Non-praxeological schools of thought mistakenly believe that relationships between certain events are well-established empirical laws when they are really necessary and logical praxeological ones. And they thereby behave as if the statement “a ball cannot be red and non-red all over at the same time” requires testing in Europe, America, Africa, Asia and Australia (of course requiring a lot of funds in order to pay for such daring nonsensical research). Moreover, the non-praxeologists also believe that relationships between certain events are well-established empirical laws (with predictive implications) when a priori reasoning can show them to be no more than information regarding contingent historical connections between events, which does not provide us with any knowledge whatsoever regarding the future course of events.
This illustrates another fundamental confusion non-Austrian schools have: a confusion over the categorical difference between theory and history and the implication that this difference has for the problem of social and economic forecasting.
I must again begin with a description of empiricism, the philosophy which thinks of economics and the social sciences in general as following the same logic of research as that, for instance, of physics. I will explain why. According to empiricism—today’s most widely held view of economics—there is no categorical difference between theoretical and historical research. And I will explain what this implies for the idea of social forecasting. The very different Austrian view will then be developed out of a critique and refutation of the empiricist position.
Empiricism is characterized by the fact that it accepts two intimately related basic propositions.20 The first and most central one is: Knowledge regarding reality, which is called empirical knowledge, must be verifiable or at least falsifiable by observational experience. Observational experience can only lead to contingent knowledge (as opposed to necessary knowledge), because it is always of such a kind that, in principle, it could have been different than it actually was. This means that no one can know in advance of experience—that is before actually having had some particular observational experience—if the consequence of some real event will be one way or another. If, on the other hand, knowledge is not verifiable or falsifiable by observational experience, then it is not knowledge about anything real. It is simply knowledge about words, about the use of terms, about signs and transformational rules for signs. That is to say, it is analytical knowledge, but not empirical knowledge. And it is highly doubtful, according to this view, that analytical knowledge should be regarded as knowledge at all.
The second assumption of empiricism formulates the extension and application of the first assumption to problems of causality, causal explanation, and prediction. According to empiricism, to explain causally or predict a real phenomenon is to formulate a statement of either the type “if A, then B” or, should the variables allow quantitative measurement, “if an increase (decrease) in A, then an increase (decrease) in B.”
As a statement referring to reality (with A and B being real phenomena), its validity can never be established with certainty, that is, by examining the proposition alone, or of any other proposition from which the one in question could be logically deduced. The statement will always be and always remain hypothetical, its veracity depending on the outcome of future observational experiences which cannot be known in advance. Should experience confirm a hypothetical causal explanation, this would not prove that the hypothesis was true. Should one observe an instance where B indeed followed A as predicted, it verifies nothing. A and B are general, abstract terms, or in philosophical terminology, universals, which refer to events and processes of which there are (or might be, in principle) an indefinite number of instances. Later experiences could still possibly falsify it.
And if an experience falsified a hypothesis, this would not be decisive either. For if it was observed that A was not followed by B, it would still be possible that the hypothetically related phenomena were causally linked. It could be that some other circumstance or variable, heretofore neglected and uncontrolled, had simply prevented the hypothesized relationship from actually being observed. At the most, falsification only proves that the particular hypothesis under investigation was not completely correct as it stood. It needs some refinement, some specification of additional variables which have to be watched for and controlled so that we might observe the hypothesized relationship between A and B. But, to be sure, a falsification would never prove once and for all that a relationship between some given phenomena did not exist, just as a confirmation would never definitively prove that it did exist.21
When we consider this position, we notice that it again implies a denial of a priori knowledge that is at the same time knowledge about anything real. Any proposition that claims to be a priori can, according to empiricism, be no more than signs on paper that are related to each other by definition or by arbitrary stipulation, and is thus completely void: it is without connection to the world of real things whatsoever. Such a system of signs only becomes an empirically meaningful theory once an empirical interpretation is given to its symbols. Yet as soon as such an interpretation is given to its symbols, the theory is no longer a priori true but rather becomes and remains forever hypothetical.
