Saturday, December 03, 2016

An econ theory, falsified

What does it mean to "falsify" a theory? Well, you can't falsify a theory globally - even if it's shown to be false under some conditions, it might hold true far away or in the future. And almost every theory is falsifiable to some degree of precision, since almost every theory is just an approximation of reality.

What falsification really means - or should mean, anyway - is that a theory is shown to not work as well as we'd like it to under a well-known set of conditions. So since people have different expectations for a theory - some demand that theories work with high degrees of quantitative precision, while others only want them to be loose qualitative guides - whether a theory has been falsified will often be a matter of opinion.

But there are some pretty clear-cut cases. One of them is the "Econ 101" theory of the labor market. This is a model we all know very well - it has one labor supply curve and one labor demand curve, one undifferentiated type of labor and one single wage.

OK, so what are some empirical things we know about labor markets? Here are two stylized facts that, while not completely uncontroversial, are pretty one-sided in the literature:

1. A surge of immigration does not have a big immediate negative impact on wages.

2. Modest minimum wage hikes do not have a big immediate negative impact on employment.

George Borjas disputes the first of these, but he's just wrong. A few economists (and MANY pundits dispute the second, but the consensus among academic economists is pretty solid.

The first fact alone does not falsify the Econ 101 theory of labor markets. It could be the case that short-run labor demand is simply very elastic. Here's a picture of how that would work:

Since labor demand is elastic, the supply shift from a bunch of immigrants showing up in the labor market doesn't have a big effect on wages in this picture.

BUT, this is impossible to reconcile with the second stylized fact. If labor demand is very elastic, minimum wage should have big noticeable negative effects on employment (represented by the amount of green between the blue and red lines on this graph):

By the same token, if you try to explain the second stylized fact by making both labor supply and demand very inelastic, then you contradict the first stylized fact. You just can't explain both of these facts at the same time with this theory. It cannot be done.

So the Econ 101 theory of labor supply and labor demand has been falsified. It's just not a useful theory for explaining labor markets in the short term (the long term might be a different story). It's not a good approximation. It doesn't give good qualitative intuition. And it's especially bad for explaining the market for low-wage labor, which is the market that most of the aforementioned studies concentrate on.

What is a better theory of the labor market? Maybe general equilibrium (which might say that immigration creates its own demand). Maybe a model with imperfect competition (which might say that minimum wage reduces monopsony power). Maybe search and matching theory (which might say that frictions make all short-term effects pretty small). Maybe a theory with very heterogeneous types of labor. Maybe something else.

But this theory, this simple Econ 101 short-run partial-equilibrium price theory of undifferentiated labor, has been falsified. If econ pundits, policy advisors, and other public-facing econ folks were scientifically minded, we'd stop using this model in our discussions of labor markets. We'd stop casually throwing out terms like "labor demand" without first thinking very carefully about how that concept should be applied. We'd stop using this framework to think about other policies, like overtime rules, that might affect the labor market.

Sadly, though, I bet that we will not. We will continue using this falsified theory to "organize our thoughts" - i.e., we'll keep treating it as if it were true. So we will continue to make highly questionable policy recommendations. The fact that this theory is such a simple, clear, well-understood tool - so good for "organizing our thinking", even if it doesn't match reality - will keep it in use long after its sell-by date. That's what James Kwak calls "economism", and I call "101ism". Whatever it's called, it's not very scientific.


In a follow-up post, I think about a couple of common responses to this post.


  1. I'm not an economist, so I'm asking this as a layman (i.e. please don't be cruel and nasty in your response because I'm not an academic)

    Isn't the Econ 101 theory of labor just a variant on the general law of supply and demand? So aren't you saying that these two facts contract the law of supply and demand, which means that supply and demand isn't true?

    And since I recall that you also believe that markets aren't efficient and actors aren't rational, aren't you basically saying that the entire foundation of academic economics is invalid?

