Winner of the New Statesman SPERI Prize in Political Economy 2016

Tuesday 24 July 2012

Can heterodox economists constructively engage the mainstream?

For heterodox economists

                In my first post on the heterodox/mainstream divide, I suggested that there sometimes seemed to be a blanket rejection of mainstream economics by the heterodox, equivalent to the de facto rejection of heterodox economists by most of the mainstream. As someone who is both supportive and critical of mainstream macroeconomics (i.e. I believe it has strengths and weaknesses), I found this attitude disconcerting.
                But impressions can be wrong, so in a second post I presented a simple piece of analysis, that used both a representative agent and rational expectations, and I asked heterodox economists how their own analysis would differ. I repeated the challenge in a subsequent post, redoing my analysis using sector financial balances. Unless I’ve missed something, no one has – in my view at least – provided an answer. (However comments on these posts have seen some welcome interchanges between the two points of view on other issues, for which I am grateful.)
                My question was what the impact of a temporary increase in government spending, financed by taxes, would do to output in an open economy. It was not a trick question – I just happened to be thinking about a presentation on exactly this issue to policymakers, so it was on my mind. Now policymakers, in response to such a question, do not want long lectures on the sins of representative agents, rational expectations or whatever, and they certainly do not want to be told it is a bad question. They want some simple macro analysis.
                So, what do consumers do if they are told that taxes are rising temporarily? First, do they take any notice of what they are told, or do they assume the tax increase is permanent? It really matters which. If the former, do they smooth the impact on consumption, as the representative agent intertemporal model (and other earlier models) suggests, which by the way means that the consumer sector goes into deficit.
                If they do, output will tend to rise. How do traders in the FOREX market react? Do they try and work out if interest rates will rise, and if they do what are the results of this calculation? Do they say that in the past when output has increased the central bank has raised interest rates, or do they recognise that (as I assumed) we are at the zero lower bound and so no interest rate increase was likely?
                There are no obviously right answers to these questions, but I’m genuinely curious about how you would answer them without some appeal to representative agents. You could avoid rational expectations by assuming something simple like static expectations, but that would imply for consumers (after the initial surprise at taxes rising) the same thing as taxes rising permanently – until they were surprised by taxes falling again. And is the idea of consumption smoothing misguided, or just the idea that it results from optimisation?
                The point to all this is my belief that mainstream macro is not some kind of inevitably interlinked construct, that you either have to buy into completely or reject completely. I believe using a representative agent is OK for some questions, but not for others. I think it’s appropriate to assume rational expectations some of the time, but not all of the time. Any analysis has to make assumptions, and it is not clear to me why the assumptions I have made are worse than others. If you think this belief is wrong, the best way to convince me is to give an alternative story of how a temporary tax financed increase in government spending would work. But of course if you are happy to just give lectures about the evils of mainstream economics, then no response is necessary.               


  1. Have you looked at Lance Taylor, Exchange Rate Indeterminacy in Mundell-Fleming ?

    1. Yes, although not in depth. He looks at UIP, which is what my exchange rate analysis is based on. As far as I can see, he does not suggest anything better. So I'm tempted to claim that he would come to the same conclusions as I do about how this policy works.

  2. If the axioms aren't sacrosanct, the edifice will fall! You are trying to be reasonable, in the popular vernacular sense, and only cold rationality works in macro Wren-Lewis!

  3. It's very difficult to engage with somebody using ancient forms of paragraph indentation and no white space.

    The world has largely moved on from that system.

    Much as it is about to from the underlying beliefs of your models.

    1. Typical MMTer style comment. Annoying, abrasive, condescending and pointless. Why do you people bother commenting in this way all over the internet? Even your most vocal advocates like Scott Fullwiler and Randall Wray seem intent on ruining their own reputations by acting like fools on the internet. Have you no self awareness?

  4. Your problem has relevance only for the world of representative agents.

    It has no real solution because it is not defined properly even in the mainstream logic. Consumers and everybody else approach taxes as tax *rates* and not as amounts. Apart from ex post / ex ante problem of budget spending and taxing you can not claim that an increase of government spending has to be accompanied by an increase in tax *rates* to balance the budget. Your problem is over-assumed, it is missing critical information and therefore has no relevance for economic analysis in this world.

    The only possible application of your representative agent problem and balanced increase of spending to the real world is if you increase government spending and then immediately tax the whole amount of it from the recipient. I just wonder why you spend your time and everybody else on such nonsensical "problems" ...

    1. A government can increase spending, and finance it by - say - not uprating tax allowances. No tax rate need change. Of course it will not get exact balance ex post, but that is not important. As the government says that it would like to increase demand without increasing debt, and this policy does just that, I think it deserves a little more respect than you give it.

    2. I give it full respect. But if people react to tax *rates* and there are no tax *rate* changes required to balance the budget, what is your argument then? Noone thinks of taxes in terms of volumes because taxes are predominantly relative. And budget result bounces up and down all the time and without any discretionary decisions on behalf of the government.

      I also say that an assumption of balanced spending increase is without any basis in the real world. It does not happen and if it happens then it will be a pure luck/coincidence. It is impossible to balance the budget ex ante regardless of all political/economic desires. What does your question really want to ask? Why should anybody even bother with answer?

      Additionally, what does it mean to "increase taxes"? In this world tax codes are probably even thicker than legal codes. There is no such thing as "increase taxes". There are zillions of them with all loopholes and attached conditions. So which taxes will increase? And here it becomes a political/institutional issue and not an economic one. You can not assume such problems away. However if you approach them properly then you can not assume a representative agent. There are all types of non-linear re-distributional effects going on that any representative agent smell test fails in the first phase. You can surely increase granularity but every such gradual increase destroys your argument of representative agent. In the end you will have to admit that you simply have to rely on averaging certain cohorts due to information/computation limitations. However averaging and representative agent approaches are not even closely the same thing.

