In spite of last week’s better-than-expected jobs report, many intelligent observers seem convinced that the unemployment rate, which fell nominally to 9.4 percent last month, will nevertheless almost certainly hit 10 percent or higher before the situation really rights itself. Here’s the New Republic’s Noam Schieber making the case. Here’s the White House. Here’s Intrade, where you can make about 2:1 on your money if the unemployment rate is below 10 percent in December.
Although it’s foolish to try to predict economic indicators, I think these observers are probably wrong: the recession likely does not have enough gas left to get us to 10 percent unemployment.
Principally, I think these analysts may simply be underestimating how much impact last week’s numbers should have on our expectations for going forward. After four straight months in which the rate of job losses had declined, the number jumped abruptly upward in June, accelerating to what was then reported as a net of 467,000 lost jobs. It wasn’t clear whether the jobs situation was still improving fairly rapidly, was improving but only slowly, or in fact the momentum was entirely sideways or was even getting worse:
After getting the July numbers, however – which included downward revisions to the reported job losses for May and June – the situation would now seem to be much less ambiguous: the jobs picture is steadily brightening, and within a couple of months, the economy will in all likelihood begin to actually create jobs:
In order for unemployment to hit 10 percent, a net of roughly 1 million more people would need to become unemployed, assuming no change in the size of the labor force (which is a big assumption and one we’ll examine in a moment). This almost cetainly won’t happen. Last month, 247,000 jobs were lost according to the payroll survey, and 155,000 more people became unemployed according to the household survey. (What’s the difference between these two numbers? We’ll discuss that too in a bit.) Given that the numbers are improving, it’s hard to see how you can squeeze another million or so job losses of 150-300K per month -– you’d need the employment picture to completely flatline for another 4-5 months, or for what now seems to be a fairly robust trend to actually reverse itself.
Note, however, what I stipulated earlier: assuming no change in the size of the labor force. Ordinarily, about 125,000 additional people each month enter the economy. So it’s not enough merely to break even on the job creation numbers; you have to be slightly into positive territory to avoid seeing the unemployment rate go up as a result of these new job-seekers.
In July, however – as well as in June – the size of the labor force actually declined, according to the household survey. This is what caused the unemployment rate to decline: although the numerator (the number of people who have jobs) in fact slightly decreased, the denominator (the size of the labor force) decreased more, causing the metric to improve. Some of this is because of so-called discouraged workers – people who do not have a job but have given up on looking for one, and (for reasons that I do not entirely agree with) do not meet the Department of Labor’s definition of being “unemployed”. This discouraged workers story has been a bit oversold though: U-4, the measure of unemployment that includes discouraged workers, went down too. Most of this, rather, is a matter of people finding somewhat longer-term alternatives to seeking employment: going back to school, early retirement, joining the Army or the Peace Corps, etc. In addition, both legal and illegal immigration are declining, which takes some pressure off the growth of the labor force. There may also be some reversion to the mean, since these metrics are subject to measurement error and since the labor force had supposedly increased by higher-than-usual amounts in April and May.
What observers like Scheiber are worried about is that as the economy shows some signs of life, the labor force will begin to increase again, perhaps even at a greater-than-normal rate, as discouraged workers have their morale boosted and go out and look for jobs. I agree that the labor force is more likely to return to growth than to continue declining – however, I think these concerns are overstated. If we look back at past recessions, we find that the labor force grew by an average of 0.44 percent in the six months after they were declared over. This compares to an average of 0.77 percent growth for all six-months periods since 1948:
People do not jump right back into the labor force the moment a recession is over. Oftentimes, indeed, they can’t, because they’ve made somewhat long-term commitments – good luck ditching the army because the local bank is having a hiring fair back at home in Topeka. These effects are fairly strongly lagged, probably by at least 3-9 months, and usually occur only once the jobs picture has gotten to the point where it’s actually pretty darn good – not just when it’s merely improving. Where we’ll see these effects is in, say, January of 2011, when the employment rate might not budge much even if a couple hundred thousand new jobs are created. But the unemployment rate should already be safely clear of 10 percent long before that.
If the labor force grows at its typical post-recessionary rate of 0.44 percent over the next six months, that means that the economy will need to avoid losing 375,000 more jobs between now and January for unemployment to stay below 10 percent. While there will almost certainly be some additional job losses this month and probably in September, I’d take about even odds on the economy gaining or losing jobs in October and November, and expect it to probably be creating jobs in December and January. In other words, I think it will win the race against the growth of the labor force. The most likely scenario is that the unemployment rate will flirt with 10.0 percent but not quite reach it.
I’ve also built a simple model that attempts to predict what the unemployment rate will look like based on lags of two variables: the unemployment rate from the household survey, and the net payrolls number from the establishment survey. These numbers measure two slightly different things: the household survey measures the number of employed people (as well as the size of the labor force) whereas the establishment survey measures the number of jobs. Someone who held down multiple jobs would theoretically be counted twice in the establishment survey but only once in the household survey. In addition, the household survey accounts for some types of employment (farm work, self-employment) that are not reflected in the establishment survey.
More to the point, though, these surveys are just that – surveys – and therefore they’re subject to the usual types of distortions and measurement errors. The establishment survey tends to be more reliable than the household survey – if they’re giving you contradictory information, that’s the one you should trust. But both provide some useful, non-redundant information.
With this admittedly rather simple model I show the unemployment rate rising slightly, but peaking at 9.6 percent in October and November. I then show it beginning a steady, and actually fairly robust decline, with about 400,000 jobs being created each month by next summer.
Don’t take this that seriously – although you can say the same thing of almost any economic forecast — but what’s interesting is how different the forecast would look if we had run it based on June’s numbers. That would have showed an unemployment rate that hit 10.0 percent in September and remained there until February, peaking at 10.2 percent around Thanksgiving. The seemingly small amount by which last month’s numbers beat expectations – the economy did between 75,000 and 200,000 jobs better than anticipated, depending on whether you look at the establishment or household numbers – makes a rather large difference in terms of what we might anticipate for the next 18-24 months. This could have some fairly huge political implications too. The September 2010 unemployment rate, which is the last one voters will see before they go to the polls for the midterms, would have been projected at 8.6 percent based on last month’s data; it’s now projected to be 7.9 percent.
(By the way, this model is decidedly more optimistic about the velocity of the jobs recovery than are most mainstream observers . I wouldn’t necessarily believe in the prospect of a more rapid, robust recovery in employment simply because this silly little model says so – but I also think the consensus forecasts are unduly pessimistic, for reasons I’ll explain in another article.)
Overall, the conventional wisdom seems to be that the odds are about 2:1 in favor of unemployment hitting 10.0 percent in at least one month between now and next spring. While that’s certainly possible, I’d posit that the actual odds are more like the reverse – 2:1 against double-digit unemployment. That doesn’t mean that I’d run out and buy stocks, which are already trading at a price that would imply they’ve fully priced in something between a U-shaped and V-shaped recovery, or that I’d cancel my plans to go back to school to go out and find work. But I think we’ve grown so accustomed to bad news that we’ve forgotten how to recognize good news when we see it; leading indicators have been turning upward for months now, both in the United States and in other countries, and now we finally have a jobs report that reflects that optimism. A lot of folks think I’m making too much of Friday’s employment numbers; I’ll take 2:1 that they’re making too little of it.