Falsification in Economics

It doesn’t make any difference how beautiful the hypothesis (conclusion) is, how smart the author is, or what the author’s name is, if it disagrees with data or observations, it is wrong.

- Richard Feynmann

Our empirical criterion for a series of theories is that it should produce new facts. The idea of growth and the concept of empirical character are soldered into one.

- Imre Lakatos

A remarkable characteristic of economics is the sheer staying power of theories, even with a lack of empirical evidence to corroborate the propositions of these theories. In my experience, it is not uncommon for lecturers to remark that the lack of evidence for a theory has been a ‘problem’ for economists (though apparently not enough of a problem for them to throw out said theory). Often textbooks, lectures and discussions of theory make no reference to evidence whatsoever, and where they do it is trivial (for example, representative agent intertemporal macroeconomic theory predicts that governments will run periods of deficits followed by periods of surplus).

In the paragraphs that follow, I’ll examine a few cases of where I believe economics has gone off the mark in this respect. Specifically, I evaluate Marginal Productivity Theory, Walrasian Equilibrium, and The Solow Growth Model. I avoid theories such Real Business Cycle models and the Efficient Markets Hypothesis, partly because they have been done to death, but more importantly to demonstrate that the bad theories in economics are not merely the result of a few ‘wild cards’ at Chicago. On the contrary, I believe an anti-empirical approach is institutionalised within mainstream economics and that economics must undergo a paradigmatic shift to move away from these theories.

Marginal Productivity Theory (MPT)

The common interpretation of MPT is that it predicts workers will be paid ‘what they’re worth.’ In fact, this is not correct; the theory predicts that average productivity of workers will be positively related to wages, rather than each worker getting precisely their ‘just desserts.’ In any case, the result is that MPT predicts that compensation will increase as productivity increases. Hence, graphs such as this one – which you have likely seen before – pose a problem for MPT:

I have seen several responses to the problems presented by graphs like this. The first is that non-wage benefits have risen, which isn’t shown in this data. The second is that the adjustments for relative prices have been incorrectly applied, and consumers have more purchasing power than it first seems. However, estimates exist which take all of these things into account, and they still come to the same conclusions: most people’s overall real compensation is not increasing, even though their productivity is.

Another response would be that marginal productivity did well until the 70s, so maybe it remains useful. This is special pleading. A theory must be equipped to explain all phenomenon within its domain (in this case the labour market), rather than selectively applied where it suits the economist. If the laws of physics suddenly stopped working, can you imagine physicists making this defence? Saying ‘MPT will work except when it doesn’t and if it doesn’t we will throw our hands up in the air and carry on’ is not science. The fact is that such a sudden and clear decoupling of wages and productivity poses a clear problem for advocates of MPT, one which requires either a thorough explanation or discarding the theory altogether.

Walrasian Equilibrium

Walrasian equilibrium is one of the more absurd pieces of theory in economics (which is saying something). There are two (rational) agents with endowments of two factors of production, which they hire out to two profit-maximising producers. The producers use these factors of production to create two consumer goods, then the consumers purchase them. Everyone behaves as if they are perfectly competitive (they can’t influence prices) and everything happens simultaneously. There is no direct trade; instead individuals trade through the market (which comes from god outside the model).

The behaviour of consumers in this model is tautological. They consume based on a predetermined utility function that cannot be observed. Hence, they consume what they were always going to consume based on the chosen, non-empirical parameters of the model. This doesn’t tell us anything.

The behaviour of producers in this model is observable in the real world and hence not tautological. It is also not what happens in the real world. Some firms maximise profits, but most don’t; those firms that do maximise profits equate MC and MR is clearly false.

The only prediction this model as a whole makes is that the initial distribution of endowments will affect what is produced, how it is distributed, how much is produced and the price of what is produced. In other words: the initial resource distribution of a market economy affects its subsequent workings. This is trivial, and easily shown by theories that are based on more realistic assumptions (such as Sraffa).

The Solow Growth Model

The Solow model, to me, seems to be a textbook case of ‘bad science.’ This is clear from the story of its development (a story anyone who has taken development or macroeconomics will know).*

The Solow model predicts that, due to diminishing returns to capital, developing countries will catch up with developed countries in terms of GDP. At a low level of capital stock, the potential returns to investment are high (e.g. irrigating/ploughing a previously unkempt field). As the stock of capital increases, the returns to investment decrease and the growth rate of a country balances out. Hence, all countries will converge to a similar long term growth rate.

That this prediction is false is no longer debated. In the 1980s, William Baumol provided evidence that seemed to support the hypothesis. This was quickly disputed by Brad Delong, who noted that Baumol had used a sampling bias – he only included countries which were developed, effectively assuming his conclusion. Delong included more countries and found no evidence of convergence.

However, economists weren’t ready to give up. The prediction of the Solow model was reframed as conditional convergence: that is, provided countries have the right institutions, social cohesion, etc. they will converge in terms of growth. This, to me, seems trivial. The entire point of development economics is that the conditions in poor countries are not conducive for them to develop and so catch up with the developed countries. The Solow model doesn’t ask how a country might achieve this, but only says that it is a necessary condition for development, something development economists have always known. Hence, the Solow model is irrelevant for the immediate problem of development economics, which is how exactly we can help poverty-stricken countries get off the ground.

Is Economics That Bad?

