The effects of monetary instability on the extent of vertical integration.
Ronald Coase  provided the theoretical foundation for all subsequent studies of the determinants of business integration. Although the transaction-costs paradigm he developed has been accused of lacking empirical content--of being tautological--even by some of Coase's admirers [Alchian and Demsetz, 1972, p. 783; Williamson, 1985, p. 4], its influence is beyond question. Coase's article remains modern because the particular causal factors that he specified as determining the size of firms continue to be emphasized in recent work.
The Coasean tradition is preserved in the present paper. The general goal is to examine the effects of monetary instability on the extent of vertical integration. A major theme of the paper is that, according to the transaction-costs paradigm, the effect of such instability on optimal firm size is ambiguous: distortions in price signals caused by price fluctuations set in motion two distinct forces. One of these forces tends to increase the extent of intrafirm integration of production activities, while the other tends to reduce this integration.
The special object of this paper is to investigate empirically the relative strengths of these opposed forces. This paper is not a test of the transaction-costs theory of the firm. Rather, the authors use this theory as the basis for an empirical investigation of the effects of monetary instability on the extent of vertical integration.
With these ends in mind, the paper proceeds as follows: Section II reviews the transaction-costs theory of the firm. Section III explains the causal relationship between monetary instability and the organization of productive activities. Section IV provides empirical evidence on the nature of this causal relationship. Ordinary least squares regressions are presented showing the magnitude and direction of the influence of monetary instability on economy-wide vertical integration. Section V interprets these empirical findings, draws out some of their implications, and poses questions for further research. Concluding remarks are offered in Section VI.
II. The Transaction-Costs Theory
Coase's most fundamental contribution to the theory of the firm is his insight that the determinants of integration are not exclusively or even mainly technological in nature. Coase recognized that resources devoted to the process of transacting must be economized upon. Theories based upon this recognition explain much more plausibly the reasons for, and the extent of, the integration of different stages of production within a single firm than do theories that abstract from transaction costs. Theories that look to purely technological factors for the explanation of business integration overlook the more essential causal forces that stem from the costliness of transacting.(1) 1The fact that the extent of business organization is not exclusively a functionn of technological forces is emphasized by Oliver Williamson. In criticizing Bain's thesis that integration is justified only when technological economies are present, Williamson does not deny that technological forces have some influence on the type and extent of integration. However, he points out [Williamson, 1985, p.87] "that decisions to integrate are rarely due to technological determinism but are more often explained by the fact that integration is the source of transaction cost economies." Coase's theory is decidedly in the transaction-cost tradition [Cheung 1983].
At the heart of Coase's theory is his distinction between two alternative types of transaction costs. These costs interact in such a way as to determine the number of stages of production that are integrated within administrative hierarchies. If either of these costs change relative to the other, the optimal sizes of firms will change as a result and in a direction that is conceptually determinate.
The first of these costs is "the cost of using the price mechanism" [Coase, 1937, p. 336]. The second category of costs are those that attend the allocation of resources by administrative command. These latter costs are borne directly by the entrepreneurs and owners of firms. It is the responsibility of enterprenuers and owners to determine the specific tasks to be performed by inputs that are within the firm. Coase [1937, p. 333] defined the entrepreneur as "the person or persons who, in a competitive system, take the place of the price mechanism in the direction of resources." The object of entrepreneurs in the Coasean model is to minimize the costs of allocating resources for any given production process by varying the number of stages of production internalized within the firm.
Let Cp(S) and Ca(S) be the two relevant categories of costs identified by Coase, where S is the number of stages of production internalized by entrepreneurs to form Coasean firms. Cp(S) is the total cost of using the price system, while Ca(S) is the total cost of allocating resources by administrative command. Cp(S) is at a maximum when all resource transfer are accomplished by reactions on the part of resource owners to differences in relative prices that are perceived by these owners as profit opportunities. Cp(S) falls as S rises. In the limit, all technically separable stages of production comprising a specific production process are integrated into a single firm and, as a consequence, Cp(S) = 0 under conditions of complete vertical integration.
However, this reduction in the costs of using the price system--i.e., reduction in the costs of contracting across ownership boundaries--is not achieved without some sacrifice. Increasing the size of firms means that the complexity of the task of intrafirm resource allocation also increases. In turn, there is an increase in the cost of achieving the optimal pattern of resource allocation by means of administrative command: Ca(S) rises as S rises.
