Saturday, October 16, 2004

And Following Up on the BLS E-mail

So now let's tabulate this group of "people who want a job now" but are not in the labor force. Using the links from the previous post, we can generate the following comparison of June 2003 and September 2004 (with the first 6 rows in thousands, and pardoning me for the poor formatting):

6/2003 9/2004
Labor Force 146,917 147,483
Employed 137,673 139,480
Unemployed 9,245 8,003
Not in Labor Force 74,097 76,458
Want a Job 4,687 4,850

Unemployment Rate 6.29% 5.43%
LF/(LF+NILF) 66.47% 65.86%
Emp/(LF+NILF) 62.29% 62.28%
Want/NILF 6.33% 6.34%
(Unemp+Want)/(LF+Want) 9.19% 8.44%

The first five rows are the raw data in thousands. The next five rows are the key ratios. The unemployment rate, labor force participation rate, and employment-population ratio are as discussed in the original post. The last two lines make use of the new category suggested by the BLS e-mail in the previous post. The share of those "Not in the Labor Force" who "Want a Job" is at about 6.33 percent in the two months. If we were to add this group to the labor force as unemployed, and then recompute the unemployment rate, we would see (in the last row) that it fell by 0.75 percentage points, from 9.19 to 8.44 percent of this augmented labor force.

With this measure, as with the other augmented measures of unemployment discussed in the original post, there is no empirical support for a proposition that the reduction in the unemployment rate (since its peak) is due to in any substantial way to a greater fraction of the potential labor force being involuntarily out of job.


Anonymous said...

More later but I wanted to highlight something:

"there is no empirical support for a proposition that the reduction in the unemployment rate (since its peak) is due to in any substantial way to a greater fraction of the potential labor force being involuntarily out of job."

This statement is simply not true. The employment/population ratio has remained unchanged in an anomalous way that does not match its pattern of behavior in the past. This is empirical evidence and to state otherwise is not correct. Professor Samwick relies on a BLS series that is not the slam dunk that he thinks it is, but I want to do more research before I call him a stubborn fool. I will observe that he made statements about the status of Ryan's status based on ignorance so I will observe, "falsis in unum, falsis in omnibus"

He still owes Krugman an apology.

Andrew Samwick said...

Thanks for your comment. You note:

"The employment/population ratio has remained unchanged in an anomalous way that does not match its pattern of behavior in the past. This is empirical evidence and to state otherwise is not correct."

In the original post, I acknowledge that over the time period in question (June 2003 - Sept 2004), the employment-population ratio is unchanged. The raw data are presented in this post and you can see them for yourself.

The key issue is to recognize that there are (at least) two ways for that ratio to stay the same, while having the labor force participation rate and the unemployment rate fall. The first is the one that Krugman pointed out: some people who are actively looking for work may stop looking. The second is the one that I postulated: some people who are working drop out of the labor force voluntarily and are replaced by unemployed people. Krugman's error is to say that only the first way is operative.

You take issue with my conclusion to this post. I'm saying that I'm having a hard time finding evidence of Krugman's hypothesized explanation for the data. I have not said it doesn't exist at all--I have not said the changes are "only" due to the pathway I posited.

The unemployment rate as conventionally measured has fallen by about 0.9 percentage point. If this decline were due to unemployed people simply giving up their search, then we would expect to see smaller declines in the unemployment rate when augmented to include discouraged workers, marginally attached workers, or anyone who would simply want a job. Those declines are all about the same size as the decline in the unemployment rate as it is conventionally measured (maybe a tenth of a percentage point smaller).

Thus, I find no empirical support for Krugman's hypothesis having any substantial role in explaining the changes since June 2003. (I think that "not substantial" is a good way to describe 0.1 out of 0.9.) If you can find any evidence of a greater share of the working age population being involuntarily out of a job over this period, I remain very interested in seeing it. It won't make what Krugman wrote correct, but it would help all of us understand the relative importance of voluntary versus involuntary reasons for the decline in labor force participation.

Roland Patrick said...

There's something else about this ongoing debate on Semi-Daily Journal that is ignored (or deleted) every time I bring it up. The late 90s were a verrrry unusual labor market. For example:

Felons, high-school dropouts, seniors, literally the lame, the halt and the blind, were drawn into the labor force by what we now know to have been a bubble economy. For Krugman to be whining that the employment to population ratio has not gotten back to where it was is disingenuous in the extreme.

Not to mention that one would normally expect an economist to know that a job is not a benefit, but a cost. And increased productivity, i.e. more output with less input (jobs)used to be thought a good thing.

Ryan Early said...

The late 90s U.S. economy was unique in another very specific way - we had full employment without inflation, which traditional Keynsian theory stated as impossible. It was also the only period in the past fifty years in which we had positive growth of real average wages. As such, it is very much indeed in all of our interests to recreate that situation and then try to prolong it by not letting our "irrational exuberance" form an economic bubble.

Anonymous said...

Interesting debate, sorry to be so late to the party. I appreciate your willingness to engage in this spirited back-and-forth with your readers.