Moreover, according to empiricism, we cannot know with certainty whether something is a possible cause of something else. If we want to explain some phenomenon, our hypothesizing about possible causes is in no way constrained by a priori considerations. Everything can have some influence on anything. We must find out by experience whether it does or not; but then experience will never give us a definite answer to this question either.
The next point brings us to our central topic of this section: the relationship between history and theory. We notice that according to empiricism there is no principal difference between historical and theoretical explanations. Every explanation is of the same type. In order to explain a phenomenon we hypothesize some other phenomenon as its cause and then see whether or not the hypothesized cause indeed preceded the effect in time. A distinction exists between a historical and a theoretical explanation only insofar as a historical explanation refers to events that already happened, something that lies in the past, whereas a theoretical explanation would be an explanation, or rather a prediction, of an effect that has not yet occurred. Structurally, though, there is no difference between such historical explanations and theoretical predictions. There is, however, a pragmatic difference which explains why empiricists in particular stress the importance of a theory’s predictive power and are not content with testing it only vis-à-vis historical data.22 The reason for this is quite evident to anyone who was ever engaged in the foolish game of data analyses. If the phenomenon to be explained has already occurred, it is easy as cake to find all sorts of events that preceded it in time and could possibly be considered its cause. Moreover, if we don't want to lengthen our list of possible causes by finding more preceding variables, we can do the following (and in the age of computers, it’s even easier): We can take any one of the preceding variables and try out different functional relationships between it and the variable to be explained—linear or curvilinear ones, recursive or non-recursive functions, additive or multiplicative relations, etc. Then one, two, three, we find what we were looking for: a functional relationship that fits the data. And you will find not just one but any amount of them that you could possibly desire.
But which of all these preceding events, or of all the types of relationships, is the cause or the causally effective relation? There are no a priori considerations, according to empiricism, that could help us here. And that, then, is the reason why empiricists emphasize the importance of predictions. In order to find out which one of these manifold historical explanations is indeed correct—or at least not false—we are asked to try them out by using them in predicting events that have not yet occurred, see how good they are, and thereby weed out the wrong explanations.
So much for empiricism and its ideas about theory, history, and forecasting. I will not go into a detailed analysis of the question whether or not this emphasis on predictive success changes much, if anything at all, with respect to the rather evident relativistic implications of empiricism. Just recall that according to its very own doctrine, neither a predictive confirmation nor a predictive falsification would help us either in deciding whether a causal relationship between a pair of variables did or did not exist. This should make it appear rather doubtful that anything is gained by making prediction the cornerstone of one’s philosophy.
I would like to challenge the very starting point of the empiricists' philosophy. There are several conclusive refutations of empiricism. I will show the empiricist distinction between empirical and analytical knowledge to be plainly false and self-contradictory.23 That will then lead us to developing the Austrian position on theory, history, and forecasting.
This is empiricism’s central claim: Empirical knowledge must be verifiable or falsifiable by experience; and analytical knowledge, which is not so verifiable or falsifiable, thus cannot contain any empirical knowledge. If this is true, then it is fair to ask: What then is the status of this fundamental statement of empiricism? Evidently it must be either analytical or empirical.
Let us first assume it is analytical. According to the empiricist doctrine, however, an analytical proposition is nothing but scribbles on paper, hot air, entirely void of any meaningful content. It says nothing about anything real. And hence one would have to conclude that empiricism could not even say and mean what it seems to say and mean. Yet if, on the other hand, it says and means what we thought it did all along, then it does inform us about something real. As a matter of fact, it informs us about the fundamental structure of reality. It says that there is nothing in reality that can be known to be one way or another prior to future experiences which may confirm or disconfirm our hypothesis.
