Comparing Recessions II

by on January 15, 2009 at 9:42 am in Economics | Permalink

Earlier I posted some graphs from the Minneapolis Fed comparing this recession to a mildest, median, and harsh recession.  Questions arose as to how the Fed was defining these categories - harshest overall? how defined? at what point?   I took a look at the underlying data and saw a sensible procedure which seemed to make sense of what the Fed was doing and I posted that in the comments.  After further questions, however, and after contacting the Fed it's now clear that the Fed is doing something else.

The mildest, median and harshest recessions in the Fed's graph are Frankenstein recessions, recessions cobbled together by taking bits of pieces of each past recession and assembling them to create a mild, median, and harsh recession - none of which ever occurred.  I do not think this is a good way of looking at the data.   To avoid some of these problems I have simply graphed all of the data below for every past recession. In the extension to this post you can also find the Fed's justification of their procedure from an email to me.


Addendum: In response to my query and post the Fed has flagged and improved their description of the data to better convey their procedure.  

From Terry Fitzgerald at the Minneapolis Fed.

You are correct that the "mildest, median, and harshest" recession lines do
not represent single recessions. Please allow me to try to justify our
procedure. We spent considerable time weighing alternative approaches.

In drawing our timeline "length of recessions" graphs, we wanted to
illustrate where the current recession lies relative to past recessions at
each month of the recession. So for each month (or quarter), the lines
would tell you what had been the largest, median, and smallest decline in
any recession to that point.

The median line would indicate that one-half of the past recessions had
experienced larger declines, and one-half had experienced smaller declines
to that point. Similarly, no recession had a larger decline to date than
the "harshest" line. (And similarly for the mildest line.)

One feature of this approach is that the mildest, median, and harshest lines do not shift over time. So we can update just the "current" line in our graphs without all the lines shifting.

…I knew that insightful readers might wonder about this point, and I hoped that the note would at least explain what we did.

We are not trying to do anything deceptive or misleading with these charts.
Our aim is only to provide some empirical context to the current recession.

1 Milena Thomas January 15, 2009 at 10:13 am

Wow – I’m noticing a trend. All the recessions eventually end!

It seems that psychologically, there is a collective obsession with the recession that magnifies its importance. What if we all just realized, it’s going to end, somehow, at some point, and just got on with our lives?

2 Andrew January 15, 2009 at 10:22 am

I think your economistitis post has something to say here. The detection of the beginning of the recession is arbitrary. Early detection makes things look better, but they aren’t better. Move your start point for this recession to the inflection point at month 8 or so and voila, it’s nasty.

What’s interesting about this recession, to me, is that the catalyst was a long induction of below-trend growth. That, to me, indicates a long process of recovery.

3 frank January 15, 2009 at 10:42 am

Why do some of the recessions start with positive growth rates (’73 but also ’69 and ’80)? Seems like I really did not get the definition of the term recession …

4 Alex January 15, 2009 at 11:28 am

I like these type of analysis, but where the recession was official dated distorts everything – especially since that, in itself, can be controversial.

Imagine if the current recession officially began in Q3:2008. The employment decline would then look more severe relative to other recessions.

5 dale January 15, 2009 at 11:31 am

Is there a way to post graphs in these comments?
I graphed the data starting where the employment growth turns negative (it is strange that some recessions show several months of increased employment – presumably this has something to do with the way that recession is defined). Even after this adjustment, the current recession appears relatively mild in terms of employment. So, I don’t get Andrew’s comment about it looking ugly after the inflection point (unless there is some particular definition of “inflection point” that is useful here).

Of course, this is only one measure and misses the potential future scenarios, circumstances unique to this recession, and the fact that employment probably lags a number of important indicators. But I think the fact is that, in terms of employment, this is currently a mild recession. To the extent that psychological factors actually influence real economic events, it would be useful for the media to present factual information as well as the stories that sell well. Perhaps the media stories are actually contributing to the recession – my speculation is that this effect is worse than it has been in the past (it makes me wonder what to make of the “information age”).

6 es3200 January 15, 2009 at 11:53 am

To some extent a lower drop in employment should be expected since the recovery generated fewer jobs than in many other post-recession periods. If the percentage of people employed falls during the recovery, there aren’t lots of new jobs that can be eliminated. Because of this, this chart is not very meaningful.

To see this, look at the following. Since there are far fewer new jobs, it is probably much harder to find jobs to eliminate. In other words, low job losses in this recession may be due to a poor recovery rather than a relatively benign recession.

employed increase lost
1981 108.7
1990 125.8 17.1 ~2
2001 143.7 17.9 ~3
2007 153.1 9.4 ~3

In this context, the current recession may be much more severe than the last two.

7 Bob Montgomery January 15, 2009 at 12:08 pm

The detection of the beginning of the recession is arbitrary. Early detection makes things look better, but they aren’t better. Move your start point for this recession to the inflection point at month 8 or so and voila, it’s nasty.