  5. The challenge for modern (aka heterodox) economists is not to engage with XIX century thinking (aka mainstream) economists but

    to engage with people in general.

  6. Maybe you can start by looking at the models presented in Monetary Economics, the textbook by Godley and Lavoie ( ?

    EViews Macros are available for all the programs in the book here:

  7. Re. "There are no obviously right answers to these questions, but I’m genuinely curious about how you would answer them without some appeal to representative agents" - this sounds to me very similar to some of the debates that come up in political science (and with which I'm more familiar).

    In those, one answer (which I often lean towards) is to say 'let's go and ask people', e.g. rather than try to create a new theoretical model, start with public opinion research into what voters say, lab experiments into how people behave presented with different options and so on - and then create a theory based on the evidence.

    Isn't that also an option for economists? Stepping away from blackboard, as it were, and instead looking for direct research evidence of people's views and attitudes? Whether rational agents end up being a good fit to then explain and extrapolate from those findings becomes a consequential decision rather than an initial one.

  8. Wow! I hope you don't teach. What a bunch of convoluted BS about the multiplier effect.

    1. In fact he does teach and is possibly the best macroeconomics lecturer Oxford has ever had. Unlike the general disorganisation of the Economics department, his macro lectures are both interesting and comprehensible.

    2. I agree, if he taught you you might not rubbish concepts you don't personally agree with as BS!

  9. The answer is to use both representative agents and perfect foresight, but not talk about them as representative agents and perfect foresight.

  10. regarding the question about the response to a tax increase, I can imagine a number of different potential 'heterodox'-type responses which would give different answers to the usual intertemporal Ricardian one, basically by downplaying the likely strength of forward-looking consumption-smoothing mechanisms, eg:

    - one could claim that, even at full employment, credit constraints are pervasive, either due to standard Stiglitz-Weiss informational stuff, or due to 'more heterodox' institutional reasons

    - one could deny strict utility maximisation: perhaps consumers treat their budget constraint as if it is hard, maybe as a heuristic to help them achieve a target level of saving, or as a committment device to get round intra-personal time-consistency problems (I know many people who act like this, and I sometimes do myself), so that future income doesn't impact on today's spending decisions

    - one could deny that the government does (has to?) behave in a 'Ricardian' manner, or at least that the public *believes* that the government will/has to behave in a Ricardian manner (this is a slightly different issue as to whether they believe a tax change will be temporary or permanent, I think); so in the case of a tax cut, say, the public won't interpret this as implying a future tax liabilty and will treat it as purely a gain to lifetime income

    as to the issue of how the exchange rate reacts, there are other 'heterodox' ways one could deny the relevance of standard UIP stuff (although I think they're less plausible in this case), eg:

    - one could strongly play up the 'political' element in exchange rate determination (maybe the home country's main trading partners are China and other 'bretton woods 2' countries, or Brazil, ie. countries which are believed by many to directly intervene in forex markets, or target the nominal exchange rate with monetary policy)

    - one could play up the role of currency speculation in driving nominal exchange rates, such that behavioural finance stuff becomes relevant

    - maybe one could avoid use of a representative forex trader by thinking of the choice between assets in two different currencies as analagous to Keynes's model of the choice between 'money' and 'bonds' in the GT, with a *distribution* of forex traders, all with different beliefs about what the 'long run'/'normal' nominal exchange rate is (perhaps because the latter isn't uniquely determined by economic 'fundamentals', but is also determined by politics etc etc), and using *this* (rather than a Tobin-style portfolio choice by a representative agent) to derive a downward-sloping relative currency demand schedule

    just a few suggestions as to how one *could* potentially proceed on a 'heterodox' basis. these are all purely speculative - I'm sure there does actually exist a heterodox literature on these issues, but I have no idea what it actually says, I just imagine it looks something a bit like the above.

    I imagine there is *something* to a number of the above points (although it wouldn't be easy to establish their empirical relevance, I would have thought). however, I would generally treat them at most as being deviations from the benchmark of the intertemporal Ricardian consumer. I guess if you really played up their real-world significance, you could argue that the latter can safely be jettisoned altogether, and it's valid to actually start from just assuming the above things as benchmark cases.

    1. Clearly credit constraints are always relevant. Only on the very cusp of the bursting of a major bubble do credit constraints cease to be binding for most economic agents, and by the nature of bubbles this does not last very long.

      Incidentally, this blows a hole below the water line in the "efficient markets" assumption, in the Blanchard model of stock market/output interaction, and generally in any model which assumes that people will short overvalued assets. Because even if they are correct in their "fair value" estimate, they risk being margin-called and sold out in a rising market before the correction they are betting on actually happens. Which greatly reduces their willingness to entertain an unhedged short position. (And if they can - or believe that they can - time the correction event with any degree of nicety, then they would be downright silly to be short rather than long at any point up until immediately prior to their expectation of the event.)

      - Jake

    2. Or as Keynes put it, the market can stay irrational longer than you can stay solvent.

  11. Christian,

    I'm told they model, for example, household whose behaviour is described by a consumption function. That's a representative agent in all but name.

