Many government-produced statistics carry the label “seasonally adjusted.” What does this mean? It indicates that the statisticians, based on a wrongheaded theory of macroeconomic fluctuations, have swept the data clean of certain variations that are said to distort the underlying reality economists are seeking to study. But neither the theory nor the practice of seasonal adjustments have a solid basis in logic or sound economics.
First, some background. Economic data observed over time are said to be determined by four factors: the trend factor, the cyclical factor, the seasonal factor, and an irregular factor. The trend is said to determine the general direction of the data over time, while the cyclical factor causes movements that are related to the business cycle. The influence of the seasons is simply that: the manner in which weather and various holidays affect the economy, while the irregular factor encompasses anything left out of the other three.
The idea is that interplay of these four factors generates final data. But one factor is said to be more important for policy purposes than all the rest, and thus more crucial for economists to discover: the cyclical influence. The purpose of “seasonal adjustments” is somehow to eliminate the part of the data that cloud our vision of cycles. The ultimate goal is to unravel the mystery of the business cycle.
Moreover, it is said to be important to observe the influence of the cyclical factor within periods of time that are as short a duration as possible. As with any disease, early detection is said to give policy planners a better chance of combating the disease. Thus once the central bank has identified the size of the cyclical influence it can offset this influence by means of a suitable monetary policy.
In economic statistics, monthly fluctuations of data are dominated by the influence of the seasonal factor. As the time span increases, from one more to one year, the importance of the cyclical influence rises while the influence of the seasonal factor diminishes.
Hence, the cyclical influence will be more powerful in quarterly data than in monthly data. The trend, it is assumed, exerts a strong influence on a yearly basis while having an insignificant effect on the monthly variations of the data. The irregular factor of economic shocks is the “wild card” but it produces result of a short duration. Consequently the effect of positive shocks is offset by negative shocks.
It follows that in order to be able to observe the influence of the business cycle on a short-term basis, all that is required is to remove the influence of the seasonal factor. The method of the removal however, must work towards presenting a crystalized picture of the cyclical factor, meaning the data reveal “underlying fluctuations” must remain unaffected.
Most economists regard the seasonal effect as constant. For example, every year people buy warm clothes before winter arrives. People tend to spend a larger fraction of their incomes before Christmas. The assumption of constancy means that removing the seasonal influence will not distort the influence of the cyclical factor. This in turn will presumably permit an accurate assessment of the business-cycle effect. Economists generate numbers for each month that provide an estimate of the seasonal effect. These data are then subtracted out the raw data.
The statistical and mathematical methods are very complicated in practice. Currently most government statistical bureaus world-wide use the computer programs X-11 and X-12 to make the estimate of seasonal influence by means of sophisticated moving averages. The computer program then uses the obtained estimates to de-seasonalize the data.
What’s wrong here? The entire framework is based on arbitrary assumptions that have nothing to do with reality. The method implies that human volition plays no part at all in economic activity. Even worse, the method implies that the data are actually what determines what human beings are going to do, as if human beings are mere robots following the commands of various statistical factors.
Human action is not robotic but rather conscious and purposeful. The data offer a very blurry look at the results of people’s assessments of reality in accordance with each individual’s particular end, at a given point in time. In other words the individual’s action is set in motion by his valuing mind and not by external factors. This in turn means that the assumption of constancy in various seasons is merely a statement about history that may or may not be correct; it certainly does not hold up as a predictor of future behavior.
Designers of the X-11 and X-12 seasonal adjustment computer programs, have attempted to address the issue of the constancy of the seasonal effect by allowing this effect to vary over time. For example, the seasonal effectfor retail sales in December will not be the same year after year but will rather change. Furthermore, these programs are instructed to use only the stable seasonal effects and to factor out changing ones.
The crux of the problem is that people’s responses to various seasons or holidays, are never automatic but rather part of a conscious purposeful behavior. The choices being driven out of the data are as much a part of the underlying reality as the data set. More fundamentally, there are no means available to quantify the individual valuations behind economic choices. There are no constant standards for measuring the act of a mind’s valuation of reality.
On this Rothbard wrote, “In order for any measurement to be possible, there must be an eternally fixed and objectively given unit with which other units may be compared. There is no such objective unit in the field of human valuation. The individual must determine subjectively for himself whether he is better or worse off as a result of any change. His preference can only be expressed in terms of simple choice, or rank. “
Since it is not possible to quantify a mind’s valuation of the facts of reality, obviously this valuation cannot be put into a mathematical formulation. This in turn means that the so-called estimates of seasonal factors generated by the computer programs must be arbitrary numbers.
Contrary to the accepted view, the adjustment for seasonality merely distorts the raw data, thereby making it much harder to ascertain the state of the business cycle. These distortions have serious implications for policy makers who employ so-called counter-cyclical policies in response to the seasonally adjusted data.
For example the strength of the seasonally adjusted employment data could cause the central bank to raise or lower interest rates. This pretense by the central bank policy makers, that they can quantify something that cannot be quantified, is itself a major source of economic instability.
Seasonally adjusted data also form the basis of applied economics. Various theories are derived by observing the interrelationships of the seasonally adjusted time series. These theories cannot be taken too seriously; the data behind the theory are statistically arbitrary and methodologically meaningless.
Now we come to the final and most fundamental criticism of this method of doing statistics. The whole idea that being able to observe the “influence” of the business cycle is fallacious. The business cycle is not inherent in the economy. Swings in economic activity are the result of central bank’s monetary policies, which falsify interest rates, thereby contributing to people’s erroneous valuations of the facts of reality.
Even if it were possible to quantify the cyclical influence, and isolate it from all other fluctuations in economic data, this would not help us to understand the source of the business cycle. Without a coherent theory, based on the fact that human actions are conscious and purposeful, it is not possible to understand the causes of business cycles and no amount of data torturing by means of the most advanced mathematical methods will do the trick.