    1. Isn't the Econ 101 theory of labor just a variant on the general law of supply and demand?

      A model can work well in one situation and not work well in another. For example, Econ 101 supply-and-demand theory (also called "price theory") might do a great job of describing the market for oranges, but a terrible job for describing the market for stocks and bonds.

      An analogy is physics. "Frictionless projectile motion" (the first theory most high school students learn) is a great model for cannonballs, and a terrible model for falling feathers.

      Every model should have a "domain of applicability".

    2. Anonymous12:13 PM

      "Even if we can assure ourselves that the employment effects are minimal, that doesn't make it a good policy

      well sure it does, its the main argument used by supply siders against minimum wage and living wage proponents

      what does make it a good policy is what is does to the lives of people who are struggling to make it on such low wages

    3. Nathanael7:19 PM

      Supply and demand equilibrium theory is *notoriously* bad at predicting prices in *any* market. (The equilibrium is the problem part here.)

      Unemployment and job vacancies are pretty much proof that labor markets are never in equilibrium, which should cause economists to study them more carefully, but 101ism still reigns.

  2. Anonymous7:10 AM

    Can't first fact be explained by immigrants bringing not only additional supply, but also new demand?

    1. Yep, that's General Equilibrium.

      I suspect that's a lot better of a theory of labor markets than the standard Econ 101 partial equilibrium approach that everyone knows and uses.

    2. Noah,I can't imagine anyone using this example in Econ101 model without also considering implications for demand shits. If one does that you can get reasonable results.So this seems a bit of a red herring.

    3. I think what you're really saying is that you prefer general equilibrium. See my post above!

  3. Does 101ism even claim that this theory is true in the short term? You might need a priest and imam to interpret 101.

  4. What market should we have in mind for your graph of stylized fact #1, where immigration does not shift demand for labor?

  5. Wait, your "evidence" about consensus amongst economists with regards to the minimum wage is a link to a lawyers blog who's pushing a book? How about this:

    More agree than disagree, and at best you can say many are uncertain.

    The econbrowser link uses a funnel graph that doesn't say ****. (aren't funnel graphs supposed to be used to detect publication bias anyway?)

    This is lazy and dishonest.

    1. "This is lazy and dishonest, mrrm hrrm hrrm"...let me buy you a stepladder so you can climb down off yourself, man. ;-)

      I linked to Kwak's post because it has a bunch of relevant links and I'm too lazy to paste them all in myself. Deal with it. Also Kwak is a good guy, and I'm happy to help him promote his book. :-)

      Funnel graphs are typically used to show publication bias toward one sign of a treatment effect, genius. If this graph were being used to show that, it would mean that studies showing minwage increases employment are out there but not getting published. Look at the plot. It shows that results are asymmetric around the peak in the direction of disemployment effects. That implies there are results in the other direction not getting published.

      Next time, spend a little more time thinking before you comment, eh? ;-)

    2. Notably, the IGM's asks about a $15 minimum wage by 2020. I can't speak for the OP, but when I read "modest minimum wage hike", I'm imagining the national minimum wage going from, say, $7.25 to $10.

    3. I'm imagining the nominal minimum wage going from, say, $7.25 to $30. That sounds pretty modest to me.

  6. Well, we could go on all day about the pitfalls of 101ism. I see it every day in how people want to do monetary policy. Partial equilibrium supply/demand is a simple tool that we can teach to someone with little technical expertise, which can help them think about the basics of economic processes. But for the questions in (1) and (2), it's not even a matter of the theory being "false" - it's just the wrong tool for the job. In fact, the tool is dictating the questions that are being asked - and they're the wrong questions.

