Neoclassical Economics and Empirical Analysis Definition

Early attempts at empirical work were the exception, not the rule. In the late seventeenth century most classical economists followed the common-sense empiricism approach. They posited their laws about how the economy worked and supported those laws with examples. Because there was no accepted test of a theory, debates as to what theory was correct were ongoing.

With the beginning of neoclassical economics in the late 1800s, that approach was called into question, neoclassical theories were becoming more formal, and there was discussion of economics becoming an exact science. That meant formalizing economists' approach to empirical work and the approach that was most discussed was statistical analysis, which itself was undergoing a revolution.

Neoclassical economists followed many different approaches to statistical analysis. For example, Stanley Jevons saw statistics as a method of making economics into an exact science that could have exact laws. Leon Walras, on the other hand, had little place for empirical work; he continued to develop his theory independent of any chance of empirically testing it. Alfred Marshall believed in empirical work but did not conduct formal statistical analysis; he saw direct observation and common-sense empiricism as the most useful way to gather empirical information.

Henry Ludwell Moore Biography and Foundation, Henry Moore Information

In the late 1800s and early 1900s, significant work was done in statistical methods and probability theory that led to their introduction into economics. One of the earliest advocates of the use of formal statistical methods in economics was Henry L. Moore (1869-1958). In the early 1900s, Moore pioneered the use of many statistical approaches that would later become standard. Moore used the statistical work of Sir Francis Galton, Karl Pearson, and others. These statisticians had demonstrated that it was possible to formally determine infer­ences from statistical data in a controlled environment using multiple correlation and contingency tables. Moore was impressed with that work and decided that it was possible to apply these statistical methods to verify economic theories.

Rather than just "eyeballing" two graphs on the same grid, as Jevons did, Moore formally compared two series of data and developed statistics that gave him information about the relationships between the two. It is important to note, however, that in doing so he made a heroic jump from the work of Pearson, who analyzed work conducted in an environment in which other physical influences could be controlled. Moore did not have that luxury, because controlled experi­ments are usually impossible in economics; he was thus assuming that statistical methods developed for use with controlled experiments would work in an uncontrolled environment.

The theory he was specifically interested in testing was J. B. Clark's marginal productivity theory of wages, which predicted that individuals would be paid their marginal product. Toward that end Moore investigated the relationship between wages and marginal productivity, personal ability, strikes, and industrial concentration. Clark's theory implied that (1) higher-ability individuals would be paid more than low-ability individuals; (2) when like-ability individuals worked in monopolistic industries and competitive industries, those in monop­olistic industries would be paid more; and (3) strikes for higher wages were more likely to be successful in concentrated than in unconcentrated industries.

Moore found a relationship between ability and wages, but there were significant problems with his analysis. In his tests Moore did not specify his theoretical structure very rigorously. For example, in one test he used average product rather than marginal product and tested not real wages but money wages. Moore also found a relationship between strikes and industrial concen­tration, but that relationship was based on limited data.

Moore's statistical work was also problematic because he was interested in more than simply scientifically testing Clark's theory. Moore had a strong interest in policy-related issues. He wanted to use his statistical analysis to argue against socialist policy proposals calling for more equality of income. He claimed that if marginal productivity theory could be proved true, then he could scientifically demonstrate to others that there were serious problems in moving to a socialist state bent on changing the distribution of income. Having an "axe to grind" does not necessarily invalidate the results of theoretical or empirical work, but it does raise questions as to whether ambiguous results will be interpreted fairly. Discovering the motives of a theorist or econometrician is not a test of the validity of a hypothesis or theory. Research sometimes takes place in funded think tanks reflecting particular segments of the ideological spectrum; as long as the results become public property that can be examined by all, significant bias is likely to be pointed out.

Moore's early work established him as a leader in integrating statistical methods with economics. His subsequent contributions, one on the empirical measurement of the demand curve and the other on the measurement of business cycles, are also important: the first forms the foundation for modern micro-economic econometrics, and the second forms the foundation for modern macroeconomic econometrics.

Moore's Demand Curve and the Identification Problem

Moore is probably best known for his work on the estimation of demand curves for agricultural goods and for pig iron. A careful analysis of his contribution is warranted, because it points out many of the problems with empirical estimation that play roles in later debates.

