The Atlanta Fed's macroblog provides commentary and analysis on economic topics including monetary policy, macroeconomic developments, inflation, labor economics, and financial issues.

Authors for macroblog are Dave Altig, John Robertson, and other Atlanta Fed economists and researchers.

« When cycles collide | Main | A look at the recovery »

July 23, 2009

Unemployment rate: Count me surprised

Brad DeLong has taken a look at the job market and is counting himself among the economists who admit that, "Well, I just got it wrong." According to DeLong:

"… the rise in the unemployment rate during a recession should be a fraction of the decline we see in GDP relative to trend. According to Okun's Law, the unexpected extra 1.2 percent decline in real GDP in 2009 should have been accompanied by a 0.5 or 0.6 percentage-point rise in the unemployment rate. Instead, we experienced a 1.5 percentage point rise in the unemployment rate. I confess this comes as a surprise to me, but it shouldn't. Because evidence has been mounting that Okun's Law is broken—especially with regard to the retention of workers in a downturn."

I share Professor DeLong's surprise at the unemployment rate's response to this recession. Though I have never had a lot of faith in Okun's Law as a predictive device, I believe DeLong may be just a little too harsh on himself (and by extension, I guess, me) for not hitting the mark on unemployment prognostications. From what we know at the moment, the unemployment/GDP correlation is going to deviate from any other postwar experience by a fair margin. As noted in today's Wall Street Journal:

"Breaking from historical patterns, the unemployment rate—currently at 9.5%— is one to 1.5 percentage points higher than would be expected under one economic rule of thumb, says Lawrence Summers, President Barack Obama's top economic adviser. Since the recession began in December 2007, the economy has lost 6.5 million jobs, 4.7% of total employment. The unemployment rate has jumped five percentage points, while the economy has contracted by roughly 2.5%."

Below is a chart that illustrates the point. It plots the peak change in the unemployment rate during recessions (which has always been the unemployment rate change from the beginning to the end of the end of the recession) against the cumulative percent loss in GDP in those recessions. (In the chart, the blue dots represent the experience in each postwar recession. The red square represents the current downturn, making the assumption that GDP growth in the second quarter will be –0.5 percent and the recession will end sometime in the third quarter. For the sake of the exercise, I pulled these figures from Macroeconomic Advisers, the forecasting group run by former Federal Reserve Gov. Larry Meyer.)


The line in the graph above represent the simple statistical estimate of the relationship between changes in the unemployment rate and the cumulative GDP loss during each recession. Using this estimated Okun's Law, you would have guessed that the unemployment rate would have risen by about 2 percentage points. In other words, the best guess for the unemployment rate would be in the neighborhood of 7 percent, not 9.5 percent.

There are a couple of caveats to this analysis, of course. One is that, as is often noted, unemployment is a lagging indicator, so the peak in the unemployment rate can come after the recession ends. This caveat changes the picture somewhat (and misaligns the unemployment and GDP data), but not by a lot.


The second caveat is that there may eventually be revisions to GDP that make the recession look deeper than it appears at the moment, which would move the current episode closer to historic norms. On the other hand, the charts above assume that the unemployment rate will peak at 9.5 percent, which is not a certainty at the moment. (The "central tendency" projections published by the Federal Open Market Committee suggest that the rate will peak in the 9.9 to 10 percent range.)

Setting aside the possibility of any substantial revision in the data, perhaps one of the questions in the end will be whether the National Bureau of Economic Research Business Cycle Dating Committee was somewhat overaggressive in choosing December 2007 as the beginning of the recession. Though currently measured GDP growth was negative (barely) in the fourth quarter of 2007, GDP did not turn persistently negative until the third quarter of 2008. If we were to assume that the business cycle peak was actually in the second quarter of 2008, the picture would look like this:


With this alternative timing for the recession, the Okun's Law miss on the unemployment rate projection would have still been to the downside, but the error is quite a bit less dramatic than you get with the official recession dating.

In any event, I'm quite sympathetic to DeLong's theme that the dynamics of U.S. labor markets coming out of recessions appear to have changed starting with the 1990–91 economic contraction. And it might be hard for many people to argue with DeLong's point that the U.S. economy is likely headed toward another so-called "jobless recovery." But until more facts are in and we're able to look back on what transpired, I think we still, at this point, must reasonably count the current run-up in the unemployment rate as a puzzle.

