<<

Robert M. Solow

How did get that way & what way did it get? Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021

My exposure to economics as a disci- ics, and I would like to provide a few pline began in September 1940 when I then-and-now snapshots. The differ- enrolled as a freshman in the elementa- ence, however, is that with economics ry economics course at Harvard College. something more is called for; the pic- I will try in this essay to make sense of tures have to be connected. I would like the evolution of economics over a span to tell a story about how and why the ar- of more than ½fty years. chitecture of economics changed. It will An analogy that comes to mind is be a sort of Whig history but without the from The Boston Globe. The Sunday edi- smugness. tion occasionally publishes pairs of pho- tographs of urban landscapes. They are There were three textbooks that were taken from the same spot, looking in used in the 1940 economics course at the same direction, but are at least thir- Harvard. One was a standard principles ty, forty, or ½fty years apart. One shows text by Frederic Garver and Alvin Han- a corner of the city as it looked then sen. Hansen had been at Minnesota with and the other as it looks now. Some Garver but by 1940 was a professor at buildings have disappeared, some new Harvard and–although we freshmen ones have been built, and some of the had no inkling–the leading ½gure in old ones are still there but with altered bringing the ideas of John Maynard facades. This description is also true of Keynes’s General Theory of Employment, the landscape and structure of econom- and Money1 into American eco- nomics. The second text was a large in- Robert M. Solow, Institute Professor Emeritus troductory book called Modern Economic and Professor of Economics Emeritus at the Mas- Society by ,2 also a mem- sachusetts Institute of Technology, has been a Fel- ber of the Harvard faculty and usually low of the American Academy since 1956. In 1987 referred to as the dean of American labor he received the in Economics. At the . The book was more about time of this essay’s publication in the Winter 1997 issue of “Dædalus,” Solow was Institute Professor Emeritus at the Massachusetts Institute of Tech- 1 , The General Theory of Employment, Interest and (London: Mac- nology. millan, 1936).

© 2005 by the American Academy of Arts 2 Sumner H. Slichter, Modern Economic Society & Sciences (New York: H. Holt and Company, 1935).

Dædalus Fall 2005 87 Robert M. economic institutions and their func- older books mention one equation, the Solow tioning than about theory. The third text Quantity Equation.) The numerical ex- was a little green volume by Luthringer, ample, hallowed in economics since the Chandler, and Cline about money and days of David Ricardo, is still in use, but banking, one of a series of little green it is no longer the analytical workhorse. books. (Lester Chandler of Princeton The older books are long on classi½ca- was the only one of the authors whose tions–kinds of , kinds of indus- name we ever heard again.) It was a pret- tries, kinds of labor–and on descrip- ty boring text, as I remember, but fortu- tions of public and private institutions. nately we only had to read bits of it. This The ½rst 260 pages in Slichter’s text are is actually an important point, and I will exclusively descriptive of the U.S. econ- Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 come back to it later. omy as it then was. I would guess that Even a quick physical comparison of fewer than one hundred of the next six a good contemporary elementary text hundred pages are devoted to the devel- with Garver and Hansen and Slichter opment of analysis or to the application tells us something. Leaving aside the of analysis. Most provide more institu- typographical changes–color, wider tional descriptions, very sensible discus- margins, larger type–the modern text sions of economic policy, and serious is sprinkled with diagrams, tables, even looks at recent history as it would be simple equations, whereas the older seen by an . No one should ones present page after page of unbro- underestimate the of these histor- ken prose. In some seven hundred ical reflections. They are, in a way, the pages, Garver and Hansen have fewer application of analytical ideas. But there than forty tables or ½gures. Some of is a not-so-subtle difference. The mod- them represent the working-out of nu- ern textbook presents and uses econom- merical examples of simple proposi- ic analysis as a tool to be directly applied tions, and the rest, maybe half, contain to contemporary or historical situations. data about the U.S. economy. Similarly, The student is shown how to map real there are ½fty-½ve graphs, again divided events into the categories that appear on between a small number of analytical the axes of the diagrams or the terms in diagrams and a larger number of graphi- the equations. The older texts are simp- cal presentations of actual data. Slichter ly more discursive. The underlying ideas is not radically different in his nine hun- are treated more like categories that res- dred pages. onate to this or that bit of history or pol- The modern counterpart, while no icy; the authors ruminate more than more intellectually demanding for the they analyze. student (perhaps even less so), is full of One sees this clearly in the way these diagrams, tables, and equations. The use two books present the idea of supply and of analytical diagrams is probably ten demand. This is the one piece of analy- times as intense, and the volume of real- sis that gets careful treatment. Charac- world data presented is correspondingly teristically, however, Garver and Hansen greater. Propositions are often stated in are very good on how one should think the form of equations, but these are al- about different kinds of commodities– most always simple statements (i.e., perishable or not, bought frequently or two intuitively understandable quanti- seldom, standardized or not–but the ties must be equal); there is not a lot of student is not encouraged to make liter- heavy mathematics in these texts. (The al use of the apparatus of supply and de-

