16 Feb 2013

I Get Empirical on Minimum Wage

Economics, Krugman 129 Comments

So Paul Krugman et al. are telling us that the empirical literature shows that minimum wage laws don’t have a discernible impact on teen unemployment. Yet another right-winger myth flushed down the toilet.

I went and looked at the paper Krugman says documents this fact. I was a little concerned because it seemed to be saying (I’m paraphrasing of course) that yes, if you just naively look at the states that have higher minimum wage laws, then they have lower employment growth, but once you correct for the broader trends in employment growth among states, then the impact of the minimum wage laws per se disappears. For example, it might just so happen that states that have bad weather, also have state-level minimum wage laws higher than the federal floor, and so earlier researchers incorrectly blamed the low employment growth on the minimum wage burden.

This sounded a little fishy to me, so I decided to do my own, very naive, straightforward look at the data. I first used this map from the Department of Labor to organize the 50 states into two groups: Those with minimum wages higher than the federal level, and all others. (The map says it’s accurate as of January 1, 2013.) Then I downloaded the Local Area Unemployment Statistics from the BLS to get state-level data broken down by age. I used this Excel table to get the unemployment rate among 16-19 year olds by state. Here’s what I found–and I am sorry but putting the two PNG files together is the only way I can figure out to present this quickly:

These results are pretty striking. First of all: Notice how high the teen unemployment rate is, across the country. If minimum wage laws have no effect, why should this be so? I am not denying that Krugman could come up with some story, but prima facie this fact in and of itself is a feather in the cap of the Econ 101 textbooks. (Specifically, we’d expect 16-19 year olds to have the lowest productivity, and so they would be disproportionately hurt by an absolute minimum wage law.)

Now then, if we look at the 19 states that have a minimum wage higher than the federal minimum, the average unemployment rate among teens is 25.2%. In contrast, if we look at the 31 states that have either no state-level minimum wage or one that equals the federal level, the average teen unemployment rate is 21.5%. A pretty big difference, and this is from sample sizes of 19 and 31. It’s not as if you’ve got California driving the result. (And of course, last time I checked, California had pretty nice weather.)

But beyond the arithmetic averages is the clustering of the states, in their respective groups, when you rank them from the highest to lowest teen unemployment rates. The most striking result to me: If you look at the top 5 and the bottom 5, you find: Four of the top five states have higher-than-federal minimum wages, while only 1 out of the bottom five does. I’m not sure how to set up the statistical problem, but I think that would be an incredibly unlikely result, if minimum wage laws had nothing to do with teen unemployment rates.

Furthermore, if you look at the top and bottom 10, you get: Out of the top 10, six of them have higher-than-federal minimum wages, while out of the bottom 10, only 1 does.

And then, just looking at the top and bottom halves: The top half has 13 states with higher-than-federal minimum wages, while the bottom half has the other 6.

Before you guys go shouting these results from the mountaintops, I really would appreciate it if somebody could reproduce my results. I did this fairly quickly Saturday night, and it’s possible I mislabeled one of the states etc. But these seem like pretty powerful results, especially since they back up what is literally textbook economics. So if someone could double check my work, that would make me more confident in publicizing this.

129 Responses to “I Get Empirical on Minimum Wage”

  1. Aaron Alghawi says:

    Just some friendly advice on how to set up the statistical problem in case you didn’t get my comment on Facebook.

    You’d want to set up the regression model where there is a positive correlation between minimum wage and youth unemployment. The youth unemployment is your dependent variable. Independent variable obviously is minimum wage, your main beta that you’re looking at, and I would also control for the state’s overall unemployment rate. If you did it in STATA I think you’d also need a panel variable for each state, which would be a z-test with a p-value instead of a t-test which is what people usually use. You might not need to since it’s all for one year I’ve done the panel variable in the past for when I had multiple metro areas over multiple years (also threw in the time variable for the years.

    I’m used to STATA and know a little bit about Excel’s statistical package and PH Stat, you might prefer SAS. But if you did it in STATA this is how I would approach it.

    Model : Youth unemployment = β0 + β1minwage + β2stateunemployment + u

    Null hypothesis would be H0: β1 ≤ 0, that there is not a positive correlation between min wage and youth unemployment while controlling for overall state unemployment.

    Alternative Hypothesis H1: β1 > 0, that controlling for state unemployment, there is indeed a correlation between minimum wage in a state and that state’s youth unemployment.

    Choose a 95% confidence interval for both models. If p 0 it’s a one-tailed test, so reject the null if p/2 < .05.

    With 50 observations DF = 50 – 2 – 1 = 47. The critical value on the z-table for this at alpha 0.05 is about 1.67.

    This should help get you started.

    • Adrian Gabriel says:

      Aaron, great way to expound the jargon you were taught at your university. Murphy states:

      “I went and looked at the paper Krugman says documents this fact. I was a little concerned because it seemed to be saying (I’m paraphrasing of course) that yes, if you just naively look at the states that have higher minimum wage laws, then they have lower employment growth, but once you correct for the broader trends in employment growth among states, then the impact of the minimum wage laws per se disappears. For example, it might just so happen that states that have bad weather, also have state-level minimum wage laws higher than the federal floor, and so earlier researchers incorrectly blamed the low employment growth on the minimum wage burden.”

      Broad correlations can lead to significant error, even within a confidence level of 95%. Regression analysis is positivisim at its worst. Murphy is trying to make a point of being a little more pragmatic about the numbers, and hence very realistic at how to derive basic and lucid conclusions. No Regression analysis is needed. What you just demonstrated is a pedantic attempt of creating excessive minutia for an analysis that can use basic deductive reasoning, in a descriptive model such as Mises and Rothbard would have done. It worked quite well for Murphy and allows to trounce Krugman, but also delineates a simple fact of how mainstream economists are silly and always arrive at more problems.

      • Major_Freedom says:

        Well this response was a little excessively harsh. Aaron was obviously just trying to help.

        • Adrian Gabriel says:

          Aaron sounded just as confused as other statists. Here’s Murphy on what I was getting at: http://mises.org/daily/1001

          • Major_Freedom says:

            Wait what? Statist?

            How in the flippertygibbet did you parse statism out of an explanation of regressions and statistical analysis?