The marginal cost of using the price system falls as a greater number of stages of production are internalized under administrative command within firms. Cp'(S) falls with increased internalization because those transactions that are most costly to carry out across ownership boundaries will be the first to be internalized; those that are least costly to accomplish by use of the price system will be the last to be internalized. This effect can be written as Cp'(S) < 0.
With regard to the cost of administrative command, those stages of production that are, on net, least costly to administer by conscious design will be the first to be internalized. As a greater proportion of the entire production process is removed from the price system and moved into a firm (or firms), the marginal cost of administratively directing additional stages increases: Ca'(S) > 0.
According to the transaction-costs theory of the firm, the optimal size of a firm is determined by equating the marginal cost of using the price system with the marginal cost of using an administrative-command system. Thus, optimal firm size requires that Cp(S) + Ca(S) be minimized, i.e., that Cp'(S) = Ca'(S). If S* minimizes the sum of these two cost functions, any change in the number of integrated stages of production, ceteris paribus, will increase the total cost of achieving the pattern of optimal resource allocation.
Changes in any of the parameters that determine the relative values of Cp(S) and Ca(S) will, of course, change the value of S*. For example, advances in the technique or practice of auditing will improve the ability of the owner of a firm to monitor effectively the activities of his employees.2 That is, the costs of administrative command decrease. Ceteris paribus, such a parameter shift will change the optimal (i.e., cost-minimizing) number of stages of production over which resources are allocated by administrative dictate. Improvements in auditing techniques decrease the cost of using administrative structures to allocate resources. The result is that a greater number of stages of production are internalized within firms. 2See Watts and Zimmerman  for a discussion of the evolution of auditing.
A fall in the cost of administrative direction relative to the cost of using the price system will cause S* to increase. By contrast, any parameter shift that causes the cost of using the price system to fall relative to the cost of administration will decrease the optimal number of internally organized stages of production.
III. Monetary Instability and the Firm
Monetary instability distorts price signals by adding noise to the relative money prices established on the market. Such distortions reduce the ability of the market to allocate resources in an efficient manner. The market may respond, however, by evolving various institutions that work to offset, at least partially, the discoordinating effects of monetary instability. As Fischer [1982, p. 169] has noted, "The costs and effects of inflation depend on the extent to which the institutional structure of the economy has adjusted to its existence." Changes in the extent of vertical integration may be one such institutional adjustment.
Greater variability in the purchasing power of money decreases the reliability of the currently prevailing set of money prices as long as individuals do not have full knowledge of the true price level or of the true rate of change in the price level. Monetary instability injects noise into the structure of nominal prices. Under these conditions, money prices will not all instantaneously adjust to their new long-run equilibrium values. Such equilibrating adjustments fail to occur because economic agents do not know what the real factors are. Thus, agents are unable immediately to adjust (i.e., before real effects occur) money prices to their new equilibrium values.
The above argument suggests that the dispersion of nominal prices will be greater the less informed are economic agents about the true purchasing power of money. And the greater the dispersion of nominal prices, the greater is the dispersion of relative prices.3 Numerous empirical investigations of the relationship between the general inflation rate and the variability of relative prices have found a strong correlation between high rates of price-level inflation and increased variability of relative prices.4 There is also a strong statistical correlation between high rates of inflation and high variability of these rates [Okun, 1971; Taylor, 1981; Holland, 1984]. The increased variability of relative prices (with respect to the general price level) represents added noise to the signals that entrepreneurs rely on to guide them in their decisions to commit resources to particular lines of production. 3Parks' [1978, p. 82] formulation of relative-price dispersion is used: VPt = relative-price variability; Wit* = weight of the ith component of the bundle whose prices are used in the index; DPit = rate of change in the price of commodity i; and DPt = general inflation rate for period t.
Fischer [1981, pp. 389-91] points out that Parks' formulation of relative-price variability is not the only possible formulation. An alternative measure that is conceptually cleaner, but impossible to gather accurate data on, involves calculating "variations of individual price levels around some appropriate path for the relative prices of the individual components [of a deflator]." 4See [Vining and Elwertowski, 1976; Cukierman 1982]. Parks  found unanticipated inflation to be positively associated with movements in relative prices.
The relevant question here concerns the effect that monetary instability--or inflation--has on the extent of vertical integration. As will presently be seen, the answer is ambiguous. One effect of monetary instability is to increase the cost of using the price system. The less reliable are money prices as guides to profitable commitments of resources, the more costly it is for entrepreneurs to use such prices. Less reliable prices translate into higher values of Cp(S) for all S.
