Volume 5, No. 1 (Spring 2002)
In this article we examine the differences between two prominent economists who lived about a century apart in time. In the nineteenth century, there was John Stuart Mill (1806–1873) who offered (in his Principles of Political Economy, bk. 2, chap. 1) the following assessment of the relationship between the phenomena of production and distribution: The laws and conditions of the production of wealth, partake of the character of physical truths. There is nothing optional or arbitrary in them. ... It is not so with the Distribution of Wealth. That is a matter of human institutions solely. The things once there, mankind, individually or collectively, can do with them as they like. (pp. 257–58) This is a somewhat mechanistic (and simplistic) exegesis of the production-distribution linkage. However, it is one with profound implications for public policy. What it suggests, particularly where collective action is concerned, is the possibility of a social “taking” of wealth (or income) from some to be then disbursed to others considered to be more deserving. According to Mill, such a redistribution process would be a zero-sum game, having no impact on the overall production of wealth (national income).There are obvious reasons to be suspicious of Mill’s premise. Written in the era that it was, it suffers from the lack of perspective that would later be pro-vided by the marginalist interpretation of the conditions surrounding the employment of factors of production, especially labor. In the case of labor inputs, Mill performs the obligatory obeisance to demand and supply forces, but lurking in the background is some form of deus ex machina that attenuates any link between labor productivity and the level of wage rates. Sometimes, it is a wages-fund notion. Sometimes, it is Malthusian pressures. Whatever, these external constraints raise the specter of “low” wages. Mill devotes a twenty-nine-page chapter (bk. 2, chap. 12) to the “low”-wage question.
Ultimately, Mill’s conception of the link between production and distribution provides an argument for social intervention in distributional outcomes, either through market devices, particularly labor unions, or through public charity that takes the form of income transfers from the more to the less affluent. The syllogism is simple. Actual distributional outcomes are rendered unsatisfactory by natural constraints, but they may be corrected by public policy at no cost to total national output. Therefore, it is appropriate to engage in income redistribution strategies. In reality, what Mill propounds is nothing but a precursor to Marx’s Capital (1867).
Move forward a century, from Mill in 1848 to Ludwig von Mises’s Human Action (1998). The opening paragraphs of chapter 32 effectively sum up the Mill-type argument, to wit:
Interventionism is guided by the idea that interfering with property rights does not affect the size of production. The most naïve manifestation of this fallacy is presented by confiscatory interventionism. The yield of production activities is considered a given magnitude independent of the merely accidental arrangements of society’s social order. The task of government is seen as the “fair” distribution of this national income among the various members of society.
The interventionists and the socialists contend that all commodities are turned out by a social process of production. When the process comes to an end and its fruits ripen, a second social process, that of distribution of the yield, follows and allots a share to each. The characteristic feature of the capitalist order is that the shares allotted are unequal. Some people—the entrepreneurs, the capitalists, and the landowners—appropriate to themselves more than they should. Accordingly, the portions of other people are curtailed. Government should by rights expropriate the surplus of the privileged and distribute it among the underprivileged.
Mises then goes on to deny explicitly the basic interventionist premise, saying:
in the market economy this alleged dualism of two independent processes, that of production and that of distribution, does not exist. There is only one process going on. Goods are not first produced and then distributed. There is no such thing as an appropriation of portions out of a stock of ownerless goods. The products come into existence as somebody’s property. If one wants to distribute them, one must first confiscate them. It is certainly very easy for the governmental apparatus of compulsion and coercion to embark upon confiscation and expropriation. ... But capitalism cannot stand such ... predatory raids. Its capital accumulation and investments are founded upon the expectation that no such expropriation will occur. If this expectation is absent, people will prefer to consume their capital instead of safeguard it for the expropriators. This is the inherent error of all plans that aim at combining private ownership and ... expropriation. (Mises 1998, pp. 800–01; emphasis added)
What Mises’s views translate into is the notion that a socially engineered process of income redistribution is not the zero-sum game imagined by Mill and the interventionists, but is, in reality, a negative-sum game. Government efforts that take income from the relatively affluent and redirect it to those at the low end of the income distribution will lead to a reduction in total production.
Which View Is Correct?
