Anti-Inflationary Demand Management in the United States: A Selective Industry Approach1
Author:
Mr. Morris Goldstein https://isni.org/isni/0000000404811396 International Monetary Fund

Search for other papers by Mr. Morris Goldstein in
Current site
Google Scholar
Close

This paper is adapted from a section of the author’s unpublished doctoral dissertation, “A Selective Key-Industry Approach to Anti-Inflationary Policy in the United States,” submitted to New York University in October 1971.

Abstract

This paper is adapted from a section of the author’s unpublished doctoral dissertation, “A Selective Key-Industry Approach to Anti-Inflationary Policy in the United States,” submitted to New York University in October 1971.

It has become generally accepted in the United States that it is the responsibility of the federal government to pursue, among other objectives, the goals of full employment, price stability, and balance of payments equilibrium. Postwar economic experience suggests, however, that it is difficult, within the present structure of the U.S. economy, to attain these three goals simultaneously or even to come close to achieving them. Accordingly, relevant policy questions center upon the extent to which it may be necessary to depart from one of these goals in order to come closer to another goal or to a combination of goals. More specifically, much attention has been paid to determining what trade-offs exist between full employment and price stability and between full employment and balance of payments equilibrium.

Enpirical studies for the United States and the United Kingdom by Phillips [42],2 Lipsey [29], Perry [39], Black and Kalejian [3], and Holt et al. [22] suggest that attempts to reduce the rate of increase of money wages or prices by reducing the level of aggregate demand will be accompanied by higher rates of unemployment. Policymakers, therefore, seem to be faced with a difficult dilemma in that higher levels of output and employment are usually paired with higher rates of inflation.

This paper explores a selective key-industry approach to anti-inflationary policy as one possible method of improving the short-run trade-off between price stability and unemployment in the United States. The essence of this approach is that policymakers exert some control over the composition, as well as the level, of aggregate demand when framing an anti-inflationary policy. The basic hypothesis is that there are substantial differences among industries in both their price and employment responses to demand changes. These interindustry differences, in turn, suggest that demand-reducing policy be directed primarily toward those industries where the combination of price and employment responses to demand changes is most favorable. In industries where this combination of price and employment responses to demand changes is relatively 3 unfavorable, greater reliance could be placed on anti-inflationary policies that do not primarily involve changes in demand, e.g., wage-price guideposts, manpower policy, import liberalization, stronger antitrust enforcement, etc. By assigning anti-inflationary policy instruments in accordance with their “comparative advantage,” it may be possible to improve the overall effectiveness of anti-inflationary policy.4

Given the above hypothesis, the primary purpose of this paper is to make an attempt (admittedly a rough attempt) to identify U.S. manufacturing industries that would be relatively good targets for selective demand reductions. The procedure involves specifying a number of characteristics for a hypothetical, good target industry, and then attempting to determine the industries that best satisfy these characteristics. To keep the study in manageable proportions, attention is confined to two-digit industries of the Standard Industrial Classification (SIC) in the U.S. manufacturing sector, but the approach could also be applied to other sectors and other economies, data permitting.

The paper is organized into four sections. Section I presents an overview of anti-inflationary policy and discusses the rationale for a selective key-industry approach to the inflation problem. Section II suggests three characteristics for a good target industry under a regime of selective demand reductions and gives reasons for anticipating that industries will differ in their price and employment responses to demand changes. Section III contains a series of tests—primarily industry price change and industry employment-demand regressions—designed to help identify those industries that best satisfy the characteristics of a good target industry. Section IV summarizes the findings on the best overall target industries, the principal limitations of the study, and the conclusions.

I. The Assumed Anti-Inflationary Policy

It will be assumed here that the policy objective is to reduce price increases to an acceptable (undefined) rate, at the lowest possible cost in unemployment.5 This implies not only that price stability is to be regarded at a given time as the primary target of economic policy but also that the instrument assigned to the target of price stability should minimize the employment losses associated with reaching that target.

As for the policy instrument, it will be assumed that: (1) it is of the demand-changing variety, e.g., monetary or fiscal policy; and (2) it can be used effectively in a selective manner, i.e., the authorities can and will control the incidence of a change in aggregate demand among the different industries or sectors in the economy. The anti-inflation decision is therefore assumed to be twofold: policymakers decide (1) how much to change the level of aggregate demand, and (2) how to distribute this change in aggregate demand throughout the economy. A change in industry or sector demand is defined as a change in active attempts to purchase at the prevailing level of prices. On the assumption that industries differ in their price and employment responses to demand changes, each demand-changing policy instrument can be viewed as potentially many instruments because (1) each instrument can distribute a given change in aggregate demand in many different ways, and (2) each of these distributions will very likely have a different effect on the policy target.6 This reasoning suggests that the Phillips curve will be closer to the origin for some distributions of aggregate demand than for others.7

The assumption that a successful anti-inflationary policy should include some control over the composition of aggregate demand can be defended for several reasons. First, as Schultze [45] and Eckstein and Fromm [11] have pointed out, where inflation results not from total excess demand in the economy but rather from excess demand or excess market power in a few sectors or industries, it is extremely difficult and costly in terms of overall employment and economic growth to slow these price increases by aggregate demand policies. Further, since the incidence of demand changes is not controlled, there is no guarantee that demand will be reduced where demand reductions are most needed.8 Second, a selective policy offers policymakers the option of singling out for special attention those industries or sectors which are particularly important for the achievement of secondary or intermediate policy targets, such as balance of payments equilibrium, rapid economic growth, or wage stability. For example, where wage bargains in a number of key industries set the pattern for wage bargains in other industries (the spillover hypothesis), a selective policy can apply pressure aimed at moderating wage bargains in the key industries, whereas an aggregate policy cannot do so without also at the same time applying pressure in the other industries.

Finally, to the extent that shifting the composition of aggregate demand is feasible, a selective policy permits policymakers to focus more of the impact of demand-changing anti-inflationary policy on those industries and sectors where the policy has the greatest relative prospects for success, i.e., industries where there is a relatively high probability that demand decreases will lead to both relatively large price decreases and relatively small employment losses.

Correspondingly, demand-reducing policy can be steered away from those industries and sectors where the policy is likely to have a low chance for success. Although the knowledge of industry price and employment behavior is admittedly imperfect, there are some indications (Kaplan et al. [25] and Hall and Hitch [19]) that there are industries in which prices respond very little to changes in demand.9 In such industries, where both the firm’s discretion over its pricing decision and its costs of changing prices are high, the anti-inflationary impact of demand reductions is likely to be low relative to industries in which price changes are more closely related to demand changes. If price decreases are to be obtained in the demand-insensitive industries, anti-inflationary policy will have to exert an impact on the cost and profit variables which primarily influence price changes. These variables, in turn, may well be most efficiently affected by policies that do not primarily involve changes in demand, such as wage-price guideposts, import liberalization, stronger anti-trust enforcement, repeal of retail price maintenance laws, and other measures which increase or approximate competition in product markets. Similarly, Oi [38] and Hamermesh [20] show that employment is not very sensitive to changes in demand or changes in output in industries where the costs of changing employment are high because of high fixed costs of hiring and training or contractual employment commitments. Therefore, the employment costs of demand-reducing policy are likely to be lower in these industries than in industries where employment changes and output changes are more closely related.

In short, because demand-reducing anti-inflationary policy is likely to be more effective and less costly when applied to some industries than to others, there can be benefits from exercising some control over the composition of changes in aggregate demand. This paper presents one possible screening procedure for identifying the preferred industry targets for selective demand reductions and applies it to U.S. manufacturing industries.

II. Characteristics of a Good Target Industry

Given the assumptions above about the policy objective and the policy instrument, three characteristics which might reasonably be assigned to a relatively good target industry under a regime of selective demand reductions are: (1) a relatively high price response to demand changes; (2) a relatively low employment response to demand changes; and (3) relatively large interindustry price and wage linkages, i.e., price or wage changes in the target industry have an important influence on price and wage decisions in other industries. While these three characteristics are not the only relevant considerations for selecting industries for demand reductions, they are important for minimizing the aggregate trade-off between price changes and unemployment.10 The rationale behind the selection of these characteristics is discussed below.

Relatively high price response to demand changes

If anti-inflationary demand-reducing policy is to lead to a decrease in prices or to a slowdown in the rate of price increase, it is necessary for industries which experience the demand reductions to reduce their prices or to reduce their rate of price increase. It is likely, however, that the expected price change for a given demand change will differ from one industry to another. As Galbraith [16], Duesenberry [7], Eckstein [10], and Eckstein and Fromm [12] have pointed out, different industries or sectors have different market structures which suggest different rates of price adjustment to changes in demand. In turn, much of this interindustry difference in price behavior may be a reflection of interindustry differences in the costs of changing prices. The brief review below of price behavior under pure competition and imperfect competition highlights these differences.

Pure competition

In a purely competitive market, no single firm is able to influence the market price of its output. The firm is a price taker, and it must accept the industry price as given. Given homogeneous products, no firm can charge a price higher than the industry price; nor is there any reason to charge a lower price, since the firm can dispose of all its output at the given industry price. Profit maximization requires that price and output be jointly determined where marginal revenue equals marginal costs.11

When demand decreases (increases) in such a market, there is a rapid decrease (increase) in price for the available supply. The price change proceeds pari passu with the demand change; i.e., there are no price adjustments to past demand changes. Since no single producer can halt the price response to the demand change and since the revenue losses from failing to equal the new equilibrium price are high, there are high costs involved in not changing prices. There are also output and, perhaps, inventory and order backlog adjustments to demand changes, but these adjustments are less important as price changes are more effective in removing supply-demand disequilibria.12 In short, in a purely competitive market where price responds to the difference between demand and supply, policies which reduce demand should produce rapid decreases in price.13

Imperfect competition

In an imperfectly competitive market where each firm’s output is slightly or markedly differentiated, each firm has some control over its pricing decision and is therefore a price setter rather than a price taker.

