Simplistic WSJ minimum wage editorial

The quantitative masterminds at the Wall Street Journal editorial board are at it again.

Back in 2007, I showed how they cooked up a bogus graphic purporting to demonstrate that "Lower corporate tax rates with fewer loopholes can lead to more, not less, tax revenue from business":

Today, the Journal denounces the 2007 law that has increased the minimum wage over the last three years, writing, "Rarely has a law hurt more vulnerable people more quickly." To support this claim, it provides the following analysis:

The nearby chart compares the three-stage increase in the minimum wage with the jobless rate for teens age 16 to 19 since 2007. The first increase, to $5.85 from $5.15, came after a decade of no increases and when the overall jobless rate was below 5% and the teen rate was 14.9%. The demand for labor was sufficiently strong in many areas that most employers were probably willing to absorb the higher wage.

But as the minimum wage increased even as the overall job market began to worsen, the damage to teen job seekers became more severe. By the time the third increase to $7.25 from $6.55 took effect in July 2009, the teen jobless rate was 24.3%, and by October it peaked at 27.6% before dropping to 26.4% in January.

The story is even worse for black teens, who often have lower than average education levels or live in areas with fewer job prospects. Their jobless rate climbed from 38.5% before the third wage hike to 49.8% in November 2009, before falling back to 43.8% in January. For black male teens, the rate climbed to 52.2% in December from 39.2% in July. The difference between the jobless rates for black teens and the entire population widened by six percentage points from June 2007 to January 2010. Even assuming those rates fall as the job market improves this year, they will remain destructively high.

While it's certainly plausible that the increases in the minimum wage over the last three years have worsened teen unemployment, correlation doesn't prove causation. Any variable that trended in one direction during the current economic downturn will be correlated with the unemployment rate among teens or any other group.

More importantly, unemployment is rising across the board, which cuts against the WSJ's hypothesis that the minimum wage is having a particularly devastating effect on teens. To illustrate the point, here is a replication of the Journal's graphic with a line added to indicate the general unemployment rate (i.e. among everyone 16 and over) -- the lines are essentially parallel (correlations of .86-.97):

In fact, the seemingly dramatic increases in the unemployment rate among teens and black teens noted by the Journal (while disturbing) are actually smaller relative to the initial rates in those groups than among adults generally. To see this more clearly, I've plotted the data as a proportion of the unemployment rate in January 2007:

Finally, the sample size is very small, but a preliminary examination shows no obvious statistical evidence of a relationship between the minimum wage and the teen or black teen unemployment rates once we account for the upward trend in joblessness.

In short, the Journal could be right, but we can't tell from their analysis, which is below Econ 101 standards. It will take a far more systematic examination of the data to know what effect the minimum wage increases have had.

Update 3/5 2:42 PM: TNR's Jon Chait links with amusing comments.

Comments

[David]

I agree with Brendan that the WSJ editorial is below Econ. 101 standards. However, even a high school level analysis earns an A+ by newspaper editorial standards. E.g., NY Times editorials don't generally bother with any kind of analysis. The Times editors make prounouncements, as if they came directly from Mt. Sinai.

I think it's impossible to do a perfect economic analysis of minimum wage effects, because other factors are changing along with minimum wage. I've seen studies showing that black teenage unemployment was far lower 50 or 60 years ago. The minimum wage was lower then, but other conditions were different as well.

Personal note: As a teenager, I got fired from a summer job working for a friend of my father because the minimum wage was raised from $1.00 to $1.10. The owner came and told me I wasn't worth $1.10. For this reason, I tend to agree that minimum wage laws are pernicious, even if the WSJ analysis was insufficient to prove that point.

[Rob]

If I'd been fired from every job where I wasn't worth $1.10 an hour, I'd have starved.

[Boris Shor]

As David notes, the presence of a plot or table makes it easier to criticize convincingly, as Brendan has done. Critiquing words is a far messier business.

[David]

Amusingly Paul Krugman's NY Times column yesterday mocked and derided an economic principle enunciated by a Republican, even though that principle is presented in Krugman's own textbook. http://online.wsj.com/article/SB10001424052748703915204575103720332317434.html?mod=WSJ_Opinion_MIDDLETopOpinion

[David]

On further thought, I believe Brendan made a statistical error. He used a base of unemployment rate, but ought to have used a base of employment rate (100% - unemployment rate.) After all, it's the employed workers who can lose their jobs, not the unemployed. The question we're studying is how many (or what percentage of) employed workers lost their jobs.