88 Dædalus Fall 2005 mand curves. Both books spend time some of them because new facts and How did discussing monopolistic elements in institutions emerged, some of them for economics get that way real-world markets, but most of the internal intellectual reasons. Not many & what way discussion is institutional. There is, of sub½elds seem to have disappeared, did it get? course, no serious treatment of monop- though there was some rearrangement oly price because there was very little as a more uni½ed ab- known at the time. sorbed segments like “business cycles.” I do not want to be misunderstood. At the most general level, however, the Garver and Hansen and Slichter were change in tone was as I have described it. serious people. Their reflections on the Many outside observers and some crit- workings of the economy are worth ics from within the profession have in- Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 reading. They inspire bursts of nostal- terpreted this development as a sweep- gia; words like “civilized” came to mind. ing victory for “formalism” in econom- The point is that the modern text takes a ics. The intended implication is that eco- different approach. Of course it explains nomics has lost touch with everyday life, more; the intervening sixty years of eco- that it has become more self-involved nomic research have not been wasted. and less relevant to social concerns as it But it is the tone that I want to empha- became more formal (and more mathe- size. The modern text treats economics matical). I think that this view of the dis- as a collection of analytical tools to be cipline rests on a misconception about applied quite directly to observable sit- the change in the way mainstream econ- uations. omists go about their work. Barking may It is plain from this comparison that well be justi½ed, but not up the wrong there was a signi½cant change between tree. 1940 and 1990 in economics as a disci- If “formalist economics” means any- pline and also in the way it sees itself. thing, it must mean economic theory Perhaps this sea change deserves to be constructed more or less after the mo- called a transformation. One way to de- del of Euclid’s geometry. One starts with scribe it is to say that economics became a few axioms, as close to “self-evident” a self-consciously technical subject, no as they can be–although this is harder longer a ½t occupation for the gentle- to do when the subject matter is more man-scholar. And I mean that literally: complicated than points and lines in a nowadays economists arrive at their plane–and then tries to work out all conclusions by using an evolving collec- the logical implications of those axioms. tion of analytical techniques, most of Formalist economics starts with a small them nonintuitive, the sort that have number of assumptions about the be- to be learned laboriously. The shift of havior of individual economic agents, the center of gravity from Great Britain and a few more about their interactions to the (and to the G.I. Bill with each other, and goes on to study veterans at that) may have helped the what can then be said about the result- process along. Judicious discussion is ing . no longer the way serious economics The past ½fty years have indeed seen is carried out. Of course, that is not all formalist economics grow and prosper. that happened in ½fty years. A lot of new But it has not grown very much. Only knowledge was acquired, most of it by a small minority within the profession virtue of those analytical techniques. practices economic theory in this style. New branches of economics appeared, To tell the truth, not many more pay any

Dædalus Fall 2005 89 Robert M. attention at all to formalist theory. Gen- tion on the willingness to work. (God Solow erally speaking, formalists write for one knows that is a reasonable thing to be another. The formalist school contains interested in.) The usual approach goes some extraordinarily able people, and something like this: Imagine a typical of course it attracts economists who person of working age who enjoys both not only are talented at mathematics of consumer goods and leisure, and whose a certain kind but enjoy it. It is not sur- tastes for them can be described in a prising, therefore, that outsiders think simple and well-behaved way. This per- that there is a lot of formalism in eco- son has a certain amount of nonwage nomics, just as half a cup of blood spread income, from property or from transfer around a bathroom can make it look like payments of various kinds. He has the Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 a scene from Psycho. Nevertheless, it is option of working any number of hours an illusion. Modern mainstream eco- at a wage rate determined by the market. nomics is not all that formal. Part of his income is taxed away accord- ing to some known schedule. We have to What the outsider really sees is model- assume that this person does the best he building, which is an altogether different can to satisfy his tastes for leisure and sort of activity. In college classrooms in for the goods that his after-tax income the 1940s, whole semesters could go by can buy. without anyone talking about building We now ask the question that led to or testing a model. Today, if you ask a this model in the ½rst place. How will mainstream economist a question about he respond to higher tax rates–by work- almost any aspect of economic life, the ing more or fewer hours? If he makes response will be: suppose we model that no adjustment, he will have the same situation and see what happens. It is im- amount of leisure time but have fewer portant, then, to understand what a goods. That may suggest that he work model is and what it is not. longer hours, giving up some leisure A model is a deliberately simpli½ed time for more goods. With the higher representation of a much more compli- tax rates, however, each hour worked cated situation. (I have no reference for brings less in the way of goods, suggest- this, but I think I remember that the phi- ing that work has become less attractive. losopher J. L. Austin wrote somewhere He may choose, therefore, to work fewer that “one would be tempted to describe hours. It may make a difference wheth- oversimpli½cation as the occupational er the tax system imposes different rates disease of philosophers if it were not on wage and nonwage income. Perhaps their occupation.” Exactly.) it depends on the details of his prefer- The idea is to focus on one or two ences; not every person need react in causal or conditioning factors, exclude the same way. This model asks for some everything else, and hope to under- deeper analysis, which it gets. stand how just these aspects of reality Notice all the casual oversimpli½ca- work and interact. There are thousands tions. Not everyone can choose how of examples; the point is that modern many hours to work. People do not buy consists of lit- “consumer goods” in general; they buy tle else but examples of this process. hundreds of different things, some of What follows are three of them, de- which go particularly well with leisure. scribed in the sketchiest terms. Suppose Some people, but not others, have some we are interested in the effects of taxa- control over the intensity with which

