Discussion of the fortnight: Knibbe, Vogel and Taylor on DSGE models

from Merijn Knibbe 

A  little while ago, I posted a ‘close reading’ of a DSGE article on this blog. It seemed right to notify the authors (see below), who sent me the next reaction (see below). I will comment on this reaction in a few days time (mainly focusing on the importance of knowing what statistical evidence is about). Until then, I will leave the floor to you – what’s the ontology of DSGE: economics, or economies? 

Dear mrs./ms. Mourougane,

herewith I draw your attention to:

1. A (quite critical) hearing of the USA congress on the use of DSGE models
2. A blog of mine on the misconceptions in Vogel and Mourougane, 2008.

You are invited to react on my comments on the work of you and Vogel: https://rwer.wordpress.com/2010/08/10/5-suggested-common-themes-for-an-economics-that-takes-its-subject-matter-seriously/#comment-2077

The reasons for my debunking of the work of you and Vogel:

A. Phlogiston, Ether, Utility. In my opinion, DSGE models take, in all their sillynomics, the science of economics back to an pre-scientific area of sloppy definitions, fuzzy mathematics and unobservable entities. I advise you to read the work of Victor Lamme, a neurologist who shows, based upon measurements, that people make choices first and convince themselves of the rationality of these choices afterwards. On consumer behavior, which outside of the economics departments has deveoped into a real science, you might consult Paul Peter, J. and J.C. Olson (2010), ‘Consumer behavior and marketing strategy’ (ninth edition), New York. Interstingly, the science of consumer behavior uses all kinds of scientific information and insights – but explicitely except the idea of rational choice and utility, as this idea ‘does not sell nylons’ (to misquote Mad Man’s Don Draper). It’s not up to the facts. It does not even exist. Take care, modern chemistry took hold when it abandoned the idea of (unobservable and, in the end, non-existing) Phogiston, modern physics took hold when it abandoned the idea of (unobservable and, in the end, non-existing) ether. The experiment that yielded the result that ether did not exist had as an completely unexpected side result that it showed that the spead of light was constant. And we all now were this finding led to.. Were will the experiments of Victor Lamme bring us? (the main book of Victor Lamme is still in Dutch, ‘De vrije wil bestaat niet’ (Amsterdam, 2010). Coincidentally but interestingly, this title can be translated in english as ‘There is no such thing as a free will’)?

B. Modern, scientific economics. Modern, scientific economics is grounded in facts – and has an important inductive side to it. When we look at what people actually do, instead of assuming that they do what is convenient for out mathematical modelling, it turns out that actual behavior (and even the essence of money) differs from our assumptions. It is only when our assumption simplify reality, instead of distorting it, that we will be able to construct usefull models which will give us at least some guidance. You (and all economists) should read: Collins e.a. (2009), Portfolios of the poor. How the world’s poor live on $2 a day’ , Princeton. This book shows that, If you don’t assume behavior but (just like modern consumer researh) actually look at what people really do, economics can be a real science, instead of a profitable mindgame. It shows that credit is not just a possibility in a monetary society, it shows (just like the historical record, you might consult sixteenth or seventeenth century probate inventories) that credit is a defining characteristic of money itself.

Economics can be a real science! 
Yours, 

Merijn Knibbe

Dear M. Knibbe,

I appreciate your interest and critical attitude. But I think you could afford a bit more modesty in the choice of words, reflecting that you might have misunderstood work, concepts and people you are criticising.

I like multidisciplinary work and outside-the-box thinking. I might even be more anarchist than you when it comes to concepts like science, truth and knowledge. Paul Feyerabend once observed that anarchists who reject any social or moral norm and authority tend to have a strange, (almost) fundamentalist believe in the authority of science, truth and knowledge. Such faith is not mine.

