More Intellectual Garbage Pickup

This is really a thankless task. But somebody needs to occasionally lay down a marker that Steve Antler writes with forked tongue.

He attempts to come to Daniel Okrent's defense:

EconoPundit: Professor Krugman says... "unemployment lasts much longer than it used to."... Daniel Okrent says: ...columnist Paul Krugman has the disturbing habit of shaping, slicing and selectively citing numbers in a fashion that pleases his acolytes but leaves him open to substantive assaults."

A plot of the BLS statistic 'average weeks unemployed' from its 1948 start to the present illustrates Okrent's point:

The first gray arrow shows upward trending from 1948 to 1983. During those years Krugman was correct. The second gray arrow shows a slight downward trend since the mid 1980's. During these years, Krugman is wrong.

So which is it? Relative to 1948, unemployment now lasts longer than it used to. Relative to 1983 recession-peak-unemployment lasts about two weeks less than it used to.

I guess the question is one of interpretation. It just has to be left to the voters. Do they share Krugman's pessimism, or do they feel optimistic about the future?

Here is Antler's figure:

What should you think of this argument?

1. You should look at the duration of unemployment during business cycle peaks, and see a steady rise: 7.8, 10, 10.8, 11.9, 12.6--today's level is 19.3.
2. You should look at the average duration of unemployment during entire business cycles, and see a steady rise: 12, 10.5, 12.8, 15, 15.4, 17.1.
3. You should cyclically-adjust the series--after all, we know that the duration of unemployment will be relatively high whenever unemployment itself is relatively high--and calculate what the duration would be if the unemployment rate were six percent, as is shown here:
   
4. You should reflect on the fact that of these four possible ways of calculating the trend in unemployment duration, only one way--and a relatively weak way--produces Antler's downward trend since 1983. And Antler's way produces a much stronger upward trend in unemployment duration in the 1970s and early 1980s than any of the alternatives.
5. You should reflect on the fact that Krugman's "unemployment lasts much longer than it used to" is part of a tradition of analysis that contrasts the first post-World War II generation--the 50s, 60s, and early 70s--with today. There's nothing in Krugman to make anyone imagine that "used to" applies to the 1980s only.

But it really isn't economics that Steve Antler is doing here, is it?

Comments

[pgl]

Even using his figure sine those "trend lines", what we see is that this recession is the 2nd worse since 1948 in this regard. Saying things are not quite as bad as they were during the worst recession since the Great Depression ain't exactly comforting.

[Frank]

Brad:
If you have time, could you explain briefly how you computed the cyclically adjusted unemployment duration series? I can take the answer in technical language.

Thanks!

[howard]

actually, this is quite revealing as to why okrent is such a nitwit.

he sees stuff like this and believes it, apparently lacking any of the necessary intellectual tools to separate propaganda from analysis. It's nothing more than a "where there's smoke there's fire" sensibility.

very helpful to see the precise nature of the pathology made clear.

meanwhile, good riddance to bad rubbish remains the appropriate farewell to okrent.

[Brad DeLong]

Simply regressed unemployment duration on a time trend and on the unemployment rate. Data from the 2005 ERP, annual. 1959-2004. No corrections for autocorrelation or anything--just what I could do in less than 60 seconds.

[jerry]

I am shocked, shocked to discover that econopundit does not allow comments on his blahg.

[anne]

Nicely done :)

[Robert Waldmann]

Steve Antler does illustrate what Daniel Okrent meant by
"shaping, slicing and selectively citing numbers in a fashion that pleases his acolytes but leaves him open to substantive assaults, (ssascniaftphablhotsa)" since he does exactly that. In particular, slicing at 1983 is an extraordinarily blatant form of ssascniaftphzablhotsa (the z is for zero). Your critique is definitely a substantive assult on Antler.

id you really do it in 60 seconds ? I don't have the data on my hard disk. some of the tiny amount of attention I have gotten blogging is based on keeping track how long it took to prove something (I still remember the humiliating 2 hours for the Nigerien Yellowcake dossier but I'm much quicker now).

I would say that to evaluate Krugman's claim (correct) it is enough to just look at the graph. The current level is one of the highest on the graph and is much higher than the average over every interval of time long enough to correspond to "used to be" which sure doesn't mean "in 1983". I mean a anyone who can understand English and knows what a graph is can see that Krugman is right and Antler is wrong. No need for fancy stuff like regressions. No need even to note that the earlier peeeks were all in recessions and we are 3.5 years into an expansion now.

Below Patrick Sullivan quotes Mankiw saying that Krugman treats people who disagree with him as if they were all fools or liars. If so, according to the normal usage of or, Krugman would be unreasonable, since some are both.

Like the effort to quote Mankiw to help Okrent, Antler's effort gives me hope that a negative can, in the end, be at least convincingly argued. If Antler is of normal intelligence and this is what he came up with, then Krugman may be the first person proven to be innocent of
"shaping, slicing and selectively citing numbers in a fashion that pleases his acolytes but leaves him open to substantive assaults,"

[pi]

The Antler piece is silly, not the least because the upward trend in unemployment durations is not something suddenly discovered by Paul Krugman. Many people have noticed it and speculated about what's driving it.

However:

(1) The *much* in Krugman's statement "unemployment lasts much longer than it used to" is also a bit silly. The change is pretty modest and seemingly part of a long-term trend.

(2) Somewhere between 0.5 and 1.0 weeks of the measured increase in mean duration is due to the 1994 CPS redesign rather than to an actual increase in durations. See http://www.bls.gov/ore/pdf/ec950090.pdf for a discussion of how dependent interviewing increased measured durations starting in February 1994.

