23 Feb 2013

Minimum Wage Analysis: Killing Econ 101 Via the Magic of Econometrics

Economics 87 Comments

I tried working my way through the Dube, Lester, and Reich (2010) paper that I think is a good representative of the “latest research” that allegedly overturns the old consensus. Now that I am pretty sure I know what happened, I have to say that if these results are allowed to guide the debate, then something really wrong is going on here.

First, let’s make sure we understand the context. Here’s their abstract:

Abstract—We use policy discontinuities at state borders to identify the
effects of minimum wages on earnings and employment in restaurants
and other low-wage sectors. Our approach generalizes the case study
method by considering all local differences in minimum wage policies
between 1990 and 2006. We compare all contiguous county-pairs in the
United States that straddle a state border and find no adverse employment
effects. We show that traditional approaches that do not account for local
economic conditions tend to produce spurious negative effects due to spatial heterogeneities in employment trends that are unrelated to minimum
wage policies.
Our findings are robust to allowing for long-term effects of
minimum wage changes.
[Bold added.]

That part I put in bold is crucial. When Dube, Lester, and Reich look at contiguous county-pairs that straddle a state line (and hence can have different minimum wage laws applicable to them), they too find that a “naive” regression will show that minimum wage hikes will retard the growth in unskilled employment.

However, they “correct” this result by putting in an additional explanatory variable that picks up the effect of the “state-specific trend” in employment. Once you do that, all of a sudden the point estimate of employment turns slightly positive, or at least is close to zero (depending on the specification). Now these authors don’t ever really spell it out in plain English, so let me quote from the John Schmitt CEPR literature review that Krugman liked so much. The CEPR paper says what Dube, Lester, and Reich did, and how their findings (allegedly) turned the original, anti-minimum-wage findings on their head:

Dube, Lester, and Reich’s study also identified an important flaw in much of the earlier minimumwage research based on the analysis of state-level employment patterns. The three economists
demonstrated that overall employment trends vary substantially across region, with overall
employment generally growing rapidly in parts of the country where minimum wages are low (the
South, for example) and growing more slowly in parts of the country where minimum wages tend to
be higher (the Northeast, for example). Since no researchers (even the harshest critics of the
minimum wage) believe that the minimum wage levels prevailing in the United States have had any
impact on the overall level of employment, failure to control for these underlying differences in
regional employment trends, Dube, Lester, and Reich argued, can bias statistical analyses of the
minimum wage. Standard statistical analyses that do not control for this “spatial correlation” in the
minimum wage will attribute the better employment performance in low minimum-wage states to
the lower minimum wage, rather than to whatever the real cause is that is driving the faster overall
job growth in these states (good weather, for example). Dube, Lester, and Reich use a dataset of
restaurant employment in all counties (for which they have continuous data from 1990 through
2006), not just those that lie along state borders and are able to closely match earlier research that
finds job losses associated with the minimum wage. But, once they control for region of the country,
these same earlier statistical techniques show no employment losses. They conclude: “The large
negative elasticities in the traditional specification are generated primarily by regional and local
differences in employment trends that are unrelated to minimum wage policies.”

This, I submit, is crazy talk, assuming I correctly understood what Dube, Lester, and Reich actually did in their study. Here’s my list objections:

#1)==> We are already supposed to be comparing apples to apples, by focusing on contiguous county-pairs across state lines, right? That’s why we did it this way. Yes, if the minimum wage laws all happened to be raised in an area that had a bunch of oranges to be picked, and then there was a bad harvest every time the minimum happened to get increased, then you could get a spurious negative result. But if you’re looking at contiguous county-pairs, then this shouldn’t happen. If there’s a bad orange crop in one county, then there should be a bad orange crop in the next county over (which happens to be in an adjacent state). To find the pairwise negative effect on employment that the old literature documented quite clearly, but then to correct for “state-wide” employment trends due to good weather, is crazy. You are overcorrecting.

#2)==> What Dube, Lester, and Reich are really saying here, is that maybe for some reason minimum wage hikes happen to be concentrated in regions that have lower than average employment growth. Hence, just because we find that teenage employment grows more slowly in regions with higher minimum wages, doesn’t mean we can blame it on the relatively higher minimum wage. But hang on a second. Minimum wage hikes aren’t randomly distributed around the country, such that we might happen to get an outcome where they tend to be concentrated in slow-growth regions. On the contrary, minimum wage hikes are implemented by “progressive” legislatures, who also (given my economic worldview) implement other laws that retard adult employment growth.