And if this meaningful proposition is taken to be analytical, that is, as a statement that does not allow any falsification and whose truth can be established by an analysis of its terms alone, one has no less than a glaring contradiction at hand. Empiricism itself would prove to be nothing but self-defeating nonsense.24
So perhaps we should choose the other available option and declare the fundamental empiricist distinction between empirical and analytical knowledge an empirical statement. But then the empiricist position would no longer carry any weight whatsoever. For if this were done, it would have to be admitted that the proposition—as an empirical one—might well be wrong and that one would be entitled to hear on the basis of what criterion one would have to decide whether or not it was. More decisively, as an empirical proposition, right or wrong, it could only state a historical fact, something like “all heretofore scrutinized propositions fall indeed into the two categories analytical and empirical.” The statement would be entirely irrelevant for determining whether it would be possible to produce propositions that are true a priori and are still empirical ones. Indeed, if empiricism’s central claim were declared an empirical proposition, empiricism would cease altogether to be an epistemology, a logic of science, and would be no more than a completely arbitrary verbal convention of calling certain arbitrary ways of dealing with certain statements certain arbitrary names. Empiricism would be a position void of any justification.
What does this first step in our criticism of empiricism prove? It proves evidently that the empiricist idea of knowledge is wrong, and it proves this by means of a meaningful a priori argument. And in doing this, it shows that the Kantian and Misesian idea of true a priori synthetic propositions is correct. More specifically, it proves that the relationship between theory and history cannot be as depicted by empiricism. There must also be a realm of theory—theory that is empirically meaningful—which is categorically different from the only idea of theory empiricism admits to having existence. There must also be a priori theories, and the relationship between theory and history then must be different and more complicated than empiricism would have us believe. How different indeed will become apparent when I present another argument against empiricism, another a priori argument, and an a priori argument against the thesis implied in empiricism that the relation between theory and empirical research is the same in every field of knowledge.
However appropriate the empiricist ideas may be in dealing with the natural sciences (and I think they are inappropriate even there, but I cannot go into this here),25 it is impossible to think that the methods of empiricism can be applicable in the social sciences.
Actions are the field of phenomena which constitutes what we regard as the subject matter of the social sciences. Empiricism claims that actions can and must be explained, just as any other phenomenon, by means of causal hypotheses which can be confirmed or falsified by experience.26
Now if this were the case, then empiricism would be first forced to assume—contrary to its own doctrine that no a priori knowledge about anything real exists—that time-in-variantly operating causes with respect to actions exist.
One would not know a priori which particular event might be the cause of any particular action. But empiricism wants us to relate different experiences regarding sequences of events as either confirming or falsifying each other. And if they falsify each other, then we are to respond with a reformulation of the original hypothesis. Yet in order to do so, we must assume a constancy over time in the operation of causes as such—and to know that causes for actions do exist is, of course, knowledge about the reality of actions. Without such an assumption regarding the existence of causes as such, different experiences can never be related to each other as confirming or falsifying one another. They are simply unrelated, incommensurable observations. Here is one, there is another; they are the same or similar; or they are different. Nothing else follows.27
In addition, there is yet another contradiction, and making it evident will immediately lead us to Mises’s central insight that the relationship between theory and history in the field of the social sciences is of an entirely different nature than that in the natural sciences.
What is this contradiction? If actions could indeed be conceived of as governed by time-invariantly operating causes, then it is certainly appropriate to ask: But what then about explaining the explainers? What about causally predicting their actions? They are, after all, the persons who carry on the very process of creating hypotheses and of verification and falsification.
In order to assimilate confirming or falsifying experiences—to replace old hypotheses with new ones—one must assumedly be able to learn from experience. Every empiricist is, of course, forced to admit this. Otherwise why engage in empirical research at all?