Why do some of the recessions start with positive growth rates (’73 but also ’69 and ’80)? Seems like I really did not get the definition of the term recession …

I graphed the data starting where the employment growth turns negative (it is strange that some recessions show several months of increased employment – presumably this has something to do with the way that recession is defined).

I’ve always read that the standard definition of the beginning of a recession, which I assume is used here, is when GDP growth is first negative, not when employment first goes down:

[A recession is] defined usually as a contraction in the GDP for six months (two consecutive quarters) or longer…

I assume that this graph is showing us, among other things, that declines in GDP don’t always correspond 1:1 to declines in % employment.

8 Dave January 15, 2009 at 1:06 pm

What people are saying about recessions and economic growth is correct. I was looking at employment as a proxy. Also, it seems that my first link is broken above. My original graph (like Alex’s) is here. My adjusted graph that frames the timeline based on employment trends (see explanation above) and shows the “V” better is here.

9 John Dewey January 15, 2009 at 1:36 pm

es3200: “the recovery generated fewer jobs than in many other post-recession periods … low job losses in this recession may be due to a poor recovery rather than a relatively benign recession.”

Is jobs gained an accurate measure of the strength of economic recovery? The number of jobs gained after a recession should depend most on the peak unemployment level for that recession. Unemployment levels for the 2001-2002 recession were fairly mild compared to most post-WW II recessions:

Recession Peak unemplyment

1948 …….. 7.9
1953 …….. 6.1
1957 …….. 7.5
1960 …….. 7.1
1969 …….. 6.1
1973 …….. 9.0
1980 …….. 7.8
1981 ……. 10.8
1990 …….. 7.8
2001 …….. 6.3

BLS historical unemployment data

10 spencer January 15, 2009 at 3:07 pm

Thank you for this post, it makes my judgment of you much better and I apologize for being so harsh.

Determining the start and end of recessions is not as simple as most people think. For example, the NBER committee was unduly criticized last year for taking so long to determine when the recession started. But few of you probably remember that the NBER committee was also criticized in 1974 for determining that a recession had began too soon. There were several schools of though about the 1974 recession that were not reflected in the official final dates selected — partially because the NBER committee jumped the gun in determining that a recession had started. For example, one school of though actually though there were two recessions with much of early 1974 a recovery between the two recessions. The point that employment rose in early 1974 was part of that argument.

But I think that Alex now agrees with me that the Fed explanation of what the original chart showed was very misleading.

11 spencer January 15, 2009 at 3:19 pm

As of yet the Minneapolis Fed has not changed their web cite to explain what they did.

My problem is not with the methodology, there is nothing wrong with what they did.
The problem is that they presented their chart as something it was not.

12 Dave January 15, 2009 at 5:37 pm

John Dewey,

I agree with you that the Fed holding its rates down so low for so long helped contribute to the asset/credit bubble. The Fed typically raised rates to tame inflation, and then lowered them to encourage growth again. This time around, after raising rates before the 2001 recession, they certainly lowered them too low for too long. Lending standards also dropped (you could argue this was in part because of the low rates or the securitization model of debt; or perhaps it was the two combined that caused the drop in lending standards). These were the biggest factors in the bubble, as I see it.

As to which precedent we most closely match, I’m picking 1957. In my graph where I line up the employment numbers at the start of their accelerations downward (see link above in my earlier comments), they line up almost neck and neck. This would suggest that we’ll get to 9.5-10% unemployment around May or June (absent any stimulus effects). I could also see it following a path similar to 1953, which would suggest hitting 8.5-9% unemployment around June or July (again, absent stimulus). But of course, this is just guessing, and none of us really know.

13 California Dan January 15, 2009 at 7:07 pm

The graph highlight the methodology change in how the data points are created. The current numbers include large ( approaching 2 million) birth/death adjustments that are added to the measured numbers. Drop these numbers out and the current recession looks a lot more damaging.

The graph is counter-intuitive because the numbers are wrong.

14 mpowell January 16, 2009 at 7:37 am

Well, in 6 months I think we’ll really know how bad this recession is. If it’s as bad as I think it is, and a few others around here do, employment will have dropped another 1%, and I think people would agree that was really quite bad. That would be the path I would expect the economy to take if overproduction turns out to have been a real problem and the disappearance of a large asset bubble severely depresses demand. Alternatively, if unemployment does not continue dropping quickly and somewhat levels off, that would suggest that the fall in demand is not that big of a problem and maybe we wouldn’t need a huge economic stimulus. I’d like to see the fed get going on a stimulus package just in case, though. If things begin to turn around quickly, we can talk about toning things done somewhat.

