There has always been an intense debate in economics concerning its nature (deductive or inductive, "soft" or "hard" science…) and the type of the laws its method could uncover. Beneath there is also a deeper debate on the type of discipline that economics should try to emulate: mathematics, logics, physics, biology or history, even philosophy (see Kolm, 1986). Most economists would agree that the discipline should be a nomological science — one that produces universal laws — but there is less consensus regarding the nature of those laws and how they relate to reality (Lawson, 1997, 2003). There is actually less consensus than expected even among so-called mainstream economists. Some consider indeed that economics is just a branch of (applied, but here also a debate is possible) mathematics (see Rosenberg, 1992; Debreu, 1991). The real economies are so complicated that only a Platonician approach with deductions from axioms of individual rationality could be used to establish simple laws relating economic variables ceteris paribus (Albert, 1970). These laws are time-invariant and valid under a very precise set of assumptions (the latter could be rather unrealistic if the derived predictions make sense, following Friedman's point of view, see Friedman, 1953). We find here among economists two different views regarding the empirical relevance of such deduced laws: for a part of the profession such laws are purely logical ones, with no necessary direct empirical measurable counterparts (Debreu in a sense defends such a position; it is also the case of people working on social choices or economic justice, aiming at a so-called normative science) but for another part of the profession such derived laws should be testable (i.e. the implied variables should have empirical, measurable counterparts, see for example Beckerian's economics). The difference here is between a pure axiomatic approach (Stigum, 1990) that views economics as logics or mathematics or even political philosophy (not an empirical science) and a more realist point of view (the deductive approach should tell us things that otherwise would have remained hidden, but that are still measurable or testable); in other words between a view of economics as "pure theory " [1] and of economics as an applied science (even if, as physics, grounded on a consistent mathematical theory). These laws that can be deduced from a body of axioms should be tested once the context has been checked by using measurement (the quantitative aspect—using econometrics).
The time-invariant nature of such economist's laws imply that these laws are also a-historical (i.e. valid for all periods, all stages of economic development, and under all institutional settings; see Hirshleifer stating that economics is the universal grammar of the social sciences). Other economists (Marx, 1867 comes to mind, for example), also consider that there are laws in economics but that the latter are only valid in a given institutional framework (e.g. capitalism). The study of institutional change becomes also a key question for the economist, who should also be a producer of laws concerning institutional change (in short advancing a theory of history and moving from a philosophy of history to a "science of history"). The study of history would play an important role in the determination of such laws concerning institutional change. The deductive approach may play a certain role (Marx for example uses a lot of tools from classical economists as Ricardo, 1817), but a degree of induction is also unavoidable (deriving from the observation of the past some key assumptions on the mechanics of economic development). The importance of historical analysis and the downplay of deductive approach could be very important for some authors (at the limit we find the classical literary economic history, "observation without theory, or a very loose one" to paraphrase Clive Granger). The two German historical schools is somewhere in-between. It is not entirely a-theoretical, because it searches for economic laws (compatible with the observed evolutions, discovered through a careful historical analysis), but the latter are mainly inductively developed (and generally of a rather macroeconomic nature; see for example the law of Wagner concerning public spending in relation with economic development). At the end of the nineteenth century, the Methodenstreit consecrated the defeat of this current of thinking and the triumph of the neoclassical vision (that which believes in laws of universal value and independent of the institutional context). Traditional economic history—outside the field of economics—continued inductive work and institutional analysis without great theoretical ambitions (although work on the new history with Braudel went beyond a simple desire to make a description). Some economists after WW2, thanks to the revival of the holistic (aggregate) perspective of the Keynesian macroeconomics, pursue works in a similar vain, using more sophisticated tools (see below). The Phillips curve or the Kuznets law (1955) relating inequalities and economic development are just a few examples.
THE ROLE OF CLIOMETRICS IN ECONOMIC THEORY
The awarding of the Nobel Prize for Economics to Douglass North and Robert Fogel in 1993 seems to have consecrated this discipline as an entirely accepted subfield of economics (developed in the USA from the 1950s to the 1970s). Once modernised by the systematic use of economic theory and econometric techniques, (new) economic history can become a fully accepted branch of economics with the same status (hopefully) as other fields. A priori quantitative economic history belongs to the ‘applied' sector of our discipline (which was for a long time not the most prestigious). Economics is indeed generally separated into a theoretical part (today, this means the development of mathematical models from rationality assumptions intended to find an answer to a fairly generic type of question that is always taken out of context; but economic theory used to be expressed in a more literary language in the past) and an applied part (in theory, if one could dare to say so, it should consist in testing in the Popperian sense — falsifying — statements whose logical consistency has been checked by theoreticians, soon considered as economist-mathematicians). If cliometrics was only a field of applied economics, it would not have much too much to say in the debate examined here (the debate on the existence of laws in economics).
