Wednesday, July 27, 2005
Misleading with numbers. Or: Is today's labor market really as weak as many claim? Let's compare...
The current unemployment rate at 5.0% is significantly better than the average for the last 30 years. But the bad news bears out there aren't having any of that. They say that the job market really is in bad shape, and since it is that unemployment number must be hiding something.
Their explanation #1 is that things are so bad that millions of people have totally given up even looking work, and thus aren't counted in the unemployment rate. That's it ... in fact, that must be it as a matter of arithmetic if we are to have a better than average unemployment rate in a bad job market.
But that's not all they cite. Paul Krugman writes...
To begin with there's an obvious problem with the claim "the job market is not that good" or is "really weak" -- it must be so compared to something, and that something isn't given.
One's own subjective impressions really aren't persuasive (especially when one's impressions carry a noticeable political tinge). I, for one, am quite sure that I work far too many hours for much too little pay ... but you may disagree.
The first way to use numbers to mislead -- even to mislead oneself -- is to deprive them of context, not provide any perspective on them. Then one can say they prove anything at all, and the credulous among one's audience who don't bother to check for themselves may buy it.
To avoid this trap let's gain some perspective by comparing today's labor market conditions today to those at the same point of the prior business cycle after the 1991 recession, the same amount of time after it's end. That's in October of 1994, just about halfway through Bill Clinton's first term.
Comparing business cycles at the same point is what economists do to avoid errors (as well as intentional misrepresentations) that result from the like of comparing peaks to troughs.
And we all know the prior business cycle was one of the all-time best -- leading into what Paul Krugman has called the "miracle economy". So it provides us with an objective standard by which we can measure whether the current job market is "good" or "not so good". If the current labor market is as good as the prior one at the same point of its progress towards the miracle economy, how bad can it be?
OK, here are the numbers for the five indicators cited by Brad DeLong, for both this recovery (as of June 2005 or latest available) and the prior recovery at the same length of time into it:
Unemployment rate
Prior recovery: 5.8%
This recovery: 5.0%
Score 1-0 for this recovery.
Long term unemployment rate (15+ weeks)
prior: 2.25%, which is 40% more than
this: 1.58%
Long-term unemployment at this point in the prior recovery was higher even in absolute terms, 2.96 million versus 2.35 million, in spite of the work force being 17 million larger today. So the current recovery is much better on this point.
Score 2-0 for this recovery.
average real weekly earnings from pre-recession high
prior: -1.9%
this: +0.25%
In fact, during the prior recovery real average weekly wages would stay below their 1990 pre-recession high for seven years, until well into Clinton's second term.
So a small gain today is decried as weakness -- while a real decline that lasted more than seven years was the happy course to the "miracle economy".
Now maybe we get an idea about why the bad news bears always present their numbers without any context, without the perspective of any comparison to the prior business cycle. In making such a comparison using DeLong's five indicators the current recovery is up 3 to 0 right now -- already we have a winner!
As a brief aside I'll mention here a cherry-picking tactic near always used by the class warriors to make it seem that workers are being paid less, and gaining less, than they really are.
This is the constant quoting of "earnings" or "wage" numbers in newspaper stories, op-eds and the like, when the average reader doesn't realize that these are only subsets of the amounts earned by workers, excluding among other things the benefits that comprise an ever growing part of compensation. They thus systematically understate both the amount and rate of growth of employee compensation.
For example...
real hourly total compensation from pre-recession high
prior: +1.9%
this: +5.6%
Cherry picking, of course, is the second prime way to mislead with numbers.
Now if Krugman was the sort of guy who'd ever been accused of having "the disturbing habit of shaping, slicing and selectively citing numbers", one might think he picked the less meaningful number for "earnings" rather than the more comprehensive and meaningful number for "total compensation" simply because one can't call +5.6% "flat".
To get back on track... The trump card (only card?) held by the bad news bears is labor force participation rate -- the percentage of people who are either working or looking for work. It's what they say is so bad today -- millions of people have left the labor force.
So this comes as something of a surprise...
labor force participation, age 20+
prior: 67.8% (unemployment rate 5.0%)
this: 67.8% (unemployment rate 4.5%)
Whoa! Surely when the bears talk of millions of people abandoning the search for work they give the impression that they are talking of adults supporting themselves and their families. But it's not so.
