When government officials asked people if they had a job last month, 137.6 million said "yes." But when employers were asked, they said they had only 129.8 million on nonfarm payrolls.
There are several reasons why the number of people on business payrolls is bound to undercount the number of workers. If more people are working at home as self-employed consultants, or working through temp agencies, they would not show up as payroll employees. And "nonfarm payrolls" ignores the fact that agriculture added 155,000 workers in August. What is nonetheless quite remarkable is that these two measures of employment are now much further apart than they were back in early 2001.
Experts decided the recession started in March 2001, two months after President Bush was sworn in, although stocks had by then been falling for a year and industrial production for seven months. According to the survey of households at that time, there were 137.7 million employed -- virtually the same as now. Yet the payroll survey then counted 132.5 million jobs -- 2.7 million more than now.
Depending entirely on which measure you choose, we have either recovered all the jobs lost during the recession or lost 2.7 million. Reporters who relish bad news and bad politics invariably tout the latter figure. Washington Post reporter Jonathan Weisman wrote hysterically of "the longest hiring downturn since the Great Depression" -- a patently absurd comparison. California Gov. Grey Davis claimed "no president since Herbert Hoover has seen job losses like this." In reality, today's 6.1 percent unemployment rate is the same as it was in 1994 or 1987 or 1978 -- years in which nobody pretended to see any similarities with the Great Depression.
The reason such a moderate rate of unemployment provokes such immoderate commentary is, of course, the looming presidential election. Yet to the debatable extent that employment might affect next year's election, it is the household survey rather than the payroll survey that surely matters. If 137.6 million people say they have jobs, what difference could it possibly make if the payroll survey implies that a few million of them are somehow mistaken about that?
Besides, it is the survey of households (not of businesses) that is used to calculate the unemployment rate. And an unemployment rate near 6 percent is much too low to influence many votes. For one thing, unemployment is highly concentrated among younger and less-educated people -- two groups with a very low voter turnout. Among married men and women, the unemployment rate is only 3.8 percent. Besides, most of those unemployed in one month have found jobs after two or three more months. The median duration of unemployment dropped from 12.3 weeks in June to 9.6 in August.
White House political advisors take their opponents' absurd comparisons with the Great Depression much too seriously. Economists have conducted numerous statistical studies regarding the impact of economic conditions on presidential elections, finding little or no importance in the level of unemployment.
Yale econometrician Ray Fair, for example, found "average unemployment rates ... were not significant." What is significant, in Fair's model, is the number of quarters before the election in which economic growth exceeds 2.9 percent. Economic growth has already exceeded 2.9 percent three times in the past two years and will again in the current quarter. Each time that happens, it adds about a percentage point to Bush's share of the vote, according to Fair's model. And the election is still four quarters away.
One study by Richard Gleisner, and another by Stephen Hayes and Joe Stone, also found election results depended in part on the performance of the stock market. Rising wealth and income for 94 percent of the population apparently carries more weight than temporary job loss among 6 percent.
Douglas Hibbs found that most presidential results can be explained by growth of real, after-income per person. That figure depends on economic growth, but also on low inflation and low taxes. Real disposable income per capita increased 5.1 percent from the end of 2001 to the second quarter of this year. But that was before the midyear tax cut. In Hibbs' model, the recent tax cuts were good politics as well as good economics. The only time rising after-tax income failed to predict the winner in presidential races, according to Hibbs, were during unpopular wars in 1952 and 1968. Iraq may yet prove more problematic for Bush than the economy.
When they are not trying to make a political issue out of cyclical unemployment, some of the president's rivals hope to make an issue out of differences in income. That issue was investigated by Harvard's Alberto Alesina and co-authors Rafael Di Tella and Robert MacCulloch. They examined 128,106 survey answers for Europe and the United States. Although the surveys showed that many Europeans really do fret about "inequality," it turns out that in the United States, this is an issue that matters "only for a sub-group of rich leftists." President Bush is therefore likely to lose Barbra Streisand's vote over inequality (not to mention her campaign contributions), but that is unlikely to matter much.
When it comes to predicting presidential elections, the pace of economic growth clearly matters. The stock market matters. Inflation matters. Local economic conditions matter. Incumbency matters. War matters. But statistics on payroll employment and income inequality matter only to guilt-ridden multimillionaires, partisan journalists and political speechwriters.