The Science of Demand (8) - Unofficial Translation of Steven Cheung's 经济解释 - 科学说需求

Section 8.  Scientific Methodology in Economics

If readers find certain areas in this chapter difficult to comprehend, don’t be upset. The methodology of science involves logic and the theory of knowledge in philosophy. These are close to the most profound knowledge in mankind’s cultural history. Though I had studied under an expert, my knowledge in this area is limited, thus am not necessarily capable of thorough explanation using simple words. So highly sophisticated is the methodology of science that experts in logic do not necessarily agree among themselves, while achievements in science often bear no relationship to this knowledge. There are indeed quite a number of scientific experts who know nothing about the methodology of science. On the other hand, specialists in the methodology of science are seldom accomplished scientists. Logic often probes to the extreme of the ivory tower, marvelous though at its sophisticated realm, a lot of sacrifice is always required in getting there.

From the viewpoint of rigorous philosophical logic, what I know of is only sketchy. It was fifty years ago when I worked on this knowledge. However, scientific methodology can also be viewed from another perspective – the empirical approach linking abstract theory and the real world. This I know more. The content of this chapter, combining philosophical logic with empirical linkage, is different from the methodologies referred to in common textbooks. Ultimately, only out of the ivory tower can science be practical.

I have named “scientific methodology” as the title of this chapter and indeed elaborated on it. It is not that this knowledge has any indispensable significance on this book. Rather, China’s cultural traditions often talk about benevolence, brotherhood and morality, yet without the spirit of empirical testing, there is a deep-rooted misunderstanding of the nature of science. In the twentieth century, the influential “Three Principles of the People” and “Marxism” – or other doctrines – added an opaque membrane to our students’ understanding of science. As earlier mentioned, this book is written for Chinese students. I consider scientific methodology having a more profound impact on Chinese than on certain Western ethnic groups. Do not drill into the dead end of the methodology of logic, albeit a rough grasp is needed.

Economic explanation is an empirical science. Its nature is identical to that of natural science, with both adopting identical scientific methods. However, in terms of the nature of content, economics is rather different from natural science, hence when approaching issues, their scientific methods have different emphases. There are two reasons. On one hand, economics’ laboratory is the real world which is neither built by economists nor controlled by researchers, thus its difficulty in observations is distinct from that in natural science. On the other hand, economics is for explaining human behavior, but economists are also human, therefore it is inevitably used to a certain extent in explaining themselves. Objective judgment is thus more difficult than in natural science.

In terms of methodology, the focus of economics is different from that of natural science. First, I believe economics should not be so deeply influenced by physics. As aforementioned, the so-called equilibrium and disequilibrium are real phenomena in physics, whereas equilibrium and disequilibrium do not exist in the real world of economics. As explained before, they are concepts at best. Certainly, some economists consider that equilibrium and disequilibrium refer to observable market phenomena. This is an embarrassingly fatal mistake. Readers should note that when I mention equilibrium in “Economic Explanation”, I mean sufficient constraints are in place in deriving a refutable hypothesis, not that there exists certain observable equilibrium point.

I regard as unimportant the use of mathematics in economic explanation, though professional economic articles nowadays apply mathematics even more than those in physics. Besides physics, other natural sciences seldom employ mathematics. I am not saying mathematics is not useful for economics, yet mathematics is no economics. Mathematics is a miraculous language: the logic of every solvable equation must be correct. However, correct logic does not necessarily match with correct content. Some people are adept in using equation to deliberate, I on the other hand regard that as a barrier hindering my emancipated style of thinking. Though not using mathematics, my logical reasoning is seldom wrong. I believe our fellow students should learn more mathematics, but in deliberation ought to consider which section of our own brain is more superior. With more practice, deliberation without using mathematics can leap to and fro more readily, hence is way better in respect of imagination.