For example, suppose that if a state legislature jacks up the minimum wage, then it is also likely to pass “pro-labor” stuff like laws giving unions more organizing power, laws allowing unfairly terminated employees to receive years of back pay, and laws granting extra perks for maternity leave. Now, these last three items I listed: Would they reduce the employers’ incentives to hire teenagers or adults, more? On the margin, they would make it costlier to hire adults, because if penalties are expressed in years of back pay, or have to do with paid leave, or strengthen unions who traditionally are going to organize adults…You get the picture. Adults make more than teenagers, and so these rules will penalize adult employment more than teenage employment.

Thus, if my model here is correct, it would produce the pattern we actually see: Looking narrowly at minimum wage laws, they seem to retard teenage employment. But then when you ask if states with high minimum wage laws have a bigger slowdown in teen employment versus adult employment, the signal becomes much weaker. It looks like, by dumb luck, for some reason all the minimum wage hikes happen in states that also have slower-than-average employment growth among adults. And hey maybe this is due to weather, even though the weather doesn’t seem to be slowing the adult or teenage employment growth in the next county, which is in a right-to-work state with no minimum wage law.

87 Responses to “Minimum Wage Analysis: Killing Econ 101 Via the Magic of Econometrics”

  1. Yancey Ward says:

    I am sure DK will have some explanation.

    • Chris P says:

      Yes. You know there’s a perfectly good explanation. I’m sure Daniel will enlighten us.

  2. Ben G. says:

    “Since no researchers (even the harshest critics of the minimum wage) believe that the minimum wage levels prevailing in the United States have had any impact on the overall level of employment, failure to control for these underlying differences in regional employment trends, Dube, Lester, and Reich argued, can bias statistical analyses of the minimum wage.”

    Really? Not even the harshest critics believe this? Am I missing something?

    • Jonathan Finegold says:

      It might be because before the more recent changes, the real value of the minimum wage has eroded with inflation.

    • Ken B says:

      I don’t believe it quite either, but as Bob says the effect will be small. It will be even smaller in contiguous counties, because of cross county line workers. Higher unemployment in county A means more applicants from county A than there would otherwise be, and by hypothesis the skill sets are low and thus comparable, so county B hires workers from A, smoothing differences a bit.

  3. Bharat says:

    Can someone explain what this means? I’ve read it a few times but don’t really understand:

    “Since no researchers (even the harshest critics of the
    minimum wage) believe that the minimum wage levels prevailing in the United States have had any
    impact on the overall level of employment”

    My initial thought is “well of course the minimum wage would affect the overall level of employment in the US” but I’m pretty sure I’m misunderstanding what Schmitt is trying to say. Does it mean the MW for a specific state doesn’t affect the employment of the entire country?

    • Bob Murphy says:

      Bharat wrote:

      Can someone explain what this means? I’ve read it a few times but don’t really understand:
      “Since no researchers (even the harshest critics of the
      minimum wage) believe that the minimum wage levels prevailing in the United States have had any impact on the overall level of employment”

      I think what he’s saying (although I agree it’s not worded the best way) is that nobody thinks that if you have two identical states, and then one state puts in a 50-cent higher minimum wage, that overall employment growth is going to be drastically reduced in that state. Even if you think the growth of employment among unskilled people will be much much lower, you still won’t think that would do much to the employment trends overall. The growth in overall employment between those two states should still be roughly equal.

      (Keep in mind something like 2% of the workforce right now makes minimum wage. So if employment growth were originally 10%, and the minimum wage hike reduced teen employment growth by half, then the one state would still have 10% total growth, while the minimum wage hiking state would now only have 9.9% growth in total employment.)

  4. ThomasL says:

    The most important thing about a conclusion is not that it be novel, but that it be true… Something not fully baked into the incentive structure of journals.

  5. Daniel Kuehn says:

    I can’t look closely at this again tonight – but just for clarification: are you saying that controlling for time-variant state specific trends in a panel analysis is going to bias the results? My first read of your post sounds wacky on its face so I want to make sure I understand and don’t falsely attribute something to you.

    • Bob Murphy says:

      I can’t look closely at this again tonight – but just for clarification: are you saying that controlling for time-variant state specific trends in a panel analysis is going to bias the results?

      I am saying that if you are looking at pairs of contiguous counties that straddle state borders, in order to isolate the effects of their differences in minimum wage laws, and then you also throw in a variable correcting for employment trends in one state versus the other, that this makes no sense to me and if it erases the first result, then I think something is screwy.