When the demand changes in an imperfectly competitive market, firms may not always adjust with a change in price. This conclusion is based on the assumptions that there are costs to changing prices in such a market and that firms weigh these costs against the costs of alternative adjustments to the demand change. The costs of changing prices arise from many sources. First, there are costs associated with uncertainty in the firm’s demand curve.14 Since each firm has discretion over its pricing decision, a firm cannot be certain about the shape of its demand curve because it cannot always predict correctly the price response of its rivals to its own price change.15 If the firm misjudges its price elasticity of demand, it faces a possible loss in total revenue. Where price does change, Adelman [1, p. 271 suggests that the price change may be small and gradual while the firm feels its way along its demand curve in search of its profit-maximizing price. This also implies that there will be price adjustments to past demand changes so that prices may increase when the current state of demand is not excessive. Second, there can be significant transactions costs associated with changing prices. When the firm produces many products, there may be printing costs to inform buyers of price changes. Also, some of the firm’s customers may desire relatively stable prices, because this facilitates their production scheduling.16 Finally, when the firm has long-run profit objectives, such as earning a target rate of return on its capital over some time period, and a desire for stability, it may be reluctant to use price changes to equilibrate supply and demand in the short run.17

In order to avoid price changes, a firm may choose any of a series of alternative adjustments to the demand change. If demand exceeds supply at the original price, the firm has the following adjustment options: (1) it can draw down its stock of inventories; (2) it can allow its backlog of unfilled orders to grow; (3) it can allow some new orders to be abandoned (when customers refuse to wait for later delivery); (4) it can attempt to increase production at the original price by more intensive use of existing factor inputs; or (5) it can adopt some combination of these four adjustments, depending on the relative costs of the different adjustment responses.18 When the costs of changing prices are high relative to the costs of other adjustments, the relationship between price changes and demand changes is likely to be weak.

The hypothesis that firms in imperfectly competitive markets consider the costs of changing prices has also received some support from interview studies of the pricing behavior of large U.S. and U.K. industrial firms. Kaplan, Dirlam, and Lanzilotti [25, pp 127-80] and Hall and Hitch [19, pp. 18-45] found that many corporations were using pricing formulas that call for relatively few price changes. For example, prices may be set so as to obtain a target rate of return on the firm’s capital stock when the firm is producing at a standard or normal level of output. Full cost pricing, another frequently mentioned formula, calls for setting price as a constant markup on unit costs evaluated at the standard level of output.19 In these pricing formulas, prices are changed when the cost of producing the standard output changes because of changes either in technology or in input prices, but prices are not changed when demand changes.20 The policy implication of this kind of pricing behavior is that demand-reducing anti-inflationary policy is not likely to be very successful in inducing price changes in industries where demand changes play only a minor role in the price-setting process.

Since industries can be expected to differ in their price responses to demand changes, there may be benefits from steering selective demand reductions toward those industries where there is a relatively high probability that demand decreases will lead to price decreases.

Relatively low employment response to demand changes

Since full employment is also a target of policy, the industries selected for the demand reductions should lay off relatively few workers. There are several reasons for expecting industries to differ in their employment responses to demand changes. Three important factors are: (1) differences in output responses to demand changes, (2) differences in factor intensities of industry production functions, and (3) differences in the costs of changing employment.

Price theory suggests that when demand contracts, the consequent decline of prices diminishes the need for reductions in output; when this price decline is pronounced, employment tends to be better maintained than when this decline is modest. Thus, a good target industry should respond to a decrease in demand with a decrease in price, for if price does not fall, output will most likely fall and employment losses will follow. The high price response characteristic should therefore be compatible with the low employment response characteristic.

Once output falls in an industry, the industry’s employment response depends on the nature of its production function, among other factors. If the industry’s production function is relatively labor intensive, then a given decrease in output will produce a larger employment loss than in an industry with a less labor-intensive production process. Since industries differ in the factor intensities of their production processes, the less labor-intensive industries can be regarded, ceteris paribus, as better targets for selective demand reductions.

Finally, and perhaps most importantly, industries differ in their employment responses to demand changes because of differences in the costs of changing employment. When an industry’s production function contains more than one factor of production and when the coefficients of production are not completely fixed, there arises a relative factor price (substitution) decision in the demand for employment. As Oi [38] and Nadiri and Rosen [36] have shown, the costs of changing employment and capital are not only a function of the relative rental charges of the factors (the wage-rental ratio) but also of the fixed transaction or adjustment costs associated with changing the levels of the factors. For employment, these fixed costs are primarily hiring and training costs. Similarly, there are searching, waiting, and installation costs for capital. Oi [38] argues that when account is taken of these hiring and training costs, labor is transformed from a completely variable to a quasi-fixed factor of production, at least in the short run. Since hiring and training costs represent fixed costs to the firm in the form of investment in its workers, the firm attempts to discount or spread these costs over a long period. The higher these costs are, the more reluctant the firm is likely to be to change employment each time demand for its output changes. Rather, the firm may hoard its workers and respond to an output change by varying either the utilization rate (hours worked per man) of its existing labor stock or the utilization rate (services per unit of capital) of its existing capital stock.21 As in the case of the price decision, the firm may change employment only if demand changes (and output changes) are judged to be large and permanent.22 The particular factor response of a firm to demand changes depends on the relative costs of these responses. Therefore, unless the entire vector of relative factor prices—the user cost of employment, the price of hours per man, the user cost of capital, and the price of services per unit of capital—are identical in all industries, the change in employment for a given change in output is likely to differ across industries.23

Since both wage costs and hiring and training costs are likely to be positively related to the skill level of an industry’s work force, a good target industry should, ceteris paribus, have a relatively highly skilled work force, because such a work force would be most consistent with a relatively low employment response to demand changes. In addition, if the target industries employ relatively highly skilled labor, those workers who are released when demand is reduced should have less trouble finding employment (although not necessarily at the same job) elsewhere in the economy.24 This position reflects the view that policymakers should be concerned not only with how many workers become unemployed due to anti-inflationary demand reductions but also with what types of workers become unemployed, because this factor in part determines the duration of unemployment.

Relatively large interindustry price and wage linkages

If the price objective is to reduce the rate of increase of prices in a few large sectors of the economy rather than in just a few industries, it is desirable that price and wage decreases in the target industries lead to price decreases in other industries—i.e., that the target industries have relatively large interindustry price and wage linkages. If the target industries possess this characteristic, then selective demand reductions may go almost as far in reducing prices, but at less cost, than more widespread demand reductions.25

Price decreases in one industry (called industry A) could be transmitted to other industries through the use of industry A’s output as an input into other industries, i.e., through the economy’s input-output relationships. A price decrease in industry A would then show up as a decrease in the cost of materials in other industries. Assuming that these cost decreases are passed forward to consumers rather than absorbed into higher profits, the price decrease in industry A should lead to lower prices in other industries. This line of reasoning suggests that the steel industry would be a better target industry than the tobacco industry, because of the greater forward linkages of steel in the economy’s input-output relationships.26

Price and wage changes in one industry can also affect price and wage changes in other industries through price or wage leadership. For example, Eckstein and Wilson [13] have shown that wage bargains in a number of key industries set the pattern for wage bargains in other industries (the spillover hypothesis).27 If wage bargains in the key industries can be moderated, then wage bargains in the imitating industries may fall in line and the prospects for price stability will be improved. This suggests that the key-wage-bargain industries will, ceteris paribus, be better targets than the nonkey-wage-bargain industries.28 Similarly, in industries where price leadership prevails, the price-leading firm would be a good target.

III. Empirical Findings

Given the three characteristics suggested above for a good target industry, the next task is to attempt to determine empirically which manufacturing industries best satisfy these characteristics.

Testing for price response

In order to identify those industries which could be expected to display a relatively high price response to demand changes, price change regressions were estimated for each of the 20 manufacturing industries for the period 1954-67.29 Annual data were used because consistent quarterly data for prices, cost of materials, and real output were unavailable for two-digit Standard Industrial Classification (SIC) industries. Given the limited number of observations, the specification of these price change regressions had to be kept relatively simple.30

Following earlier investigators, the a priori hypothesis about industry price changes is that price changes are a function of cost changes and of excess demand within the industry.31 However, since the costs of changing prices may differ from one industry to another, it is unlikely that price changes will relate to the same cost and demand variables in all industries; i.e., in industries where the costs of changing prices are high, prices may respond to cost changes considered permanent but not to temporary cost changes or to demand changes. Therefore, several alternative price regressions were estimated for each industry by a two-step procedure. Industry price changes were first regressed on several alternative combinations of changes in labor costs and in material costs. These regressions were then re-estimated after adding various alternative demand-change proxies.

The regressions relating price changes only to cost changes took the following forms:

Δ ln P = ln α 0 + α 1 Δ ln C M + α 2 Δ ln U L C + u ( 1 a )
Δ ln P = ln α 0 + α 1 Δ ln C M + α 2 Δ ln C M H α 3 Δ ln ( R Q / M H ) + u ( 1 b )
Δ ln P = ln α 0 + α 1 Δ ln C M + α 2 Δ ln U L C N + u ( 1 c )
Δ ln P = ln α 0 + α 1 Δ ln C M + α 2 Δ ln C M H α 3 Δ ln ( R Q / M H N ) + u ( 1 d )
Δ ln P = ln α 0 + α 1 Δ ln C M N + α 2 Δ ln U L C + u ( 1 e )
Δ ln P = ln α 0 + α 1 Δ ln C M N + α 2 Δ ln U L C N + u ( 1 f )

where P is the implicit price deflator of the total output of an industry, CM is the price deflator of the cost of materials used by the industry, ULC is the industry’s unit labor cost, CMH is compensation per man-hour for the industry, RQ/MH is real output (real value added) per man-hour for the industry, α is a constant, and u is the error term. ULCN, CMN, and RQ/MHN are approximations for standard unit labor cost, standard cost of materials, and standard real output per man-hour, respectively. In most cases, standard N denotes a weighted, three-year moving average of the variable with weights of 0.5 for year t and 0.25 each for years t-1 and t-2.32

A precise description of the method of construction and data sources for each of these variables is given in Appendix I; Δ is the first difference operator, and ln is the natural log of the variable. All variables except the standard variables relate to the same time period. The specification was in terms of changes in the logarithms of the data. First differences were used because the concern is with industry price changes rather than with price levels. The double log-linear form derives from the assumption of Cobb-Douglas industry production functions. The percentage change in the price of output is therefore related log-linearly to the percentage change in the prices of inputs. The double log-linear form also means that the estimated coefficients in the price equations can be interpreted as elasticities. Ordinary least squares was the method of estimation in all cases.