90 Dædalus Fall 2005 they work. There are customs and norms reasons for holding and changing inven- How did economics that affect the behavior of different tories. And there are potentially impor- get that way groups. All of this sort of talk is cheap. tant conditioning factors that have been & what way The point of the exercise is to simplify completely left out: relations with sup- did it get? and see where it leads. Alternative sim- pliers and customers and ½nancial con- pli½cations are possible, and making straints, for instance. those choices is the art of the model- Modeling inventory fluctuations is builder. How do we judge success? It a matter of ½nding a way to represent is a good question, and I will return to some or all of these motives so that they it soon. can be weighed against one another in Here is a different type of example. much the same way that a pro½t-seek- Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 Anyone who has looked at the history ing ½rm will have to weigh them as it of business cycles knows that invest- decides what to do. Notice that last ment in inventories by businesses is month’s unintended inventory fluc- highly volatile and can easily account tuations will have an effect on this for most of the top-to-bottom change month’s plans, so that the behavior in production during a . It is to be described has a dynamics of its therefore a matter of some importance own. How do we judge success? Good that we understand the nature of inven- question, and I will come to it soon. tory fluctuations. There are plenty of Lastly, I give yet a third example reasons for ½rms to hold inventories because it illustrates a quite different and to change the amount of invento- point. Ten years ago, Elhanan Helpman ries they hold. Production schedules modeled a group of countries trading are ef½cient when they are smooth, but with one another under very special cir- sales can fluctuate unpredictably (or cumstances. Each country specialized predictably, as from season to season). completely in producing a single varie- Inventories of ½nished goods provide a ty of good. In the eyes of consumers, buffer, enabling ½rms to meet a fluctuat- each country’s “own” variety served ing demand with smooth production. as a symmetrically imperfect substitute Inventories of goods-in-process and, to for each other country’s variety. Con- a lesser extent, raw materials and com- sumers, however, all had the same set ponents may be tied fairly closely to cur- of tastes, no matter where they lived. rent production. Some ½rms build up in- Under these restrictive assumptions ventories in anticipation of future sales, and a few others, he showed that there or they may try to run their inventories would be a simple formula relating the down if they expect sales to be slack. In- volume of a country’s trade to its size. ventories of raw materials may provide In reality, countries do not specialize in a way to speculate on the prices of raw producing a single good, and consumers materials, buying more than needed do not have the same tastes wherever when the price is low and using up the they are. Nevertheless, Helpman’s for- surplus when the current price is high. mula seemed to work quite well for a Finally, ½rms may ½nd themselves with group of (i.e., advanced) coun- inventories that are lower or higher than tries. The moral might be that, in reali- they actually want: higher because sales ty, production patterns are a lot more have been disappointing, lower if sales specialized than tastes. have been unexpectedly strong. Even Recently, however, other economists this list is not a complete inventory of tried out the Helpman formula on a