Update: Casey Mulligan (University of Chicago) writes:

There are a host of public policies that discourage the earning of income, and do so more than they did before the recession. IMO, theat's why Okun's law is broken.

More at his blog.

By David Altig, senior vice president and research director at the Atlanta Fed

July 23, 2009 in Business Cycles , Data Releases , Labor Markets | Permalink


TrackBack URL for this entry:

Listed below are links to blogs that reference Unemployment rate: Count me surprised :


But even when the economy was growing and times were good, business was laying people off left and right. Places like IBM and others would shutdown 5k worth of 'skilled' labor almost overnight.

It seems like steady destruction is the name of the game for many people in this new service economy.

Posted by: Former SS Resident | July 23, 2009 at 07:22 PM

Could simply mean that in the future there will be big downward revisions of gdp data...

Posted by: Daniil | July 24, 2009 at 01:54 AM

Of course you are surprised. Economists have been cheering the sending of American jobs overseas for a couple of decades now. It was supposed to make things better, for the Chinese and other poor folks. It also made Americans poor and unemployed. Oops.

America: no functioning job market, health care market, or education system. America: too corrupt to fix serious problem. America: the fattest 'no-can-do' country in the world.

Posted by: lark | July 24, 2009 at 02:52 AM

In each of the graphs, we can find at least one earlier recession - 1980 - in which the dot is as far from the regression line as in this case. In one graph, there appear to be 2 misses the size of the one we are working on now, with 1990 tossed in with 1980. In those cases, the jobless rate rose too little rather than too much, but it does give the impression that the fit isn't as tight as we are asking it to be. If we insist on one direction - the jobless rate too high - then we have a unique case on our hands. If we allow for the jobless rate being too high or too low, then we simply have one of the more extreme cases.

Posted by: kharris | July 24, 2009 at 08:54 AM

Another "DUH" moment!

Most GDP growth is going to the top spectrum of the economic ladder. That is why a jobless recovery...

Posted by: dr | July 24, 2009 at 12:58 PM

Mulligan is right that there is a disincentive to make earned income relative to investment income. His suggestion that this is new within the time period in question, is dubious, however.

Marginal tax rates on earned income were much higher prior to the reforms in 1986, and there were still ample incentives for investment income (particularly capital gains) relative to earned income at that time. The tax rates themselves also understate that preference because there were, believe it or not, far more loopholes (like passive loss tax shelters) available then than there were immediately following the reform. Estate tax rates have changes a lot during the relative time period as well, but are quite neutral between earned income and unearned income.

While the preference for unearned income increased during the George W. Bush Administration (i.e. after the 1990 and 2001 downturns and before the current one), with reduced rates on dividends from C corporations, reductions in top capital gains tax rates, increased importance for stock options and carried interest arrangements, etc. This isn't the whole story.

The portion of the population that is in the income taxed part of the workforce has greatly decreased, and the overall economic impact of income taxes on the working class and middle class has declined dramatically, starting with the reforms of 1986 and continuing since then. This sentiment is reflected in greatly reduced priority for the issue of lowering taxes in opinion polls.

Unemployment, historically, had a far disporportionate impact on members of these social classes. Less educated and less skilled and less experienced members of the workforce are usually hit the hardest, by a large margin, by unemployment. Yet, these are the people most indifferent to the earned-unearned income disparity. The only place where income taxes create much of an incentive is right on the threshold of the line between poor and working class, where the phase out of the earned income tax credit, FICA taxation, and the phase out of eligibility for government benefits like Medicaid, Free/Reduced School Lunch, rent assistance, heating assistance and a host of other programs conspire to reduce a whole host of government tax and non-tax benefits with the highest marginal tax rate of any group. Just above that phaseout range, one hits a point of near zero average taxation, near zero government means tested benefits fairly low marginal taxation rates of earnings by historical standards, and abundant opportuntities to use contribution limited tax preferenced vehicles like IRAs, education savings accounts, HSAs, residential real estate tax preferences, etc. to gain tax benefits that favor investment income, a trend that continues all the way up the line in ever evolving forms.