A few scattered, general and, maybe, slightly polemic remarks:

Most fundamentally we seem to disagree on epistemology, specifically on the purpose of (economic or whatever) models. I don’t see the advantage of large models. There is the famous Einstein word that a theory or model should be as simple as possible and as complicated as necessary. Now, what is necessary depends clearly on what your interest, your question is that your want to look at.

An economic model that includes all agents instead of making suitable aggregations is, from an analytical point, as useful or useless as a map of the world of scale 1:1.

I do not see how a big model necessarily leads to more realism. The more elements and agents you include, the more (ad hoc) assumptions you will have to include as well (about how the entities interact, where they are located, what they know about each other, in which order they make theirs decisions, whether they are permanently immobile or can move, that they always behave in similar ways or fully random) and the more information you need (which is sparse in real time).

Large models are not genuinely better at forecasting. The simple “model” that GDP growth in two years time will be the same as today is likely to beat most or all complex models. The large model built on history doesn’t tell you the future shocks, and these large estimated models are particularly bad at projecting turning points in the cycle. “Old-style” macro econometric models that are estimated on data from rather stable historical episodes are unlikely to forecast severe crises.

On a side note: Consumer or producer confidence data may have some predictive power. But then what determines or “explains” the confidence and changes thereof? As such they don’t provide a explanation for cycles and macro volatility. You would have to go further and explain the dynamics of confidence – as a consequence of erratic moods, habits, forward-looking behaviour, or whatever – and see how robust the explanation is across agents, countries and time.

A model, old or new, small or huge, will (and should) never fit “all facts”. It should deliver a good approximation to the facts and variables that are relevant in a certain context. No more and no less. A model of climate change does not have to replicate the interest rate on French government bonds.

Not surprising that models with little structure are more flexible and fit historical data better than models which (for other virtues) impose important restrictions. The closest approximation to an elephant is an elephant. But is it a useful model if one looks for a general theory of the evolution of mammals?

Now, the Lucas critique of traditional macro econometric models is exactly that such models have or impose too little structure. You estimate a model on past data which were generated in a specific historical and institutional context. Now this context changes (e.g. fall of the iron curtain, euro introduction, deregulation). Are your parameter estimates still reliable in the “new world”? Are they still informative enough to guide policy makers? Which are, which are not? Does stronger competition between producers change the impact of wage setting or oil prices on Dutch CPI inflation? Does the euro introduction affect the size of fiscal multipliers? The traditional estimates won’t tell you. A more structural model, however, may give some hint. The DSGE approach tries to distinguish between variables and parameters that are likely to change (e.g. the conduct of monetary policy with euro introduction, the labour supply and wage pressure in the West after the fall of communism, labour demand in the context of technological progress and automation) and those that are likely to be more stable (degrees of risk aversion, expectation formation). The DSGE model may be full of simplifications, but it gives you an intuition how macro economic relationships will change if the global, political, institutional environment changes: Is the Phillips curve becoming flatter? May exports become more price elastic? Is employment likely to become more volatile? And the structural approach allows analysing counterfactuals of the “what if…” type for which there are no data because they are counterfactuals.

On the US Congress hearing, you may have liked the Solow piece, but I recommend reading the Chari piece as well. Chari makes pretty clear that much DSGE-bashing grounds in misperceptions about the approach, or, if you prefer, diverging semantics.

Solow gives the impression that DSGE is synonym for the simplest, most stylised dynamic neoclassical model. Chari illustrates that DSGE is an approach, not a specific toy model. It is a concept of modelling. It has the notable advantage of incorporating the dynamics that is inherent in many kinds of human behaviour (planning, expectations, experience, adaptation, risk aversion) and the interdependence of decisions of households, firms, governments and countries (not least via theirs budget constraints). It illustrates feedback effects that may weaken or amplify the initial impact of private decisions and government policies.

A DSGE model can be as complex as you wish. You can build a DSGE with 80 million German households, if you want (and have computer power to solve it). You can model households as rational, schizophrenic or psychopaths, if you believe this is superior. The DSGE approach is in principle very flexible. You just have to specify the behavioural pattern or decision rules for your agents.