    1) In this case, what we really want to determine are the effects of large-scale immigration. First, it's a big general equilibrium problem, not a partial equilibrium problem. The results are going to depend on the characteristics of the immigrants - age, skills, etc. In a democratic society, these people are not going to distribute themselves uniformly - they're going to concentrate in particular areas of the country. There will be adjustment costs due to strains on infrastructure and housing supply. These people will matter not only for the supply side of the labor market but for the demand side in markets for goods and services. Maybe they will come with skills that they will pass on to domestic workers. Lots of things going on here. In fact, wages may be the last thing we would want to think about.

    2) Even if we want to treat the minimum wage as a partial equilibrium problem (which it's not - this is macro policy), you can see that S/D doesn't get past first base. In competitive equilibrium, everyone behaves under the assumption that they can trade all they want at market-clearing prices. But if prices are constrained not to be market-clearing, the model can't tell us what happens. Further, we're again asking the wrong question. There's much more to this than what happens to employment. We're thinking about a policy that is purported to make low-skilled workers better off. But in principle it's going to affect everyone - it can lead to substitution of other inputs for low-skilled workers by firms, to firms going out of business, to changes in the wages of higher-skilled workers, etc. If the goal of the policy is as stated, maybe there are better policies we could think about. Even if we can assure ourselves that the employment effects are minimal, that doesn't make it a good policy.

    1. Agree.

      Did you like the first two paragraphs? I imagined you reading them while I was writing them... :-)

    2. Well, that's a little scary. You got it about right, though.

    3. Awww, thanks. :-)

  7. As I (and others) have argued before, Marshall's scissors do not seem like the best organizing framework for the labor market. The scissors assume anonymous spot markets. In contrast, most labor markets involve relationships.

    I talk a bit more about this here:

    It will just take time for the old tool to be replaced by the new. I already discuss the shortcomings of the scissor approach in my undergrad classes. But then, I've worked in the field of labor market theory. These materials eventually have to find their way to textbooks, but it will take time.

    A legitimate question one might ask is whether falsified models in econ are not rejected as rapidly as they are in other fields (like sociology, political science, physics, etc.). And if so, why.

    1. Already in at least one book:
      See Chapter 6.

    2. oldster1:12 PM

      "These materials eventually have to find their way to textbooks, but it will take time."

      I've been reading similar comments every since the global financial crisis. As a non-economist I wonder why change is so... inefficient! Is it how economics like no other discipline is linked to economical and political power in society? Mankiw type advisors, overlap between financial sector interests and researchers, and so on. We've had many books like Quiggin's Zombie Economics, Blyth's Austerity and now Kwak's Economism.

      Has there been attempts at more directed, sustainably funded campaigns? A site that list the 10 currently worst high profile textbooks and generate fair but effective critical pressure for change.

  8. Increases in the federal minimum wage are always followed by large increases in BLS estimates of the number earning less than the minimum wage. So employment need not fall, but average wages usually do. At the least, this suggests the minimum wage is not binding.

    1. I think you've got your arithmetic wrong, Alan. If minwage goes up, the percent of people earning less than minimum wage (because of exceptions in the law, or because of law violation) will tend to go up, but so will average wages, because the minimum wage is now higher than before.

      Evidence clearly shows that minimum wage hikes raise wages; that isn't controversial. For example:

    2. That increase in "estimates of the number earning less than the minimum wage" would appear to refer to the higher minimum wage, not the previous one. So your conclusion that average wages would fall is not supported by the graph. Also it's clear that there's a bump in the curve clearly corresponding to the imposition of a new minimum wage - it appears that there is some lag time for compliance to the new law.
      If you go past the bump, there appears to be a gain of 200,000 sub-minimum wage workers. I seriously doubt that this would be enough to imply that "average wages fell" when the minimum wage was raised from $5.25 per hour to $7.25 per hour. That's nearly a 40% increase in wages for a very large number of people.

    3. When I suggested growth in average wage stagnated after 3 of the last 4 minimum wage hikes that was just an observation, not a theory. The number paid the federal minimum is smaller than the number paid less, so the net effect is ambiguous.