The expected signs of the coefficients for all the cost variables are positive except for the productivity variables Δ(RQ/MH) and Δ(RQ/MHN), which have negative signs—i.e., α1, α2 > 0 and α3 < 0. These costs-only price regressions relate price changes to several variants of changes in variable costs.33 Equations (1a) and (1b) use changes in actual labor costs and materials costs,34 while equations (1c), (1d), and (1e) use combinations of changes in actual costs and changes in standard costs. The standard cost variables are included because several writers suggest that prices respond only to those cost changes considered permanent. Since annual cost changes are likely to be viewed as more permanent than quarterly cost changes, the a priori expectation is that the standard cost variables will not be found as important as they were in earlier studies based on quarterly data.35 Equation (le) relates price changes to changes in standard costs alone.

After the costs-only price regressions were estimated for each industry, they were re-estimated on the basis of the addition of various demand-change proxies, which were entered into the regressions (one at a time for reasons of collinearity) to capture the influence of demand changes on price changes.36 The demand-change proxies tested in the regressions were: the change in the ratio of unfilled orders to shipments, both current and lagged one period—Δ(UO/S)t and Δ (UO/S)t-1; the change in the ratio of inventories to shipments, current and lagged one period—Δ(INV/S) t and Δ(INV/S)t-1; the level (CUW) and the change in the Wharton capacity utilization index (ΔCUW); the change in real output (ΔRQ); and the change in standard real output (ΔRQN). The expected signs of the coefficients for these demand-change proxies are all positive, except for the inventory variables which have negative expected signs. Only three industries (fabricated metals, machinery, and electrical machinery) had available data for all of the demand-change proxies. In the other industries, the regressions contained only demand-change proxies for which data were available.

The 20 industries were then divided into two groups approximating their relative price sensitivity to demand changes. The first group, the cost-plus-demand responsive group, consists of those industries in which demand changes seem to affect price changes directly. The second group, the cost responsive group, consists of those industries in which demand changes seem either to have no effect on price changes or to affect price changes only through changes in costs.37 An industry was placed in the cost-plus-demand responsive group if it satisfied two conditions: (1) if a cost-plus-demand price regression could be found for that industry which explained price changes better than the best costs-only price regression as measured by the adjusted R¯2, and (2) if the coefficient on the demand-change proxy in the best cost-plus-demand price regression was statistically significant at the 90 per cent probability level as measured by the coefficient’s t value. If an industry satisfied only one or neither of these two conditions, it was placed in the cost responsive group.

Of the 20 manufacturing industries, 14 fell into the cost-plus-demand responsive group and 6 fell into the cost responsive group. The industry composition of the two price groups may be seen in Table 1. Ceteris paribus, the cost-plus-demand responsive industries are assumed to be better targets for selective demand reductions than the cost responsive industries.38

Table 1.

Comparison of Cost-Only and Cost-Plus-Demand Explanations of Industry Price Changes, 1954-671

article image
article image
article image

Symbols: N = “standard” form of the variable; L = lagged form of the variable; * = period 1955-67. See Appendix I for explanation of symbols used. All variables are expressed in natural logs, and the numbers in parentheses below the coefficients are t values; Δ is the first difference operator; R2 is the coefficient of multiple determination; R¯2 is this coefficient adjusted for degrees of freedom; SEE¯ is the standard error of estimate (also adjusted for degrees of freedom); and DW is the Durbin-Watson statistic for serial correlation in the residuals.

The industry price regressions for the two industry groups appear in Table 1. For each of the industries in the cost-plus-demand responsive group, two regressions are presented: the best costs-only explanation of price changes and the best costs-plus-demand explanation of price changes. For the cost responsive industries, only one regression is presented: the best costs-only explanation of price changes, which is also the best explanation of price changes.

Analysis of the industry price change regressions

The industry price regressions corresponded well with a priori expectations. For most industries, changes in cost and demand factors explained a large fraction of the total variance of price changes. For 19 of the 20 industries (miscellaneous manufacturing being the exception), the regressions explained more than 60 per cent of the total variance of price changes; for 15 industries, the explained variance was greater than 80 per cent. Given the rather simple nature of these regressions, as no account was taken of capital costs, profits, taxes, wage-price guideposts, etc., the goodness-of-fit of the regressions was generally satisfactory.39 Change in the cost of materials (ΔCM) was almost invariably the most powerful explanatory variable, but its influence is no doubt exaggerated due to some overlapping in the price indexes for output price and cost of materials.40 The vast majority of the labor cost variables were significant and carried the expected signs. The demand-change proxies were significant in the regressions for 70 per cent of the 20 industries, but their influence on price changes was always less than that of the sum of cost factors.41 This can be seen in Table 2, where the beta coefficients for the cost and demand factors are presented.42 Almost all of the demand-change proxies carried the expected signs.43 As was anticipated, the standard cost variables did not perform as well as in price studies based on quarterly data, but in 7 of the 20 industries the best regression contained a standard cost variable (usually standard unit labor cost or standard real output per man-hour).44 The number of observations is too small to indicate much about autocorrelation in the residuals, but the Durbin-Watson statistics suggest that the autocorrelation was not serious in most industries. Autoregressive corrections did, however, seem to be necessary in the regressions for apparel, leather, textiles, electrical machinery, printing, and instruments.45 Multicollinearity between the cost variables (usually ΔRQ/MH or ΔCM) and the demand-change proxies emerged as a serious problem in a few industries (for example, apparel and leather), but in most others it did not appear to be very pronounced.46

Table 2.

Relative Importance of Cost and Demand Factors in Explaining Industry Price Changes, with Beta Coefficients for the Industry Price Change Equations 1

article image

The beta coefficients are computed from each industry’s best price equation (as given in Table 1). See Appendix I for symbols used.

The demand factor refers to the demand-change proxy used in the industry’s best price equation.

For the purposes of this study, the most encouraging aspect of the price change equations is the support which they lend to the hypothesis that industries differ in their price response to demand changes. For 30 per cent of the 20 industries (the cost responsive group), an independent influence of demand changes on price changes could not be found. While in some of these industries demand factors may have been operating through the cost mechanism, there appear to be other industries in this group where demand factors have very little, if any, influence on price changes. Primary metals is a case in point.47 Further, within the group of industries where demand changes are significant, there appear to be differences in the degree to which prices respond to demand changes. For example, the price responses to demand changes seem to be stronger in the petroleum and textile industries than in the fabricated metals or machinery industries. In each of the former two industries, three demand-change proxies were found with larger and more significant coefficients than that of the best demand-change proxy in each of the latter two industries.

In short, while it is readily acknowledged that these industry price regressions are only a first step toward determining whether and how demand changes affect price changes in different industries, they do suggest that demand-reducing anti-inflationary policy is likely to be more effective in reducing prices in some industries than in others.

Testing for employment response

In an attempt to identify industries with relatively low employment responses to demand changes, employment demand regressions were estimated for each of the 20 industries for the period 1954-67. Annual data were used once again so that the employment responses would have the same time dimension as the price responses.

The basic hypothesis about the demand for industry employment is that this demand is composed of a scale decision, a relative factor price (substitution) decision, and an adjustment decision.48 The employment demand function is treated in two parts: (1) a function for desired employment, and (2) an adjustment process which relates actual employment to desired employment.

Assume that the desired level of employment (E*) of an industry is a function of its real output (RQ), the ratio of the wage to the rental price of capital (W/r), and the state of technology (T). The desired employment function can be written as: †

E t * = a + b 1 R Q t + b 2 ( W / r ) t + b 3 T ( 1 )

where b1 > 0, b2, b3 < 0. Next assume that there is an adjustment process by which the actual change in employment (EtEt-1) is related to some fraction (α) of the difference between desired employment in the current period and actual employment in the preceding period. This familiar stock adjustment equation can be written as:

E t E t 1 = α ( E t * E t 1 ) . ( 2 )

Equation (2) is equivalent to assuming a Koyck-type distributed lag and can be rewritten in levels form as:

E t = α E t * + ( 1 α ) E t 1 . ( 3 )

Substituting equation (1) for Et* in equation (3) yields an equation containing only observable quantities:

E t = a α + b 1 α R Q t + b 2 α ( W / r ) t + b 3 α T + ( 1 α ) E t 1 . ( 4 )

A log-linear form of equation (4) was used to estimate the industry employment equations. The coefficients in equation (5) are then equivalent to elasticities:49

ln E t = ln B 0 + B 1 ln R Q t + B 2 ln ( W / r ) t + B 3 T t + ( 1 α ) ln E t 1 + e t . ( 5 )

In equation (5), E is the annual level of all employees in an industry, RQ is the annual level of real value added for the industry, W is annual compensation per man-hour for the industry, r is the annual implicit price deflator of producers’ durable equipment (assumed to be equal for all industries) from The National Income and Product Accounts of the United States [55], T is time, the proxy for technological change (where T = 1 for 1954, T = 2 for 1955, etc.), and e is the error term. The variables are described in greater detail in Appendix I.