      To put it in other words, what kind of trend occurs in the one state versus the other, that wouldn’t already be captured by the fact that we’re looking at pairs of contiguous counties? The only thing that makes any sense is state government policies. Weather etc. should already be accounted for.

      So now it seems we’re saying: Yes, minimum wage policies by themselves appear to hurt teen employment, if we control for everything except other state policies that hurt overall employment. So only if you can explain why those other state policies might hurt adult employment more than teen employment, does the Econ 101 logic survive.

      And so, I’ve said that actually I think that’s true. So I’m going with the Econ 101 logic, and the pairwise comparison of contiguous counties, both of which say making labor more expensive leads employers to buy less of it.

      • Joseph Fetz says:

        If I am getting this right, then it is kind of like if you were to observe a match between two teams A and B, and that team B won as a result. But since this doesn’t conform to our bias, lets make an adjustment to our model such that the points scored by teams A and B are adjusted in such a way that they are equal. Then we form our conclusion thenceforth: team A actually won the match!

        Ah science, you’ve gotta love it.

      • Ken B says:

        “I am saying that if you are looking at pairs of contiguous counties that straddle state borders, in order to isolate the effects of their differences in minimum wage laws, and then you also throw in a variable correcting for employment trends in one state versus the other, that this makes no sense to me and if it erases the first result, then I think something is screwy.”

        +1

        • Bob Murphy says:

          Ken B. your existence makes me doubt my objectivity. When you disagree with me, I am sure you are nuts. When you agree with me, especially on something mathematical, it confirms to me that I’m right.

          • Ken B says:

            Well we agree on something else then Bob — your objectivity is doubtful!
            🙂

            Seriously, this post is just about perfect Bob.

            • Joseph Fetz says:

              Apparently this (^) bonobo is a swinger.

              • Ken B says:

                I’m the perfect date Joe: I get drunk quickly and I’m easy.

              • Joseph Fetz says:

                So you’re a floozy named Ken? This is all too much, because now I can’t get that trip to Thailand out of my head (long story).

              • Ken B says:

                What happens in the Thailand Empire Motel stays in the Thailand Empire Motel.

              • Joseph Fetz says:

                There are a lot of things that I experienced/observed in Thailand that are going to stay there forever.

      • Ken B says:

        “To put it in other words, what kind of trend occurs in the one state versus the other, that wouldn’t already be captured by the fact that we’re looking at pairs of contiguous counties? The only thing that makes any sense is state government policies. Weather etc. should already be accounted for.”

        Exactly. Plus one.

        Look, you can always show anything you want with an extra causal factor so ideally you look for places where all else is equal. That’s what the contiguous counties do here, as Bob says. If you compared random pairs of counties you would use state level correction.

        Say I want to test the effect of a upbringing on a mental trait, and I used identical twins raised apart as my test subjects. Would I then also correct for current level of education, which is related to inheritance, as a way of correcting for inheritance? No, that’s what twins gave me already.

        • Bob Murphy says:

          Ken B. wrote:

          Say I want to test the effect of a upbringing on a mental trait, and I used identical twins raised apart as my test subjects. Would I then also correct for current level of education, which is related to inheritance, as a way of correcting for inheritance? No, that’s what twins gave me already.

          Yes, exactly! My analogy-generator was backfiring on this one.

          • Anonymous says:

            So if this was not a detailed study of contiguous counties, then the correction for regional growth differences would be relevant right? But the way the study is set up, any negative effect of minimum wages is “controlled away”. More or less?

            • Bob Murphy says:

              Anon wrote:

              So if this was not a detailed study of contiguous counties, then the correction for regional growth differences would be relevant right?

              My objections would be much less vociferous in this case, for sure. I would still think they might be missing something with the fact that min wage laws are correlated with other anti-growth laws, but at least their approach would make sense. (And in fact, I think perhaps in one part of their paper they *do* exactly this, looking at state-level data.)

            • Ken B says:

              More or less. Design beats correction any day. And the correction should be for other govt policy for which is a poor proxy anyhoo.

      • Ryan Murphy says:

        If it were “overcorrecting,” then the state-level trend would just be statistically insignificant. I believe the way Daniel phrases the question is correct.

        • Bob Murphy says:

          Ryan Murphy wrote:

          If it were “overcorrecting,” then the state-level trend would just be statistically insignificant. I believe the way Daniel phrases the question is correct.

          Ryan, you wanna get nuts? Let’s get nuts!

          I sense this is going to be the Great Blog Debate of 2013, comparable to the Debt Debate of 2012.