In the interests of balance, it is worth noting some predictions made in economics that have been either empirically verified or dropped subsequent to falsification. Quantity of money targeting was tried, and failed, in a few countries, which led to Milton Friedman himself repudiating it (though economists still erroneously use the same framework which led to it). The lifetime consumption hypothesis (and non-utility based consumer theory in general) display good empirical corroboration and have all the hallmarks of a ‘good‘ scientific approach. The Phillips Curve as used by economists was modified in light of evidence in the 1970s. Both the multiplier and the Giffen Good are good examples of non-trivial, clear, falsifiable predictions, though I will not comment on evidence for them because that would take a post for each one.

Nevertheless, the record as a whole is not good. Theories from over a century ago look, and are taught, the same way as they were when they were initially adopted. New ideas that are not even disputed by economists, such as behavioural economics, are slow to be adopted, and when they are adopted are presented as a ‘special case’ and in a way amenable to the core framework, which is, of course, still taught alongside them. As far as I’m aware, there is no clear cut case of a neoclassical theory being completely thrown out and never mentioned again. This alone should be an indicator that the scientific method is not at work in economics.

*I am ignoring the internal problems with models of this ilk, and the fact that Solow himself seemed to agree with many of these criticisms.

About these ads

, , , , ,

  1. #1 by Cameron Murray (@Rumplestatskin) on March 15, 2013 - 10:50 pm

    The thing I find most interesting is that the economics discipline seems to ignore the important questions that the general public expects them to be studying – where do profits come from, what is growth and where does it come from, where do prices come from, etc. My general feeling is that the profession seems to only ‘speak’ with one mathematical model (optimal control) and the theory is squeezed into that particular type of maths. This mutual reinforcement makes it hard to cut through because you need to adopt new mathematical tools with new theories THEN EXPLAIN BOTH THE THEORY AND THE MATHEMATICAL MODELLING FOUNDATIONS TO THE PROFESSION!

    What I have learnt is that the audience for economists is… other economists.

    There are encouraging signs – agent based models, network models, etc, and lots of empirical work revealing some fairly robust relationships in the data, even if they usually don’t support the fundamental theories. While I think they are encouraging, they don’t speak to the core of the profession, since they are speaking a different language, thus I worry that they will always be regarded as fringe ideas.

    I also put part of the blame in textbook writers who seem to go out of their way to define economics as the study of marginalism and equilibrium, even if they are applying theories improperly and incompletely anyway. For example, in general equilibrium there is a set of prices that clears the market. Once one market is out of equilibrium, all other markets can be anywhere and the impacts of these markets can make the dynamics of that one market very complex. Yet the textbooks brush it aside an show how markets handle “shocks”.

    Rant over.

    • #2 by Unlearningecon on March 16, 2013 - 11:52 am

      Indeed, I went to an event recently, set up by students, where many agreed that economics had not been what they expected. The relevance of, for example, a utility-maximising individual choosing between two goods, has not yet struck me.

      As for equilibrium: typically, economic theories simply assume equilibrium. How we get there is an open question, one which textbooks will note poses a problem, then carry on as usual. Rod Hill and Tony Myatt call this the “note and forget” approach, and it is found throughout economic models.

    • #3 by Luis Enrique on March 19, 2013 - 12:05 pm

      ” economics discipline seems to ignore the important questions … where do profits come from, what is growth and where does it come from, where do prices come from, etc. ”

      umm, what? economics ignores the question of where prices come from? don’t be bloody silly.

  2. #4 by Josie on March 15, 2013 - 11:57 pm

    “The common interpretation of MPT is that it predicts workers will be paid ‘what they’re worth.’ In fact, this is not correct; the theory predicts that average productivity of workers will be positively related to wages, rather than each worker getting precisely their ‘just desserts.’”

    Where does this come from? All the stuff I’ve read DOES say that workers will be paid ‘what they’re worth’. Surely that is necessary for the neoclassical theory of incentives, the “perfect” market maximising utility etc etc.

    Otherwise I don’t see how the theory says anything very much at all. It tells you nothing about income distribution, and the average of how many workers? You could average over a team, a firm, a country, the world…?

    Nice post as usual.

    • #5 by Unlearningecon on March 16, 2013 - 12:46 am

      In unguarded moments, economists will also assert that the theory means workers are paid what they are worth.

      In fact, the demand schedule is determined by all the different productivities within the labour market. The resultant equilibrium is at one, somewhat arbitrary point, which would be an average of the points around it. Each worker will be paid this wage only, so it doesn’t make sense to say that they are all paid their marginal productivity.

  3. #6 by Hedlund on March 16, 2013 - 12:24 am

    Is this post new? It doesn’t appear at the top of the main page; in fact, it appears to be dated February 28.

    • #7 by Unlearningecon on March 16, 2013 - 12:38 am

      Thanks a lot for noticing that – strange! I have changed it now.

  4. #8 by Peter on March 16, 2013 - 7:37 am

    Academic economists certainly tend to prefer theory to reality, as noted by Oxford economics Prof. Simon Wren-Lewis here:

    http://mainlymacro.blogspot.co.uk/2012/04/microfoundations-and-evidence-2.html

    However you face a problem in attacking MPT, namely what better theory do you have? And if you can’t produce anything better, then MPT wins, for all its defects.

    MPT is actually based on a very simple and common sense idea namely that employer in deciding whether to hire someone will base the decision simply on whether the employer thinks a profit will be made as a result. I.e. if an employer thinks the total extra costs that result from employing someone will be less that total extra sales that result from the extra employee, the employer will hire.

