Friday, November 02, 2007

U.S. Economic Numbers Much Weaker Than They Seem

The last few days we have seen reports about the U.S. economy which at first glance have appeared strong-or at least not weak enough to indicate recession. Yet if you look more closely they really are much weaker than they appear-

1. The GDP report which published GDP growth of 3.9% in the American way of expressing growth. Even setting aside the good reasons for believing the GDP price deflators underestimate inflation -and so overestimate real growth- it is still far less impressing than it would seem. First of all, GDP numbers are generally revised down. 2004 GDP growth was for example first published as 4.4%. Now after having been revised down three times in the annual revisions of 2005, 2006 and 2007-it is estimated to be 3.6%. Secondly, the domestic price deflator rose 1.6% versus the 0.8% increase for the GDP deflator (If you're not sure why estimating real income gains on the basis of the prices you pay rather than the prices you receive see here). The likely revision and terms of trade factors together thus indicate a real number 1.6% lower than the headline figure, leaving us with 2.3% in growth.

Moreover, some of the growth was related to increased government demand for goods and services (note this does not include transfer payments like social security or welfare) which rose from 19.4% of GDP to 19.5%-the highest since early 1993.Adjusting for that, private sector growth were more like 1.8%. That is not a recession, but is far from the boom numbers the headline numbers suggest.

Of course, it was not to be expected that the third quarter would be the first of the recession. But it seems increasingly likely that the fourth quarter will have negative growth, as the two reports analyzed below-and many other indicators- suggest.

2.The ISM Manufacturing report fell back to 50.9 in October from 52 in September. That indicates weak growth, but not a contraction. However, this headline number is calculated by weighing together different sub-components, such as new orders, production, employment and prices paid. Guess which sub-component had the highest value and which increased significantly in October? Well, if regular readers remember how many times I've been mentioning stagflation, they should be able to guess that it is the prices paid index. And if they guessed that, they guessed right, as the prices paid index rose from 59 to 63. Excluding that, the overall index would be below the 50 level which is the level which is the border line between expansion and contraction.

3. The employment report were widely interpreted as indicating a strong job market. The main basis for this claim is that the number that financial journalists always focus on, the increase in payrolls in the payroll survey was a full 166,000.

However, first of all, with the mere 0.2% increase in nominal wages, it seems almost certain that real wages fell significantly given the surge in food and energy prices.

Moreover, there are good reasons to believe that this job growth number is extremely misleading and that job growth was in fact negative. The household survey for example indicated a 250,000 job loss.

The payroll survey is widely considered more reliable than the household survey. This is however only to some extent true with regards to monthly changes which have been somewhat erratic in the household survey. However, there are good reasons to believe that the household survey better reflects trend movements in employment. During the housing bubble, employment growth according to the payroll survey was suspiciously
low, with employment increasing by a mere 4.2 million or 3.2% in the four years between December 2001 and December 2005. And the initially reported number was actually even lower than that as the payroll survey growth for that period has been upwardly revised several times.

The household survey's increase of 6.8 million or 5% looks a lot more reasonable.

Now the tables are turned, with the 12 month increase in the household survey being 0.5% versus 1.3% in the payroll survey.

The reason why the household survey is more reliable in tracking trend changes in employment, while the payroll survey tend to underestimate growth during booms and overestimate in during downturns is that the household survey automatically tracks changes in for example in employment of illegal immigrants and start-ups of new businesses or business deaths. The payroll survey is useless in covering either. In order to account for employment in new businesses, the Bureau of Labor Statistics has a "birth-death" model of new businesses. The trouble with that is that it is being applied without any consideration of cyclical fluctuations. The result is that most of the payroll survey growth is now a result of the assumption of new businesses being created, whereas during the housing bubble only a small part of payroll survey growth was a result of this. The absurdity of this is most apparent in the construction sector, where the Bureau of Labor Statistics would have us believe that 164,000 more jobs were created in new businesses than lost through bankruptcies in the last seven months, despite the housing bust.

The payroll survey is thus simply not credible as an indicator of cyclical fluctuations. And if we look at the household survey the bottom line is that employment was 133,000 or 0.1% lower in October than in June. This implies a direct contraction in the labor market.


Post a Comment

<< Home