Dædalus Fall 2005 91 Robert M. group of non-oecd countries, includ- details. There is something to be said for Solow ing some in Latin America and Africa. both. It seemed to work pretty well for them The interesting question is why eco- too. Paradoxically, perhaps that success nomics stopped being clubbable and be- casts some doubt on the Helpman mod- came technical sometime in the 1940s el: one would not expect the less ad- and 1950s, and why model-building took vanced countries to exhibit the same over as the standard intellectual exercise. specialization in production and com- I think one has to allow for the possibili- monality of tastes that is plausible for ty that it is, after all, the best way to do oecd countries. After all, there may economics, and we are just seeing the be quite different models that imply a survival of the ½ttest in another context. Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 similar relation between the size of a That would be Whig history with a ven- country and the volume of its trade. It geance. I confess to some sympathy with appears that measuring success may not that view, but only within limits. I would be a simple notion. add that the model-building approach is A good model makes the right strate- peculiarly vulnerable to unproductive gic simpli½cations. In fact, a really good controversy of a particular kind. I will model is one that generates a lot of un- discuss this later when I talk about meas- derstanding from focusing on a very uring success (one model at a time). small number of causal arrows. Model- I have a different hypothesis to sug- building is not a mechanical process. gest–that technique and model-build- Some people are better at this sort of ing came along with the expanding thing than others. Economic models availability of data, and each reinforces are usually stated mathematically, but the other. Each new piece of informa- they do not have to be. They can be de- tion about the economy, especially if scribed in words, as I have been doing, it is quantitative information, practi- or in diagrammatic form, or in comput- cally sits there and begs for explana- er flow charts for that matter. But math- tion. Someone will eventually be clever ematics turns out to be a very ef½cient enough to see that it is now feasible to way to express the structure of a sim- construct a model. Reciprocally, alterna- pli½ed model and it is, of course, a mar- tive models have to compete on some velous tool for discovering the implica- basis. They are not usually fancy enough tions of a particular model. That is prob- to compete on the basis of elegance or ably why outsiders tend to think of mod- depth or the intellectual equivalent of el-building as just more formalism. That pectoral development. They compete on is a mistake. The mere use of mathemat- the basis of their ability to give a satisfy- ics does not constitute formalism. May- ing account of some facts. Facts ask for be the sharpest way to make this point explanations, and explanations ask for is to say that the mathematics in these new facts. models is almost never deep. There are There is another partner in this evo- exceptions, of course. Nevertheless I lutionary spiral: the development of venture the estimate (safe because it is new methods of data analysis and statis- unveri½able) that there is little or no cor- tical inference. The other highly visible relation in fact between the dif½culty change in the style of academic econom- or mathematical depth of an economic ics since 1940 has been the explosion of model and its value as science. God is in from an esoteric minority the details, or perhaps in the absence of taste to an essential part of a Ph.D. edu-

92 Dædalus Fall 2005 cation–at least one chapter in most of el-builder himself, but he opened up a How did economics the dissertations produced in a major gold mine for those who came after. get that way department. I will say a little about this Suddenly there were models of aggre- & what way vertex of the triangle later. gate consumption and aggregate invest- did it get? The spread of model-building coin- ment, small but complete models of ag- cided in time with the development and gregate output and employment, and diffusion of . This data against which they could be tested was an accident, but an accident with and perhaps improved. Econometricians consequences: the heyday of Keynesian had new problems of statistical infer- economics provides a wonderful exam- ence to solve. And it all seemed so impor- ple of the interplay among theory, the tant. Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 availability of data, and the econometric The General Theory was and is a very method. The General Theory dates from dif½cult book to read. It contains sever- 1936; ’s book on national al distinct lines of thought that are never income accounting appeared in 1938.3 quite made mutually consistent. It was Both were no doubt related to the de- an extraordinarily influential book for pression of the 1930s, but that is just his- my generation of students (along with tory. The point is that Keynesian theory ’s Value and Capital4 and Paul needed the national income and product Samuelson’s Foundations of Economic accounts to make contact with reality, Analysis5), but we learned not as much and the availability of national income from it–it was, as I said, almost unread- and product accounts made Keynesian able–as from a number of explanatory macroeconomics fruitful (and helped to articles that appeared on all our gradu- shape it). ate-school reading lists. These articles When I mentioned at the very begin- reduced one or two of those trains of ning that my freshman textbook of thought to an intelligible model, which money and banking was a bore, this is for us became “Keynesian economics.” what I had in mind. It was the only ves- The most important of those articles tige of macroeconomics that we were were by John Hicks and Oskar Lange, taught–although the but there was a whole series of them, by rate in 1940 was still about 14 percent! Brian Reddaway, David Champernowne, –and it consisted of a few details about and others. This story provides a differ- the fractional-reserve banking system ent sort of illustration of the clarifying and the way it provides credit and gen- power of the model-building method. erates cash. There was such a thing as It is very likely that the war, as much business-cycle theory, but it was regard- as the depression, worked in the same ed as a sort of special topic. The text- direction. The panoply of wartime poli- book writers before 1940 had neither cy–Treasury ½nance, price and produc- the theory nor the data required to give tion controls, logistics of various kinds– a coherent account of macroeconomics involved economists in social engineer- as part of the core of the subject. Keynes more or less invented macro- 4 Sir John Richard Hicks, Value and Capital: An economics. He was not much of a mod- Inquiry into Some Fundamental Principles of Eco- nomic Theory (Oxford: Clarendon Press, 1939).

