and Stigler Barkai on Ricardo's Profit Rate Theory
Some methodological considerations 35 years later *
by Andrew Kliman, Ph.D. Associate Professor Department of Economics Pace University Pleasantville Pleasantville, NY 10570 USA Phone: [phone] Fax: [phone] akliman @ pace.edu Andrew_Kliman @ msn.com
To be presented at the Ninth Annual Miniconference of the International Working Group on Value Theory, at the Eastern Economic Association meetings, Boston, March 15, 2002. 1st Draft, March 2002. Please Do Not Quote Before Obtaining Author's Permission.
* I wish to thank Keith Gibbard for his extremely helpful references to and discussion of the literature. The usual caveat applies.
This paper briefly surveys the hermeneutic criterion that interpretations should aim to understand a text as a unified whole. It then suggests that this criterion was applied to the interpretation of analytical works by Stigler (1965), when he proposed that these interpretations be tested according to their ability to derive the author's main conclusions. I next show that other eminent historians of economic thought, Barkai and Hollander, have also embraced and employed Stigler's principle, and I provide some new arguments in support of it. Finally, I apply Stigler's principle, as well as the general hermeneutic criterion, to the new controversy over the interpretation of Marx's value theory, and show that their application allows the controversy to be definitively resolved. Marx's theory is cleared of the charges of internal inconsistency that have been leveled against it, and the temporal single-system interpretation (TSSI) emerges as the preferred one.
I. The Holistic Criterion of Interpretive Adequacy With the possible exception of deconstructionists, those who have dealt seriously with issues of textual interpretation during the last two centuries all seem to accept what I shall call the "holistic criterion of interpretive adequacy." As originally formulated by Friedrich Schleiemacher, it holds that "as the whole is of course understood from the individual, so too the individual can only be understood from the whole" (quoted in Connolly and Keutner 1988:10). This principle distinguishes genuine understanding from misunderstanding; we genuinely understand a text (for example) only if the individual parts are reconciled, brought together as moments of the whole.1 The principle gives rise to a test of interpretative adequacy: "the adequacy of a given textual interpretation depends on the extent to which it can show the text's coherence as a unified
2
whole" (Warnke 1993:21). Accordingly, part of Kosk's (1976:95) distinction between "a substantiated interpretation" and "textual distortions or modifications" is that the former will "leave no opaque, unexplained or `accidental' passages in the text." The basic point is simply that a substantiated interpretation must not be selective, but must explain the text "as a whole." If an interpretation leaves some passages unexplained, or if there remain "accidental" passages passages that contradict it then it must be regarded as unsubstantiated. Despite its simplicity, this criterion of interpretive adequacy has some important implications that have not always been readily apparent. It provides a prima facie reason to doubt claims that a text is internally inconsistent. It also implies that an interpretation that can resolve apparent inconsistencies within a text that can, in other words, understand the text as an internally coherent one is superior to one that cannot. Of course, a particular text may indeed be internally inconsistent. But that is so only if there is no interpretation according to which it forms a unified, coherent whole. Some interpreters, however, judge a text to be internally inconsistent merely because it becomes so under their own interpretation, even though an alternative interpretation can understand it as a coherent whole. In such cases, the holistic criterion clearly implies, first, that the latter interpretation is the better one, and second, that we must reject the claim of internal inconsistency. Thus Warnke (1993:21) argues that allegations of internal inconsistency should not be taken at face value: "if certain parts of the text seem to contradict others, the initial presumption of the critic has to be that they do so because one or the other set has been misunderstood [by the critic]." The point is that a text has not been demonstrated to be internally incoherent merely because a particular interpretation finds it so. A claim of self-contradiction is substantiated only
3
if a variety of possible interpretations repeatedly fail to eliminate the apparent inconsistencies. If, on the other hand, some interpretation does succeed in eliminating them, the claim of selfcontradiction is thereby falsified. This is a particular instance of the elementary logical point that a claim is true only if it holds in every possible instance, while even one counterinstance suffices to falsify the claim.
II.