The 20 industries were divided into two groups on the basis of the size of each industry’s estimated elasticity of employment with respect to output. The 9 industries with the lowest elasticities of employment with respect to output were placed in the low employment sensitivity group and the 9 industries with the highest elasticities were placed in the high employment sensitivity group.50 Food and tobacco were excluded from these groups because their estimated elasticities of employment with respect to output were statistically insignificant. However, the size of the coefficients on the real output variable in these two industries suggests that they would probably fall into the low employment sensitivity group. The industry composition of the employment groups may be seen in Table 3. Ceteris paribus, the industries in the low employment sensitivity group are assumed to be better targets for selective demand reductions than the industries in the high employment sensitivity group.

Table 3.

Industry Employment Demand Regressions, 1954-67, All Employees 1

ln E t = ln B 0 + B 1 ln R Q t + B 2 ln ( W / r ) t + B 3 T + ( 1 α ) ln E t 1
article image
article image

All variables, except time (T), are expressed in natural logs, and the numbers in parentheses below the coefficients are t values. The summary statistics (R2, R¯2, SEE¯, and DW) are defined as in Table 1. For those industries in which the wage-rental ratio (W/r) was statistically insignificant at the 90 per cent probability level, the regression presented above excludes this variable.

The industry employment demand regressions are presented in Table 3.51 The industries are listed from lowest to highest in the order of the size of their estimated elasticities of employment with respect to output.

Analysis of the industry employment regressions

The results of the industry employment demand regressions are encouraging in a number of respects. First, the goodness-of-fit of the regressions was generally satisfactory, and the explanatory power did not rest too heavily on the lagged dependent variable. For 18 of the 20 industries, the regressions explained more than 90 per cent of the total variance of industry employment, and for 14 industries, the explained variance was greater than 95 per cent. The real output variable was statistically significant at the 90 per cent probability level and carried the expected sign in the regressions for 18 of the 20 industries. This supports the well-established conclusion that there is a strong scale effect in the demand for employment. The wage rental variable, despite its simple nature, was significant and carried the expected negative sign for 8 of the 20 industries. This suggests that relative factor prices (substitution effects) also play a role in the demand for labor.52 The simple proxy for technological change (time) was significant for over half of the industries and usually carried the expected negative sign. The lagged dependent variable was significant for 14 of the 20 industries, suggesting that the adjustment of actual to desired employment is not complete for many industries even over a period of one year.53 For 17 of these 20 industries, the adjustment coefficient (α) fell within its expected range of 0 < α < 1.54 Multicollinearity was not a serious problem in these regressions, but there was some imperfect evidence of autocorrelation in the regressions for a few industries (rubber, petroleum, electrical machinery, chemicals, and instruments).55

Most importantly, these regressions provide support for the hypothesis that industries differ in their employment responses to output changes. If there is a close relationship between production (output) changes and demand changes across all industries, this means that industries will also differ in their employment responses to demand changes. Although all of the estimated elasticities of employment with respect to output were less than unity, there was considerable variation in the size of this elasticity across industries.56 For example, a 1 per cent decrease in output will lead to 0.2 per cent decrease in employment in printing and publishing but to 0.8 per cent decrease in employment in lumber. It therefore appears, as in the price behavior case, that significant incentives exist for choosing some industries rather than others for selective demand reductions, i.e., for exercising some control over the composition of changes in aggregate demand.

Testing for interindustry wage linkages 57

To obtain an approximation of the relative extent to which wage changes in a given industry affect wage changes in other industries, reliance has been placed on the findings of two previous studies of industry wage behavior: Eckstein and Wilson [13] and Perry [40]. These two studies selected a key-wage-bargain group or visible industry group by either examining the pattern of industry wage bargains or consulting a group of expert economists. The key-wage-bargain group in the Eckstein-Wilson study is a group of high-wage industries with strong industrial unions. These industries tend to prosper as a group because of considerable input-output connections among them. Also, a wage pattern is known to exist in these industries. As Eckstein and Wilson demonstrate, wage bargains in this group have an important influence on wage bargains outside this group. Perry’s visible industry group contains those industries that were supposedly most affected by the 1962-66 wage-price guideposts. Presumably, policymakers were concerned about potential bad pattern-setting effects of wage bargains in these industries. The industrial composition was very similar for Eckstein and Wilson’s key-wage-bargain group and for Perry’s visible industry group. Given the findings of these two studies, the 20 industries were divided into a key-wage-bargain group and a nonkey-wage-bargain group.58 An industry was placed in the key-wage-bargain group if it was included in both Eckstein and Wilson’s key-wage-bargain group and in Perry’s visible industry group. The industry composition of the two wage-bargain groups is shown in the last column of Table 4. Ceteris paribus, the industries in the key-wage-bargain group are assumed to be better targets for selective demand reductions than the industries in the nonkey-wage-bargain group. While some industries in the key-wage-bargain group, such as primary metals and motor vehicles, are surely more influential wage leaders than others in this group, all industries within each group are assumed to be equally desirable targets.59

IV. The Best Overall Target Industries, Limitations of the Analysis, and Conclusions

The best overall target industries

To determine which industries appear to be the best overall targets for selective demand reductions, it is necessary to bring together the results of the tests for the characteristics of a good target industry. Table 4 presents the distribution of good target characteristics by industry.

Table 4.

The Best Target Industries Grouped in Order of Distribution of Target Characteristics

article image
Symbols: + = has the target characteristic; − = does not have the target characteristic.

The best target industry is one that has all the characteristics of a good target industry, and the poorest target industry is one that has none. A brief examination of Table 4 indicates, however, that most industries fall into neither of these two categories; i.e., only petroleum and electrical machinery have all the target characteristics, while only furniture and miscellaneous manufacturing have none of them. Therefore, to evaluate the relative attractiveness of the remaining 16 industries, it is necessary to weight the target characteristics. A simple ordinal weighting pattern has been adopted. It is assumed that: (1) the price response characteristic has the highest weight, the employment response characteristic has a lower weight, and the key-wage-bargain characteristic has the lowest weight; and (2) the price characteristic has a higher weight than the sum of the weights for the other two characteristics. This weighting pattern is consistent with the assumed policy objective of reducing the rate of increase of prices at the lowest possible cost in unemployment.60

Given this weighting pattern for the characteristics, the 20 industries have been divided in Table 4 into 7 groups, each corresponding to a different degree of attractiveness as a target for demand reductions and composed of a set of industries with the same target characteristics. Group 1 consists of the most attractive target industries, Group 2 of the next most attractive industries, etc.

The screening procedure used in this study suggests that the 7 industries in Groups 1 and 2 would be the best overall targets for selective demand reductions. These 7 industries—petroleum, electrical machinery, food, tobacco, paper, leather, and instruments—appear to be the only manufacturing industries in which selective demand reductions are likely to lead to both relatively large price decreases and relatively small employment losses. Among these 7 industries, electrical machinery and petroleum emerge as especially attractive target industries, because they satisfy another characteristic (key wage bargaining) of a good target industry as well. The 11 industries which comprise Groups 3-6 represent examples of the familiar dilemma where either relatively large price decreases are paired with relatively large employment losses or where relatively small employment losses are paired with relatively small price decreases. If the price target took almost sole priority, then textiles, apparel, rubber, fabricated metals, machinery, lumber, and stone, clay and glass would be suitable targets. On the other hand, if employment losses were of chief concern, then printing, chemicals, primary metals, and motor vehicles would be suitable targets. The poorest targets for selective demand reductions appear to be furniture and miscellaneous manufacturing, the only two industries combining the undesirable characteristics of relatively low price response and relatively high employment response.

If the results of both the industry price regressions and the industry employment regressions are reliable, then estimates of the aggregate trade-off between inflation and unemployment in the U.S. manufacturing sector may be obscuring important interindustry differences in the extent of this trade-off. In other words, the aggregate relationship between price changes and the unemployment rate may represent a weighted average of quite varied individual industry relationships. More specifically, for the case of demand-reducing policy, the evidence suggests that this trade-off is likely to be less favorable for the 13 industries in Groups 3-7 than for the 7 industries in Groups 1 and 2. It may therefore be possible to obtain some marginal, but not necessarily insignificant, improvement in the trade-off between inflation and unemployment by modifying the industrial composition of changes in aggregate demand.

Important limitations of the analysis

The screening procedure used in this study to identify the best target industries for selective demand reductions suffers from the familiar limitations of partial-equilibrium analysis. It is therefore necessary to discuss the more important limitations of the industry price and employment analysis in order to stress the exploratory and tentative nature of the study’s empirical findings.

Interdependence of decision variables

One of the principal limitations of the industry price and employment analysis used in Section III is that an interdependent, simultaneously determined set of decision variables has been treated in a rather fragmented manner, so that only a partial picture is obtained of the process by which industry prices, employment, output, and demand changes interact. From a statistical viewpoint, the single equation approach may also mean that the estimated coefficients in the price equations are biased and inconsistent;61 therefore, it is not certain that all the industries with relatively high price response to demand changes are correctly identified. Also, because the true relationship between industry demand changes and industry output changes is unknown, the industry employment equations give only an approximation of industries’ relative employment responses to demand changes.

To illustrate the interdependence problem, consider the industry price change equations. In these equations, price changes were treated as the endogenous dependent variable, while inventory changes, unfilled order changes, output changes, and the capacity utilization rate were treated as alternative exogenous demand-change proxies. The difficulty with this procedure is that just as the demand-change proxies can and do affect price changes, so can price changes affect each of the demand-change proxies. For example, given an industry with a downward-sloping demand curve, price changes in the industry can be expected to lead to changes in the demand for the industry’s output. These demand changes, in turn, are likely to lead to production changes, inventory changes, unfilled order changes, and possibly to new price changes. Thus, while in some cases demand changes may be considered exogenous (due to factors other than price, like income changes), in other cases they may be induced by price changes.62 In the latter case, the feedback from price changes to the demand-change proxies violates one of the assumptions underlying single equation estimation of price changes, thus producing biased and inconsistent estimates of the effect of demand changes on price changes (see Johnston [24, pp. 231-34]).

