Winner of the New Statesman SPERI Prize in Political Economy 2016

Tuesday, 17 July 2012

The heterodox versus the superhuman representative agent

                Following this post, I’ve been reading the blogs of quite a few heterodox economists. There is a lot that I have read which is challenging, and which has made me think about things in different (for me) ways, which is good. But there is also lots of stuff that seems less helpful, which when it is repeated over and over becomes (for me) somewhat annoying.
                Stuff like we cannot possibly take microfounded macro seriously, because it is based on an all-embracing representative actor equipped with superhuman knowledge and forecasting abilities. To which I feel like shouting – where else do you start? I always say to PhD students, start simple, understand the simple model, and then complicate. So we start with a representative agent. What else could we do? We could start with aggregate relationships, but unless these are purely statistical, they will almost certainly appeal to theory about what individuals do. 
                What about superhuman knowledge and forecasting abilities? That seems like an extreme position. But the alternative is to assume we know what kind of mistakes agents will make. Where does this knowledge come from? I’m sure different agents are using different models from the one I’m using, but I have no idea what these models are. To keep things simple, I therefore assume I do not know what mistakes they will make, which implies rational expectations. If I want to be more realistic, I can look at the huge mainstream literature on models of learning. It is not a field I know well, but if there is a message there that we should go back to assuming adaptive expectations, I have missed it.  
                Some of the commentators on recent posts seem genuinely puzzled about why it is useful to have a representative agent or assume rational expectations. Others seem to believe that doing so must inevitably lead to laissez-faire results. The ultimate test is empirical relevance of course, and I have tried to make that case elsewhere. However I thought it might be useful to give an example of why I find thinking about a representative agent and rational expectations gives more plausible answers than using pre-microfoundations textbook analysis. It is an example where I’m genuinely curious about what heterodox economists who condemn using representative agents with rational expectations would do instead. As this is a post I’ll keep things as simple as I can and leave out most caveats and qualifications.
                The question is quite topical: what is the impact on output of a balanced budget increase in government spending in an open economy stuck at the zero lower bound (ZLB)? Well the first thing we have to do is ask whether the increase is permanent or temporary. If it’s permanent, if the import content of government spending is similar to consumer spending, and if taxes are lump sum, the answer is nothing. As the tax increase is permanent then consumption falls by the same amount, with no net impact on the demand for domestic output. That is pretty obvious, although anyone using a first year textbook would get this wrong (because they would have the mpc<1).
If the government spending increase is temporary, then the tax increase is also temporary. Thinking about an optimising consumer immediately gets us the result that consumption will initially fall by less than government spending, so there is a short run net increase in demand. (I am assuming that investment is unchanged, for standard reasons described here.) Higher demand raises output and income. If inflation does not change we get a multiplier that would be one if there were no imports. The analysis without imports is here, but we do not need it to show that output must rise.  
                In an open economy, we need to ask what will happen to the exchange rate if we have a temporary balanced budget increase in government spending, lasting no longer than the period interest rates are stuck at zero. Everyone remembers their Mundell Fleming – under flexible exchange rates fiscal policy is ineffective, because the exchange rate appreciates to crowd out the additional demand. But that is completely wrong in this case. Agents in the foreign exchange markets will note that there is going to be no increase in interest rates and no change in the steady state (so no long run appreciation or depreciation), so there is no reason for the exchange rate to move in the short run either. There is no crowding out through the exchange rate, so the analysis in the previous paragraph stands.
                This is pretty simple stuff, but it gives different – and in my view better - answers than many undergraduate textbooks. And both the representative intertemporal consumer and rational expectations were central in getting the answer. Now you may want to complicate in various ways, but that normally means building on this analysis rather than overturning it.
                So this is microfounded intertemporal macro telling us that a balanced budget fiscal expansion works at the ZLB. If you think this is obviously wrong because I’ve assumed all-embracing representative actors equipped with superhuman knowledge and forecasting abilities, tell me how you would do the analysis differently.  


  1. "I always say to PhD students, start simple, understand the simple model, and then complicate. So we start with a representative agent."

    The problem being that more advanced models tend to introduce further assumptions rather than relax them.

    Many assumptions in economics are 'domain' assumptions, where the conclusion only applies when the assumption does. Perfect information is an example of this - once you abandon that, you must start again a la Akerlof. Full employment is another one (Keynes believed neoclassical theory only applied at full employment).

    The point is that models must have clear boundaries for what to do when assumptions when they are relaxed. If all you can do to make the model remain consistent is add more assumptions then I'm sorry but you are performing an exercise in Ptolemy.

    "It is an example where I’m genuinely curious about what heterodox economists who condemn using representative agents with rational expectations would do instead."

    Steve Keen effectively uses representative agents. The difference? He splits them into classes: banks, firms, households, governments. Make a lot more sense in my opinion.

    1. It is perfectly possible to deviate from rational expectations by looking at mistakes - I do not know what you mean by starting again.
      In the example I used I actually had two representative agents: a consumer and a FOREX trader. Of course everyone uses representative agents, which is what annoys me about some heterodox rhetoric.