            Chill out?

    • Ken B says:

      Ehhh, regression is one approach. Look below at Landsburg’s quite different test. At a blush Landsburg’s test, which is a test on order data, is a more natural match to the kind of data we have. But in any case you want when possible to apply multiple logically distinct tests.

  2. zak says:

    just say no to triangular intervention

  3. Bruce says:

    Couldn’t time series data be used in comparing change in unemployment rates before and after an increase in state effective minimum wage for a sample of states with such an increase and with no change.

    • Bruce says:

      Having read more, the I find the Card study seems to have done just that. It seems to me that opportunities to test the theory against data will be pretty limited by the data. I’m skpetical of apply the Card work to policy.

      Ultimately, it comes down to priors I suppose.

  4. skylien says:

    Well, how about Greece’s and Spain’s youth unemployment rate of now over 55 and 62% which is in stark contrast to their average employment rates ?

    How does an economist explain this if he throws out textbook economics? I thought those economists were arguing that the EUR works as a straight jacket (like the Gold standard) and doesn’t allow monetary devaluation in Greece and Spain to bring down their price level (that is to say WAGES!) to cure high unemployment. Yet when it comes to low wages this logic suddenly doesn’t apply? …

    • skylien says:

      The 55% are for Spain, and 62% for Greece..

    • Lord Keynes says:

      Labour markets all over the world are still in the aftermath of the worst real output collapse since the 1930s.

      • skylien says:

        So that is how you explain the discrepancy between overall unemployment rate of 27% to youth unemployment of 62% in Greece?

        • Christopher says:

          It’s because the evil speculators who caused Greece’s problems in the first place decided to go after young people first as part of their evil strategy.

      • Major_Freedom says:

        This doesn’t explain or address why WITHIN the unemployment category, we see such large differences according to such sub-categories like age groups.

        Yes, we all know full well that Keynesians can’t think in terms of anything other than crude aggregates, and that you send a hammer for every problem, but at least try to stay on topic.

  5. Lord Keynes says:

    (1) ” First of all: Notice how high the teen unemployment rate is, across the country. If minimum wage laws have no effect, why should this be so? “

    Because labour markets are still in the aftermath of the worst real output collapse since the 1930s.

    (2) assume that your method is free from problems with method, statistical bias and problems, and data is all correct (big assumptions). Even here the crucial factor would likely make the difference: an economy with fiscal interventions to create full employment.

    • Major_Freedom says:

      Red herring.

    • Tel says:

      Get government to pass a law declaring by fiat the everyone is employed. Then, unemployment will by definition be zero at all times, and we can save the effort of measuring any more. After that it will guaranteed be the case that minimum wage no longer has any effect on unemployment.

      Next step is for government to pass a law declaring by fiat the everyone is happy.

    • Nick says:

      And since the banking mess, we have had billions of deficit spending and borrowing. So if Keynes is correct, that will produce massive increases in employment.

      ie. The state borrows and spends – cures the problem

      Except it hasn’t.

      The reason is that the problem is borrow and spend. It’s what caused the banking mess when people didn’t pay back their debts.

      More of the cause isn’t going to be the fix.

  6. Matt Wetta says:

    There are too many variables to Mathematically prove Unemployment is positively correlated with Minimum wage rates. The bottom line is Krugman et al hide behind guises and gimmicks, basically just making up stuff. Economics is law 1+1=2. Can you prove 1+1 = 2? It’s ridiculous. Try proving Santa Clause isn’t real?

    Let’s call Krugman what he is, a bumbling fool who somehow got a ride to the top from ridiculousness he spouted enough so that drugged up Europeans believed him.

    Just Simply Show a graph of the Labor supply and Demand and how Minimum Wage puts a Floor across the curves causing the number of employed to drop drastically, across the board. Since Teenager make up most of the unskilled labor suply(calculate a percent) then that’s how many teenagers are affected.

    Low skilled jobs have high Employee turn-a-rounds, I think the average worker stays at Mcdonalds 4 months tops. Another Variable, so many people quitting and finding different jobs or quit working period, it’s impossible to tell.

    Krugman’s statements don’t deserve actual economic analysis. They are just completely absurd. 9 Dollars an hour Min Wage would destroy our economy and jobs. The only jobs available would be Tip based.

    Another point, most teenagers get jobs making tips. Waitressing, Bartending, so the Minimum wage isn’t even applied to those. The biggest industry in America is Food. In the food industry a lot just make tips, sub contract etc..

    • Ken B says:

      No, you cannot do this. The logic of the Econ 101 argument is not at issue. What’s at issue is whether it fits the world. That cannot be assumed, so empirical tests matter. PK says this data suggests the effect is small or non existent. Landsburg makes a strong NUMERICAL AND QUANTITATIVE case that PK is wrong here, and that the data vindicates 101. Murphy also makes an EMPIRICAL case. The dispute here cannot logically be resolved by an appeal to the theory.

      • Dan says:

        Austrian economists would disagree with you about this. Dr. Murphy is simply doing what he does quite often, playing by the rules of his opponents to beat them at their own game. As he points out in this article,

        “Economists in the Misesian tradition stress the primacy of theory in the social sciences. When trying to figure out the Great Depression, for example, we can’t approach the topic with a blank slate and let the facts “speak for themselves.” Mises argued that in order for us to even know which facts to consider as relevant, we need to have an antecedent body of deductive insights.”

        http://mises.org/daily/4648

        • Ken B says:

          You cannot settle an empirical test of a theory by appealing to the theory. PK made an empirical claim. Bob and Steve have challenged his reading of the numbers. They do so on the basis of statistics, and probability ie evidence.

          You cannot dismiss the empirical test of a theory by appeal to the same theory. You cannot dismiss a violation of Newton’s Laws by citing them. I think most Austrians understand this, and your Mises quote in no way at all contradicts my point.

          • Dan says:

            You said you believe empirical tests matter in proving economic theories. Austrians don’t agree with that. There is no empirical test that would convince an Austrian that minimum wage laws are a good thing.

            • Joseph Fetz says:

              That’s not entirely true, Dan. Austrians use data to prove theory all the time, the difference primarily lies in the source of the theory (deductive vs inductive), and that data is very limited in showing the full picture of what is happening. It’s only a snapshot of the economy at a given time, and not all factors can be controlled for or even known.

