===== PAGE 1 ===== NOTES AND REPLIES Statistical Malfeasance and Interpreting Economic Phenomena Richard Vedder t took seven decades, but most people now accept what Ludwig von Mises explained three quarters of a century ago, namely, that centrally directed socialistic economies cannot succeed in coordinating vast numbers of interrelated decisions, in large part because of the information problem arising from non-market forms of resource allocation (Mises 1920). No amount of input-out- put models generated on vast computers can overcome the problems of directing resources under changing conditions of wants and scar- city. The information problem that plagued socialist states, like the old Soviet Union, persists in another form today in so-called “mixed” economies like the United States. While the price data generated by markets, as consumers and producers interact in a productive, if seemingly chaotic, discovery process, allow decen- tralized economic agents to make complex and ever-changing economic decisions without any central direction; governments try to generate data which aggregate economic activity over entire econo- mies to assist the softer forms of economic planning that persists in most of the industrialized democracies—fiscal and monetary policy, Richard Vedder is professor of economics at Ohio University. This paper was presented at the 1997 Austrian Scholars Conference, Auburn, Alabama, April 5, 1997. I am indebted to Lowell Gallaway for valuable insights utilized in this paper and for making the oral presentation at Auburn. Review of Austrian Economics 10, no. 2 (1997): 77-89 ISSN:0889-3047 77 ===== PAGE 2 ===== 78 Review of Austrian Economics 10, No.2 (1997) environmental rules, governmentally mandated distortions in the use of energy resources, and so on. On methodological grounds, Austrian economists reject the logi- cal-positivism of most contemporary economics. Even if one were raised in a neoclassical tradition that places a high utility on evaluat- ing economic phenomena in terms of some analogue of the scientific method present in the physical sciences, however, one should be wary of many modern research findings, owing to the inherent and probably insurmountable difficulties of aggregating economic data. Bad data lead to bad conclusions, even if one accepts an activist economic philosophy that centralized decision making can improve on the spontaneous decisions made in the market economy. This paper presents five examples of how data problems can lead to a misinterpre- tation of economic phenomena, or at least promote great uncertainty in evaluating the direction and scope of economic change. Example One: Are Wages Rising or Falling, or, Are Workers Being Exploited? Public figures as politically diverse as former Labor Secretary Robert Reich, Pat Buchanan, and Ross Perot have argued that the standard of living of American workers has stagnated in recent decades. Reich blames it on greedy businessmen, while Buchanan and Perot claim that as a consequence of ill-considered free-trade policies, low-paid foreign workers are robbing Americans of their affluence. Is this unholy trinity right? Using the method of mainstream economics, let us turn to the purported evidence to evaluate the scenarios of three economists whose names began with “M”: Malthus, Marx, and Mises. Which economist do you like? Using contemporary data, I can give you evidence to support the views of any of them. Go to the allegedly trusty source of current economic informa- tion, the Economic Report of the President (U.S. Council of Economic Advisers, various years). Turning to page 352 in the 1997 edition, the ===== PAGE 3 ===== Vedder: Statistical Malfeasance and Intrepreting Economic Phenomena 79 hourly average wage of private sector American workers in 1973, expressed in dollars of 1982 purchasing power, is stated to be $8.55. For 1996, it was only $7.43. Workers were making 13.1 percent less in 1996 than a generation earlier. On a weekly basis, the wage decline was even greater. This is evidence that Malthus was right. With population growth and the law of diminishing returns, wages are moving toward subsistence. Or maybe Marx was right. Turning the page, we learn (page 354) that the output per hour in the same period has risen more than 30 percent.’ Productivity was rising significantly even while wages were falling—we have a total refutation of the Austrian claim (Mises 1963, p- 597) that workers are paid according to their marginal productiv- ity. We have progressively greater exploitation of the proletariat by greedy capitalists. If this scenario is true, Bill Gates and Sam Walton make capitalists like the Vanderbilts and the Rockefellers look like Mother Teresa. Why did Marxism largely wither away (except in universities) when we needed it to explain the growing absolute and relative misery of the American worker? Alas, there is another side to the story. Michael Boskin and some other distinguished number crunchers have concluded that the con- sumer price index used to calculate real-wage change overstates inflation by about 1.1 percentage points a year, and has done so for along time.’ Earlier, the price gurus at the Bureau of Labor Statistics as much as admitted that their CPI-U index suffered significantly by overstating housing prices after 1967, and concocted a CPI-U-X1 index. It was generally believed around 1980 that adopting the X1 'Output per hour in the “business sector” shows an increase of 30.2 percent from 1973 to 1996. The 1996 figure was assumed to be the average of the second and third quarters of the year. Using data for the “nonfarm business sector,” the increase is only 25.8 percent. Implicit in the difference between these two estimates is farm-sector productivity growth of a huge magnitude (perhaps 100 percent), given the relatively small size of that sector. 2For a discussion by mainstream economists of problems with the consumer price index, see the 1997 Economic Report of the President, pp. 67-72. ===== PAGE 4 ===== 80 Review of Austrian Economics 10, No.2 (1997) procedures would improve factual accuracy, but our government did not do so because, among other things, it would have reduced Social Security cost-of-living benefit increases. Hence, for political rea- sons, the U.S. government continued to use an index that everyone says was wrong. Using the X1 index and reducing annual inflation rates by 1.1 percentage points annually, I calculated a new and presumably improved measure of changing rates of inflation (see appendix). “Boskinizing” the data, real wages in fact rose 9.4 percent an hour from 1973 to 1996, rather than fell as officially reported. Using the method of contemporary mainstream economics, we can say that, em- pirically, the Malthusian—Marxian scenario described above has now taken a hit, but, alas, the wage growth is still well below the reported productivity growth of 30.2 percent. In other words, Marx is closer to the truth than Mises, or, for that matter, than Alfred Marshall. Don’t despair yet. Marx is about to take a big hit. Page 354 of the Economic Report covers “real compensation per hour” in the “business sector.” This table incorporates into employee compensation the fringe benefits excluded from the wages measure, and thus is a more comprehensive measure of the remuneration that workers receive from their employers. Even using the flawed BLS data on inflation, it is revealed that real compensation per hour rose 9.1 percent from 1973 to 1996. So much for Malthus. Applying the X1 and Boskin adjustments to the CPI (see sppendix), I calculate that real compensation per hour in fact rose 42.8 percent from 1973 to 1996. Workers are doing much better than their parents did a generation ago. Has Mises been vindicated by the very quantitative approach that he disdained? Not exactly. Wages, broadly defined, are now re- corded as having risen faster than productivity. If correct, this implies that corporations are being financially squeezed by labor, either by accident or design. We have reverse Marxism—the prole- tariat is squashing the capitalists—the withering away of capitalism, if you will. Alternatively, a benevolent, “kinder and gentler” breed of entrepreneur is voluntarily turning over income to workers. ===== PAGE 5 ===== Vedder: Statistical Malfeasance and Intrepreting Economic Phenomena 81 This latter conclusion, however, is murky, since it compares rising real compensation to productivity change. Productivity is de- fined in terms of real output per hour of work. Thus, the calculation of productivity involves using a price index, and if that index has been understated, then the recorded productivity growth has similarly been below reality. A large number of students of productivity data believe there is an understatement of modern productivity growth. Correcting for that misstatement, it is plausible and indeed likely that the real-wage data and the productivity data would show very similar upward trends, consistent with both Austrian and neoclassi- cal traditions in economics. Mises is thus vindicated. All of this reinforces Austrian concerns about attempting to verify or falsify economic propositions based on aggregative eco- nomic data. At the same time, as Professor McCloskey (1985) tells us, the rhetoric of modern economics includes heavy use of numbers and econometric manipulation of them. Only partly tongue in cheek, I think it is legitimate to use Austrian praxeological principles to clarify some of the existing statistical mayhem. To illustrate, accept the Austrian proposition that wages are determined by the marginal productivity of labor. The rate of growth in aggregate prices over time, then, would equal that growth necessary to be consistent with this Austrian proposition. If Boskin’s 1.1 point adjustment to the CPI leads to wage growth exceeding productivity change, while no adjustment leads to productivity changes exceeding wage growth, the correct adjustment is one that equates these two measures, perhaps 0.5 or 0.7 percentage points. Since, given the rhetorical passion of economists for quantitative measures, we are going to use price indices; why not use Austrian insights to calculate them, even if Austrians themselves are disdainful of their use? I say this with some trepidation, ever mindful of Mises’s magisterial injunction: “In the field of praxeology and economics no sense can be given to the notion of measurement”> (Mises 1966, p. 222). 