    2. "Steve Keen effectively uses representative agents. The difference? He splits them into classes: banks, firms, households, governments. Make a lot more sense in my opinion."

      By that logic, in a DSGE model, there are firms and households. In a price-rigidity model, some firms may adjust prices each period. So by that logic, there is no difference between Keen and DSGE?

    3. Come on zopolan, are you just trolling? The difference is that Keen's models are not based on preference driven agents and equilibrium. Banks are more than intermediaries so the financial sector is intertwined with the real sector, as are asset prices.

    4. " Perfect information is an example of this - once you abandon that, you must start again"

      um, you do know textbook DSGE does not assume perfect information?

    5. Yep, I was just using it as an example.

    6. not a terribly good example, I feel.

      Simon says: start simple then complicate
      You say:
      1. more advanced models introduce new assumptions rather than relax them
      2. Perfect information is an example of a 'domain' assumptions ... once you abandon that, you must start again

      but in text book macro, the assumption of perfect information is relaxed, and nobody has to start again. And of course the conclusions change once policy makers lack perfect information. This example supports Simon, not your supposed critique - you can start by teaching how things would work under perfect information, then relax that assumption and see what happens then.

    7. 1. Remains true and you haven't disputed it.

      2. I was unclear - by imperfect information I meant information asymmetry more than anything, rather than the kind of lack of knowledge about other firm's strategies/lack of knowledge about the model itself faced by many DSGE market participants, and the subsequent use to heuristics and guesswork.

      When transactions are characterised by information asymmetry even the standard economic assumptions lead to different behaviour. When I say 'start again' I don't mean reinvent macro; I just mean the kind of approach Akerlof used.

    8. 1. no... I don't see what your point is. Say one relaxes the assumptions behind a representative household and introduces some heterogeneity. These new elements are going to require some modelling, some new assumptions. And that is supposed to be a flaw in mainstream economics is it? Personally I don't see how any model - formal or otherwise - that introduces new elements can avoid new assumptions, implicit or otherwise.

    9. You'd expect the theory to get more realistic as it developed. You're defending an approach that starts with a vast array of incredibly unrealistic assumptions, relaxes some, introduces others and always has a large number at every level.

      You might need to introduce new assumptions when you introduce a new aspect to a model. But this should always be with a view to - and a clear idea of how - they would be relaxed. I just don't see how DSGE models pass this test.

    10. You are not making any sense.

    11. "Personally I don't see how any model - formal or otherwise - that introduces new elements can avoid new assumptions, implicit or otherwise"

      A new model with new elements supposedly has to incorporate the old model with all its assumptions. If it does not then it has to reject it.

      If you start with specific assumptions and then push them towards a general case you can end up anywhere you want in the sense of Kuhn. Your final result can be pretty much arbitrary and that is how Kuhn described scientific progress. If mainstream admitted this fact I guess everybody would be fine.

      However, while applying the Kuhnian logic of scientific push-progress of "relaxing assumptions" mainstream assumes the Popperian logic of "pull progress", i.e. gradual and inevitable movement towards absolute truth. That is a clear contradiction. It is challenged on correct grounds and it can not be defended.

  2. Aren't there Marxists who use bounded rationality in their analysis?

    Still though I think the argument that because the microfoundations have poor microfoundations themselves is reason good enough to drop neoclassical microfoundations.

    Additionally, the argument that there are emergent properties when you aggregate lots of micro actions is fairly convincing, and that's something that current microfoundations clearly leaves out.

    But I think Unlearning Economics' argument that the problem with the neoclassical model, and neoclassical model foundations with complexity tacked on at the end is that it treats reality like a special case of theory. It leads to conclusions and policies that try to force the real world to better mirror neoclassical micro theory (or worse like neoliberal policies).

    1. But these are yet more general claims. I gave a specific example because I wanted to know what was wrong with the analysis there. If these general claims are true, then they should apply to any application of microfounded macro, including the one I gave.

  3. I would use a reduced form model, where minimal theoretical restrictions are directly imposed on IRF (i.e. sign restrictions) to see what happens

    1. So you would more or less abandon theory-founded economics. That's possible, but I am not sure that's a good idea.

  4. with the data of course.

  5. What is the impact on output of a balanced budget increase in government spending in an open economy stuck at the zero lower bound (ZLB)?

    At ZLB, aren't we supposed to have I<S ? (i.e. money sleeping in banks' vaults or under mattresses)
    And thus, if government raises spending and taxes (even permanently), there is a rise in output, as the consumer will maintain his consumption and pay the additional taxes with the money hidden under the mattress?

    As output rises, and the gap between S and I narrows, the rates would move away from ZLB, and the government may reconsider its budget.

    It looks to me that you're applying an equilibrium analysis to a non-equilibrium situation. I'm saying this with much caution, because I don't doubt your competences.