The remainder of this article will be devoted to assessing the correctness of these two contrasting viewpoints. We begin by recognizing the recent emergence of a substantial literature that is highly consistent with Mises’s views on the impact of the confiscatory activities of government on levels of total out-put. The level of government predation can be measured either by its spending or by its actual tax collections.1 One of the more notable works along these lines is that of Charles Ballard, John Shoven, and John Whalley (1985), who estimate the deadweight costs of taxation in the United States using simulations generated by a computable general equilibrium model. They find that, at the margin, the deadweight losses lie somewhere between twenty and fifty cents per additional dollar of taxation. Similar conclusions are reached by Martin Feldstein, dealing with the special case of Social Security imposts, in two National Bureau of Economic Research papers (1995 and 1996). Also worth noting on the tax side are the works of Gerald Scully (1988, 1992) and of James Gwartney and Randall Holcombe (1998).
As to spending-measure analysis, our own work (Gallaway and Vedder 1995; Vedder and Gallaway 1998) is strongly consistent with the tax-oriented studies. Our 1995 piece found a marginal deadweight loss of output associated with federal government spending of thirty-eight cents on the dollar, midway in the range of tax impacts estimated by Ballard, Shoven, and Whalley. The methodology used in this analysis is of the more conventional econometric variety.
In summary, a range of scholarly studies using different measures of the impact of government activities and a variety of statistical methodologies reaches conclusions that are strongly consistent with the Misesian perception of the effect of government actions on the total level of economic activity in an economic system.
Some Simple Statistics
At this point, we propose to narrow the focus of our inquiry to the specific question of how government attempts to redistribute income by taking from some and giving to others affect total levels of output. To set the stage for a subsequent, more formal analysis of the Mill-Mises dichotomy, as it applies to the United States, some simple statistics are useful. We focus on the third of the twentieth century beginning with 1966 and concluding with 1998.2 We turn first to the matter of the level of income transfer payments in the United States. From the National Income and Product Accounts (NIPA) data, we have calculated the real level of income transfers per capita over the period in question.3 The results are shown in Figure 1. Observe that over the thirty-three year span of time real income transfers per member of the population have increased by more than three-and-one-half fold (from $589 to $2,165 in 1982–84 dollars). Also, the pattern of their growth has been quite variable. Initially, there is a relatively rapid increase in such transfers. This continues on into the mid-1970s, to be followed by a more than decade-long substantial slowing in their growth. Finally, in the late 1980s and the 1990s, there is a reacceleration in the increase in real income transfers.
Figure 1: Real Per Capita Income Transfer Payments United States, 1966–1998
The pattern of movements in levels of growth in transfer payments in the United States closely parallels (albeit in a negative fashion) variations in the rate of growth in real per capita gross domestic product (GDP). Between 1966 and 1975, when transfer payments were surging, real output per head rose at an average annual rate of 1.61 percent. However, over the interval 1976–1987, when the increase in transfers slowed markedly, the rate of growth in real per capita GDP accelerated to 2.44 percent a year. Finally, in the period 1988–1998, when transfer growth burgeoned once more, the output growth rate fell back to 2.04 percent. Figure 2 illustrates these changing per capita GDP growth percentages.
Figure 2: Average Annual Growth Rate, Real Per Capita Gross Domestic Product, United States 1966–1975, 1976–1987, and 1988–1998
A More Detailed Analysis of the Data
These cursory statistics further support the Misesian perception of the production-distribution linkage. Also, they suggest a need for a more in-depth empirical analysis. We have conducted such an examination using multiple regression techniques, with real per capita GDP as the dependent variable and with real per capita income transfers, the average productivity of labor, and the unemployment rate as independent variables. Data from the years 1966–1998 are used. Two different versions of a multiple regression equation have been estimated, one using levels of the variables and the other employing changes in them. A few comments concerning the average productivity of labor and unemployment variables are in order. The productivity measure is designed to control for long-term trends in growth in output per head. On the other hand, the unemployment rate is used to take into account cyclical variations in output.
