, , , , , , , , ,

I’ve been so preoccupied lately with reporting on the Federal Reserve’s manufacturing production figures each month that I’ve forgotten to point out a big data problem that they’ve been suffering for a long time, and that still distort output levels – to the upside.

The problem: the Fed figures, which are inflation-adjusted, overstate output in the information technology hardware (computers, semiconductors, telecoms equipment, and the like). And these high tech products for many years have accounted for all the real growth registered by U.S. manufacturing.

Just one figure is needed to show how important they are. Since the December, 2007 beginning of the last recession (which was a humdinger for manufacturing), overall after-inflation manufacturing production is up by 3.43 percent. But if you take away the high tech hardware sector, it’s down by 1.34 percent. And this means that all those manufacturing cheerleaders who keep claiming that the sector’s output has never been higher are peddling fakeonomics.

What exactly is the problem with the infotech hardware data? It has to do with strong deflationary trends in these products. When prices are rising (inflation), and we want to know what output levels are independent of these price increases (which according to most economists provides the most accurate picture of growth), the adjustment process brings the output level down. That is, the inflation-adjusted output level is lower than the pre-inflation level. But when prices are falling (deflation), and we want to strip out price effects, the adjustment process brings the output level up. That is, the deflation-adjusted output level is higher than the pre-deflation level.

Now it starts getting tricky, but if you want to understand the real state of domestic U.S. manufacturing, ya gotta bear down! If inflation is actually higher than the official estimates, then the actual inflation (price)-adjusted output levels published by official sources are going to be too high. That’s because by definition there’s inflation that isn’t getting recognized and plugged into the inflation-adjusted data. But if deflation is actually stronger than the official estimates, then the same price-adjusted output levels are going to be too low.

And here it gets tricker still! Some very careful studies claim that prices for high tech hardware products are falling faster than the official score-keepers can record them. But doesn’t that then mean that the price-adjusted, real U.S. output figures that are published for these sectors are too low? No. And the reason has to do with something we all know about these products but that we all seem to forget when judging manufacturing output: They’re largely made from parts and components produced overseas. That is, they’re made up of lots of what economists call “imported inputs.”

And it’s the prices of imported inputs – mainly from East Asia, which produces so many manufactures to the point of glut – that have been falling fastest of all. And if the prices of those foreign-made parts and components are falling especially fast, then it means that their price-adjusted share of the final “Made in America” high tech products is larger than the official data recognize. And therefore the price-adjusted share of the American-made content of these products – which is what the Fed figures measure – must be lower. Therefore, since the infotech products have monopolized recent manufacturing growth – and then some – overall American manufacturing output is considerably lower than the officially reported levels.

To get incredibly (but importantly) technical about the issue, the exact data affected by these pricing problems are not the production data published by the Federal Reserve, but a data category that’s called “real value-added.” I used the Fed figures because the real value-added numbers put out by the federal government aren’t detailed enough to show what share is accounted for by high tech goods.

At the same time, the feds do recognize the problem, as indicated by a statement here from a senior official from the Bureau of Labor Statistics (which tracks price trends for the U.S. government). Don’t, however, expect a more accurate picture of manufacturing any time soon, for Congress has so far refused to fund the data gathering effort requested by BLS. Because its estimated cost is all of $11 million.

FYI, if you’d like to read more from the economists who have done the pioneering work on this issue, click on this link.  It takes you to the papers written for a recent conference on the subject. I attended and it was absolutely terrific.