              • Joseph Fetz says:

                * the understanding that data is very limited …

              • Dan says:

                Do you think there are empirical studies that can prove the Austrians are wrong about minimum wage? If not, I’m not sure how we could say empirical studies can prove Austrians right, either.

              • Dan says:

                I just don’t see how an empirical study can prove or disprove that unemployment will be higher than it would’ve otherwise been if minimum wages are risen. How do you show what unemployment was going to be absent the minimum wage increase?

              • Joseph Fetz says:

                No, I am not saying that.

                What I am saying is that the Austrian argument against data is based upon the foundation of theory, not its use in every sense.

                Austrians understand the limited nature of data, but that does not preclude them from using data to confirm their theories.

                However, it is this same understanding that the data is limited that is the basis behind the Austrian conclusion that data should not be the foundation of economic theory.

              • Bharat says:

                Joseph, I’m pretty sure you’re wrong on this. The Austrian tradition is known for its deductive method of praxeology in discovering economic laws, but these laws cannot be proven or falsified by empirical evidence. They can only be proven or falsified by logic.

                What you are talking about is when Austrian economists use data side by side with theory. This is economic history supplementing theory. It is not being used to prove or falsify the theory.

              • Dan says:

                Unless I’m just misunderstanding what you are trying to say I’m with Bharat on this.

              • Joseph Fetz says:

                Perhaps “confirm” was too strong a word, maybe “supplement” is a better word. In any case, it now looks like my questioning was unwarranted, because looking back I can see that the word that Dan used was “prove”.

              • Ken B says:

                “economic laws .. cannot be … falsified by empirical evidence. They can only be … falsified by logic.”

                If Austrians really believe this nonsense then Krugman is right about them. Fortunately not all do.

              • Bob Murphy says:

                Ken B. wrote:

                If Austrians really believe this nonsense then Krugman is right about them. Fortunately not all do.

                Ken, chill out. Are you aware of what the nuanced position is? Economic principles like “people respond to incentives” and “other things equal, an increase in the quantity of money leads to higher prices” are not something you go test.

                In contrast, to say, “Obama’s decision will throw millions of people out of work” isn’t an economic law, it’s an empirical claim.

              • Ken B says:

                Yes Bob, I am. But let’s look at the claim that started this subthread.
                “There are too many variables to Mathematically prove Unemployment is positively correlated with Minimum wage rates”
                Let’s set aside prove as we know you cannot ‘prove’ a theory empirically. But you can test its predictions empirically. Imagine all studies and tests failed to show an correlation of the minimum wage on employment. For decades. Would you still go forth can claim the economics is a science that tells us raising the minumum wage has a positive correlation with unemployment and the facts be damned?

                This subthread is about a particular claim: that you can dismiss the relevance of any empirical tests of a theory by appealing to the same theory. That’s wrong. You can argue that a test is inconclusive or based on a misunderstanding but you cannot say “Theory X tells me theory X is always true so theory X is always true.”

              • guest says:

                … but you cannot say “Theory X tells me theory X is always true so theory X is always true.”

                Believing that data can prove or disprove the effect of the Minimum Wage on employment is to assume that the data reflects a state of the economy in which all other factors have remained constant.

                But if some “greedy” capitalist has greatly streamlined some production process, thereby lowering the costs of production for some previously expensive item, and more people are able to buy it, thereby creating more job opportunities, then a rise in the minimum wage may not have a noticeable effect on unemployment.

                Assuming that people want to increase their standard of living, it follows logically that if an increase in the cost of production has been imposed on a business, then people’s standards of living will not increase as much as they want;

                And so they will act to regain the progress toward their increased standard of living by lowering the costs of production insofar as they have control over it.

              • Ken B says:

                Believing that data can prove or disprove the effect of the smoking on lung cancer is to assume that the data reflects a state of the lungs in which all other factors have remained constant.

                Not very convincing is it?

              • guest says:

                Believing that data can prove or disprove the effect of the smoking on lung cancer …

                Both the act of smoking, and lung cancer, are empirically observable.

                The Minimum Wage is observable, as are people”s actions; but this is not the case with the decision-making process, itself – the valuations behind the actions.

                This is unobservable by way of the Scientific Method, and therefore cannot be tested by it.

                The Minimum Wage issue is a philosophical issue, rather than a scientific one.

          • Bala says:

            “You cannot dismiss the empirical test of a theory by appeal to the same theory.”

            This presupposes that the empirical “test” is actually capable of being a valid test of the theory. This presumes that empirical tests can indeed invalidate a theory derived as a logical corollary of a sound axiom. I think this is still the same fundamental methodological problem surfacing all over again.

            • Ken B says:

              I do not mean a particular test. It would have been clear had I written “the testing of”. I mean you cannot dismiss the relevance of empirical tests of a theory by saying “and my theory tells me empirical tests cannot disprove it.” That’s not science, and it’s an admission the theory has no value for predicition.

              • guest says:

                I mean you cannot dismiss the relevance of empirical tests of a theory by saying “and my theory tells me empirical tests cannot disprove it.” That’s not science, and it’s an admission the theory has no value for predicition.

                See? This is why I took issue with Herbener’s claim that Austrian Economics is a Science.

                I see where you’re coming from;

                But consider that the Scientific Method isn’t “scientific” by the standard you’ve expressed.

                The Scientific Method must be logically sound before it can be used to gain information from data.

                The Scientific Method is a philosophical claim about the method of gaining information from empirical data.

                No amount of data in the world can falsify the Scientific Method. Only theory can do that.

              • Bala says:

                Ken B,

                You are clearly wrong big time.