3 Simon Smith Kuznets, National Income and 5 Paul Anthony Samuelson, Foundations of , 1919–1935 (New York: Na- Economic Analysis (Cambridge, Mass.: Harvard tional Bureau of Economic Research, 1938). University Press, 1947).

Dædalus Fall 2005 93 Robert M. ing. Any routinization of policy, even ed in the original, narrowly focused Solow nonintrusive policy, leads inevitably to model. Trouble arises because data are technical questions. What will be the scarce in economics; more cannot be consequences if we do A? Vague gen- generated by experimentation. So there eralities will not do for an answer; de- is a danger of “over½tting”: adding vari- mands for quanti½cation are just around ables that work in the data at hand but the corner. Policy A can usually be un- will turn out to be irrelevant in the next dertaken with more or less intensity. batch, making them therefore deeply The powers that be will not only want irrelevant. to quantify the consequences but they This plays into the second and more will have the power to measure where fundamental problem. In the nature of Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 no one had measured before. Models the case it will often happen that two happen. quite different models can ½t the facts If they happen in connection with the just about equally as well. No doubt the availability of data, as I have suggested, right way to proceed is to think of cir- then success will be measured by the cumstances in which the two models ability to “explain” the data. Fitness give widely different predictions and is goodness of ½t. (I put “explain” in to look around for real-life situations quotes to emphasize that there need be that offer the opportunity to discrimi- no claim to fundamental explanation. nate between them. But that may not A model of inventory accumulation be possible. (Chemists can do experi- will likely eventuate in an equation that ments. There is a movement that does relates inventory spending to a small , but I cannot number of observable variables. If that guess how far it can go.) So naturally the equation actually holds to a fair degree temptation becomes irresistible to com- of approximation, the model explains pete by adding variables, making slight the data.) changes in formulation, looking around There is, however, a twist, and I think for especially favorable data, and other- it is important. If the logic of model- wise using the tricks of the trade. It can building, in economics anyway, is a become very dif½cult ever to displace drastic simpli½cation, then one cannot an entrenched model by a better one. expect any model to ½t the facts in every Clever and motivated–including ideo- detail. There are examples of models– logically motivated–people can ½ght and not only in economics–that have a rearguard battle that would make Rob- been judged to be very successful be- ert E. Lee look like an amateur. (And, of cause they manage to account for fair- course, they may turn out to be right.) ly gross, large-scale patterns that are ac- Old models never die; they just fade tually observed and measured. In prac- away. So the model-building approach tice, two consequences seem to follow. to economics has its problems. But it is The ½rst is a persistent temptation what we have: not formalism, and not to add explanatory variables in order to the more discursive approach that began improve the ½t. The variables do not fol- to break up in the 1940s and is now long low from the model, or else they would gone. already be there. But it is usually easy to As this description suggests, model- think of reasonable auxiliaries, things building economists tend to be natural- that “should” plausibly affect inventory born, loose-½tting positivists. Progress even if they were not includ- will come from weeding out empirically

