March 24, 2009

Where Did the Steelworkers Go?

Attention conservation notice: back-of-the-envelope calculations about why the US has only about a fifth as many steelworkers now as it did in 1960. Not backed by any actual knowledge of the steel industry. Utterly untimely, it was, I think, prompted by a comment thread on Unfogged, but so long ago I can't remember which.

In 1960, US primary steel production was 91 million tons, of which 2.95 million tons were exported; it also imported 3.24 million tons. This part of the industry employed 530,000 people in all capacities, for an annual output of 170 tons/employee.

In 2007, US primary steel production was 98.1 million tons, with exports of 10.1 million tons and imports of 30.2 million tons. Employment was only 97,540 people, coming to 1005 tons/employee.

Exports and imports in 1960 were a wash, nearly enough, so let's suppose trade patterns had remained comparable and say that all of the net imports were to be made up by higher domestic production: (20.1 million tons)/(1005 tons/worker) = 20,000 extra workers. This would be a substantial increase, but it would still leave employment in steel at only 22% of its 1960 level. Where did the other four-fifths of the industry go?

The most obvious explanation is productivity. The industry in 2007 produced more than it did in 1960, with many fewer employees. In fact, output per employee grew 5.9 times over that period. A six-fold increase in productivity divided by a slight rise in total demand equals a roughly five-fold fall in employment.

Now, this calculation understates the effect of trade because it only considers net imports of steel. But steel is used as an input to producing many other things, and a washing machine made of steel shows up in this sort of official statistic as an import of a manufactured good, not an import of steel. So to really see what US steel production would be if we retained 1960 trade patterns, we'd need to see what the change in the (foreign*) steel content of US net imports has been. Since I don't have Leontief input-output matrices for the US and its trading partners in the two years, I can't do this.

Failing actual knowledge, I'll turn to guesswork. Suppose the steel content of imports was equal to net direct imports; this seems high, but what do I know? This would just add another 20,000 jobs, and bring us up to 26% of the size of the industry in 1960. To get the same level of employment in steel production now as in 1960, the net increase in the foreign steel content of our imports would have to satisfy

(530,000 workers) - (117,540 workers for domestic production and direct imports) = (increase in net indirect imports)/(1005 tons/worker)
or 414,522,300 tons, i.e., about 3.5 times total production plus net direct imports. This is highly implausible.

I conclude that domestic employment in steel production has collapsed largely because increases in productivity have not been matched by increases in demand. If someone can point out where this reasoning goes wrong, I'd appreciate it.

*: Foreign steel content, because if the washing machine is made abroad of steel exported by the US, replacing that washing machine with a US-made one will not increase the demand for American steel.

Sources: 2007 employment figure from BLS (NAICS code 3311). 1960 employment figure from Table 1 on p. 2 of Lebergott. (It does not, however, appear to be affected by some of the well-known problems with Lebergott's series for the 1930s.) Annual production, import and export figures from USGS.

Manual trackback: The Inverse Square Blog; Nothing Funny About Feldspar (with more facts; go read)

The Dismal Science

Posted by crshalizi at March 24, 2009 10:49 | permanent link

Three-Toed Sloth:   Hosted, but not endorsed, by the Center for the Study of Complex Systems