    1. You are right that as the size of the BB fiscal expansion increases, the chances that the ZLB constraint still applies will fall. I'm just looking at the case where that does not happen. I'm not sure what describing being at the ZLB as non-equilibrium means.

    2. I'm not sure to use the right words (my modern macro education is not quite recent; besides I struggle a bit with the English language).
      What I'm trying to say is that if ZLB => I<S, then the economy is not in equilibrium (or you may call that a sub-optimum equilibrium).
      Now, maybe I'm wrong, but it seems to me that a rationale representative agent would always drive the economy to an optimum equilibrium.
      And so my question could be phrased "is the use of the representative agent consistent with a ZLB situation?"

    3. At the ZLB, the real interest rate is too high to get us to the optimum. The nominal interest rate cannot fall any further by definition. So to get to the optimum the expected rate of inflation must rise. But if that means that the inflation rate goes above the inflation target, the monetary authorities will stop that happening, so we get stuck - even with rational representative agents. I explain this in more detail here:

  6. I don't understand what your final argument is supposed to prove. Yes, in a simple microfounded model, there will be a zero multiplier for a permanent change in spending balanced by a permanent change in taxes. Are you saying that because we have independent reason to think this is true of real economies, this is evidence that the simplifying assumptions of the model reasonably capture what happens in real economies? Or are you saying that since we have proven a priori that the multiplier must be zero in those conditions, we can ignore any empirical evidence on the question.

    The fact that strong conclusions follow from a model isn't, as far as I can see, an argument for using that model. Why are your answers "better" than the undergraduate textbook ones? Just because they are microfounded?

    Anyway, to your question of how else to do it. A Keynesian approach starts with the fact that forming "true" beleifs about many events in the distant future is not possible. This is partly due to the inherent complexity (non-ergodicity if you like) of the economy, and partly due to the fact that future outcomes depend on agents' beliefs about them, with multiple equilibria possible. In addition, there are no markets for most future-dated goods, so no contracts possible and no price discovery. (Among other things, it is impossible to contract with people who do not yet exist.) But people still have to take actions with important outcomes in the distant future. The result is that people adopt conventional beliefs about the future, and assume that current conditions will be more persistent than they are in fact likely to be.

    I think this is just as reasonable a starting point as your "assume no one makes mistakes," and it fits what we see in the world much better. Take the stock market. Do we see stock prices generally giving an unbiased estimate of the present value of future dividend flows? Or do we see large swings between pessimism and optimism and overweighting of recent news, leading to large price deviations from fundamentals that are not arbitraged away due to liquidity constraints and a lack of contracts spanning more than a short period. Pretty clearly, the latter. (Shiller is definitive.) And if expectations in the stock market looks so much more Keynesian than "rational", how much more is that likely to be true for market that are much less complete?

    1. As I said the ultimate test is empirical, and I tried to address that point - as much as you can in a single post - elsewhere. What I wanted to tell was a simple story that used representative agents and rational expectations, and ask why that was wrong. You say that you would prefer to assume that people adopt conventional beliefs. So what is conventional in this case in the FOREX markets? What do you think would happen to the exchange rate? I too think that markets often get fundamentals wrong, but that does not tell me what to assume in this particular case.

    2. I think the balanced budget multiplier would be positive, for a variety of reasons. (I don't think your "proof" that it is zero is valid.) I think that the main effect on the exchange rate would be via higher incomes drawing in more imports, so we would expect a depreciation. More generally, I think Mundell-Fleming and the elasticities approach are examples of macro using simple aggregate relationships (based of course on plausible accounts of micro behavior, but not explicit optimizing) that explain the behavior of trade flows and exchange rates much better than your preferred approach does.

      Income and price elasticities are measurable, stable and quite useful. Nothing is gained by "deriving" them from explicit microfoundations.

  7. I think that the heterdox economics that I am still becoming familiar with (MMT) might start by asking if the government you are describing issues its own fiat currency. If that is the case, they might question your question's relevance. If the economy was suffering from high unemployment and had low inflation (presumably the reason for the zero rates), the government should spend without offsetting that spending with taxes. In MMT, the purposes of taxes are to give the currency its value as a currency and to slow the economy if spending causes inflation. That being said, I dont know how they would analyze your particular question in which the government inexplicably tries to stimulate the economy and contract it at the same time. Perhaps they would say that the policy makers don't know what they are doing.

    1. But that is a non-answer. Governments often do silly things, and its still good to know what effect it has.

    2. No expert here but your comment is exactly why MMT makes so much more sense to me than anything else I have read (not meant to disregard anyone's work).

      MMT makes the effort to understand (very convincingly) how the monetary system actually works. From that point they have drawn conclusions to correct problems in the current environment.

      For me they have been very insightful.