                “I mean you cannot dismiss the relevance of empirical tests of a theory by saying “and my theory tells me empirical tests cannot disprove it.” ”

                The “relevance” of empirical testing to economic theorising is a methodological issue. It is after settling that and choosing deductive reasoning starting from the axiom “Man acts” that Austrians develop economic theory. The “Laws” of economics are those propositions that are logically, necessarily true given this basic axiomatic proposition. The basis of the truth of these propositions is the strength of the axiom and the soundness of the reasoning. Empirical data cannot – do you hear me?…. cannot – confirm or negate a conclusion arrived at by such a process. So, it has no real place in economic theorising. And…. I do not refer to the theory to dismiss empirical methods. I go “pre-theory” to do so. I say that the empirical method is nonsense. That’s why
                1. I never chose it in the first place
                2. I do not make a monkey of myself by using it to “validate” my theory (which I cannot)

                “That’s not science, ”

                Oops… Wrong again. You are talking like a scientologist. What you describe is the method of the natural sciences where empirical methods are very much applicable. To insist that those are the only methods to apply to all “science”, you identify yourself as a scientologist.

                “and it’s an admission the theory has no value for predicition.”

                What if the fundamental issue is that the very attempt to “predict”, that too precisely, is asinine because you just cannot predict the behaviour of the subject???

                What you said is the very reason I am disappointed that Bob went empirical instead of dismissing the insane attempts at empiricising a science where empirical methods has a very limited place and no place in theorising.

              • Ken B says:

                You two guys really just don’t get science.

              • Bala says:

                “You two guys really just don’t get science.”

                So says the head priest of The Church of Scientology.

              • Ken B says:

                You know Bala, you can call me a Scientologist all you want. I suppose it saves the effort of making a point.

              • Bala says:

                “You know Bala, you can call me a Scientologist all you want.”

                Wrong as usual. You said that the empirical method should be used to validate economic theory and, worse, that if proponents of the theory state that the demand is incorrect because the empirical method is inapplicable, their theory is not a science?

                And then you whine when you are identified as (Note that I said IDENTIFIED AS, not CALLED) a scientologist. You have some temerity.

                Incidentally, this is what I got on dictionary.com for “scientism”

                “the belief that the assumptions, methods of research, etc., of the physical and biological sciences are equally appropriate and essential to all other disciplines, including the humanities and the social sciences.”

                I guess that does make you a scientologist.

              • Bala says:

                Ken,

                You’ve also got a lot of nerve saying

                “I suppose it saves the effort of making a point.”

                after you said just this much in place of a valid point

                “You two guys really just don’t get science.”

                I guess you were saving yourself the effort of making a point.

        • Tel says:

          Economists in the Misesian tradition stress the primacy of theory in the social sciences.

          Sure, but if theory never delivers a workable prediction then it is useless theory.

          • Bharat says:

            Austrian theory can predict, but only by use of ceteris paribus. Austrian economics can say, for example, that if a price rises, then quantity demanded will fall, all else equal. This is a prediction, not of the regular sort, but it is still in a way a prediction.

            Meanwhile, mainstream economics thinks it can predict whether all else will be equal or not. It thinks it can predict human action and human preferences, which are constantly changing.

            Sounds better to me to have a theory that is correct than a theory that claims it can predict but cannot. If predictions of the sort the natural sciences make are your standard, then I’d say all economics are useless.

          • Bala says:

            What if the nature of the subject of the study is such that it is not open to precise predictions, especially of the numerical variety? Would you still consider the lack of precise predictions reason enough to call the theory useless?

            • Bala says:

              I forgot to mention this.

              Don’t forget to note that you have coloured the question with loaded terms when you use the phrase “workable predictions”.

            • Tel says:

              Imprecise predictions of the statistical variety are still good enough for betting on.

              • Bala says:

                What if the subject matter of the study is not amenable to imprecise predictions of the statistical variety?

              • Ken B says:

                What are the predictions then? What decisions can be made on the strength of them? I mean I can predict the effect of turning up the thermostat: “the temperature will change or it won’t.” That’s my prediction. How is that not an example of what Tel said, a useless prediction.

              • Bala says:

                “What are the predictions then?”

                This is rich, especially after I said this

                “What if the nature of the subject of the study is such that it is not open to precise predictions, especially of the numerical variety?”

                ” What decisions can be made on the strength of them?”

                How much simpler can it be? Just do not meddle in the market’s price mechanism. Do not set a minimum wage. Is this that hard to get?

                ” I mean I can predict the effect of turning up the thermostat: “the temperature will change or it won’t.” That’s my prediction.”

                Scientology at its best. You show an example of prediction in the natural sciences and ask “Why don’t you also make similar predictions irrespective of the fact that the science you are studying and using does not and cannot follow the same method and, specifically, that applying the method of the natural sciences to this science is asinine?”

              • Ken B says:

                Which doesn’t answer the question Bala. Tel’s point is that if you insulate your predictions from any possible test they become useless. Like the thermostat example I gave. Cannot be refuted, totally useless. You on the other hand seem to maintain useful but non-falsifiable predictions can exist. So, how does that work?

              • Bala says:

                “Tel’s point is that if you insulate your predictions from any possible test they become useless.”

                Looks like your comprehension skills decided to take a vacation. My answer simply was that in certain spheres, to expect predictions out of your knowledge can be fairly asinine and that Economics happens to be one of them. We therefore need to be content with explanations and not expect predictions. To claim that that makes it useless is fairly silly because it is the framework with which you understand everything else. It’s like saying “My eye glasses help me see the world clearly but since they do not help me predict what’s in front of me, they are useless.”

              • Ken B says:

                Did you notice we said the *predictions* become useless not the theory? Just as a matter of reading comprehension and logic I mean?

              • Bob Murphy says:

                I predict this debate will be useless.

              • Ken B says:

                Ah Bob, if only you showed such acument in your bets!

              • Bala says:

                Ken,

                Here’s what Tel said which I responded to.

                “Sure, but if theory never delivers a workable prediction then it is useless theory.”

                And then you say THIS???

                “Did you notice we said the *predictions* become useless not the theory?”

                WOW!! And then you talk of my comprehension and logic!!! Maybe you should just read first.

  7. A. Grigorova says:

    Hello there my Keynesian slaying friend.

    This is a very lovely example of how increases in minimum wage can affect the unemployment of youth in America. I took the liberty to check your work, not that I was skeptical, but more or less that I carry the attitude of a person who feels the best thing that are done, are done yourself.