          Before we go down this path, wherein I plan on coming up with simulated economies in an Excel spreadsheet on which I will run regressions, let me ask you this:

          Why are we looking at contiguous county pairs that straddle state borders, in the first place? What’s the point of that procedure?

          When you answer that question, then you’ll see why including a variable for “state employment trends” is crazy.

          • Joseph Fetz says:

            I would really love to see how DLR would handle this situation if it were between Tijuana and San Diego.

          • Ryan Murphy says:

            I think increasing minimum wage will increase unemployment. I just don’t think your criticism was correct.

            The point of these studies is that you get to pretend that it’s a natural experiment. From what I can tell, you can control for everything that would ever make sense to control for and still get more unemployment. There may still be something hypothetically that the unemployment rate may be endogenous to (perhaps magic progressive pixie dust). These comparisons are supposedly the closest we have to a truly “exogenous” experiment.

            The “state fixed effect” of these things may be *correlated* with the imposition of the minimum wage, but I don’t see why there would be multicollinearity. In any case, it wouldn’t overshoot the result; it would just make estimation imprecise.

            • Bob Murphy says:

              Ryan Murphy wrote:

              The “state fixed effect” of these things may be *correlated* with the imposition of the minimum wage, but I don’t see why there would be multicollinearity. In any case, it wouldn’t overshoot the result; it would just make estimation imprecise.

              Ryan, believe me, I get what you’re saying here. But I still think what I said is perfectly defensible.

              At first we run a regression, and we find that factor X2 has an coefficient of 2.5. Then we realize there might be an omitted variable, so we “correct” for this by adding in X3, and now the coefficient on X2 falls to 2.1. So we corrected for this missing thing, and it reduced the coefficient.

              Now let’s say we put something else in there that we shouldn’t have, thinking that we are further “correcting,” and it pushes down the coefficient to 0. But suppose in reality, the coefficient is indeed 2.1. I think it’s fine to say we “overcorrected,” meaning we pushed down the calculated influence below what it really is, thinking we were “correcting” it.

              If that’s not what “overcorrecting” means in this context, I don’t think the word could ever be used.

              If you are granting that the word choice is OK, but I simply have not described what is going on with the minimum wage studies, then I think you’re wrong, and maybe next week I can come up with an Excel example to illustrate my point.

              • Ryan Murphy says:

                But what I meant was that if the method did a good enough job of correcting without including the state trend line, it would no longer be correlated with the state trend line!

                If the endogeneity had already been corrected for, X3 would just be white noise thrown into the regression. It wouldn’t be statistically significant. It’s only if the endogeneity had NOT been corrected for would it be statistically significant.

                Concrete example. Suppose you adjust for a cross-country comparison using PPP. If you use that adjusted value of wealth and another cost-of-living adjustment on the RHS, the cost-of-living adjustment would only be statistically significant if the PPP adjustment wasn’t capturing everything.

          • Daniel Kuehn says:

            re: “Why are we looking at contiguous county pairs that straddle state borders, in the first place? What’s the point of that procedure?”

            Time-invariant trends could be taken care of by a pre-post analysis.

            Comparing across contiguous counties eliminates time-variant non-state-specific trends.

            Additional matching accounts for time-variant state specific trends.

            If this is screwy then I need to submit some errata to some editors and government agencies.

            • Ken B says:

              Questions.

              Are contiguous counties straddling the Georgia Florida border more alike, aside from state laws, than rural Georgia counties and Atlanta?

              If time trends stop at a state line, why? Law and state govt are all I can see. Since the effect of law and state policy are what we are trying to isolate?
              They have orchards on both sides of the line.

              Why should rising employment in eastern Tennessee matter more to truck stops in western Tennessee than the economy just across in Mississippi?

              • Daniel Kuehn says:

                re: “Are contiguous counties straddling the Georgia Florida border more alike, aside from state laws, than rural Georgia counties and Atlanta?”

                Probably not.

                re: “If time trends stop at a state line, why? Law and state govt are all I can see.”

                Also state-specific trends. i.e. – contiguous state counties. I’m next DC. I’m also next to a bunch of Virginia counties. D.C.’s economy is less similar to mine than the contiguous Virginia counties. But obviously the policies are the idea too.

                re: “Why should rising employment in eastern Tennessee matter more to truck stops in western Tennessee than the economy just across in Mississippi?”

                Nobody said it should matter more. I doubt it does.