Unfortunately, there are no easy solutions to the simultaneity problem between price changes and the demand-change proxies. To obtain a more complete picture of the dynamics and linkages in the firm’s production decisions, it would be necessary to specify and estimate a simultaneous equations model. In such a model, each of the firm’s decision variables (price, production, inventories, and unfilled orders) would be a function of the other decision variables, its own lagged values, and certain exogenous variables. Models of this type by Mills [32, pp. 82-260] and Hay [21, pp. 531-45] exist in the literature, but they are still rather exploratory. Also, given the large number of variables in such models, the extensive data requirements preclude estimating such a model for a large number of two-digit SIC industries. Alternative methods of dealing with the simultaneity problem also involve serious difficulties. If an alternative estimating technique is employed to remove the simultaneous equations bias, such as two-stage least squares, there is the problem of obtaining good equations for inventory changes, or unfilled order changes, or the capacity utilization rate for many industries. A simple alternative procedure is to use a proxy for demand changes which is not a function of industry price changes and which is also not under the control of the industry, such as changes in GNP. Obviously, the difficulty with this approach is that changes in GNP, or other aggregate demand-change proxies, are not likely to be good measures of changes in industry demand, because the change in industry demand and the change in aggregate demand can often be in opposite directions.

If it is conceded that the industry price equations may be subject to some simultaneous equations bias, the relevant question becomes how serious this bias is likely to be. There are a number of reasons for suspecting that the size of this bias will be small. First, the available empirical evidence suggests that the effect of price changes on inventory changes or unfilled order changes is small in U.S. manufacturing; i.e., the feedback from price changes to the demand-change proxies should be small.63 Second, for the few industries where the best demand-change proxies were lagged one period, there should be no feedback from price changes to demand changes, since this period’s price change should not affect last period’s demand change. Third, since the price equations are conducted at the broad, two-digit industry level, substitution possibilities induced by price changes may be quite limited; e.g., if the industry price of fabricated metals increases, the only alternative source of supply may be imports. Finally, bias in an estimator is a less serious problem when something is known about the direction and size of the bias. It has already been argued that the size of the bias should be small. With respect to the direction of the bias, the coefficients on the demand-change proxies should be biased downward, assuming downward-sloping demand curves. Thus, the negative effect of price changes on demand changes should offset some of the positive effect of demand changes on price changes; i.e., if no feedback from price changes to demand changes occurred, the coefficients on the demand-change proxies would be higher. One consequence of this downward bias would be an underestimation of the number of industries in which demand changes affect price changes.

The industry employment analysis also describes only one step in a dynamic, interrelated process by which demand changes lead to employment changes. One potentially important link excluded from the analysis is the relationship between demand changes and output changes. This link becomes important in determining whether the industry employment equations indicate employment responses to demand changes or merely employment responses to output changes. The problem is that changes in output may not be a good proxy for changes in demand. More specifically, when demand changes, the industry can activate all of its decision variables, only one of which is its control over the rate of production. If an industry responded to a demand change primarily with price, inventory, and unfilled order adjustments, and only residually with output changes, then that industry could have a low employment response to demand changes even if its employment response to output changes was very high. Therefore, if the industry employment equations are to be an accurate measuring rod of interindustry differences in employment responses to demand changes, it is necessary to assume that the relationship between demand changes and output changes is quite uniform across industries. No conclusion about the empirical validity of this assumption can be offered in this study.

Since the industry price change and employment demand equations capture only part of the dynamic interaction among demand changes, price changes, employment changes, and changes in other product and factor market decision variables, the identification of the best target industries should be regarded as a first approximation.

Interindustry linkages

A second limitation of the price and employment analysis is that only partial account has been taken of interindustry price, wage, and employment linkages in the economy.64 Ideally, the estimates of the price and employment effects of reducing demand in a given industry should include not only that industry’s price and employment responses but also the induced price and employment responses in related industries. The equations in Section III give estimates of only the first round of price and employment changes, i.e., direct price and employment effects. If the industry ranking of indirect price and employment effects differs from that of the direct effects, then the selection of the best target industris would have to be revised. If, for example, an industry’s own elasticity of employment with respect to output is low, but output reductions in that industry lead to output reductions in industries with high employment elasticities, then that industry will be a bad target; however, that industry would have been placed in the low employment sensitivity group in Section III.65 Therefore, until proper account can be taken of these interindustry price and employment linkages, the selection of the best target industries should be regarded as preliminary.

Data limitations

Last, but not least, it should be recognized that the data used for the industry price and employment analysis suffer from a number of deficiencies. Aside from the limited number of annual observations, which precluded very sensitive analysis, the primary data problem was the overlapping in the price indexes for output price (the dependent variable) and cost of materials. The cause of this overlapping, which is more serious in some industries than in others, is that the output series at the two-digit SIC industry level sometimes includes as final product its own basic raw material input. (See Houthakker [23] on this point.) Thus, for example, in the lumber industry, the largest component of both raw material input and final product is unfinished lumber. Similarly, leather is counted in the output and input price indexes in the leather goods industry. The obvious consequence of this overlapping is that the high correlation between output price changes and changes in cost of material is often spurious; i.e., the coefficient on the cost-of-materials variable is biased upward. The solution to the overlap problem is to exclude intraindustry flows from both the output and input deflators; however, such industry price indexes are not generally available.66 In addition, there is a pressing need for wider industry coverage of the data for the demand-change proxies (new orders, unfilled orders, inventories, shipments, etc.), as well as for more complete measures of the costs of changing factor inputs. Until more accurate and comprehensive industry data become available, the empirical results of this study should be interpreted cautiously.

Conclusions

This paper has examined several aspects of a selective key-industry approach to anti-inflationary policy. The selective policy under investigation was a demand-reducing policy in which the authorities were assumed to exert some control over the composition, as well as the level, of aggregate demand. It was found that a number of reasonable arguments could be made for such a policy. One argument which has been neglected in the literature is based on the hypothesis that industries differ in both their price and employment responses to demand changes. If such interindustry differences exist, it will be possible to improve the aggregate trade-off between inflation and unemployment by directing a higher proportion of a given aggregate demand reduction into industries with relatively favorable trade-offs between price changes and employment changes. Correspondingly, in industries where price reductions cannot be obtained through demand reductions without incurring relatively large employment losses, greater reliance (but not exclusive reliance) could be placed on anti-inflationary policies of the nondemand-changing variety (wage-price guideposts, import liberalization, manpower policy, etc.). By assigning anti-inflationary policy instruments in accordance with their “comparative advantage,” it should be possible to improve the overall effectiveness of anti-inflationary policy.67

Given the above a priori case for a selective anti-inflationary policy, an attempt was made to identify those U.S. manufacturing industries that would be the best targets for selective demand reductions. Three criteria were suggested for a relatively good target industry, and an attempt was then made to determine empirically which industries best satisfied these criteria. The empirical results indicated that, although almost all of the 20 industries displayed the familiar trade-off between price changes and employment changes, the combination of price and employment responses to demand changes was significantly more favorable in some industries than in others.68 In other words, the evidence suggested that industries differed in both their price and employment responses to demand changes. In 14 industries, demand changes were found to have a significant influence on price changes. In the remaining 6 industries, no such influence of demand changes on price changes could be detected with the available measures of demand changes. The industry employment equations likewise illustrated interindustry differences. Although the estimated elasticity of employment with respect to output was less than one in all industries, this elasticity ranged from about 0.2 in printing and publishing to about 0.8 in lumber. Therefore, it seems that there is scope for altering the composition of aggregate demand for stabilization purposes. Not only would such a policy help to reduce the inequality or dispersion of excess demand in product and labor markets among different sectors but it would also take better advantage of the uneven effectiveness of demand-reducing policy across sectors and industries.

It should also be recognized, however, that a selective approach to anti-inflationary policy might involve significant costs. In particular, the information, administrative, and resource allocation costs of such a policy would probably be higher than those for a more general policy. Important equity questions are also involved in subjecting different industries to different types of inflationary controls. Further, although a wide variety of selective policy instruments exist—such as government expenditure changes, sales from government stockpiles, tax credits for plant and equipment expenditures, accelerated depreciation allowances, limited sales taxes, selective credit provisions (for consumer durables, housing, agriculture), tariff and import quotas, wage-price guideposts, retail price maintenance laws, and various manpower policies—the flexibility of some of these instruments may be too low for countercyclical purposes.69 In short, the expected costs of a more selective anti-inflationary policy would have to be carefully weighed against the expected benefits.

Finally, it must be emphasized that this paper represents only a first step toward evaluating some of the potential benefits of exercising greater control over the composition of aggregate demand. Many important questions relating to the implementation of a selective anti-inflationary policy have been left unanswered, and the techniques employed for identifying the best target industries are admittedly too imprecise for direct policy application. Additional industry price, wage, and employment studies using better data and more dynamic simultaneous models of price, production, and employment behavior are needed to help remedy these deficiencies. However, even on the basis of the preliminary results reported in this paper, it is clear that future proposals to reconcile more closely the twin objectives of high employment and price stability should pay greater attention to the properties and characteristics of the different sectors and industries in the economy. This, in turn, implies that policymakers will have to seriously consider supplementing traditional aggregate stabilization policies with selective policy instruments which better take into account interindustry differences in price, wage, and employment behavior.

APPENDICES

I. Definition of Variables

The variables in the price change equations

Prices (P): Implicit price deflators of the total output of an industry taken from the annual data of the Office of Business Economics (OBE) are the dependent variables. These prices are deflators of the value of shipments, adjusted for changes in inventories of finished goods and goods in progress; the deflators are essentially a wholesale price index for the industry concerned. These price indexes were selected rather than deflators of gross product originating (value-added deflators), because little success has been achieved in explaining value-added deflators in previous industry price studies [23, 46]. Since total output deflators are list prices, they are likely to be less flexible and less responsive to demand changes than are transactions prices.70 Therefore, the estimates of the influence of demand changes on price changes are likely to have a downward bias. Data for 1953-64 are found in Gottsegen and Ziemer [17, pp. 339-47]; data for 1965-67 were furnished to the author by OBE.