Stigler's "Principle of Scientific Exegesis" The great merit of George Stigler's (1965) "Textual Exegesis as a Scientific Problem" is
that it showed how to concretize the holistic criterion when considering differing interpretations of a theoretical work. His "principle of scientific exegesis" holds that an interpretation of a deductive theoretical work can be regarded as correct only if it can deduce the author's main analytical conclusions from her definitions and premises. A University of Chicago economist who later won the Nobel Prize, Stigler articulated this principle in the midst of a debate over the meaning of Ricardo's theory of the demand for corn. Haim Barkai (1965) had just argued that, according to Ricardo's theory, a rise in the price of corn would cause the amount demanded to fall. Stigler argued, to the contrary, that Ricardo had in effect asserted that the demand for corn is perfectly inelastic. A rise in its price would not lead to a reduction in the quantity demanded. I am not a Ricardo specialist, so I am not in a position to say which author, if either, was right about the substantive issue. My concern is rather with the test of interpretive adequacy that Stigler proposed. He objected to Barkai's selective use of quotations in order to make his case. "Why," Stigler (1965:448) asked, "should we allow the hand-picked quotation to carry an interpretation
4
when we would reject the hand-picked fact as an empirical test of a hypothesis? In fact the two problems are basically the same." This point is important because it pinpoints precisely what is wrong with quoting out of context, and why "battles of quotations" fail in the end to clarify matters. These are "unscientific" ways of deciding among interpretations. As Stigler noted, issues of interpretation are empirical issues, textual evidence is empirical evidence, and so the "scientific," scholarly procedure is to evaluate competing hypotheses about a text's meaning in essentially the same way that one evaluates competing hypotheses about the external world. One needs to test whether they fit with the empirical evidence taken as a whole. Yet the really brilliant aspect of Stigler's paper is his apparently novel understanding of what constitutes empirical evidence in the case of a text. He recognized that the textual evidence is not limited to passages in which an author sets out her definitions and premises; another part of the evidence consists of her theoretical conclusions. Stigler thus proposed that a textual interpretation be judged according to whether it can hold both types of evidence together and form from them a unified whole. The test of an interpretation, in other words, is whether it can deduce the author's theoretical conclusions from her definitions and premises: textual interpretation must uncover the main concepts in the man's work, and the major functional relationships among them. ... ... We increase our confidence in the interpretation of an author by increasing the number of his main theoretical conclusions which we can deduce from (our interpretation of) his analytical system. The test of an interpretation is its consistency with the main analytical conclusions of the system of thought under consideration. If the main conclusions of a man's thought do not survive under one interpretation, and do under another, the latter interpretation must be preferred. (The analogy to maximum likelihood is evident.) ... This rule of consistency with the main conclusions may be called the principle of scientific exegesis. [Stigler 1965:448].
5
A few points are worthy of emphasis or elaboration. First, Stigler did not say that "consistency with the main conclusions" is one desirable feature of an interpretation. He proposed it as the test of interpretive adequacy. What this means is that an interpretation that one believes to have other desirable features, but which fails the test of consistency with the main conclusions, must be rejected as incorrect. Conversely, an interpretation that one regards as undesirable for other reasons, but which is consistent with the author's conclusions, must be accepted. One may not, for instance, favor the former on the ground that the latter "distorts our understanding" of the text. Clearly, if an interpretation fails the test of interpretive adequacy, its "understanding" of the text must be regarded as a misunderstanding. Second, just as the holistic criterion in general implies that an interpretation which finds the text to be internally coherent is better than one that doesn't, so too does Stigler's application of it. "If the main conclusions of a man's thought do not survive under one interpretation, and do under another, the latter interpretation must be preferred." Finally, although Stigler's "analogy to maximum likelihood" may not be as evident as he thought, he appears to have been suggesting that one should work backwards. One should use a theory's conclusions as evidence of what its premises and definitions actually are. One should infer from the conclusions how to interpret otherwise ambiguous textual evidence pertaining to definitions and premises. In maximum likelihood estimation, one works backwards in a similar manner, beginning with the results the sample observations and inferring the values of the parameters from them. The parameter values one chooses are those which are most likely to have produced the observations. The conclusion of Stigler's paper, as well as its method, lends support to this interpretation of his analogy. He wrote in conclusion: "Let us recognize the fact that the