Theory has more prestige than the empirical fields in our discipline. Young doctoral students are often asked to demonstrate their skills and virtuosity by developing models requiring advanced mathematics. Data gathering and the testing of theoretical models using econometric techniques do not have the same prestige, probably because they are not considered as demanding in terms of creativeness and innovation as the theoretical analysis (the more so that econometric software are now widely available to do the dirty job). More fundamentally, there is perhaps also the fact that economists "do not do what they say they do". The Popperian approach may not really be theirs, whatever Blaug wrote on the subject. Theoretical discussions definitely settled by econometricians are fairly rare and it is to be feared that if one should apply the Popperian methodology to the full (rejecting a theory whenever we find only just one sample or one sub-period for which it seems to be rejected), nothing would remain of economic theory [2]. In any case, since the Lucas critique (Lucas, 1976), economists have been extremely sceptical regarding the ability of (macro-) econometrics to settle a theoretical debate. The macro-models did not explicit the link between micro and macro-behaviour; they indeed assumed fixed coefficients, i.e. they assumed passive agents, copying the engineering model, and did not allow for changes in behaviour linked with changes in the economic policy. We could also speak of other problems as aggregation bias, data quality and the small time horizon of econometric tests. It has also been contended that theoreticians and "empiricists" are actually interested in different things. Much theoretical work in the 1990s (for example) concerned the theory of industrial organisation (principal-agent models, information theory, incentives, etc.) or endogenous growth, whereas applied economics (or econometrics) was devoted to the study of other phenomena (exchange rates, finance, etc.). On top of that, econometrics seems to increasingly depart from its initial vocation as a field closely related to the testing of economic theories and to develop as a separate branch, closer to statistical theory than to economics per se. The success of time series analysis –causality, cointegration, VAR models and so on– tends to confirm the view that the figures will tell their own story; the success of such methods is indeed linked with the underlying idea that everything is in everything in economics and that it is difficult to distinguish truly exogenous variables as soon as the real world is addressed. Some economists go so far as talking of measurement without theory ("or a very loose one" as Granger said, see Darnell, 1990) since the triumph of the econometrics of time series at the expense of structural modelling. The defenders of such an approach contend that it is a more successful one when dealing with real-world (business or policy) issues. It should also be added that the work of certain applied economists is not so much designed as the test of economic theories and laws (which would involve close relations between theoreticians and empiricists—in fact, mathematicians consider the separation between pure and applied mathematics to be largely artificial) but rather as the use of statistico-econometric tools and economic data for decision making, whether public (see the use by large administrations of macroeconomic models that are still made up of an impure mixture of classical theories and more Keynesian ingredients) or private (all the work on the econometrics of exchange rates and financial series—even if in the latter field the link with theory is doubtless closer (see below)).
It is clear that from a strictly empirical point of view the quality of our data is not that of our colleagues in the exact sciences in controlled laboratory experiments, that economic laws tend to be probabilistic rather than deterministic, that these relations are interlinked within a more overarching system of extreme complexity (hence the appearance from time to time of physicists in our discipline who contend that our modelling approach is both too simplistic and too old, based upon ideas of equilibrium that belong to a physics that is too Newtonian); that agents are not passive like units of matter (atoms) but are definitively intelligent beings that anticipate and analyse information (the theory of rational anticipation has been a response to these objections, putting the theoretical legitimacy of the large Keynesian macroeconomic models in difficulty, although these are still used for practical reasons), that culture is involved (the interaction between rational individuals can give rise to the emergence of structures such as institutions or organisations — see North, for example — that in turn will constitute a constraint upon individual behaviour) and that there is a historicity dimension (or path dependence as economists like Paul David say. See David, 1997). It is interesting to note here that all these criticisms stem either from a historian's approach (historical critique to question data origin and quality), from economists who first received training in history (such as Lucas with regard to rational anticipation) or from economist/historians like North (the role of institutions or culture) or David (a specialist in research on R&D). All these features enable us to glimpse a greater degree of complexity than that assumed by Keynesian theory (copying the engineering model—see Philips' hydraulic machine in the cellars at LSE). In the face of this increased perceived complexity, one tends to abandon structural modelling in econometrics for an approach leaving increasing scope to the idea of "making the data speak for themselves" (even if historians admit that the facts never tell their own story) using techniques that are more based on the statistical properties of the time series than on an economic theory — perhaps viewed as insufficient or too simplistic (this is above all true in applied macroeconomics). We have nevertheless to point out the development of experimental economics that explicitly aims at testing propositions from microeconomic theory.