The entire drop in the labor force participation rate has been among teenagers -- something I've never seen any bear note even once.
Using the adult rate we have an exact tie between the two recoveries, making the "DeLong Indicators" score 3.5 to 0.5.
But let's make the bears happy by doing what they want -- computing a "real" unemployment rate for today by adding all those who have left the workforce to the number of "unemployed", as dubious an idea as this may be.
total labor force participation rate
prior: 66.7%
this: 66.0%
That's a 1% difference -- and under the bears' methodology it moves the unemployment rate for today from 5% up to a "real" 6%.
Comparing this 6% rate to the 5.8% of the prior recovery, and noting the 0.3% margin of error in the unemployment rate, we find the two rates are statistically the same. So doing their worst the bears do no more than wind up with the same unemployment rate for the same time in the two recoveries. A push.
And that is the worst possible case for today, because a decline in the labor force participation rate need not be a bad thing -- especially for teenagers.
To the extent it represents something good -- perhaps teenagers are finally beginning to listen to their elders and deciding to stay in school longer to get better jobs, regardless of the state of the job market -- then we have a push no more, and the current labor market is better again.
So ... are teenagers dropping out of the work force due to discouragement because they can't get jobs? Or because they have decided to stay in school instead?
Fortunately for us the Bureau of Labor Statistics has investigated this and tells us in Declining Teen Labor Force Participation ...
Using the July rate cited by the BLS, this good cause reduced their participation by more than 3 points during the boom. Assuming that the same trend would have continued absent a recession (and without an historic boom this trend should have been stronger) then even granting "discouragement" among teenagers due to the recession, we can conclude conservatively that about half the total decline in the teenage participation rate since 1994 is due to this good cause and should not be counted as "unemployment". This reduces the bad news bears' "real" unemployment rate from 6.0% to 5.5%, again less than the corresponding actual rate of the prior recovery.
And this itself is a worst case at another level. For if an adult with the need to support self and family is unemployed due to a weak economy, that's truly bad. But if a teenager decides to gain further education due to a weak job market, is that really comparably as bad? I don't think so.
So ... with the very same participation rate and a better unemployment rate for everyone over 20, and a reduced participation rate for teens that is at least in substantial part due to beneficial causes, I'm going to score this as a "push" in the "DeLong Indicators" tally, bringing the score to 3.5 to 0.5 in favor of the current recovery over the prior one.*
Now here are two challenges for those selling the "millions of discouraged workers" meme...
1) Compared to this point in the prior recovery, with unemployment lower today among those looking for work, long-term unemployment much lower, and pay having risen faster, why would so many more people be discouraged about finding a job? What's discouraging them?
2) If there really are so many more discouraged workers today, why have they stopped describing themselves as such to the BLS?
Unemployed and discouraged workers as a percent of the labor force and discouraged workers
Prior: 6.1%
This: 5.3%
There's a puzzle to solve, if there are so many more discouraged workers today.
Ooops, with the contest settled I almost forgot the fifth indicator cited by Prof. Delong, payroll employment. This can be quickly ceded even by fans of the current recovery -- payroll indeed grew faster during the prior one.
The significance of this, however, is dubious. It ignores the unique nature of the 2001 recession, which involved the bust of an employment bubble as much as of a stock market bubble -- with payroll employment entering 2001 about two million over the long-term trend line. To say that employment should have continued upward at the "normal" rate after the end date of the recession is little different than saying the stock market should have continued upward at the "normal" rate of 7% per year after the trough of the recession, and so should today be 30% or so higher than at the end of 2001 -- in spite of the stock bubble.
Even Prof. Delong himself has stated this elsewhere...
So our final tally is 3.5 to 1.5 in favor of the labor market of today over that at the same point in the prior "Clinton recovery", which was on track to the miracle economy and full-fledged boom. That is context and perspective.
Note, I'm not saying the job market today is "strong" -- strong was from 1997 to 2001.
But compared to the prior business cycle, which was just about the best on record, it is at least as strong -- and how many times have you heard the bad news bears of the left econosphere say that?
To imply otherwise as they do, perhaps innocently, by omitting context and perspective -- and also frequently do, um, disingenuously, by comparing today's economy to the boom years while using the boom as a base line norm** -- is simply wrong and misleading.