6
interpretation of a man's position especially if the man has a complex and subtle mind is a problem in inference, not to be solved by the choice of quotations" (Stigler 1965:450, emphasis added). And in his discussion of what Ricardo had assumed about the demand for corn, Stigler indeed worked backwards, attempting to infer Ricardo's view from his theoretical conclusions. Beginning with the conclusion that the rate of profit tends to fall with the progress of capital accumulation, plus some other premises of Ricardo's theory, Stigler tried to show that it implies a perfectly inelastic demand for corn. That must be so, he argued, since Ricardo's conclusion that the profit rate falls would otherwise not follow from his premises. Stigler's principle is an application of the holistic criterion in two senses. First, he proposed that interpretations be tested according to whether they can establish a coherence between two different aspects of a text definitions and premises on the one hand, conclusions on the other. Second, his principle requires a holistic rather than a linear method of reading a theoretical text. It denies that the meaning of an author's premises (and definitions) can be determined by focusing solely on passages that discuss them directly. It thereby also implicitly denies that one can judge whether a work is internally coherent by determining whether its conclusions follow from an interpretation of its premises (and definitions) that has been worked out prior to and without regard to the conclusions. The meaning of the premises is instead established when an interpretation is able to take passages that contain conclusions, and passages that set out premises, and make them coherent with one another. And while we must no doubt proceed from premises to conclusions in order to ascertain whether an argument is internally consistent, Stigler's principle stipulates that we must also proceed from conclusions to premises in order to ascertain what an author's premises really are. This is a classic example of the so-called "hermeneutic circle."
7
III. Barkai (and Hollander) on Stigler Barkai (1967) responded to Stigler about a year later. That he continued to affirm his own interpretation of Ricardo's theory against Stigler's critique is not surprising. What may be surprising is that Barkai nevertheless endorsed the principle of scientific exegesis. After quoting Stigler in the opening sentence of his reply "The test of an interpretation is its consistency with the main analytical conclusions of the system of thought under consideration" Barkai (1967:75) remarked that "[t]his is undoubtedly a useful criterion, and I propose to apply it here." Thus Barkai accepted the terms of the debate as Stigler had just (re)formulated them. He agreed that it was not possible to decide which interpretation of Ricardo's theory of the demand for corn was correct simply by examining isolated passages that address the issue directly, or even by examining the totality of such passages. To be considered correct, Barkai conceded, an interpretation of Ricardo's theory must be able to deduce his conclusion that the rate of profit falls with the progress of accumulation: I do not dispute the strategic position of this `law' [of the falling profit rate] in Ricardo's conceptual structure. Consequently, I would have to concede that my interpretation of Ricardo's position on demand is untenable if it were true that a `conventional' demand relation and the law of the falling profit rate are incompatible, or that the latter is `weakened' when the former applies. [Barkai 1967:76] Barkai (1967:76) thus acknowledged that he needed to "show ... that `the law of the falling profit rate' can be deduced rigorously from the premises of the Ricardian model, even if one assumes a negatively sloped (and not a zero) elastic demand curve." According to Ricardo's theory, both authors agreed, an increase in capital will lead to population growth and thus an increase in the demand for corn. They also agreed that Ricardo had held that a rising (relative) price of corn would lead to a falling profit rate. The dispute was thus reducible to the following: if Ricardo had in effect assumed that the demand curve for corn
8
was downward sloping, rather than vertical as Stigler maintained, would an increase in the demand for corn still cause its (relative) price to rise? Barkai set out to show that it would. He was of course able to show this it is the most elementary prediction of supply and demand analysis. To be sure, Barkai acknowledged, the price will rise less and thus the profit rate will fall more slowly over time if the demand curve is downward sloping than if it is vertical. He argued, however, that this "has no bearing on what is at issue" (Barkai 1967:77). What was at issue was only the direction of change in the profit rate, not the rate of change. Since I am not a scholar of Ricardo's work, I am unqualified to judge whether this last contention is correct, and thus unqualified to say which author, if either, was right about the substantive issue. (To my knowledge, Stigler never responded to Barkai's rejoinder.) For those of us who are not Ricardo specialists, what is noteworthy about this exchange is not that Stigler