Samuelson was a major force for writing down economic models using mathematics in North America. When academic economists today disagree, its primarily over empirical magnitudes of parameters in mathematically articulated economic models. He was right to be proud of his achievement.
Stephen Resnick tells the story of going in to Samuelson’s office when he was a grad student at MIT (this would be the early ’60s) and asking him if there was anything important in Marxism that you couldn’t talk about using conventional economics. Samuelson’s answer: “Class struggle.”
When academic economists today disagree, its primarily over empirical magnitudes of parameters in mathematically articulated economic models.
I venture to say this is not true of any important argument in economics, in any school. Mathematical models, sure, but the arguments are invariably over what model is appropriate, not parameter values.
Krugman, DeLong and their critics such as Mankiw are arguing over the effectiveness of the fiscal stimulus using the Keynesian macro model. In this model, fiscal policy is or is not effective depending on parameter values. But since credible estimates of the magnitudes of the parameters of interest are missing, the protagonists argue over models. I would put it as the model with their favorite parameter values.
As an aside, the Keynesian model was formalized by Hicks, a British economist. The NYT is misleading to give the intellectual credit to Samuelson. Samuleson popularize the Keynesian model in his principles textbook.
Showing mainstream economists that their conclusions do not follow from their assumptions seems to have no impact on the beliefs of many of such economists. Likewise, demonstrating that their ideas are logically inconsistent does not seem to have any impact. But, as a matter of norms, they seem to feel required to put their hazy notions in mathematics, without making them any less hazy.
I hope the obituarities in professional journals put as much emphasis on Samuelson’s debates with Joan Robinson as the New York Times does on his debates with Milton Friedman.
I hope the obituarities in professional journals put as much emphasis on Samuelson’s debates with Joan Robinson as the New York Times does on his debates with Milton Friedman.
Hell, I bet most economists in this country would ask “Joan Who?”
But still, I view Samuelson with a respect and kind of awe that does not pertain to many or any other American post-war economists. He was the very center, wasn’t he?
Just to clarify that it was Resnick’s co-author in those years, the late Stephen Hymer, who went into Samuelson’s office. Hymer told Resnick about this story while both were working on writing an article celebrating the work of Charles Kindleberger and they didn’t know how to start/develop it, so Hymer told Resnick this story as a way to get them started.
My understanding (I stand to be corrected of course) is that as the originator of the flawed Revealed Preference approach, Samuelson unfortunately did a lot to enable the hegemony of the loose coalition of neoclassicists that has been marching in carefully syncopated two-step ever since. It looks as though he clinched things when he fudged a face-saving dismissal of Wong’s Devastating Critique in 1950.
Disillusion with Rev Pref (along with among other things the use of Pareto criteria), when once it became clear that these weren’t retractible pedagogic ladders from A- to undergrad- level, was a significant factor in abandoning PPE in favour of philosophy. I could see it was a wrong ‘un, but the tangled web of neoclassical price theory etc makes a comprehensive criticism elusive. From what I can make out, Wong concentrates on taking the theory on in its own terms, and goes off on a needless excursion into philosophy (actually more like sociology) of science.
The last of the links here has a brief and partisan (but as far as I can make out accurate) summary. The promise of further clarification is still simmering on a mid-to-back burner, but may yet make an appearance here (try to contain your indifference, CTers!) if I can pare it down to decent proportions.
@8: Having looked at the first few quite unenlightening (to me anyway) pages of Mirowski’s “Wong’s Foundations after 25 years” (which is where your link leads), my preference would be for a short-ish and comprehensible (to a non-economist) explanation of all this, if such is possible.
I have always admired Paul Samuelson as a mensch. But for customer economics — the branch of econ in which I muck about — he was a scourge.
1. It in the 30s and 40s, he led the march away from heterogeneity as the starting point for thinking about how customers choose and how markets operate.
2. He turned away from the opportunity cost of customers’ money as a key determinant of how people make trade-offs in their purchases of differential value in products.
3. He legitimized what we have come to call the Samuelsonian Skipover: the handwave the lets you skip from the goodness a customer gets from a product to what that customer will pay for it, without giving a moment’s thought to how benefit actually converts to worth.