  9. Anonymous1:35 PM

    An employer may respond to a legislated increase in hourly pay by reducing fringe benefits, holding workers to higher expectations, sacrificing workplace safety, investing less in training, etc. All of these things factor into the price of labor, not just hourly pay.

    This may sound like introducing Ptolemaic epicycles into the econ 101 model of the labor market, but regardless of its bearing on this issue, surely we can agree that the price of labor includes all of these things besides hourly pay. Moreover, if you take seriously the sticky wage view of the unemployment cycle, then you are committed to the idea that inflexibility in "wages" leads to mass layoffs, which is precisely what the elastic labor demand model suggests. On this view, by sticky wages we really mean sticky price of labor. If the minimum wage was a minimum price of labor (somehow), and we found that it had little impact on employment, I agree that this would be a puzzle, but I'm not sure we've found anything like that.

  10. People don't oppose increasing minimum wage because of econ 101. They deploy econ 101 because they oppose increasing the minimum wage, and the opposition to it does not have a prior justification that has anything whatsoever to do with economics.

    1. derrida derider6:34 PM

      Bingo. Noah's implicit assumption is that self-interest nowhere enters into the scientific process. But it is abundantly clear that political power very strongly influences the falsification process, even among the academics.
      As I get older, I'm gaining much more respect for postmodern critiques of science - especially the social sciences.

  11. My theory of the limited effect of the MW is the fact that the sort of places that pay MW (fast food restaurants, Walmart) are also the sort of places where MW people spend their money. So, increasing the MW just means they have more money to spend where they work, negating the additional cost of labor.

  12. Question: suppose a firm had a demand curve (or marginal gain curve) such as the following function, and applied "supply = demand":

    d(n) = $50 if n is less than 20 (n = number of employees)
    = $25 if n = 20
    = zero if n is greater than 20

    In other words, each of the first 19 employees bring in $50/hour, the 20th brings in $25/hour, and any more employees bring in zero (because their work can be distributed among the first 20).

    The obvious conclusion is that increasing the minimum wage from $7.50 to $15.00 won't change the employment for that firm. Is this different from Econ 101, or an application of Econ 101 supply = demand intersection?

    1. Your obvious conclusion is correct. Now change it to
      = $10 if n = 20
      Raise the minimum wage to $15 and employee #20 is now unemployed

    2. Yes. And if the 20th employee were worth $5, he wouldn't have been employed in the first place. In general, the change in employment due to the minimum wage change affects a small fraction of that last employee, the fraction within the change in the cost of the employee (possibly more than just the wage).

  13. I'm a bit surprised no-one mentions the obvious fact (and the reason that the theory is wrong) that demand for labour is not independent of either the number of people in work or the level of wages.

    The higher wages are and the more people in work, the more demand there is for goods and services and the more employment that is needed.

    Put that into your model and it works.

    1. Sorry, I read it in a hurry. I now see the passing reference to a general equilibrium model (with respect to immigration at least).

      But even this won't help if you include the assumption of full employment that appears to be standard. The assumption that demand is always high enough, except for short periods after shocks, is one that stops the necessary fiscal spending being done.

      Could anyone really empirically back up the idea that unemployment is either voluntary, or it is the fault of wages that are artificially high.

      For a start, you could look what happens when you cut wages to try to increase employment.See Greece for extreme example. It seems to partly work only in extremely open economies where demand from abroad and impoverishment of workers leads to small real growth, albeit in a nominally shrinking economy with a devalued currency.

      The whole idea is not only false, but it is a really damaging for policy.

    2. Anonymous4:35 PM

      You're so wrong you might as well be Egmont.

  14. An Econ 101 argument about price controls is that the market often find another way to get to equilibrium, and it is difficult for government to prevent these adjustments. If you place a ceiling on a loaf of bread the size, freshness, quality of ingredient, convenience of purchase are all still open to the force of competition.