    3. Thank you for your reply. Yes, it was a non-answer of sorts. I think MMT holds that government spending decisions should be considered separately from taxing decisions. But the question would be relevant at the state or local government level. And, as you say, governments often do silly things, and its still good to know what effect it has.
      So, in this particular question with these particular circumstances that you have proposed, I dont have a better way to do the analysis. Maybe thats the way an MMT economist would do it if the economy was at full employment. Or if they were advising a local government. I will try and ask.
      I just linked to your post "Savings Equals Investment" and think its excellent. One question though- what part of the paradox of thrift was microfounded?

    4. I'm glad you liked the S=I post. The paradox of thrift was not based on a microfounded model. But the fact that you can also get it from a microfounded model makes me much more confident that its a robust result!

    5. Well, I have asked a few "heterodox" economists how they would approach your question. I have not received a direct reply, which is understandable. However, Bill Mitchell of billy blog, who is one of the pre-eminant MMT economists, explained in his subsequent post that sectoral balance analysis is his preferred method. There is, of course, little reason to assume he was responding to my or your particular question. Since then you have done a post on sectoral balances which I found quite interesting. And that makes me wonder why that wouldn't be your first approach.

      Anyways, since that time there have been a lot of posts by a lot of economists on this subject. Thank you for allowing me to be a small part of this discussion. But please remember that this economy is not working for millions of people in my country and yours. So come up with an answer, whatever model is used.

    6. In case you are still interested, I found this while looking for something else. "What is the balanced-budget multiplier?"

      I think it is fair to say that he uses a sectoral balances approach to this analysis. The marginal propensity to consume is also a factor- not sure if that is considered microfounded. If I have understood him correctly, a balanced-budget increase in spending would be mildly stimulating to the economy.

  8. Aren't there unspoken assumptions here about the monetary target and the CB reaction function, both of which agents must know, e.g. to determine the fx rate?
    And because you have assumed no CB reaction, you have assumed your result?

    e.g. if fiscal policy moves (expected) future demand, length of the period at the ZLB should shorten, which raises real interest rates and causes current fx rate to appreciate in anticipation, which is an offsetting drag on demand.

    1. Of course, but a little pedantic. I'm obviously assuming that the fiscal expansion is not sufficient to lift the ZLB constraint. The interesting thing you say is that I have assumed my result. But only because agents know that the fiscal expansion will not change interest rates - they anticipate the central bank's lack of reaction. If instead they believed (inappropriately) the Mundell Fleming result applied here, the exchange rate would appreciate.
      The point I am trying to make is that in this case the assumption of rational expectations seems quite natural, and not superhuman at all.

    2. OK, I see your point.

      What didn't occur to me before: what if it requires a superhuman to know the monetary reaction function?

      And I can't help thinking about Japan, with deficits, ZLB, and Yen appreciation - Mundell-Fleming seems to have won there.

  9. "So we start with a representative agent. What else could we do?"

    An economy doesn't just have representative agents. It has institutions such as production firms, banks/financial institutions, the government, central bank. So you start with households and institutions.

    "The question is quite topical: what is the impact on output of a balanced budget increase in government spending in an open economy stuck at the zero lower bound (ZLB)? "

    The question itself is at fault. There is no "balanced budget increase in government spending". By starting with such an assumption, you also imbibe the culture/false thinking in PhD students that the government budget be balanced.

    "If the government spending increase is temporary, then the tax increase is also temporary."

    How about a permenant increase in government expenditure by some amount and a slight decrease in tax *rate* (as opposed to tax) by a slight amount permanently?

    "Everyone remembers their Mundell Fleming"

    The Mundell Fleming is based on an assumption of exogenous money right?

    "Now you may want to complicate in various ways, but that normally means building on this analysis rather than overturning it."

    Well, economists generally don't see the reverse causality and no amount of model building is going to help. Of course model building is a good thing but there are certain things which can be understood even without models and I see economists continuously failing to see it. Take for example, the saving investment relationship. Economists continue to argue that more saving is needed for investment. I do not know how they can make a model which assigns the right causality to this!

    1. "they" as in mainstream economists.

    2. Tom Hickey says I should reply, so lets take these points in turn:
      1) In my problem I need to know how consumers react to a tax cut, and how FOREX traders react to a change in government policy. How would calling consumers households change things?
      2) Nonsense. Of course the government can increase spending and raise taxes at the same time. At no point is there any implication that they should do so.
      3) You are absolutely right about MF, which is why I said it does not apply in this case. So it would be irrational for FOREX traders to think that result applied. But there I go again, assuming rationality.
      4) I do not really understand the last paragraph so will not comment.

    3. As for 2 the challenge was that government spending is done in absolute terms while taxes are fixed in relative % terms. Therefore it does not follow that increasing government expenditure automatically requires increasing taxes to balance the budget again. As Ramanan said increasing government spending might require *dec*creasing taxes in order to balance budget.

      Another point is that the problem setup is pretty irrelevant for any economy. Intentional government spending increases happen ex ante while taxes are realized ex post. It is irrational to expect that ex ante will be equal ex post because reality says that realized budgets are never like planned ones. And so your problem does not exist and therefore any answer to a non-existent problem does not prove anything relevant for the real world (even assuming it is correct). Does it imply that heterodox and mainstream are like euclidean and lobachevsky?