Defining Cp(S) as the new cost function under conditions of monetary instability, Cp(S) > Cp(S) for all S, and Cp'(S) < Cp'(S). This last inequality implies, of course, that ~ Cp'(S) ~ > ~ Cp' (S) ~. Without any change in Ca(S), the optimal number of internalized stages of production will increase as monetary instability increases.
However, monetary instability also has an effect on the cost of administrative command. The decreased reliability of money prices increases the costs of allocating resources within the firm, at least to the extent that administrators rely on market prices as benchmarks for their own internal-to-the-firm pricing (and resource-transfer) systems.(5) Monetary instability causes the values of Ca(S) and Ca'(S) to rise for all S. 5In a study analyzing the use of actual market prices as guides to intrafirm decision making, Whinston  discovered that, whenever possible, firms rely on outside prices as guides to their own internal allocation decisions. See also the interesting discussion by Carlton .
With both of these costs increasing due to the decreased reliability of money prices, the net effect on the optimal number of internally organized stages of production is ambiguous. Thus, empirical investigation is necessary to determine what effect additional noise in money prices caused by inflationary monetary policy has on vertical integration.(6) 6There is another respect in which the less-integrated economy differs from the more-integrated one. This is in the demand for money. Because the number of transactions carried out with money is greater the less integrated is an economy, the transaction velocity of money will be higher in the less-integrated economy. For the sake of simplicity, this effect is ignored here.
IV. Empirical Model and Results
The empirical model employed is of the following general form: GSR = f(PVAR, ANTI, TIME, e), where: GSR = the ratio of GNP to total business sales (TBS);7 PVAR = the variance of the annual inflation rate over the three-year interval, t -- 2 through t; ANTI = the number of antitrust cases instituted annually;8 TIME = a linear time trend; and e = the regression error term. The data cover the years 1947 through 1982.9 Some summary statistics are displayed in Table 1. 7An economy in which all production processes are integrated from beginning to end will have total sales equal to GNP. Of course, no measure of the extent of economy-wide vertical integration is entirely satisfactory. The most common alternative to the proxy is the ratio of value added to sales, but this too is subject to bias [Adelman, 1955; Laffer, 1969, Tucker and Wilder, 1977]. For other attempts to characterize vertical integration quantitatively, see [Gort, 1965; Livesay and Porter, 1969]. 8ANTI is used as a proxy for the effect that antitrust-enforcement has on the incentive of firms to integrate. 9These data are available from the authors on request.
Before estimating the regression equation, the authors investigated the causal relationships in the data using the procedure suggested by McMillan and Fackler , which relies on the definition of causality developed by Granger .10 In brief, a variable X is said to Granger-cause Y if the current value of Y is better predicted from past values of Y and X than from past values of Y alone. In implementing the test, the series were first made stationary: the GNP-to-sales ratio, inflation-rate variance, and antitrust cases were stationary in their first differences (the variables so formed were labeled DGSR, DPVAR, and DANTI, respectively). The authors also took account of the annual growth rate in real GNP, RGROW, and the variance of the M1 monetary aggregate, M1VAR, both of which were stationary in their levels. 10An application of the technique to causality between taxing and spending is provided by Anderson, Wallace, and Warner .
After converting the variables to stationary series, univariate models were formed in which three series, DGSR, DPVAR, and DANTI, were each regressed successively on a maximum of six lags of itself, and the optimal lag length was determined by selecting the lag that minimized the model's final prediction error, FPE [Akaike, 1969]. Three bivariate models were then formed.
In one of these bivariate models, for example, DGSR was regressed on the optimum lags of itself determined in the previous step plus up to six lags of the inflation-rate variance. The optimum lag length of DPVAR in this regression was again determined by the minimum-FPE criterion. The process was then reversed, treating DPVAR as the dependent variable. In each case, if the minimum FPE of the bivariate model was less than the FPE of the univariate model, then the added variable was provisionally accepted as causally related in the sense of Granger. A set of trivariate models was then estimated, and so on until the multivariate FPE was made as small as possible. 10An application of the technique to causality between taxing and spending is provided by Anderson, Wallace, and Warner .