4. He helped legitimize thinking about the demand space within a market as one single slope instead of the polyscape that (in most cases) it is.
As a person, I’ve always held Paul Samuelson in the highest regard. But as a foundation layer for the branch of economics where I’ve worked for last 20 years, his mode of thought for thinking about the basics of customer choice and the structure of markets has not been a force for good. It held this sub-field back for something close to 60 years.
Richard Harmer @10: No 3 looks familiar – is that basically revealed preference theory? (Likewise 1?) FWIW I don’t really get point 2, but then I dropped economics so what do I expect.
LFC @9 – the Sen articles are worth a look – Rational Fools is quite a good overview IIRC (it’s hard, for me anyway, to disentangle individual assumptions from the overall project of neoclassical economics).
Anyway, here’s a very rough first attempt to articulate some problems I think I see in RP theory (reviewing it, I see it’s perhaps couched in somewhat intemperate terms) :
The idea of revealed preference was the rather suspicious ‘free lunch’ of providing a sound empirical basis for something like utility (the measure of subjective and thus objective value), without requiring any psychological investigation or indeed really anything but the projection of a vanishingly slim rationality assumption onto whatever people actually do. The pernicious effect of this unseemly haste to let all those equations get to work on us was really to make absolute economcis’s divorce from empirical reality – for little more than convenience, we are transformed for economic purposes into machines for achieving the intersections and tangents mapped out for us in the space of conveniently tangible and easy measurable goods within a specific set of property institutions (Excuse the slightly overblown turn of phrase – I’m just a bit fed up of underblowing it).
First, one of the main roles of RP is as input to a social welfare function, IIRC. But it is unable to fulfil that function, since it is supposedly based only on actual actions interpreted as choices, which radically underdetermine the structure of preferences. For example, the range of outcomes over which a person conceives themself as choosing cannot be determined; the single organising assumption of consistent preferences over time is unrealistic, and if it is relaxed then there is no organising principle for the interpretation of observable behaviour and the degree to which a person is successful in taking action to achieve their goal is not known. And none of these can be disentangled from each other in interpreting behaviour.
Second, it is illicitly restricted to market behaviour, thus ruling out a wide range of choices as impossible.
Third, the use of preferences (even if we knew which choices people conceived themselves as making, and had a stock of counterfactual worlds – not limited to those in which market behaviour is effectively enforced – to observe in constructing indifference curves) is only a suitable proxy for indvidual welfare given some very strong – and ideologically determined – assumptions about people’s ability to predict what is best for them, as well as the possibility of those preferences being suboptimal, prisoner’s dilemma style, given coordination problems.
Finally, as I believe Stanley Wong has pointed out to deafening silence, the single axiom of RP theory, its one a priori assumption – that preferences are consistent in the sense that the ‘…is preferred to…’ relation is assymmetric – is either useless (if it is supposed to be operative only at a point in time) or untrue (if it is taken to state that preferences are stable over time.) It has therefore been fudged and messed around with but AFAICT, basically retained when it comes down to it.
All of these have more-or-less ad hoc answers, some of which are arguably good enough, taken in isolation (economists are not idiots). For example, ‘it’s an operationalised proxy measure’ may be OK for some limited purposes – the approximation may be a reasonably close one in some circumscribed circumstances. And ‘we can observe behaviour of a range of people in a range of circumstances, and thus use revealed preference to derive standing assumptions for interpreting behaviour’ is perhaps true, but beside the point since the resulting construct is not revealed preference – probably more like a way of determining rankings of desiredness, and introduces psychology, avoiding which was supposed to be a major motivation for RP. And some answers, such as ‘we must use preference rather than welfare because the individual must be the arbiter of his own interests’ may have some plausibility, but also some implausibility. That example is hostage to some very complicated philosophical arguments which are by no means settled – it is a political, moral, normative position.
So taken together, this kind of problem undoubtedly amounts to a radical failure to connect with reality, despite the fact that the sole motivating factor behind the adoption of revealed preference as a welfare measure is that it alone is objectively determinable. And there is a good deal of ideology buried in there, too. Basically, the theory in which rational preference is embedded might seem rigorous – and even rigorously empirical – when the literature on individual topics is taken in isolation, but the disconnected refinements to (or utter transformations of) the theory’s elements have to be discarded when it’s time to join them back together to get actual normative results. (And normative results must be got – this is not airy-fairy stuff, but the whole point of economics – individual and/or social welfare.)
