Pawel Ciompa and the meaning of econometrics: a comparison of two concepts

Abstract

Pawel Ciompa’s original and forgotten conception of econometrics is compared with Ragnar Frisch’s interpretation of the term. Ciompa conceived of econometrics as being entirely descriptive, facilitating the conveyance of information in the field of accounting. In contrast, Frisch regarded it as a quantitative and empirical approach to economic theorising. Arguments against Frisch’s conception of econometrics have been brought forward among others by Ciompa’s countrymen of the Austrian School. Ciompa’s own conception is immune to that criticism. It is concluded that modern econometrics should be confined to the narrower descriptive character that Pawel Ciompa originally attached to the term.

Keywords:

• Econometrics
• Pawel Ciompa
• Ragnar Frisch
• Austrian economics

JEL CODES:

• B23
• B31
• B53
• C10

Acknowledgements

I thank Dr. Łukasz Dominiak from Nicolaus Copernicus University in Toruń, Poland, for his assistance in looking up Polish references on Pawel Ciompa. I also thank an anonymous reviewer for constructive and helpful comments on earlier drafts of this paper.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 In the German original: “die geometrische Darstellung des Wertes” (Ciompa 1910, 5). We will go into much more detail in the second section of this paper.

2 All quoted passages from the Grundrisse in this paper are my own translations. In the German original

(Ciompa 1910, 5) we read:

Wie die mechanischen, akustischen, dynamischen und dgl. Erscheinungen durch die Physik oder wie die Massenerscheinungen durch die Geometrie, so sollten auch die volkswirtschaftlichen Erscheinungen durch eine Lehre, die ich mir als eine Art Oekonomographie vorstelle, zur Darstellung gebracht werden. Diese Oekonomographie waere eine Art darstellende Nationaloekonomie, sie muesste auf der Nationaloekonomie, Mathematik und Goemetrie aufgebaut werden. Einer solchen Lehre wuerde dann vor allem die geometrische Darstellung des Wertes zufallen. Diesen Teil der Oekonomographie nenne ich Oekonometrie. Die praktische Anwendung dieser Oekonometrie auf die mathematische Darstellung der Werte und deren Veraenderungen ist dann die Buchhaltung. Umgekehrt ist dann die Oekonometrie nur die Theorie der Buchhaltung.

In the remainder of the paper, the original German quotations are only provided, when my English translations run the risk of becoming too inaccurate for a proper understanding of what Ciompa actually wrote.

3 This classification scheme is known for example from Ciompa’s fellow countryman Carl Menger (2007 [1871], 121ff. and 226ff.), but at least the distinction between use and exchange value has a much longer tradition in the history of economic thought, for example over Pufendorf (1744) to the scholastic doctors of the Middle Ages, all the way back to the old Greeks (Rothbard 2006 [1995], ch. 1; Schumpeter 2006 [1954], part II, ch. 2).

4 This is an example of an imprecision in Ciompa’s exposition. It is not quite clear what he means by proportions of value. Although, he made the above-mentioned distinctions into different kinds of values, he is in the remainder of his book usually only writing about “the value” as if there was only one. These inaccuracies notwithstanding, here again, we might argue that Ciompa follows his countrymen C. Menger and E. von Böhm-Bawerk, although he is not referring to them and their work directly. Böhm-Bawerk stated that “organized exchange gives almost every good a second value” (Böhm-Bawerk 1930 [1891], xxxiii), which is of course the exchange value, and then emphasized the indissoluble connection between price and exchange: “The law of Price, in fact, contains the law of Exchange Value” (132). Furthermore, Menger (2007 [1871], 257) has presented an account of the emergence of money as just another good with certain properties from the original state of barter exchange.

5 The careful reader realizes that at least a person’s own labor according to the proposed definitions must be seen as a part of wealth, as it falls under the definition of services and therefore is an economic good.

6 It is clear that in practice, and from a broader perspective, it might be very difficult to decide what a productive use of wealth is–an argument which has been brought forward against the idle resources argument for macroeconomic policy interventions during crises and recessions (Hutt 1939). However, for Ciompa’s purposes, it seems, the concepts are sufficiently well defined, as we could relatively easily find out which economic goods are used, and therefore constitute capital, in any specific and well-defined economic action, for example, the production of a wooden chair.

7 In the German original, we read:

Die Buchhaltung hat die Aufgabe, die aus dem wirtschaftlichen Leben resultierenden Erfolge des Vermögens und Kapitals mathematisch resp. ökonometrisch zu berechnen resp. darzustellen. Das wirtschaftliche Leben besteht aus Prozessen und Handlungen, welche altes Vermögen $v_1$ verbrauchen, um neues Vermögen $v_2$ zu schaffen. Dadurch nimmt das ursprüngliche Vermögen $V_1$ ab, es bleibt ein Vermögen $V_2=V_1-v_1,$ welches wiederum zum Vermögen $V_3=V_2+v_2$ anwächst. Dasselbe geschieht mit dem Kapital dieses Vermögens; im ersten Falle nimmt es gleichzeitig ab und es verbleibt $C_2=C_1-c_1,$ im zweiten Falle nimmt es mit dem Vermögen gleichzeitig zu: $C_3=C_2+c_2.$ (Ciompa 1910, 14)

8 The Krone became the official currency in the Austro-Hungarian Empire in 1892 when the gold standard was adopted. It replaced the Gulden, which was defined in terms of silver.