  12. It is difficult to answer this question within a framework where money and debt are not modelled. It is not sufficient to say that everything in a DGSE model 'adds up' as money is not modelled, is money is not modelled, and neither are assets or liabilities, then there is no means of of assessing the impact of holding an inventory of money or taking out additional debt, essential at the zero lower bound. Which is why both a sectoral balance approach and a stock flow consistent model is essential.
    As to expectations most heterodox modellers would take as a starting assumption the Keynes assumption that under radical uncertainty behaviour will remain unchanged unless agents have to evolve their expectations in the light of changing conditions. So in the case of a tax rise agents will be holding excessive cash balances because of uncertainty they will be forced to either draw down these balances or reduce consumption in the case of a tax rise, but the tax rise is exactly matched by government consumption which results in an end to the net withdrawl from money in transactional circulation to savings. This will boost aggregate demand. If consumers thought there real wage was due to decrease because of economic contraction they might reduce consumption (intertemporality is not the issue here it is discretionary spending at the margin, intertemporality is only really an issue with lifetime asset purchases such as housing)but why should they as total spending in the economy stays the same (and as government expenditure is more likely to be spent domestically may have a positive BOP impact), the velocity of transaction has risen and so has growth, leading to a decline in liquidity preference and an increase in demand for money (in the Wray sense) leading to a restoration of credit growth, and with increased profitability bank lending power will be restored (see my blog). All of this can be explained without any recourse to rational expectations etc. Baggage dump.

    1. Thank you for addressing my question!

      As I said, consumers need to form expectations about future taxes. So they either need to decide whether to believe the government, or rely on some simple expectations formula. By doing the first, they are doing what rational expectations says they will do.

      As to what they do if they decide the tax rise is temporary, my discussion was consistent with a sectoral balance approach (see my post on this) - indeed that is a rather easy way of getting to the answer, but only once you have factored in behaviour i.e. consumption smoothing. There is no need to think about money, but if you have to, it should make no difference to the answer.

      I know you will not like me saying this, but it is very hard to think about this problem without using rational expectations or representative agent behaviour. If you think about your analysis, it is using both e.g. "If consumers thought their real wage was due to decrease ..." The comment by vimothy is I suspect generally true.

    2. Model-consistent expectations go out the window as soon as you move from equilibrium models to proper nonlinear dynamic models (which you must if you want to model business cycles, and you have to model business cycles if you want your research tradition to be policy-relevant). Because in nonlinear dynamic models, determining the model-consistent expectation and optimal response functions becomes both computationally and analytically intractable very, very quickly.

      Which means you have to rely on assigning the agents decision-making strategies, and metastrategies for how to modify strategies. These strategies and metastrategies can plausibly be determined from empirical studies in microeconomics, sociology, psychology and organization theory. But they are not going to look like best-response strategies under model-consistent expectations.

      - Jake

  13. "It is not sufficient to say that everything in a DGSE model 'adds up' as money is not modelled"


  14. " It is not sufficient to say that everything in a DGSE model 'adds up' as money is not modelled"

    where do people get this idea? I think they get it from Steve Keen. I happen to have a copy of Ljungqvist and Sargent, which I think is the mainstream DSGE textbook. This is not my field so I have not (yet) read it, but just looking at the preface, they discuss which varieties of model have a role for fiat money, and they promise "several models of fiat money".

  15. "I believe using a representative agent is OK for some questions, but not for others. I think it’s appropriate to assume rational expectations some of the time, but not all of the time."

    I believe you are very right here.
    But in that case, before developing and exposing a model, an economist (including you in the balanced gov spending example) should always state why he or she believes the representative agent / rational expectations may apply or not.

    (In my understanding, supply-side economists usually don't argue whether rational expectations may apply or not, but claim that they "must" apply)

  16. The key here is that in your request Government will fund their additional expenditure by raising taxes.

    The central question is than which taxes are raised. E.g. taxes on labor or consumption (sales tax increase), or raising taxes on asset values (capital tax) / land & real-estate.

    The key point is that current private debt used to finance the latter category and that has raised asset values, is then rerouted toward government investment/spending, but probably hardly noticed by people with relatively low incomes and low assets.

    So a rather generic statement about "raising taxes" to fund government is oversimple. The key question is whether taxes are raised on overvalued items / assets (stock) or on flow goods (consumption) and flow income (wages).

    It is also relevant to understand whether the government uses the additional tax income to invest it (say infrastructure) or to transfer it / redistribute it (raising the number of civil servants, paying entitlements).

    Only with that level of specification it is possible to create a model that has something meaningful to say about impacts.

    1. This is a point about abstraction, which is a different issue. Macroeconomists, including me in this case, often use the fiction (unless there is a poll tax) of lump sum taxes, because they want to abstract from the particular impacts of particular taxes. The key point is whether the abstraction is a useful simplifying device. If the complications caused by using particular taxes involve additional incentive effects, without altering the income effects, then it is useful. If people do not notice the income effects of some taxes rather than others, it is not.

    2. Under most realistic conditions, the bulk of the effect of taxes depends on their incidence and the details of their implementation. So if you abstract from incidence and the details of implementation, then you are abstracting from most of the tools which make taxes useful to the conduct of public policy.

      You can do that, of course. But then you need to keep in mind that you're studying the tail, not the dog.

      - Jake

  17. If I were going to model this problem, I'd start by asking "who are you planning to tax, and what are you planning to spend on?" If, as in your original writeup, you are going to tax everyone a lump sum and pay them back the tax in consumer goods that they would have otherwise bought, then you will get no effect (beyond the employment of a number of bureaucrats to manage the scheme). And this is going to be true whether the policy is temporary or permanent.

    But this is obviously a silly way to do fiscal policy.

    The second question you need to ask is "which part of the business cycle are you in?" You say that you are "at the zero policy rate bound." Under contemporary macrostabilization doctrine, this means that you are in a Fisher debt-deflation crisis, because that is the only time central bankers operating under contemporary orthodoxy are at the zero bound and stay there when you start making fiscal interventions.

    And now we're cooking with gas, because this allows us to identify the two most relevant groups to model: Net creditors and net debtors. The debt-deflation cycle is a pathology that occurs because net debtors owe net creditors too much money. The solution is to make net debtors not owe net creditors too much money.

    There are two ways in which balanced-budget fiscal policy can go about that: You can tax net creditors and pay net debtors cash in hand. This will expedite the resolution of the crisis, but will not provide immediate relief to output and employment, because all the money will go towards paying down debt (or near enough as makes no matter). Or you can tax net creditors and spend the money on buying stuff from net debtors (such as by engaging in useful public works). This will provide immediate relief to output and employment and expedite the resolution of the crisis, by giving net debtors money to pay down their debts.