The point I have in mind here is that neoclassical economic theory taken as a whole is unavoidably normative, since the Invisible Hand style of argument remains the only significant argument for the free market. The attainment of Pareto optima (which by the way is even more obviously market-society bound, amounting as it does, near as dammit, to the statement that given the way things have gone, there aren’t any more trades to be made – but don’t get me started) is taken as a welfare measure capable of demonstrating the superiority of capitalism over other possibilties, but also more restrictedly as a best (‘efficient’) outcome in any situation.
This kind of issue has a strong Somebody Else’s Problem Field around it – to offer a nested quote from Olof Johansson-Stenman:
the link from fundamental ethics towards policy
recommendations is often very weak and that it depends on several very strong
assumptions. This is so even though the economics literature is becoming
fairly rich in questioning these restrictive assumptions separately. However,
in applied work it is unfortunately much more difficult to avoid these strong
assumptions. As expressed by Goodwin (1991: 286):
Hence we have a relatively unscientific applied welfare economics, and a scientific
theoretical welfare economics, with no better bridges between
the two today than was to be found at the beginning of the century in
Marshall’s system of inconsistent, or ambiguous, inclusiveness.
I am surprised that people describe Hicks as a “Keynesian”. Hicks was the first significant Neoclassicist, and Samuelson broadcasted the non-Keynesian Neoclassical economics, being largely responsible for its hegemony from the 1960s on.
{ 14 comments }
Canadian 12.13.09 at 7:05 pm
Samuelson was a major force for writing down economic models using mathematics in North America. When academic economists today disagree, its primarily over empirical magnitudes of parameters in mathematically articulated economic models. He was right to be proud of his achievement.
lemuel pitkin 12.13.09 at 8:07 pm
Stephen Resnick tells the story of going in to Samuelson’s office when he was a grad student at MIT (this would be the early ’60s) and asking him if there was anything important in Marxism that you couldn’t talk about using conventional economics. Samuelson’s answer: “Class struggle.”
When academic economists today disagree, its primarily over empirical magnitudes of parameters in mathematically articulated economic models.
I venture to say this is not true of any important argument in economics, in any school. Mathematical models, sure, but the arguments are invariably over what model is appropriate, not parameter values.
Canadian 12.13.09 at 9:26 pm
Krugman, DeLong and their critics such as Mankiw are arguing over the effectiveness of the fiscal stimulus using the Keynesian macro model. In this model, fiscal policy is or is not effective depending on parameter values. But since credible estimates of the magnitudes of the parameters of interest are missing, the protagonists argue over models. I would put it as the model with their favorite parameter values.
As an aside, the Keynesian model was formalized by Hicks, a British economist. The NYT is misleading to give the intellectual credit to Samuelson. Samuleson popularize the Keynesian model in his principles textbook.
Robert 12.13.09 at 11:00 pm
Showing mainstream economists that their conclusions do not follow from their assumptions seems to have no impact on the beliefs of many of such economists. Likewise, demonstrating that their ideas are logically inconsistent does not seem to have any impact. But, as a matter of norms, they seem to feel required to put their hazy notions in mathematics, without making them any less hazy.
I hope the obituarities in professional journals put as much emphasis on Samuelson’s debates with Joan Robinson as the New York Times does on his debates with Milton Friedman.
bob mcmanus 12.14.09 at 2:21 am
I hope the obituarities in professional journals put as much emphasis on Samuelson’s debates with Joan Robinson as the New York Times does on his debates with Milton Friedman.
Hell, I bet most economists in this country would ask “Joan Who?”
But still, I view Samuelson with a respect and kind of awe that does not pertain to many or any other American post-war economists. He was the very center, wasn’t he?