  10. I should add, everyone who does macro modeling in a non academic context, professional forecasters working for businesses and governments and so on, uses simple models of aggregate relationships, not models derived from optimizing individuals. There is simply no way to get an operational model out of explicit optimizing behavior without imposing so many simplifying assumptions that it is useless for analyzing real economies.

    If I'm sitting at the OECD or wherever and I need an estimate for British exports in 2014, I use a reduced form model that relates trade flows to incomes and relative prices, with parameters estimated directly from macro data. Or take your original example. Nobody who takes the question of fiscal multipliers at all seriously tries to answer it in that a priori way. Anybody interested in the balanced-budget multiplier knows there are lots of potentially important effects (e.g. on income distribution) that aren't captured in the simple model. And a more "realistic" model wouldn't be tractable or give determinant answers. So you have to look for direct empirical evidence, and not just estimate the supposed "structural" parameters. In practice, NOBODY does macro the way you describe here when it comes to concrete policy questions.

    1. I know what you mean - I have been there. But when I was an economist in government, we did look to academic macro for guidance, and now as an academic I provide it. And some central banks do now use microfounded macromodels, which assume representative agents and RE, although whether they should is an interesting question.
      But my point was much simpler. Theory almost always comes from thinking about representative agents, so to argue that theory should not is a bit strange. And rational expectations, as I hoped my example would show, is often the common sense thing to use.

  11. As a layman I view the problem with this model as follows: it basically assumes the solution.

    If you modeled this way a response of an exploding population to resource/food scarcity you would get that the rational, infinitely forward looking individuals would start killing themselves as newborns the moment the population reaches the max planet capacity. Your model would "predict" that the population would top at this maximum and stay there. This solution is clearly absurd. The problem with econ is that its solutions are as absurd but in a way invisible to economists.

    A government representing a growing population and a growing economy would surely find a way to inject more money into the economy with growing debt. Your assumption of "rationality" is actually irrational - you assume the government in the future will be irrational and choke the economy along your prescriptions and that the population knows this already and will refuse to spend the received money. You assume your thinking to be rational, and then impose this "rationality" on the agents, and voila - you get whatever you had thought was "rational" in the first place.

  12. Simple macroeconomic models can produce complex, chaotic dynamics. I don't see why one would insist agents in such models to converge on rational expectations. For an argument along these lines, see:

    Anna Agliari, Carl Chiarella, and Laura Gardini (2006). "A Re-Evaluation of Adaptive Expectations in Light of Global Nonlinear Dynamic Analysis", JEBO, V. 60: pp. 526-552.

    Our host seems to insist on an unfounded special case, a special case whose assumptions he cannot even state.

    1. ??? I have no idea what your last sentence means. No one has yet told me where my simple analysis of a balanced budget multiplier is wrong, and what their analysis would be instead.

    2. "No one has yet told me where my simple analysis of a balanced budget multiplier is wrong"

      Because your assumption of a balanced budget is wrong. It is *ir*rational to assume it will be balanced.

  13. I guess that is our host's way of thanking me for directly addressing his comment:

    "if there is a message there that we should go back to assuming adaptive expectations, I have missed it."

    On his blog, Lars P. Syll directs you to a result of a laboratory experiment. I also like that chaotic dynamics emerge in the simple game of Rock-Paper-Scissors. See:

    Youzuru Sato, Eizo Akiyama, and J. Doyne Farmer (2002). "Chaos in Learning a Simple Two-Person Game", Proceedings of the National Academy of Sciences. V. 99, no. 7: pp. 4748-4751.

    When does behavior consistent with rational expectations emerge in a world in which quite simple models of economic agents generate chaos?

    Our host seems to insist on an unfounded special case, a special case whose assumptions he cannot even state. But he is conforming to the anti-scientific norms macroeconomists have imposed - sometimes with bullying - for decades.

    1. Any analysis of the impact of a balanced budget fiscal expansion has to make some assumption about how expectations change. If you do not assume rational expectations, what do you assume?

    2. We know people do not have rational expectations. No economist thinks they do. No economists takes the idea of a known probability distribution over future events as being literally true.

      How do these objections differ from simply pointing out what we already know?

      We also know that models have to make assumptions that aren't literally true, the question is always whether an assumption leads us seriously astray or not, and renders the model useless.

