The quip “There are three kinds of lies: lies, damn lies, and statistics” is variously attributed to Benjamin Disraeli or Mark Twain. What should the public make of government statistics, such as the monthly employment report released today, Thursday, July 2, by the U.S. Bureau of Labor Statistics (BLS)?
There is no lying in US government statistics. But there are always commentators who will use the numbers to make whatever point they want. One should learn enough to be able to interpret the numbers for oneself. That is the only way to prevent being misled.
Of the many numbers contained in the BLS reports, I view three as especially important.
The most salient figure politically is the unemployment rate, which hit 9.5% in June, according to Thursday’s report. This was the highest level since August 1983 and clearly reflects the current extent of distress in American labor markets.
Critics of the official statistics like to point out that the unemployment rate does not capture discouraged workers who have dropped out of the labor force because they couldn’t find a job. True. But the government isn’t trying to make the unemployment number look smaller. Rather, it is just too difficult to decide who is a “discouraged worker,” as opposed to simply being out of the labor force. So the BLS always defines only those who have looked for a job recently as being in the measured labor force. This still allows us to compare changes in unemployment over time, which is the purpose of the unemployment rate. The agency does compute a measure that attempts to include discouraged workers and part-time workers — the U-6 series — but I don’t think it is right to call this the “real unemployment rate.”
The second important number in the labor market reports is employment, that is, the number of workers who have jobs, which was down another 467,000 in June. This is the statistic to which the financial markets and macroeconomic forecasters pay the most attention on a monthly basis. (In that sense, the question of discouraged workers is a red herring.) Employment peaked in December 2007, the start of the recession. Since then, we have lost 6 million jobs altogether. The current recession is now both the longest-lasting and the deepest since the 1930s. But at least the period of the steepest rate of job loss – November 2008 to March 2009 – appears to be behind us.
Two details about the jobs number. First, the statisticians get the “employment” number through one method, by surveying establishments (employers), while the unemployment rate uses a measure of employment derived through a different method, by surveying households. The employment number is generally considered more reliable because it is based on a wider survey — another reason to prefer it.
The second point is that, for purposes of comparison across different business cycles, we still need to divide employment by something. If not the labor force, then what? We must, at a minimum, allow for population growth. So it is useful to divide employment by total population. This way we don’t have to attempt distinctions about which Americans might be prepared to take a job under the right circumstances. The fraction of the population (civilian non-institutional) with jobs peaked at the end of the Clinton Administration, exceeding 64% in January 2001. It has now declined below 55%.
Although the financial markets pay most attention to the number of workers with jobs, employment is not much good for forecasting the overall economy, because it tends to be a lagging indicator. Even when firms see economic activity starting to pick up, they delay hiring, because it is costly to find, hire, and train new workers – not to mention to fire them again if the recovery turns out abortive.
For this reason, the third indicator is my personal favorite for gauging the business cycle in real time: the rate of change of total hours worked in the economy. Total hours worked is equal to the total number of workers employed, multiplied by the length of the workweek for the average worker. The length of the workweek can be expected to respond at turning points faster than does the number of jobs. When demand is slowing, firms tend to cut back on overtime, and then switch to part-time workers or in some cases cut workers back to partial workweeks, before they lay them off. The phenomenon is called “labor hoarding.” Conversely, when demand begins to rise, firms tend to increase the workweek, before they hire new workers. (To take two historical examples, the “change in total hours worked” improved in both April 1991 and November 2001, which on other grounds were eventually declared to mark the ends of their respective recessions.)
The workweek reached a historically short level in June: 33.0 hours. Not a good sign. As one consequence, total hours worked fell 0.8% that month, continuing the same rapid deterioration we have seen since last September, the month when Lehman Brothers failed and the recession worsened sharply.
The bottom line for the economy: despite signs in other areas that the recession is leveling out – most importantly, production and sales — the labor market indicators in themselves are not yet signaling a turning point. Thus the June numbers confirm the evaluation I made a month ago, based on hours worked in May, that the apparent good news in the widely reported May employment number was probably an insignificant blip. The bottom line for newspaper readers: master your statistics, so that they can’t master you.
Originally published at Jeff Frankel’s Weblog and reproduced here with the author’s permission.