and Barkai disagreed about substance, but that they agreed about method. Both authors acknowledged that the test of an interpretation its whether it can derive an author's conclusions from (its understanding of) her premises, and both willingly applied this test to the case at hand. In recent years another prominent historian of Ricardo's thought, Samuel Hollander, has also returned to and endorsed Stigler's view that the test of an interpretation of an analytical work is whether it is consistent with the text's main conclusions. It is true that Hollander takes issue with Stigler's 1965 paper, but his criticisms pertain solely to other aspects of Stigler's position on interpretation. As Hollander (1990:730-32) interprets Stigler, the latter suggested that we should not test interpretations against the text's main conclusions as the author herself formulated them. Our goal should not be to understand the what author really believed, but "to maximize the value of a theory to the science." We should thus formulate a text's "central theoretical position ... in a
9
strong form capable of contradiction by the facts," even if what the author herself wrote must first be "amended" in order to produce the falsifiable hypothesis we desire.2 Although Stigler (1990) replied to Hollander's paper, he did not contest Hollander's interpretation of his position. So it is presumably the intended one. In any case, I believe that Hollander (1990:731, 733; emphasis in original) was right to insist that "[w]e must isolate the central theoretical position from the texts without amendment" and to reaffirm that "the primary requisite of exegesis ... is to get the model right on the author's own terms." What is in dispute here is quite limited. The only point of disagreement is whose version of the "main conclusions" to use, the author's original ones or the interpreter's possibly "amended" ones. With regard to the main issue, whether an interpretation needs to be consistent with the main conclusions in order to be considered correct, Hollander accepted Stigler's position. "The `scientific rule of exegesis' is ... acceptable provided it is limited to a test of interpretation understood simply as consistency with the main analytical conclusions" (Hollander 1990:131).
IV.
In Defense of Stigler's Principle As far as I am aware, no one has yet advanced an argument against Stigler's view that
"[t]he test of an interpretation is its consistency with the main analytical conclusions of the system of thought under consideration." Moseley (2000) is the only author I know of who explicitly rejects it, and he fails to offer any argument against it.3 Yet this test of interpretive adequacy appears to be rejected implicitly by the mainstream Marxist economists, inasmuch as they fail to employ it and decline to discuss its validity. It will therefore be helpful to offer some arguments in its favor.
10
Stigler himself, as we saw above, justified his principle of scientific exegesis by means of an analogy to maximum likelihood estimation. He drew this particular analogy in order to stress that the test of interpretations he proposed was simply an application of a test commonly employed in scientific work generally. The title of his paper and the name he gave to his principle provide further evidence that he wished to stress the consonance of his principle with generally accepted scientific norms and practices. In the tradition of textual hermeneutics, the holistic criterion of interpretive adequacy has often been defended by means of a different kind of analogy. Beginning with Schleiermacher, it has been noted that in daily life "wherever there is anything unfamiliar ... in the expression of thoughts through speech" (quoted in Connolly and Keutner 1988:9) we apply essentially the same criterion. We try to understand the speaker's utterances as a coherent whole, in two senses. First, we interpret the individual words, phrases, etc. in the context of her statement as a whole. Second, we choose to interpret her statement in such a way that it makes sense if that is at all possible. As Warnke (1993:21) would put it, if we at first cannot make sense of it, our initial presumption is to chalk that up to our own misunderstanding rather than to charge the speaker with internal inconsistency. At least this is how we behave when we listen in good faith. The point is that we behave in the same way if we are interpreting in good faith. Just as we can defend the holistic criterion by pointing to our practices in daily life, so too can we defend Stigler's principle of scientific exegesis by referring not only to scientific norms and practices, but also to daily-life ones. Indeed, I myself defended the principle in this way when, unaware of Stigler's work on the issue, I first proposed it independently in 1996. In Kliman (1996), I wrote Although acknowledging that the TSS [temporal single-system] interpretation dispels the appearance of inconsistency in key aspects of Marx's value theory, some of its
11