3Mises’s views were anticipated by the English economist J.E. Cairnes (1888). See also Rothbard (1993), chap. 5. ===== PAGE 6 ===== 82 Review of Austrian Economics 10, No.2 (1997) Example Two: The Worsening Post-World-War-I1 Depression’ Problems with price indices can lead to grotesque and changing interpretations of historical phenomena. According to the official national-income-account data, the nation had a huge downturn in 1946. Moreover, unique among downturns in American history, it continues to get worse—even after the downturn is over. In 1960, the U.S. Department of Commerce reported that the national output decline for 1946 had reached an extraordinary 14 percent. With the historical revisions reported in 1995 (U.S. Council of Economic Advisers), the calculated output decline for the year 1946 was 20.6 percent. This is greater than the accumulated reported decline for 1931 and 1932 during the darkest part of the Great Depression. Yet, all of this happened while consumer spending was rising sharply, the unemployment rate was under 4 percent, and the stock market was registering double-digit gains, with the Standard and Poor industrial index reaching the highest level since 1929. The statistical fiction that official national-income indicators show re- flects the switch from largely command, non-market-based output in 1945 to a much more market-determined output with a dramati- cally downsized public sector. The end of the wage-and-price con- trols meant that inflation moved from being disguised to being explicit. Perversities in the way the aggregate GDP price deflator is calculated meant that the shift from public to private activity sub- stantially increased the recorded GDP price deflator for the econ- omy. Over time, the difference in the reported increase in prices in the public and private sectors meant that the post-war shift back to private enterprise increased the aggregate price index independent of price movements. Thus, revisions in statistics years after they are originally com- piled do not always lead to greater accuracy. Given the fundamental %See Vedder and Gallaway (1991, 1997) for a more detailed analysis of this example. ===== PAGE 7 ===== Vedder: Statistical Malfeasance and Intrepreting Economic Phenomena 83 problem of evaluating government activity that is not sold in markets, any aggregate output statistic is subject to considerable debate. As Robert Higgs (1992) has shown, under one very reasonable method of accounting, the Great Depression actually persisted until the mid- 1940s, rather than decisively ending with America’s entry into World War II. Example Three: Are We Undergoing Deindustrialization? The contradictory data are present within a single edition of the Economic Report of the President. For example, some people have spoken about the “deindustrialization” of America, presumably referring to a sharp decline in the relative importance of manufacturing in the American economy. Using nominal data from page 312 of this year’s report (U.S. Council of Economic Advisers 1997), this observation is confirmed, with manufacturing’s share of gross domestic product falling by nearly one-fourth in just 17 years from 1977 to 1994 (from 22.81 to 17.27 percent). Looking at the next page (p. 313), where the data are expressed in real terms, one observes manufacturing’s share of output falling only very modestly, from 18.61 to 17.68 percent of GDP Hardly major deindustrialization. Example Four: Is Government Growing or Declining in Relative Size? With respect to government, the conclusions are just the opposite. With the data expressed in real terms, the government by 1994 was about a 20 percent smaller proportion of the economy than in 1977; with nominal data, the decline was only one-third as great. Adding to the confusion, the measured change of relative size in government varies depending on whether one looks at tables B-1 and B-2, B-8 and B-9, or B-10 and B-11. Turning to data on total current expen- ditures of government on page 394, we learn that government as a ===== PAGE 8 ===== 84 Review of Austrian Economics 10, No.2 (1997) percent of GDE grew from about 30 to 32 percent from 1977 to 1994, reflecting the impact of income transfers not included in the basic GDP classifications. Moreover, none of these measures picks up the impact of coercive governmental regulation or mandates on the private sector, which almost certainly have grown in relative importance over time. As with earlier examples, the hazards of price indexation contrib- ute importantly to the contradictory findings. For example, the statistics supposedly correcting for inflation use an index for govern- mental services to deflate, which is, at best, a highly speculative exercise, given the non-market nature of governmental activity. In general, governmental services are valued in the GDP accounts by adding up compensation paid to employees, which is to say on payments to inputs rather than a valuation of output. Given the very high levels of economic rent present in much government employ- ment, this is a doubly dubious procedure. After all, one group of employees is not paid according to its marginal contribution to society, it is government employees (Cox and Brunelli 1994). Example Five: Is the Economy Doing Well or Poorly? Even most mainstream economists acknowledge significant difficul- ties with the measurement of aggregate economic performance. The official GDP statistics suggest that the worst post-war year in terms of total output change was 1946, while the best was 1951. The former year saw the nation convert from a wartime to a peacetime economy, and from a price-controlled economy to one in which prices were more or less free to fluctuate with market forces. The exact reverse happened in 1951. The 1951 boom came in part by forcing human resources into employment at below-equilibrium wages—the military draft. The allegedly good performance came from coercive