Cost of materials (CM): Cost of materials indexes represent annual OBE price deflators of the cost of materials for an industry. More specifically, they are the U.S. Bureau of the Census cost of materials, supplies, components, semifinished goods, fuel, and electricity actually consumed or put into production during the year and the cost of goods purchased for resale. Since these indexes include semifinished goods which may also appear in the output price indexes, it is to be expected that the coefficients attaching to cost of materials will have some upward bias. Data for 1953-64 were found in Gottsegen and Ziemer [17, pp. 339–47], and data for 1965-67 were furnished to the author by OBE.

Standard cost of materials (CMN): Standard cost of materials is a three-year weighted moving average of actual cost of materials for an industry. The weights were chosen, not estimated, as 0.50 for current year (t), 0.25 for year t-1, and 0.25 for year t-2. The purpose of this variable, as well as of the other standard variables, was to obtain a more permanent measure of cost changes. Since the standard variables change less markedly than do their current counterparts, they call for less frequent price changes—a desirable development where the costs of changing prices are high.

Compensation per man-hour (CMH): Annual compensation per man-hour for each industry was derived by dividing total employee compensation for the industry by estimated total man-hours worked for the industry. Annual total man-hours worked were estimated by multiplying the number of employees by normal weekly hours of production workers, and then multiplying this total by 52. Total employee compensation was used rather than total wages, because it is a more comprehensive measure of labor costs. Data for total employee compensation for 1953–67 are from the U.S. Bureau of the Census [55, pp. 91–93; 56, p. 41]. Data for all employees and normal weekly hours of production workers are from the U.S. Bureau of Labor Statistics [57].

Real output (RQ): Annual real output for an industry is the OBE estimate of real value added for the industry, i.e., industry value added in constant (1958) prices. An industry’s real value added is its contribution to real GNP. Real value added was used rather than real value of production (real total output), because the basic decision-making framework in the area of factor costs would appear to be most consistent with a concept of real output net of purchases from other industries. Data for 1953–64 are from Gottsegen and Ziemer [17], while data for 1965–67 were furnished by OBE.

Standard real output (RQN): Annual standard real output is a three-year weighted moving average of actual real output for the industry. The weighting pattern is identical to that for the other standard variables. This variable was used as a proxy for expected demand in the price equations.

Actual unit labor cost (ULC): Annual actual unit labor cost was derived by dividing total compensation of employees for the industry by real output for the industry.

Standard unit labor cost (ULCN): Annual standard unit labor cost was derived by dividing total compensation of employees for the industry (in year t) by a three-year weighted moving average of real output for the industry. The weights chosen were identical to those selected for other standard variables (0.50 for year t, 0.25 for year t-1, and 0.25 for year t-2). The purpose of this variable was to correct actual unit labor cost for short-run swings in labor productivity and thereby to obtain a measure of more permanent unit labor costs.

Real output per man-hour (RQ/MH): Annual real output per man-hour for the industry was calculated by dividing real output for the industry by estimated total man-hours worked for the industry.

Standard real output per man-hour (RQ/MHN): Annual standard real output per man-hour is a three-year weighted moving average of actual real output per man-hour for the industry. The weighting pattern is identical to that used for the other standard variables.

Capacity utilization rate (CUW): Annual industry capacity utilization rates were computed as annual averages of the quarterly Wharton School utilization rates for each industry. Data for years 1953–65 are from Klein and Summers [27, pp. 52-73]. The data for 1966-67 were furnished by the Economics Research Unit at the University of Pennsylvania.

Unfilled orders to shipments ratio (UO/S): Annual unfilled orders to shipments ratios for each industry were calculated as annual averages of the monthly unfilled orders to shipments ratios (seasonally unadjusted, in current dollars). Data are from the U.S. Bureau of the Census: for 1953–62 [53]; for 1963-67 [54].

Inventories to shipments ratio (INV/S): Annual inventories to shipments ratios for each industry were calculated as annual averages of the monthly inventories to shipments ratios (seasonally unadjusted, in current dollars). Data are from the U.S. Bureau of the Census: for 1953–62 [53]; for 1963-67 [54].

Variables in the employment equations

Total employees (E): U.S. Bureau of Labor Statistics annual estimates of all employees (in thousands) in an industry are the dependent variables. Data are from the U.S. Bureau of Labor Statistics [57].

Real output (RQ) and Compensation per man-hour (CMH): These are defined exactly as in the industry price change equations. However, for notational convenience, compensation per man-hour was labeled the wage (W) in the employment equations.

Rental price of capital (r): Annual estimates of the rental price of capital are implicit price deflators of producers’ durable equipment from The National Income and Product Accounts of the United States [55]. Rather naïvely, the rental price of capital was assumed to be identical in all industries. Data are from the U.S. Bureau of the Census: for 1954–65 [55, p. 160]; for 1966-67 [56, p. 14].

Time (T): Time was introduced into the employment equations as a simple proxy for technological change. Technological change is assumed to occur at a linear rate in all industries, so that T takes the value of 1 for 1954, 2 for 1955, etc.

II. The Costs of Changing Employment

In this paper, it has been argued, following Oi [38], that the higher are the costs of changing employment to the industry, the lower will be the industry’s employment response to output changes, i.e., the lower will be the industry’s elasticity of employment with respect to output and the slower will be the industry’s speed of adjustment of actual to desired employment. This hypothesis suggests that the costs of changing employment should be higher for the low employment sensitivity group of industries than for the high employment sensitivity group.

Since there is no single measure which encompasses all the costs of changing employment, a number of proxies are used for these costs. It is assumed that the costs of changing employment vary directly with the rental charges of employment, the fixed costs of changing employment, the skill level of the industry’s work force, and the institutional barriers to changing employment. The average level of compensation per man-hour for the industry in the 1954-67 period is used as an approximation for the rental charges of employment. The fixed costs of changing employment are primarily hiring and training costs. The proxy for these costs is Eckaus’ [9] estimates of the specific vocational preparation required for an average performance of the jobs in each industry for 1950. The two skill-level proxies are the average (1954—67) proportion of nonproduction workers to all workers in an industry and Eckaus’ estimates of the general educational development required for an average performance of the jobs in each industry for 1950. Finally, the institutional barriers proxy is Douty’s [6] estimates of the extent of unionization in each industry for 1958. The a priori expectation is that compensation per man-hour, specific vocational preparation, general educational development, the ratio of nonproduction workers to all workers, and the extent of unionization will all be higher for the low employment sensitivity group than for the high employment sensitivity group.

The two industry employment groups, with the proxies for the costs of changing employment, are shown in Table 5.71 In accord with the adjustment costs hypothesis, the low employment sensitivity group does display higher costs of changing employment for each of the five cost proxies. A t test was used to determine whether the group means for each of the five cost proxies were statistically different. The null hypothesis that the group means were equal was rejected at the 98 per cent probability level for compensation per man-hour, at the 95 per cent level for the nonproduction-worker proxy for skill levels, at the 80 per cent level for both the specific vocational preparation and the general educational development proxies, and at the 75 per cent level for the unionization variable.72 Since interindustry differences in the elasticity of employment with respect to output are due to many factors, only one of which is differences in the costs of changing employment, these results provide strong support for the adjustment costs argument.73 In addition, these results support the argument that industries with relatively highly skilled labor will, ceteris paribus, be relatively good targets for selective demand reductions.

Table 5.

Costs of Changing Employment for Two Industry Employment Groups

article image

Total employee compensation is taken from U.S. Department of Commerce, Office of Business Economics, The National Income and Product Accounts of the United States, 1929-65, pp. 91-93, and Survey of Current Business, Vol. XLVIII (July 1968), p. 41. The data for total man-hours worked are from U.S. Department of Labor, Employment and Earnings Statistics for the United States, 1909-68, pp. 48-716.

R. S. Eckaus, “Economic Criteria for Education,” The Review of Economics and Statistics, Vol. XLVI (May 1964), pp. 188-89.

U.S. Department of Labor, ibid.

Eckaus, ibid.

H. M. Douty, “Collective Bargaining Coverage in Factory Employment, 1958,” Monthly Labor Review, Vol. LXXXIII (April 1960), p. 347.

Gestion anti-inflationniste de la demande aux Etats-Unis : une approche sélective par industrie

Résumé

Dans cette étude, l’auteur examine s’il serait possible, en appliquant une politique anti-inflationniste fondée sur une approche sélective par industrie, d’atténuer le problème que pose, en courte période, aux entreprises américaines, l’alternative entre la stabilité des prix et l’emploi. Cette méthode permet essentiellement aux responsables de la politique anti-inflationniste d’exercer un certain degré de contrôle sur la composition de la demande globale autant que sur son volume. Elle est fondée sur l’hypothèse qu’il existe des différences très notables dans la façon dont les diverses industries réagissent aux variations de la demande, notamment en ce qui concerne leurs politiques en matière de prix et d’emploi. Etant donné ces différences, il semblerait par conséquent que les mesures destinées à réduire la demande visent principalement les industries dont la réaction combinée en matière de prix et d’emploi est la plus favorable. En ce qui concerne les industries où celle-ci est relativement défavorable, on pourrait recourir davantage à des mesures anti-inflationnistes qui ne visent pas principalement à modifier la demande, comme l’établissement de directives en matière de prix et salaires et la libéralisation des importations.

A la lumière de l’hypothèse ci-dessus mentionnée, l’auteur s’est attaché à identifier, parmi les industries manufacturières américaines entrant dans la classification type à deux chiffres, celles qui se prêteraient relativement bien à une compression sélective de la demande. Il définit les trois caractéristiques que devraient présenter les industries qui, par hypothèse, constitueraient des objectifs appropriés, et essaie ensuite de déterminer empiriquement celles qui répondent le mieux à ces caractéristiques. Les deux caractéristiques les plus importantes sont une réaction aux variations de la demande sous forme de variations relativement fortes des prix et non pas de l’emploi.