Consider the difficulty of empirically measuring a demand curve. Market observations are of combinations of prices and quantities at which trades take place. If the market is in equilibrium, the observed prices and quantities are points on both the supply and demand curves; if the market is not in equilibrium, the observed prices and quantities could be on the supply curve, on the demand curve, or on neither curve. How can the researcher know which is the case? If the researcher could do a controlled experiment and hold everything else equal, as in the equations

Qd = f(RPq,, price of all other goods, tastes, income,...)
Qs = g(PQ,, prices of factors of production, technology,. . .)
Qe = h(f,g)

where everything but the price and quantity is being held constant, then one could measure the true relationship between price and quantity. But where that

cannot be done (and with economic statistics it cannot be done), the researcher must somehow relate the observed data on prices and quantities to the theory. Therein lies the heart of the econometric problem: relating observed data from uncontrolled experiments to theory.

The Equilibrium Assumption

In his analysis of agricultural markets, Moore was willing to accept that the markets would move to equilibrium, so that the observed prices and quantities could be assumed to be equilibrium prices and quantities, Pe and Qe, and that they would be points on both the supply and demand curves. This assumption can be seen in Figure. It lets us assume that the observed point is a point such as (Pe, Qe) rather than a point (P1, Q1) where the market is in disequilibrium in the process of adjusting to equilibrium.

Moore was also willing to assume that for agricultural commodities, supply was determined exogenously by summer rainfall and therefore would be unaf­fected by price in the current harvest period. He further implicitly assumed that past events had no effect on supply and demand and that changing expectations played no role in determining the actual data. These assumptions changed the graph of the model to one represented by Figure. Because quantity supplied is assumed to be determined exogenously, the estimated points (P1, Q1) and {P2, Q2) must be points on the demand curve.

In carrying out his analysis, Moore expressed the data as percentage changes around a trend and derived his demand relationship in terms of percentage changes. He proposed both a linear and a cubic equation for his demand curve. A linear demand curve would have the general form of P = a - bQ, where P is price, a is the price intercept of the demand curve, b is the slope of the demand curve, and Q is quantity. The negative sign for the b coefficient indicates a downward-sloping demand curve. Moore estimated two different curves with the following coefficients:

Moore’s Assumption About Exogenous Supply

AP/Pt-1 = 7.8 - 0.89AQ/Qt-1
R2 = 0.61,s=16


AP/Pt-1 = 1.6- 1.1AQ/Qt-1 + 0.02(AQ/Qt-1)2 - 0.0002(AQ/Qt-1)3
R2 = 0.71, s=14

Notice that in both cases the demand curve has the negative sign predicted by theory (it is downward-sloping) and that there is a fairly high coefficient of determination.

Moore's estimated demand curve did not bring him immediate acclaim; many did not understand his accomplishment, and others (such as Edgeworth) who did understand asserted that the empirical demand analysis was far too simple, given the complexity of the underlying theory. Edgeworth maintained that the many untested assumptions that underlay the conclusions were so great that the formality of the estimate gave little benefits. These criticisms, although substan­tial, are still leveled in various degrees against econometric work and should not demean Moore's contribution. He was one of the first economists to measure a demand curve statistically, although, as Nancy Wulwick points out,3 it is not clear that Moore intended to estimate a traditional demand curve.

The cool reception of Moore's estimate of an agricultural demand curve was equivalent to positive adulation compared to the frosty reception of his estimate of demand for pig iron. He claimed the demand curve for pig iron to be positive-sloping, so that when price went up, the quantity demanded went up. He proposed the following demand equation:

AP/Pt-1 = 4.48 + 0.5211AQ/Qt-1

Moore's claim of having discovered a positively sloping demand curve went directly against microeconomic theory and provoked strong critical re­sponses.

Given Moore's sophistication as an economic theorist, it is now suggested by Wulwick that Moore's positively sloping demand curve was not the result of error or failure to understand the identification problem (the need to hold supply or demand constant in order to estimate the other curve). According to Wulwick, it represented an attempt to address the data limitations and allow those limitations to direct his analysis, rather than letting the theoretical analysis direct his empirical work. This view is supported by the fact that, in his writing, Moore was clear that his demand curve was not a typical demand curve that followed from Marshallian theory but was, instead, a dynamic demand curve that related empirical regularities involving many interactive changes.