              • Ken B says:

                The point id Daniel, you are suggesting “corrections” based on all these, to compensate for an effect already well controlled in the “experimental design” ie the contiguous bit. You need to show ahy the contiguity doesn’t control for stuff, and why your correction does. If the counties were compared in all random pairs your task would be easy. Here I think it’s very hard.

            • Bob Murphy says:

              DK wrote:

              If this is screwy then I need to submit some errata to some editors and government agencies.

              If you guys are allowed to throw out the textbooks, I’m allowed to challenge journal editors and government agencies.

      • Daniel Kuehn says:

        Matched DIDs are very common precisely because counterfactuals in DIDs aren’t perfect. I’m not sure I’m understanding the concern.

        I need to look more closely at exactly what they do but I don’t really think there’s a problem here.

        • Bob Murphy says:

          DK wrote:

          Matched DIDs are very common precisely because counterfactuals in DIDs aren’t perfect. I’m not sure I’m understanding the concern.

          See, this is why I’m persisting here Daniel. I think Ryan Murphy gets what my problem is, and is just disagreeing with me.

          It doesn’t sound like you even see the prima facie problem.

  6. Rob Rawlings says:

    “On the contrary, minimum wage hikes are implemented by “progressive” legislatures, who also (given my economic worldview) implement other laws that retard adult employment growth.”

    Wouldn’t these other laws likely retard both minimum-wage and other employment growth ? If so then it seems reasonable that other laws that reduce employment growth should be taken into account when looking at how MW-laws reduce employment growth in counties that differ not only in MW legislation but in other labor-laws as well.

    • Bob Murphy says:

      RR wrote:

      Wouldn’t these other laws likely retard both minimum-wage and other employment growth ?

      Yes, but if they had a disproportionate impact on higher-skilled employment versus low-skilled employment, then we’d see the pattern we actually observe.

      So now ask yourself: These other types of legislation, do they (on the margin) penalize adult or teenage employment more? I think adult. E.g. any kind of rule having to do with back pay, is costlier the higher the wage or salary.

  7. Major_Freedom says:

    This is an epistemological battle between a priorism and a posteriorism.

    Statistical analysis simply cannot be done without a priori assumptions. For complex statistical studies, the a priori assumptions tend to be deeply buried and difficult to discern.

    Dube, Lester, and Reich are a priori assuming that mininum wage laws have no adverse effect on employment, and then they pretend to be unbiasedly analyzing the data and concluding that minimum wage laws have no adverse effect on employment. The a priori assumption is buried in the “local economic conditions”.

    • Ken B says:

      No, it’s bad statistical reasoning.

      • Bob Murphy says:

        Ken B. wrote:

        No, it’s bad statistical reasoning.

        Yes, but do you think it’s a coincidence that Krugman and Daniel Kuehn see no problem with including the state employment trend variable, whereas MF and I think it’s problematic? I don’t think our econometric concerns are randomly distributed.

        • Joseph Fetz says:

          Confirmation bias?

        • Daniel Kuehn says:

          OK that’s complete bullshit Bob. Let’s stick to the actual arguments.

          • skylien says:

            You didn’t do it either:

            “People who won’t hear of any study that contradicts their econ 101 story are of course going to continue to disagree, but I don’t think that’s a very reasonable segment of the economics community.”

            • Daniel Kuehn says:

              Acknowledging it happens and accusing me of it are two different things. I don’t think Bob is doing this and I’m definitely not.

          • Major_Freedom says:

            Nobody said it’s being done intentionally.

          • Bob Murphy says:

            DK wrote:

            OK that’s complete bullshit Bob. Let’s stick to the actual arguments.

            !! This is like the Russ Roberts Moment of Unbridled Honesty again.

            Daniel, I am admitting that *I* am hot to trot on this point, because of the ramifications. I didn’t bother digging through the work of the guys who said the literature confirms the result I like, because I think they’re right.

            But since Dube et al. are the strongest paper saying something I think is nuts, I spent a while studying it, looking for holes.

            And, I submit, because you think their result is solid, you aren’t even “seeing” what my problem is with the county pairs thing, even though it’s perfectly obvious to Ken B.

        • Ken B says:

          Yeah but aren’t they doing it because the data doesn’t match what they think should happen? They have theory that says no effect, so …

      • Major_Freedom says:

        No, because In economics, “bad” is redundant.

    • Tel says:

      http://clubtroppo.com.au/2013/02/16/measurement-in-social-science-hard-but-worth-it/

      Not that I admit to seriously reading such non-Austrian blogs, but yeah, lots ot people are struggling with similar problems. The other alternative would be to admit there’s a lot we don’t know and try to just follow some common sense where possible.