    Doubtless many employers don’t do detailed scientific MRP calculations before deciding to hire – their decision is probably based more often on hunch than on detailed calculation. But profit (or at least avoiding loss) is still the rule they follow.

    Moreover they are FORCED to follow the latter rule: if they ignore the bottom line, they go bust. It’s all brutally simple. In short, strikes me that MPT is an inevitable and unavoidable part of market economies.

    As to the divergence between productivity and hourly compensation shown in the above chart, that does not seriously dent MRP theory. You’d expect productivity and hourly compensation to be directly related ALL ELSE EQUAL. But it’s perfectly possible for “all else” to change. For example there has been an obvious tendency over the last thirty years or so for senior executives and entrepreneurs to screw every penny the can out of their businesses: executive pay relative to average pay has rocketed over that time.

    Plus the levels of fraud and criminality in the banking industry have clearly risen since the 1950/60s. I.e. the culture has changed.

    • #9 by Unlearningecon on March 16, 2013 - 10:44 am

      Well, if a theory is wrong then in principle it doesn’t matter whether or not there is an alternative, it’s still wrong and not useful. Of course in reality I agree that often theories are only abandoned once alternative theories supercede them.

      In the neoclassical world of perfect competition what you say might be true. However, in the real world businesses tend not to adjust each input according to its productivity, which is hard or even impossible to measure individually (see also Arnold Kling). Instead, they will make decisions in terms of projects, new branches etc, which involve expanding all inputs, and then calculate total profitability (actually, more likely returns than profit) based on this.

      I agree that ‘rent-seeking’ makes sense as a response. But I don’t see how that preserves marginal productivity theory, which doesn’t mention rent seeking. We’d need to modify it somehow. Personally, I’d just go for the idea that income distribution is in large part determined by political power – with the decline of unions and the dismantling of BW, capital had a lot more power and wages have been suppressed ever since.

    • #10 by gastro george on March 16, 2013 - 2:20 pm

      “MPT is actually based on a very simple and common sense idea …”

      Isn’t this one of the main problems with economics? That ideas may seem to be common sense but are, in fact, not true. Economists need to be cleverer than “simple common sense”. But often aren’t, and don’t learn.

      • #11 by Dan on March 17, 2013 - 5:03 pm

        I kind of feel like economists are already clever enough and that is part of the problem. Physicists and astronomers aren’t clever, they are observant. But economists are continually finding some reason or another to explain away why their theories are bullshit. Other sciences don’t have that liberty and are confined to studying reality rather and making theories around that. Economists on the other hand make up theories and then try to come up with excuses why reality doesn’t reflect it.

    • #12 by josie100 on March 16, 2013 - 5:20 pm

      Sorry to bang on about it but this is how I understood MPT as well and it does suggest that it is individuals who are (supposedly) paid in accordance with their “contribution”, together with the other side of it – that if a firm was paying under someone’s “value”, competition would (supposedly) ensure that that individual got a better offer elsewhere. I don’t understand what you (unlearning) are saying about averages.

      All of it, of course, assumes that the rate of profit is zero, which makes you kinda wonder why anyone would bother starting a company.

    • #13 by Olle J. on March 26, 2013 - 9:20 am

      “Moreover they are FORCED to follow the latter rule: if they ignore the bottom line, they go bust. It’s all brutally simple. In short, strikes me that MPT is an inevitable and unavoidable part of market economies.”
      Or, in other words, the as-if argument. It might sound feasible to argue for this approach with regard or MPT. However, in several cases this is not really of any importance. A lot of the implicit or explicit assumptions are rarely met on a labour market and the problems of search- and information costs are often dire. My experience, having studied firm strategies and behaviour on the labour market, is that there’s usually plenty of room for divergence from paying wages in line with MP, adoption takes years or even decades.

      • #14 by Unlearningecon on March 26, 2013 - 11:44 am

        In many cases I don’t think it’s possible for firms to know the individual productivity of a worker, or even for it to exist alone. See, of all people, Arnold Kling with the best exposition of this idea I have seen.

      • #15 by Karl on April 3, 2013 - 3:09 pm

        It seems that if there’s a broadway play, every actor’s marginal productivity is 100% of the play’s productivity. (At least among the actors without whom there will be no show.)

        …unless one of those actors has an understudy, in which case, the actor’s marginal productivity is only the difference between him and the understudy. If the understudy is better than the actor who got the part — it happens — then the actor’s marginal productivity is negative! In this case, the actor will receive negative wages.

        ====

        The assumption that a worker can be taken from one job and put in another and have identical productivity is just plain stupid. We all know that’s not true: people learn specialized knowledge for their specific roles. Ten years experience on one company’s product is not identical to ten years experience doing something else.

        And yet that blatantly stupid assumption seems to be the entire basis to consider marginal productivity not just an upper bound, but also a lower bound on compensation.

        Yeesh.

  5. #16 by QP on March 16, 2013 - 10:46 am

    From the MPT plot It looks like something changed ~1971 – end of dollar link to gold?
    Perhaps in practical terms the expansion of the finance sector and non-wage earnings (capital gains, dividends, interest) wrt labour? The trigger to increasing inequality since the poor are typically reliant on wage earnings whereas the rich take the former?

    • #17 by Unlearningecon on March 16, 2013 - 11:48 am

      I think it was the dismantling of BW which allowed corporations to outsource jobs more effectively.