I hope the following explanation isn't too unclear:

The WSJ measured the arithmetic change in unemployment rate, in number of percentage points. They showed that the change for black teenagers (BTA) was much larger than the change for the population as a whole. Brendan instead looked at the unemployment rate as a percentage of the 1/07 unemployment rate. That base rate was about 29% for BTA but only 5% for the general population. Subsequently, BTA unemployment rose by about 15 percentage points, and general population unemployment rose by about 5 percentage points.

By using ratios, Brendan made it appear that the rise in unemployment was lower for BTAs than for the general population, because general population unemployoment doubled while BTA unemployment only rose 50% or so.

To see why Brendan's approach was so biased, suppose he had used 2009 as a base. BTA unemployment was nearly 50%. The worst it could do is double. Meanwhile, the general population rate of 8% could go up by a factor of as much as 12 times.

Let's look at Brendan's figures using proper ratios.
-- In 1/07, BTA unemployment was 29%, so 71% of BTAs were employed.
-- In 2010 44% were unemployed so 56% were employed.
-- There was a drop of 15 percentage points in employment.
-- If we look at the job loss as a ratio to those employed, this drop meant that BTAs lost 15/71 = 21% of jobs.

That is, 21% of previously employed BTAs lost their jobs during this period. Meanwhile, for the general population, only 5/95 = 5.3% lost their jobs. Proper use of ratios shows that WSJ was actually conservative in their approach.

Note that Brendan's center graph ought to have shown the 3 employment rates, rather than the unemployment rates. That would have made clear just how badly BTAs were hurt.

[Rob]

David, I do believe you just took Brendan down to Chinatown.

[bnyhan]

David, I think that's wrong on a couple of points.

(1) Unemployment isn't just about people losing jobs; it's about the number of people in the labor market who are looking for jobs and can't find one, which can include new entrants or immigrants, etc. Thus David's calculations are incorrect. But in any case, I'm just following the WSJ in focusing on unemployment; I make no claims that this is the most appropriate way to study the issue.

(2) More importantly, the teen unemployment rate does seem to change in proportion to its base - i.e. a three-point increase in the general unemployment rate tends to have a larger effect on the teen unemployment rate, as my second graph suggested. See the graph here for data since 1948: http://oikonomeo.blogspot.com/2009/09/teenage-unemployment-followup.html

[David]

Brendan, I'm well aware of your point #1. I wrote the explanation that way to try to make my point more clear. Evidently I didn't succeed.

To see how deceptive it is to relate unemployment rate change to unemployment rate, consider the following example: Suppose

-- the base black teenage unemployment rate were 50%
-- the base Actuaries unemployment rate were 1%.

(In fact, these were pretty close to the actual rates in late 2009 -- an amazing fact in itself.)

If every black teenager then lost his/her job, so the BTA unemployment rate increased to 100%, then the BTA unemployemnt rate would double from 50% to 100%. Brendan's approach would say that this doubling was somehow comparable to an increase in the Actuaries unemployment rate from 1% to 2%. This is obviously a silly approach.

OTOH my approach would say that all BTAs losing their jobs (employoment rate goes from 50% to 0%) is comparable to all Actuaries losing their jobs (employment rate goes from 99% to 0%). In both cases, the

(change in unemployment rate)/(base employment rate)

equals the same 100%.

I agree with Brendan that an increase in unemployment is generally accompanied by a larger increase in teen unemployment (and probaly an even larger increase in black teenage unemployment.) However, I'm unclear why this point supports Brendan's criticism of the WSJ. The WSJ asserted that black teenagers are hurt worse than the general population when unemployment increases due to minimum wage increases. Brendan seems to not to disagree with this, but points out that black teenagers are also hurt worse when unemployment increases for reasons other than minimum wage increase. I'll buy Brendan's point, but it doesn't contradict the WSJ.

Brendan says he's just following the WSJ. To a degree he is. But, Brendan and Jonathan Chait (and Chait's commenters) are also mocking the WSJ, based on this different approach, which turns out to be incorrect. They think they're laughing at the WSJ, but the joke is on them.