If on the one hand certain economists tend to privilegiate the analysis of the properties of long run economic time series (hoping that they will reveal some unknown truth), other economists tend to retreat from the empirical fields to favour the pure world of abstraction. Instead of confronting the complexity and uncertainty of econometric tests, such economists tend to favour the certitude of logical analysis. This is not the posture of the least and most obscure of the profession (Debreu, 1991). They think that economics is not mainly an empirical science (and so the laws produced by economists would be closer to mathematical theories or the results produced by logicians; Kolm even contend that economics is in a sense a mathematised field of political philosophy, i.e. a normative and not a positive discipline). They feel that economics should concentrate on deducing certain propositions through the established procedures of mathematical logics. When they have to choose between saying too much, at the price of accepting certain doubtful ideas or theories as true and that of saying too little, or concentrating on a segment that is too narrow but certain, economists (faithful to Wittgenstein in this matter) do not hesitate and choose the second alternative (analogy with first and second type risks in statistics). In the debate between truth and precision (see Mayer, 1993) they always choose precision. The main objective pursued by economists following this line of thought is to build a logical, a priori non-empirical science consisting in the analysis of all the possible consequences of a rationality principle (maximisation of personal interest under constraints). In the same way that geometricians specify a series of axioms (one and only one straight line runs between two points, two parallel straight lines join at infinity, etc. to limit the approach to Euclidian geometry) and then use mathematical logic to deduce all the possible consequences, economists would do the same thing using rationality axioms. This axiomatic procedure (defended since the 1950s but tending to gain importance, see Stigum, 1990) gambles that in the very long term this non-empirical approach will provide us with tools that will asymptotically lead us towards a more fundamental understanding of economic phenomena. It is a somewhat Platonic vision that considers that the switch to mathematical idealities and the approach of a mathematician (but not that of an engineer, like the approach that presided over Keynesian economics) will give us the best chance to say something serious about the working of an economy. The role of empiricism is here secondary — at most — and quite logically comparatively depreciated. The world constructed by pure theory [3] is often just a sketch of the real-world, an idealisation, and very often is even not a representation of the real-world but an abstract representation of an idealised world where it is possible to carry out thought experiments ( Fables ). A number of the concepts developed here have simply no empirical counterparts. Game theory can be viewed as a typical field for such abstract exercises, as used to be the case of the theory of general equilibrium. Even if one departs from the view of economics as a normative science, using those tools from game theory and industrial economics in more realist models where imperfect competition and increasing returns occur often leads to multiple multiple equilibria and are no longer useful in providing us with certainties regarding the working of economies. Theory then tends to become a succession of small, somewhat ad hoc models.
WHAT ROLE FOR CLIOMETRICS?
To come back to Cliometrics, is it a discipline that would help us in closing the gap between theory and empirical analysis in economics? For many authors (and many of its protagonists) cliometrics appears to be first of all a branch of history, using economic tools and theories to provide answers to historiographical debates and not so much to economic debates? Might therefore this new discipline make it possible to discover laws in economics, as in physics?
The meaning of the word "empirical" for (American) economic historians has varied with the passing of time. One can observe a shift from a concept of empirical fact as understood by the classical historian (for whom anything can be used in his demonstration and not only—initially not at all—quantitative phenomena) to one as understood by (applied) economists (the empirical aspect consists of numerical time series) and a convergence of theoretical viewpoints of historians and economists thanks to a common interest in building theories of development. Here, Kuznets seems to play a key role by emphasising the importance of macroeconomic analysis based on the major quantitative changes before possibly identifying certain sectors that are central for economic development. One should note that even in this view to combine history with economic analysis, the economics remains founded inductively on the observation of the major stylised events enlightened by the work of economic historians. This view is therefore intimately linked with the historicism current in economics (The German Historical School, see Schmoller, 1904, Hodgson, 2003) despite of the use of more sophisticated techniques. It could be said that the two disciplines are becoming closer, but probably within the frame of ‘inductive' economics, that is today, traditional economic history. So cliometrics seeks to provide answers to historiographical questions (for example, the role of the railways in American economic history or the profitability of slavery). Econometric techniques may be used, with the reconstitution of time series and identification of missing figures by interpolation or extrapolation—something that annoys professional historians—but these cliometric procedures have nonetheless a historical vocation, that of shedding light on historical questions considering economic theory or econometrics as auxiliary disciplines of history. And when cliometrics is used to serve as a basis for the building of a development theory based upon clearly measured facts, this is induction. The conflict between Kuznets and Rostow regarding stages in economic development actually concerns the empirical foundations of Rostow's theory and not at all a defect in the theoretical background (no micro-foundations, for example, and no use of growth theories, which would doubtless be the subject of the bulk of criticism today). In short, either cliometrics is still a (modernised) branch of (economic) history—in the same way as the modernisation of methods in archaeology, from carbon-14 dating to the use of statistical techniques such as discriminant analysis, does not turn the discipline into a branch of natural science—or cliometrics is used for obtaining results that more or less share this vision of a science that must be founded on facts and result from the generalisation made using an accumulation of empirical proofs that all go in the same direction. Here it contributes to an economic science that is related to the German historical school.