    There is no need to invoke expectations here, model-consistent or otherwise, because in this scenario net credit contraction drives economic behavior, gross credit contraction is not discretionary (it is determined by already existing contracts), credit creation is demand-constrained, and the demand side of the credit market has no discretionary power for expectations to influence (because it is insolvent).

    The model does use representative agents (or can at least be trivially recast in terms of representative agents without altering the conclusions). But then, I don't object to the use of representative agents per se, only to improper construction of representative agents.

    Improper construction of representative agents is typically due to either excessive or improper aggregation. One example of excessive aggregation is modeling the whole economy as one representative agent, and then concluding that because he owes all debts to himself debt is irrelevant to the analysis. An example of improper aggregation would be arguing that higher wages reduces employment because the representative firm will hire fewer workers (the aggregation error here is the implicit assumption that the independence of the firm's budget constraint on wages is preserved under aggregation).

    - Jake

    1. Jake - thank you also for answering my question. I agree with what you say about representative agents. All I would add is that mainstream macro at the moment is all about addressing this issue, so there really should be common ground here.

      I do not agree with what you say about the business cycle. I would argue that, using conventional New Keynesian analysis, being at the ZLB means we have a problem of deficient demand. Hence output is demand determined. I do not understand why you say it must imply a debt-deflation crisis.

      In my analysis, I do not consider the debtor/creditor question. I think it is a very interesting question to address. Can you can address it without thinking about expectations? That would surprise me, as most economic decisions require thinking about the future, but not all do. But I would then just ask, when you do have to think about a case where agents do have to form expectations, what do you assume?

      But I also had a genuine query. Suppose in your analysis the government spending was on public works that employed the unemployed, who were neither creditors nor debtors, and it was financed by raising taxes from creditors. I think you would say that this did nothing to resolve the crisis. But in my analysis it would raise demand and therefore output, because at least the creditors smooth consumption. Does it in yours?

      - Simon

    2. I would argue that, using conventional New Keynesian analysis, being at the ZLB means we have a problem of deficient demand. Hence output is demand determined.
      We don't disagree on that, since a debt-deflation crisis implies deficient demand. (The converse, however, is not true: Debt-deflation is not the only type of event that gives you deficient demand.)

      I do not understand why you say it must imply a debt-deflation crisis.
      That's black empiricism: A debt-deflation spiral is the only sort of event which has been historically observed to prompt central banks to go clear to the zero bound and stay there for any extended period of time. Imputing a debt-deflation event from the mention of the zero bound in the original question is a historical inference, not a theoretical one.

      In my analysis, I do not consider the debtor/creditor question. I think it is a very interesting question to address. Can you can address it without thinking about expectations?
      I would argue yes. In a severe credit event, enforced savings (enforced by contracts of debt service and amortization - essentially contractually enforced chasing of sunk costs) exceed desired savings. So the private sector has no agency: Its expectations don't matter, because it is forced to save more than it can plausibly be made to desire to either save or invest.

      when you do have to think about a case where agents do have to form expectations, what do you assume?
      Generally that they do the same thing they did yesterday, with minor, opportunistic modification. Until and unless they get a sufficiently strong, sufficiently immediate negative feedback to prompt a wholesale reevaluation of their strategy. The key point here is that wholesale reevaluation of your strategy is costly and uncomfortable, both for individuals and institutions. So unless something drastic happens, you only explore a small parameter space around that strategy, looking for opportunistic improvements.

      Now, this does not preclude agents from using the same model I am using to obtain their expectations for the future. "Use the model Jake is using to make a forecast" is a valid strategy. But for nonlinear dynamic models that's not model-consistent expectations, because such models typically do not have analytical solutions and do have complex numerical behavior. Which means that the best you can do, if the parent model is to execute in finite time, is to compute some finite number of steps ahead with some finite precision, both of which can usually scale only logarithmically with available computational power.

      Suppose in your analysis the government spending was on public works that employed the unemployed, who were neither creditors nor debtors, and it was financed by raising taxes from creditors.
      It does provide relief for output and employment by raising demand. It wouldn't reduce the debt overhang (to first order - to higher order it does, of course, because the newly employed spend their wages buying stuff from people who do use it to pay down debts).

      Take the US federal government spending during the last world war: If the war had stopped in 1942, and spending been immediately cut back to prewar levels, it is probable that there would have been a recession, because not enough of the 1920s debt overhang had been rolled back. So you had an economy at (arguably over-)full employment, but the underlying pathology had not been resolved (stock-flow consistency is your friend here - employment and output are flows, debt overhang is a stock).

      This talk by Richard Koo makes the point very nicely:

      - Jake

    3. We agree on the first point, and if your argument about the ZLB going with a debt crisis is empirical I have no problem with that either.

      On the fourth point, you seem to be saying that assuming model consistent expectations is OK (unless perhaps after a structural change), but not in the kind of non-linear model you use to describe business cycles. No problem with that, but I am not sure Lars Syll would agree!

      On the final point, again I do not think we disagree. What I think I would argue is that time may eventually cure a debt problem, and if the government is not going to do anything else to help, stimulating demand while the problem works itself out is better than doing nothing. Martin Wolf puts it well in his blog today ( for debt financed fiscal expansion, and a balanced budget fiscal expansion offsets the saving by debtors with deficits run by creditors.

      Sorry for the delay in your last comment appearing. I think it was including the video that made Blogger think it was spam.

    4. Time will obviously reduce a debt overhang on its own as long as you can keep your economy out of deflation (for which fiscal policy enforced full employment should suffice). But insofar as you want to return to the "normal rules of capitalism" as soon as possible (which is, granted, a political statement, but a very popular one), it makes sense to pay attention to how fast you can expect the debt overhang to go away.