    Likewise, a minimum wage can change the nature of the job even holding worker quality constant. For example, the employer could require that you be available to work at the employers discretion even if your total monthly hours stay the same.


    Also, even Kwak writes "detailed empirical analysis is inconclusive." Yes, that calls into question the simple employment will fall claim, but my assessment of the literature is there's not a consensus that the claim is falsified. I agree there's enough doubt that we shouldn't teach it as truth. It can be taught as a starting point for a discussion.

  15. How does this relate to Keynes's view of the labour market, namely that it is often out of equilibrium? Hence his seemingly convoluted definition of involuntary unemployment.

  16. Anonymous11:46 AM

    The graphs used by economists seem inspired by the desire to emulate physics, which when economics was developing in the 19th century was essentially dynamics, thermodynamics and equilibrium systems. Graphs like Charles' and Boyle's laws were the models. But these are systems approaches based on (as later recognised) the behaviour of vast ensembles of molecules whose sheer numbers ensure statistical behaviour that enables us to separate out systems properties like 'temperature' and 'entropy'(these become very problematical when only a few entities are involved). And also to separate out bulk properties from surface ones, like surface tension.

    Economic behaviour involves only relatively small numbers. In fact some markets although they may involve vast sums can depend on only a few actors. So other factors become important. I seem to recall that Mandelbrot pointed out some work that showed that the actual fluctuations in price in a particular market were significantly determined by the structure of the market (payment periods, for example).

    So, from a scientific point of view, I would say that all those demand and supply curves etc. are invalid: they are hypothetical descriptions of an imaginary world that bears no relation to our real one, but (even if implicitly) assumes that there are an infinite number of actors in any market. If we do find real, empirical smooth curves in economics then *those* are results that need to be explained with realistic information about how actors interact and the structure of the markets they are operating in. This seems a completely different mindset from 101ism.

    Supposed economic laws are 'ceteris paribus' but because of the small numbers there are no situations where in reality you can have an expression of a 'law' or deterministic curve where everything remains the same. This is why we can't say that minimum wages will necessarily cause unemployment, because (for example) we could argue that the lower paid account for a much disproportionate amount of the cash flow in an economy, even though they have only a small fraction of the wealth. Therefore it's an entirely plausible suggestion that the minimum wage will have a significant effect on the demand for various goods and push employment up (although it would likely cause changes as between different sectors). Whether it will or not - and I don't know the answer to this, obviously - is a time-bound technical problem, not an expression of a timeless 'law' or curve.

  17. Noah,

    Are you arguing that if you use S&D in the product market in conjunction with S&D in the labor market, then that amounts to a refutation of just using the latter? I've always thought of it as an obvious extension since policies in the labor market can have their effects be entirely in the product market. For example, we have good reason to believe that the costs of the minimum wage may just be born by consumers due to rising prices. But I don't see that as a refutation of 101ism.

    1. What you're really saying is that general equilibrium, not partial equilibrium, is what we should be using as our baseline model to think about labor markets. Which I agree with.

    2. If you're interested on the idea of price pass-through, then I recommend the following three sources.

      First, the works of Daniel Aaronson at Chicago who shows that minimum wages cause consistent small jumps in prices:

      Second, here is Michael Reich arguing that the San Jose minimum wage was primarily born by consumers:

      Third, my favorite paper in all of minimum wage studies is the following one, based on a massive minimum wage hike in Hungary. This paper is awesome and will knock your socks off:

      The competitive model lives!

  18. Anonymous6:12 PM

    Our host is clearly unaware of the existence of Opocher and Steedman's 2015 book on Full Industry Equilibrium. A coherent theory to be falsified has not existed for about half a century.

  19. Nathanael7:15 PM

    Imperfect competition has MASSES of evidence for it. We all know that businessmen operate in a bit of a cartel and avoid competing on wages.

    Oligopolistic competition is the standard theory for this, and it makes sense to start with that if you're trying to make a labor theory market.