15 John Dewey January 16, 2009 at 8:28 am


Changes in the workforce over the past 60 years reduces the confidence we can place in comparisons of recessions, IMO. Certainly manufacturing employment is significantly lower, primarily due to automation and work process improvement. When factories are idled today – in order to clear out inventories – the impact on national employment should be lower.

On the other hand, outsourcing (not offshoring) and term contract labor make it psychologically easier for a large firm to cut labor costs. A CEO can proudly claim that his firm did not layoff anyone, even while service contracts are reduced in scope and “temporary” labor contracts are not renewed.

Those differences came to mind pretty quickly. I’m sure there are others.

16 Michael January 16, 2009 at 4:40 pm

FINANCIAL CRISIS RESPONSE : IMF Spells Out Need for Global Fiscal Stimulus (, an interview with Olivier Blanchard, Economic Counsellor, and Carlo Cottarelli, Director of the Fiscal Affairs Department.
Blanchard:There are indications that the contraction in demand could exceed anything seen since the Great Depression in the 1930s. … In normal times, the Fund would indeed be recommending to many countries that they reduce their budget deficit and their public debt. But these are not normal times … What is needed is not only a fiscal stimulus now but a commitment by governments that they will follow whatever policies it takes to avoid a repeat of a Great Depression scenario.

And answer with only one graph in a turkish blog:

17 Steve January 20, 2009 at 3:28 am

Excellent work following up on the source of the data!

From what I recall employment usually lags other indicators such as real GDP and retail sales when you come out of recessions. Say 3-6 months on average. I believe in the last 2 mild recessions unemployment lagged falls in output and real GDP by well over a year. Clearly employment has shrunk very slowly in the first months of this recession meaning, from this perspective, it looks much like a typical or “median† recession to this point. However, look at the rate of change in employment in the last 6 months – its rate of fall is much faster than earlier and this rate is accelerating.

Retail sales have been decreasing 2-3% a month (yes a month!) for the last three months and that this rate of decrease is also accelerating. For more detail on retail sales see So given employment’s lag we should expect the employment to keep falling at the current rate or faster for at least 3-6 months even if retail sales turned around in January. Given that that rate of change in retail sales or real GDP is inevitably not going to turn positive in the next 3 months and probably not for at least 6 months (and possibly much longer though stimulus will give some positive bumps) we’re looking at a that employment graph heading south at the current rate of decline for 6 to 12 months. That will take the current decline in employment to worse than the harshest recessions on this graph in terms of total loss of employment. It will also mean the length of time from the start of recession until it begins to sustainably turn around (maybe somewhere between months 21 and 27 on that graph) would be much longer than longest recession. In the two most recent recessions the unemployment rate actually peaked around 15 months behind real GDP so if that pattern continues it’s even worse. So employment held up well early but it’s going to get uglier than in living memory and looks like living up to the “worst post war recession† hype.

I’ll be posting more on this adn implciations for equity markets at my blog at

18 jad jar February 10, 2010 at 7:35 pm

thank you for this information.sis jarMy local telecom is a monopoly, and it is out-of-control as far as wiretapping, eavesdropping, hacking, controling e-mail programs, phishing, spoof websites, etc.
No company should be immune from law suits and especially companies that control our communications.To give telecoms immunity will make “big brother”free nokia 6600 games“In this incidence and extent of formal coauthorship observed in economics against that observed in biology and discuss the causes and consequences of formal coauthorship in both disciplines. We then investigate the economic value (to authors) of informal comments offered by colleagues. This investigation leads us naturally into a discussion of the degree to which formal collaboration through coauthorship serves as a substitute for informal collaboration through collegial commentary. Data on manuscript submissions to the Journal of PolzticalEconomy permit us to shed additional empirical light on this subject. Finally, we demonstrate that while the incidence and extent of formal intellectual collaboration through coauthorship are greater in biology than in economics, the incidence and extent of informal intellectual collaboration are greater in economics than in biology. This leads us to search for evidence (which we find) of quids pro quo offered by authors to suppliers of free nokia n70 games

19 Stan Johnson August 20, 2010 at 4:13 pm

Thanks for the post! Very interesting, I like the graph, it really puts things into perspective.

20 louis vuitton backpack September 10, 2010 at 10:20 pm

I find this graph really surprising, I don’t think it reflects how bad things really are! The rate of foreclosure is astronomical, and layoffs are coming in waves more frequently. I’d take this graph with a grain of salt, they obviously don’t have all their numbers in.“>channel management

21 jordan shoes October 26, 2010 at 4:15 am

Here again we see the short delicate impressionist brush strokes, and the focus on color and light over exact lines. These are among the defining traits of the impressionist art.

22 Pillows February 27, 2011 at 2:47 pm

Thanks for all this great information, I’m doing a speech and this post helped me a lot. I think recession it’s caused by the lack of responsibility of the countries central banks. The good thing is that recession encouraged companies to do things more efficiently and effectively.

Comments on this entry are closed.

Previous post:

Next post: