Asymmetrical Information

Jane,

Your arguments are ridiculous. Questions of minimuim wage and social policy are to be decided based on their impact on "THE PEOPLE", not on any sneaky economic manipulation of statistics.

You are just trying to impose your bourqeouse capitalist free market religion, when we are in the strict arena of social policy and its affect on the disenfranchised masses of our elitist racist country.

Note: Sarcasm sensor is on.

It strikes me that labor force participation might be the better statistic here but for some reason I can't find that statistic at the BLS web site.

To expand a little on my last comment: it would seem to me that if (per hypothesis) states are raising the minimum wage in part because of low unemployment, they would do so not on the basis of one particular month's unemployment figures but on the state's long term experience with the labor market, and labor market participation statistics would better reflect that experience. In particular, IIRC, Massachusetts is having an off period for unemployment but its labor force participation numbers over time have been quite good.

Well, with respect to Michigan (at the top of your list), the raise in the minimum wage did not cause the high unemployment. The cratering of big-3 manufacturing (and associated parts suppliers) did that over the course of a few years. The rise the minimum wage above the Federal minimum, on the other hand, was just signed into law this March. Not a good idea when unemployment is already high, but not (yet) the cause of high unemployment.

You're not Tarzan. You're Jane.

Jane,
I have a question. It seems to me that one of two conditions must exist (in general)

1. The minimum wage is below the real market wage clearing point
2. The minimum wage is above the real market wage clearing point

If 1, then raising it helps poor people.

Doesn't the fact that we have many jobs at the minimum wage prove that 1 is true? Why else would there be jobs that paid this amount in particular?

To show this, assume that there is no minimum wage law and there is a free labor market. There would be a wide variety of wages and PRACTICALLY NO ONE WOULD MAKE EXACTLY THE WAGE DEFINED IN THE CURRENT MINIMUM WAGE LAW. It is only with a minimum wage law and a non free labor market and that #1 above is true that you get a large number of jobs clustered at the minimum wage point.

Note by "no free labor market", I mean that the employer and the employee do not meet on a level field to negotiate wages.

Robb,

Your phrase "real market wage clearing point" implies that there's a single market-clearing wage for unskilled labor. This isn't true: worker A1 may be prompt and polite, where B1 is chronically tardy and surly. Or worker A2 may have poor English, where B2 has excellent English, and C2 is biligual.

Raising the minimum wage may in fact help B and C -- they'll probably keep their jobs, and they'll get the upward bump immediately, not upon receiving a raise or promotion. But the same raise is likely to hurt the A's of this world -- if it's illegal to pay them what an employer thinks their labor is worth, they're not going to find jobs.

Additionally ... why should all $2/hour jobs be illegal? If a game company wanted to hire me for 4 hours 1 night a week to playtest its games, I'd think $2/hour was a pretty sweet deal. (In practice, they give you previews and sometimes free copies of the stuff you playtest.) I couldn't afford to do it full-time, but if someone offers to pay me to do something I enjoy in the hopes that I'll show up on time more reliably, why not?

Robb -

You ask: "Doesn't the fact that we have many jobs at the minimum wage prove that 1 is true?"

I don't understand. Are you saying that if there are, say, three employers offering a category of work, and they believe the wage necessary to attract good workers is $6.00, $6.50, and $7.00 per hour respectively, that they'll all *lower* their wage to $5.15 just because that's the minimum wage?

Seems to me much more likely that if they valued the work at $4.00, $4.50, and $5.00 per hour, they are forced to pay the minimum of $5.15 per hour, and that this explains the number of jobs paying exactly that amount.

Unless I'm misunderstanding the wage clearing point?

Showing Maryland and DC at opposite ends of the employment spectrum is clearly an anomaly-- the areas of highest and lowest unemployment are nearly contiguous geographically. More generally, comparing economic statistics state-by-state, e.g., comparing California to Rhode Island, just isn't serious.

Robb, you have it exactly backwards.

If 1 were true, then employers would be paying MORE than the minimum wage to keep their employees. The clearing point is higher.

But if 2 were true, then employers have to pay more than they otherwise would. The result is that they hire as few unskilled employees as they can manage to still function and pay them all the minimum. Many jobs would pay minimum wage (the least employers can pay) and more unskilled laborers would be without jobs (since employers lose money hiring more people at this price.)

So to answer you question, "Doesn't the fact that we have many jobs at the minimum wage prove that 1 is true?" The answer is no. It proves the opposite.

CAL

Robb:
"Doesn't the fact that we have many jobs at the minimum wage prove that 1 is true? Why else would there be jobs that paid this amount in particular?"

Few workers are paid minimum wage. According to the BLS, only 0.7% of all workers received the national minimum wage of $5.15 per hour in 2004. An additional 2.0% of all workers were paid less than $5.15. As the BLS points out, this latter group does not represent any illegal activity because:

"there are numerous exclusions and exemptions to these minimum wage statutes."

http://www.bls.gov/cps/minwage2004tbls.htm

You get lots of clustering at one of a couple wages without any outside intervention: in New Orleans right now, I believe the going wage for unskilled fast-food labor is $10 an hour (plus $250 biweekly, maybe?). This is much more than the minimum wage, but pretty much everywhere is charging that. I think it's just a natural effect of employers basically agreeing on how much the labor is worth, and of the fact that you have to pay as much as everyone else to attract any workers.

What would be more surprising was if everyone was paying different wages. How does the low-wage-payer manage to hire anyone?

Michigan's minimum wage law has not taken effect. So it's three states are in the ten best employment states, versus four in the worst employment states.

Oregon's minimum wage took effect Jan 1, 2006. From the data on http://stats.bls.gov/eag/eag.or.htm, unemployment is slightly lower now (5.6%) than it was before than law went into effect (5.7%).

Jane,

For what its worth, I ran a (VERY) highly simplified correlation analysis on the actual data (using Excel correl functionxwith your earlier suggestion: modest positive correlation (0.18263). Of course the real test here wouldn't be overall unemployment, but maybe something like African American males age 18-24.

As Ezra pointed out on his site, the key words in his previous post are "Any attempt to correlate minimum wage increases with joblessness falls on its face." He never made the argument for a correlation at all.

Sam L,

As I pointed out, even using this *INCREDIBLY* simplified model, you do get correlation. Not tremendous, but you do get correlation. My understanding is that when you use something like a real model the correlation persists. And when you start looking at employment patterns, the effects are even more noticable. Whatsmore, Klein does suggest that there is no correlation and we can safely ignore employment effects. The evidence doesn't support that, no matter how much you throw in offsetting factors.

Pass along the price increase to the customer.

Is that hard to understand? My company hires part time high school kids to come in and put stickers/labels on various packaging. It works out to be cheaper to do that than to buy automated machines which would be able to handle a variety of sizes.

They are actually paid more than the min. wage, but if we are required to pay them more than we do now, we would either fire them and figure out a way to make the machine alternative work or pass along the increase in labor costs, payroll taxes, etc. to the customer.

Perhaps what some love about min. wage increases is it also increases the amount of tax revenue the employee as well as the employer pays to the state and fed. gov. A sneaky way to force an increase tax revenue on a certain percent of the population?

Bill Dalasio, what are you using for your data?

"How does the low-wage-payer manage to hire anyone?"

The world is a complicated place, and the package of goodies (and baddies) that one gets from a job can be only approximately related to the formal paycheck.

As a modest example, I worked as a computer programmer when I was in high school. I was pretty good at it, if I may say so myself, and I was not only paid more than the minimum wage, but also got very flexible working hours and a job that I enjoyed and that taught me a lot. If you had held constant the flexible hours and the enjoyment, but taken away how good I was:-) so that I could only have been paid half the minimum wage, it would still have been a good deal for me compared to working at the minimum wage at some ordinary part-time job.

This is one of the reasons I sometimes have a mite of trouble crediting the sincerity of pro-minimum-wage intellectuals. They are fabulously inventive in imagining subtle ways in which the market could be failing people in ways that the minimum wage could fix, but seem to uncritically embrace the extremely simplistic axiom that a nominal wage increase is an actual increase in welfare. You don't need to be fabulously inventive to cook up situations where when you outlaw what was someone's first choice, then even if they can find an alternative with a nominally higher wage, their welfare is not increased by as much as their new paycheck suggests, and is likely decreased. Like, dumb question, I know, but why wasn't it their first choice before? Of course! The oppressive monopsony is broken and so the minimum wage has created exciting new all-around-better opportunities which are their new first choice, so outlawing their old choice is immaterial! But if you doubt the monopsony or other fancy constructs as Ms. Galt and I do, the dumb question is harder to set aside; and if you just look around in your everyday experience, you can probably find numerous examples where people are clearly choosing to work for a lower nominal wage than they could, often for quite natural reasons. Not just comfortable high school kids like me, either; if you're poor, it can still be a pretty iffy tradeoff to pay to move and then pay more for housing and commuting in order to work in the highest-wage area in the country, e.g.

Michigander,

I took the data from the tables cited by Klein to make his point. I assumed the federal minimum as a floor. Specifically, I pulled the data from

http://en.wikipedia.org/wiki/List_of_U.S._state_minimum_wages
and
http://money.cnn.com/pf/features/lists/state_unemployment/index.html

In all cases, I used the minimum outstanding as of today (first cite) and the cited unemployment rate.

Basically, I was creating a back-of-the-envelope calculation. If I messed up with the data at all, let me know.

Also, a couple of additional notes: the correlation is between minimum wage rate & unemployment rate, not unemployment rank.

bill I'm fairly certain the correlation you've found would fail a test of statistical signifigance. If we did this three months ago before Michigan raised the minimum wage, it would look even more trivial. This is a non-finding.

I am not sure what regression Bill ran, butmy results largley agree.

I ran it based on the data that Ezra presents and get a small positive correlation.

When I include the states which have a minimum wage determined by Federal law I get that an increase of $1 in the minimum wage is assosiated with an increase of .236 percentage points in the unemployment rate.

When I exclude those states I get that an increase is associated with an increase of .7501 percentage points.

The first is not significant, the second is signficant at the 1% level.

Bill, I think back-of-the-envelope calculations are a great tool for giving assumptions a reasonableness test.

Did you include Michigan as a state without a minimum wage?

They're significant at 80%(barely): again not great, but this is acknowledged as a misspecified model (I don't think anyone claims that unemployment is solely a function of the minimum wage rate). Of course you throw out Michigan, you get results consistent with your conclusions. That said, why not throw out tourist states Jane references. If you do that (also throwing out Michigan per your demand), all of a sudden the model is significant at the 95% level. Again, my major claim rests less on total unemployment, than on the distribution of unemployment.

Oh I see that Bill did correlation - not a regression my correlations are as follows

Including floored states: .2031
Excluding floored states: .3359

I prefer the excluded method because the scatter plot with the floored really looks like two totally seperate distributions.

Michigander: the data set has 51 items in its population, and the relevant number of states with a higher wage floor than the federal minimum is 21; throwing out Michigan is unlikely to totally tank the correlation.

More broadly, of course you can't prove correlation with something as crude as these figures; for one thing, they're a snapshot of a single month, and for another, as Bill suggests, there are a lot of underspecified variables.

My point was not that the data somehow proves that the minimum wage is bad for employment; that would be ridiculous. My point is that I cannot for the life of me understand how Ezra Klein could look at data which seem to show a modest correlation between high minimum wages and high unemployment, and say that this proves that high minimum wages aren't bad for unemployment.

Jane,

Said much better than I could ever hope to

/tips hat

Bill,

I did a regression on the same data. I threw out all of the states that had just the federal minimum wage - because they were all the same, they essentially contain no data about how minimum wage levels would effect unemployment.

I used excel for this.

The r squared for this was 4%, in other words the variation in minimum wages explained about 4% of the variation in unemployment rates. The adjusted r squared was actually negative.

Additionally, the t-statistic on the variable was not significant, so we would 'fail to reject the null hypothesis' which in this case means we cant reject the idea that they were completely uncorrelated.

I then did it again including the states with a minimum wage.

R squared of 2.6% with a t-statistic of 1.15, once again not significant.

mickslam,

I'm seeing a bit of a disconnect between your results and karl smith's. He's getting a lot more significance than you. Interesting. You seem to be using the same methodolgy (based on your descriptions). I'm curious, did you get a significant F-statistic? At what level was the t-stat insignificant? Finally, I really think we're making a mistake in making too much of the model here. We know its misspecified. I cited the correlation simply to show that Klein was being more than a little disingenuous with the data. Picking out a couple of data points and saying "See, they have a high minimum wage and low unemployment." is bordering on con artistry.

It seems like state income taxes ought to figure into this somehow.

My point in all of this is that when I looked at the data, my first thought was that there might be weak correlation between the two pieces of information, but that very likely this correlation was less than significant. It just jumps out at you.

When 3 of the 5 lowest unemployment rates have higher than the federally mandated minimum wage, its a huge sign that its really unlikely for there to be a very strong or significant correlation between the unemployment rate and a minimum wage, at least in the way JG might want it to be.

So I can easily see how Ezra might have thought that there might be no correlation. Without running a regression or anything like that, it makes intuitive sense just scanning the information. Its weak, but likely not significant. If you really look at it, that is.

Now there might be a stronger, but positive and significant relationship between minimum wage laws and state per capita GDP. That is the higher the minimum wage, the higher the GDP per capita.

mickslam,

I do hope you weren't trying to imply a causal relationship there. :7D

...throwing out Michigan is unlikely to totally tank the correlation.
Of course not, but it will weaken a correlation that was weak to begin with. As Sam L. pointed out, Erza doesn't claim to make an inverse correlation between high minimum wages & employment as you say he does, he says the correlation fails.

I can't imagine that a minimum wage has no effect on employment. But the effect seems small.

Be glad to send anyone my work. mickslam (at--) netscape.net If I start getting spam, well we know where its from.

Plus I was talking about R squared while I think he was talking about something else. The reason I talked about R2 is because the data looked to me like one of those where there might be a slight correlation, but there wasn't going to be anything really significant.

I didn't get any numbers anywhere that matched his. I did a regression analysis, he may have done some variation on correlation - I don't know. I didn't get his number on anything I did here. Thats not to say he isn't correct.

Bill,

:) Nah-lol

Jane, what are you using for a source for state unemployment rates?

From http://money.cnn.com/pf/features/lists/state_unemployment/index.html , both New York and Illinois are at the national average of 4.6%.

You have both of them in the unhappy list.

They're above the median, but at the mean; the bottom few outliers are skewing the average.

Also, I was using last month's list; the numbers seem to have changed slightly

On the list I was using, Hawaii was #1, for example; it has now dropped back to #2.

Robb, almost no one does earn the minimum wage, it's about 0.5% more people actually earn under minimum wage...

Total workers at or below minimum wage ar 2.7%

http://www.sudan.net/news/posted/12778.html

rasing minimum wage absolutely affects unemplyment, it absolutely effects immigrants and minorities most, the question needs to be whether the benifits outweigh the gains - seeing we do have certain social umbrellas in this country.

J-Deal - forgive my spelling playing with a toy as I type!

http://www.bls.gov/cps/minwage2004.htm

posted the wrong link, sorry

Pass along the price increase to the customer.

Is that hard to understand?

Well, it's not hard to understand, but I don't think that it's right. If one could raise prices without negative repurcussions (smaller volume), he wouldn't wait for a wage hike. In other words, unless there's something I'm just not getting, your idea fails first-year micro.

My R2 was something really tiny too but my T-stat was significant without the states with the federal floor.

Now when I say without the federal floor I mean I took out all the states with $5.15, lower or nothing.

I did this because here in NC the state minimum wage is 5.15 but it is just lifted from federal law, so it might as well be the federal floor.

Again, I admit that this is probably picking up a lot other stuff too. Like states with a minimum wage higher than the federal are likely to be liberal and therefore have other market restrictions.

Also, there might be some correlation between a high minimum wage and being a manufacturing state.

None of this, as Jane said, is an attempt to show that the minimum wage causes unemployment just that what the idea that there is a simple inverse correlation is patently false.

Besides I don't think unemployment is the relevant measure anyway. If unemployment gets to high kids will drop out of the labor force. The relevant measure is probably some cross between payroll growth rates and labor force participation.

Karl Smith,

Interesting variable choice (Okay, so I tend to geek out on this sort of thing). I was actually thinking that minimum wage per se was the wrong variable choice here. I was considering a ratio of (2080 * minimum wage)/median income. The basic idea being to try to adjust for local labor market differences.

Small R squared - for the non-stats folks out there - means that the differeneces in minimum wage laws only explain a very small amount of the differences in unemployment. I got 2% and 4%. And Will might very well be correct that we should be looking the other way.

But really the point here is the incredible mocking tone that Jane took in her original post. 'Tarzan confused' ? You would think that Ezra deliberatly spun the data to show no correlation - when its pretty clear from any reading, the correlation is small, explains only a little of the changes, and I found it to be insignificant.

If anyone is spinning here, its Galt, not Ezra. He didn't claim there was no correlation, only that it was small at best. Well, thats where we are with this data - small at best.

Then Jane tries to use the 'outliers' excuse. Well, outliers are something that are completely out of line. You can't have a decent distribution like we have here and have 3 of the top 5 be called outliers, unless you throw out all of the points above that level. Its simply not the way you do things. I know about outliers, as what I've done professionally is test trading systems. I am completely familiar with outliers and how they can completely skew your results.

I suspect that Jane meant 'outliers' in the sense that 'these points doesn't agree with my theory - I should throw them out', which of course is the entirely wrong to approach examining data, if you want to reach a practical, real world conclusion. If you want theory confirmation at any cost, try Galts way.

The point isn't the data, its the condesending tone. This is highly suspect data at best, and like Jane said above, you can't use 1 month for something like this. Well, if thats the case, you cant use this data to mock someone either.

For the record, I am not for an across the board minimum wage raise. There is simply too much difference in income levels across this country, from region to region. You can't and shouldn't compare what happens in NYC with rural Alabama.

Bob Dobalina,

I'm not so sure I agree with you. Price is a factor of both demand and supply. Yes, if the business owner could raise prices absent a minimum wage hike they would do so. But the factor holding them back is competition. If they raise prices, their competitors will undercut them. But, with the minimum wage hike, the competitors also face higher costs. If we assume a competitive market in the first place, the competitor's ability to undercut them would have been diminished by the same cost increase.

I was much nicer to Jane than I should have been, after looking over her post at Ezras place.

Check this out, in light of all thats been posted here:

"I was honestly just trying to figure out how on earth you got that conclusion from this data. I'm afraid that as far as I can tell, the answer is: you didn't."

This, after a long rant about how Ezra was somehow using this as a dishonest plot to slam republicans. Well, Tarzan, you seem to be doing exactly the same thing, but to Ezra. The data simply doesn't support your views, Tarzan.

Then this:

"More broadly, you are doing something very common among people trying to score political points against opponents: you're setting up a straw-man that no one believes in, refuting the super-strong case--"high unemployment will inevitable be found in lockstep correlation with high minimum wages--and then claiming that this somehow refutes the ordinary case"

Just as she is doing here, setting up an Ezra strawman, and then demolishing it.