The labor market today is operating today in "normal" territory. Not very weak, not very strong, in the middle range, pretty much like at this time in the last business cycle ... only somewhat better.
~~~~~~~~
* Ah, but what about the much quoted Bradbury policy brief (.pdf) and its statement...
Well, the curious thing about those above whom I've read (I haven't gone through the whole list) who've quoted this "8.7%" number is that not one of them has also quoted Bradbury's own acknowledgment that this number might be wrong, too high, because it is based on projecting forward from an atypical labor market boom, as noted above.
If fact, projecting "normal" labor force growth from 1996, Bradbury says the bears' "real" unemployment rate increase could be as little as 1.3 points, giving a 6.3% rate for today.
And if Bradbury instead of selecting the rather arbitrary 1996 had used the same-point-in-the-cycle 1994, she'd be right back to what's been described above.
** Paul Krugman is relentless at making these boom-as-a-baseline comparisons, of course. As, for example, in this very column...
Hey, what "disturbing habit of shaping, slicing and selectively citing numbers"? ;-)
The current unemployment rate at 5.0% is significantly better than the average for the last 30 years. But the bad news bears out there aren't having any of that. They say that the job market really is in bad shape, and since it is that unemployment number must be hiding something.
Their explanation #1 is that things are so bad that millions of people have totally given up even looking work, and thus aren't counted in the unemployment rate. That's it ... in fact, that must be it as a matter of arithmetic if we are to have a better than average unemployment rate in a bad job market.
But that's not all they cite. Paul Krugman writes...
... adjusted for inflation, average weekly earnings have been flat for the past five years ... As Berkeley's J. Bradford DeLong writes on his influential economics blog, "We have four of five indicators telling us that the state of the job market is not that good and only one - the unemployment rate - reading green."... quoting Brad DeLong's ....
Four Out of Five Indicators Say the Job Market Really Is WeakSuch claims can be found today all over the left-liberal side of the econo-blogosphere. But are they true?
It's not just employment-to-population ratios. It's real wage growth. It's the relative amount of long-term unemployment. It's payroll employment. We have four of five indicators telling us that the state of the job market is not that good and only one -- the unemployment rate -- reading green.
To begin with there's an obvious problem with the claim "the job market is not that good" or is "really weak" -- it must be so compared to something, and that something isn't given.
One's own subjective impressions really aren't persuasive (especially when one's impressions carry a noticeable political tinge). I, for one, am quite sure that I work far too many hours for much too little pay ... but you may disagree.
The first way to use numbers to mislead -- even to mislead oneself -- is to deprive them of context, not provide any perspective on them. Then one can say they prove anything at all, and the credulous among one's audience who don't bother to check for themselves may buy it.
To avoid this trap let's gain some perspective by comparing today's labor market conditions today to those at the same point of the prior business cycle after the 1991 recession, the same amount of time after it's end. That's in October of 1994, just about halfway through Bill Clinton's first term.
Comparing business cycles at the same point is what economists do to avoid errors (as well as intentional misrepresentations) that result from the like of comparing peaks to troughs.
And we all know the prior business cycle was one of the all-time best -- leading into what Paul Krugman has called the "miracle economy". So it provides us with an objective standard by which we can measure whether the current job market is "good" or "not so good". If the current labor market is as good as the prior one at the same point of its progress towards the miracle economy, how bad can it be?
OK, here are the numbers for the five indicators cited by Brad DeLong, for both this recovery (as of June 2005 or latest available) and the prior recovery at the same length of time into it:
Unemployment rate
Prior recovery: 5.8%
This recovery: 5.0%
Score 1-0 for this recovery.
Long term unemployment rate (15+ weeks)
prior: 2.25%, which is 40% more than
this: 1.58%
Long-term unemployment at this point in the prior recovery was higher even in absolute terms, 2.96 million versus 2.35 million, in spite of the work force being 17 million larger today. So the current recovery is much better on this point.
Score 2-0 for this recovery.
average real weekly earnings from pre-recession high
prior: -1.9%
this: +0.25%
In fact, during the prior recovery real average weekly wages would stay below their 1990 pre-recession high for seven years, until well into Clinton's second term.
So a small gain today is decried as weakness -- while a real decline that lasted more than seven years was the happy course to the "miracle economy".
Now maybe we get an idea about why the bad news bears always present their numbers without any context, without the perspective of any comparison to the prior business cycle. In making such a comparison using DeLong's five indicators the current recovery is up 3 to 0 right now -- already we have a winner!