94 Dædalus Fall 2005 unsuccessful models and improving and sion, and made the (somewhat) formal- How did economics extending those that survive empirical ist appeal that mainstream economists get that way tests. This is not to say that mainstream were in the position of teaching on Tues- & what way economists think explicitly about meth- day and Thursday a macroeconomics did it get? od. Philosophical tendencies may come that was fundamentally incompatible and go. They are attended to only by a with the taught on tiny fringe of economists who care about Monday and Wednesday. I do not think formal methodology. Their arguments that the appeal to “” make no dent in the mainstream, which amounted to much. Macroeconomic goes on making and testing models. hypotheses had always been justi½ed by It may be useful if I tuck in here a brief some sort of appeal to microeconomic Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 commentary on recent and current con- reasoning. troversy within academic macroeco- Nevertheless it was in part an appeal nomics, as seen from the point of view to formal criteria, and there was consid- advanced in this essay. (David Kreps’s erable force to this logic. But the “New discussion is similar, but not identical.) Classical Macroeconomics” then faced This was not part of my original plan, the problem of explaining, or explain- but the controversy is highly visible. ing away, the fluctuations in aggregate The conference held at the Huntington income and employment that consti- Library in March of 1995 seemed to have tute the everyday history of prosperity a lively interest in the details. Most in- and recession. And this task had to be triguingly, however, the controversy is performed within the framework of a often presented as a dispute between formal theory that seemed to exclude formalists and informalists. It would not even the possibility of the events to be damage the argument I have been mak- explained. I will not recount the ingen- ing if there were an element of truth in ious proposals that were invented to that characterization. Maybe there is, a perform this feat because they were ul- little. But in fact I think the story is ulti- timately felt to be implausible. No em- mately a strong con½rmation of the the- pirical successes were forthcoming. sis of this essay. This approach languished. In the mid-1970s, the standard text- The original New Classical Macroeco- book treatment of macroeconomics was nomics evolved into, or was superseded recognizably “Keynesian” or “American by, a related style of modeling called Keynesian.” It aimed speci½cally to pro- “Real Theory” (“real” vide an aggregative model of the whole means “nonmonetary”). And now we economy that could give some sort of cut to what I take to be the chase in this analytical account of unemployment, narrative. The goal of Real Business Cy- excess capacity, and recession (and cle Theory was the same: to show that their opposites) as pathologies of the the everyday experience of economic market economy. Opposition comes fluctuations could indeed be accounted from a school of thought that did in fact for within the framework of formal gen- invoke the equally standard formal theo- eral equilibrium theory, without the ry of a capitalist economy (the theory of “impure,” “ad hoc,” “Keynesian” viola- general competitive equilibrium). This tions of standard principles. In doing so school pointed out that this (microeco- it proceeded to abandon formalism in all nomic) model had no room for unem- but name by canonizing one very sim- ployment, excess capacity, and reces- ple, very special, and very maneuverable

Dædalus Fall 2005 95 Robert M. version of competitive general equilibri- amount to a signi½cant move toward Solow um–in fact, by adopting a highly speci- formalism. The new doctrine does try to ½c model. It is a model of an economy appropriate an air of “rigor”–a standard populated by a single immortal family ploy–but this is mostly advertising. with perfect foresight. The industrial and of the economy is My reading of the current state of af- such that it carries out, step by step, the fairs is much like Kreps’s. In the course in½nite-horizon optimal plan of the sin- of massaging the model to make it con- gle “representative consumer.” The gen- form to the facts, the more adventure- erality that is the hallmark of formalism some advocates of Real Business Cycle Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 is gone. Theory have found it necessary to mod- The gimmick is that the economy is ify many of the clean but extreme as- disturbed by irregular, unforeseeable sumptions that give formal general equi- changes in the preferences of the repre- librium theory its arti½cial vanilla flavor. sentative consumer and/or in the tech- As the representative-consumer-with- nology available to the industrial sec- perfect-foresight model has been ex- tor. Economic fluctuations are thus not tended to allow for elements of imper- pathological at all; they are the best that fect information, imperfect competi- can be done by way of adapting to these tion, imperfect flexibility of prices, in- pleasant and unpleasant surprises. The complete markets, and a teeny bit of het- model itself was already there to be erogeneity among the inhabitants, it has used. The bulk of the intellectual effort come closer and closer to the more or goes into the ways of showing that the less “Keynesian” model it was supposed data of observed fluctuations are com- to discredit. It is possible–though sure- patible with the demands of the model. ly not inevitable–that in another decade This is not easy because the key driving all that will be left are some purely tech- forces–irregular changes in tastes and nical differences in modeling strategies technology–are not directly observable. plus an underlying difference in spirit, So this is formalism, in more or less a with one side regarding all those imper- window-dressing sense. In practice it is fections as (removable?) flaws in the a little bit of model-building and a lot system and the other side regarding of fairly sophisticated data analysis. It is them as the essence of the system itself. not a revolution or transformation in the Seen this way, the macroeconomic way macroeconomics is done. controversy (made more intense by ide- To be sure, there has been a dramatic ological freight) only makes the model- change in doctrine. One genealogy of building tradition seem pretty irrevoca- models has replaced another. One set of ble. But then what about a historical implications has replaced another. This approach to economics? Is such a thing was a genuine shift of ideas, perhaps re- viable? The issue is worth discussing, lated to events of the 1970s that were, at and it will shed some further light on the least temporarily, hard to explain with main argument. In one sense, economics older models, perhaps related to the gen- is history. I have been insisting that the eral mood of conservatism and suspi- modern approach to economics is most- cion of government action that affected ly about accounting for data. It is hard economists as well as others. Such shifts to imagine where else data can come occur from time to time, in macroeco- from but the past. So economics is about nomics and elsewhere. This one did not accounting for the past. Most of the time