9 The two tables correspond to Fig. 21 and Fig. 22 in Ciompa (1910, 17).

10 Ciompa notes that “the truth and clarity of our balances is sinking further into a muddy subsoil. Today, it takes considerable effort to rescue the theory of accounting from the clutter that surrounds it” (136). He argues that differences in accounting rules for different types of businesses (privileges for corporations) and different types of equity, such as operational capital [Betriebsgegenstände] and goods for sale [Veräusserungsgegenstände] are unjustifiable on economic grounds. Everything should be based on universal principles.

11 For biographical information on Frisch, see for example Bjerkholt and Dupont (2010, section 1), Bjerkholt (1995) and Frisch (1970c).

12 An English translation of the article has been published in Chipman (1971, 386–423) under the title “On a Problem in Pure Economics”. Having no access to the volume, direct quotations provided in the article are my own. The French original is provided only for passages that are of particular importance. All quotations come from the republication of the original French article in Metroeconomica in 1957.

13 In the French original we read:

Intermédiaire entre les mathématiques, la statistique et l’économie politique, nous trouvons une discipline nouvelle que l’on peut, faute de mieux, désigner sous le nom de l’économétrie.

L’économétrie se pose le but de soumettre les lois abstraites de l’économie politique théorique ou l’économie ‘pure’ à une vérification expérimentale et numériques, et ainsi de constituer, autant que cela est possible, l’économie pure en une science dans le sens restreint de ce mot. (Frisch 1957 [1926b], 79)

14 After Irving Fisher had arranged a visiting professorship from his personal funds, Frisch came to Yale University in the early 1930s. In one of his lectures, he praised astronomy as being one of the most scientific fields, as “astronomical observations are filled into the theoretical structure […] Economic theory has not as yet reached the stage where its fundamental notions are derived from the technique of observations” (Frisch 1930, ch. 1.1, as cited in Bjerkholt and Dupont 2010, 57). It would be false, however, to reduce Frisch’s ideal conception of econometrics to empiricism and little else, although, this is exactly what this particular statement suggests. In his lectures there are several other references to Newtonian physics and Einstein’s theory of general relativity.

15 According to Bjerkholt (1995), Frisch’s two great mentors were Marshall and Wicksell (he refers to Frisch and Marshall 1950; Frisch 1952). Apparently, their books were the only really worn-out ones in his personal library. Frisch credits Marshall for having combined the Walras–Jevons–Menger subjective notion of value with the cost of production viewpoint (Frisch 1970a, 16). He specifically refers to Menger as the head of the Austrian economists, but misspells his first name as Karl instead of Carl. Karl Menger (1902–1985), son of Carl Menger, was a mathematician and probably more to Frisch’s liking. Obviously, Frisch was more drawn towards the formal mathematical presentation in Walras (1874) and Jevons (1965 [1871]), rather than the entirely verbal presentation in Menger (2007 [1871]). For an interesting de-homogenization of these three thinkers, see Jaffé (1976). Frisch studied a French translation of Fisher’s dissertation thesis Mathematical Investigations in the Theory of Value and Prices during his time in Paris (Bjerkholt and Dupont 2010, 28, fn. 10). Fisher himself refers to Jevons’ Theory of Political Economy as one of two books that had the biggest influence on him (Fisher 1892, 3). Although, Walras and his successor at the University of Lausanne, Pareto, both made important contributions to the mathematical formalization of economic theory in general, and value theory in particular, it is important to note that Pareto rejected a cardinal interpretation of value and utility, but thought of them as being ordinal, and, in fact, based his famous welfare economics on ordinal utility (Aspers 2001). In this respect, Frisch departed from Pareto.

16 Schumpeter may be seen as an exception. Although he did engage in empirical work, for example, in his Business Cycles (Schumpeter 1939), he was not actively involved in the mathematization of economic theory. He did, however, enthusiastically support the efforts of his mathematically inclined colleagues. Schumpeter, together with Frisch and Fisher, was a founding member of the Econometric Society in 1930 and its president from 1940 to 1941.

17 On a recent and systematic reinterpretation of economic laws as essentially being of counterfactual nature, see Hülsmann (2003).

18 For a complete list of the axioms outlined in Frisch’s Poincaré lectures (Frisch 2013 [1933], Lecture 1), see Bjerkholt and Dupont (2010, 42, Table 2).

19 Similarly, he would later explain in his Nobel Memorial lecture that expert economists could derive a preference function for policy optimization from interrogating the politicians in charge, which is a conclusion that he reached “not only on theoretical grounds but also because of […] practical experience” (Frisch 1970a, 23).