    • #18 by Dan on March 16, 2013 - 8:40 pm

      I was wondering that too, the early to mid 1970s did see a shift in monetary policy with the death of Bretton Woods. However my first thought was that it could signify the growing use of computers in the workplace. From the mid 1970s onward they began playing a larger role in the workplace, and undoubtedly increased productivity. Of course the 70s also saw an increase in unemployment in the western world from the decline in heavy industry and manufacturing which would also tend to depress wages due to the “reserve army” of unemployed. Though while the increase in unemployment could explain stagnating wages, it doesn’t seem to explain increase in productivity. My guess is that even though computers increased productivity, they didn’t increase overall workload, its not like we work harder than we did in the 60s, we can just get more done in the same amount of time, so you could make the argument that the “value” of the worker never really did increase at all. In fact if anything it could have decreased, kind of like how musket men may be worth less than knights on an individual level, even though they are overall far more effective on the battle field. But anyways, I think that the rise of computers and the efficiency they bring to the workplace could explain why productivity continued to rise despite wages leveling out.

    • #19 by Draco T Bastard (@DracoTBastard) on March 17, 2013 - 7:23 am

      First oil shock and then pay resumes its climb at the historic rate and then the second oil shock hits. After that worker pay stays the same and the income of the top 1% begins its exponential climb.

      It was during/after the second oil shock that governments around the world went hard on the free-market dogma. Reagan in the US ’81 – ’89, Thatcher in the UK over a similar time period etc.

      You’ll note that the 1% had wanted to go back to such a system for quite some time but conditions hadn’t allowed it. The problems that Keynesian economics began to have in the 1960s and into the 1970s coupled with the oil shocks allowed the partial dismantling of the welfare state(s) in most developed countries which tilted the political power over to employers and away from employees.

      • #20 by Unlearningecon on March 18, 2013 - 10:18 am

        Precisely. I find this kind of explanation is unfortunately, anathema to economists, who hold ideas like the NAIRU and lump of labour ‘fallacy’ at the top of their heads and are hence unable to see why the actions of the Free Market could do anything but create more jobs and higher wages, at least in the long term.

  6. #21 by Dan on March 16, 2013 - 4:56 pm

    “Walrasian equilibrium is one of the more absurd pieces of theory in economics (which is saying something). ”

    While I do agree that equilibrium economics is stupid, I am not one to completely dismiss the underlying ideas behind them. While static or even dynamic equilibrium doesn’t seem to be true, it doesn’t make sense to say that all economic actors are acting randomly and that markets don’t have any general direction. Clearly if you look at history there is something that is “pulling” markets somewhere. Almost in the way that natural selection works, genetic mutations are random, but they only stick if they provide some kind of usefulness to the species, and the final product is supposed to be a species which is an adequate fit for the environmental challenges it faces. In other words, biological evolution seeks an equilibrium as well, but since it is always changing, it never does reach it, new challenges emerge and it must keep adapting.

    Human economies are the same way, economic strategies must be viable to survive, and eventually successful ones become obsolete. The fact that mixed market welfare state capitalism has become the standard all over the world tells me that it is successful in responding to the natural pressures facing our species. In other words, even though Walrasian equilibrium is not real, in some ways economies behave as if it is real. Natural selection pushes our species towards an equilibrium, but the changing pressures of natural selection continually change what that equilibrium should be. So the Walrasians may be full of shit, but they are not completely full of shit.

    • #22 by gastro george on March 16, 2013 - 7:31 pm

      “… biological evolution seeks an equilibrium as well …”

      I need to re-read my chaos theory, but wouldn’t “islands of metastability” be a better description?

      • #23 by Dan on March 16, 2013 - 8:49 pm

        Islands of metastability probably would be a better description when it comes to biological evolution. Although at first glance I thought you had said homeostasis, which I think would also be an interesting term. Biologists tend to talk about homeostasis for an organism, a state in which an organism is able to regulate its internal processes adequately (like the human body regulating temperature). It is certainly different from equilibrium which sounds more like balance, while as homeostasis tends to keep things within measure so its total stability remains. There are certainly times when economies look to be in a homeostatic state, however just like the human body, homeostasis is never a permanent thing.

  7. #24 by gappy on March 17, 2013 - 2:31 am

    Can you provide an academic reference to the statement “average productivity of workers will be positively related to wages”? I assume by “wages” you mean “average wages. I don’t see how it follows from marginal disaggregated relationships.

    • #25 by Unlearningecon on March 18, 2013 - 10:25 am

      See here:

      The marginal productivity theory caused something of a little tornado around the turn-of-the-century, which deserve some attention…

      The first and most straightforward error (which is sometimes repeated today) is to assume that the marginal productivity theory says that factor prices are determined by marginal products… It has never said that, regardless of whatever [J.B. Clark] let himself say in unguarded moments. Factor prices and factor quantities are determined by the demand and supply of factors, period.

      Or Marshall:

      The doctrine that the earnings of a worker tend to be equal to the net product of his work, has by itself no real meaning…

      The supply and demand schedules interact to produce an equilibrium wage in a labour market. Though this schedule is related to the observed productivity in the market and an increase in productivity will shift the demand curve and hence increase the wage, it is not true that a worker’s product is equal to the wage, because:

      (A) You also need supply.

      (B) The wage will be equal to the marginal product of the last worker employed, but the same for all workers.