There is a common misconception. Some people consider it more accurate using mathematics or statistical equation for reasoning or testing. This is not correct. Measurement is the ranking of numbers, and accuracy in measurement depends on common acceptance of such ranking. This is another topic of philosophy, and I will demonstrate how this is to be approached when analyzing transaction costs in Volume III.  

Second, I have not undertaken hypothesis testing in natural science, yet hypothesis and testing in economic explanation always start from changes in constraints, which is equivalent to starting from changes in test conditions. Saying the occurrence of “A” implies the occurrence of “B” is in fact saying changes in “A” will lead to changes in “B”.

As earlier said, to predict is the same as to explain, though they differ as to ex-ante or ex-post. To predict is to infer what phenomenon will occur after noting changes in constraints; whereas to explain is to trace the cause of the occurrence of a phenomenon to identify the changes in constraints. Logical structures being the same, to predict is therefore the same as to explain. Given different approaches in investigation, it is hard to judge which method is more demanding. Let’s have a think. Noting a phenomenon, in order to explain, we need to trace its cause to identify the changes in constraints. But there are countless constraint changes in this world, which item or which combination of items should we choose? Tracing changes in constraints ex-ante requires theoretical guidance – not an easy task. What about prediction? Noting changes in constraints, we therefore theorize to infer what phenomenon would follow suit. The problem is that changes in constraints ex-ante are not necessarily stable. They could keep on changing to overturn the prediction which was originally failsafe. The inference I made in 1981 that China would go down the road of market economy was based on certain observable and reasonably stable changes in constraints. Nevertheless, to have been correct, needed was God’s blessing in keeping the changes stable.

This raises another important yet related topic. The aforementioned ex-ante inference or ex-post explanation requires a good application of underlying theories and concepts. However, in the so-called applied economics textbooks that we find, their approach is first proposing a theory, then stuff into it real-world examples. This basically in search of “right” contravenes the principle of empirical science: in search of “wrong” in the hope of not being refuted by facts. In a stricter sense, empirical science looks for refutability, i.e., a theory or hypothesis has to be potentially refutable by facts. Refutable and testable are the same. Having different starting points and different intentions, it is therefore no easy matter to learn a great deal from these “applied” textbooks.

The last point is “if invisible, then non-testable”. Economics these days frequently contravenes this simple philosophy, leading to disastrous developments. If we say the occurrence of “A” will lead to the occurrence of “B”, both A and B must exist or at least be visible or touchable in principle. We have said that the starting point of a theory is often abstract, allowing the existence of in-principle non-observable variable. We should avoid non-observable or non-realistic variables as much as possible. After years of investigation, there is only one unreal variable we have to accept as inevitable: quantity demanded. My research has gone through thousands of miles, and this is the only exception.

Throughout today’s developments in economics, there are innumerable non-observable variables or behavior: game, motive, shirking, blackmailing, threatening, concealing, laziness, opportunism, etc. Since in practice they are non-observable, non-measurable and non-testable, only God knows if they exist. As stories, they can be told very logically and credibly, and may contain certain religious flavor. Yet invisible means non-testable. Science that cannot be tested has no explanatory power.

The Science of Demand (7) - Unofficial Translation of Steven Cheung's 经济解释 - 科学说需求

Section 7.  Realism of Theory

Since facts are not explainable by facts, using theory to explain phenomenon must to a certain degree be abstract. Abstract concept is non-factual. This led many people to consider theory, having lost touch with reality, merely empty rhetoric with no applications. “Realism” therefore grew to be a big controversy. Today such controversy has died down, yet this issue deserves to be clarified.

“Reality” has several meanings. If we are unclear as to which aspect we are focusing on, controversy will never end. Abstract concept is certainly not factual, therefore it is acceptable to say that “theory” is not reality. But the ultimate goal of a theory with explanatory power involves its empirical testing and prediction. We can therefore also say that practical theory has its share of reality. For certain theories, no testable implications can be derived (e.g., the various economic development theories in the 1950s and 60s), hence they could at best be counted as “games” that have no connections with the real world whatsoever.