Having said that, I think you are seriously oversimplifying the issue through your constant recourse to BLS statistical indicies. The last few years have seen massive changes in states' unemployment insurance policies, with some states dramatically raising eligibility requirements and bureaucratic hurdles in a deliberate effort to reduce claims. An additional factor that warrants attention is the growth of independent contracting and other employment arrangements not eligible for UI coverage. The notion that you can do any sort of real time-series analysis without controlling for these two downward biases strikes me as naive.

The American workplace has been transformed in the last few years. These days, labor market analysis is as much art as science. Krugman may well have been wrong on this (or at least sloppy), but it will take more than downloading a few figures from the BLS website into Excel to prove it. I also have to wonder, did you ever write to him to ask for more details? It seems unsporting to sharpshoot a 750-word article written for a lay audience...did you expect footnotes?

Andrew Samwick said...

Thanks for your comment, and welcome to the discussion.

Instead of the word "only," I expected a word like "primarily" or "partially," depending on what the data could actually show, and a recognition that not all of the reduction in the labor force participation rate is an appropriate critique of the economy's performance since June 2003.

A couple of times in the past, I have written to Krugman personally or as letters to the editor in the NYT. The former have not been answered, and the latter have not been published.

The augmented measures of the unemployment rate do oversimplify the broader discussion of how important the voluntary versus involuntary pathways have been, but not the original critique of what Krugman wrote. With each augmented measure that fails to show a substantial difference in its June 2003 to September 2004 decline relative to the standard unemployment rate, it becomes harder and harder to find any empirical support for the involuntary pathway to which Krugman attributed the whole decline. That does not mean there are no groups that are involuntarily out of the labor force that would rather be in (see Ryan's second comment contending that he would not be tracked even in this broadest group), but it does mean that we don't find support for even the weaker version of Krugman's conjecture in the places we would expect it to appear.

Your points about the changes in unemployment insurance programs are interesting. The impact they would have on my critique would depend on whether the people who are now not able to obtain unemployment insurance (and be in the labor force) would be classified as "wanting a job" under the augmented measure of unemployment discussed in this post.

I am also sympathetic to the need for additional measures of under-employment or under-compensation of those in the labor force (as suggested in some comments on earlier posts) to better understand the change in the labor market since its peak in 2000/2001.

Thanks again for your comments.

Roland Patrick said...

Today's Krugman column is even worse:

Adam O'Neill said...

Krugman argues that the declining employment rate over the last year is due to declining labor force participation (the rate is defined as employment/labor force participation) and not to job growth and says that if the labor force were constant at its 2001 level unemployment would be 7.4 percent. There is nothing factually wrong with this analysis.

Given that the employment-to-population rate is fixed over the time period in question at 62.3 percent, it is a mathematical identity that the unemployment rate is determined by the labor force participation rate. (If the population doesn't change and employment doesn't change, but the unemployment rate falls, isn't there only one potential culprit?)

Samwick tries to solve things (a tough task when hitting up against the axioms of algebra) by throwing more statistics at the problem. He examines discouraged workers and marginally attached workers (individuals usually left out of the traditional definition of the labor force) and notes that even when you factor these guys in the "modified" unemployment rate still falls.

But this analysis still glosses over the fact that the unemployment rate fell because fewer Americans are in the labor force (or in the labor force plus discouraged workers or the labor force plus discouraged workers plus marginally attached workers...) and not because of improvements in employment. Where are these people? Some get fired and retire. Some get fired and raise families. Some get fired and go back to school. These people aren't in the labor force even under Samwick's definition. But it's not clear it's because they want to be. However you label them, these are people who worked when the market was good and don't now. Krugman's point still holds; if these people were still counted, the unemployment picture would look a lot worse.

More here.

Andrew Samwick said...

Adam also posted his comment to his blog, so I responded briefly there.

Adam O'Neill said...

I think the differences between our interpretations of the unemployment statistics boils down how we interpret falling labor force participation rates. Just in terms of the raw numbers, when the unemployment rate goes down but the employment/population rate is constant then we know by construction that people are leaving the labor force. This is what I took Krugman to mean when he wrote "...unemployment declined only because some of those without jobs stopped actively looking for work..."

Are they leaving because of adverse economic conditions or for another reason? Krugman obviously thinks it's the former and that the discouraged are leaving the workforce. Samwick asserts that individuals are leaving the workforce for other reasons and the evidence he brings to bear is the data from the CPS identifying marginally attached and discouraged workers (questions designed to eliminate the controversy over who is or isn't in the workforce). His alternative measures of the unemployment rate mirror the reported unemployment rate. This suggests to me that people are leaving for non-economic reasons (Samwick's point I believe) or that the census bureau's questions are not capturing all the gradations of willingness to work (my point).

The difference boils down to how you interpret the individuals who are no longer in the labor force and no longer tell the census bureau that they are marginally attached or discouraged. My take (and I think Krugman's take) is that these are people who work when it's good and don't when it's bad. Like I said in my post, I think these are people who are substituting retirement, schooling, or raising a family for current employment because current employment prospects aren't what they were 4 years ago.