A number of interrelationships could make his dynamic demand curves consistent with static demand theory. For example, when the price of pig iron went up, aggregate income and economic activity were likely to increase, which would be associated with an increase in demand. Because it was impossible to exogenously specify the supply of pig iron, as would be necessary to estimate a static demand curve, Moore contended that his dynamic demand curve, which captured an empirical regularity, would be a useful tool in making predictions about the economy.

Moore argued that even though one could not exogenously specify supply, one could nonetheless estimate a curve that incorporates normal reactions to interrelated shifts in supply that can be measured. These normal reactions can include shifting the static demand curve in a time-consistent manner; they can make the measured dynamic demand curve incorporating these interdependen-cies upward-sloping. If these later relationships are true, then whenever we see the supply of major industrial goods exogenously increasing, we should expect the price of these industrial goods to rise, not fall. That was Moore's conclusion. Moore felt little need to relate this dynamic demand curve to underlying static theory, because doing so would be only an exercise and would not be convincing. He wrote:

According to the statical method, the method oicaeteris paribus, the proper course to follow in the explanation of the phenomenon is to investigate in turn theoreti cally, the effect upon price of each factor, caeteris paribus, and then finally to make a synthesis! But if in case of the relation of each factor to price the assumption caeteris paribus involves large and at least questionable hypotheses, of final completely lose himself in a maze of implicit hypotheses when he speaks of a final synthesis of the several effects? We shall not adopt this bewildering method, but shall follow the opposite course and attack the problem of the relation of prices and supply in its full concreteness.

The fruitfulness of the statistical theory of correlation stands in significant contrast to the vast barrenness of the method that has just been described, and the two methods follow opposed courses in dealing with the problem of multiple effects. Take, for example, the question of the effects of weather upon crops. What a useless bit of speculation it would be to try to solve, in a hypothetical way, the question as to the effect of rainfall upon the crops, other unenumerated elements of the weather remaining constant? The question as to the effect of temperature, caeteris paribus} How, finally, would a synthesis be made of the several individual effects? The statistical method of multiple correlation formulates no such vain questions. It inquires, directly, what is the relation between crop and rainfall, not caeteris paribus, but other things changing according to their natural order.

The problems addressed by Moore's justification of his work relate to some still unresolved issues in econometrics. They provide perspective on the empirical approach of the institutionalist school, which argued that the data should direct the theoretical analysis rather than the theory's directing the empirical work. Recent justification for Moore's work has developed in an atmosphere in which the profession is much more aware of the limitations of static analysis and relating that analysis to empirical observation. Such justification was not forthcoming in Moore's time or in the mid-1900s. Moore was attacked from both sides—by those against formal theoretical and empirical work, who felt his statistical methods were too complicated, and by those in favor of formal theoretical and empirical work, who felt that he did not pay enough attention to theory.

The ridicule he endured about his upward-sloping demand curve ultimately led him to abandon his econometric work, although not without having left his mark on the profession. It remained for his students to carry forward the empirical revolution. Of these students, the most famous was Henry Schultz, whose Statistical Laws of Demand and Supply (1928) and Theory and Measure­ment of Demand (1938) would play major roles in the development of modern microeconometrics.

Henry Schultz and Independent and Dependent Variables

Henry Schultz's (1893-1938) contribution came as a derivative of his analysis of tariffs, which required him to estimate a demand curve. As Schultz was attempting to do so, he made an interesting discovery: one could obtain quite different elasticities by regressing quantity on price rather than price on quantity, as Moore had done. In discussing these issues, Schultz argued that if one had a prior belief about which variable is the correct one to regress (which variable is dependent and which independent), that would determine the correct choice. If, however, one did not have a prior belief, there was no way of choosing between the two. In such a case, Schultz argued, it was best to choose the regression that had the better fit as determined by a Pearson chi-squared test.

Schultz's insight was an important one; it means that statistical measurement cannot be considered independently of theory. What you see is partly determined by what you believe. This insight led to the current practice in econometrics that requires researchers to carefully distinguish independent from dependent vari­ables.

Of course, to say that statistical measurement changes in relation to theory is not to say that measurement is totally dependent upon theory. It does not say that theory is determinant; it simply provides a limited range of interpretation that one can draw from statistics.