(People may make choices and then build up an ex post justification for the sake of cognitive consonance. But this does not mean that people chose out of nothing, without preferences, without experience, and that they don’t learn from obvious mistakes. The person that had her flat burgled repeatedly may think about changing neighbourhood instead of making up a story to convince herself that there is something nice about being burgled. Less drastically this may also apply to investment decisions and other choices in the economic sphere.)

If you want to explain why A drinks Heineken, B Leffe and C Paulaner, the representative household is a bad modelling device and you may resort to marketing theories. But maybe the representative household assumption does quite well in replicating the impact on average consumption of a VAT increase?

The analytical purpose of your specific model should guide your modelling choice. DSGE models can be very complex. They can easily have 1000s of equations and e.g. include:

– Skill differentiation and labour market segmentation (if you are looking at skill-specific employment, wages and inequality, where the simplistic “representative household” is clearly not enough),
– Many different but interlinked countries (to think about contagion),
– Many sectors (if you want to look e.g. at the impact and efficiency of carbon taxes, specific subsidies and the like),
– Many features (contract duration, indexation schemes, information lags, copying peer behaviour…) that add substantial persistence and lag length to the model to obtain more realistic short-term dynamics.

It is also incorrect to claim that the modelling approach as such requires the economy to end up in equilibrium. The approach actually allows you to pin down the conditions under which variables either return to a stable path or do not. In fact, important parts of the DSGE literature on monetary policy investigate the requirements for stability as to avoid, e.g., hyperinflation or deflation. There is also no problem with including, e.g. unemployment hysteresis (due to skill-loss among unemployed) which makes the equilibrium history-dependent.

If stability is a feature of most estimated or calibrated DSGE models, it is so just because the data suggest that typical OECD economies are rather stable. Hyperinflations in the Netherlands are rare nowadays and, to my knowledge, a contractions of Dutch GDP to zero did never happen. The majority of DSGE models focuses on rich industrialised countries, be it for the reason of data availability or narrow-mindedness, or both. But availability and reliability of data are as much a problem for “traditional” estimated macro models.

I am aware of the Arrow theorem and the impossibility of a unique preference aggregation (unless all agents have equal preference and endowments). But this does not preclude deriving decision rules (for certain groups) from (group-specific) preferences. The impossibility of having a “correct”, unique aggregation of preferences is a problem for normative economics and economic policy. It’s ok as long as you are explicit about the welfare function that your normative analysis is conditioned on. In practice it is a matter of political or social choice. Should the government give equal weight to all households, or prioritise the impact on the poorest, or predominantly try to avoid big losses, or focus on containing inequality?

Don’t forget that the national accounts are an artefact at least as much as economic models. There might be “households” in the classification that don’t fit the narrow picture. But, e.g., institutional households are normally part of government in the models. Shall we model words and definitions or some empirical phenomena?

Think of government spending on education and health care. Probably and hopefully it has some positive impact on a society’s literacy, health status, skill levels, labour productivity etc. Therefore, DSGE modellers prefer to treat it (in a sufficiently complex model) as some form of public investment (“human capital”) or technology shifter. National accounts classify education and health spending as (rather wasteful) government consumption, while at the same time they count each useless bridge and highway as (valuable) investment in infrastructure. Which classification is less “fuzzy and sloppy”? Following your remark, the national account definition must be “the truth”. Not to mention that the current accounts get revised after quarters and years, and from time to time redesigned according to new (unifying) standards. Do the revisions change economic history or just our picture thereof?

As your remarks are about the DSGE approach as such, less about a specific paper, I leave it here.

Best regards,

Lukas Vogel

 

from Merijn Knibbe

Some days ago I posted a blog containing an answer of Lucas Vogel on comments of me about an DSGE article baring his name (see above). I promised an answer. This is it. For the time being, I’ve made my points and will not bother you again.