tactics, and by valuing an important part of output not by the subjective evaluations of consumers and producers, but by the non-mar- ket prices paid to governmentally directed inputs. ===== PAGE 9 ===== Vedder: Statistical Malfeasance and Intrepreting Economic Phenomena 85 Even if one accepts the concept of GDP as a reasonable way of evaluating the performance of an economy, however, there are sig- nificant practical difficulties. I randomly selected the year 1993 for evaluation. In the 1994 Economic Report of the President, it is reported that the 1993 GDP rose 2.9 percent—To illustrate this, it is best to go to the Statistical Abstract of the United States (U.S. Department of Commerce, Washington, D.C.: U.S. Government Printing Office, 1997). Compare the data on changing financial assets in Table 771 with those in Table 697. Household net financial assets in, say, 1993, rose by $752 billion in Table 771, but “personal savings” in Table 697 was only $216.4 billion—a difference of a factor of more than three. ===== PAGE 11 ===== Vedder: Statistical Malfeasance and Intrepreting Economic Phenomena 87 numbers on imports from the U.S. More generally, the “errors and omissions” component of balance-of-payments statistics is often huge in magnitude. Some aggregate statistics are no doubt reasonably reliable. For example, I trust the numbers on total steel production. Changes in the factor shares in aggregate national-income data probably give better insight into whether labor is earning its marginal product than the wage-productivity data discussed above. Basic data on individual interest rates or individual commodity prices are probably pretty good. Moreover, the private economy in its planning wants, and is willing to pay, to get some aggregate economic data, even mediocre data. That explains why otherwise sensible businessmen pay econo- mists to do forecasts. If GDP didn’t exist, someone would invent it. The statistical mayhem described above suggests that one can find data to support almost any economic theory. The notion that mainstream economists practice science while Austrian economists are more like theologians is, at the very least, an exaggeration. Many mainstream economists selectively use data to defend preconceived positions. They try to add legitimacy to often bankrupt intellectual positions by invoking the mantle of science. On average, I think, Austri- ans are more straightforward and intellectually honest. The moral of my little story, then, is: beware, government statistics may be injurious to your economic health. Treat them gingerly and with suspicion, as Aus- trian economists have long done. Indeed, when it comes to govern- ment statistics, ignorance often may well be bliss. ®In the 1997 Economic Report of the President (p. 415), the median statistical discrepancy (ignoring the direction of the discrepancy) in the years 1989 through 1995 was $31.5 billion, nearly as much as the nation’s purported “unilateral transfers.” ===== PAGE 12 ===== 88 Review of Austrian Economics 10, No.2 (1997) Appendix Below is the official CPI-U (1982-84 =100) and the author’s recon- struction of it incorporating the X1 and Boskin price adjustments. Date CPI-U CPI-U-X1-BOSKINIZED 1973 44.4 50.3 1974 49.3 - 55.4 1975 53.8 59.7 1976 56.9 62.0 1977 60.6 65.3 1978 65.2 69.5 1979 72.6 75.8 1980 82.4 85.2 1981 90.9 93.1 1982 96.5 97.8 1983 99.6 99.8 1984 103.9 102.4 1985 107.6 104.9 1986 109.6 105.8 1987 113.6 108.5 1988 118.3 111.7 1989 124.0 115.8 1990 130.7 120.8 1991 136.2 124.6 1992 140.3 127.6 1993 144.5 129.4 1994 148.2 131.3 1995 152.4 133.3 1996 156.9 136.1 ===== PAGE 13 ===== Vedder: Statistical Malfeasance and Intrepreting Economic Phenomena 89 References Cairnes, J.E. 1888. The Character and Logical Method of Political Economy. 2nd ed. London: Macmillan. Cox, Wendell, and Samuel A. Brunelli. 1994. America’s Protected Class III. Washington, D.C.: American Legislative Exchange Council. Higgs, Robert. 1992. “Wartime Prosperity: A Reassessment of the U.S. Economy in the 1940s.” Journal of Economic History 52:42—60. McCloskey, Donald N. 1985. The Rhetoric of Economics. Madison: University of Wisconsin Press. Mises, Ludwig von. [1920] 1990. Economic Calculation in the Socialist Common- wealth. Auburn, Ala.: Ludwig von Mises Institute. ——. 1966. Human Action: A Treatise on Economics. 3rd rev. ed. Chicago: Regnery. Rothbard, Murray N. 1993. Man, Economy, and State: A Treatise on Economic Principles. Rev. ed. Auburn, Ala.: Ludwig von Mises Institute. U.S. Council of Economic Advisers. Various Years. Economic Report of the President. Washington, D.C.: U.S. Government Printing Office. U.S. Department of Commerce, Bureau of the Census. 1992. Current Popu- lation Reports, Series P-60, No. 182. Measuring the Effect of Benefits and Taxes on Income and Poverty: 1979 to 1991. Washington, D.C.: U.S. Government Printing Office. —. 1960. Historical Statistics of the United States, Colonial Times to 1957. Washington, D.C.: U.S. Government Printing Office. ——. 1975. Historical Statistics of the United States, Colonial Times to 1970. Washington, D.C.: U.S. Government Printing Office. —. 1997. Statistical Abstract of the United States. Washington, D.C.: U.S. Government Printing Office. Vedder, Richard, and Lowell Gallaway. 1991. “The Great Depression of 1946.” Review of Austrian Economics 5 (2): 3-32. —. 1997. Out of Work: Government and Unemployment in Twentieth-Century America. Rev. ed. New York: New York University Press.