Teen employment and the minimum wage: sixty years of experience

Kevin Erdmann relates:


There is much more here.  Kevin concludes: “Is there any other issue where the data conforms so strongly to basic economic intuition, and yet is widely written off as a coincidence?”


A couple things jump out - first is the lack of a Y axis label, second is the inconsistent increments on the X axis. I also don't understand the "during a minimum wage hike" terminology. Did minimum wages rise every year from 74 - 82?

Until the late 70s the correlation between MW hikes and whatever the Y axis is appears relatively short lived and I think particularly after 1999 there is an overall trend that the graph does a poor job with. The two year rise in 06-07 looks like a blip in a 12 year decrease, but he chose that as a trend. The 99-04 slope looks to my pre-coffee eye very similar to the 07-11 slope.

Trend line definition (from the linked post [my comment]):
* Pre-trend lines [black] are for period from 27 to 3 months before MW hike.
* MW Trend lines [red] are for period from 3 month before to 27 months after initial MW hike.

As to if MW was really increasing in the shaded periods, see:

"for period from 27 to 3 months...for period from 3 month before to 27 months"

I don't understand why all the trend lines are different lengths if each signifies a 30 month period.

Nevermind, they're all the same horizontal length--haven't had my morning coffee yet.

"27 to 3 months before MW hike" = 24 months
"3 month before to 27 months after initial MW hike" = 30 months

Did you mis-type and mean to write "27 months before to 3 months after"?

Also, why do some trend lines above appear to be much greater lengths of time than either? There is one red trend line that appears to go from slightly before 9/1/1990 to 1/1/1994. That's more than 30 months.

the x-axis increment is an unusual (?) but constant 20 months

Thanks - I missed that earlier.

And what is this "during minimum wage hikes" bit about when the minimum wage has been consistently raised? That part would only make sense if there were periods of time when they the minimum wage was lowered. This chart is bollocks.

Several minimum wage hikes were phased in over several years, like the most recent one. Those are shaded.

The 7 times where the law changed were used as a natural experiment on the impact of the minimum wage by viewing the change in law as an exogenous shock to the low wage labor market (approximated by the teenage labor market)

The finding of the post / experiment is that there was almost uniformly (6/7) a dramatic downward inflection point (change in slope) in low wage employment coinciding with the change in minimum wage legislation.

Pardon for the long response ;

Sometimes one has nothing to say except to point to something someone else said. To this point, beowulf, a blogger and commenter much respected in MMT and Modern Monetary realism (MMR) circles, added a number of lively comments about the desirability of raising the minimum wage that I think are worth blogging here. He said:

“Minimum wage laws are like hummingbird wings. In theory they shouldn’t work at all, in the real world they work pretty well.
Australia’s minimum wage was just bumped to A$15.96 an hour, US$16.84.hr at today’s exchange rate. Unemployment rate is 5.2%.
Think about that, their U3 rate is three points lower AND their minimum wage is more than double ours. Either the Coriolis effect makes neoclassical economics work backwards in the Southern hemisphere, or mainstream economists are a bunch of astrologers who think they’re Carl Sagan.
$16.84/hr is high enough that a full time worker making that here would be means-tested out of food stamps, section 8 and other income security programs.
So what’s going on is ***Australia puts the cost of a living wage for the working poor on their employers instead of taxpayers***, enabling govt spending to be focused on other needs– like universal Medicare and a Social Security system so broad it would impress even Rodger Mitchell.”

*** (mine)

Then later on he added:

“One other thing, this John Stossel post last month may be the most mendacious thing I’ve read all year.

—-”Statists say that Australia is proof that minimum wage laws help workers. They point to Australia’s 5.1% unemployment rate… But statists ignore the details.
“Most people who earn minimum wage are young, unskilled workers. How are they doing in Australia? In June, Australia’s unemployment rate for workers age 15 to 19 was 16.5%.”—–

“That’s digging pretty deep for an unemployment stat. Curious that Stossel neglected to mention the comparable US stat (for workers 16 to 19). In June, their unemployment rate was…26.6%.
That Coriolis effect is CRAZY.”

And then he added a bit more:

”OK, this is really the last one…
“According to the Heritage Foundation/WSJ “2011 Index of Economic Freedom”, Australia’s “government spending as a percentage of gross domestic product (GDP)” is less than that of the United States; 34.3% vs 38.9%.
“Stop and think about that… universal Medicare (with dental!), a jumbo size Social Security system, a defense policy of jumping into the same wars we do (including Vietnam and Iraq) and Australia still spends less on government than we do. At risk of sounding hyperbolic, I”d say that Coriolis effect is strong enough to move hurricanes (and cyclones). :o)”

I can add nothing to that comment…… maybe greatest internet comment of all time!

This seems pretty weak. Take away the recessions in 74, 90 and 07 and it's hard to see the significance of the correlation. Maybe the lesson is that hikes in the minimum wage are more likely to come at the top of the economic cycle.

I agree, there are recessions during a lot of these periods. Too many confounding factors, I mean I suppose now we're going to have peope claiming that minimum wage increases caused these recessions (and not Fed interest policy or Asset bubbles bursting).

Eyeballing, there is a greater than 45 degree drop in trend line (which to me is a GIGANTIC effect in the policy world) in every instance except for 97 (only positive change) and 62 (still a noticeable drop in trend and still a larger negative angle than 97 is positive).

But even 1967 looks like just some kind of wierd blip.

This was my first thought as well, and I tend to believe the economic intuition at issue.

Good point. Even if there was a strong correlation, my response would be "so what?" Hikes in the minimum wage are there to benefit people earning minimum wage plus margin as well as people earning just the minimum wage. Rising tide lifts all boats.

I don't understand. In the case of a strong correlation between raising the minimum and reduced employment, it would be hurting some people it was intending to help, and definitely not lifting all boats. Whence the "so what?"?

This is a correlation between higher minimum wage and reduced TEEN employment. So, again, "so what"? I don't care about teens--I care about everyone. And this chart doesn't tell me what happens to everyone, but tons of studies about minimum wage NOT reducing unemployment, do. http://www.cepr.net/documents/publications/min-wage-2013-02.pdf

You care about everyone, but not teens who comprise everyone. Let me know how that works out for you.

"Rising tide lifts all boats."

The whole point of this is that raising the minimum wage hurts some employees very much (the one's who lose their job or don't get hired) and helps others a moderate amount (the one's who were closest to minimum wage).

So, it most explicitly is not a case of "a Rising tide lifts all boats".

"The whole point of this is that raising the minimum wage hurts some employees very much (the one’s who lose their job or don’t get hired)"

And this point remains unproven. This chart doesn't even implicate that. At all.

Actually, you mean a rising tide lifts most boats, but sinks a predictable number of others. But hey, a single death is a tragedy, but a million is just a statistic.

You should attribute this. I believe it was J. Stalin.

Thanks for my first /facepalm of the day.

It is about appearing to care about the poor, even if you do more harm than good.

Can't speak for Michael. But the argument is often made that raising the minimum wage helps teens (who may be well-off), not the poor. (http://www.heritage.org/research/reports/2013/02/who-earns-the-minimum-wage-suburban-teenagers-not-single-parents) If the goal of the policy is to help poor single parents (as Heritage and others suggest) then data showing that it hurts teens seem off-base, and even undercuts a standard criticism of the policy.

And thus prevents more boats from ever getting in the water.

Every two variable analysis is instantly suspect.

Yes, the effect of the recessions on employment certainly confounds the analysis - omitting important variables causes serious bias. But the correlation serves as a suggestion of what might be causing it. I agree that the evidence is certainly refutable.

On the other hand, another peculiarity jumps out at me. Note that Congress tends to raise minimum wage just before recessions. What does that tell us?

It doesn't mean that raising minwage causes recessions, but rather in the course of a business cycle, Congress sees everyone doing well and it declares that those at the bottom must also do well. They do this just in time to catch the wave on the downroll.

Aside from being poorly timed, pro-cyclical policy, the policy seems to be reacting to fleeting conditions.

It is time to take the keys from these drunk drivers.

So this is why they should just link the thing with inflation and then it wouldn't end up getting raised haphazardly by a slow reacting Congress.

I dont agree with the policy, but it we have the policy, linking it to inflation seems to be the right way to go about it.

On the other hand, inflation itself is part of the business cycle, so that might exacerbate the problem - especially if priced are sticky downward. The key to eliminating unemployment is facilitating the repricing of labor downward during recessions.

I wouldn't mind that, but the political battle would be over what level to start with. If you adjust for inflation, there is already an enormous difference between an inflation-adjusted minimum wage based on the original starting point (1938) which is about $4.07 in 2012 dollars, versus the version that I see the left advocate, which is to adjust based on the highest level in the history of the minimum wage (1968), which would be $10.56 in 2012 dollars.

the biggest declines occurred during recessions and the other instances are flat or rising. not very convincing.

The point is not that employment declines, it's that there's a marked downward shift in the trend line, eg from sharply up to flat. (Black is above red)

Admittedly, to control for recession like effects, one would like to see comparative trend lines for other labor segments not as closely tied to MW.

But theory (supply/demand) is unusually (for macro) well matched to data in this case.

Closely related - visualizing the last 20 years of the minimum wage in the U.S. Complete with a demand curve!

And if anyone's upset by our excluding the years between 1997 and 2003 from the regression analysis, the years of the Dot-Com Bubble in the U.S. stock market, we're going to directly address why that's appropriate later this week....

(1) Explain to me how 2002 and 2003 (post-9/11) are part of the Dot-Com Bubble. (2) Refute the alternative model that technology made a permanent rightward shift in demand that was then shifted back by 9/11 (see 2001-2002)

Bear in mind that my prior is "Raising the US minimum wage is ineffective given a global labor market and increasing automation." I'm just poking at the graph.

1) The dot com bubble has a trough at roughly mid 2002. In other words, they aren't part of the bubble, but rather signify the bursting of said bubble.

2) Provide a model to refute

I want to see the percentage of households with a less than living wage (or maybe below the poverty line) on top of this. Maybe if minimum wage grew in a more stable manner, various types of unemployment would be more stable.

Great. Where's your crystal ball?

Doesn't this call into question the whole idea of economics as a science? If you can't even get agreement on something this basic?

Kent hits the nail on the head.

Do a one-word replacement and you get the exact argument of a global warming skeptic.

What about this is basic?

If the price of something goes up, people buy less of it.

Minimum wage jobs are basically entry-level jobs. Youth unemployment is high. It's more important that 10 people catch on in the job market than that 9 people make an extra $1 per dollar for their first six months while the tenth is shut out.

I think you misunderstand what "economics" actually is. It is the act of taking a blindingly obvious truth, and desperately trying to find (and twist) information so as to persuade others that it isn't true. To wit:

>If the price of something goes up, people buy less of it.

This is problematic, because raising the minimum wage makes liberals feel really, really good. It just does. Twice actually, because they also get to call their opponents evil when they point out the obvious truth you mention.

Liberal "economists" are therefore highly-paid to tell everyone that if the price of something goes up, people will buy the same amount of it. Maybe even more! Their readers nod sagaciously, feel great, and go about their day. Unemployed teens play video games and get another bowl.

Yeah, working at McD's is the start of a bright career.

I can't stand this sort of elitist bullshit. Most people on this comment board got a nice degree from a nice university. That is not how the other half lives.

Hey Average if Over, a nice degree from a nice university is just the ticket to future work at McDonalds these days

It certainly can be. Ask the people in McDonald's management.

It may have not worked out for you, but others have had different results. On a serious note, it is weird how an upper class trait has seeped down into the managerial class. In another era, elites did not want to see people toiling. They had a class of servant who supervised those who tilled the earth, prepared the food, made the hard goods and so forth. The disdain for the horny handed sons of toil was not just about the people, but a disdain for their work. Today, the mediocrities in the managerial class ape the same attitude. It's not that they are too good for these jobs; they are too good to even look upon the jobs being performed!

It was for me.

Exactly where does a job at McD's lead? I had a job like that as a teenager. Did shit for my career. Would have been better off studying harder or even socializing more. Both would have been better for my long run career.

The people I worked with at these jobs weren't going anywhere. They were going to the next shit job that was just the same. You learn no skills and it actually wears you down as a person.

In the event somebody actually would use that job to become a manager or something do we really expect that person to get laid off because of the minimum wage? If they are going to cut people at jobs like that there is a vast amount of functionally retarded people that show up to shifts high to fire first. Bootstrap McGoodworker isn't going to be the first one out the door.

To be honest, I don't want my kids to work fast food (like I did), I'd rather they cut lettuce or cleaned toilets.

But I will be strongly encouraging them to hold a sh*t job at least once in their lives. Might be a different country right now if every American had, and did.

Might not.

asdf, you might not be aware, but McDonald's is very big on hiring from within, and particularly on hiring management and corporate positions from people who worked in the restaurants. If you have commitment, you can turn it into a very good career. I know a guy who did it.

No, because:

1) Policies are often calibrated to avoid blatant mistakes. So the debate continually shifts to stay within the threshold of statistical uncertainty.

2) We can't run experiments to test the counterfactual.

By comparison, what kind of agreement do we get on healthy diet?

"2) We can’t run experiments to test the counterfactual. "

If we were to pass a minimum wage rate, the opponents should insist it take effect all at once. They won't for various reasons, the biggest being they'll attempt to delay the pain, but it would be better in the long run to settle the point. One of the reasons this is always at the margin of uncertainty is that the raises are generally stretched out of multiple years. If the minimum wage hike were take effect the month after it passed at the full amount, you'd be able to sort the signal out from the noise to a much greater degree.

Isn't the counterfactual the periods of growing employment when there was no change in the minimum wage?

No. It is difficult to get agreement because there are lots of incentives for people to promote bad economic policy that deny basic economic theory.

Actually I think the whole minimum wage debate does, and this chart actually reinforces the fact that these "general economic principles" do not in fact hold in general. We see instances in this chart where increasing the minimum wage doesn't appear to correlate with drops in employment, we see in the late 70s and int he 90s employment increasing while the minimum wage is continuing to be raised.

The most important thing I see from this chart is that historically even if the minimum wage raise in accompanied by some initial dip in employment, the employment situtation recovers and surpasses the old high with more workers being employed at new higher wages. Teen employment peaks in the late 1970s and never regains that peak but perhaps this is due to changes in society. This trend is breaking down in the 2000s, which have basically been a crappy decade and a half for job creation anyway as it corresponds to an acceleration of globalization dumping legions of VERY low wage workers onto the world market hence eliminating a huge number of US jobs.

One of the big reasons why, at least in my opinion, raising the minmum wage has started to become a big issue on the left is because the old mechanisms by which workers would force employers to increase wages (collective bargaining and unions) and actually make gains have completely broken down. The raising of the minimum wage might not be ideal but it appears to be the last option left.

"I dropped this coin while the rocket was taking off, and it kept going up. Gravity: disproven"

Nonsense, and don't compare economics to the sciences. I mean look at this chart even, it's not at all consistent that employment is going down when the minimum wage goes up, it happens sometimes but there's exceptions and blah blah blah blah. Hell you have 7 Nobel prize winning economists claiming that the minimum wage has little to no effect on employment. I don't see Nobel Prize winning physicists denying gravity.

Seems like economics just doesn't come to conclusions especially when we are discussing macro policies.

Seven Nobel prize winners who didn't make their careers studying minimum wage and whose political/economic ideology demands denial of objective reality. This, by the way, is an Appeal to Expertise -a logical fallacy.

Science is the process of observation and prediction with suitable controls to isolate the factor under scrutiny. Human behavior is constantly observed and quite predictable. Entire fields of study (psychology, anthropology, sociology, economics, finance, linguistics, etc) are based on this.

No no I'm not appealing to them, you misunderstand. I'm not saying that they're right but for all the people claiming economics is a legitimate science and that this is basic basic economic theory, and it's a slam dunk - how do you explain this?

The denial of objective reality line is just more BS, like only conservative economists are truth tellers - the entire field is just a mess and I can't accept any of this "basic economic theory" when some of the field's leading practitioners can't seem to agree.

Nonsense, and don’t compare economics to the sciences.

Aside from pointing out that one thing can be bigger than another in economics and sciences, I wasn't, really. Simple and obvious point was simple and obvious.

Yes, we'd better not think about this question the same way we think about physical sciences, because then we'd see the very strong negative impact on employment, and realizing that increases in the minimum wage increase unemployment would be A Bad Thing

But we don't see that at all from this chart. At best we see some dips when the minimum wage is raised (some of those dips are due to recessions) and we see a recovery from this and an overall strong increasing trend in Teen unemployment EVEN THOUGH the minimum wage is increasing, until the late 1970s. We see a big downward drop, also corresponding with the time of the early 1980s recession and then weaker growth as we enter the period of neoliberalism and globalization.

This Chart does not say much at all in fact, and as someone who thinks that raising the minimum wage might be a good way to help generate higher wages amongst even somewhat higher paid workers, the fact that we see recovery after the initial dip doesn't really lead me to believe this is a very very bad policy.

Well no shit you can't see that one factor is bigger than another from a chart that charts one thing over time. Welcome to things beyond your first year of college.

The most interesting thing I see in this chart is why we see a long-term increasing trend in Teen employent before he late 1970s (even though the minimum wage was increasing) and afterwards a long-term trend of teen employment decreasing.
Other than this I don't see this chart as being helpful to understanding the overall effects of a minimum wage increase.

Not necessarily. If science is inconvenient to a political point of view, it will still be scrutinized.
For example: Climate change to the right wing, and fracking to the left wing.

There are many more examples of politics going against a scientific consensus, but I can't list them all during my break.

Lets just say, "Politics ruins science".

Not necessarily. The end product of science is the manipulation of the observed phenomenon. In other words, science is meaningful only because it is useful.

What we call politics is confounded by the fact that we give most people a say in policies that affect them. Even if minwage has zero effect on labor hours, one can still oppose it as an infringement of the property rights of employers. It is redistributive no matter what the effect on total surplus.

If the thesis is that minimum wage increases lower teen employment then the minimum wage must have decreased from 1955-1975. This is in real dollars, right?

I learned a long time ago that economists are bad at reading graphs. And they fancy themselves scientists!

No one is claiming that minimum wage policy is the only factor affecting teen employment. Your claim that people can't read graphs is stupid.

No I claim that economists can't write graphs, stupid.

Pretty selective line-drawing, as others have noted. For example the long period from 1999 to 2007 would feature falling real minimum wages (a long time without a raise), but also falling employment. More broadly,from 1979 to today real minimum wages have declined significantly (fn1), as has teen employment. It would be very easy to tell the opposite story if you wanted to.

(fn1) e.g.: http://www.fas.org/sgp/crs/misc/R42973.pdf

Selective? The trend lines are drawn before and after a MW hike, its not cherry picking. If it is so easy please tell the opposite story, that every time a MW hike occurs there is a rise in absolute or real teen employment.

In reality, I agree with Kevin here but can he note some of the other impacts occurring here? He should have included the late 1970s in the analysis which he can note was an unusual situation with a high teenage population population and high inflation. Otherwise the chart shows the minimum wage has an impact but the impact of large recessions appear to be larger. Of the last several (jobless) recessions we have seen teenage workers drop a lot.

Let's see the same chart now with recession bars...

I would argue that a the graph would be more useful if it tracked percentage of teens employed, rather than the raw population of employed teens.

We know the over these time spans, US demographics have changed drastically, and the population of teens has changed along with it.

Having, for instance, 20% fewer teens around, is likely going to mean that 20% fewer teens are going to be employed (and unemployed, and tall, and short, etc)

Also, control for the supply of illegal immigrants in min wage jobs.

Yes, I was thinking the same thing, but the trend lines sharp breaks help one navigate that a bit.

Follow the link, he has graphs for % of teens as well, same result. Its not as if there are less teens around today therefore less will be employed.

Breaking News: a flotilla of unemployed teenagers from the world's highest minimum wages nations of western europe have set off in a flotilla of makeshift rafts headed from various nations of africa where there is no minimum wage, thus guaranteeing them a economic paradise.


A flotilla of African's have set off for Europe, where they are willing and happy to work for less than minimum wage in the gray market. Because the minimum wage has effectively left millions of jobs untapped.

the best nations in the world have the highest min wage.

the worst have no min wage.

res ipsa loquitir

The best nations in the world also have the highest rate of diabetes. So I guess the key to prosperity is not having properly functioning insulin.

please leave your door unlocked today while you are at work---your plasma tv is "untapped". I would like to tap it by taking it down to the pawn shop.

That's really a pathetic comment. If you were going for funny, you missed.

I guess you missed my point. Let me make it reaaal clear for you: "tapping" millions of jobs in a western nation by importing cheap foreign labor is theft. And treason. Just my ever humble opinion!

You have successfully illustrated that having no minimum wage is not the only requirement for a good economic policy. However, no one has ever thought that was true, ever.

Many European nations were relatively late adopters of a min wage, and many still don't have a hard set national min wage.

This is just one of many great posts on minimum wage at that blog. I encourage you skeptics to read all of Erdmann's posts before tossing it aside. You might be able to dismiss one graph, but not pages of compelling arguments backed by data.

"You might be able to dismiss one graph, but not pages of compelling arguments backed by data."

I think that is wishful thinking on your part. There are plenty of people who aren't going to be convinced by any data that doesn't agree with their preconceived bias.

maybe because they understand that academic studies and data are mostly bogus. And they will remain such until academics go to jail for their fake studies.

Economicists mainly serve to generate propaganda for the upper class and corporations, propaganda based on fake studies.

Ah yes, if it goes against your beliefs, it must be fake. I wonder what lobby Darwin was paid off by......

The better question might be why Darwin's work has been so misrepresented by Darwinists - a point brought out by Kropotkin in Mutual Aid (free download here - http://www.gutenberg.org/ebooks/4341 )

'Two aspects of animal life impressed me most during the journeys which I made in my youth in Eastern Siberia and Northern Manchuria. One of them was the extreme severity of the struggle for existence which most species of animals have to carry on against an inclement Nature; the enormous destruction of life which periodically results from natural agencies; and the consequent paucity of life over the vast territory which fell under my observation. And the other was, that even in those few spots where animal life teemed in abundance, I failed to find—although I was eagerly looking for it—that bitter struggle for the means of existence, among animals belonging to the same species, which was considered by most Darwinists (though not always by Darwin himself) as the dominant characteristic of struggle for life, and the main factor of evolution.'

If the guy was unable to find competition over mates, he really wasn't looking very hard

Yeah, but I don't ever assume anyone with whom I disagree is anything other than an honest truth-seeker. There are dishonest ideologues out there, but I don't have disagreements with them - I try not to talk to them about anything at all, except maybe the weather. :)

Ok, fair enough and that's indeed a good point.

The minimum wage only matters if it's a binding constraint. Wage should be listed as "Percent of median wage". That would give a better sense of the impact of minimum wage laws.

Further, while youth used to be in low end service jobs--food counter work, lawn care--these tasks today in, say, Princeton are overwhelmingly provided by immigrants (Hispanic and Indian, primarily) older than 19 years. So the issue is not teen employment, but minimum wage employment.

And then you have to control for overall employment / population trends.

Tricky analysis. Erdmann's work is partial at best.

And then you have to suggest the "optimal" minimum wage.

"Wage should be listed as “Percent of median wage”. That would give a better sense of the impact of minimum wage laws."

That seems like it could be a useful metric. Perhaps part of the problem in determining the effects of minimum wage is that we've been using the wrong metrics to look at the data.

Minwage is binding only subject to its own market equilibrium. I dont see how using a summary measure of all wages in the economy is going to inform us whether it binds for, say, fast food or hotel services. You can indirectly observe the actual or expected profits of firms that hire minwage workers to see if it is binding. The average wage in the industry should also rise in the CEW data.

A lot of this would fall apart of he shaded recessions instead of the period of the increase.

There's a good reason why we have standards for model identification. This exercise illustrates those reasons perfectly.

What? No that doesn't match the facts. We know the dates of all the recessions during the period. They don't coincide with the start of minimum wage increases and only 3 of the 7 were during a recessionary period.

2007 - recession

1996 - no recession

1989 - no recession

1974 - recession

1966- no recession

1961 - recession

1956 - recession

I meant 4 of the 7.

4 out of the 7....yeah exactly, it's really not such a great set of data

If you totally exclude periods that they overlap with recessions then you are left with 3 examples. 2 of which show a sharp decline in teen unemployment.

But even the downturn that APPEARs to begin around 1989 coincides with the early 1990s recession. It's really not very clear.

His first post-MW period looks roughly '55-'60, with the '57-'58 recession right in the middle.
His second post-MW period looks roughly '61-'64, no recession coinciding
His third post-MW period looks roughly '66-'70, with the '69-'70 recession right in the middle
His fourth post-MW period looks roughly '74-'78, with the '73-' 75 recession in the middle and other macro problems towards the end.
His fifth post-MW period looks roughly '90-'94, with the '90-'91 recession coinciding.
His sixth post-MW period looks roughly '95-'99, with no recession coinciding.
His seventh post-MW period looks roughly '07-'11, and I just can't recall if there have been any major disruptions in the last five years or so.

So that's five of the seven post-MW red lines occurring with downturns in the middle of them. And guess what - the only two that don't have recessions coinciding with them have increasing red lines! (only one of those coinciding with a recession has an increasing red line, and that was a mild recession in a high growth decade.

It seems to matter A LOT, JWatts.

This is one of many reasons why we care about identifying models.

Now I'm just going off the picture here. But if even eye-balling it things look fishy that's a good reason to be more careful than this.

"It seems to matter A LOT, JWatts."

Only if you are trying to prove that minimum wage increases cause recessions or vice versa. Other wise, it's noise in the model.

Periods after minimum wage increases coincide with recessions in a high number of cases. When they don't the decline in employment isn't there. That's not "noise" that's an alternative explanation for the descriptive statistics presented here that must be accounted for. It's like a middle school science student noticing that most of his plants with fertilizer A are sitting in the sun and most of his plants with fertilizer B are sitting in the shade, shrugging his shoulders, and saying "it's just noise in the model". Even they know that's not right!

"Periods after minimum wage increases coincide with recessions in a high number of cases."

I don't disagree with the premise that minimum wage increases are factors contributing to recessions, but I don't see any strong evidence for that. There is strong evidence for minimum wage decreasing the trend line for teen unemployment.

That wasn't my premise at all.

My point is that when you make that inference about minimum wages' effect on teen unemployment you are doing it badly because you are not accounting for non-random assignment of the minimum wage.

Actually, none of it would "fall apart". The trend changes are pretty well established at the points of increased MW. One could say that in a few instances, MW increases occurred just prior to recessions (make of that what you will, but I am sure no one really wants to examine that).

Recessions are called by the NBER based on numerous factors, not all of which agree. A decline in teen employment could begin earlier or later than the start of a recession.

Using the dates of economic recessions could be as much of a specification error as not considering them at all. An instrument such as automobile or home sales could be used. These would be positively correlated to declines in economic activity but not endogenous to the employment of teenagers.

But the point is nobody should look at this graphic and be swayed one way or another (no matter what your position was) from the MUCH better identified studies we have.

I think that a good debate should be between those who think that say: 9 people working at $10/hour and one involuntarily unemployed is better than 10 people working at $7.50/hour and those who think the opposite. I think that there are non-momentary benefits to having a job so I would choose 10 people working at $7.50/hour over 9 people working at $10/hour and one involuntarily unemployed. I also do not feel any need to punish people who employee workers at low wages.

Except top economists can't even seem to agree if minimum wage increases affects employment much at all so we aren't even at that point in the discussion yet.

So CBBB if it could be proven that there was a trade off like I described above what side would you be on?

Another point that I made on the last MW post is that in the USA at least many people end up working for less than minimum wage. Of course better enforcement could help with that but it is often unclear.


And, I think someone should write an article about the segment of the population that actually is earning the minimum wage. Even McDonald's tends to pay higher. So, for those who still make minimum wage: Where do they work? Where do they live? What regions of the country? What type of jobs?

The first year I was able to work was the year I was 16 years old. That was also the last year of my life I ever worked for minimum wage. Since then, I've been earning more than the minimum even at low-skill, menial jobs.

I think the support for the minimum wage is not simply about some immediate effect for minimum wage workers. Raising the minimum wage might create pressure to raise the wages of other low (but slightly above minimum) wage workers. Since the traditional avenue of using labour unions to create this pressure has been destroyed, this is likely the only instrument for starting to create some long absent growth in wages for lower income workers.

"long absent growth in wages for lower income workers"

The average person currently on MW does not make the MW a year later, so no this is quite unlikely the only instrument if it is one at all.

Yes but that doesn't tell you much. It could just be that many of these MW workers are just teens who are only working temporarily in those sorts of positions. Overall there's been stagnation in wage growth - and not just at the very low end, and increasing the minimum wage could lead to a bit of chain reaction pushing up wages up the ladder.

It doesnt matter. There are flows of workers into and out of MW employment, and employment shows the stock. A declining stock means, on net, that the outflow is greater than the inflow. You might have a point if there was some reason to believe that a higher MW increases average human capital such that the outflow is actually inflow to higher wage levels.

Good point to think about.

"I think the support for the minimum wage is not simply about some immediate effect for minimum wage workers."

Labor unions have never been about creating growth in wages for lower income workers. They're about raising the wages and benefits of their membership. The lower the wages of others, the more money that's available for the rank and file. That's not economics, it's simple arithmetic.

Aren't we ignoring the geographical component? A fast food restaurant in western North Dakota is now paying a higher wage than a similar outlet in West Virginia. Paying attention to that would upset the macro paradigm so let's not.

+50, I always think that the biggest argument against a national minimum wage is the varying cost of living. The average wages in New York are much higher than the average wages in Mississippi. Most minimum wage hikes probably have no significant employment effect on high wage states, but probably have a significant negative employment effect on low wage states.

It would seem trivial (in the difficulty sense) to perform the ground work with respect to low wage states, but I never see an analysis that breaks it down that way. Granted, it may just be my own lack of knowledge.

You are conflating two concepts here.

Yes, a researcher should calculate the real wage based on the local cost of living, but...

The high wage in ND is an equilibrium wage well above minimum wage, and hence MW is not binding. Firms buy the (higher) equilibrium quantity rather than the lower MW quota. If anything, this would detract from the observed impact so the conclusion is stronger; the predicted effect shows despite the handicap of extra workers in ND.

ND is also far too small to affect national employment rates as shown here.

Why don't we have a futures market on the unemployment rate or employment rate of teens? Let people put their money where their mouths are.

Here's the same data, plotted against recessions: http://research.stlouisfed.org/fredgraph.png?g=rcX . All of the drops in teen employment, except for the 1968 one, coincide well with recessions.

Amazing, another article opposing the minimum wage that assumes that the business cycle does not impact teen unemployment and that all changes in teen unemployment are due to changes in the minimum wage.

Since WWII about half of the minimum wage increases have been during a business expansion and about half have been in recessions. The data shows that minimum wage hikes during a business expansion have essentially no impact on teen unemployment.

Every rise in teen unemployment following a minimum wage hike occurred during a
recession. If you regress teen unemployment against adult unemployment --to capture the impact of the business cycle-- and changes in the minimum wage, the regression shows that virtually all the changes in teen unemployment is due to changes in adult unemployment and that minimum wage changes had a minimal impact.



Tyler Cowen, you are a better economist than this to publish such biased economic research that completely ignores the impact of the business cycle on teen unemployment. You should be ashamed of yourself. Would you really give a student an A for such analysis?


sorry -- here are the correct links



Amazing, another article opposing the minimum wage that assumes that the business cycle does not impact teen unemployment and that all changes in teen unemployment are due to changes in the minimum wage.

No one here has come close to making the argument that "the business cycle does not impact teen unemployment". What's the point in posting such an out right straw man argument?

If you leave recessions out of the graph (or out of your model), then you are indeed claiming "the business cycle does not impact teen unemployment." That's what model specification means.

He didn't include a whole lot of potentially confounding factors either. For example, during the 1980's to 2000's many localities in the US restricted hours and nights for teenage employment. This would obviously have a dramatic effect on teen employment, almost certainly more than recessions or minimum wage increase. But you can't include everything.

It's fair to criticize the output because it doesn't include the effects of recessions. Indeed, you'll note that I asked the author to include them in his detailed graphs.

However, to proclaim that not including recessions in the graphs is the equivalent to denying their effects is a ridiculous argument. The author didn't make that claim. You are putting words into his mouth in a classic straw man fashion.

When you develop a method of creating multidimensional charts, return with your criticisms.

I don't even have to read the research to suspect that these economists controlled for business cycles irrespective of what the chart shows.

Even if you include recession shading, these time series variables may be nonstationary. There might be endogeneity. The chart is SUGGESTIVE of a potential relationship. You still need to do the hard econometric work, and I presume the authors are not dolts; they fo want to publish this, right?

Amazing how people react when their pets are kicked. I suppose I do the same.

"Tyler Cowen, you are a better economist than this"

No he isn't.

Kevin, thanks for kicking off such and interesting debate (and hat tip to Tyler of course).

Do you feel up to the work of marking the start and end points for the relevant recessions on your detailed graphs?

Somebody up higher in the comments posted the employment data on a Fred graph, with recessions. The interesting statistical question might be, over a 60 year period, how did it happen that recessions and minimum wage hikes happen so coincidentally? I don't know if that can be answered objectively, but I do intend to have some financial hedges in place the next time a major MW hike is implemented.

"The interesting statistical question might be, over a 60 year period, how did it happen that recessions and minimum wage hikes happen so coincidentally"

Honestly, I don't think that the interesting question nor a statistical one. I think it's the classic case that the people want Washington to "Do something now!" and the standard response is to spread money around like a fire hose (stimulus), raise the minimum wage and hold a Congressional hearing.