    I think you may be working with old information with respect to youth unemployment, as with the figures I have thinks are more dismal. Anyway, my referrence for age 16-19 unemployment rates used firgures from http://www.bls.gov/lau/ptable14full2012.pdf and I used same minimum wage law url as you (http://www.dol.gov/whd/minwage/america.htm)

    The picture is even more bleek when you couple this info with the piss poor unemployment rates of age 20-24 as well! Yikes!

    Here is what I got:

    Youth unemployment rates as a percentage by state vs. states with > federal minimum wage.
    (Order: > age 16-19 unemployment rate, decending. The * is to represent state with a minimum wage greater than federal minimum)

    Califronia- 34.6*
    District of Columbia- 34.0*
    Sout Carolina- 31.7
    Arizona- 30.9*
    Hawaii- 29.7
    Georgia- 29.6
    Washington- 28.6*
    New York- 28.4
    Louisiana- 27.8
    Illinois- 27.1*
    Colorado- 26.2*
    Arkansas- 26.2
    Rhode Island- 25.9*
    Virginia- 25.9
    Nevada- 25.6*
    North Carolina- 25.4
    Connecticut- 25.4*
    New Jersey- 24.7
    Maine- 24.6*
    Kentucky- 24.2
    Mississippi- 24.0
    Missouri- 23.9*
    Delaware- 23.8
    Florida- 23.5*
    Oregon- 23.3
    idaho- 22.5
    Indiana- 21.9
    Maryland- 21.8
    Michigan- 21.5*
    Kansas- 21.3
    Texas- 21.1
    Utah- 20.7
    Alaska- 20.6*
    Wisconsin- 20.4
    Tennessee- 20.4
    Massachusetts- 20.2*
    New Mexico- 18.9*
    West Virginia- 18.8
    Minnesota- 18.6
    Alabama- 18.5%
    Vermont- 17.7*
    Wyoming- 17.1
    New Hampshire- 17.1
    Ohio- 16.9*
    Iowa- 16.8
    Pennsylvania- 16.8
    Oklahoma- 14.5
    Nebraska- 13.5
    North Dakota- 12.4
    South Dakota- 11.4
    Montana- 10.8

    Keep up the good work!
    Cheers.
    A. Grigorova

    • A. Grigorova says:

      Oops, forgot my averages. Looks like I had a blonde moment there! Here we go:

      Average unemployment rate of age 16-19 unemployment rate of states with > than federal minimum wage: 26.25-%

      Average unemployment rate of age 16-19 unemployment with no law increasing minimum wage over federal minimum wage: 21.24-%

      Number of states with > than federal minimum wage (or *):

      In the top 5: 3*
      Bottom 5: 0*

      Top 25: 12*
      Lower 26: 6*

      1-10: 5*
      42-51: 1*

      • Peter says:

        Just curious: Are your averages weighted or raw based on the number of states?

    • Tel says:

      http://lnx-bsp.net/murph_min_wage.png

      That’s your data, not Murphy’s because yours was easier to cut and paste.

  8. Mike Sax says:

    The Europeans listen to Krugman? Really? In fact they listen to Bob-look at all that wonderful austerity.

    Bob what is that on the right about “from 1-5, 4 years, from 46-50 1 year?”

  9. Bharat says:

    If you run a simple regression state by state for the year 2011, with the minimum wage as your independent variable and youth unemployment (the average for the year) as your dependent variable, you’ll get a coefficient of 3.955 (significant at 10%; so increase in minimum wage of $1 increase youth unemployment by 3.955%). If you run a multiple regression, correcting for growth rates, the coefficient of the minimum wage comes out to be 4.619 (significant at 6%).

    Of course, if I correct for more variables, I would certainly get different values and it’s possible I could even get negative values. Again, this is why econometrics is such a poor methodology compared to praxeology. You can use an econometric study to show whatever you want. And unlike the natural sciences, you cannot properly control for variables.

    • Bruce says:

      It’s dangerous to look ignorant, but I suppose I will.

      What is: praxeology???

      • Bharat says:

        It’s the deductive method of Austrian economics. It starts with the premise that persons act, that actions are purposeful, they involve the use of a means to attain a end and are motivated by that end. It goes on to deduce conclusions from this premise through the use of logic: if a premise is true, and valid logic is applied, then the conclusion must also be true. Logical chains can be formed by using the old conclusion as a new premise and then again using valid logic. Austrians have used this method to explain marginal utility, the law of demand, how the business cycle occurs, and much more.

        If you’re looking for an introduction to the type of reasoning used by Austrian economists, you can look at An Introduction to Economic Reasoning by David Gordon or look at praxgirl‘s videos on YouTube. If you’re interested in reading about Austrian economics in depth, you can look at Man, Economy, and State by Rothbard.

        • guest says:

          This video by Jeffrey Herbener is good, too:

          Is Austrian Economics ‘Unscientific’?
          http://www.youtube.com/watch?v=6bF_gPioOoc

        • Tel says:

          Praxeology comes to a bit of a dead end when you get to “actions are purposeful” but the purpose is, well whatever you want it to be.

          Combining praxeology with Darwin’s theory of evolution results in a stronger set of axioms. Actions are purposeful and the purpose is invariably either survival of the individual, or reproduction (which in the long run are the same thing). Pretty much all of modern biology has taken this as a basis.

          • Ken B says:

            Hence my earlier jab at Bob: “he thinks Austrian economics is a science but biology isn’t.”

          • Bala says:

            “Praxeology comes to a bit of a dead end when you get to “actions are purposeful””

            Actually, it doesn’t and you are wrong.

            “but the purpose is, well whatever you want it to be.”

            Economics does not care about what the purpose it. It only identifies the existence of a purpose.

        • Bharat says:

          I highly recommend the link ‘guest’ posted.

  10. Daniel Kuehn says:

    re: “especially since they back up what is literally textbook economics.”

    I find this “textbook economics” point really odd.

    In my labor textbook, by Borjas, which is an undergraduate textbook, they talk about monopsonistic labor markets. They probably bring it up in some micro principles textbooks.

    So when people say things like “textbook economics” around this, what they really mean is “what we teach students in the first two weeks of class”. It seems to me that shouldn’t necessarily be the obvious answer to a question. I’m not even persuaded that the minimum wage can necessarily have a positive minimum wage effect (although I’m usually more persuaded by natural experiments than IVs). I’m of the “no or very small negative effect” camp. But it seems silly to act like the answer is beyond the pale or some how strikes at the heart of good economics.