[Rob Sperry]

Why not reverse the analysis and take out the effect of unemployment duration on unemployment level and frequency? What insight on the current unemployment level and frequency would we gain from this?

[The unemployment rate is a very good indicator of the overall macroeconomic health of the economy. Unemployment duration is not.

If you did that, you'd conclude that unemployment today is *way* lower than it should be given how difficult the unemployed seem to find it to get new jobs.]

These are all dynamically changing variables, unless there is some strong theoretical reason to assume that they must be independent then I don't see why artificially holding level and frequency constant is supposed to yield any information.

[James]

I am the blogger who originally started this discussion. Although Mr. DeLong did not address my argument, and probably didn’t even read it, I will insert my little bit regarding the original point I was trying to make. As far as Mr. Antler’s analysis I will leave that up to him if he so chooses.

“You should reflect on the fact that Krugman's "unemployment lasts much longer than it used to" is part of a tradition of analysis that contrasts the first post-World War II generation--the 50s, 60s, and early 70s--with today. There's nothing in Krugman to make anyone imagine that "used to" applies to the 1980s only.”

Yes there is, he said so. Mr Krugman stated, “Over the past 25 years the lives of working Americans have become ever less secure.” And then proceeded to list 5 supposed economic indicators of this phenomenon. He did not state “Since WWII the lives of working Americans have become ever less secure.” Maybe Krugman is only guilty of vague sentence structure, but this is the impression he leaves with that statment. My original article was asking why he picked that time period, whether the numbers were accurate was only a follow on issue. I can find no particular significance to this number, other than he probably wanted to blame Reagan for the start of this supposed trend, not for any particular economic reason. This is politics, not science. The whole gist of my original post was that this was a ridiculous time period to compare the present situation with. Although things may not be stellar now, can any serious economist, or layman, look back on the era of 10% unemployment, and 20% mortgage rates, and say people felt better off and more “secure” then?

"You should cyclically-adjust the series--after all, we know that the duration of unemployment will be relatively high whenever unemployment itself is relatively high--and calculate what the duration would be if the unemployment rate were six percent, as is shown here:"

This is an idiotic way of looking at it. You have to remember, Krugman was not merely making a scientific observation of the relation of unemployment rates and unemployment duration, he was trying to make a socio-economic point as to how this factor effects people. Adjusting this to some arbitrary 6% not only obscures this effect, it intentionally distorts it. The fact is (once again as I said in my original blog post) that unemployment has trended down significantly in the last 25 years, something that Krugman intentionally avoided mentioning. Which would effect the “security” of the average worker more, 10,000 workers being unemployed for 50 weeks, or 10 million workers being unemployed for 25 weeks? The latter obviously would have a much greater economic impact, although the “duration” as Krugman selectively cites is much greater in the former. The most accurate way of measuring this effect, in this context, is to measure the amount of long term unemployed in terms of the whole workforce, as I did on my blog, not just in terms of those unemployed. If you look at it in this manner, it is not only not “much longer” than 25 years ago, a claim even you cannot defend, but it trends significantly downward.

If Krugman were, for example, trying to examine the historical trends of unemployment duration in order to study certain factors such as education, workplace mobility, outsourcing, unemployment insurance etc. then such a way of adjusting would make sense, but he is not. Instead he is trying to make a political point, so he handpicks certain data that will back up his view. This is the biggest bone I have to pick with him. He used to be a serious economist, now he is nothing more than a political attack dog, like Michael Moore or Ann Coulter with an MIT degree. I know economics is called the “dismal science” but it is still a science. You are supposed to observe the data and use that to form theories, not come up with a theory based on a political ideology, and then selectively pick data to back it up, ignoring all common sense and contrary data.

[spencer]

How about thinking about looking to see if there are structural reasons for the duration of unemployment changing and second looking to see if the data supports that analysis rather then just playing with the numbers.

In the 1950s recessions were almost completely inventory adjustments in the durable goods sectors -- autos and housing. Workers losing their jobes in these industries typically expected to return to their old job. And for the most part they did. But now we have few lay-offs in this sense. Now, most workers lose their existing jobs and have to find a new job. In this context one would expect search time to lengthen,
and that is what the data shows.

[Ari]

James

Of course your argument, however lacking in substance, was addressed.

[James]

I was thinking more about Krugman's claim, and something dawned on me.

"Over the past 25 years the lives of working Americans have become ever less secure. Jobs come without health insurance; 401(k)'s vanish; corporations default on their pension obligations; workers lose their jobs more often, and unemployment lasts much longer than it used to. "

[The "lose their jobs more often" makes a distinction between temporary layoffs and permanent dismissals that you are missing. Paul's fine here.]

[howard]

no, James, it wouldn't be true: what makes you think it would?

[spencer]

Antler says that since 1983 the trend has been lower.

If you do a trend since 1983 the trend is positive --
the equation is y=0.003X + 15.426.

Even though Antler is right that the peaks have been lower, the bottoms have been higher. The regression results that gives the upward slope takes both into account. But it clearly shows that Antler's claim that the trend is down is incorrect.

[spencer]

James -- we have been over this many times. All you have to do is look at the labor force data and see that the participation rate is down.

Besides we are not talking about the unemployment rate, we are talking about the duration of unemployment.

[spencer]

To be techniocal, the unemployment rate by itself does not demonstrate if more of less people or losing their jobs. The unemployment rate is the product of several trends, including the numer of jobs lost and the number of jobs gained. But we do not have data on that back far enough to determine the history of whether or not job losses have increased or decreased.

[Jennifer]

James

Please do not be insulting to Brad. This is always uncalled for and makes a mockery of any argument you care to use. I thank Brad every day for the wonderful teacher he has been and is. Please, no more rudeness.