The results of this exercise are presented in Table 2. In the first row, the model that best predicts the GNP-to-sales ratio contains one lag of itself, and six lags of the inflation-rate variance. The preliminary conclusion is, therefore, that changes in the inflation-rate variance Granger cause changes in the GNP-to-sales ratio. Similarly, the results displayed in the second and third rows indicate that the inflation-rate variance and antitrust enforcement are each best predicted by lags of itself only.11 (Neither of the two control variables, RGROW and M1VAR, added significantly to a reduction in the FPE's of the single-equation models.) 11The Box-Pierce Q statistics calculated for the first 20 autocorrelations of the residuals of the single-equation models are shown in the last column of Table 2. Comparing Q(20) with a 5 percent value chi-square variable with 19 degrees of freedom, one obtains X(2).05.19 = 30.14. One can conclude that there is no strong evidence of autocorrelation in the residuals.
The next step is to test whether the variables selected for inclusion by the FPE criterion jointly add to the explanatory power of the various models. The results presented in Table 3 suggest that, taken as a group, the six coefficients on the lagged values of DPVAR in the equation, using DGSR as the dependent variable, are significantly different from 0 at the 1 percent level on the basis of the calculated F statistic. Similarly, in the equation where DPVAR is the dependent variable, the four coefficients on lagged values of itself approach significance at the 5 percent level. Thus, there is fairly strong evidence of unidirectional causality running from changes in the inflation-rate variance to changes in the GNP-to-sales ratio.
Given the causal relationships in the data, the regression equation was estimated by ordinary least squares. The results, shown in Table 4, suggest that increases in the inflation-rate variance increase the extent of vertical integration in the economy: Using the significant coefficients on DPVAR at lags one, two, five, and six, the cumulative effect is 75.443. That is, over the interval running from year t-9 through year t, an increase in the variance of the inflation rate reduces the cost of administering prices within the firm relative to the cost of using the price system. Thus, the optimum number of stages of production that are internalized within firms increases, ceteris paribus.
As a final test, the authors estimated the model explaining DPVAR suggested by the second row of Table 3. The predicted values of the inflation-rate variance, PVHAT, and the associated residuals, R, were then used as proxies for anticipated and unanticipated price variability, respectively. The results, shown in Table 5, indicate that the GNP-to-sales ratio is affected entirely by those changes in the inflation-rate variance that are unanticipated on the basis of past changes in price variability. The extent of vertical integration in the economy is not influenced by anticipated changes in the inflation-rate variance. Overall, the model explains on the order of 65 to 67 percent of the variation in economywide vertical integration of production over the 36 years since the end of World War II.
The effects of inflation, or of manipulations of the supply of money, on the extent of vertical integration can occur only as far as inflation distorts relative prices. Only through changes in the structure of actual nominal prices can monetary policy have effects on real economic variables. Economic decision makers must be led to do that which they would not do if they had more-accurate knowledge of real variables.
Because the bulk of resource-allocation decisions is made on the basis of relative prices as these actually exist in markets [Hayek, 1945], any effect of inflation on the pattern of resource allocation must be transmitted through whatever effects inflation has on the structure of relative money prices, for only then is it plausible that agents will act on the basis of information that is faulty. The noise in the price system which has the potential to bring about adjustments in the extent of vertical integration is the noise that interferes with the signaling function of prices.
The finding that lagged and contemporaneous changes in the variance of the inflation rate has a statistically significant effect on the extent of vertical integration must be understood as implying that changes in the inflation-rate variance add noise to relative-price signals. This noise is distortive enough to cause agents to alter the institutional environment within which they conduct their economic activities.
Put another way, inflation alters the contractual structure of the economy. As an illustration of the nature of these institutional changes, consider two alternative mechanisms for dealing with the risk attendant upon increased price variability. On the one hand, resource owners and entrepreneurs can themselves bear the additional risk of capital loss on the resources under their command that are brought about by changes in the inflation-rate variance by holding larger resource inventories.
On the other hand, the entrepreneur can disintegrate, purchase his required resources on the market, and use forward or futures contracts to shift the risk to someone else. Thus, higher rates of price variability can, by reducing the number of resource owners and entrepreneurs who wish to specialize in predicting future spot prices of resources, shift the pattern of contractual relations in the economy toward increased reliance on transactions carried out across the boundaries of firms. It is an empirical question as to which of these effects dominates.
A more direct empirical investigation into the effects of monetary policy on vertical integration might use relative-price variability (or some measure of changes in this variability) as the independent variable. However, as mentioned above, previous studies have found a strong positive relationship between the variability of the average rate of inflation and the variability of relative prices. Therefore, the inflation-rate variance is a close proxy for relative-price variability. Also, a market economy operating free of any distortions caused by inflation will always have some positive amount of relative-price variability.