  8. kelly warden says:

    so they have two comparable counties in two seperate states whose employment is basically static assuming they would be using statewide employment trends to confirm this static rate then raise the minimum wage in one state and when the employment rate declines they reintroduce those same statewide trends which now is necessarily in decline because of the introduction of the minimum wage increase attributing it to some other cause to compensate and minimize the results of the net loss of employment?

    • Bob Murphy says:

      Kelly Warden: Maybe, I’m not sure exactly what you’re saying. I think it’s more like this:

      If you look at counties that are touching, but happen to be separated by a state line, then when the minimum wage is raised in one of them (by their state legislature), you see (over the entire sample of such events) a definite negative effect on employment growth in unskilled labor in the county that has the min wage hike.

      However, this jump is not much different from the slower employment growth in that county’s entire state. For some reason, when a state raises its minimum wage, that tends to happen in a state with sluggish employment growth overall. So whatever is causing the overall sluggish growth is presumably affecting teenage employment too. Nobody thinks a minimum wage hike would make it harder for $40/hour auto workers to get a job, so there’s something else going on, which we’ve been unfairly pinning on minimum wage laws for decades.

  9. Martin says:

    I haven’t looked at the paper, and it is early, but if you’re comparing apples to apples where the only difference is due to the rise in the minimum wage, and then you find that when you decide to control for any differences between those two apples that the effect is not that different from zero, then that should not come as a surprise.

    If in one state people wear red shirts and in another they wear blue, and they’re otherwise the same, then if the minimum wage is raised in the former and you add a control for the color of the shirt in the latter, we’ll end up with a blue shirts theory of unemployment. Graduate students everywhere will now try to find the right shade of blue to reduce unemployment.

    Am I missing something or is this the gist of their argument?

    • Tel says:

      That sums up my understanding of the situation.

      Since no researchers (even the harshest critics of the minimum wage) believe that the minimum wage levels prevailing in the United States have had any impact on the overall level of employment,

      OK, so we know it can’t be minimum wage, must be something else.

    • Bob Murphy says:

      Martin I think it’s a little more sophisticated than that. If people always celebrated the raising of a minimum wage by wearing blue shirts, but then 6 years later they switched back to wearing red shirts, then I think you’ve got it.

      • Martin says:

        Thanks, Yes I think I see it now. 🙂

  10. Tel says:

    I will say one thing about the minimum wage debate. Let’s say employees are earning equal to their marginal productivity (which is the classical position). Then minimum wage goes up by say 10%. We expect that the business would fire pretty much everyone who was on minimum wage before, because after all they are all earning right at their marginal productivity, and with an increase as big as 10% it stands to reason they must all have just hit negative productivity.

    In other words, yes I agree that increasing the minimum wage does logically reduce employment, but not by sufficiently large amounts for the classical theory to be correct.

    • Ken B says:

      Tel, they won’t all be at thr threshold. Imagine a sloping line upward not a cliff, and redo your calculation. You will see a sub linear effect, which is what the neo classics predict.

      • Tel says:

        If they aren’t all at the threshold, then at least someone is getting paid less than they actually produce (or maybe more than they produce). Approx 1.3% of US workers are on the minimum wage, but just a tiny step lower than that and suddenly it goes to 0% (ignoring illegality).

        It isn’t a slope, it has to be a cliff because there is a legally enforced discontinuity.

  11. Bala says:

    Bob,

    I still don’t get why you are literally accepting the legitimacy of the empiricist approach to economics by even engaging those claiming that empirical data can validate or invalidate laws identified through reasoning from sound axioms. Isn’t the cause of truth served better by steadfastly rejecting the legitimacy of that approach? Just take a look at the (precious) man-hours that are being wasted in refuting what is fundamentally nonsense-on-stilts. It’s nothing short of criminal.

    Wouldn’t it be better to spend that time explaining why the empirical method of validating economic theory is utter nonsense?

    • Bala says:

      I am sorry. I should have said “implicitly granting some” instead of “literally accepting the” in my first sentence. Realised it the second I clicked on “submit”.

    • Bharat says:

      Because arguing on that basis is like speaking to a brick wall.

      Here’s an anecdote from Murphy that summarizes the mainstream reaction to that.

      Murphy also says (7:57):
      I’m glad I went the route I did because now when I’m dealing with mainstream people, I really do understand their models. I can literally prove Arrow’s Impossibility Theorem and that kind of stuff, so when mainstream guys throw out stuff and try to scare off Austrians with their formalism and all that kind of stuff, I can know exactly what they’re saying cause I did all that stuff.

