The BLS reported today that 598,000 nonfarm payroll jobs were lost in January and the unemployment rate climbed to 7.6%. No one is surprised by this.
For anecdotal evidence, as I move around the region listening to and reading the local news, I am hearing more stories of what seem to be classic demand-driven layoffs. The ripple effects of the housing slowdown and the global demand slump are being felt in more places than they were 6 months ago.
Nearly 600,000 jobs lost sounds like a lot. Certainly, we haven't seen those kind of absolute numbers for a very long time. However, the labor market is much larger than it was in past recessions. So let's go behind the numbers and look at the job losses as a percentage of total employment.
Let me be very clear about the methodology behind these charts. I am calculating the cumulative net job losses during all post-war recessions as a percentage of peak employment near the start of the recession. In some cases, employment dipped slightly and rose again before taking a larger downturn. In those cases, I considered the peak to be the month before the larger downturn.
Though all of these downturns are associated with NBER recessions, not all of the peak employment dates coincide exactly with NBER business cycle peaks. (For example, the NBER dates a business cycle peak in November 1973, but employment did not turn down until July 1974, so I used the July 1974 date for this chart.)
As an example, employment peaks in July 1953 at 50,536,000. Thirteen months later, in August 1954, a total of 1,711,000 jobs were lost, which was about 3.39% of peak employment (dark green line).
It is a fairly busy chart, but we can see the current recession (orange) is very similar to the 1981 recession (light green) in terms of job losses as a percentage of peak employment. But we have had sharper downturns in percentage terms.
If you believe that this recession is not fundamentally different from other demand-driven post-war recessions, then a forecast of job losses continuing for another 6 to 9 months would not be out of line. Furthermore, looking at past cycles, one would expect it will be at least a year (possibly more if the recovery looks more like that after the 2001 recession) before employment reaches the previous peak. Personally, my expectation is that it will take 18 to 24 months (from now) to get back to the previous peak.
Finally, here's a version of the chart with the last four recessions (including the current one) that is a little less cluttered. I've also included the 1957 recession as an example of a sharper downturn in payrolls.
UPDATE: Spencer at Angry Bear looks at household survey (CPS) data and concludes that the current downturn is the worst in the post-war period. Actually, according to the household data the 1953 recession is worse, but that's a minor point. But there is another issue to consider. I'm not sure what to make of the fact that the household survey is subject to less variation (in both directions) than the payroll survey, especially during the '70s and '80s. It is widely acknowledged that the household survey is a less reliable indicator of the labor market overall than the payroll survey. So I am understandably nervous about the possibility of understating the severity of the employment declines in the '70s and '80s if we were to rely on that data.
For anecdotal evidence, as I move around the region listening to and reading the local news, I am hearing more stories of what seem to be classic demand-driven layoffs. The ripple effects of the housing slowdown and the global demand slump are being felt in more places than they were 6 months ago.
Nearly 600,000 jobs lost sounds like a lot. Certainly, we haven't seen those kind of absolute numbers for a very long time. However, the labor market is much larger than it was in past recessions. So let's go behind the numbers and look at the job losses as a percentage of total employment.
Let me be very clear about the methodology behind these charts. I am calculating the cumulative net job losses during all post-war recessions as a percentage of peak employment near the start of the recession. In some cases, employment dipped slightly and rose again before taking a larger downturn. In those cases, I considered the peak to be the month before the larger downturn.
Though all of these downturns are associated with NBER recessions, not all of the peak employment dates coincide exactly with NBER business cycle peaks. (For example, the NBER dates a business cycle peak in November 1973, but employment did not turn down until July 1974, so I used the July 1974 date for this chart.)
As an example, employment peaks in July 1953 at 50,536,000. Thirteen months later, in August 1954, a total of 1,711,000 jobs were lost, which was about 3.39% of peak employment (dark green line).
It is a fairly busy chart, but we can see the current recession (orange) is very similar to the 1981 recession (light green) in terms of job losses as a percentage of peak employment. But we have had sharper downturns in percentage terms.
If you believe that this recession is not fundamentally different from other demand-driven post-war recessions, then a forecast of job losses continuing for another 6 to 9 months would not be out of line. Furthermore, looking at past cycles, one would expect it will be at least a year (possibly more if the recovery looks more like that after the 2001 recession) before employment reaches the previous peak. Personally, my expectation is that it will take 18 to 24 months (from now) to get back to the previous peak.
Finally, here's a version of the chart with the last four recessions (including the current one) that is a little less cluttered. I've also included the 1957 recession as an example of a sharper downturn in payrolls.
Great post! ...I just stumbled upon your blog, good stuff, thanks.
I prefer the household survey. I contend that it is a better indicator of cumulative gains or losses, because it is not subject to various biases that accumulate in the payroll survey.
Month to month, it has a larger error, but methodologically, you can't get much simpler than contacting a random sample of people and asking them "are you employed".
I like simplicity. It results in a more meaningful cumulative number (versus simply counting payroll employment when there is cyclical and secular movement among payrolls, contractors, self-employment, and so on).