  11. The Existential Christian says:

    What’s with all the italicized comments?

    • The Existential Christian says:

      Including mine!

      • Bharat says:

        Lies

      • Joseph Fetz says:

        It happens about once a month.

        • Ken B says:

          Oh gawd Joe, not the lunar conspiracy!

        • The Existential Christian says:

          It’s like everybody is really trying to make a point.

          • Ken B says:

            It’s the Italiratti. They are the secret society Bob warned us of months ago. And I scoffed! Mea culpa, mea culpa, mea maxima culpa.

          • Joseph Fetz says:

            It’s almost worse than the all caps, because it’s easy for me to read every word as if it was yelled. But emphasis? That’s a different animal.

            It’s like I have to read everybody’s comment as if it were spoken by Captain Kirk.

      • joshua says:

        Somebody forgot to close an italicized tag. Sometimes you can fix it by closing or double-closing your own.

        • Bharat says:

          Oh lmao is that how it happens? I tried purposely making my comments italicized to see if it turned out un-italicized but it didn’t work out so well =/

  12. Daniel Kuehn says:

    I am not sure about the empirical exercise, not that I’ve provided any better. What matters of course is the real minimum wage and if we’re looking across states we probably want to look at variations in local prices. I also see Nevada and California in the top five and no big housing bust states in the bottom five (and the Dakotas are there, with their oil boom). I’m just not sure what to make of it.

    State level variation is exactly where we ought to go after C&K and that’s exactly where N&W did go and they found a small negative effect.

    I feel the need to reread a lot of this after looking at some of Don’s links the other day (the Adie and Galloway review of Myth and Measurement, specifically). One of the complaints about C&K was that they looked at employment and not hours. This may imply that the minimum wage has different intensive and extensive effects. It would be interesting to see if that is what is driving a lot of the disagreement. My work on job creation tax credits last semester found that too – negative extensive effects, positive intensive effects.

    It was easy to see why that was the case for the tax credit…. I’m still noodling over why it might be different for the minimum wage. I’m not sure.

  13. Steven E Landsburg says:

    Okay, so there are 19 “red” states. If they were ranked in a randomly chosen order, I get:

    Probability that there are at least 6 red states in the top 10: about 11%

    Probability that there is at most 1 red state in the bottom 10: about 4%

    Probability that there are at least 6 in the top 10 AND at most one in the bottom 10: about 1%

    This is all based on some quick mucking around with Stirling’s formula. Subjective probability that I got it right: about 85%.

    • Adrian Gabriel says:

      Steve, they weren’t chose in random order. Murphy used basic praxeolgoical common sense here. Maybe we should all just go have lots of sex “randomly” and it would be safer sex in the end. You made much more silly assumptions that Murphy my friend.

      • Bob Murphy says:

        Adrian wrote:

        Steve, they weren’t chose in random order.

        No, Adrian, Steve was just trying to put some numbers to my point. He was saying IF it were random, then you’d only expect to see the outcome of “at least 6Ys in the top 10 and at most 1 Y in the bottom 10” 1% of the time. So since we do see that, it makes us think that the Ys aren’t randomly dispersed through the list.

  14. Daniel Kuehn says:

    Here is a question for you Bob: if a state naturally had a high teen unemployment rate, would they be more likely or less likely to institute a higher minimum wage (assuming it really is higher – as I noted above, a federal minimum wage in Mississippi may very well be higher in real terms than a state minimum wage in California).

    • Ken B says:

      Daniel, that question sounds like you admit the strong correlation. Now as ever correlation isn’t causation, but in the context of the strength of the Econ 101 argument, I think admitting the correlation is close to the whole game.

      • Major_Freedom says:

        Not if one feels guilty about sanctioning others entering into voluntary $2.00/hr wage contracts, and then conflate one’s feeling of guilt with some sort of mystical insight that enshrines one as now having a social duty to explain to the congregation why sanctioning it is a bad idea.

        Then we cannot, indeed should not, be so hasty in putting the correlation together with the theory and painting a picture that is guilt inducing.

  15. Yancey Ward says:

    Fixed?

    • joshua says:

      Your extra closing tags are apparently powerless here.

  16. Jordan says:

    Youth Unemployment = B0 + B1(# of Republicans in Congress) – B2(# of Krugman blog posts) + B3(# of Murphy Youtube videos) + e

  17. Walter Wessels says:

    Most labor economists dismiss using the unemployment rate as a measure of the effect of the minimum wage. Why? Because economic theory says the minimum wage should reduce employment, so you should look at the employment rate. The “unemployment rate” does not measure the number of people who lost their jobs due to the minimum wage (or anything else, for that matter. It measures the number of persons actively looking to work. As Jacob Mincer showed, a minimum wage could increase the unemployment rate if it caused more people to want to look for work (which he called the “pull” effect. Alternatively, the minimum wage could reduce unemployment if the lack of jobs and turnover discouraged workers from looking for jobs. What is worse, an increase in unemployment could, but need not, be consistent with the minimum wage increasing the value of looking for a job, thus making all workers better off, and alternatively, a decrease in unemployment could (but need not ) be consistent with a minimum wage lowering the value of looking for a job (making everyone worse off). To put it another way, the effect of minimum wage on unemployment tells us nothing, at least without other data. So once again, one can become labeled by the Bureau of Labor Statistics as unemployed without losing a job, and one can lose a job, leave the labor force, and not be “unemployed” at all.

    • Jordan says:

      I find the bulk of this point insightful and helpful. I’ve never thought about how the minimum wage may affect unemployment rates through other channels.

      I got lost a little starting with “What is worse”. Can someone help me out? Specifically what does the following mean: “an increase in unemployment could, but need not, be consistent with the minimum wage increasing the value of looking for a job, thus making all workers better off…”

      So the value of looking for a job increases because the prospect of landing a job brings higher wages than before for marginally unemployed persons. More discouraged workers begin re-entering the labor force and therefore the unemployment rate goes up. How does this benefit “all workers”? If the value of looking for a job increases, why does this help those who don’t need to look for a job?