But if one can learn from experience in as yet unknown ways, then one admittedly cannot know at any given time what one will know at a later time and, accordingly, how one will act on the basis of this knowledge. One can only reconstruct the causes of one’s actions after the event, as one can explain one’s knowledge only after one already possesses it. Indeed, no scientific advance could ever alter the fact that one must regard one’s knowledge and actions as unpredictable on the basis of constantly operating causes. One might hold this conception of freedom to be an illusion. And one might well be correct from the point of view of a scientist with cognitive powers substantially superior to any human intelligence, or from the point of view of God. But we are not God, and even if our freedom is illusory from His standpoint and our actions follow a predictable path, for us this is a necessary and unavoidable illusion. We cannot predict in advance, on the basis of our previous states, the future states of our knowledge or the actions manifesting that knowledge. We can only reconstruct them after the event.28
Thus, the empiricist methodology is simply contradictory when applied to the field of knowledge and action—which contains knowledge as its necessary ingredient. The empiricist-minded social scientists who formulate prediction equations regarding social phenomena are simply doing nonsense. Their activity of engaging in an enterprise whose outcome they must admit they do not yet know, proves that what they pretend to do cannot be done. As Mises puts it and has emphasized repeatedly: There are no empirical causal constants in the field of human action.29
By means of a priori reasoning then, one has established this insight: Social history, as opposed to natural history, does not yield any knowledge that can be employed for predictive purposes. Rather, social and economic history refers exclusively to the past. The outcome of research into how and why people acted in the past has no systematic bearing on whether or not they will act the same way in the future. People can learn. It is absurd to assume that one could predict in the present what one will know tomorrow and in what way tomorrow’s knowledge will or will not be different from today’s.
A person cannot predict today his demand for sugar in one year any more than Einstein could have predicted the theory of relativity before he had actually developed it. A person cannot know today what he will know about sugar one year from now. And he cannot know all the goods that will be competing against sugar for his money in a year. He can make a guess, of course. But since it must be admitted that future states of knowledge cannot be predicted on the basis of constantly operating causes, a person cannot pretend to make a prediction of the same epistemological type as, for instance, one regarding the future behavior of the moon, the weather, or the tides. Those are predictions that could legitimately make use of the assumption of time-in-variantly operating causes. But a prediction about future sugar demand would be an entirely different thing.
Provided social and economic history can only come up with reconstructive explanations and never with explanations that have any systematic predictive relevance, another extremely important insight regarding the logic of empirical social research follows. And this amounts to another decisive criticism of empiricism, at least regarding its claim of being an appropriate methodology for social science research.
Recall what I said earlier about why it is that empiricism so strongly emphasizes the predictive function of explanatory theories. For every phenomenon to be explained there are a multitude of preceding events and a multitude of functional relationships with such preceding events by which the phenomenon in question could possibly be explained. But which of these rival explanations is correct and which ones are not? The empiricist answer was: Try to predict, and your success or failure in predicting future events will tell you which explanation is or is not correct. Evidently, this advice won't do if there are no time-invariantly operating causes with respect to actions. What then? Empiricism, of course, cannot have an answer to this question.
Yet even if actions cannot be predicted in any scientific way, this does not imply that one reconstructive historical explanation is just as good as any other. It would be regarded as absurd if someone explained the fact that I moved from Germany to the United States by pointing out, for example, that the corn in Michigan, prior to my decision, was experiencing a growth spurt and that this had caused my decision. But why not, assuming here that the event regarding Michigan’s corn indeed happened prior to my decision? The reason is, of course, that I will tell you that Michigan’s corn had no relevance for my decision. And insofar as anything is known about me at all, it can be recognized that this is indeed the case.
But how can you recognize this? The answer is by understanding my motives and interests, my convictions and aspirations, my normative orientations, and my concrete perceptions resulting in this action. How do we understand somebody and, moreover, how do we verify that our understanding is indeed correct? As regards the first part of the question—one understands somebody by engaging in a pseudo-communication and interaction with him. I say pseudo because, evidently, we cannot engage in an actual communication with Caesar in order to find out why he crossed the Rubicon. But we could study his writings and compare his convictions expressed therein with his actual deeds; we could study the writings and actions of contemporaries and thereby try to understand Caesar’s personality, his time, and his particular role and position within his time.30
As regards the second part of the question—the problem of verification of historical explanations—one has to admit from the outset that there is no absolutely clear-cut criterion that would allow one to decide which one of two rival explanations, both equally based on understanding, is definitely correct and which one is not. History is not an exact science in the same sense as the natural sciences are exact sciences or in the very different sense in which economics is an exact science.