96 Dædalus Fall 2005 it is the recent past, but there is no rea- economic historians are very likely to be How did economics son why the more distant past cannot be model-builders. I have the disappointing get that way treated in the same way, if only the rele- impression that they are far too will- & what way vant data are available or can be recon- ing to accept the models devised by late did it get? structed. twentieth-century economists and ap- When I studied as a ply them uncritically to the data of oth- graduate student at the end of the 1940s, er times and places. There are certainly my teacher was A. P. Usher. I read his some sterling exceptions who do proper- History of Mechanical Inventions,6 Clap- ly exploit the advantages offered by ex- ham on British economic history, and otic data; my insider informant says sections of various works on monetary they are few and far between. There are Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 history. They were long on narrative and a few who use the study of the evolution short on analysis, a lot like the elemen- of institutions as a laboratory for eco- tary textbooks of a decade earlier. It did nomic analysis under unconventional not occur to me then, as it has since, that assumptions. There are even some who the more distant past provides some- continue to do narrative economic histo- thing potentially valuable to the model- ry. On the whole one has to report that building economist. A good model em- the historical approach to economics has bodies accurately a representation of the lost most of its distinctiveness and is los- institutions, norms, and attitudes that ing the rest. This seems to be a case of govern economic behavior in a particu- not being able to lick ’em, or not want- lar time and place. There is no reason ing to. to presuppose that a successful model The case is much the same with re- of the supply of labor in the second half spect to the other social sciences. I am of the twentieth century will apply un- tempted to guess that economics has changed to the nineteenth century when drawn further away from the other so- institutions, norms, and attitudes were cial sciences in the past half-century. different. Long runs of history offer the But the truth is that there was little or economist or historian or economic his- no interchange even in 1940 or 1950. torian the chance to ½gure out how Despite the existence of the occasion- changes in the “noneconomic” back- al sport like or George ground factors have an influence on Akerlof, who learned from the other behavior in the narrowly economic social sciences, most of the flow of ideas realm. It is a little like being able to ex- is in the other direction. There are sub- tend the range of temperatures or pres- cultures in political science and sociolo- sures available in a laboratory. gy that seem to want to adopt the meth- I am not sure that is how it has worked ods, the terminology, and sometimes in practice. One thing is certain: the the assumptions of economics. Those same progression from discursiveness Wahlverwandtschaften are best discussed to model-building has happened in eco- from within the other disciplines. Rich- nomic history as in economics. Econom- ard Swedberg’s Economics and Sociology is ic historians have become a lot more like full of interesting material,7 but I am less economists than economists have be- optimistic than he is about any system- come like economic historians. Today’s atic development. 7 Richard Swedberg, Economics and Sociology: 6 Abbott Payson Usher, A History of Mechanical Rede½ning Their Boundaries (Princeton, N.J.: Inventions (New York: McGraw-Hill, Inc., 1929). Press, 1990).

Dædalus Fall 2005 97 Robert M. Some sociologists and political scien- taught me that I should say explicitly Solow tists are drawn to the way economics that I have no neocolonialist designs: so- uses rationality–in effect, constrained ciology may be right to stay away from maximization–as an organizing princi- model-building as a mode of thought. ple and as a source of ideas for model- Adjacent territories may adopt different building. You could do a lot worse. But track gauges for good and suf½cient rea- there is an irony tucked away in that sons, but their railroads will have prob- remark. Some economists, though not lems at the border crossings. many, would like to look to sociology as It might be useful for me to say some a way of escaping from the narrow idea fairly informal things about the analo- of rationality. Actually, that way of put- gies between economics and the natu- Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 ting it is not quite right, so I shall try ral sciences. It is an uncomfortable task. again. The program of constrained max- I have read the usual quota of layman imization has to rest on a careful state- books and, after forty-seven years on ment of what is being maximized and the faculty at mit, I have a lot of friends what the constraints are. Mainstream who are physicists, chemists, and biolo- economics takes a narrow view of both; gists, not to mention engineers. But it is some hardy souls would like to try out a perfectly clear to me that I have no real wider range of assumptions. They look sense of what goes on in a physicist’s or to sociology and social psychology as a biologist’s mind. Still, it is a topic that source of alternative ideas. On the whole often comes up in cross-disciplinary dis- they do not ½nd what they are looking cussion. for, though again there are notable ex- There is no doubt that economists are ceptions. attracted to the style of explanation they This is not anybody’s fault. The writ- see (or think they see) in physics. This ings of people like Jon Elster, Mark Gra- is at least clear in the externals. Econo- novetter, Arthur Stinchcombe, and Aage mists feel at home with equilibrium Sørensen–just to take those who are conditions deduced from ½rst principles closest to the economists’ wavelength– or from reliable empirical statements. are full of interest to those all too few Similarly, they are used to deducing economists who read them. But they dynamics from local assumptions or do not provide the usable raw material generalizations; economics is full of (or intermediate product) that is being differential or ½nite-difference equa- sought. Even a book like Elster’s The Ce- tions. All this seems fairly harmless, ment of Society,8 intelligent as it is and as long as it works. It will occasionally on exactly the right subject, does not turn out that some piece of economics send an economist racing to the draw- is mathematically identical to some ing board. I suppose, though without piece of utterly unrelated physics. (This much con½dence, that this failure to has actually happened to me, although connect may arise because the other so- I know absolutely nothing about phys- cial sciences have not adopted the mod- ics.) I think this has no methodological el-building philosophy that motivates signi½cance but arises merely because and guides economists. Experience has everyone playing this sort of game tends to follow the line of least mathematical 8 Jon Elster, The Cement of Society: A Study of resistance. I know that Philip Mirowski Social Order (Cambridge: Cambridge University believes that deeper aspects of main- Press, 1989). stream economic theory are the product