Dear Lucas,
Thank you for your response to my comments. They confirm my fears. We should not be commenting to much on details (that reminds me too much of Goethe: “intelligent people are smartest when they are wrong”) but I will respond to the ‘between the lines’ information:
1. Introduction
A. Data are important. Though I do admit that we still have a lot of work to do, I do not share your somewhat cynical attitude towards data. A nice example of the importance of inductive analysis of the data is, on this blog, https://rwer.wordpress.com/2009/12/10/4-years-of-calling-the-gfc/ in which Keen describes how he only became (rightly) alarmist on debt after examining data on debt (“my jaw fell to the ground”, debt as a percentage of GDP did grow exponentionally for decades). In most sciences, students spend much (if not most) of their time on how to gather data. In most sciences, data gathering is glorious. and discovering new facts which change our view of the world might earn you a Noble price. Not so in economics, though the tide seems to be turning a little, albeit not yet in the curricula I know about. Many economists are, as scientists, at a grave disadvatage as they were not tought how to gather data and often do not ‘grasp’ the data. An example of how this influences economics as a science: the 1947 – 2008 USA data on quarterly GDP and productivity are often used for analysis. Superficial investigation of these data however yields that between 1947 and 2008, fourth quarter productivity increased only half as much as in the other quarters. Futher investigation yields that this happened especially between 1965 – 1980 and was located on the GDP side of the (GDP/labor) equation. What happened? I haven’t the slightest idea – at present it’s an anomaly. Why has nobody noticed this before? The answer is simple: economists are not trained to look at raw data. Economists should (like the biologists and agricultural engineers with whom I work together) be trained to draw explicit attention to such anomalies. My point: if we do not teach our students to look at the data in various ways, including inductive ways, and to scrutinize the quality of the data (including extensive discussions on how data are gathered) and to compare the definitions of measured variables with the theoretical definitions, we jeopardize economics as a science. Without data, however, we are no science.

B. An extension of this point: let’s write an article together. You will be familiar with the (PT = MV) formula of macro economics. Economists however measure V by (V = (GDP)/M) (Abel, Bernanke, Croushore, 2011(?!), do neo-classicals indeed have a problem with time?), p. 259). But as GDP is not PT but: (PT – PaTa), the last term being the value of intermediate inputs, it is more precise to use the Gross Production series of the National Accounts instead of the Value Added series to calculate V. You will sense my hidden agendas: National Accounts data are important to economists and to be a good economist you need to know about definitions. When writing the article, we will have to think hard on M: which data do we use: January? December? A monthly average? An average based upon december (t-1) – november (t) or december (t-1) – december (t)? I suggest sending the article to Abel e.a. (see below). To be clear: the real reason to do this exercise is of course that a more precise measurement might give better insight into the economy, thoug to gain this insight we also will have to modify the definition of M2 (‘credit rights’ instead of actual credit on checking accounts and credit cards should be counted as part of M2, in my view)
Abel, A.B, B.S. Bernanke and D. Croushore (2011), ‘Macroeconomics’, seventh edition, Boston.
C. To simplify the discussion: I’m not against mathematical models: simple when possible, complicated if that’s needed. I’m against models who do not only simplify, but who also do distort. Maps can be extremely abstract – and still in certain circumstances guide you in a life saving way. If maps give a distorted view of reality, however…. This discussion is not about models. The discussion is about our models reflect reality or not: using outdated maps can be life threatening – a very real option to somebody like me (see C.). You might consult the (somewhat tedious) discussions about this on this blog.