      My main issue with model-consistent expectations, even in models where they are analytically tractable, is that they preclude endogenously generated macroeconomic surprises. If all agents have perfect foresight arbitrarily far into the future, then a macroeconomic imbalance cannot "sneak up on them." And yet in the real world, macroeconomic imbalances come as a professed surprise to economic agents with dreary regularity. This indicates a fundamental problem with model-consistent expectations models.

      If you are using model-consistent expectations despite this reservation, I think you should at a minimum make sure to check their robustness against introduction of agents with non-model-consistent expectations. If your model blows up under any arbitrarily small volume of trading based on idiosyncratic heuristics (such as the Blanchard model does), then you need to go back to the drawing board.

      - Jake

  18. I cannot answer your specific question but laud the effort. Nick Rowe has done a number of posts on MMT and with the help of many commenters i think made some headway. Unfortunately, my observation is that the ability of the heterodox economists to engage is limited by the fact they they do not *want* to engage. The use of humpty-dumpty lingo becomes like an Abbott& Costello "whos on first routine." My own observation is that once you get past the translation, there are elements of so-called heterodox economics in mainstream economics (for example, many of Steve Williamson's comments bear a striking resemblance to MMMT positions, but at least its easy to identify the working assumption like interest on reserves that the conclusion hinges on).

    Criticism is healthy, however, I have to say i have an immediate, instinctive, and deep skepticism for any anything that *chooses* different language when it appears the only purpose is to obfuscate. That applies to heterodox economists, Austrians, and Christian Scientists alike (among others). So maybe i am part of that distressing divide you mention, but i see much of heterodox economics as going a step beyond mere criticism into willful ignorance.

  19. I think my problem with your challenge, as a semi-detached heterodox economist, is that presumably without realising it, you've stacked the deck - it's rather like the fable of the time that the mole decided to run a "king of the animals" contest and spent all day designing a superbly fair set of rules for a digging competition.

    What language do they speak in this open economy? Which side were they on in the Second World War? In the Cold War? Who is in government and how big is their majority? What are their biggest export industries? Are they a net importer or exporter of food and energy? etc etc.

    What I mean here is - because you've set the challenge in terms of describing the parameters of an abstract model of a stylised economy, you've immediately dealt institutionalists like me out of the game, and put a big handicap on anyone who deals with questions like the demand elasticity of a temporary tax rise by using ad hoc empiricism. I remember once being at the Bank of England and needing to model the effect of the Halifax demutualisation windfall; nearly all of Monetary Analysis went off into deep discussions of permanent-income and overlapping-generations, and I would guess that not one in five of them was particularly interested to know that 50% of the Abbey National demutualisation windfall had been saved, six years earlier.

    When you said that "I just happened to be thinking about a presentation on exactly this issue to policymakers", you weren't really - you were thinking about a specific tax-financed temporary spending increase, in a specific open economy.

    So I think my answer to your question is that I would model this question in any context I cared about by using heuristic or empirical work on the actual consumption behaviour of the economy I was looking at. I suppose you could describe it as a representative agent approach but actually I would be trying to work with the characteristics of the aggregates themselves. In terms of answering this question for a generic economy in the abstract, I kind of end up responding that this is something I would recommend not doing.

    1. Footnote: reading that back makes it look like I'm against modelling per se and I'm not; only that all macro models of interest have a reduced form susceptible to empirical estimation, and that for any such class I would tend to just say - why bother with anything beyond the reduced form?

    2. Let me give you a very topical example of why its important to do mainstream macro theory as well as empirical analysis. Alesina and others have argued that the best way to reduce government debt is by cutting government spending rather than raise taxes, and they justified this primarily using empirical analysis. They work was highly influential - yes, of course in part because that is what some people wanted to here, but I know it also influenced more open minds.

      Now it turns out that doing the empirical work a bit more carefully gives you different results - I've written about this here: But that took time, and the results are still argued over. And to be honest, the more country and time specific detail you put into that analysis, the less clear things can be made to seem. (Sense intended.)

      However their claim makes very little sense in terms of macro theory - theory that is just a variation on balanced budget fiscal expansion. That is a point that (if I had been blogging at the time) I could have figured that out and written about it immediately.

      So it is important to do this kind of macro analysis as well as the empirical work, and then add in the kind of stuff you discuss as well. If your Bank of England colleagues really were not interested in past evidence, that was clearly wrong. But they were right to want to back the evidence up with a plausible story.

    3. I think the reply "a combination of ad hoc empiricism and specific institutional knowledge" is perfectly sensible, and think any macroeconomist who, faced with this question, would think only in terms of formal models and would ignore these approaches is going wrong. But it strikes me your approach has similar sorts of shortcomings as mainstream economics.

      As far as I know, ad hoc empiricism is not any better at supplying a guide to the future than those supplied by an appropriate DSGE model.

      And what do you do with your specific institutional knowledge? You have to insert it into some kind of model of the world, albeit an implicit informal model. Again, this isn't necessarily a bad thing: doing so can allow us to think about things that we are incapable of modelling formally. At the same time, isn't your implicit formal model populated by representative agents of a sort?

      [like others on this thread, I think a good answer to this question might require more than a single representative household - a model that allows you to think about different kinds of household getting taxed and/or spent on]

  20. "I'd start by asking "who are you planning to tax, and what are you planning to spend on?" "

    + 1000

    The representative agent assumption has different weaknesses with respect to different questions but in this case, is it not a question about whether heterogeneous agents really will behave in the same way (and be affected the same way - given a more realistic scenario)? I.e. will they behave (and be affected) in the same way whether they are rich or poor, credit constrained or not, creditor or debtor, old or young, getting their income from work or capital gains etc. etc.