  8. #26 by commenter on March 17, 2013 - 7:48 pm

    “I believe an anti-empirical approach is institutionalised within mainstream economics”

    Sorry, but your belief is wrong. We can test your belief empirically. I selected the most recent volume (Feb 2013) of the AER (the top ranked economics journal), and read through the titles and abstracts of all the papers that were not comments or replies. It was readily apparent that 12 of the 19 papers were empirical, with another four that looked like they interacted with empirical facts but were mostly theory based. Only three papers appeared to be only theoretical.

    • #27 by Unlearningecon on March 17, 2013 - 9:27 pm

      Our comments aren’t mutually exclusive. In many ways there is a disconnect within economics. One side (econometrics, loosely speaking) is empirically based. The other side – what is taught to students, DSGE etc. – is quite a rigid theoretical framework that is rarely put to falsification, or not altered despite falsification.

      • #28 by commenter on March 18, 2013 - 2:09 am

        The disconnect that you have identified is mostly temporal – economics was relatively athereotical 30 years ago, but that is not the case any more. The problem is that what is taught to undergrads is too far disconnected from modern research.

      • #29 by Unlearningecon on March 18, 2013 - 10:12 am

        OK well at least we can agree on that.

        But IMO DSGE has similar problems. Most (all?) DSGE papers don’t reference evidence but merely attempt to explain well known phenomena, one at time. There is no real attempt to see if the predictions of a DSGE model are consistent with all available evidence; they only have to be loosely consistent with one part of the data.

      • #30 by commenter on March 20, 2013 - 4:59 am

        I’m not a macro guy, so I’m certainly not up with all the finer points of DSGE. From the few DSGE papers that I have read they tend to try to match a broad range of first and second order moments of the data – certainly not one at a time.

        But I certainly don’t envy macro economists – they have ONE independent data point to work with. Everything within their time series is endogenous, and they have to calibrate their model with the same data series that they are trying to predict. Not to mention that there are near impossible to predict breaks in the time trends.

        It’s a tremendously difficult problem that they face, and people who try to hand wave over the difficulties and propose magical solutions (Keen, for example) are not helping. It’s not like macroeconomics isn’t making any progress (try comparing a volume of Journal of Monetary Economics from today with one from 30 years ago), but its just damn hard.

      • #31 by Unlearningecon on March 20, 2013 - 10:22 am

        Okay well I’m not trying some sort of Galileo Gambit here, but don’t you think that, compared to Ptolemy, Newton’s Laws seemed ‘awfully simple?’

        It just strikes me that economists are incredibly hamstrung by their utility maximising individuals, market clearing etc. and resultant aggregation problems. All Keen really does is drop these restrictions, add endogenous money and uses classes as agents, and he has a business cycle that seems to corroborate well with the data. I don’t see why his model has so much bile spit at it.

      • #32 by commenter on March 20, 2013 - 5:03 am

        Also, when it comes to teaching students outdated material, it doesn’t help that our students are woefully underprepared when it comes to mathematics. I have had to TA second year game theory courses where we couldn’t use calculus because not all of the students had taken a full year calc sequence. Until we have more rigorous pre-reqs for taking Econ courses then we have no choice but to accommodate the lowest common denominator.

      • #33 by commenter on March 20, 2013 - 5:31 pm

        It’s not Keen’s model that is the problem (at least for me, can’t speak for everyone) it is the ridiculous claims that go along with it. His model is not novel, and passing it off as an amazing innovation is intellectual fraud. Heck, passing it off as any sort of innovation would be intellectual fraud. There are other models that do very similar things, and match the business cycle just as well. It’s definitely not my area, but I think that the Journal of Economic Dynamics and Control would be the place to start looking for this stuff.

    • #34 by moiracathleen (@moiracathleen) on March 17, 2013 - 9:42 pm

      I think you mischaracterize the assertion that economics is “anti-emprical.” It would be a leap to construe the statement that economics is anti-empirical as meaning that no economics papers contain empirical evidence.

      I think UL’s claim economics is anti-empirical points to problems with the empirical methodology employed more so than anything else. He addressed this issue in another post titled “Institutions and Economics.”

      http://unlearningeconomics.wordpress.com/2013/01/28/institutions-and-economics/

      See the Section: “Problems with this Methodology” and the preceding paragraph. The problem with empirical analysis in economics is well accepted by most people in the field. See Economics vs. Physics post: http://unlearningeconomics.wordpress.com/2013/02/02/economists-versus-physics/ Economics is subjective valuation. Empirical evidence used to back various theories and contentions in 16 out of 19 papers will not change this fact.

  9. #35 by Luis Enrique on March 18, 2013 - 2:28 pm

    the point of the Solow growth model is to demonstrate that you cannot account for observed differences in income across countries by appealing to differences in savings rates. I can’t remember the exact maths of the top of my head, but the point is that if you change s by some amount, call it ds, then the steady-state level of income will change by something like ds^(a/(1-a)) where a is output elasticity of capital. Hence, for all permissible a and s, the observed differences in income across countries are far to large, and you need to appeal for variations in total factor productivity or the Solow residual, or as he put it, “the measure of our ignorance”. This is the point. That’s how this model should be taught.

    the Solow model is not and never has been a model to provide insights into the “immediate problems of development economics” – it performs the service of drawing attention to what we don’t know. So now economists search for theories of TFP, institutions, factor misallocation etc. etc. and the whole shebang of development and growth economics.