critics have suggested that it may nonetheless not be what his texts "really meant." What this suggestion overlooks is that an interpretation's repeated ability to replicate a text's theoretical results is itself decisive evidence that the interpretation corresponds to the "real meaning" of the text. Let me illustrate this by means of a parable. Many people have been trying to put together a jigsaw puzzle, but they continually fail. Some say: "The puzzle has no solution. Let's throw away some pieces and see if we can solve the puzzle." Others say: "Let's take some pieces from a different puzzle and use them here, to see if we can solve the puzzle." And some say: "Let's throw out the puzzle and do a different puzzle." Suddenly a few other people come along and say: "The puzzle's instructions read: `if "joining pieces" is equated with "interlocking the pieces," it is always possible to go wrong'." This puzzle lets you join pieces by putting straight edges together. Look, we've done so, and the result is just like the picture on the box." Whose interpretation of the instructions is superior? Don't the results speak for themselves? 4 The following year, I returned to the issue, offering another analogy that can be elaborated as follows (Kliman 1997:10). One person takes a pudding recipe and, reading it in a certain way, sets out to make pudding. He fails. The result is not pudding, but an inedible, foulsmelling, and unsightly concoction. Another person takes the same recipe, but reads it differently way. She succeeds in making a tasty, aromatic, and attractive pudding. Whose interpretation of the recipe would we consider to be the right one? The proof of the pudding is in the eating. The first parable is admittedly somewhat fanciful. Yet I suggest that both parables defend the principle of scientific exegesis by appealing to the criterion of interpretive adequacy we regularly employ in daily life. We decide among differing "interpretations" of instructions, cookbook recipes, etc., by looking at the results they produce. Those that achieve the intended result are deemed successful; those that do not are deemed failures. The purpose of both parables is to point out that the premises and definitions of an analytical argument are like the instructions to a puzzle or the steps of a recipe, and the argument's conclusion is like any other
12
result finished puzzle, pudding, etc. that one should obtain if one understands the directions and follows them properly.
V.
Applying Stigler's Principle to the New Value Theory Controversy It may of course be impossible to move successfully from a puzzle's instructions or a
cookbook recipe to the intended result; it may likewise be impossible to move successfully from the premises and definitions of an analytical argument to its conclusion. If that is the case, we must conclude that the directions are flawed, or that the argument is internally inconsistent. Yet that was not the case in my parables, nor is it the case in the renewed controversy of the past decade over how to interpret Marx's value theory. His theory has been consistently subjected to allegations of internal inconsistency for more than a century (see, e.g., Bortkiewicz 1952). Indeed, most interpretations of Marx's theory are, to varying extents, unable to deduce his conclusions from (their understanding of) his definitions and premises. But some puzzle-assemblers and cooks in my parables also met with failure. The point is that if success is at all possible and the existence of an interpretation that deduces the conclusions, assembles the puzzle, or makes pudding is proof of that possibility then the claim of internal inconsistency must be rejected. In the case of Marx's theory, the existence of the temporal single-system interpretation is proof that success is possible.5 Table 1 examines the extent to which it and other interpretations are able to deduce Marx's theoretical conclusions (for further elaboration, see Kliman and McGlone 1999, and Kliman 2001). The others are the standard (simultaneous dual-system) interpretation (e.g., Morishima 1973) and the recent simultaneous single-system interpretations (e.g., Lee 1993, Moseley 1993).
13
Table 1 Interpretations of Marx's Value Theory: Contrasting Implications
Interpretation Standard (simultaneous dualsystem) Simultaneous SingleSystem
Marx's Theoretical Conclusions
Temporal SingleSystem
Equalities and Inequalities profit rate = s/(c + v) total price = total value total profit = total surplus-value values always > 0 Relations of Determination profit always > 0 if surplus-labor > 0 surplus-labor always > 0 if profit > 0 mechanisation itself can reduce profit rate variations in living labour performed affect profit rate1 profit rate invariant to distribution of profit1 profit rate affected by luxury industries1 inputs lacking value before production transfer no value unit values invariant to real wage rate unit values invariant to length of working day Conclusions deduced Conclusions negated
1 2 2 3
3 3
3 3
2 11 4 9
13 0
= Marx's conclusion deduced
1 2 3
refers to functional determination of uniform profit rate not deduced unless postulated not deduced even without joint production
14
Thirteen different conclusions are listed. One may argue that this or that conclusion is not among those that Stigler called the "main analytical conclusions" that an adequate interpretation must be able to deduce. Others, however, clearly fall within that category, such as Marx's exploitation theory of profit ("profit always > 0 if surplus-labor > 0" and "surplus-labor always > 0 if profit > 0") and his law of the tendential fall in the profit rate ("mechanization itself can reduce profit rate"). In any case, the point of listing all of the conclusions is to show that interpretive success is possible in every instance. Wi...