It’s the job numbers, stupid

Readers totally uninterested in labour force figures should proceed no further. Me? Well, I love them. They really tell us what is going on. National accounts, retail sales figures, and so on are usually fudgy or partial and, in any event, are at least one step removed from the main game, which is jobs.

If everyone who wants a job can find a job, then we have the economics version of nirvana. It can’t get much better than that. Governments can crow all they like about this or that economic statistic but if people aren’t finding jobs they won’t buy it.

The only problem for the dedicated observer is that labour-force figures are collected by sample survey, so individual monthly figures can’t be relied upon too much. You wouldn’t know this if you paid attention to economists. They try to pick a specific number in advance and then comment on the actual result as though it precisely reflected reality. This gives the whole monthly exercise the appearance of an accuracy it simply doesn’t have.

It is much better to even out the figures over three to six months to detect a trend. Then the figures become a much more reliable guide to the state of play in the economy. Politically portentous labour force figures were released late last week in the in the United States.

One of Mitt Romney’s recent slogans has been 43 months with unemployment over 8 per cent. He was fairly confident of being able to say 44 months after the September figures were released. Sadly for Romney, and taking some (albeit minor) gloss from his debate victory, the Bureau of Labor Statistics (BLS) claimed that unemployment in September had fallen from 8.1 per cent to 7.8 per cent.

The BLS also reported that non-farm employment had grown by (only) 114,000 jobs. The two figures don’t add up. Employment growth was less than the growth in the available labour force and fairly anaemic; therefore, if anything, unemployment should have edged up. Conservative commentators were apoplectic.

Jack Welch ex-CEO of General Electric suggested that the books had been cooked. Puzzlement and disgruntlement abounded on the right; and, after the great debate debacle, gasps of relief on the left.

The key in the US is that two surveys are untaken by the BLS. The more reliable one, just showing non-farm employment, samples business and government establishments. The less reliable one, out of which the rate of unemployment falls, samples households. Amazingly, this latter survey claimed that employment grew in the month by 837,000.

Even accounting for the broader coverage of the household survey (which includes agricultural employment and working from home), the disparity between the reported increases in employment in the two surveys is undoubtedly down to sampling error (unless, of course, Welch is right). One way to test the figures is to iron them out by comparing employment growth in the two surveys over, say, sixth months to September.

The establishment survey shows growth in employment of 637,000 over that period. The comparable figure for the household survey is 940,000. Clearly the household survey understated employment growth in the run up to September, but then much more than made up for this with a grossly inflated September result.

Mind you, even with the inflated September number unemployment would have remained above 8 per cent if labour force participation had not fallen over the six month period. This all worked out well for Obama. But, as the next release is due on November 2, four days before the election, it might not work out too well if, as is likely and usual in these kinds of surveys, untoward individual monthly figures revert to norm (unless, as I’ve said, Welch is right and there’s a plot afoot in the bowels of the BLS to re-elect Obama).

For us fans of labour force figures there is more grist for the mill this week when the September figures are released for Australia. Something to look forward to I think you will agree, because only fellow fans would have had the interest and strength to read this far.