    I think it’s just wrong to claim the model was “reframed” as a conditional convergence model. These models (the Solow model is not the only one with the characteristic of predicting convergence to a balanced growth path) only predict convergence to an absolute level of income if you assume all countries share the same structural characteristics, including the same level of TFP. In other words, these models have always been models of conditional convergence, the level of income (or the level of the balanced growth path) to which economies converge is conditional on that country’s characteristics.

    • #36 by Unlearningecon on March 18, 2013 - 7:00 pm

      Honestly, I’ve never seen the Solow model characterised as saying that before, but I’ll accept that it is a falsifiable, non-trivial prediction. However, I also think it is falsified by Mankiw et. al, who found about 60% of the difference is explainable by savings and population. Their idea of what the Solow model predicts also seems to be completely opposed to yours. I’m confused about how you got that prediction?

      Perhaps it’s fair to say the model always implicitly had conditionality in it. However, the early empirical attempts did not take this into account, which is what I meant by ‘reframed.’

  10. #37 by Luis Enrique on March 18, 2013 - 3:11 pm

    you want economics to have abandoned the idea that worker’s compensation is somehow tied to their productivity, whether as individuals or in aggregate. If economics did abandon the marginal productivity theory of income, you would congratulate economists for being good scientists who change their theories when the data tells them they need to.

    so …. have you checked out any labour economics over the last couple of decades? Ubiquitous approach is some variation on Mortensen Pissarides model, been around since 1990, wages determined by bargaining over surplus of production, determination of bargaining power between worker and capitalist left an open question.

    • #38 by Unlearningecon on March 18, 2013 - 6:57 pm

      That’s interesting, could you point me to some particulars?

      • #39 by Luis Enrique on March 19, 2013 - 10:35 am

        I have to confess, I don’t know this topic so well. Afaik, the main focus of these models is to explain unemployment dynamics, so I’m not sure how well developed they are as models of worker compensation. For example in the original M&P paper they just write ” For lack of better information, equal bargaining power was assumed by setting [the bargaining parameter]=0.5″ so workers and firms just split the surplus.

        Google scholar locates papers like this

        http://www.jstor.org/stable/3598795

        which I’m not going to read.

        I guess you could explain the disconnect between productivity and wages by saying worker bargaining power collapsed for various reasons. I cannot immediately point to a model/papers that address that, this isn’t my field.

      • #40 by Luis Enrique on March 19, 2013 - 11:05 am

        fwiw, in conversation perfectly mainstream economists have suggested to me that to explain recent changes in income distributions, you almost need a separate model for what’s going on with the incomes of the 1% from everybody else – the modern day aristocracy – superstar economics, rent extraction, it’s all about having very high bargaining power and being sat on top of a big pie you can help yourself to a large slice of. 99% of us aren’t in that sort of position.

        more generally, remember that one reason why the median worker might see stagnant incomes whilst economy-wide productivity has increased could be that the individuals responsible for the productivity gain have been rewarded by higher wages, but these individuals are not the median worker, but a small group of highly skilled. Now I don’t imagine for a minute that the distribution income across workers is simply a mirror of the distribution of productivity across workers, but that’s not to say things like skills-biased technological change and globalisation play no role. So maybe it is possible to reconcile what you call MPT with what we’ve seen happen to the wages of the average worker. See here:

        http://www.voxeu.org/article/new-evidence-international-trade-offshoring-and-us-wages

        http://economics.mit.edu/files/8512

        of course if we’re talking about inequality more generally, we need to think about capital incomes.

      • #41 by Unlearningecon on March 20, 2013 - 1:05 pm

        I agree that recent progress in labour market economics veers so far away from the perfectly competitive ‘ideal’ that it could be said to be completely separate from it.

        However – and I admit I have only skimmed the papers as I have mid terms – it strikes me that economists are still hamstrung by their approach. The idea is that, in absence of ‘rents’ ‘frictions’ etc, the labour market would function as in the econ101 textbooks. To paraphrase Noah Smith, the whole idea that you start with an ‘ideal,’ then to get anywhere close to the real world have to introduce so many ‘frictions’ that they overwhelm the supposed ‘underlying’ mechanics, doesn’t bode well for the basic model.

        Obviously some economists, such as the ones you mention, agree with the potential need for a new model.

      • #42 by Luis Enrique on March 19, 2013 - 2:55 pm

        here’s what mainstream economics has to say about imperfect competition in the labour market. This is (hopefully) an ungated version of a chapter from the mainstream-defining Handbook of Labor Economics

        http://eprints.lse.ac.uk/28729/1/dp0981.pdf

        at what point do you say that economics has abandoned what you call the MPT theory of wage determination?

  11. #43 by Luis Enrique on March 18, 2013 - 8:23 pm

    Here is Solow, quoted from his piece “reflections on growth theory”

    “In the beginning, one of the surprising implications of the neoclassical growth model was not merely that the steady-state growth rate was independent of the saving-investment rate, but perhaps even more that the (moving) equilibrium level of output per person apparently responds very weakly to changes in the saving-investment quota. To be more precise, the elasticity of output per worker with respect to the saving-investment rate is the ratio of the capital share to the labor share (in steady-state equilibrium). Back then the conventional numbers were 1/4 and 3/4, giving an elasticity of 1/3. The message appeared to be that as big an increase in the rate of investment as policy could manage would yield only a disappointing in- crease in productivy”

    If you write down the expression for steady state income, differentiate w.r.t. savings rate, that’s what the Solow model predicts about how much changes in s can buy you.