      That’s not to say there’s no role for pointing out flaws in the methodology of empiricism when it comes to economics. But it’s certainly not Murphy’s role. Libertarians like to bring up the division of labor when asked what individuals should do to advance the cause of libertarianism. And what some have suggested, such as Tom Woods (if I remember correctly), is that the path for two individuals do not have to be exactly the same. Maybe some should pursue politics, other should do education, etc. The same is the case here. Murphy’s expertise as an Austrian isn’t just in his knowledge of Austrian economics, it’s in his knowledge of the intricacies of mainstream economics. He can argue against others on their own terms, and in many cases that makes for a more convincing argument.

      Sure, some economists will be convinced by methodological arguments, but certainly not most or all. They have spent decades and decades doing economic research with an empiricist framework. They are not going to change that mentality just because someone points out that in the social sciences it’s much more difficult to control for variables or because there’s a deductive methodology for economics (along with the other various reasons).

      • Bob Murphy says:

        Bharat, oh my gosh. Did I really say “stuff” four times in the same sentence?! I’m not sure I deserve $9/hour.

        • Major_Freedom says:

          Stuff happens on youtube.

        • Bharat says:

          Haha, I considered using ellipsis, but ultimately decided to quote you verbatim to preserve your spirit in the words!

        • Bala says:

          Your argument on your blog is quite well worded. Thanks.

          • Bala says:

            Oops… That was meant for Bharat 🙂

          • Bharat says:

            Thanks! Are you talking about the minimum wage post?

            • Bala says:

              Yes. I really liked that one.

      • Bala says:

        Bharat,

        Thanks for your response. However, when you say this

        “Because arguing on that basis is like speaking to a brick wall.”

        just bear in mind that everyone (Bob included) arguing with the mainstream economists on the basis of all their mathematical ex-post analysis is doing just that. In my mother tongue, one would say “sevudan kAthula sanga Uthikittu irukkInga” (Sorry, everyone else and Bharat too in case you do not get Tamil. That’s a very popular idiom in the language I speak and I find it the most appropriate. I means that you are blowing a conch right into a deaf man’s ear).

        Listen to Bob, especially at 37:00. The world is a more connected place today. You would gain much more by spending you time spreading education among people about Austrian Economic Theory than by precious-time wasting arguments with those who do not matter. When you can educate a larger population, how do academic economists matter? Because they hold the pulpits now and can preach to the hoi polloi from there?

        No matter what you argue with them, they are not going to change their ways. They have armed themselves with enough weapons to deflect any argument and waste precious lifetimes over useless discussions which will have no output. The State will be what it is. Governments will keep taxing and borrowing spending and terrorising. Banks will keep stealing. Nothing you argue with them is going to change any of that.

        But what if you could draw away more of their current ignorant support base that thinks they are dropping pearls of wisdom all over the place? Why not attract the public at large through spreading the Austrian Economic Theory? I mean it’s so easy and intuitive that almost anyone can understand it. Increase your support base among ordinary people. Let a billion people do your shooting down by laughing at their arguments rather than you trying to fell them using their weapons.

        • Bharat says:

          Don’t you see some value in small victories though? For example, you obviously don’t want the minimum wage to be raised right? We see the damage it would currently inflict on mainly teens but on some of the adult population as well. What Dr. Murphy is offering in this particular case is the illegitimacy of a study from its own methodological framework. And that certainly will not fall to deaf ears. A large segment of the population do listen to people like Krugman and use his opinions when they’re trying to get some legislation passed. But if Krugman is citing paper X, and we can say both 1) paper X uses a wrong-headed methodology for economics, and 2) even on its own terms, paper X is incorrect, shouldn’t we do both? Isn’t -anything- we can add to our arsenal good?

          It’s like when Thomas Woods argues in favor of nullification based on constitutional arguments. We can shout “Spooner Spooner” or “Rothbard Rothbard” all we like but it won’t make a real effect in our lives right now. But if we point out “no even by Constitutional standards nullification should be allowed” we have an addendum to our base of arguments.

          The reason I point this out is because Murphy and Woods aren’t always arguing on their opponent’s own terms. They do both. They spread Austrian and libertarian ideas AND argue against opponents on their own terms. I don’t think it’d be wise to commit all our resources to the former when the latter offers some benefit as well.

          I am Tamil too by the way! But born and raised in the US.