But this is not the end of the story. Some recent work in cliometrics performed by economists ( stricto sensu ) reveals the possibility of a cliometrics that could also be an auxiliary discipline of economics and as such should form part of the competences of all economists. However, it cannot exist as itself here unless it is slightly outside the field of economics in the strict sense. It must be a compound of the application of the newest econometric techniques and economic theory with the old institutional and factual culture characteristics of old economic history. History is indeed always a discipline of synthesis. It should also be the case for cliometrics. If not, this research would seem to be "institution-free" to a certain degree and it would be difficult to see the contribution of these retrospective exercises in econometrics (besides the longer time span on which econometric tools are applied). In a way this would mean — a conventional belief of economists (in fact, that of Lord Kelvin) — that "qualitative is poor quantitative". But might not "quantitative is also poor qualitative" sometimes be true? A feature that traces a clear border between economists and historians is the sense of historical criticism and the desire to avoid any anachronism. In addition to close examination of sources, this involves the close examination of the institutional context that forms the framework of the players' behaviour and the taking into account of the economic context. It is true that (new) economic history will not construct a general theory (it shares too strongly the belief in the necessity of examining economic phenomena in their context) but it could suggest a few useful lines of thinking and insights to economists who are seeking to draw up laws (unlike history, economics is still a nomological science). This is a view shared among others by Acemoglu and Robinson (2004), trying to use the material derived from traditional history to build new ideas useful for economic theorists.
When it stops defending its specificity by forbidding to itself the use of either non-quantitative data or to perform syntheses using extra-economic knowledge, does financial cliometrics (for example) not become simply an exercise in financial econometrics with a longer time horizon than usual? The same thing can be said about applications to the economics of education or of labour. What contribution can it make when it is empirical in the pure economist sense of the word besides the provision of original databases for a longer period and long run econometric exercise?
Summing up, it could be contented that aporias can be distinguished in the cliometric approach. Wishing to be too strictly "economist", it would not be possible for cliometric to answer certain questions that would require for example more information about the microstructure of financial markets or on the actual functioning of stock exchanges during the period under scrutiny — it would only measure phenomenon that it cannot explain. It would require the specific approach of a historian to describe the reasons for the lack of relevance of such economic theory in a given context (precise date and period). It is perhaps only in this regard that cliometrics can provide something for economists by suggesting lines of research. However, this would be like a message from elsewhere, in the same way as the knowledge of empirical phenomena revealed by political science can help economists to develop better models.
There is nevertheless a last possibility, viewing the new economic history as the science of the emergence of institutional and organisational structures (Greif, 2006; North, 1990, 2005). Economic history would use the old techniques of the discipline coupled with the state of the art arsenal of econometrics in order to reveal stylised facts about the efficiency of various institutional arrangements as well as on the causes and consequences of institutional change. It would help the theorist in developing a true theory of institutional change, i.e. one that at the same time would be general (serving the needs of policy makers today for example) while solidly grounded on empirical regularities as put forward by a joint economic and historical analysis. This analysis of institutional morphogenesis would be the true theoretical part of a cliometric science, besides the sole playing ground of long run econometricians.
It is clear that economists' desire for generality (their vision of laws similar to the situation in physics) encourages them not to pay too much attention to contextualisation. However, neo-institutionalist economists like North or Greif warn us to take institutions seriously. This article is thus also aimed at encouraging economists to examine more systematically these theories founded on history and nonetheless aiming at the determination of general laws (on the creation of institutions or of institutional changes). Beyond the studies of long quantitative data sets, a branch of cliometrics is more and more focused on the role and evolution of institutions by aiming at combining both the desire for generality of economists and the taking into account of the precise context in which economic players act. This middle road between pure empiricism and disincarnate theory (as it ignores the role of institutions) might perhaps open the door to an economic theory producing laws useful for understanding the world (an intellectual aim) and for political decision makers (aimed at action).
Jean-Luc DEMEULEMEESTER & Claude DIEBOLT, for AFC.
February 2006.
[1] We should point out that some partisans of economics as a pure deductive theory do not advocate the use of mathematics but instead of logics, even a verbal ones (referring to a third type of reasoning between pure rethorics and pure mathematical logics. See the works of Perelman on legal logics).
[2] One should note that the Popperian methodology may be more suitable for testing deterministic propositions ("always true") and that economics generally propose (empirically) probabilistic propositions ("true on average"). The new field of experimental economics nevertheless consistently rejects the a priori purely instrumental rationality assumptions — questioning the very foundations of standard microeconomic analyses and calling for a much more complex modelling strategy.
[3] Let us take, for example, the simple example of pure and perfect competition.