Where are the data? First, I think falling labor force participation rates in and of themselves signal that individuals are finding the labor market less attractive (at least relative to what they were a few years ago). Beyond that, I think you could just go back to the CPS and see what people are doing now (Are 18-24 year olds more likely to be in school? Are women 20-40 raising children? etc). But frankly, I don't think there is data that will give you an explicit answer.

A link back to my post and your comment.

Anonymous said...

Adam O'Neill has been able to state the case much better than I was. Given it is a matter of interpretation and that Samwick has shown himself already to have been somewhat ignorant as to how the labor statistics are derived, I believe his assertions that Krugman made a false statement and that there is no empirical evidence for Krugman's statement. He simply overstates his case, but insists he's right. He won't apologize and if this is consistent with his past behavior then it is no surprise to me that Krugman ignores him.

As to the series that Samwick references, I promised I would comment after more research:

1. This series tracks the unemployment numbers and the employment/population ratio through the 90's but then does the same thing as the unemployment numbers which stop correlating with the employment/population ratio so citing it does not do anything but return to the original disagreement about the reliability of the unemployment number to accurately reflect the state of the job market. (By the way, if you replace the "S" with a "U" and the first "1" with a "0" then you get the unadjusted series, but I'm not sure that illuminates anything.)

2. According to footnote 1 on the relevant table in the cps, not everyone "not in the labor force" is asked this question which I don't understand because elsewhere in the handbook, it asserts that everyone is asked this question. (something like, Are you interested in working now?)

3. The phrasing of the question does not identify those (like Ryan) who left the workforce due to economic reasons, but are not ready to reenter it at this time.

Andrew Samwick said...

I think we can isolate where Adam and I disagree, and I don't really think Anonymous and I will ever agree.

Adam says:

Just in terms of the raw numbers, when the unemployment rate goes down but the employment/population rate is constant then we know by construction that people are leaving the labor force.So far we all agree. People are leaving the labor force. My post shows the magnitude of these flows. Adam continues:

This is what I took Krugman to mean when he wrote "...unemployment declined only because some of those without jobs stopped actively looking for work..."Here's the issue: that may be what you take Krugman to mean, but that is not all that the words he has written mean. The statement in quotes assigns one (and exactly one--that's why his use of the word "only" is critical here) reason for people leaving the labor force. What he has written completely excludes the other possible explanation that I posited--some of the decline may be due to people voluntarily leaving the labor force and being replaced by unemployed workers.

My critique is that such a complete exclusion is unwarranted. Where is the evidence that none (or, as Richard Schwartz has commented, no statistically significant amount) of the decline in the labor force participation rate between June 2003 and September 2004 is due to voluntary exits of employed workers with replacement by unemployed workers?

Krugman used an extreme statement to make an extreme point. If he had said that the decline is "partially" due to people giving up an active search, then it is clear from the data and anecdotes that we all have been able to bring to bear on this topic that such a statement could not be falsified. But "partially" would not have had the same effect as "only" in his editorial.

If this wasn't clear in the original post, I hope it is clear by now.

JG said...

"Krugman argues that ... if the labor force were constant at its 2001 level unemployment would be 7.4 percent. There is nothing factually wrong with this analysis."

Ah, but 2001 was a "bubble year" for employment as it was for the stock market! Hardly a year that an impartial observer would use as a base to gauge a "normal" labor market. Instead of just looking at what happened after that bubble, how about looking at what happened before?

Let's remember what Krugman said earlier, in 1996: that five years of 4% average economic growth was impossible because it would reduce the unemployment rate to only 1.5%. And only ignorant fools who can't do simple math could think that possible. (He was as tactful then as ever!)

Thereafter, of course, we had five years of 4% growth that reduced unemployment to only 4%. How could Krugman have been so wrong?

Because he didn't realize that the boom would draw people who *weren't* discouraged workers into the labor market due to all the new atttactive job opportunites it created. [See Roland Patrick's comment]

Now let's look at that pre-bubble labor market in 1996 -- a fine Clinton economy year -- when Krugman wrote and declared the employment situation to be about the best that was sustainable, near NAIRU, and compare it to this year. (8/96 and 8/04)

Then the labor force participation rate was 66.7% and employment rate 62.4% for an unemployemnt rate of 6.4%. In 2004 the numbers were 66% and and 62.4% with an unemployment rate of 5.5%

Now, as per Krugman, if the labor force participation rate today was constant at the healthy, sustainable, pre-bubble 1996 level, the unemployment rate today would be the same 6.4% as it was then -- which Krugman then considered *good*. How does he justify it as being so very bad now?

As for the 2001 peak in the participation rate -- is Krugman willing to say that the 2.5 point drop in the unemployment rate that he falsely predicted didn't occur because so many *discouraged* workers that *he didn't recognize as such* in 1996 entered the labor force by 2001?

If not then how can he declare them discouraged when they left the force after 2001. Has he become any better at recognizing discouraged workers today? Maybe they are as content now not looking for a job as they were in 1996.

Where's the evidence to the contrary?, is what I think our host is asking. And it's a fair question.