  18. Ken B says:

    Bob, this is very compelling but I think you muddy the water a bit by including your average calculation. You are presenting a data set which is order data, right? That’s why Steve’s test is apposite. I get the sense some readers don’t grok the difference between this and the more usual regressions. I guess I am suggesting you might want to explain this.

  19. Peter says:

    Remember Alan Krueger’s “study”? He supposedly proved that raising the minimum wage actually increased employment. What he forgot to mention was that even though employment indeed increased, the average number of hours worked declined by a higher percentage.

    • Bruce says:

      Could this be response that would reduce the overall effective wage by reducing benefits by using more part-time that often have none.

      • Tel says:

        Yeah, it’s called unpaid overtime (and generally undocumented too).

    • Robert Fellner says:

      The Krueger study is flawed.

      The data they used was wrong and re-evaluations of the study showed the opposite effect. They compiled their data based on phone interviews. Other economists re-evaluated the study using actual data – payroll data, specifically – and the results were the opposite of what Card & Krueger found. Read the first 3 paragraphs of this 1996 paper, The crippling flaws in the NJ fast food study , which aptly demonstrates how the 1992 study is fundamentally flawed. http://epionline.org/studies/epi_njfastfood_04-1996.pdf

      Another re-evaluation of the Card/Krueger study found the same flaws:

      “Second, estimates of the employment effect of the New Jersey minimum wage increase from the payroll data lead to the opposite conclusion from that reached by Card/Krueger.”

      http://www.nber.org/papers/w5224

      • Jordan says:

        http://prospect.org/article/cooked-order

        I’ve been told numerous times that the EPI’s payroll data was handpicked and that when Neumark / Wascher use their own payroll data without including EPI they actually find no effect. Can anyone verify the above article or link to a better source on the issue? Was the EPI a neutral authority on the topic?

        • Economic Freedom says:

          See Neumark/Wascher’s follow-up in the American Economic Review from 2000, at:

          http://www.davidson.edu/academic/economics/foley/eco324_s06/Neumark_Wascher%20AER%20(2000).pdf

          There’s extensive discussion of the data sets, both EPI’s and N/W’s. On PDF page 24 (article page 1384) of this 36-page article, footnote #34, the authors write:

          “Note also that the issue of the larger standard errors in the subsample we collected, which has been raised in a polemical essay by John Schmitt (1996), is essentially a “red herring.” Given that the EPI sample was drawn from areas where one or two franchisees (or a corporation) owned all or most of the restaurants (and thus was nonrandom), the greater uniformity of employment changes in the EPI sample should not be surprising, and the statistical tests performed by Schmitt-regarding whether the two subsamples are random samples of the same population-are uninformative. This point is reinforced by the fact that the representation of chains in the two samples is very different, because the data on all company-owned restaurants in a chain arrived together. Thus, for example, the data collected by the EPI include no Roy Rogers company-owned restaurants, whereas these constitute 99 of the 154 restaurants in the data we collected.

          The other issue Schmitt discusses is the evolution of estimates through the early drafts of this study. This also seems to us to be a nonstarter. Although the earliest draft received a lot of attention because of the minimum-wage debate at the time, it was a preliminary version (and so labeled), and the estimates have changed because of corrections and additions to the data, changes in the definition of variables, and more generally responses to comments and advice from colleagues.”

          You might be interested in learning that the article you linked to above is by a far-left economist named John Schmitt, writing for a far left magazine called “The American Forum” (TAP), which often takes funding from far left sources like the Schumann Center for Media and Democracy (controlled by Bill Moyers); the “Open Society Institute” (controlled by George Soros); etc. Neumark and Wascher refer to the Schmitt article in TAP as a “polemical essay”, but it was really a “hit piece,” intending to impugn their characters.

          Schmitt also belongs to, and writes for, a “progressive” think-tank in Washington, DC called the “Center for Economic and Policy Research” (CEPR), which not only has published many articles supporting increases in the minimum wage — ostensibly because “no data exist” showing any harmful effects whatsoever of wage floors on employment — but whose co-director, Mark Weisbrot, has spoken in positive glowing terms of the wonderful economic reforms in Venezuela implemented by Hugo Chavez.

          I don’t think any article or “study” produced by The American Forum, the Center for Economic and Policy Research, or the people working there like John Schmitt, Dean Baker, Mark Weisbrot, et al., need to be taken too seriously.

          Finally, most of these regression meta-analyses supposedly showing “publication bias” in minimum-wage studies are highly flawed; what the authors are hoping for is that no one will spend the necessary time looking too closely at their data, their methods, or their priors too closely, which is why they have to dump lots of data on readers in the first place.

          • Jordan says:

            Great, thanks for reply, I’ll take a look

  20. matt says:

    kind of wierd how Missouri has such a high unemployment rate, considering how low that of its neighbors is.

  21. Andrew_M_Garland says:

    Our society of (many) thinking, responsible people is set on task after task to analyze each new government program. Often these programs are announced and implemented with lightning speed and at astronomical cost.

    Obama wants higher minimum wages on all states, tied to inflation.

    We are reduced to arguing about what studies appear in the world which support or criticize that policy proposal. But, the relevant studies and analyses are the ones that our government is relying on. What are those?

    This is the United States Government, of, for, and by the People. The public cannot participate in a government that runs on policies that are hidden. The government owes us a “white paper” giving an explanation of the thinking behind each program, policy, and bill.

    We need proposed results, proposed evolution, methods, justifications, comparative studies, past successes, funding sources, the works. These must already exist in the background; the government does not want to release its reasons and thinking. The press and public should be able to review these undelivered documents. The government should be proud to display its carefully researched and supported policies. We deserve this as a free people.

    The absence of these policy papers is appalling. Instead of open and proud policy, we have closed, imperial government. Sadly, we are controlled by a tyranny, if our government sees no reason to explain and justify its actions. Are our politicians ashamed or afraid of what they propose?

  22. ken says:

    Some years ago I did a plot of minimum wage increases against young black male unemployment rates. That group of 16-24 year olds is the hardest hit by minimum wage regulations. I don’t have what I put together anymore (I sent it to Thomas Sowell in 2003, I think), and I think it would be worthwhile for someone to put together again. The unemployment rate among that demographic is rarely below 40% nationwide, and will increase to 50 or 60% right after a minimum wage increase. This means that a huge number of kids who want work, and are looking for it, can’t find it. So what else are they going to do to learn valuable job skills and get out of the welfare trap?