      I think, and afaik most macroeconomics think, and I'd guess Simon agrees, that moving away from RatEx would be fruitful. To use the recent Nobel winners as examples, Chris Sims is working on rational inattention and Thomas Sargent works on Knightean Uncertainty (ambiguity)

      How far the assumption of Rational Expectations has led macroeconomists astray, I couldn't say and I don't see how anybody else could either, with much confidence. We'll have to wait until we see how results differ under improved models of expectations formation. (Although we can already compare adaptive expectations models with Rat Ex - I don't know what that comparison shows us).

      afaics, none of these heterodox critics are really engaging with the point that the assumption of rational expectations is a sensible place to start from a modelling point of view. We need models in which capture in some sense how the economy works, how policy takes affect and how agents form expectations and choose actions, and starting from a position in which these things are mutually consistent is sensible.

      also, afaik, many macroeconomists have an appropriate sense of the "reality" of their models, and see them as essentially parables in maths which are useful for taking arguments to their logical conclusion, useful thought experiments and such like, and don't seem them as simulcra of reality from which we can simply read-off policy conclusions. And whilst economists might put forward their best shot at a model capable for forecasting, they are well aware of how short they fall. It often seems to me like heterodox critics proceed as if economists think otherwise.

    3. I want to add - I think one sense in which representative agents + rational expectations has led macroeconomists astray is that it has perhaps meant less attention has been applied to problems like coordination amongst heterogeneous agents, that cannot be tackled in such a set-up. Of course some economists do work on such things, but they are not the dominant way of thinking about business cycles and stabilisation policy. Some people, like Solow if I understand him correctly, think that might have been a better route to go down, and my guess is the same.

    4. Luis,

      I think your first comment makes some important points, but that the last is misguided. The idea that representative agents and rational expectations forces us to ignore coordination problems is simply a misunderstanding. Here, let me take an example: You seem literate enough to know about the classic game theoretic device of the "prisoners' dilemma". Both prisoners are in fact "representative agents". They are identical in all aspects! Yet, there is a clear coordination failure in which one of the Nash equilibria are suboptimal.

      Rep. agent does not imply that everyone acts in harmony. Everyone acts in his or her own interest. These interests are identical across agents, but they would all like something to happen at the loss of someone else. We don't need heterogeneity to reach that point!

      I'm personally a macroeconomist who has worked a lot on heterogeneity. But my conclusion from this work has been that there is very little to learn about macroeconomic aggregates from heterogenous agents. Cross-section, though, is another game altogether.

    5. Pontus,

      Oh i am surprised to hear that. I think I see what you mean ....I don't have any expertise in this area, although I have read some surveys that take the position that heterogenous agents yield new insights. Could you lift the veil of Internet anonymity and link to your work? I had in mind studying the importance of things like heterogenous beliefs, but even if, as you say, it is possible to use representative agents to think about strategic behaviour where agents are identical and face symmetric problems, there are other potentially important avenues of heterogeneity. Multiple sectors, for example, modelled as multiple representative agents, I suppose.

    6. Luis,

      I had in mind a very particular type of heterogeneity (that's the problem isn't it? You can add heterogeneity in infinitely many ways, but impose homogeneity in only one way!) in which agents vary by their skills, and ultimately by their wealth. Just the fact that some are rich and some are poor matters very little for aggregate movements. But there might of course be situations which I am unaware of where this type of heterogeneity might matter.

      With respect to beliefs and, more precisely, information heterogeneity matters a lot! See for instance the work by Guido Lorenzoni. But Prof. Lorenzoni would still insist that the agents he models are indeed "representative agents". Why? The agents are all identical in all aspects apart from information. This means that if two agents receive the same information, they will behave identically.

      To be honest, it bugs me a little when post-keynesians shit on representative agents. First of all, there is never one rep. agent. There is always a household, a firm, a central bank, a government, a banking sector and so on. Coordination failures are everywhere (see for instance the Diamond and Dybvig model on bank runs which relies on representative agents and rational expectations! In fact, their idea hinges on rational expectations), and they would still be there even if only ONE type of agent was considered. Second, most post-keyenesians speak about different groups having different "marginal propensity to consume" and so on. But this is representative behavior! And representative behavior is an even stricter assumption than a representative agent!

    7. "If you do not assume rational expectations, what do you assume?"

      One could assume complete ignorance by having the agent make choices completely at random. While this might not generate behaviour that is any more realistic than that generated by the perfect omniscience of agents with rational expectations, it would at least cover the opposite end of the spectrum and demonstrate how much or how little the rational expectations assumption affects the predictions of the model concerned.

  14. All that analysis in in the textbook Macroeconomics in Emerging Markets, by Peter Montiel.

  15. In my view, a valid economic model forbids the use of "representative" agents, even classes of them. However you might use "agents" which can make desicions, whatever they may be. This decision maker could be me, you, a firm, a bank, governments, the mafia, a computer program or even my cat. All of them can make "decisions" with real economic outcome and need to be considered. Of course not all agents are the same. You might introduce a factor named "power" to indicate there economic abilities, whatever sort they may be. Of course it would be absurd to assume something like rationality, a deterministic future, perfect information and such. so you don't include them in your model. Think of "agents" as evolutionary computer programs, pursuing different goals (e.g. "Don't die", maximise all your wealth, provide for your family, ...) with different strategies (e.g. get good education, steal, become a politician, found an enterprise, get in debt). The environment in which these agents are operating is of course constantly changing, partly because of their own economic activities. Therefore you could introduce a factor called "risk", which covers insureable and non-insureable risk, also cash/(debt)- and equity-flows between agents and of course technological and capital structure change.