    I really do not see why the different agents behavior somehow would cancel out in the aggregate. I also really do not see why the study of a Robinson Crusoe economy is called "makro".

    PS: I recently had a colleague who concluded that income didn't seem to affect the demand for tobacco - which I guess is true when it comes to the top 1 % who acquired most of the resent gains in income - but I guess that you would get a different result if you looked at specific subgroups.

  21. PS 2: If you made a forecast of demand for tobacco from the above mentioned analysis, you would be right in that increased income did not affect the demand for tobacco as long as the income of the relatively poor did not increase - but this relationship would not be policy invariant.

    If there only existed a well known critique of these kind of analysis :)

  22. which heterodox texts have your read, professor? just blog posts?

    1. Lars Syll gave a list recently: Hyman Minsky, Michal Kalecki, Sidney Weintraub, Johan Åkerman, Gunnar Myrdal, Paul Davidson, Fred Lee, Axel Leijonhufvud, Steve Keen

      I cannot recall reading papers or texts by Akerman or Lee, but I have read at least something of all the others. I was also taught Neo-Ricardian economics when young, so I have read plenty by Robinson, Sraffa, Pasinetti etc. I do not know much about the Austrians.

      But actually my concern is not what any particular author thought, but with the divide I talked about. Mainstream economists can and in some cases have learnt a lot from some of these authors, and a number of mainstream economists have acknowledged this. (One of these days I want to write a paper arguing that Leijonhufvud was the first New Keynesian economist.) Anyone claiming that mainsteam analysis has nothing to learn from those labelled heterodox would be foolish, in my view. But the corresponding claim, that everything mainstream economists do is flawed because they assume superhuman agents, is just as wrong, and that is the point I'm trying to make.

    2. Leijonhufvud as NK ? I thought he was against the idea that sticky prices/wages was "what Keynes meant." Didnt he embrace the "Keynesian" idea of fundamental uncertainty, which surely has no flavour of NK about it. I suppose Roger Farmer has tried to run with that idea, but he is not NK either is he ?
      As for the original question, I guess the state of the art heterodox PK model now is that of Godley and Lavoie. Have you tried to use that ? I imagine the set-up costs may prevent many from doing it. It would be useful to see a suite of models trying to answer your qn including G&L

    3. Unless my memory is completely wrong, I remember his argument that the price that was basically wrong in a Keynesian regime was the real interest rate, rather than wages or the price level. I think that is very New Keynesian.

      ON G&L, I've only looked at some stuff, but would it be fair to characterise it as Old Keynesian behaviour with stock/flow consistency?

    4. Old Keynesian with stock/flow consistency sounds to me more like Tobin and Brainard and Backus than G&L. I wish some of the PK people would chime in.

    5. I don't think you'd get far with putting Leijonhufvud as NK. In a speech last September he said: "Economics has got things backward. We insist on forcing our subject into our preconceived methods - optimization and equilibrium. This creates an utterly distorted view of reality." So no NK there!
      Robert Clower said in an article in the 1990s that macro econ is trapped in a Walrasian vortex that prevents mainstream macro from being useful. He was arguing for agent-based modeling - different again from post-Keynesian.
      I do think you may be underestimating just how much of a crisis macro theory is in today. The macro of the future will surely have little contact with today's mainstream macro. Maybe some combination of network theory/ agent based modeling/ stock-flow consistency.

    6. On Leijonhufvud, Lars Syll also has a go at me about that passing remark. I think the problem is that my view of what New Keynesian economics is really about is not sticky wages and prices, or optimisation and representative agents, but about the consequences of having the wrong real rate of interest. (See this recent post of mine:

      In that respect I think Leijonhufvud's criticism of Keynesian (IS/LM) economics as being too obsessed with wage and price stickiness is correct. In addition, I also think his discussion of intertemporal coordination failures in which people’s expectations about long-term interest rates differ from the marginal efficiency of capital anticipate the key role of deviations from the natural rate of interest in NK analysis.

      So I think why my remarks may seem strange to you and Lars is that I have a different view to you (and probably a lot of NK economists) about what is really important in NK economics. I would also argue that my view is consistent with the writings of Michael Woodford, who wrote the book on NK economics, but I have no idea if he would agree.

      The problem with this kind of thing is that it is very easy to see what you want to see when reading others work. What is really important (to me, anyway) is the ideas, and not who was thinking what when.

  23. Dear Simon,

    I am starting my PhD at a heterodox school in the fall, and I received my master's in economics from another heterodox school (My macro prof was a student of Leijonhufvud) . Your balanced budget multiplier example's logic seems pretty tight. Maybe I don't understand the debate clearly enough, but the reaction by other heterodox folks about rational agents is bizarre.

    I was always taught that models were insight pumps or "gadgets" like Paul Krugman calls them. Moreover, I was taught that models are only as good as their empirical validity. Perhaps I had not received a "true" heterodox macro education during my MA, but the IS-LM and the WS-PS model a la Blanchard seem to be useful approximations. Plus, both the IS-LM and WS-PS are close to what Joan Robinson was thinking about with her inflation barrier to full-employment and Abba Lerner's low and high full employment. In fact, Abba Lerner's "Economics of Employment" more or less uses the IS-LM throughout.

    As a heterodox economist, I guess my approach is somewhat different. I'm not sure if you're familiar with Stephen Ziliak (also my prof) of the "Standard Error of Regressions" fame, but I favor his approach. I would rather criticize the empirical proof than the models themselves. Bad models, like Lucas's, have perpetuated because of statistical significance, and that is what we should all strive to change. And, I'm not interested in playing the victimn- I just want to find mechanisms that make the life's of every day people better!

    Plus, your reasoning and open-mindedness are what is important. You are a thoughtful and open individual that heterodox folks could use on the team!

    1. I'd add that although Simon is receiving lots of responses, I am not sure any of them are from actual heterodox economists.