    • Economic Freedom says:

      >>>The unemployment rate among that demographic is rarely below 40% nationwide, and will increase to 50 or 60% right after a minimum wage increase. This means that a huge number of kids who want work, and are looking for it, can’t find it. So what else are they going to do to learn valuable job skills and get out of the welfare trap?

      And Obama has previously said (to much audience approval and enthusiastic applause) that he wants a “Civilian National Security Force”. Hmmm. I wonder what demographic might be unemployed and available for realizing this noble vision?

      • ken says:

        You’re kidding! Liberals love black people, from a distance and so long as they mostly stay in their neighborhoods. Giving them guns and putting them in uniforms and placing them on American streets would be the progressive version of apocalypse. We have a drug war, prisons, and welfare to keep them where they belong. “We” meaning those well-intentioned liberals, of course.

        • Economic Freedom says:

          >>>Liberals love black people, from a distance and so long as they mostly stay in their neighborhoods

          That’s the liberal definition of “love.”

          >>>Giving them guns and putting them in uniforms and placing them on American streets would be the progressive version of apocalypse.

          Not if they’ve been adequately brainwashed by school, media, and Dear Leader. In any case, they need not be armed. I see the Civilian National Security Force as being a group of informants and nanny-state busybodies; e.g., legislation might be passed requiring you to open the door of your home to a “squad” of them so they can ascertain if you’re living the politically correct Green Lifestyle. Naturally, they would have to check if you’re a gun owner; have you made your Obamacare payments, etc.

          In a perverse sense, that might even empower some of us; e.g., “Say, guys, if you leave me alone, I can tell you which of my neighbors has a really BIG stash of styrofoam.”

  23. William Anderson says:

    Bob, I have mentioned this post in my latest KIW piece on Krugman and the minimum wage.

    http://krugman-in-wonderland.blogspot.com/2013/02/raise-that-teenage-unemployment-rate.html

    As I see it, Krugman is trying to simultaneously argue some mutually-exclusive points, and ultimately turns to his ad hominem fallacy:

    Republicans are evil
    Republicans oppose increasing the minimum wage
    Therefore, anyone who opposes the minimum wage is evil.

  24. Harold says:

    Am I missing something? Bob quotes a paper that says “if you just naively look at the states that have higher minimum wage laws, then they have lower employment growth, but once you correct for the broader trends in employment growth among states, then the impact of the minimum wage laws per se disappears”. Bob then decides to ” to do my own, very naive, straightforward look at the data” and finds something very similar to that found in the original paper before “broader trends” were corrected for. Isn’t this exactly what we would have expected?

    The paper concludes: “Economists have conducted hundreds of studies of the employment impact of the minimum wage. Summarizing those studies is a daunting task, but two recent meta-studies analyzing the research conducted since the early 1990s concludes that the minimum wage has little or no discernible effect on the employment prospects of low-wage workers.” If this is an accurate description of careful meta-analysis, then should we give more weight to a back-of-the-envelope study conducted overnight and conclude the opposite?

  25. Mike says:

    Why do you think some states have higher than federal minimum wages? because, on average, they have higher wages too. duh

    States with higher minimum wages are (most of them) those with higher median wages:
    http://www.careerinfonet.org/wages3.asp?soccode=493023&stfips=06

  26. Ken B says:

    This debate has become other wordly, but perhaps someone can explain something to me. Say I define a new quantity U defined as the sum of the squares of other real-values quantities. I propound the new law of nature “U is always greater than or equal to 0”. Have I propounded a useful new law? I say no, and Richard Feynman, whose example this is, said no. I will leave it to you to figure out why. But my question is, why the hell does this mean Richard P Feynman is a follower of L Ron Hubbard?

    • Bala says:

      That’s because a “law” is only an identification of a general condition of existence. When you speak of scientific “laws” (as in the natural sciences), they are usually “laws” when they cover existence in general or a broad class of existents.

      In the specific case you have taken up, it is the identification of the property of a particular expression defined as it is. That is not a general condition of existence. It is the property of a particular, specially defined expression.

      And then you claim you understand “science”? So, do you understand why it is justifiable to identify you as the head priest of The Church of Scientology?

      • Ken B says:

        “it is the identification of the property of a particular expression defined as it is. That is not a general condition of existence. It is the property of a particular, specially defined expression”

        No because the form of the proposed law is the same as the form of the second law of thermodynamics, where the expression is for entropy and the rule is that it is always >= 0 in any process. So that’s not the problem. The U >= 0 is not a useful law for quite a different reason. I’m sure most here see why. Let U = s^2 where s is the entropy change just discussed for example.

        I propose another law. Well not a law, more of an ordnance. Anyone who spends weeks confusing scientism and scientology is in no position to lecture those who quote Feynman on what is or is not science.

        • Bala says:

          The difference between you and me is that I am ready to accept my mistake and you are not. Feel free to engage in more scientism or whatever you wish to call it. As long as I am concerned, the very fact that you think that the method of the natural sciences are the ONLY valid method of science disqualifies you from ANY discussion on economics.

          • Bala says:

            Apart from that, the mistake I made was tangential to the core issue I was addressing – your error in claiming that empirical testing can validate or invalidate conclusions drawn through deductive reasoning and claiming that if theory claims to to be immune from empirical testing, it is not science.

            Any attempt to use my (fairly irrelevant) mix-up as an excuse to run away from that debate would be a rather shameful one.

    • Bala says:

      OK! I get where I am wrong. I should have said “head priest of The Church of Scientism”.

      Thanks for correcting me, but that still leaves you right there – with a miserably wrong understanding of science and the scientific method. You are taking “the scientistic method” as “the scientific method”. Once you get rid of that baggage, you may make more sense.

  27. Xan says:

    Anyone seen Krugman’s latest? Apparently, now, labor demand curves DO slope downward. Spain’s wages are too high, which is causing “incredibly high unemployment.”

    http://krugman.blogs.nytimes.com/2013/02/25/a-tale-of-two-adjustments/

Leave a Reply