The start date of the policy has coincided with labor market declines, so the legislative passage has taken place before the labor market declines, not in response to them.

Are you saying that minimum wage legislation has consistently been enacted before recessions? Hmm, when I get some time I might have to research that. And if so, then your comment "I do intend to have some financial hedges in place the next time" bears thought and eventual action.

I decided to look at it in more detail:

03/1956: UE=4%, recession starts in late 1957
09/1961: UE>6% but recovering, recession had ended in early 1961
02/1967: UE<4%, no recessions
05/1974: UE=5%, recession started at end of 1973, unemployment shot up in 1974
04/1990: UE~5%, recession started soon after and unemployment shot up 1990
10/1996: UE~5%, no recession
07/2007: UE<5%, recession started soon after and unemployment shot up in 2008

So, 3 of the 7 were enacted just before a recession, one at the beginning of a recession but before unemployment had shot up, and three unrelated to recessions. So, as a precursor to recessions, it's 3 or 4 out of 7.
Looking at this closer, while there is clearly some significant effect on employment, depending on whether the post-MW context is recessional or not, the timing of the recessions cannot explain why there is a consistent downward kink in employment that begins in the few months before the MW hike. For recessions to explain this pattern, MW hikes would have to come on the heals of recessions. 1974 is the only episode that fits that timing.

2nd attempt at better formatting:

I decided to look at it in more detail:

03/1956: UE=4%, recession starts in late 1957

09/1961: UE>6% but recovering, recession had ended in early 1961

02/1967: UE<4%, no recessions

05/1974: UE=5%, recession started at end of 1973, unemployment shot up in 1974

04/1990: UE~5%, recession started soon after and unemployment shot up 1990

10/1996: UE~5%, no recession

07/2007: UE<5%, recession started soon after and unemployment shot up in 2008

So, 3 of the 7 were enacted just before a recession, one at the beginning of a recession but before unemployment had shot up, and three unrelated to recessions. So, as a precursor to recessions, it's 3 or 4 out of 7.
Looking at this closer, while there is clearly some significant effect on employment, depending on whether the post-MW context is recessional or not, the timing of the recessions cannot explain why there is a consistent downward kink in employment that begins in the few months before the MW hike. For recessions to explain this pattern, MW hikes would have to come on the heals of recessions. 1974 is the only episode that fits that timing.

Kebko, thanks for doing the homework.

Thanks for the input everyone. I did a new post where I changed the indicator for teen employment to teen Employment to Population Ratio. And, I added the unemployment rate (inverted) and recession indicators to the graphs to more specifically account for business cycles.


MW went up during the boom so we could all share in the 'prosperity.' The political impetus, unfortunately, came to fruition just as the booms were ending.

As I said earlier, recession dating is the wrong way to look at this because of endogeneity. You have to use IV techniques.

Thanks for all the input. This is obviously just one simple graph, and is not the last word on the topic.

The correlation between MW hikes and recessions is definitely interesting. I haven't found any signatures in the data that clearly delineated MW related job losses from cyclical job losses. Both tend cause larger disruptions in teen employment than in total employment. You would think that you would see some substitution from teens to older workers as the MW is increased, but the recent bls data indicates a fairly proportional rise in MW employment across the demographic categories. So, I haven't found a signature that I thought was a satisfactory way to separate these factors out. Let me know if anybody knows of research where they found clear way to separate these factors.

I think it is really important to combine the series of hikes (like 2007-2009) into episodes. This means that there are only 7 observations, which makes statistical analysis difficult, but if there is a pattern here, it's a pattern that mostly follows an initial hike.

In every case except 1956, there was at least one additional MW hike within one or two years of the initial hike. If there is a disemployment effect here, it's visually clear in the graph that employment usually dropped markedly from trend after the initial hikes, and that employment trends on the followup hikes reflected a recovery as the labor market moved in time away from the MW episode. Any analysis that treats all the individual hikes as separate observations is likely misidentifying recovery from the policy as a positive result of the policy.

Seven discrete events is enough to conduct an event study. At the very least, it will be suggestive.

The same technique could be used in a panel of all the state minimum wages, using the federal MW where states have none.

In that case, 'real' MW has to be adjusted for the local price levels.

Employment is not the correct variable, though. The traditional model predicts a reduction in labor HOURS.

Use an IV approach to control for economic downturns. Home sales/prices or auto sales may work. Take a close look at the lag structure too. In the very short run, one shouldn't expect an instantaneous response. On the other hand, if the policy is anticipated, the labor response would flow from the announcement instead of the effective date.

The problem with conducting an "event" study from the trendlines in the graph is that of "model mining" or "data mining" or some hybrid. Why choose 3 months before the hike takes affect and 27 months after the hik takes affect for the trend lines? When there are multiple hikes, why not chose 24 months after the last hike to end the trend line rather than 27 months after the first?

So if we lower minimum wage would we expect more teen unemployment? Interesting to think about why that might happen.

Chart shows employment, not unemployment.

doh! I completely misread. I thought Tyler was showing something counter-intuitive and misread the header.

Here I chart the proportion of the labor force that is at or less than minimum wage, compared to the minimum wage level:


Here, I use the same chart to take a broad stab at estimating the disemployment effects (which are split between unemployment and labor force exits). As with the chart Tyler linked to, this won't settle anything, but it's a conversation starter.


You might also want to some basic demographics of the baby boom moving through tthe age span as well.

For example the 18 to 19 year old population was 10,566,000 in 1960 and rose to 17,227,000 in 1978 before falling back to 13,762,000 in 1992.