However, there are at least four denotations in respect of the non-realism of theories with explanatory power, three of which are very trivial. First, the theory itself must have its abstract component. To say it is not reality is certainly right, but it is wrong to say since it is not reality, it has no explanatory power. Given facts cannot explain facts, without an abstract starting point, real-world phenomena generally cannot be explained. Second, portrayal of all facts or observations must be simplified – such simplification makes fact “unreal”. This is a middling yet finicky viewpoint. Let’s use apple as an example. Suppose we need to comprehensively portray an apple, we will fail even after exhausting all the papers in the world. Just portraying the colors and shapes of the apple – not to mention its tastes or the vitamins it contains – is difficult to be exact! Under a finicky yardstick, no portrayal of a phenomenon or a fact in the whole world is reality. However, criticizing empirical science this way – such people do exist – is not scientific.

The third type of non-reality also comes from simplification. The world is so complicated that simplifying assumption (different from the notion of an abstract assumption) is necessary. But the objective of simplification is purely for ease of handling. Since simplification has no apparent effect on the outcome, it is therefore allowed. For instance, let’s assume there are only two countries in the world (in fact more than this, so not reality) and examine the outcome when they trade with each other, etc. Changing two countries to three or four would yield roughly the same outcome. For certain peculiar topic, however, changing two to three would yield different outcome. So in researching such peculiar topic, the distinction between two and three cannot be ignored, though some other simplification is also necessary.

The last type of non-reality is non-trivial. The aforementioned additional test conditions are treated by many as a kind of assumption. Such assumption would certainly become unreal due to simplification, yet we cannot treat that as a castle in the air like abstract thinking, and sever ties with the real world. The assumption of test conditions must be traceable, and test conditions in their simplified form must be essentially congruent with reality. For instance, given that a chemistry experiment requires a clean test tube (clean is a test condition), we cannot use a dirty test tube and assume it is clean.

In economics, test conditions are usually called constraints. Economics does not have theories “with no constraints”. Like other scientific theories, economic theories necessitate test conditions, otherwise they would have no explanatory power. Suppose we say, under a situation of no transaction costs (a constraint, may be grudgingly called an assumption), the occurrence of “A” would lead to the occurrence of “B”. To test this implication, we must work under a real situation with negligible transaction costs. In other words, the “assumption” of constraints cannot lose touch with the real world. That is, other than inevitable simplification, test conditions must have their realism.

We can therefore draw the following conclusion. For scientific theory using abstract thinking or concept as a starting point, since fact is not self-explanatory, it is necessary to be “unreal”. “Impossibility to be too detailed” and “simplification” are allowed. Yet test conditions losing touch with the real world is a fatal mistake. Truthful investigation and simplification of constraints (test conditions) are the most arduous processes in economic explanation. Real-world phenomena are like chess games in that every game is novel. It is only common that a few years’ effort is needed before getting a little basic knowledge in certain constraints. People age as time flies, therefore economists who undertake empirical studies often have to be sure of the importance of the issue before committing their last bet.

In the great methodological debate during the 1950s – 60s, there was an embarrassing fallacy on the realism of theories. That was: if the occurrence of “A” leads to the occurrence of “B”, then we can go on to say if “Not B”, then “Not A”. However, we cannot say if “Not A”, then “Not B”. The fallacy of the latter has earlier been discussed. In that great debate, many economists forgot about this first lesson in logic that “Not A” has no impact on what “B” would be. The gentleman who investigated Boston’s transportation companies considered the hypothesis of “A” unreal, then made a big fuss to predict how “B” would be. Such mentally-deficient analysis was originally not worth responding to, yet so inexplicable was scientific advancement that feedbacks from numerous scholars led to a vastly constructive debate.