As a brief aside I'll mention here a cherry-picking tactic near always used by the class warriors to make it seem that workers are being paid less, and gaining less, than they really are.
This is the constant quoting of "earnings" or "wage" numbers in newspaper stories, op-eds and the like, when the average reader doesn't realize that these are only subsets of the amounts earned by workers, excluding among other things the benefits that comprise an ever growing part of compensation. They thus systematically understate both the amount and rate of growth of employee compensation.
For example...
real hourly total compensation from pre-recession high
prior: +1.9%
this: +5.6%
Cherry picking, of course, is the second prime way to mislead with numbers.
Now if Krugman was the sort of guy who'd ever been accused of having "the disturbing habit of shaping, slicing and selectively citing numbers", one might think he picked the less meaningful number for "earnings" rather than the more comprehensive and meaningful number for "total compensation" simply because one can't call +5.6% "flat".
To get back on track... The trump card (only card?) held by the bad news bears is labor force participation rate -- the percentage of people who are either working or looking for work. It's what they say is so bad today -- millions of people have left the labor force.
So this comes as something of a surprise...
labor force participation, age 20+
prior: 67.8% (unemployment rate 5.0%)
this: 67.8% (unemployment rate 4.5%)
Whoa! Surely when the bears talk of millions of people abandoning the search for work they give the impression that they are talking of adults supporting themselves and their families. But it's not so.
The entire drop in the labor force participation rate has been among teenagers -- something I've never seen any bear note even once.
Using the adult rate we have an exact tie between the two recoveries, making the "DeLong Indicators" score 3.5 to 0.5.
But let's make the bears happy by doing what they want -- computing a "real" unemployment rate for today by adding all those who have left the workforce to the number of "unemployed", as dubious an idea as this may be.
total labor force participation rate
prior: 66.7%
this: 66.0%
That's a 1% difference -- and under the bears' methodology it moves the unemployment rate for today from 5% up to a "real" 6%.
Comparing this 6% rate to the 5.8% of the prior recovery, and noting the 0.3% margin of error in the unemployment rate, we find the two rates are statistically the same. So doing their worst the bears do no more than wind up with the same unemployment rate for the same time in the two recoveries. A push.
And that is the worst possible case for today, because a decline in the labor force participation rate need not be a bad thing -- especially for teenagers.
To the extent it represents something good -- perhaps teenagers are finally beginning to listen to their elders and deciding to stay in school longer to get better jobs, regardless of the state of the job market -- then we have a push no more, and the current labor market is better again.
So ... are teenagers dropping out of the work force due to discouragement because they can't get jobs? Or because they have decided to stay in school instead?
Fortunately for us the Bureau of Labor Statistics has investigated this and tells us in Declining Teen Labor Force Participation ...
... the labor force participation rate for teens dropped from 65.4 percent to 62.3 percent between 1994 and 2000. This happened even as the unemployment rate for teenagers was falling to its lowest level in three decades. Thus, adverse job market conditions, the usual explanation for declining teen work activity, were not the cause during this period.... and concludes that indeed the teen labor force participation rate has been dropping because more teens have decided to stay in school for longer periods.
Using the July rate cited by the BLS, this good cause reduced their participation by more than 3 points during the boom. Assuming that the same trend would have continued absent a recession (and without an historic boom this trend should have been stronger) then even granting "discouragement" among teenagers due to the recession, we can conclude conservatively that about half the total decline in the teenage participation rate since 1994 is due to this good cause and should not be counted as "unemployment". This reduces the bad news bears' "real" unemployment rate from 6.0% to 5.5%, again less than the corresponding actual rate of the prior recovery.
And this itself is a worst case at another level. For if an adult with the need to support self and family is unemployed due to a weak economy, that's truly bad. But if a teenager decides to gain further education due to a weak job market, is that really comparably as bad? I don't think so.
So ... with the very same participation rate and a better unemployment rate for everyone over 20, and a reduced participation rate for teens that is at least in substantial part due to beneficial causes, I'm going to score this as a "push" in the "DeLong Indicators" tally, bringing the score to 3.5 to 0.5 in favor of the current recovery over the prior one.*
Now here are two challenges for those selling the "millions of discouraged workers" meme...