The statistical results of the analysis are shown in Table 1. Basically, they confirm the aggregate data patterns. The income transfer variable is statistically significant with a Misesian negative sign. In the levels regression, over 99 percent of the variation in the real per capita output variable is explained.4
The changes (or first-difference) regression is equally powerful, explaining 93 percent of the changes in real GDP per capita. Again, the transfer payment variable is statistically significant with a negative sign. Thus, both of these regression equations confirm the Misesian insight and contradict Mill’s vision.5 In the interest of sensitivity considerations, one other version of the regression model has been estimated. In it, the income transfer measure is transfers as a percentage of GDP. The statistical results are very similar to those observed in the other regression results.
Deadweight Loss Estimates
The regression results reported in Table 1 indicate a statistically significant negative effect of real income transfers on per capita economic growth. But, how strong is this effect quantitatively? To explore this issue, we have developed two different estimates of the quantitative loss of real output per person in 1998, arising from the presence of income transfers in the American economy. The first employs an indirect method, which consists of applying the per capita growth rate for the low transfer payment growth era, 1976–1987, to the entire period 1966–1998, creating a simulated real growth data set. Such a series is presented graphically in Figure 3, where it can be contrasted with the actual real-output-per-capita data set. By 1998, the annual loss per person is estimated at $4,110. At this time, income transfers per head amount to $2,165. Therefore, on average, the estimated costs, or deadweight losses, associated with an additional dollar of income transfers are $1.90 in 1998.
Figure 3: Real Per Capita Gross Domestic Product and Simulated Real Gross Domestic Product, United States, 1966–1998
This may seem to be a large value. However, it is consistent with the direct estimates provided by the regression coefficients already reported. In the levels regression, the coefficient is -1.69, indicating that an additional dollar of income transfers reduces output per head by $1.69. In the first difference form of the regression, this coefficient is -2.03. The average of the two coefficients is -1.86, very close to the -1.90 calculated from the output growth data.
Conclusions
Our analysis shows that Mises’s critique of “confiscatory interventionism” is the appropriate paradigm for interpreting events in the American economy
during the last third of the twentieth century. As the “transfer state” has grown in scope, the cumulative deadweight losses to the economy have increased until, during the bulk of the decade of the 1990s, they have driven real per capita levels of output to levels that are about one-eighth below what they might have been. Losses of this magnitude are quite inconsistent with the popular view held in significant political circles to the effect that government activity of the tax and transfer variety is, at its worst, a zero-sum game. This is a position popularized by Lester Thurow (1980) in his book The Zero-Sum Society.
Why do losses of this sort occur? Mises (1998) answers that question on page 803 of Human Action where he points out that economists ask, “what are the effects of confiscatory taxation on capital accumulation?” His answer says it all: “The greater part of that portion of the higher incomes which is taxed away would have been used for the accumulation of additional capital.” Without that capital, economic growth slows, and the deadweight losses mount. Enough said.
- 1Spending levels probably are the more appropriate. They are a more precise indica-tor of the distortion of resource utilization patterns generated by government activities in that they include those financed by means other than tax revenues. However, even the spending measure is incomplete in that it does not capture the cost of government regulations, variously estimated to amount to several hundred billion dollars (Hopkins 1996).
- 2Data for 1999 are available. However, they are less reliable due to the likelihood that they subsequently will be revised by the data collecting authorities.
- 3Real transfer payments per capita are calculated by dividing total transfer payments by population and deflating the result using the Consumer Price Index for Urban Consumers (CPI-U). In that index, 1982–84 = 100.
- 4Some econometric observations are in order. In any time-series analysis, the issue of the stationarity of the variables arises. To explore this, we have calculated Augmented Dickey-Fuller statistics for the dependent variables and the transfer payment variables. In the levels regression, the test statistic for real per capita gross domestic product exceeds the Mackinnon critical value at the 1-percent level while the statistic for real per capita income transfers is equal to the Mackinnon value at the 10-percent level. In the first difference regression, the statistics for both the change in real per capita gross domestic product and change in real per capita income transfers variables exceed the Mackinnon value at the 5-percent level. Thus, there are no serious stationarity problems.
- 5Some additional econometric detail is appropriate. A variety of diagnostic tests have been performed. The regressions reported here pass standard serial correlation (LM) and heteroskedasticity (White) tests. In addition, they pass Ramsey Reset model specification tests and the Recursive Residuals, Cusum, and Cusum Squares model specification tests. Details will be provided by the authors on request.