          • Bharat says:

            Whoops, don’t know if Krugman ever actually cited the Dube, Lester, and Reich (2010) paper that Murphy criticizes above, but he does use the John Schmitt paper (which itself refers to the DLR paper). Just want to clarify since I kind of used a false example.

          • Bala says:

            “Don’t you see some value in small victories though? ”

            No. I don’t because I don’t see them as victories. To me, they seem more like pissing contests.

            “For example, you obviously don’t want the minimum wage to be raised right?”

            I couldn’t care less. In fact, I would like to see them raise the minimum wage because it would hasten the move towards the crack-up boom.

            “What Dr. Murphy is offering in this particular case is the illegitimacy of a study from its own methodological framework.”

            Oh!! Don’t worry!!! They will soon find wriggle room. A fancy, incomprehensible mathematical argument would be thrown at you. Alterrnately, a host of new studies would be commissioned to explore a host of new angles so that by the time Bob sifts through them, he will be 83 (I’m assuming he is 43 now).

            “And that certainly will not fall to deaf ears.”

            Oh no!! It will. Academia is deaf. The hoi polloi knows and understand little and defers to the wisdom of the speakers standing at the pulpits.

            “but it won’t make a real effect in our lives right now.”

            I think it is naivete to expect any real effect in our lives. The only thing we can do is to bequeath our children or our grandchildren a better world to live in. Short-term goals are quite pointless. So I think it better to carry the torch of truth and liberty where it really matters – to the people. For the worst of tyrants still depends, for his legitimacy, on the implicit sanction of the victim.

            “They do both.”

            Except that they too have a limited lifespan and what they can achieve fighting the Agents is far less than what they can achieve by getting more people out of the power plant and increasing the population of Zion.

  12. Steven E Landsburg says:

    I have not read Dube, Lester and Reich but assuming that your account is accurate (and I have little doubt of this), your first point is devastating.

    • Nick Rowe says:

      I think I agree with Steven.

      Also, maybe what’s happening is that they are estimating the “political supply curve” of minimum wage laws. If a state government saw increasing labour demand and employment in the state, it might want to increase minimum wages, and if it saw falling labour demand and employment in the state, it might want to reduce minimum wages.

      • Major_Freedom says:

        Good point.

        A similar point is that we tend to find larger (smaller) absolute sized governments in larger (smaller) economies.

        Depending on one’s pre-existing convictions, this could signal a “causal” or a “despite” relationship.

  13. Bob Murphy says:

    Thanks Steve Landsburg and Nick Rowe. One point just to cover myself: Even though in their abstract Dube et al. say they are looking at county pairs, at points in the article it sounds like they also just look at state-level data, and do the regional correction. And for sure, in that Schmitt block quote in my post, he says they also drop the county-pair thing.

    So, even if it’s true that I’m right about that one part of their paper, it’s possible they can still get their result without that goofy procedure.

  14. kelly warden says:

    ok i finally read the paper complete and i see numerous flaws in their reasoning which they seem to concede by saying they assume the results will have no affect mainly in the tip credit section of the full service restaurants. in the fifties the minimum wage was not a big factor because it was so far below the going rate of employee pay anyway. you can apply the same reasoning to the tipped worker minumum wage which is far lower to the point that the wages arent the drawing factor to employment. also they state that they only compared states with tip credit which is true but not all of the states have 100% tip credit some are partial. they claime they used the two types (full service and fast food types) to gain a general assessment of the entire industry but if you look at the individual data it is clear the higher minimum wage recieved by the hourly non tipped worker and the employment thereof decreases while the tipped and lower waged employment increases.

  15. kelly warden says:

    they also say they do see increases in costs and prices of the restaurants which would surely apply to all other industries which employ the minumum wage workers such as manufacturing and maintenance which surely would drive up the overall costs of doing business and sinces we are considering border states and counties we would also assume that they will be competing in price assuming there is no tarriff or tax for the interstate commerce they will necessarilly have to keep their prices low and to achieve this some costs will have to be reduces either by reducing the amount of cheaper laborer in larger number or by decreasing the more senior and expensive labor in smaller number.
    my argument is simply the higher minimum wage decreases low skilled employment and the additional costs of that minimum wage across the region decreases the amount of higher skilled and higher cost employees.

  16. Chris P says:

    Bob,
    This paper by Neumark and Salas spends several pages discussing whether it is legitimate to correct for state level trends. It may be of some interest.

    http://epionline.org/studies/Neumark-01-2013.pdf

Leave a Reply to Ken B

Cancel Reply