D. There is a slight suggestion that I’m some kind of anarchist. I’m not. It might be worse to you, but I’m a law abiding father of three, dedicated teacher and certified outdoor guide, wo loves to go on extended night hikes with his oldest . I’m driving a 1.2 Fiat Doblo. As a teacher, I try to teach people a craft instead of ‘giving them the opportunity to develop their own unique perception of reality and to develop their own system of values’. No post modernism scepticism about data for me: reality is not just in the eye of the beholder. You will understand that I, as a neo-romantic ‘burgher’, don’t share your (slightly post modern) disgust of data. Much more so than models, data are often ‘larger’ than the individual researcher, often showing us that our own ‘unique perception of reality’is, well, dead wrong. Of course I’m critical about the quality of data – after all, I’m a scientist. But (and may be there indeed is some anarchism in this) I do trust good data much more than the opinion of fellow economists. For some short but devastating comments on post-modernist academics as well as to gain some insight in the dynamic side of households you might read (though De Vries neglects the fact that much present day consumption is highly dependend on public investements like bridges etcetera):
De Vries, J. (2008), ‘The industrious revolution. Consumer behavior and the household economy, 1650 to the present’, Cambridge.
For an example of an economist who is often dead right, just because he takes a staight forward (but informed, he knows about statistics) look at the data, see the columns of Dean Baker.
2. Comments.
Methodological remark: I’ve ordered my remarks according to ‘five suggested common theme’s’, which can be found on this blog, to investigate if this is a usefull way of approaching modern, scientific economics.
1. Economic phenomena are social phenomena
You use the concept of utility. People do try to measure utility, thoug they do not succeed – they do not measure utility, they measure choices (Abdellaoui, Ettema and Bleichrodt, 2009). When one reads articles on such attempts, it strikes the mind that these experiments force (small numbers of) people to act like atomistic choice machines – but even then, te outcomes defy the standard utility model you use. When one reads such articles, his jaw also repetedly falls to his chest. A much more interesting real life experiment is the zipp code lotery: people who live in a winning street, but who did not buy a ticket, also do change their expenditure behaviour, in line with the winners. Same income, different expenditures. Preferences can only be understood in a social context. And people are not rational in the economic sense. Real life indifference curves just do not exist – they are the jar with gold at the end of the rainbow. Real life social behavior does exist – it is the rainbow. I, for one, am not searching for the jar with gold. I’m marvelling at the rainbow.
Abdellaoui, M., A.E. Ettema, H. Bleichrodt (2009), ‘Intertemporal Tradoffs for Gains and losses: an experimental measurement of discouted utility’. Rotterdam. http://www.omroepbrabant.nl/?news/140746932/Verliezers+Postcodeloterij+kopen+toch+nieuwe+auto.aspx
2. Evidence should be paramount
You seem to be sceptical of the measurement of economic phenomena. When one uses, like you do, a model based on: (households maximize utility from here to eternity and have perfect foresight and knowledge) this is understandable. However. We do have an awfull lot of information on economic phenomena – and you might try to increase your grasp of the why, how and when of this information: scepticism as a way out of this tedious endeavour is too cheap. An example: you seem to want to ‘explain’ consumer confidence (or do you mean: to put it into the neoclassical mold?). Quite some economists have tried to explain consumer confidence, and this is interesting and important (though te explanations do not fit into the (max (U) mold). But this statistic was not developed to give academics the opportunity to write articles about it. It was developed as a kind of McStatistics. A standard criticism of Keynesian politics is that data on the business cycle are always lagging, while it also takes politicians time to act on it and even more time to implement new policies. To adress this problem, fast ‘short’ indicators like consumer confidence have been developed. And this one is fast! To quote the conference-board (July 2010):
