Fixed weighted simply measured the output of things produced or performed by each industry for a time period and multiplied it by their price to get the total. Subsequent periods simply re-measured what was produced and applied the same fixed price to get a new period total. Chained weighting does the same, but makes a price adjustment for inflation across all units each period.
The inflation adjustment in Chained Weighting yields discrepancies that grow larger over time, but that’s not the primary source of divergence between Chain Weighted and Current Dollar GDP figures. Again, the difference is what’s being measured. Simply speaking, it’s the difference between measuring stuff done over time and the value of stuff done over time.