One plausible alternative explanation is that a lot of the economically important GDP growth before the financial crisis hit, and hence a lot of economically important GDP decline afterwards, never made it onto the GDP books. What is missing from the GDP books is unrealized perceived gains and declines in asset value. The consumption growth we saw during the preceding boom was driven by wealth effects from "on paper" appreciation in financial and real estate assets that was only partially monitized with debt driven spending and cash out transactions. People thought that this wealth was real anyway -- real estate appraisals were often underestimates of real market prices just months after they were made, and financial valuation data was available in real time on the Internet or CNN and could be converted to cash at a moment's notice (or at most, a few day's notice).

If you were to look at decline in apparent GDP, including estimates of housing stock valuation and financial market valuation that weren't reflected in actual transactions at those prices, I think that you would see that the GDP decline in this recession has been deeper than GDP declines in previous recessions, relative to the conventional GDP number.

Employer decisions to lay off workers in this recession have frequently been driven by a desire to control costs due to weak expectations for future growth (something that also explains the rising share in this financial crisis of the ranks of the involuntary part-time work force which can be shed "just in time," with hour reductions, more easily). For employers making these forward looking decisions, who are the group most likely to have experienced on paper wealth losses in the financial crisis, perceptions of economic decline are more important than actual economic decline. So, if paper losses make GDP decline seem much greater than it actually is (the losses too are often unrealized due to the discipline of financial planners who tell people to stay in the market so they don't miss the boom that follows), employers are likely to reduce their workforces and unemployment is likely to be greater.

Just a heuristic explanation, but a more plausible one than the earned-unearned income distinction (particulalry because the pre-tax return on unearned income is very low and the impact of low pre-tax returns swamps any tax preferences).

Posted by: ohwilleke | July 24, 2009 at 03:15 PM

Just as a quick footnote, I confirmed that GDP definitions do expressly and intentionally exclude unrealized capital gains in the secondary financial and housing markets (i.e. sales involving non-IPO financial instruments and sales of homes to someone other than their first owner). These sectors of the economy were, of course, precisely the sectors of the economy that bubbled and then collapsed in the current financial crisis; this off the GDP books action is at the core of what was going on this time.

GDP impacts that have flowed from these off this off the GDP books events have essentially a mere shadow of the driving events in the current economy, so it is not surprising that GDP measures don't work very well to show their true impact.

The distinction between the "real economy" and the "financial/secondary market" economy inherent in the very definition of GDP has never been more relevant than in the 2001-current recession boom. Financial sector firms reaped the lion's share of the economic growth in this boom, and the senior manager-production worker gap in compensatioon was eclipsed by a gap in financial industry senior manager v. real economy senior managers. Not one but two waves of secondary market collapses -- first in the secondary housing market and then in the secondary financial market, preceded the demand driven real economy effects.

Posted by: ohwilleke | July 24, 2009 at 03:37 PM

Personally, I think the IMF unemployment and GDP forecasts are the only ones worth reading. They've been pretty good in their last reports.
Is there any way to see what they use for projections?

Posted by: Dave | July 25, 2009 at 11:17 AM

okun's law does not take into account changes in tech, changes in financial management from longer to shorter time (don't need employees related to future products), nor outsourcing.

Nor does it deal with what requires a more narrative approach, the increased concentration of wealth and what rich people are likely to do with it (buying land in Chile, for example.) Math alone is not a good predictor in economics. A good reporter i also necessary.

Posted by: Doug Carmichael | July 26, 2009 at 03:38 PM

The comment by SS Resident is correct that even in the good times that companies would think nothing of trimming several hundred in headcount at a time in the pursuit of "efficiency" or "core competency."

The thing most puzzling to me is that considering the jobless recovery from the last recession that can be seen in the low participation rate for that time, that there are/were so many jobs left to cut to boost the unemployment rate that much further. This is particularly true in light of the methods used to calculate U3 as opposed to U6.

The questions about GDP and future revisions remain to be seen. I would also question the importance of credit and leverage in the period between the last and this recession and their effects on GDP.

Posted by: Mr.Sparkle | July 26, 2009 at 05:11 PM

Post a comment

Comments are moderated and will not appear until the moderator has approved them.

If you have a TypeKey or TypePad account, please Sign in

Google Search

Recent Posts



Powered by TypePad