“The Conference Board Consumer Confidence Index® which had declined sharply in June, retreated further in July”.
Or, to quote Bloomberg (July 2010):
“July 30 (Bloomberg) — Europeans’ confidence in the outlook for the economy dropped the most since the Sept. 11 terrorist attacks as soaring energy costs and the euro’s advance against the dollar rattled consumers and executives”
These data are even faster then the data on M1 and M2 (which became available yesterday)! We do know, to an extent, why these confidence variables change and that’s indeed interesting– but the explaining variables are only known with a lag of several months. In my opinion, this does address part of the criticism on Keynesian politics (though I do agree with you when you would state that governments are not always too rational). By the way: these data do, like the data on M1 and M2, point to a double dip.
About two months after these very fast indicators, other ‘short’ indicators like those on international trade, production in manufacturing and consumption are gathered, which give further information and which are matched with each other in the quarterly national accounts which (necessarily) are available only later. And indeed, these are revised when more information becomes available – that’s the name of the scientific game. But does that mean that we should not gather them anymore? As far as I know, the only way to measure the amount and impact of increased government spending in 2008 and 2009 is by constructing national accounts – we might need this information
You also mention that National Accounts do not count government consumption on education etc. as investment. You might want to consult Bos, 2003, 6.6, ‘Modules on social policy, human capital and for olicy analysis’.
Bos, F. (2003), ‘The National Accounts as a tool for analysis and policy; past present and future’, Berkel en Rodenrijs.
Knibbe, M. (1988), ‘Indices of Cyclical Indicators’, interne nota 272-88-KI-E8, CBS, Voorburg, 1988.
3. Accepting the complexity of economic phenomena
Economists love markets. I do. Markets played a very, very large role in economic progress since the industrial revolution. To give a ‘Schumpeterian’ example – my cheap nine year old car ‘embeds’ technology which, about twentyfive years ago, was cutting edge in an expensive new Mercedes. That’s the capitalist market! However.You might not be aware of the ‘Komlos-paradox’. This paradox states that, when isolated, non-monetary come into contact with the global market economy, this indeed leads to specialization and trade. But it also leads to higher mortality and morbidity and smaller, less well fed children as people trade away precious food for highly regarded utensils and gadgets. This happened, for instance, in parts of the Scotland of Adam Smith, where people traded oats and fish for whiskey and cottons. And this is just one of many instances. A somewhat comparable process might be happening today – it’s not too difficult to find alarming reports on day care centers for small children which are explicit on this connection: “With the need for women to get involved in economic enhancement of the family and with career mothers in the increase in our society, there has been increase in the need to have day care centers to keep the little baby while both parents go to work. Often children come home from some of the Day Care centers with episodes of diarrhea, gastroenteritis and skin diseases. A number of these centers are established in our country without any supervision as to standard of the environment.” (Olaitan and Adeleke, 2007).
And yes, that’s economics too. Just consult the work of Noble prize winner Fogel. According to him, the biological standard of living only started to improve above pre-industrial levels after 1895. According to my work, this might (might!) before about 1925 have been the result of higher real wages for men, enabling women to give more breastfeeding and other aspects of care, and not of investments in sewer systems and the like. Economic phenomena are complex. The relation between government spending on health and human capital is complex too. We will have to measure it.
Olaitan, J. and O. Adeleke: Bacteria in Day Care Environment. The Internet Journal of Microbiology. 2007 Volume 3 Number 1.
Knibbe, M., ‘De hoofdelijke beschikbaarheid van voedsel en de levensstandaard in Nederland, 1807 – 1913’, Tijdschrift voor Sociale en Economische Geschiedenis 4-4 (2007) 108-135.