Pour vérifier ces deux réactions aux variations de la demande, l’auteur a calculé, pour chaque industrie, en se basant sur des données annuelles pour la période 1954-67, la corrélation, d’une part, entre les variations de la demande et les variations des prix de chaque industrie et, d’autre part, entre les variations de la demande et celles des offres d’emploi dans chaque industrie. La première corrélation indique que les variations de la demande ont une assez forte répercussion sur celles des prix dans 14 des 20 industries. Dans les 6 autres industries, on n’a constaté aucune répercussion directe des variations de la demande sur les variations de prix, sur la base des données utilisées pour mesurer les variations de la demande. De même, la corrélation entre variations de la demande et variations des offres d’emploi indique que l’amplitude de l’élasticité de l’emploi par rapport à la production varie considérablement d’une industrie à l’autre. Bref, bien que presque toutes les industries aient eu à faire le choix bien connu entre une modification des prix et une modification de l’emploi, on a constaté que la réaction combinée vis-à-vis des prix et de l’emploi se présenterait de façon plus favorable dans certaines industries que dans d’autres. Une des principales conclusions qui s’impose en matière de politique, c’est que les responsables de la politique économique devraient sérieusement envisager de compléter les mesures classiques générales de stabilisation par des mesures sélectives qui tiennent davantage compte des différences dans le comportement des diverses industries en ce qui concerne les prix, les salaires et l’emploi.

La política antiinflacionista y la demanda en los Estados Unidos: el sistema selectivo de industrias

Resumen

En este estudio se analiza un sistema selectivo de industrias para la política antiinflacionista, como posible método para mejorar la contraposición a corto plazo entre la estabilidad de precios y el desempleo en los Estados Unidos. En esencia, el sistema consiste en que al formular la política antiinflacionista se ejerza un cierto control sobre la composición, así como sobre el nivel, de la demanda agregada. El sistema se basa en el supuesto fundamental de que hay considerables diferencias de reacción de precios y empleo ante las variaciones de la demanda entre las industrias. A su vez, estas diferencias interindustriales sugieren que la política de reducción de la demanda se oriente primordialmente hacia las que presentan la combinación más favorable de reacción de precios y empleo frente a las variaciones de la demanda. Para las industrias en que esta combinación de reacción de precios y empleo es relativamente desfavorable, podría recurrirse en mayor grado a tipos de política antiinflacionista que no supongan primordialmente variaciones de la demanda, por ejemplo, las directrices de salarios y precios y la liberación de las importaciones.

Con esta hipótesis, se trata de identificar las industrias manufactureras de los EE.UU., comprendidas en la categoría de dos dígitos de la clasificación industrial uniforme (CIU), que podrían resultar relativamente adecuadas como objetivos para reducciones selectivas de la demanda. Se sugieren tres características de este tipo de industrias y, a continuación, se trata de determinar empíricamente las que mejor se ajustan a ellas. Las dos características más importantes son: una reacción de precios relativamente fuerte y una reacción de empleo relativamente débil ante las variaciones de la demanda.

A fin de verificar la reacción de precios y de empleo ante las variaciones de la demanda, se calcularon regresiones de la variación de precios y de empleo con respecto a la demanda para cada industria, a partir de datos anuales para el período 1954-67. Las regresiones de las variaciones de precio indican que las variaciones de la demanda tienen un efecto significativo en las variaciones de precios en 14 de las 20 industrias. En las 6 industrias restantes no pudo encontrarse un efecto directo de las variaciones de la demanda en las variaciones de los precios mediante los datos disponibles que se utilizaron para representar las variaciones de la demanda. Análogamente, las regresiones del empleo ponen de manifiesto que la elasticidad del empleo con respecto a la producción varía considerablemente entre las industrias. En resumen, aunque casi todas las industrias presentan la habitual contraposición ante las variaciones de los precios y las variaciones del empleo, en algunas la combinación de la reacción de precios y la reacción de empleo resulta más favorable que en otras. Una de las principales consecuencias de política que se deduce de este estudio consiste en que al formular la política puede considerarse con fundamento la posibilidad de complementar las políticas tradicionales de estabilización global con instrumentos de política selectivos que tengan más en cuenta las diferencias interindustriales en lo que se refiere al comportamiento de los precios, los salarios y el empleo.

BIBLIOGRAPHY

  • Adelman, M. A., Steel, Administered Prices and Inflation,” The Quarterly Journal of Economics, Vol. LXXV (1961), pp. 1640.

  • Archibald, G. C., The Phillips Curve and the Distribution of Unemployment,” American Economic Association Papers and Proceedings of the Eighty-first Annual Meeting, The American Economic Review, Vol. LIX (May 1969), pp. 12434.

    • Search Google Scholar
    • Export Citation
  • Black, S. W., and H. H. Kalejian, A Macro Model of the U.S. Labor Market,” Econometrica, Vol. 38 (1970), pp. 71241.

  • Brechling, F. P. R., The Relationship between Output and Employment in British Manufacturing Industries,” The Review of Economic Studies, Vol. XXXII (1965), pp. 187216.

    • Search Google Scholar
    • Export Citation
  • Darling, Paul G., and Michael C. Lovell, Factors Influencing Investment in Inventories,” The Brookings Quarterly Econometric Model of the United States, ed. by James S. Duesenberry, Gary Fromm, Lawrence R. Klein, and Edwin Kuh (Chicago, 1965), pp. 13162.

    • Search Google Scholar
    • Export Citation
  • Douty, H. M., Collective Bargaining Coverage in Factory Employment, 1958,” Monthly Labor Review, Vol. 83 (April 1960), pp. 34549.

    • Search Google Scholar
    • Export Citation
  • Duesenberry, James S., The Mechanics of Inflation,” The Review of Economics and Statistics, Vol. XXXII (1950), pp. 14449.

  • Dunlop, John, Guideposts, Wages and Collective Bargaining,” Guidelines, Informal Controls, and the Market Place, ed. by George P. Shultz and Robert Z. Aliber (University of Chicago Press, 1966), pp. 8196.

    • Search Google Scholar
    • Export Citation
  • Eckaus, R. S., Economic Criteria for Education,” The Review of Economics and Statistics, Vol. XLVI (May 1964), pp. 18190.

  • Eckstein, Otto, A Theory of the Wage-Price Process in Modern Industry,” The Review of Economic Studies, Vol. XXXI (1964), pp. 26786.

    • Search Google Scholar
    • Export Citation
  • Eckstein, Otto, and Gary Fromm, Steel and the Postwar Inflation,” Study Paper No. 2, U.S. Congress, Joint Economic Committee, Study of Employment, Growth and Price Levels (Washington, 1959).

    • Search Google Scholar
    • Export Citation
  • Eckstein, Otto, and Gary Fromm, The Price Equation,” The American Economic Review, Vol. LVIII (December 1968), pp. 115983.

  • Eckstein, Otto, and Thomas Wilson, The Determination of Money Wages in American Industry,” The Quarterly Journal of Economics, Vol. LXXVI (1962), pp. 379414.

    • Search Google Scholar
    • Export Citation
  • Eckstein, Otto, and David Wyss, Industry Price Equations,” Harvard Institute of Economic Research, Discussion Paper No. 158 (January 1971).

    • Search Google Scholar
    • Export Citation
  • Ferber, Robert, and P. J. Verdoorn, Research Methods in Economics and Business (New York, 1962).

  • Galbraith, John K., Market Structure and Stabilization Policy,” The Review of Economics and Statistics, Vol. XXXIX (1957), pp. 12433.

    • Search Google Scholar
    • Export Citation
  • Gottsegen, Jack, and Richard Ziemer, Comparison of Federal Reserve and OBE Measures of Real Manufacturing Output,” in The Industrial Composition of Income and Product, ed. by J. Kendrick, National Bureau of Economic Research (Columbia University Press, 1968), pp. 225346.

    • Search Google Scholar
    • Export Citation
  • Griliches, Zvi, A Note on Social Correlation Bias in Estimates of Distributed Lags,” Econometrica, Vol. 29 (1961), pp. 6573.

  • Hall, R. L., and C. J. Hitch, Price Theory and Business Behavior,” Oxford Economic Papers (May 1939), pp. 1245.

  • Hamermesh, D., Spectral Analysis of the Relation between Gross Employment Changes and Output Changes,” The Review of Economics and Statistics, Vol. L (1968), pp. 6269.

    • Search Google Scholar
    • Export Citation
  • Hay, George A., Production, Price and Inventory Theory,” The American Economic Review, Vol. LX (September 1970), pp. 53145.

  • Holt, Charles C., C. Duncan MacRae, Stuart O. Schweitzer, and Ralph E. Smith, The Unemployment-Inflation Dilemma: A Manpower Solution (The Urban Institute, Washington, 1971).

    • Search Google Scholar
    • Export Citation
  • Houthakker, H., The Statistical Foundations of an Incomes Policy,” American Statistical Association, Proceedings of the Business and Economic Statistics Section, 1968 (Washington, 1968), pp. 13035.

    • Search Google Scholar
    • Export Citation
  • Johnston, John, Econometric Methods (New York, 1963).

  • Kaplan, A., Joel B. Dirlam, and Robert F. Lanzilotti, Pricing in Big Business: A Case Approach (The Brookings Institution, Washington, 1958).

    • Search Google Scholar
    • Export Citation
  • Klein, Lawrence R., and Michael K. Evans, The Wharton Econometric Forecasting Model (University of Pennsylvania Press, 1967).

  • Klein, Lawrence R., and Robert Summers, The Wharton Index of Capacity Utilization (University of Pennsylvania Press, 1966), p. 94.

  • Lee, K. S., and C. Holt, Employment Decisions of Manufacturing Firms in Response to Demand Fluctuations: Theory and Empirical Results” (Social System Research Institute, University of Wisconsin, 1969).