    It is possible that in an empirical application, the savings rate could explain more of the variation in the data than the theoretical model says it ought to. Read the para in MRW starting ”Yet all is not right for the Solow model… ” It gets the magnitudes wrong, too large. So they augment the model with human capital.

    • #44 by Unlearningecon on March 19, 2013 - 9:24 am

      I still find it odd that they seem to have an entirely different idea of what the model means to Solow and yourself. Not saying that you’re the one who’s wrong…

      In any case, even after augmentation, the predictions for the significance of s are still significant enough to qualify as evidence against the model, surely?

      • #45 by Luis Enrique on March 19, 2013 - 10:21 am

        I don’t follow. I mean, I think I know the MRW paper pretty well, and I don’t recall any discrepancy between how they regard the Solow model and what I describe above (which isn’t really disputable, it’s just what the model says and the interpretation I give it’s in the text books – i.e. Romer). Why do you say they have an entirely different idea?

        MRW show that when estimating the basic model, in cross section, savings and population growth explain more of the variation in the data than the Solow model says they ought to – i.e. they are saying the same thing as I described above and the Solow quote does too – in the model, you cannot explain observed income differences by appealing to variation in savings rates, but MRW show that apparently empirically you can, thus “all is not right for the Solow model”. They then show that if you add human capital, in the thus augmented Solow model, variations in savings rates have a more powerful effect on income, consistent with that estimated in the data.

        In the second half of the paper, MRW estimate a conditional convergence equation. Crucially, they use a fixed-effects model combined with the assumption of a global rate of technological progress, so the country-specific fixed effect plays the role a country-specific initial level of TFP.

        More generally, I think it’s a bit odd that you cite the Solow model as an example of a model that economists ought to have abandoned because it has been “falsified”. In one sense, they have abandoned it, long ago. Asked to explain the development process, no economist is going to tell you the Solow model is all you need to know. The Solow model typically appears in chapter 1 of a growth textbook, and the rest of the book can be seen as step by step trying to make up for all the shortcomings of that model. For example, look at the contents page of the canonical growth text book by Acemoglu:

        http://press.princeton.edu/TOCs/c8764.html

        However the Solow model remains an excellent introduction to growth. It’s probably the first time students will have encountered transition dynamics, resource constraints, using log differentiation to manipulate growth rates, steady-state growth, it raises all sorts of questions for discussion, it can even be used to talk about poverty traps. It makes perfect sense to me that the Solow model is still taught.

      • #46 by Unlearningecon on March 20, 2013 - 12:59 pm

        I don’t think it has been falsified – I initially said the prediction of conditional convergence is trivial. I tend to agree with Solow:

        I have never thought of the macroeconomic production function as a rigorously justifiable concept. … It is either an illuminating parable, or else a mere device for handling data, to be used so long as it gives good empirical results, and to be abandoned as soon as it doesn’t, or as soon as something else better comes along.

        my problem is that even though conditional convergence isn’t exactly a unique prediction, and even though Solow leaves many questions open, the model hasn’t yet been superseded by something more comprehensive.

        I get that MRW reduce the impact of savings by adding human capital, but it’s still significant enough to cast doubts on the model. I suppose I just have a problem with the way they find the data doesn’t agree with the model, then use incredibly hazy and hard to measure concepts like human capital and TFP to try and correct for this. The whole thing seems spurious to me.

      • #47 by Luis Enrique on March 20, 2013 - 2:12 pm

        no, MRW increase the impact of savings by adding human capital to the model. The problem was that in the basic Solow model the impact of savings is too small, relative to that they estimate in the data.

        Of course human (and organizational) capital is a nebulous concept and hard to measure. But would you suggest jettisoning such concepts from economics? Empirical analysis has to make use of proxies, such as measures of investment in education, knowing full well this means only loosely grasping concept. Nobody said empirical economies was easy.

      • #48 by Unlearningecon on March 20, 2013 - 2:20 pm

        Sorry, that was badly phrased – what I meant was that they reduce the degree to which the core model must explain savings by adding human capital to explain it.

        But would you suggest jettisoning such concepts from economics?

        Does it surprise you that the answer is yes? Empirical measurement is always difficult in economics, sure, but something like TFP is such a fudge (only really measurable as a residual, and whose units we must never discuss) that I don’t regard it as justifiable.

      • #49 by Luis Enrique on March 20, 2013 - 3:34 pm

        well yes TFP is “whatever it is that explains why one firm, or country, produces more output with this combination of inputs than another firm or country with the same inputs” and you can eliminate TFP by doing things like arguing the inputs aren’t in fact the same, or by saying they are misallocated, or whatever. TFP stands for “we need explain this”. But just because TFP stands for something unexplained and only implicitly, doesn’t mean we should start ignore it. In fact I think empirical analysis that doesn’t allow for unobserved variables is usually misleading.

        but I asked about human and organizational capital, not TFP. I really don’t think you’re going to get very far without such concepts, indeed I’d have guessed you’d actually favour explanations for the wealth of nations that emphasized things like social capital.

        (p.s. the core model does not “explain” savings – the savings rate is left as exogenous. The issue is how much variation in income is explained by variation is savings. But let’s leave that there!)

        p.p.s if you want to endogenize savings, I’ve a neoclassical growth model to sell you

        p.p.p.s I think the units worry about TFP is a red herring. Whatever units you define output and inputs in gives you your units for TFP.