Just eyeballing the data suggest this change in the teen population probably played a major role in driving the rise and fall in teen age employment.

Again, a major omitted variable that is likely to be as important as recessions and much more important than the minimum wage in explaining the changes in teen unemployment you have charted.

I'll stand by my original claim that this is such shoddy research that you should be ashamed of publishing it.

sorry I wrote this too fast;

It is the 16 to 19 year not the 18 to 19 year old population that rose and then fell.

You're right. Employment to Population Ratio is probably the better way to present this. I updated the original post to add a graph that uses EPR.


"such shoddy research that you should be ashamed of publishing it"

To call it "research" is a stretch.

I think we are seeing a cultural change here, not an economic one. When I was in high school it was almost universally accepted that high school kids would get a part time after-school job as soon as they could (usually after getting a drivers license at 16). The only exception were those who were on sports teams, and even they would get jobs out of season. By the time I graduated college (1992) this began to change, and it did change-- fast. It's now unusual for middle class teens to hold part-time jobs, especially during the school year, but even in the summer.

Kevin Erdmann's data actually does address this to a degree. The % min wage workers as % of labor force has been "mostly" climbing since mid 90s if I read his graph correctly. Doesn't speak directly to teens, but probably relevant nonetheless.


I don't see that this addressed at all.
Unless someone is too young to have watched this trend develop, it should be very obvious that teens have been exiting the work force in droves for about a generation now. In a big part due to the fact that middle class kids are no longer pushed into jobs as soon as they can be. Instead they are slated for college, and with increasing competition for slots at colleges, their non-school hours are spent doing school work, or racking up time with extra-curricular activities of the sort colleges smile on (which almost never involves working a minimum wage job).
An additional factor: there have always been regulations limiting how much, and how late, high schoolers can work, even on weekends and vacations. These were once honored mainly in the breach (I remember working well past midnight on weekends when I was 17), but nowadays they tend to be enforced-- and we've also had the trend of restaurants and retail (two of the big employers of the underage) staying open later at night, and often being not mom-and-pop operations but corporate run meaning there are people at headquarters who are aware of the laws and do not want to risk flouting them. All in all those laws create a hassle for managers who need staff they can count on, and who can work late and long at need, not kids they have to send home at 9:00pm, if not earlier.

How do we know that recessions are not the cause of raises in the min wage? At least in recent years, a bad economy has made for political pressure from the left to raise the min wage. Maybe this has always been so?

What hackery!

Some of this appearance is due to the choice of points to use to draw the trendlines. There is a large minimum wage hike in 1950 that was followed by an uptick in employment of teens. In 1956/57 the peak is reached before the minimum wage became effective. In 1967 the employment is heading down when the wage increases and in 1968 after the minimum wage hike the slope increases upward slightly. Another hike in late '63 is follwed by a slight dip and then an increase in the slope within a year.

There are also nuances in the law prior to 1966 so that not all firms were covered.

What makes you think that youth unemployment is a bad thing?
Studies show that children that work during school do worse in school.
We don't want teenagers to have jobs.
What we care about is unemployment as a whole.
This is a red herring in a straw man.

I believe they choose teens as a proxy because only ~5% of the workforce (majority being teens) make the MW so using unemployment as a whole probably wouldn't show how changes in MW affect it since it would be gobbled up in larger trends.

Also, can you cite your evidence on the teen employment affecting school performance? I believe it was well understood that college students who have part-time jobs perform better in college, why would it be different in HS?


That is a reference I found on goggle, although it isn't the one I was thinking of. The point is that when you raise min. wage, you get older people coming into the job market (People respond to incentives) and they force out the teenagers. This is a win win. We want people over 20 to work and we want people under 20 to be in school.

He seems to think that this is an argument for not raising the min wage when it clearly is an argument for raising the min wage. It is all very Orwellian.

Two points: Correlation is not causation. The massive drop in employment from 2007 to 2012 occurred during a sever recession. Nobody in their right mind would suggest that raising the minimum wage caused the recession.

This employment trend shows only "teen unemployment" not total minimum wage employment. Its quite possible that the drop in teens employed for minimum wage is offset by an increase in the number of older people taking those jobs during wage increases or in recessionary periods. In other words, the teens aren't getting minimum wage jobs during recessions because their parents are taking those jobs instead.

"Its quite possible that the drop in teens employed for minimum wage is offset by an increase in the number of older people taking those jobs"

No, it's not. The number of min. wage workers increases proportionately among all age groups as the minimum wage shifts.


If we include the 2007 episode, the average loss in the teen employment to population ratio in the 7 events is 5.9%, compared to trend.
If we only include the first 6 episodes, it's 5.6%.
If we only include the 3 most favorable episodes - 1961, during a recovery, 1967, with no recessions at all, and 1996, where there wasn't a decline, it's a 2.6% average decline.

Thanks for the input everyone.

I did a new post where I changed the indicator for teen employment to teen Employment to Population Ratio. And, I added the unemployment rate (inverted) and recession indicators to the graphs to more specifically account for business cycles.


Australia's minimum wage is as successful as America's minimum wage: both of them create unemployment by a factor of two for disfavored (largely dark-skinned) workers over favored (white) workers.

Minimum wage laws and their proponents are racist.

What is the point of this exercise? Where is the chart showing the impact on overall employment? Does the higher minimum wage, lower employment effect only apply to teenagers? Does it have the inverse effect on older people? The implication of this post is that it does?

In the raw data, the effects are similar, but smaller - the proportion of all workers that work at MW is about 1/4 the proportion of teens that work at the minimum wage. I find that the scale of employment changes in the total employment data is about 1/4 of what it is in the teen data. I didn't mean to imply that it only shows up in the teen numbers. It just shows up more clearly there.

I did some more work looking at the influence of the business cycle on the relationship.


Comments for this post are closed