The Science of Demand (6) - Unofficial Translation of Steven Cheung's 经济解释 - 科学说需求

Section 6.  Non-factual and Unrestrained

I have repeatedly stated the significance of “a theory refutable by facts”. I have also pointed out that tautologies, ambiguous, or contradictory theories are not refutable by facts. There remain two other types – of “theories” without explanatory power – that are not refutable by facts. One is when the phenomenon used for empirical testing is non-factual; the other is when the phenomenon inferred to occur is unrestrained.

Suppose I say: “If it rains, then there are clouds in the sky.” In this statement, “rain” and “clouds” are factual and observable. If “rain” and “clouds” are merely castles in the air and non-factual, then the statement cannot be tested. This embodies a non-trivial philosophy of empirical science. For every inference with explanatory power, its testing must have the following implication: if “A” happens, then “B” follows suit – and both “A” and “B” are observable facts. No matter how costly or how much time it takes to do the testing, in principle at least, the existence of “A” and “B” have to be confirmable. Albert Einstein’s theory of relativity involved the gene theory in genetics. Though its implication was initially difficult to be empirically tested, it was nonetheless confirmed subsequently.

The question is, as aforementioned, facts cannot be explained by facts. The occurrence of “A” cannot explain the occurrence of “B”. Any regularity between “A” and “B” can only serve to confirm the implication of a certain theory. No matter how abundant the facts are and how obvious the regularity is, they are not self-explanatory. Therefore, a theory with explanatory power often starts from abstract thinking and non-factual postulates, then through logical reasoning to derive testable implication – the latter is the “If it rains, then there are clouds” hypothesis.

This is no easy project. A testable implication has to be refutable by facts; however, facts are not self-explanatory, and abstract theory itself cannot be tested. In the subtle transition from abstract reasoning to empirical testing, the difference in capability between a master and a mediocrity will be clearly revealed.

Let me quote a fundamental example. In economics, the renowned law of demand says: if the price of a good falls, its quantity demanded will increase. Price and its movement are observable, but quantity demanded is non-factual! Quantity demanded, referring to consumers’ desired or intended demand, is an abstract term. Therefore, the law of demand itself cannot be empirically tested. However, this law is indispensably significant in economics. The average person often treats quantity transacted in the market as quantity demanded. Similar to treating a deer as a horse, this is undoubtedly wrong. The correct approach is very different. We should say: if the law of demand is right, then by logical reasoning, under certain observable circumstances, the occurrence of “A” will lead to the occurrence of “B”, and both “A” and “B” are observable facts (this is a testable implication inferred by the non-testable law of demand). If “A” occurs yet “B” does not occur, then the law of demand is highly problematic – it either requires addition of other conditions, or be counted as refuted by facts. If “Not B” implies “Not A”, then the law of demand is not refuted, and can be interpreted as having explained the regularity between “A” and “B”.

Indeed – if amazingly done, such prediction and empirical testing can be extremely laudable. This is the beauty of science. In this book I will go to great lengths to demonstrate the astonishing explanatory power of the law of demand. The aforementioned additional conditions could be ever-changing, aplenty or just a few. In scientific methodology, additional conditions are called test conditions, while in economics they are termed constraints. Sometimes we could say the occurrences of “A” and “B”, or the occurrence of “A” or “B”, will lead to the occurrence of “C”. We could also say the occurrence of “A” will lead to the occurrences of “B” and “C”. These variables could be aplenty or just a few; could occur simultaneously; one or several of them could occur in several or many possible observations. These all fulfill the requirements of theories with explanatory power. No matter how many phenomena are involved in testing, one restraint is necessary.

Suppose the occurrence of “A” will lead to the occurrence of an infinite list of “B”, or “C” or “D” or “F” …, then such an implication is neither deniable nor refutable. Strictly speaking, this is the so-called disequilibrium in economic theory. Equilibrium, on the other hand, is attained when a refutable implication is derived from restrained hence definite phenomena.