    2. I'm an economist. Well, a Master student. Whether I qualify as heterodox, shrug. If "orthodox" means "using only equilibrium models and letting them define what you can study," then I'm heterodox. But that's an unkind demarcation of the central tradition, which most people working it would not accept.

      - Jake

  24. Let me admit I am side stepping your question, but I think it is the most honest response I can offer. As a heterodox economist reading comments and responses over the last few posts, I felt compelled to accept your challenge and play the game. I believe your posts, and engagement in the comments section, are completely sincere so I don't call it a "game" to denigrate your efforts, but the priors we take into this conversation prevent it from being a real exchange. What game are we playing? Based on what I see, I am supposed to give an account of the effects of fiscal policy without mentioning people. If I can do so successfully, I have won - a coherent heterodox economics without microfoundations is possible. If I slip up and mention people, I lose, suggesting microfoundations are inevitable.

    Because I like games, I wanted to play. I thought - Ok, I can do this! He won't find single mention of any type of person! Luckily I realized this game was stacked. It is absurd. I will lose. If you get me talking long enough I'll mention someone - BOOM microfoundations! But this is unfair, and obscures the actual methodological debates at stake. The issue is not whether people, in some sense, exist. The issue is not whether, or how often, we refer to individuals. The issue is how we think about them, and their relationship to the other complex processes that exist in the world.

    Structuralism in linguistics does not forbid discussing individual speech acts. "Structuralism" in social science does not forbid recognizing the existence of individuals. However, if you want to know how a child picks up a language from mess of sounds they are exposed to, or how individuals work to reproduce a messy economy, structuralism says you start with structures (broadly understood).

    I'm sure you would find it unfair if the roles were reversed, with heterodox economists suggesting that any reference to structures/institutions in your account of fiscal policy proves that structure-founded economics was inevitable. And you would be justified. Structuralism should not be reduced to "structures exists," as microfoundations should not be reduced to "people exist." All of these terms are loaded. What do we man by structures or people? What do we mean by exist? There are post-structuralist answers to these questions, but please note that the questions themselves are posed by every social ontology. For proponents of microfoundations, the individual has a ontological, epistemological, and/or methodological privilege over the structure. Their social existence is different than the way an institution exists. Structuralists, and a variety of post-structuralists, would disagree.

    I think you want to make the point that multiple microfoundations exist. That is fine. Heterodox economists who reduce the diversity of mainstream economics to a caricature are wrong and doing themselves a disservice. But just as the existence of multiple variants of fascism does little to placate anti-fascists, the relative flexibility of microfoundations is intellectually interesting, but not a compelling defense to those whose problem is "microfoundations" itself. Internet Miscommunication Warning: I am not suggested any link between fascism and microfoundations - just trying to make clear that the existence of different types of some thing is not a compelling defense when the issue is the thing itself.

  25. In my day job, I do non-macro but related research, but I have started dabbling in heterodox modelling, similar to the models in Godley & Lavoie's reprinted text. (Or Steve Keen's.) My comment echos those given above - the problem as you state largely assumes a mainstream DSGE-like economy. Thus, a DSGE-style model is best fitted to give an answer, by construction.

    - You do not specify how the government spends, only that its consumption implicitly can replace consumer sector consumption. This really only makes sense in a single-good economy. A stock-flow consistent model (SFC for short) aggregates spending by sectors, but the real economy effects are implicitly aggregated across varied goods. Changing the composition of spending will have distributional effects. (My last toy economy model is highly sensitive to a switch in the composition in spending, which cannot be captured in a single-good economy model, or the "continuum good" Calvo-style models, where the infinite number of goods are integrated down into a single good.)
    - You do not specify the structure of the tax, and it is unlikely to hit capitalists and workers equally. This is OK for a representative agent model where there is only one consumer, but won't work once we start splitting up the consumer sector into workers & capitalists (for example). The SFC models have parameters akin to a propensity to save for sectors (and subsectors), but they are meant to be averages. How the tax is structured will affect the observed impact on spending.
    - In a DSGE model, the fact that the tax is temporary fits directly into the model structure. Other than a few exceptions (e.g., corporate reactions to temporary changes to depreciation allowances), there's no compelling evidence that a tax being temporary has any impact on real world consumer behaviour, so nobody embeds that distinction into the model structure. (If you look at the U.S., there is a big political fight over "temporary" tax cuts ending. It seems naive to assume that saying a tax hike is temporary will be viewed as a credible commitment.)
    - The SFC models I have seen work on tax rates, not dollar tax amounts, so there is no easy way to determine what is a fiscally neutral change.
    - I seriously doubt that a currency level (price) would react to this sort of news (unless it was very large). However, the relative propensity to import will be affected by changing the mix in spending from consumer to government. The outcome thus depends on having information on the structure of imports.
    - I'm not an expert on DSGE modelling, but I've never seen anyone actually solve the nonlinear model with its fancy-pants microfoundations. They invariably wave the magic "log-linearisation" wand around the trivial solution with no dynamics. They end up with a linear model that can answer sensitivity questions like this. In all the SFC models I've played with, my objective was to start off at an unstable equilibrium and then end up with business cycle dynamics. The sensitivity of the model to an input depends on where you are in that trajectory.

    The summary answer: it depends. We need to have information akin to the "multipliers" to determine the response.

  26. Prof Simon, Your ask, “what . . . the impact of a temporary increase in government spending, financed by taxes, would do to output in an open economy.” That seems to be your central question. As self-appointed spokesman for the heterodox community, my answer is thus.

    If the objective is stimulus, what on earth is the point of raising taxes given that the effect of the latter is deflationary (or “anti-stimulatory”). I.e. why not just create new money and spend it (and/or cut taxes)?

    How’s that for an extreme heterodox view?