William:
I posted this over at "Angry Bear" in response to you.
I was looking at your Blog, found it very detailed, and added it to my resources. If you will endulge my lack of knowledge on the topic for a bit and also the subsequent questions, I am hoping you can answer a couple of points. I do agree with the percentage determination in comparison to other years.
I am assuming Labor Market is the equivalent of the BLS Civilian Labor Force. In which case, you are measuring within the NonInstiturtional Civilian Population. Since the sixties/seventies(?), there has been an ever increasing percentage of the population as a part of the Civilian Labor Force, an ~30-40 year upward trend. In 2001 and immediately after the official end of the recession, Participation Rate stood at 66.7/8%, which is about what it was pre-Dot.Com bubble. Participation Rate now stands at 65.5%, not seen since 1986? Does the model capture that decrease in percentage of the population as a part of the Civilian Labor Force?
I think you know where I am going with this and maybe I am missing something or do not understand something portrayed. It could be that the graph is not meant to take it into account or I could be blowing smoke.
Thanks for your return.
"If you believe that this recession is not fundamentally different from other demand-driven post-war recessions, then a forecast of job losses continuing for another 6 to 9 months would not be out of line."
The problem is that the second derivative of the curve still appears to be negative (although it may be approaching zero). Seeing an inflection point would be more reassuring ...
It would be useful to include Great Depression data in this graph, since the current recession is so often compared to that period. How steep, how deep, and how long is that crater in the graph compared to the others?
"If you believe that this recession is not fundamentally different from other demand-driven post-war recessions, then a forecast of job losses continuing for another 6 to 9 months would not be out of line."
I agree with Mr. Bell - where's the inflection point? Using the July 1981 recesssion as a model of an 18-24 month downturn, the rate of unemployment growth had already turned by this point.
Here we have a triple-whamy of a still increasing number of upside-down mortgage holders, toxic debt and a malfunctioning credit market, anf finally declining demand and employment. It seems to me to be more than a business cycle recession. Convince me otherwise.
To Robert and Jim,
Of course it is true that seeing the inflection point would be reassuring. And any of my regular readers certainly know that I am not suggesting that I believe that the current recession is a carbon copy of '81-'82. Far from it. The point is that this recession seems more like '81-'82 than it is like 2001 (or even 1991). Part of what I am trying to do here is provide some context. Given that many of my readers were pretty young in '81-'82, it may surprise them to know that the numbers we are seeing now are even close to what we saw then. The dynamics at work then and now were quite different, to be sure. But the message is that what we have seen so far is not orders of magnitude different from the realm of our experience.
So could it go deeper than '81-'82? Yes, I believe that is certainly possible. In fact, if I were literally suggesting that this recession is a copy of '81-'82 (which I am not), then I would have said 4 to 6 months more of job losses. I thought by saying 6 to 9 months more I was already acknowledging that this has a significant probability of being a bit more protracted. But I'm not ready to say that I think it will be drastically more protracted at this point. I do stand ready to re-evaluate that stance periodically. But do I think we're heading for another Great Depression? No.
Do I think that the recovery will be take as long as in '91 or '01? Hard to say, but possibly shorter. I think that one of the things contributing to the longer recoveries in the last two recessions were that the job losses were more structural than cyclical. (See here, for example: http://www.newyorkfed.org/research/current_issues/ci9-8/ci9-8.html) Some of the layoffs I hear about are from manufacturers of intermediate goods (parts) facing a drop in orders from big manufacturers--and not just the auto companies. There is no good reason in my mind to expect that the demand in those sectors will not rebound. Hence, I would expect the recovery to be somewhat quicker than in '01--all else being equal.
But of course, all else is not always equal, which is why we will revisit this periodically over the next few months. And I invite you to stick around and continue the discussion.
"There is no good reason in my mind to expect that the demand in those sectors will not rebound. Hence, I would expect the recovery to be somewhat quicker than in '01--all else being equal."
I hope to stick around.
Let me raise a point, little discussed as yet:
The demand of 2002-2006/7 was driven, according to consensus accounts, by easy credit and rising home values. Absent another bubble (but see Erix Janszen in Harpers), we would not expect demand to return to previous levels in the near future - 3 to 5 years - unless we get an emerging economic sector with extremely strong growth and the linkages to other sectors to spread the growth broadly.
Are you counting on some such engine to restore demand, or am I missing some reason why it isn't necessary?
Dan,
Monthly data that far back is hard to come by, but we can look at annual data (see here: http://www.census.gov/statab/hist/HS-31.pdf) to find the following...
From 1929 to 1932, nonfarm payroll employment declined from 31,324,000 to 23,615,000. That is a decline of 25% over three years.
So far, we have experienced a decline of about 2.5% over 13 months.
Or put another way, in the last 13 months, the total number of jobs lost has been just under half of all the jobs lost during the period from 1929 to 1932 while the total number of jobs today is more than 4 times the level in 1929.