    This would, in my view, constitute a reasonable "heterodox" model. Of course you can't make a lot of fancy predictions with it, which is, however no different than current economic models. But it gives you a framework to *think* about the economy in a way, which is not reliant on explicitly false assumptions. That's one of the first things you learn in math: If your assumptions are wrong you can prove anything, which means your calculation is wrong.

    1. "If your assumptions are wrong you can prove anything, which means your calculation is wrong."

      No, it doesn't mean your calculation is wrong. It just means your assumptions are wrong.

  16. "But the alternative is to assume we know what kind of mistakes agents will make. Where does this knowledge come from?" From my limited understanding of psychology, certain kinds of cognitive errors are more common than others. Has anyone tried to model economic issues by incorporating likely cognitive errors?

  17. oh and now you know my "model" i can't wait to give you my *opinion* on your question. I've never heard this phrase before, but I believe you mean by "balanced budget increase" a balanced budget, which is expanded both on the income side and spending side. This, of course assumes that a) the government is in control of its budget (which it is not fully, because taxes are dependent on economic activity and spending not entirely discretionary) and b) there was a balanced budget before. Which is not the actual case. ZLB I interpret as low economic activity.

    Now there are several possible ways:
    First, consider all new income comes from newly printed money and all spending is done by burning money. This of course has no effect at all, I assume. Consider a different case: All money is printed and given to everyone in a way which forces everyone to spend. This would presunably increase economic activity, but may have negative effects on inflation, capital structure, labour availibility, scarce resources, trade deficit and so on. So you can't be certain.
    Consider yet another case: All money is taxed via a super regressive head tax (the poorer you are the more you have to pay) and all money is going to buy stocks and bonds, which are then burned. This may actually also increase economic activity as the majority of the people is now more desperate and corporations are freed from liabilities. It may also increase crime and political instability.
    As we see, the question 'temporary or not' is not the first question one should ask, especially because you can't be certain about political decisions. The question should rather be: who gains and how?

    As we've considered these 'extreme' examples, one might guess that the best way is somewhere in between. In the end it requires political judgement, not economic "science".

  18. Im an econ undergraduate with an interest in Post-Keynesian economics, heres my attempt to give a heterodox answer to the question.

    A common concept in Post-Keynesian economics is endogenous money; money is created by the banks with loans, and destroyed when loans are repaid. Therefore S does not equal I.

    Because S does not equal I, from Steven Keen we get GDP = Income + change in debt.

    At the zero lower bound I assume the private sector is deleveraging (S > I); in aggregate more loans are being repaid than new loans created. Increasing taxes targeted at people who save, such as upper tax brackets, can reduce the amount of savings, while higher government spending increases income. So it may be possible that higher government spending with a balanced budget may increase GDP, but deficits are better.

    "What about superhuman knowledge and forecasting abilities? That seems like an extreme position. But the alternative is to assume we know what kind of mistakes agents will make. Where does this knowledge come from? I’m sure different agents are using different models from the one I’m using, but I have no idea what these models are."

    I believe Post-Keynesians and maybe other heterodox schools of thoughts, reject rational expectations and adopt uncertainty. People are uncertain about the future and rely on feedback from the economy about what the future economy will be like.I think an example is Minsky's financial instability hypothesis.

  19. Representative agent models are simpler to work with, but I cannot really trust results obtained with such a model. At this day and age there is no reason why not to use heterogeneous models.

  20. "If you do not assume rational expectations, what do you assume?" You assume only very local knowledge and very basic choice -- kind of like gas molecules bumping around, converging on a global equilibrium. Eric Smith and Duncan Foley have done this in "Classical thermodynamics and economic general equilibrium theory" JEDC 32, 7-65.

    Turns out you can still get pretty much all the equilibrium results you really need without unrealistic assumptions.

    Foley has elsewhere written much more on this kind of model.

  21. Have a look at James K. Galbraith, "The Final Death (and Next Life) of Maynard Keynes" (Keynote address, May 13, 2011) for suggestions about other approaches, like ones that actually predicted the crisis and explained whys and wherefores. These approaches also explain the solution.

  22. Benedict@Large19 July 2012 at 09:38

    "[If] the tax increase is permanent then consumption falls by the same amount, with no net impact on the demand for domestic output. That is pretty obvious, ..."

    It is? If the tax increase is targetted at components that are added to make up a higher monetary aggregate, and the spending increase is targetted at the components that make up a lower monetary aggregate, why wouldn't we expect an increase in velocity?

  23. With respect professor, while I agree with you agents do not need to be omniscient to model economies (observable patterns emerge, which is all agents need to model with approximate predictive accuracy), I think you might be missing the underlying, principal point of heterodox economists. There is no reason to have a balanced budget increase for some brief, temporary increase in output. It appears you're operating under a paradigm which assumes budgets ultimately need balancing. To the contrary, I would propose to you balanced budgets are useful only within a specific economic context, i.e., within the context of healthy aggregate levels of demand, employment, and output (which are largely three sides to the same coin). In this context, to run a government surplus would reduce consumer incomes, and hence acts to reduce aggregate demand. Conversely, within this same context, running a deficit risks inputing excessive income. Excess incomes lead to excess demand and hence to potentially harmful levels of inflation. Naturally then a balanced budget is the sensible choice. This choice however is specific to this economic context, not to all.