Gibson, A.J.S. and T.C. Smout (1995), ‘Prices, food and wages in Scotland 1550 – 1780’, Cambridge.

4. Rejecting analytic formalism where this distorts good analysis
I told you that your assumption on consumers without access to credit is false – credit is an essential property of money. In my view, this means that you’re essentially writing science fiction, not science.
5. Recognising the need for clusters of related models of many kinds and levels
See 3. Reality has many different sides – but these are connected. The famous analogy of the 6 blind men touching different parts of an elephant, non of them recognizing the species as an elephant, does in fact not hold up in reality. One of my colleageas – a biologists – told me that in a German Zoo where he worked a blind employee could distinguish the feathers of over a hundred species of birds. We do have to get in touch with reality. We have to use different methods to do this. But we also have to understand about the connections between these methods.

An anarchist statement at the end: Bernanke has to know better.
In Abel e. a. (2011, p. 239), a (very common) myth about money is retold: “having a medium of exchange also rises productivity by allowing people to specialize … in a barter economy … the difficulty of trading would leave people no choice but to produce most of their own food, clothing and shelter”. In the real world, barter is at best a side show in non-monetary economies – the comparison is void. And we’ve already seen about the Komlos paradox – even if specialiation results in an increase in productivity, it might deteriorate peoples lifes. In my view, myths as told by Abel e.a. should be replaced by real world stories (Of course, I do have to admit that myths can be tempting. Sometimes, I also fantasize about the famous economists myth of the ‘coconut island’, living on an island with another or, sometimes, even two others. My fantasy is however not about coconuts). Diamond, D. W. and P. H. Dybvig (1983), Bank runs, deposit insurance, and liquidity, Journal of Political Economy, 91(3), 401-419.

Summarizing: economists often don’t know. That’s science. But economists sometimes try to fill this void with myths, disgust of data, incoherent use of data and weird assumptions. That’s not science. That’s very human, understandable but, in the end, irrational behaviour.

 

from Dave Taylor

Lucas Vogel says “There is the famous Einstein word that a theory or model should be as simple as possible and as complicated as necessary. Now, what is necessary depends clearly on what your interest, your question is that you want to look at.

“An economic model that includes all agents instead of making suitable aggregations is, from an analytical point, as useful or useless as a map of the world of scale 1:1.”

I agree. So let’s compare a couple of models which include neither aggregations nor scale. The world being like a football rather than a river tells us an awful lot about the mathematics needed to locate objects on a map of it, and the non-straight-line rivers on it being both static and continuously in motion tells us a lot about the relationship between continued existence and recycling.

Let’s also compare models of logic. Deductive logic proves nothing: it merely demonstrates consistency between axioms and conclusions, the whole process being linguistic. (Whitehead and Russell discovered a century ago that it is incapable, without suitable axioms, of explaining even arithmetic, never mind Pythagorean right angles, Cartesian coordinates and Hamiltonian rotations). The inductive logic of Bacon (taking real things to bits to see how they work, as against Hume’s statistical elimination of differences in reported observations) acquired a different interpetation (half-anticipated by deductivist Karl Popper) from C E Shannon and N Wiener. Where Bacon’s applied scientist eliminated arrangements that he could see didn’t work, Shannon’s error-correcting logic and Wiener’s cybernetics act on what deduction prescribes, but to sufficiently achieve it rely on continuous elimination of observed errors (both predictable and unpredictable) via feedback from real outcomes. Likewise, I Lakatos would have had us abandon mainstream economics as a research program on the ground that the numerous errors in its foundations were now manifest, but still neither corrected nor correctible.

 

from Merijn Knibbe 

Dear Dave,

to avoid misunderstandings:

A. It’s not about complex or simple models. I did not attack the article of Vogel and Mouroukane on their use of simple or complicated models and I agree with the Einstein-quote.

An example: the crudest of all models enables me to make a sure bet on the prediction that the temperature in Amsterdam, January 23, 2011, 14.27 (GMT) will be lower than the temperature in Amsterdam, July 23, 2011, 14.27 (GMT). The most complex of all models (or about the most complex of all models, I’m not sure about this) does not enable the weather forecasters to make a sure bet on the prediction that the temperature in Amsterdam, September 14, 2010, 14.27 (GMT) will be higher or lower than the temperature in Amsterdam, September 4, 2010, 14.27 (GMT). By the way: I love simple models and concepts – these can give tremendous insights: Dean Baker justs looks at the ‘basket’ used to calculate the consumer price index – and draws the ex-tre-me-ly important conclusion (until some time ago neglected by almost all other economists, who were talking about the ‘service economy’) that the housing market is the largest of all markets. Love you too, Dean.