The aforementioned concepts of “equilibrium” and “disequilibrium” are different from the traditional equilibrium theory adopted by economists. I consider the traditional concepts fundamentally wrong. “Equilibrium” in traditional economics was copied from physics. Equilibrium in physics refers to a pendulum, after oscillating, reaching a static state at the center; or an egg, after rolling on the floor, reaching a stationary point; or an incessant object, after entering an orbit, displaying regularity. This type of “equilibrium” refers to certain phenomena that are indeed observable.

“Equilibrium” in economics is a different matter. For instance, economists say that the intersection of the demand curve and the supply curve is an equilibrium point. Yet there is neither demand curve nor supply curve in this world – these are merely conceptual tools conceived by economists. Without economists, these conceptual tools will not exist. Similarly, “equilibrium” and “disequilibrium” in economics are merely concepts that do not exist in the real world. Being neither phenomena nor facts, they cannot be observed.

In the spring of 1969, Coase and I drove from Vancouver to Seattle. In the over two-hour journey we debated about the “equilibrium” concept in economics. He considered “equilibrium” and “disequilibrium”, being castles in the air, should be abolished. I assented to the viewpoint of castles in the air, but considering so popular were “equilibrium” and “disequilibrium” in economics, I could provide a little remedy for these concepts.

I pointed out to Coase that “disequilibrium” could be interpreted as a theory lacking a refutable implication since inferred phenomena are not restrained; whereas “equilibrium” refers to a testable theory since inferred phenomena are restrained. This is the difference between the aforementioned “unrestrained” and “restrained”. Coase at that time agreed this interpretation could salvage the economic “equilibrium” and “disequilibrium” concepts which ought to have been trashed. That was more than forty years ago. Today, the number of economists who understand and concur with this concept can be counted on the fingers of both hands.

Abstract theory itself cannot be empirically tested. In order to have explanatory power, abstract theory must possess one or more implications that can be tested. For implication to be testable, it must be refutable by facts. The implication can have many additional conditions and inferences of phenomena, and they can be linked by the affirmative “and” or the not-so-affirmative “or”, though they cannot be infinite. Undoubtedly, the affirmative “and” has stronger explanatory power than the not-so-affirmative “or”, and the simpler the abstract reasoning and test implication, the more convincing they are. Gifted scientists are capable of marvelously simplifying complexities.

The Science of Demand (5) - Unofficial Translation of Steven Cheung's 经济解释 - 科学说需求

Section 5.  Ambiguity and Contradictory

For a theory to be capable of explaining phenomenon, it must be refutable by phenomena (facts). This is a maxim in empirical science. I have elaborated earlier that a “theory”, like tautology, that cannot be falsified has no explanatory power since it is not potentially refutable by facts. However, other than tautology, four situations could render a theory not refutable by facts. We’ll explore the first two here; in the next section we’ll discuss the remaining two.

The first one is what I teasingly term “the second Coase theorem”. In his 1960 seminal article (from which the “Coase theorem” was originated), Coase advocated a familiar yet had never before been distinctly proposed philosophy. After exhausting all possible means to comprehend Arthur Cecil Pigou’s economic analysis, but still unable to figure out what it referred to, Coase wrote: “An idea not clearly stated can never be proved clearly wrong.”

Indeed – ambiguous concept or analysis, being not clearly wrong, is impossible to be clearly refuted by facts. To be refutable by facts, a precondition is: the theory itself has to clearly indicate a possibility of being falsifiable. “If it rains, then there are clouds” can be falsified (but has never been falsified); “if it is spring, then a bud blossoms” can be falsified (but has never been falsified, either). Yet if we are uncertain what clouds are or what spring is, how can we determine right from wrong?