Ian J. Seda Irizarry 12.14.09 at 5:29 am
Just to clarify that it was Resnick’s co-author in those years, the late Stephen Hymer, who went into Samuelson’s office. Hymer told Resnick about this story while both were working on writing an article celebrating the work of Charles Kindleberger and they didn’t know how to start/develop it, so Hymer told Resnick this story as a way to get them started.
indian 12.14.09 at 9:28 am
end of an era ………………………… ECONOMICS will never be the same
Tim Wilkinson 12.14.09 at 3:05 pm
My understanding (I stand to be corrected of course) is that as the originator of the flawed Revealed Preference approach, Samuelson unfortunately did a lot to enable the hegemony of the loose coalition of neoclassicists that has been marching in carefully syncopated two-step ever since. It looks as though he clinched things when he fudged a face-saving dismissal of Wong’s Devastating Critique in 1950.
Disillusion with Rev Pref (along with among other things the use of Pareto criteria), when once it became clear that these weren’t retractible pedagogic ladders from A- to undergrad- level, was a significant factor in abandoning PPE in favour of philosophy. I could see it was a wrong ‘un, but the tangled web of neoclassical price theory etc makes a comprehensive criticism elusive. From what I can make out, Wong concentrates on taking the theory on in its own terms, and goes off on a needless excursion into philosophy (actually more like sociology) of science.
The last of the links here has a brief and partisan (but as far as I can make out accurate) summary. The promise of further clarification is still simmering on a mid-to-back burner, but may yet make an appearance here (try to contain your indifference, CTers!) if I can pare it down to decent proportions.
LFC 12.15.09 at 7:32 am
@8: Having looked at the first few quite unenlightening (to me anyway) pages of Mirowski’s “Wong’s Foundations after 25 years” (which is where your link leads), my preference would be for a short-ish and comprehensible (to a non-economist) explanation of all this, if such is possible.
Richard Harmer 12.15.09 at 12:13 pm
I have always admired Paul Samuelson as a mensch. But for customer economics — the branch of econ in which I muck about — he was a scourge.
1. It in the 30s and 40s, he led the march away from heterogeneity as the starting point for thinking about how customers choose and how markets operate.
2. He turned away from the opportunity cost of customers’ money as a key determinant of how people make trade-offs in their purchases of differential value in products.
3. He legitimized what we have come to call the Samuelsonian Skipover: the handwave the lets you skip from the goodness a customer gets from a product to what that customer will pay for it, without giving a moment’s thought to how benefit actually converts to worth.
4. He helped legitimize thinking about the demand space within a market as one single slope instead of the polyscape that (in most cases) it is.
As a person, I’ve always held Paul Samuelson in the highest regard. But as a foundation layer for the branch of economics where I’ve worked for last 20 years, his mode of thought for thinking about the basics of customer choice and the structure of markets has not been a force for good. It held this sub-field back for something close to 60 years.
Robert 12.15.09 at 3:15 pm
For what it’s worth, my name links to an attempt to summarize Wong’s book on Samuelson’s Revealed Preference theory.
LFC 12.15.09 at 6:22 pm
@11: thanks
Tim Wilkinson 12.15.09 at 6:35 pm
Richard Harmer @10: No 3 looks familiar – is that basically revealed preference theory? (Likewise 1?) FWIW I don’t really get point 2, but then I dropped economics so what do I expect.
LFC @9 – the Sen articles are worth a look – Rational Fools is quite a good overview IIRC (it’s hard, for me anyway, to disentangle individual assumptions from the overall project of neoclassical economics).
Anyway, here’s a very rough first attempt to articulate some problems I think I see in RP theory (reviewing it, I see it’s perhaps couched in somewhat intemperate terms) :
The idea of revealed preference was the rather suspicious ‘free lunch’ of providing a sound empirical basis for something like utility (the measure of subjective and thus objective value), without requiring any psychological investigation or indeed really anything but the projection of a vanishingly slim rationality assumption onto whatever people actually do. The pernicious effect of this unseemly haste to let all those equations get to work on us was really to make absolute economcis’s divorce from empirical reality – for little more than convenience, we are transformed for economic purposes into machines for achieving the intersections and tangents mapped out for us in the space of conveniently tangible and easy measurable goods within a specific set of property institutions (Excuse the slightly overblown turn of phrase – I’m just a bit fed up of underblowing it).