B. It is about models which do or do not distort reality. To use the (well known) map analogy introduced into this discussion by Vogel: when I drive from the Netherlands to southern France, I might use a 1: 2.000.000 map. Whenever I’m cycling in France, the 1:200.000 Michelin-maps do fine. When I’m hiking in Les Ardennes Francaises, I use the 1:25.000 maps. My outdoor activities in the Waddensea area require me to use an official topographic map as well as a kind of ‘song lines’ mind map based on recent experience with the ever changing shallows. All these maps give a non-distorted view of reality and enable me to find my way. But I never use a map which is drawn by somebody who did not bother to investigate the area, just imagines cities and roads and assumes away mountains and shallows. That’s distorting reality. That’s what DSGE models do – as they use the “max (U) from here to eternity concept”.

C. There is no such thing as ‘Utility’. DSGE models use the concept of social indifference curves, which is based upon maximizing individuals as well as upon the concept of Utility. That’s bad. There is no clear definition of ‘Utility’, even neo-classical economists who do try to measure individual ‘indifference curves’ do not bother to define what they are measuring, though the titles of their articles suggest they do (at least, I haven’t encountered attempts to do this). The neo classical textbooks I consulted on this just introduce the concept without discussing it. This lack of attention to the very concept of ‘Utility’ has an obvious reason (one can consult: Andersen, H.C. (1837), ‘The emperor’s new suit’, 1872 translation of the original Danish Version, http://hca.gilead.org.il/emperor.html on this). Utility is badly defined (if at all), can not be measured in any direct way, is not used by scientists doing research on the behavior of real consumers while modern neurological research (as well as common sense) indicates that preferences are not stable, often follow choices instead of preceding them and are often contradictory. Despite this unscientific mess, DSGE economists even use ‘social indifference curves’, which add up utility of individuals and, as Vogel readily admits, require the additional assumption that everybody is exactly equal, i.e. has the same preferences. Kornai has compared this kind of thinking with Stalinist thinking (Kornai, J. (1971), ‘Anti-equilibrium. On Economic Systems theory and the tasks of research. Amsterdam): one size fits all. And this is not just theory!

D. Using ‘social indifference curves’ is ‘outright dangerous’ – an empirical example how using ‘social indifference curves’ can harm society.

The ‘one size fits all’ approach has recently been used by the Centraal Planbureau (CPB) of the Netherlands, and economic think tank. In their analysis of the housing market they used the ‘representative consumer’. with a Cobb-Douglas Indifference Curve (why Cobb-Douglas? No empirical reasons are given). As everybody is equal, 300,– spent on a Gucci Gadget by a spoiled boy is as ‘useful’ to society as 300,– spent on rent for a family of five – the representative consumer knows everything and is perfectly rational. Or, in this case, 300,– spent on interest on a mortgage is as useful as 300,– spent on rent. As rich people generally own houses, have mortgages and (surprise!) spent more on housing than poor people, this CPB-study boils down to the conclusion that lower income people have to pay more for smaller houses as this enables the well to do to occupy larger dwellings (in the process assuming that space can be transferred from the houses of people with lower incomes to the houses of the well to do….). However. Not everybody can get a mortgage and buy a house. Houses differ. People differ. Regions differ. Ages differ. Family situations differ. Space can not be transferred from small houses to large houses… I’ve read quite some studies on the Dutch housing market. Many of these are written by people with a background in engineering, or architecture. The economics of these reports (according to the ‘literature’ parts of their reports routinely ignored by economists writing on the Dutch housing market …) are often more refined (i.e. comparing different approaches, having better databases) than those of economists. At this moment, one of the conclusions of the CPB model (increase real rents with 60% during the next decades) has become part of the program of one of the mayor political parties in The Netherlands – while costs of housing in The Netherlands are already the highest of Europe and historically have never been higher than in 2009 (well, maybe in 2010).

CPB (2010), ‘hervorming van het Nederlandse woonbeleid’, Den Haag.

G. Romijn, ‘Regulering en subsidiering verlagen de huren, maar ontwrichten de woningmarkt’ in: CPB-nieuwsbrief juni 2008 pp. 3-4.