So when you ask, "How steep, how deep, and how long is that crater in the graph compared to the others?" My answer is, so steep, so deep, and so long, that adding it to the chart would relegate all the other downturns to a little spot in the corner of the graph.
Jim,
Short answer (and all I have time for now):
While easy credit and rising home values did play a role, I don't think it is accurate to say that the demand of 2002 to 2007 was "driven" by that bubble. I hadn't read Janszen's piece, but I just did now, and I don't find his story all that compelling.
I, for one, do not think that the long term prospects for growth of GDP and total employment have diminished. Are some big changes ahead? Sure. But at what time in our history would that NOT be true?
Of course, if you're trying to sell a story that only a green energy revolution can save us... well, I guess I'm not buying. (Not that a certain amount of investment in green energy would be a bad thing.)
That'll have to do for now...
This may be a fooish question (my economics background is limited to writing checks), but I wonder if there's any sort of effect involved in the change (if any) over time in the relative employment by small business (maybe 100?) and government? In other words, how was that 25% in '29-'32 distributed compared with the present 2.5%? I have an anecdotal feeling that we have a greater concentration of business depending on a larger pool of purchasers/payers, such that a smaller percentage of job losses may still amount to a massive, possibly critical, level of money loss to business. Also, it would be interesting to see that in some sort of context - what salary figure distributions are we looking at now vs. then, and also also how do we take into account (or is it necessary?) how stretched both employees and businesses are now vs. then? As I recall, there was a lot of hand-wringing over the past decade or so that the US economy was now primarily a 'service' as opposed to a 'production' one. Is there any easy way to factor that in somehow? I would think that this might affect at least the unemployment recovery rate, but then this is wildly unimformed speculation..
Think of job losses as slices taken from a pie. The Great Depression removed 1/4 of the pie. The current recession has so far removed a sliver of pie 1/10 as large as that. The distribution of job losses are certainly a little different, but not drastically so. Look at the link I provided in my 3:43PM comment. It shows the job loss by sector as well. Manufacturing jobs shrunk by 1/3. Construction by more than 40%. Services lost less than 20%. Government (then about 10% of jobs; today about 16%) was steady. The distribution of losses is not exactly what it was then, but I don't think that you can claim that the differences in the distribution account for the disparity in the overall magnitude.
Comparisons of the income distribution then and now are tricky, but it can be done. (Look at the census data.) But if you're suggesting that we are somehow worse off today because of the income distribution, I'd have to say no. As the Great Depression set in there were practically no government social programs. People were on their own to be cared for by extended families or private charities. Not to downplay the plight of the poor today, but just read some of the accounts of the widespread poverty of the Great Depression. There's no comparison.
Furthermore, just consider one aspect of the current crisis--foreclosures. Many (though admittedly not all) of the foreclosures are happening to people who are not in poverty. While it is certainly disruptive to go through the the process, many people will return to renting as they probably were doing just a couple of years ago. It's not a pleasant circumstance, but it's a far cry from what happened in the Great Depression. Again, I am not insensitive to real cases of poverty--and it is appropriate for government to address that--but I am growing weary of comparisons to the Great Depression.
I suppose I have just made a case for why we should teach more and better history (and economic history) in the schools and universities.
Can you post an update to these graphs please so we can see how it is trending over time.
Mr. Polley,
With a couple more months of job loss data out now, do you have an April update? How is this downturn looking now relative to priors?
I'd also be interested in your comments regarding two of my observations:
1. What we can notice as a general trend is that (starting in '57 and moving forward) the shapes of the downturns has flattened out. (All those from '48 to the '70s were similar in shape, though different in magnitude.) This could be explained in saying that the market has gained size and thus inertia; as a smaller portion of the economy heads south, the overall number isn't taken as far down, it heads back north again as new economic activity replaces the old.
I would propose that the reason for the lengthy recovery period in the 2001 downturn is because we didn't have a large emerging industry to quickly add jobs to the pool. I would further propose that the reason is because we off-shored a substantial portion of our manufacturing base, and any new manufacturing ideas have tended to get established overseas rather than here, where labor costs are highest. So in summary it's been hard for American entrepreneurs to add industry here at home, and thus as a group we're not able to create value and jobs.
This does not portend well for us unless we pretty quickly come up with a really nifty idea that we do or manufacture here in the USA. We can't just sit around and hire each other to shine each others' shoes; we need to make something new, something so extraordinarily valuable that it can profitably overcome the lofty salary expectations of the American worker, and which the rest of the world can't copy. A tall order...
2. Save for 1957, the slope of job losses this time around is steeper than ever. All other factors being equal, this should have been all the more unlikely in a larger (higher inertia) market. I think that a lot of us sense that this downturn is more severe than prior ones. That today we should be losing jobs at such a rate IN SPITE OF the apparent tendency for downturns to be gentler in slope as the job pool has grown is what I find most remarkable; THIS is why I think we should have all hands on deck. In the absence of a new valuable economic activity to replace what we're losing, I suspect that we're going to see losses continue for a while, bottom substantially lower than where they are today, and then take a long, long time to recover.