First, one of the main roles of RP is as input to a social welfare function, IIRC. But it is unable to fulfil that function, since it is supposedly based only on actual actions interpreted as choices, which radically underdetermine the structure of preferences. For example, the range of outcomes over which a person conceives themself as choosing cannot be determined; the single organising assumption of consistent preferences over time is unrealistic, and if it is relaxed then there is no organising principle for the interpretation of observable behaviour and the degree to which a person is successful in taking action to achieve their goal is not known. And none of these can be disentangled from each other in interpreting behaviour.
Second, it is illicitly restricted to market behaviour, thus ruling out a wide range of choices as impossible.
Third, the use of preferences (even if we knew which choices people conceived themselves as making, and had a stock of counterfactual worlds – not limited to those in which market behaviour is effectively enforced – to observe in constructing indifference curves) is only a suitable proxy for indvidual welfare given some very strong – and ideologically determined – assumptions about people’s ability to predict what is best for them, as well as the possibility of those preferences being suboptimal, prisoner’s dilemma style, given coordination problems.
Finally, as I believe Stanley Wong has pointed out to deafening silence, the single axiom of RP theory, its one a priori assumption – that preferences are consistent in the sense that the ‘…is preferred to…’ relation is assymmetric – is either useless (if it is supposed to be operative only at a point in time) or untrue (if it is taken to state that preferences are stable over time.) It has therefore been fudged and messed around with but AFAICT, basically retained when it comes down to it.
All of these have more-or-less ad hoc answers, some of which are arguably good enough, taken in isolation (economists are not idiots). For example, ‘it’s an operationalised proxy measure’ may be OK for some limited purposes – the approximation may be a reasonably close one in some circumscribed circumstances. And ‘we can observe behaviour of a range of people in a range of circumstances, and thus use revealed preference to derive standing assumptions for interpreting behaviour’ is perhaps true, but beside the point since the resulting construct is not revealed preference – probably more like a way of determining rankings of desiredness, and introduces psychology, avoiding which was supposed to be a major motivation for RP. And some answers, such as ‘we must use preference rather than welfare because the individual must be the arbiter of his own interests’ may have some plausibility, but also some implausibility. That example is hostage to some very complicated philosophical arguments which are by no means settled – it is a political, moral, normative position.
So taken together, this kind of problem undoubtedly amounts to a radical failure to connect with reality, despite the fact that the sole motivating factor behind the adoption of revealed preference as a welfare measure is that it alone is objectively determinable. And there is a good deal of ideology buried in there, too. Basically, the theory in which rational preference is embedded might seem rigorous – and even rigorously empirical – when the literature on individual topics is taken in isolation, but the disconnected refinements to (or utter transformations of) the theory’s elements have to be discarded when it’s time to join them back together to get actual normative results. (And normative results must be got – this is not airy-fairy stuff, but the whole point of economics – individual and/or social welfare.)
The point I have in mind here is that neoclassical economic theory taken as a whole is unavoidably normative, since the Invisible Hand style of argument remains the only significant argument for the free market. The attainment of Pareto optima (which by the way is even more obviously market-society bound, amounting as it does, near as dammit, to the statement that given the way things have gone, there aren’t any more trades to be made – but don’t get me started) is taken as a welfare measure capable of demonstrating the superiority of capitalism over other possibilties, but also more restrictedly as a best (‘efficient’) outcome in any situation.
This kind of issue has a strong Somebody Else’s Problem Field around it – to offer a nested quote from Olof Johansson-Stenman:
the link from fundamental ethics towards policy
recommendations is often very weak and that it depends on several very strong
assumptions. This is so even though the economics literature is becoming
fairly rich in questioning these restrictive assumptions separately. However,
in applied work it is unfortunately much more difficult to avoid these strong
assumptions. As expressed by Goodwin (1991: 286):
Ted 12.16.09 at 3:20 am
I am surprised that people describe Hicks as a “Keynesian”. Hicks was the first significant Neoclassicist, and Samuelson broadcasted the non-Keynesian Neoclassical economics, being largely responsible for its hegemony from the 1960s on.
Comments on this entry are closed.