Even if two historians agree in their description of facts and their assessment of factors of influence for a given action to be explained, they might still disagree on the weight that should be assigned to such factors in bringing about the action. And there would be no way to decide the matter in a completely unambiguous way.31
Yet let me not be misunderstood here. There is nonetheless some sort of truth-criterion for historical explanations. It is a criterion that does not eliminate all possible disagreements among historians, but that still excludes and disqualifies a wide range of explanations. The criterion is that any true historical explanation must be of such a kind that the actor whose actions are to be explained must, in principle, be able to verify the explanation and the explanatory factors as being those that contributed to his acting the way he did.32 The key phrase here is: in principle. Naturally, Caesar could not possibly verify our explanation for his crossing the Rubicon. Moreover, he might actually have strong reasons not to verify the explanation even if he could, since such a verification might conflict with some other objectives that he might have.
Also, to say that any true explanation must be verifiable by the actor in question is not to say that every actor is always best qualified to be his own explainer. It may be that Einstein can explain better than anyone else why and how he came up with the theory of relativity when he did. But this might not be so. As a matter of fact, it may well be possible that a historian of science may understand Einstein and the influences leading to his discovery better than he himself did or could. And this would be possible because the influencing factors or the rules that determined one’s actions might only be subconscious.33 Or they might be so obvious that one would fail to notice them simply on account of this.
The following analogy may be quite helpful in understanding the curious fact that others might understand a person better than the person himself. Take, for example, a public speech. Of course, to a large extent the person giving the speech can probably give reasons for saying what he says and formulate the influences that led him to see things the way he does. He can probably do so better than anyone else. And yet, in saying what he says, he follows rules habitually and unconsciously that he could hardly or only with great difficulties make explicit. He also follows certain rules of grammar when he says what he says. But quite often he would be completely unable to formulate these rules even though they clearly influence his actions. The historian who understands someone’s actions better than the person himself is quite analogous to the grammarian analyzing the sentence structure of the public speaker. Both reconstruct and explicitly formulate the rules that are actually followed, but that could not, or only with extreme difficulties, be formulated by the speaker himself.34
The speaker may not be able to formulate all the rules that he follows and may need the professional historian or grammarian to help him. But it is of great importance to realize that the truth criterion for the grammarian’s explanation would nonetheless be that the speaker would have to be able—in principle—to verify the correctness of the explanation after what was previously known implicitly was made explicit. In order for the grammarian’s or historian’s explanations to be correct, the actor would need to be able to recognize these rules as being those which indeed influenced his actions. So much for the logic of historical research as necessarily reconstructive research based on understanding.35
The argument establishing the impossibility of causal predictions in the field of human knowledge and actions now might have left the impression that if this is so, then forecasting can be nothing but successful or unsuccessful guessing. This impression, however, would be just as wrong as it would be wrong to think that one can predict human action in the same way as one can predict the growing stages of apples. It is here where the unique Misesian insight into the interplay of economic theory and history enters the picture.36
In fact, the reason why the social and economic future cannot be regarded as entirely and absolutely uncertain should not be too hard to understand: The impossibility of causal predictions in the field of action was proven by means of an a priori argument. And this argument incorporated a priori true knowledge about actions as such: that they cannot be conceived of as governed by time-invariantly operating causes.