Mickslam, I don't know what the hell your problem is, but I used the title "Tarzan Confused" as a gentle gibe at myself, not a put-down to Mr Klein. At that point I was working on the assumption that he had done something sophisticated that I was missing, or that I was misreading the chart. An assumption that turns out to have been incorrect. Ezra Klein used data to illustrate a point that said the exact opposite of what he claimed . . . and then tried to defend it by excluding all the data points that contradicted him from the sample. That's a data no-no, as you so eloquently declaim above.

As for using the "outliers" excuse, it wasn't an excuse . . . I didn't try to use any of the outliers to exclude them from the set. I simply pointed out that the very high rates of unemployment in the bottom three cases skew the mean unemployment rate, so that it is higher than the median rate, which is, I think any analyst would agree, a better indicator of whether your state has a high unemployment rate. As many people are fond of pointing out in re income distribution, if everyone gets poorer except Bill Gates, who gets fantastically richer and thus raises "average" incomes, this does not mean that the country as a whole is better off.

I can certainly use this data to mock someone who misuses this data, which is all I am doing (though not at the time of the post). I cannot use this data to prove, or disprove that the minimum wage is a good idea, but while the latter is quite an overstretch, the former is logically risible.

Mr Klein screwed up. I will continue to mock him until he acknowleges that his cursory reading of the data led him astray, or, much more likely, until I get bored. Which is probably going to happen any minute unless someone can come up with an interesting rebuttal to my mockery.

Mickslam, you are absolutely correct. The relevent measure is not the unemployment rate.
Rather it should be the economic well being of the population.

According to the Census bureau the median real
household income from 2001-03-- a moving average to reduce the impact of one year -- was $43,531.
If you take the average income of the 19 states that have a minimum wage above the federal minimum their average income was $46,527.

Moreover, if you take the average rank of these 19 states it is 12.8 and only three of these 19 states were below average --they were D.C., Oregon and Florida.

Now we can argue to the cows come home about which way the causal relations flow, but the evidence is extremely strong that a higher minimum age is strongly correlated with a higher standard of living as measured by median household income by state.

I will be glad to send anyone who want to play with the data a file of the raw data.

The argument that the only way to evaluate the minimum wage is a great strategy by the right wing to determine the terms of the debate in a way that they think they have an advantage. But if you bring other factors into the discussion you can clearly see that the unemployment rate is a minor factor that should not play a significant role in the debate.

The argument the opponents of the minimum wage are trying to turn the debate on is even if an increase in the minimum wage improves the economic well being of 2 million people, if it causes one person to suffer a drop in hours worked this is an unacceptable traded off. While this may not be a Pareto Optimal, I know of no political issues that are required to meet that standard.

I was considering a ratio of (2080 * minimum wage)/median income. The basic idea being to try to adjust for local labor market differences.

Yeah this stuff gets really hard which is why great empirical work gets you a Bates Medal

Local labor market conditions are influenced by the minimum wage so want to be careful not to take too much out.

If there is any high min wage -> high employment connection it probably comes through the minimum wage inducing people to become more skilled or something like that.

To say that someone is better off because the median household income is higher in a particular state just doesn't wash. You have to know what the cost levels are in the particular state to say that they are better off. In other words, you can live better on less in Mississippi.

Will -- the data was real income so it had the adjustment you suggested.

But you did just violate my basic rule that any proposal that claims we could raise our standards of living by emulating missippippi and/or alabama has to be wrong.

Will,

That wasn't so much what I was proposing. My argument is more based on the level of the market clearing wage. Basically, I'm considering proxying minimum wage as a portion of the median earnings for some sort of measure of how high the minimum wage is relative to the market clearing wage. After all, a $7 minimum wage in New York City or San Francisco is going to mean something a little different from a $7 minimum in rural Mississippi. The $7 minimum would be imposed on a lot fewer labor contracts (%-wise) in NY/SF than MS.

This data doesn't mean anything because there are too few observations and a huge number of factors besides the minimum wage which would cause some states to have higher or lower "unemployment."

Comments are Closed.