  27. No with all respect, Simon, i can´t really understand how Axel Leijonhufvud could be called a New Keynsian in either in his early years
    or in more recent work.Here is some from Axel well worth notice.
    Axel Leijonhufvud: Debt: Inflation and Austerity
    Ouside The Mainstream: An interview with AXEL
    Leijonhufvud, Axel The Wicksell Connection: Variation on a Theme.

  28. in Godley/Lavoie's SFC models, agents react to changed circumstances with partial adjustment functions. This is consistent with the G/L view that agents act under uncertainty know that they are acting under uncertainty.

    If such agents aim to consume a target proportion of their post-tax income, and post-tax income falls, they will adjust their consumption, but not by as much as the fall in post-tax income.

    So a one-off increase in tax to fund government consumption would cause the household sector to reduce its consumption, but by a smaller amount than the amount by which government consumption had increased: GDP would rise.

    So (in this very simple example) we've got representative agents in the household sector contributing to a rise in GDP precisely because they don't have perfect foresight, and know that they don't.

  29. Throughout human history, at least as long as there have been market oriented economies, there have been cycles of boom and bust corresponding with credit-fueled speculation into assets financial and otherwise. Without fail during these periods of time, large quantities of savings have been directed toward investment with little or no economic value (or at least, none where no such spigot is wide open), guaranteeing socially expensive losses. In practice, these episodes are of paramount importance to any understanding of the functioning of an economy, both because their answers hold out the promise of more enlightened policy regarding boom and bust, but also because these extremes should be revealing about the nature of things, like a particle accelerator is to physicists.

    Given that is the case, one might think it meaningful to a post such as this that these most pertinent and beguiling of economic phenomena are entirely inconceivable in a world where RE determines the expectations of representative agents. And yet, this is nowhere to be found in Professor Wren-Lewis’s case for “one of [mainstream economics’] greatest achievements” (granted, that could’ve been damning with faint praise). Something tells me that when Buitler talks about, “Research tended to be motivated by the internal logic, intellectual sunk capital and esthetic puzzles of established research programmes rather than by a powerful desire to understand how the economy works – let alone how the economy works during times of stress and financial instability”, exactly this is what he had in mind.

    Of course to the extent that they do, heterodox economists are wrong to focus on RE, in part for the issues raised here. How information is assimilated by the representative agents of represent agent models does not account for the catastrophic model failure of those models nearly so much as the (mis)specification wrought of the original abstraction. Put another way, there's a reason why hoocoodanode counts as sardonic humor.

  30. For what it’s worth, Keynes appears to pretty clearly reject the ‘rational expectations’ of ‘representative agents’ as a worthwhile exercise in his response to Viner’s criticism of The General Theory (as quoted by Hyman Minsky in his book, “John Maynard Keynes”):

    "By uncertain knowledge…, I do not mean merely to distinguish what is known for certain from what is only probable. The game of roulette is not subject, in this sense, to uncertainty; nor is the prospect of a Victory bond being drawn. Or again, the expectation of life is only slightly uncertain. Even the weather is only moderately uncertain. The sense in which I am using the term is that in which the prospect of a European war is uncertain, or the price of copper and the rate of interest twenty years hence, or the obsolescence of a new invention, or the position of private wealth owners in the social system in 1970. About these matters there is no scientific basis on which to form any calculable probability whatever. We simply do not know. Nevertheless, the necessity for action and for decision compels us as practical men to do our best to overlook this awkward fact and to behave exactly as we should if we had behind us a good Benthamite calculation of a series of prospective advantages and disadvantages, each multiplied by its appropriate probability waiting to be summed. ” [QJE, pp. 213-14]

    Thus the use of certainty equivalents,- much beloved by academics- is to practical men a convention, to which lip service may be paid, but which is abandoned when evidence inconsistent with the polite convention emerges.

    In the face of uncertainty and “the necessity for action and for decision” (QJE, p. 214), we devise conventions: we assume the present is a “serviceable guide to the future” we assume that existing market conditions are good guides to future markets and “we endeavor to conform with the behavior of the majority or the average” (QJE, p. 214). Given these flimsy foundations, the view of the future “is subject to sudden and violent changes” (QJE, pp.214-15). “All these pretty, polite techniques made for a well-paneled Board Room and a nicely regulated market, are liable to collapse” (QJE, p. 215)

    Also, in The General Theory, “Business men play a mixed game of skill and chance, the average results of which to the players are not known by those who take a hand” (GT, p.150)

    George Soros obviously got the memo Of course, he is no philosopher, and certainly no economist, but one should never underestimate the power of accountability to ground an intellectual exercise in reality.

  31. You ask "What happens if consumers are told there is a temporary tax increase ? " I don't know the answer to that but I do know what happened when US consumers were told that there was a temporary tax cut. They were told this in March 2009 when the ARRA (Obama stimulus) passed. It included a temporary roughly lump sum tax cut of $400 per year for individuals filing alone and $800 per year for couples, scheduled for 2 years (for employees in employment (wage and salary workers) the cut was applied monthly as a reduced whitholding from paychecks). Notably there were no Federal tax increases during those 2 years. Now one might suspect that the tax cut was not really going to be temporary -- indeed it was replaced with a partial payroll tax holliday for and additional 2 years). But I am going to address a much simpler question. During the 2 years, did US consumers believe (correctly) that the Federal taxes of most US consumers had been reduced. This question was asked in 2 polls. In one 12 % or respondents answered correctly. In the other 8% did.

    You think that it is sometimes useful to assume rational expectations when considering fiscal policy, because it is unreasonable to assume that people are almost totally clueless about what will happen in the future. This is very odd given the overwhelming evidence of almost total cluelessness abut what had happened in the past (and was continuing to happen every month, month after month).


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. In addition, I cannot publish comments with links to websites because it takes too much time to check whether these sites are legitimate.