There are lots of ambiguous concepts in economics. Theories not susceptible to refutation by facts – impossible to be clearly refuted by facts – emerge one after another. Karl Marx’s “Capital” is an example. What exactly is residual value? Some scholars say it is rent, some say interest, some say profit, while some say there is nothing like that at all. Despite all the rhetoric, it is still ambiguous. According to Marx, “residual value” was the residual after capitalists had paid wages. However, other production costs had not been completely deducted, how could that be said to be residuals after exploiting workers? The “capital” concept in the “Capital” was also ambiguous. It was only until 1930s that such concept was clearly explained by Irving Fisher (see Volume II of this book).

No one has ever tried to empirically test Marx’s theory. It is not surprising that there was no verification in China then, but why didn’t Western scholars put Marx’s theory to the test? The answer is: ambiguous theory cannot be empirically tested. Very unfortunately, a non-falsifiable theory is often believed by some blind adherents as “not falsified means absolutely right”.

Ambiguous concepts or theories are of course not unique to Marx. David Ricardo, an earlier genius who had an enormous influence on Marx, was confused about the concepts of “capital” and “cost”, resulting in his analysis on wages and rent not understood by later generations. Modern gurus like Frank Knight – with five of his students winning Nobel Prizes in Economic Sciences – also fell into the trap of ambiguity. Knight split risk and uncertainty into two, yet after much deliberation we still cannot tell the difference.

John Maynard Keynes’ “The General Theory of Employment, Interest and Money” was ambiguous, too. Consequently, for certain important areas of that theory, no one could boast about having done any verification. The originator of the utility theory, Jeremy Bentham, subjectively using utility as a proxy for happiness, baffled later generations. Therefore Alchian became famous by asking the question “what is utility?”. Bentham’s utility theory, being ambiguous, is not refutable by facts. However, after Alchian, studies on testing of the utility theory started cropping up here and there. (Being Alchian’s student inside the chambers, I refrain from applying this utility concept for explanation. This will be elaborated later.)

Since ambiguous concepts or analyses cannot be proved clearly wrong, they possess no explanatory power. Another type of theory that is not refutable is the meaningless type. Meaningless refers to neither devoid of content (unlike tautology) nor ambiguous, but statements of this type are contradictory and logically inconsistent, puzzling people as to their meaning, hence they become meaningless.

Let’s cite a few examples. If I say: “There are black dots on an all-white wall.” This sentence is not devoid of content. In fact it is extremely clear, though “black dots” and “all-white” are contradictory and cannot co-exist. This sentence is therefore meaningless. Logic can certify that an all-white wall that has black dots could lead people identifying a deer as a horse, or a wall as God! (This is no trivial logical reasoning, but since it is outside the scope of economics, we will not elaborate here.) Contradictory statement can have content, can be crystal-clear, yet cannot have meaning.

There are plenty of contradictory theories in economics. Like tautologies, contradictions may not necessarily be readily detected. My thesis, “The Theory of Share Tenancy”, overthrew the previous view by pointing out its contradiction. For instance, the theory of Charles Issawi was based on every individual maximizing his self-interest, yet he wrote: “In this paper I implicitly assume that the landlords will not respond speedily to economic gains, nor attempt to increase investment in order to increase their income.” What is it if not contradiction? And in a separate case, in guru Marshall’s analysis of share tenancy, he did not allow the landlords to choose the system of fixed rent, even though he knew well the income of fixed rent was higher than that of share tenancy, and the two systems could co-exist.

Similar contradictory analyses are often found in the publications of economic masters. William Baumol said that a monopolized enterprise did not seek the highest profit but the highest sales, yet his theory did not allow the enterprise to give up a little sales for a much bigger profit. John Hicks pointed out that when a person’s income increased, his demand for certain goods would decrease. This is not incorrect, but in his analysis, the model that he used was a world with only two kinds of goods, and in that particular world, any increase in income would not cause the quantity demanded of one of the two goods to decrease. Contradictory problems are commonly seen in any science, and economics is no exception. Direct contradictions are not difficult to identify, yet indirect ones – those after one or multiple inferences – often cannot be avoided even by masters.