Thus, while economic forecasting will indeed always be a systematically unteachable art, it is at the same time true that all economic forecasts must be thought of as being constrained by the existence of a priori knowledge about actions as such.37
Take, for example, the quantity theory of money, the praxeological proposition that if you increase the quantity of money and the demand for money stays constant, then the purchasing power of money will fall. Our a priori knowledge about actions as such informs us that it is impossible to predict scientifically whether or not the quantity of money will be increased, decreased or left unchanged. Nor is it possible to predict scientifically whether or not, regardless of what happens to the quantity of money, the demand for money to be held in cash balances will go up or down or stay the same. We cannot claim to be able to predict such things because we cannot predict future states of knowledge of people. And yet these states evidentiy influence what happens with respect to the quantity of money and the demand for money Then, our theory, our praxeological knowledge incorporated in the quantity theory, has a rather limited usefulness for one’s business of predicting the economic future.
The theory would not allow one to predict future economic events even if, say, it is an established fact that the quantity of money had been expanded. One would still be unable to predict what would happen to the demand for money And though, of course, concurrent events regarding the demand for money do affect the shape of things to come (and cancel, increase, decrease, accelerate, or decelerate the effects stemming from the source of an increased money supply), such concurrent changes cannot in principle be predicted or experimentally held constant. It is an outright absurdity to conceive of subjective knowledge, whose every change has an impact on actions, as predictable on the basis of antecedent variables and as capable of being held constant. The very experimenter who wanted to hold knowledge constant would, in fact, have to presuppose that his knowledge, specifically his knowledge regarding the experiment’s outcome, could not be assumed to be constant over time.
The quantity theory of money then cannot render any specific economic event, certain or probable, on the basis of a formula employing prediction constants. However, the theory would nonetheless restrict the range of possibly correct predictions. And it would do this not as an empirical theory, but rather as a praxeological theory, acting as a logical constraint on our prediction-making.38 Predictions that are not in line with such knowledge (in our case: the quantity theory) are systematically flawed and making them leads to systematically increasing numbers of forecasting errors. This does not mean that someone who based his predictions on correct praxeological reasoning would necessarily have to be a better predictor of future economic events than someone who arrived at his predictions through logically flawed deliberations and chains of reasoning. It means that in the long run the praxeologically enlightened forecaster would average better than the unenlightened ones.
It is possible to make the wrong prediction in spite of the fact that one has correctly identified the event “increase in the money supply” and in spite of one’s praxeologically correct reasoning that such an event is by logical necessity connected with the event “drop in the purchasing power of money.” For one might go wrong predicting what will occur to the event “demand for money.” One may have predicted a constant demand for money, but the demand might actually increase. Thus the predicted inflation might not show up as expected. And on the other hand, it is equally possible that a person could make a correct forecast, i.e., there will be no drop in purchasing power, in spite of the fact that he was wrongly convinced that a rise in the quantity of money had nothing to do with money’s purchasing power. For it may be that another concurrent change occurred (the demand for money increased) which counteracted his wrong assessment of causes and consequences and accidentally happened to make his prediction right.
However, and this brings me back to my point that praxeology logically constrains our predictions of economic events: What if we assume that all forecasters, including those with and without sound praxeological knowledge, are on the average equally well-equipped to anticipate other concurrent changes? What if they are on the average equally lucky guessers of the social and economic future? Evidently, we must conclude then that forecasters making predictions in recognition of and in accordance with praxeological laws like the quantity theory of money will be more successful than that group of forecasters which is ignorant of praxeology.
It is impossible to build a prediction formula which employs the assumption of time-invariantly operating causes that would enable us to scientifically forecast changes in the demand for money. The demand for money is necessarily dependent on people’s future states of knowledge, and future knowledge is unpredictable. And thus praxeological knowledge has very limited predictive utility.39
Yet of all forecasters who correctly forecast that a change such as an increase in the demand for money will take place and who equally correctly perceive that an increase in the quantity of money has indeed happened, only those who recognize the quantity theory of money will make a correct prediction. And those whose convictions are at variance with praxeology will necessarily go wrong.
To understand the logic of economic forecasting and the practical function of praxeological reasoning, then, is to view the a priori theorems of economics as acting as logical constraints on empirical predictions and as imposing logical limits on what can or cannot happen in the future.