    As is no secret, economies throughout the West do not find themselves in the aforementioned healthy economic environment. Rather, they suffer from insufficient aggregate demand, high levels of unemployment, and lower than potential output. At their root, these conditions are a function of insufficient consumer incomes. Increase incomes, demand increases, employment increases. What heterodox economists (HEs) have realized is this: Central governments which create their own, non-convertible currency--aka currency issuers--operate under an entirely different paradigm from currency users (i.e., entities which cannot create money, e.g., households, businesses, state governments within the U.S., countries in the EU, and so on). HEs have recognized the means to increasing consumer incomes depends on currency issuers recognizing they have a responsibility to deficit spend funded from direct money creation when demand/employment/output levels are unhealthy.

    In this underproductive environment, historically central governments would often borrow money to increase spending. This is to say they run debt financed deficits. The fact is, there is no need to borrow money to fund deficit spending. Central governments, such as that found in the U.S. as well as found in your country, are not operating within the paradigm of the currency user. They can create money on demand (there may be legal hurdles, but in theory they can do this) to fund deficit spending.

    Often the policy prescription which I'm describing is met with anxiety (or vehement antipathy) over inflationary side effects. This fear typically arises from an a-priori, axiomatic acceptance of the quantity theory of money. It is held that any level of money creation will have a direct, proportionate, inflationary effect on aggregate prices. There is no other way to put it than this inflation theory is false. Increases in income do not necessarily lead to a proportionate rise in prices or any rise in prices at all. Instead, in economic environments such as our own, an increase in incomes raises employment and output with relatively stable prices. The objective of the central government is to avoid too much money creation, which would drive demand higher than healthy levels of employment and output could support, thereby causing prices to rise at harmful rates.

    It is important for central governments and their economic advisors to begin to recognize there is not just a single paradigm, but two--that of the currency user and that of the currency issuer. Balanced budgets are much more the concern of the latter than the former, with the exception being a healthy economy where maintenance is the objective.

    I appreciate your consideration of heterodox economics and the opportunity to discuss these important issues with you. Your open-mindedness is in the spirit of real science ever so vital to our civilization.

    1. I appreciate your long response, and your closing thoughts. But it is really important to distinguish - whether you are mainstream or heterodox - between policy analysis and policy advocacy. As a macroeconomist I need to be able to analyse the impact of what a government might do, whether the policy is wise or not. But having said that, I would take a temporary increase in output right now, if temporary was as long as it takes for a the economy to recover under its own steam.

    2. Understood. Analysis seems to invariably lead to advocacy/implementation, so perhaps more than theorists and/or empirical analysts realize, they have a very real impact (in fact, my argument above follows this form). I would regard theorists and/or empirical analysts as the very foundation of civilization. Human life exists in the context of theory.

      To emphasize through analogy; pure mathematics often leads to the applied, e.g., non-euclidean geometry and its relation to Einstein's relativity.

      Here in the states I'm afraid a temporary increase, while nice, would be a band-aid on a ruptured artery. I hope we can recognize the underlying soundness of heterodox theory as described above and use it as a long term prescriptive solution. Science would undoubtedly be the benefactor. Like most things, it requires funding, and there's an easy place to acquire it from what I believe to be sound economic theory.


  24. This comment has been removed by a blog administrator.

  25. Professor,
    I am not an economist, but I think your example shows very well the problems of standard macro. What is wrong with it?
    If I understand you correctly, your example assumes away the most important problems in this case, namely income and wealth distribution, propensity to spend and - most importantly - the time lag between spending and income/taxes.
    When you want to analyze policy options, I think you cannot just ignore these aspects.
    The government may decide to spend more by say increasing unemployment benefits or providing showel ready projects for the unemployed to work at, boosting their income. This increase of income is almost certainly going to be spent on reasonable things (food, rent, clothing) and thus generate other incomes (Kalecki: spending determines income!). This will, almost certainly, increase GDP as well as inflation BEFORE new taxes are collected. Especially if the tax is not targeted at the recipients of government spending (but at, say, the higher echelons of society who have enough wealth and income but are not spending it for one reason or other). At the end of the day the government may be imposing a tax of nominal value equivalent to its prior spending, but since GDP and inflation have probably increased, the real ratio of government spending and taxing will probably have shifted in some way. As a result your balanced budget assumption does not really make much sense to me, since it ignores the time lag between spending and taxing. In that lag, a lot of things can happen.


Unfortunately because of spam with embedded links (which then flag up warnings about the whole site on some browsers), I have to personally moderate all comments. As a result, your comment may not appear for some time.