1) Compared to this point in the prior recovery, with unemployment lower today among those looking for work, long-term unemployment much lower, and pay having risen faster, why would so many more people be discouraged about finding a job? What's discouraging them?
2) If there really are so many more discouraged workers today, why have they stopped describing themselves as such to the BLS?
Unemployed and discouraged workers as a percent of the labor force and discouraged workers
Prior: 6.1%
This: 5.3%
There's a puzzle to solve, if there are so many more discouraged workers today.
Ooops, with the contest settled I almost forgot the fifth indicator cited by Prof. Delong, payroll employment. This can be quickly ceded even by fans of the current recovery -- payroll indeed grew faster during the prior one.
The significance of this, however, is dubious. It ignores the unique nature of the 2001 recession, which involved the bust of an employment bubble as much as of a stock market bubble -- with payroll employment entering 2001 about two million over the long-term trend line. To say that employment should have continued upward at the "normal" rate after the end date of the recession is little different than saying the stock market should have continued upward at the "normal" rate of 7% per year after the trough of the recession, and so should today be 30% or so higher than at the end of 2001 -- in spite of the stock bubble.
Even Prof. Delong himself has stated this elsewhere...
Asking that Bush leave "the job market no worse than he found it" is setting the bar too high by perhaps 2 million jobs or so. It would have been very hard for any set of economic policies to have sustainably kept the economy at the high-pressure state it was in in 2000... and keeping that fair caveat in mind he can collect the full point for the payroll employment indicator, for all it's worth.
So our final tally is 3.5 to 1.5 in favor of the labor market of today over that at the same point in the prior "Clinton recovery", which was on track to the miracle economy and full-fledged boom. That is context and perspective.
Note, I'm not saying the job market today is "strong" -- strong was from 1997 to 2001.
But compared to the prior business cycle, which was just about the best on record, it is at least as strong -- and how many times have you heard the bad news bears of the left econosphere say that?
To imply otherwise as they do, perhaps innocently, by omitting context and perspective -- and also frequently do, um, disingenuously, by comparing today's economy to the boom years while using the boom as a base line norm** -- is simply wrong and misleading.
The labor market today is operating today in "normal" territory. Not very weak, not very strong, in the middle range, pretty much like at this time in the last business cycle ... only somewhat better.
~~~~~~~~
* Ah, but what about the much quoted Bradbury policy brief (.pdf) and its statement...
While the official unemployment rate has fallen from a peak of 6.3% in June 2003 to 5% in June 2005, the labor force participation rate remains close to 15-year lows of 66% .... If labor force participation rates had improved as much during this recovery as typical, between 1.6 million and 5.1 million more people would be in the labor force ... If those people were counted in the labor force but not working, the jobless rate would have been somewhere between 6.5% and 8.7%...... that's been cited by bad news bears everywhere (including, according to Econobrowser, "Brad DeLong, Paul Krugman, Angry Bear, Economist's View, William Polley, [...more...] among others").
Well, the curious thing about those above whom I've read (I haven't gone through the whole list) who've quoted this "8.7%" number is that not one of them has also quoted Bradbury's own acknowledgment that this number might be wrong, too high, because it is based on projecting forward from an atypical labor market boom, as noted above.
If fact, projecting "normal" labor force growth from 1996, Bradbury says the bears' "real" unemployment rate increase could be as little as 1.3 points, giving a 6.3% rate for today.
it might be unrealistic to expect participation to rebound cyclically relative to such high base levels ... narrowing the gap by using 1996 participation rates rather than those of March 2001 as the basis of comparison for "change from peak,” [increases] the unemployment rate by 1.3 percentage pointsTo quote the "8.7%" with high enthusiasm while not bothering to mention this caveat about it fails the candid honesty test, one might think.
And if Bradbury instead of selecting the rather arbitrary 1996 had used the same-point-in-the-cycle 1994, she'd be right back to what's been described above.
** Paul Krugman is relentless at making these boom-as-a-baseline comparisons, of course. As, for example, in this very column...
even the most favorable measures show that employment growth has lagged well behind population growth over the past four years.... four years ago being when the employment ratio reached it's boom all-time high, a good two million above sustainable levels according to Krugman's own good friend and cited authority Prof. DeLong.
Hey, what "disturbing habit of shaping, slicing and selectively citing numbers"? ;-)