98 Dædalus Fall 2005 of a profound imitation of nineteenth- rapid development of evolutionary game How did economics century physical theory. That thesis theory, there may be an opening for real get that way strikes me as false, but I would not claim progress. The loop closes, because what & what way expert knowledge. is now needed is a body of data to be ex- did it get? To the extent that economists have the ploited by the evolutionary game-mod- ambition to behave like physicists, they eler. face two dangerous pitfalls. The ½rst is the temptation to believe that the laws Nothing that has been said in this es- of economics are like the laws of phys- say is complicated enough to require ics: exactly the same everywhere on summary. Since part of my aim has been Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 earth and at every moment since Hector to dispel a misperception, I should con- was a pup. That is certainly true about clude by making another pass at explain- the behavior of heat and light. But the ing how the misperception has come to part of economics that is independent be so widely accepted. Many observers of history and social context is not only in the other social sciences and in the small but dull. wide, wide world perceive that econom- I want to suggest that a second pitfall ics has become formalistic, abstract, comes with the imitation of theoretical negligent of the real world. The truth physics: there is a tendency to underval- is, I think, that economics has become ue keen observation and shrewd gener- technical, which is quite different. (No- alization, virtues that I think are more body regards computer-aided tomogra- usually practiced by biologists. There phy as formalistic.) Far from being un- has long been a tendency in economics worldly, modern model-builders are to promote biology as an analogy. Even obsessed with data. a genuinely great man like Marshall took How could this confusion arise? I this line. Most of what is said on this have already suggested that it may be subject is a hopelessly vague use of unex- the trappings of mathematical model- amined analogy, uninformed by biologi- building that gives the wrong impres- cal theory. I am making a much weaker sion to outsiders. Now I will try out an- point, that there is a lot to be said in fa- other thought. There is a tendency for vor of staring at the piece of reality you theory to outrun data. (This includes are studying and asking, just what is go- statistical theory as well as economic ing on here? Economists who are enam- theory.) Theory is cheap, and data are ored of the physics style seem to bypass expensive. Much the same thing seems that stage, to their disadvantage. to happen in high-energy physics. I am There is another respect in which a told that the very latest ideas in particle broader biological analogy might be rel- theory could not come close to being evant. Many economists have noted that tested with any current accelerator or the evolutionary paradigm ought to be a even with the superconducting super- useful way of doing business. In isolated collider if it were to be built. No one instances it has already been valuable in even knows how enough energy could economics, but perhaps a little less so be mobilized to do the experiments that than might have been expected. I attrib- might con½rm today’s most advanced ute this to the absence of any close paral- speculations. lel to the quantitatively analyzable trans- In economics, model-builders’ busy- mission mechanism provided in biology work is to re½ne their ideas to ask ques- by population genetics. Now, with the tions to which the available data cannot

Dædalus Fall 2005 99 Robert M. give the answer. Econometric theorists Solow invent methods to estimate parameters about which the data have no informa- tion. And, of course, people are recruited whose talent is for just these activities, whose interest is more in method than in substance. As the models become more re½ned, the signal-to-noise ratio in the data becomes very attenuated. Since no empirical verdict is forthcom- ing, the student goes back to the draw- Downloaded from http://direct.mit.edu/daed/article-pdf/134/4/87/1829009/001152605774431518.pdf by guest on 03 October 2021 ing board–and re½nes the idea even more. Oscar Wilde described a fox hunt as the unspeakable in pursuit of the in- edible. Perhaps here we have the overed- ucated in pursuit of the unknowable. But it sure beats the alternatives.

100 Dædalus Fall 2005