Market forces are dynamic. Consequently, there is a continual need for the market to alter its existing pattern of resource allocation. The primary way it accomplishes this task, of course, is by altering relative prices. Any changes in preferences, technology, or resource constraints can be accommodated most efficiently through these resulting changes in relative prices. As such, it is not desirable to rid the economy of all relative-price changes. What is desirable is to rid the economy of relative-price changes that are not consistent with long-run equilibrium (such as that induced by money-supply manipulations).
Despite the fact that this natural amount of relative-price variability is beneficial, it nonetheless influences the institutional structure of the economy. For example, even in an economy that does not experience arbitrary changes in the purchasing power of its money, this natural amount of relative-price variability may still affect the extent of vertical integration: If for some reason the natural amount of relative-price variability changes, there is every reason to expect that the optimal degree of vertical integration for many of the economy's different industries will also change.
Studies that use relative-price variability as the crucial independent variable (in a regression having the extent of vertical integration as the dependent variable) are unable to distinguish between the amount of vertical integration caused by inflationary monetary policy and the amount brought about by the ordinary forces of the competitive market process. The fact that inflation is largely a monetary-policy variable--a phenomenon whose cause is exogenous to the operation of the economic system--makes it desirable to isolate the effects of this policy. The use of the inflation-rate variance as a crucial independent variable allows for such an isolation.
Another closely related point is that the direction and extent of vertical integration caused by the economy's natural amount of relative-price variability may differ from the direction and extent caused by the variability of relative prices caused by discretionary monetary policy.
For instance, the findings indicate that an increase in the variance of the inflation rate increases the extent of vertical integration. But, one cannot conclude from these findings that all increases in relative-price variability are associated with increased vertical integration, even though the variability of relative prices is positively associated with the variance of the inflation rate. Because relative-price variability brought about by discretionary monetary policy has a different source than does the natural amount of relative-price variability, the effects on integration of each of these two types of relative-price variability may well be quite different. Obviously, more theoretical and empirical work must be done to see if this is, in fact, the case.
It is impossible to determine, purely on the theoretical grounds cleared by Coase and subsequent transaction-costs theorists, the net effect that noisy prices have on economywide vertical integration. This inability is due to the fact that monetary instability increases both types of costs that Coase identified as the relevant margins on which the optimal sizes of firms are determined. At this stage of theoretical understanding, then, the effect of inflation on the extent of vertical integration is an empirical question. The purpose of this paper was to provide a simple framework for interpretation as well as to determine empirically the direction in which monetary instability affects business integration.
Using the variance of the inflation rate as the crucial independent variable, and the ratio of GNP to total business sales as the measure of economy-wide vertical integration, it was found that changes in the variance of the inflation rate have a statistically significant effect upon vertical integration with a one-to-nine year lag. This lag is plausibly attributable to the time it takes for the entrepreneurs and managers to recognize and respond to the inefficiencies caused by the noise injected into the price system by monetary manipulation.
In addition, the cumulative correlation between the lagged inflation-rate variance and vertical integration is positive. Ceteris paribus, this implies that an increase in the variance of the inflation rate will, following a lag, increase the extent to which the economy relies upon administrative hierarchies to allocate resources. It follows from this that increased inflation-rate variance causes the cost of using the price system to rise by more than the increase in the cost administrative command.
The findings have two important normative implications. First, the result that increased price-level variability tends to increase the extent of vertical integration in the economy brings to the fore an additional real cost of inflation not identified previously. Moreover, the evidence suggests that this effect is cumulative, and occurs as a result of price variability that is not anticipated by resource owners.
Secondly, given that increased price variability raises the cost using the price system by more than it raises the cost of administrative command, it appears that the value of a monetary rule which promotes price stability is greater to a market economy than to one guided by central planning. Thus, the effects of inflation go beyond those normally associated with changes in the purchasing power of money to encompass changes in the boundaries of the institutions within which economic transactions are undertaken.
Table : Summary Statistics
Table : Final Single-Equation Models
Table : Tests of Sets of Coefficients Against Zero
Table : Dependent Variable: First Difference of GNP-to-Sales Ratio, 1947-1982
Table : Anticipated vs. Unanticipated Price Variability
|Printer friendly Cite/link Email Feedback|
|Author:||Boudreaux, Donald J.; Shughart, William F., II|
|Publication:||Atlantic Economic Journal|
|Date:||Jun 1, 1989|
|Previous Article:||Annuities and inflation.|
|Next Article:||Campaign contributions by political parties: ideology vs. winning.|