20 Frisch includes intercepts in the equations, but this is not relevant at all.

21 Assume the first initial random observations were $(x_{11}=1.5, x_{21}=-0.6)$ and $(x_{12}=0.2, x_{22}=0.4).$ Applying $T_3$ leads to the following transformed observations: $(y_{11}=1x_{11}+2x_{21}=1.5-1.2,y_{21}=2x_{11}+4x_{21}=3-2.4)$ and hence $(y_{11}=0.3, y_{21}=0.6),$ and $(y_{12}=1x_{12}+2x_{22}=0.2+0.8,y_{22}=2x_{12}+4x_{22}=0.4+1.6)$ and hence $(y_{12}=1, y_{22}=2).$ In general, $y_{2i}=2x_{1i}+4x_{2i}=2\left(x_{1i}+2x_{2i}\right)=2y_{1i}$ holds, and hence, no matter what the initial random data, the transformation leads to a straight line in the $\left(y_1,y_2\right)$-plane with slope 2. It should also be obvious that, simply by choosing a different singular transformation matrix, we enjoy full freedom of changing the slope. If we had $T_3=\left(\begin{array}{cc}1& -3\\ 2& -6\end{array}\right)$ instead, the slope would be −3.

22 The transformation matrices $T_1$ and $T_2$ have inverse matrices, so that it is always possible to go back to the initial data from the transformed data. For example: $$T_1{T}_1^{-1}=\left(\begin{array}{cc}1& 2\\ 0& 4\end{array}\right)\left(\begin{array}{cc}1& -\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$2$}\right.\\ 0& \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$4$}\right.\end{array}\right)=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right)=I.$$

23 One must keep in mind that the illustration is not really chaotic. For all we know, human beings are not capable of purposefully creating chaos. It might look like chaos on the surface, but there is always an underlying deterministic structure. For the sample in the above example (Figure 4), I have used the statistical programming language R, which generates pseudo-random numbers using the Mersenne-Twister based on the Mersenne prime number $2^{19937}-1.$ This pseudo-random number generator has been developed by Matsumoto and Nishimura (1998). The same process is used for the following example (Figure 5).

24 In fact, simulating the exact scenario from Figure 4 repetitively (1000 times), has shown that in 96% of the cases the correlation coefficient after transformation by $T_1$ is larger than 0.82. It has never been lower than 0.76.

25 For the simulation of variable $x_1,$ the mean of the normal distribution has been randomly selected (with replacement) from the integers between −25 and +25. The standard deviation has been randomly selected (with replacement) from the integers between 1 and 10. Both parameters have been randomly changed after every 10th observation. For variable $x_2,$ the lower bound of the uniform distribution has been randomly selected (with replacement) from the integers between -25 and 25, the upper bound has then been randomly fixed between 1 and 10 integers above the lower bound (again random selection with replacement). After every 10th observation the bounds have been randomly changed. After the simulation of a sample of size $n=100,$ all observations have been divided by 10, in order to bring them roughly onto the same scale as the observations in the previous simulation.

26 Now, it is clear that in principle, the resulting correlation coefficient could also be higher than 0.87. The point is, that one can never know, when there is a “chaotic” underlying structure. Repeated simulations (1000 times) show that more than half of the time (53.7%) the correlation coefficient is below 0.5, and in only 0.4% of the simulations, it is above 0.87.

27 The position taken here is contested by some authors. Miller (1994, ch. 2), for example, argues that falsificationist approaches are not inductive and tries to refute a series of objections to the claim that Karl Popper had solved the problem of induction. I thank an anonymous reviewer for pointing my attention to Miller’s analysis.

28 It is not to be confused with the axiomatization in the tradition of the Austrian school, which came to full fruition in Rothbard (2009 [1962], p. lvi), who claims:

The present work deduces the entire corpus of economics from a few simple and apodictically true axioms: the Fundamental Axiom of action–that men employ means to achieve ends, and two subsidiary postulates: that there is a variety of human and natural resources, and that leisure is a consumers’ good.

In fact, this might exactly be the kind of “metaphysical verbiage” that Frisch refers to. Action itself is an unobservable concept, and hence, useless from the point of view of Frischian econometrics. Certain consequences or aspects of actions are observable, such as the movement of the human body or the rearrangement of objects in the external world, but not action itself.

29 One of the most prominent and important cases in modern macroeconomics that provide a well-known example for what has been considered a structural change, is the relationship between price inflation and unemployment. The stagflation of the 1970s has led to a rejection of the large-scale Keynesian macroeconometric models à la Evans and Klein (1967), Klein and Goldberger (1955) or Klein (1964, 11–58) that were essentially elaborated IS-LM models augmented with a politically exploitable Phillips Curve (Webb 1999). Interestingly, Hurtado (2014) gathered empirical evidence that the modern New Classical and New Keynesian DSGE models (Galí and Gertler 2007; Galí 2008; Woodford 2003) that emerged out of the New Classical Critique (Lucas 1983 [1976]) would not have performed much better, if they had been used back in the 1970s instead of the old models. True enough, those models suffer from the same fundamental problems.