  12. #50 by randommarxist on March 20, 2013 - 12:36 am

    I don’t think that the problem with economics is its empirical incongruity. All sciences, even physics, have empirical data that current theories cannot explain, and I’m not going to throw out general relativity because it cannot explain certain cosmological phenomena. Now, the magnitude of incongruities is much, much larger in economics, and there seems to be few serious attempts to explain them, unlike physics. Correction to the models are usually ignored in actual work and policy prescriptions; they exist more as a rhetorical bone to throw to critics. Your article is only a criticism of the carrying out of economics, not economics itself. If this was all that was wrong with economics, we could fix it by changing the behavior of economists i.e. getting them to seriously modify models in order to explain the data.

    But it is not just the discipline of economics that is flawed, it is economics itself. The entire framework of economics is fundamentally broken. It rests on ahistorical and just plain false premises and is logically contradictory (as you have pointed out throughout this blog). The project of economics is made useless by these criticisms, and no amount of tinkering with models to fit the data will fix this.

    Note: I should clarify that by economics I am referring to neoclassical economics, specifically all approaches trying to explain economic phenomenon by individual “rational” actors maximizing utility. I do think the above applies to other schools though.

    • #51 by Unlearningecon on March 20, 2013 - 12:42 pm

      I see what you mean although my main point in this post is to point out that, while economists point to Friedman’s article when questioned about assumptions, arguing that only predictions matter, they don’t seem to have a rigorous empirically driven framework. In such a framework, something like Walrasian equilibrium would surely have been abandoned long ago.

      • #52 by Dan on March 20, 2013 - 7:59 pm

        You would think so, but when you see all the sopping moralism and normativism that is rampant in economics it all starts to make sense. What Friedman meant by “predictions” was confirmation bias and essentially equivocating your way out of falsification by coming up with excuses why your model still holds. It makes sense since many of the early economists started off as lawyers.

      • #53 by randommarxist on March 20, 2013 - 11:06 pm

        Yeah, I think your post was a good refutation of that. Just wanted to make sure you’re not endorsing the same crude empiricism as Friedman

  13. #54 by veganarchonomics on April 1, 2013 - 3:45 am

    “The common interpretation of MPT is that it predicts workers will be paid ‘what they’re worth.’ In fact, this is not correct; the theory predicts that average productivity of workers will be positively related to wages, rather than each worker getting precisely their ‘just desserts.’ In any case, the result is that MPT predicts that compensation will increase as productivity increases.”

    I remember my micro lecturers telling me that workers are paid their marginal revenue product never anything about ‘just deserts’ or ‘what their worth.’ However, economists seem perfectly willing to acknowledge that there are situations in which this will not be the case. In any event, labor producivity stats are calculated using some output measure, say GDP, and dividing by some labor supply measure, say hour worked. This would give the average product of labor, not the marginal gain (loss) from one more (less) hour worked. If the shares going to capital and labor are relatively constant, then average and maringal productivity will remain about equal (see this: http://www.unige.ch/ses/ecopo/kohli/uslp0430N.pdf) so there will be the positive relationship between average productivity and wages you predict. Labor’s share oscillated about a more or less constant long term level until about 1980 (http://research.stlouisfed.org/fred2/series/PRS85006173?rid=47&soid=22), which is close to the time that productivity and compensation start to diverge. So what exactly is the empirical problem?

    Just speculating at this point but it would seem to me the problem is that while labor is extemely productive on average its marginal productivity is, for most workers at least, low. This is not to say that worker’s bargaining power is not relevant, in fact I would bet bargaining power and marginal productivity are related (if the withdrawal of an individual worker or work hour will not diminish revenue or productivity much then workers do not have much individual bargaining power, equally lack of bargaining power diminishes ability to capture surpluses created by increased effort and so lowers productivity). Still, alot of it I think has to do with technology, and segmentation of the labor market. Probably the easiest solution would be to increase low income people’s access to capital income.

    and when you suggest that compensation has been stagnating, what do you make of these:

    http://research.stlouisfed.org/fred2/series/COMPRNFB?rid=47&soid=22

    http://research.stlouisfed.org/fred2/series/ULCNFB?rid=47&soid=22

    a lower rate of growth: sure stagnation: not really and increases that may well be in line with marginal productivity and are consistent with a declining share of labor anyway.

    • #55 by Unlearningecon on April 2, 2013 - 1:59 pm

      Labor’s share oscillated about a more or less constant long term level until about 1980 (http://research.stlouisfed.org/fred2/series/PRS85006173?rid=47&soid=22), which is close to the time that productivity and compensation start to diverge. So what exactly is the empirical problem?

      I’m confused. That is the empirical problem! Productivity was correlated with wages for a while, then stopped. So we need something other than merely productivity to explain wages.

      Still, alot of it I think has to do with technology, and segmentation of the labor market.

      I have been racking my brains for an explanation of why technology has been capital-biased since the 1970s. Surely capital biased technological change would be a steady divergence, rather than such a sudden decoupling? What was invented – or applied in the 1970s? The computer didn’t start to catch on in offices until, well, the turn of the millenium.

      As for your last data: it seems to me you are using average rather than median compensation. Average includes financiers, managers, CEOs and so forth, all of whom represent ‘capital’ more than labour.

  1. Falsification in Economics | Fifth Estate
Follow

Get every new post delivered to your Inbox.

Join 1,046 other followers