    • Search Google Scholar
    • Export Citation
  • Lipsey, Richard G., The Relationship between Unemployment and the Rate of Change of Money Wage Rates in the U.K., 1862-1957,” Economica, New Series, Vol. XXVII (1960), pp. 131.

    • Search Google Scholar
    • Export Citation
  • McGuire, Timothy W., and Leonard A. Rapping, The Role of Market Variables and Key Bargains in the Manufacturing Wage Determination Process,” The Journal of Political Economy, Vol. 76 (September/October 1968), pp. 101536.

    • Search Google Scholar
    • Export Citation
  • Means, Gardiner, Administered Inflation and Public Policy” (Kramer Associates, Washington, 1959).

  • Mills, Edwin S., Price, Output and Inventory Policy (New York, 1962).

  • Mortensen, Dale T., Job Search, the Duration of Unemployment and the Phillips Curve,” The American Economic Review, Vol. LX (December 1970), pp. 84762.

    • Search Google Scholar
    • Export Citation
  • Mundell, Robert A., The Monetary Dynamics of International Adjustment under Fixed and Flexible Exchange Rates,” The Quarterly Journal of Economics, Vol. LXXIV (1960), pp. 22757.

    • Search Google Scholar
    • Export Citation
  • Nadiri, M. I., The Effects of Relative Prices and Capacity on the Demand for Labour in the U.S. Manufacturing Sector,” The Review of Economic Studies, Vol. XXXV (July 1968), pp. 27388.

    • Search Google Scholar
    • Export Citation
  • Nadiri, M. I., and Sherwin Rosen, Interrelated Factor Demand Functions,” The American Economic Review, Vol. LIX (September 1969), pp. 45771.

    • Search Google Scholar
    • Export Citation
  • Nerlove, Marc, and Kenneth F. Wallis, Use of the Durbin-Watson Statistic in Inappropriate Situations,” Econometrica, Vol. 34 (1966), pp. 23538.

    • Search Google Scholar
    • Export Citation
  • Oi, Walter, Labor as a Quasi-Fixed Factor,” The Journal of Political Economy, Vol. 70 (December 1962), pp. 53855.

  • Perry, George, The Determinants of Wage Rate Changes and the Inflation-Unemployment Trade-off for the United States,” The Review of Economic Studies, Vol. XXXI (1964), pp. 287308.

    • Search Google Scholar
    • Export Citation
  • Perry, George, Wages and the Guideposts,” The American Economic Review, Vol. LIX (June 1969), pp. 36569.

  • Perry, George, Changing Labor Markets and Inflation,” in Brookings Papers on Economic Activity 3, ed. by A. Okun and G. Perry (Washington, 1970), pp. 41141.

    • Search Google Scholar
    • Export Citation
  • Phillips, A. W., The Relationship between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861-1957,” Economica, New Series Vol. XXV (1958), pp. 28399.

    • Search Google Scholar
    • Export Citation
  • Pierson, Gail, The Effect of Union Strength on the U.S. ‘Phillips Curve’,” The American Economic Review, Vol. LVIII (June 1968), pp. 45667.

    • Search Google Scholar
    • Export Citation
  • Salant, Walter S., and Beatrice N. Vaccara, Import Liberalization and Employment (The Brookings Institution, Washington, 1961).

  • Schultze, C. L., Recent Inflation in the United States,” Study Paper No. 1, U.S. Congress, Joint Econmic Committee, Study of Employment, Growth and Price Levels (Washington, 1959).

    • Search Google Scholar
    • Export Citation
  • Schultze, Charles L., and Joseph L. Tryon, Prices and Wages,” in The Brookings Quarterly Econometric Model of the United States, ed. by James S. Duesenberry, Gary Fromm, Lawrence R. Klein, and Edwin Kuh (Chicago, 1955), pp. 281334.

    • Search Google Scholar
    • Export Citation
  • Stigler, George, and James Kindahl, The Behavior of Industrial Prices, National Bureau of Economic Research (Columbia University Press, 1970), p. 202.

    • Search Google Scholar
    • Export Citation
  • Waud, Roger N., Man-hour Behavior in U.S. Manufacturing: A Neoclassical Interpretation,” The Journal of Political Economy, Vol. 76 (May/June 1968), pp. 40727.

    • Search Google Scholar
    • Export Citation
  • Wilson, George W., The Relationship between Output and Employment,” The Review of Economics and Statistics, Vol. XLII (1960), pp. 3743.

    • Search Google Scholar
    • Export Citation
  • Yordon, Wesley J.Industrial Concentration and Price Flexibility in Inflation: Price Response Rates in Fourteen Industries, 1947-1958,” The Review of Economics and Statistics, Vol. XLIII (1961), pp. 28794.

    • Search Google Scholar
    • Export Citation
  • Zarnowitz, Victor, Unfilled Orders, Price Changes and Business Fluctuations,” The Review of Economics and Statistics, Vol. XLIV (1962), pp. 36794.

    • Search Google Scholar
    • Export Citation
  • U.S. Department of Commerce, Bureau of the Census, 1967 Census of Manufactures, Vol. I, Summary and Subject Statistics (Washington, 1971).

    • Search Google Scholar
    • Export Citation
  • U.S. Department of Commerce, Bureau of the Census, Manufacturers’ Shipments, Inventories and Orders, 1947-63 Revised, Series M3-1 (Washington, 1964).

    • Search Google Scholar
    • Export Citation
  • U.S. Department of Commerce, Bureau of the Census, Manufacturers’ Shipments, Inventories and Orders, 1961-68, Series M3-1.1 (Washington, 1968).

    • Search Google Scholar
    • Export Citation
  • U.S. Department of Commerce, Bureau of the Census, Office of Business Economics, The National Income and Product Accounts of the United States, 1929-1965 (Washington, 1966).

    • Search Google Scholar
    • Export Citation
  • U.S. Department of Commerce, Bureau of the Census, Survey of Current Business, Vol. 48 (July 1968).

  • U.S. Department of Labor, Bureau of Labor Statistics, Employment and Earnings Statistics for the United States, 1909-1968 (Washington, 1968).

    • Search Google Scholar
    • Export Citation
  • Economic Report of the President (Washington, 1971).

*

Mr. Goldstein, economist in the Special Studies Division of the Research Department, is a graduate of Rutgers University and received his doctorate in economics from New York University. Before coming to the Fund, he was a Research Fellow in Economics at the Brookings Institution.

1

The author wishes to acknowledge his indebtedness to staff colleagues, to Hourmouzis Georgiadis, Peter Albin, and M. I. Nadiri of New York University, and to Charles Holt of the Urban Institute and Otto Eckstein of Harvard University for their helpful comments and suggestions on earlier drafts of this paper. The author emphasizes that he remains solely responsible for the present text.

2

The numbers in brackets refer to items listed in the bibliography, which appears at the end of this paper.

3

Throughout this paper, “relative” is interpreted to mean “relative to other industries.”

4

Mundell [34, p. 250] has called this solution to the assignment problem the “principle of effective market classification,” which asserts that an instrument should be assigned to the target on which it has the greatest direct influence.

5

The price index to be reduced could be the gross national product (GNP) deflator, the wholesale price index, or the consumer price index. The wholesale price index would, however, be most consistent with the empirical tests carried out in Section III of this paper.

6

The argument is analogous to asserting that different levels of consumption may be obtained at a given level of disposable income by redistributing this income level among groups with different marginal propensities to consume.

7

Archibald [2] has in fact shown that the rate of change of money wage rates will be higher in the United States and in the United Kingdom at a given rate of unemployment, the greater is the dispersion or inequality of unemployment among different regions or different industries. Schultze [45] and Lipsey [29] also have argued that if the relationship between price (or wage) changes and demand changes (or the unemployment rate) is nonlinear for all sectors, then a change in the composition of demand (or unemployment) at a given level of demand (or unemployment) can lead to a change in the aggregate rate of price (or wage) change. More recently, Perry [41] has shown that changes in the age-sex composition of unemployment, at any given unemployment rate, can affect the trade-off between wage inflation and unemployment.

8

Dunlop [8, pp. 92-96] has recently proposed a selective type of anti-inflationary policy which might better cope with inflations of the bottleneck or problem-sector variety. In his policy, the major activities of policymakers would include: (1) the identification of the major priority bottlenecks in the economy, and (2) the development of private and public policies in bottleneck sectors to reduce inflationary pressures by increasing supplies and reducing demands.

9

If demand changes are large enough and permanent enough, they will ultimately affect prices in all industries. However, the demand changes induced by stabilization policies will often be only moderate in size and temporary in duration and direction.

10

Other considerations which could be important in choosing the target industries are the industry’s recent rate of price increase, the industry’s weight in the aggregate price index, and the ease (cost) with which demand can be reduced in that industry with available policy instruments.

11

This assumes that the second-order conditions are satisfied, i.e., that the marginal cost curve intersects the marginal revenue curve from below.

12

Presumably, the output adjustment to a given demand change will be greater the flatter is the slope of the marginal cost curve.

13

While no industry satisfies all the assumptions of the pure competition model, Means [31, p. 8] and Galbraith [16, pp. 125-26] argue that this model approximates price behavior in some branches of agriculture, bituminous coal mining, lumber, hides and leather products, and cotton textiles.

14

This uncertainty is sometimes expressed by the kinked demand curve, where the kink represents the collective equilibrium price.

15

In industries where price leadership or other forms of collusion exist, this cost of changing prices will be low; however, there may still be other significant costs of changing prices, such as government and public censure.

16

If two producers of intermediate goods display the same average rate of price change over a given time period but the variance of the average price change differs, then the producer with the more stable price record can be at a competitive advantage.

17

Galbraith [16, pp. 127-28] has argued that if the firm does attempt to maximize short-run profits by making frequent price adjustments, it faces the dangers of (1) government and public censure, (2) increased entry of new firms in response to the higher profit rate, and (3) more aggressive wage demands by unions in response to the firm’s increased ability to pay.

18

For two good analyses of the costs associated with changing production, inventories, and order backlogs, see Lee and Holt [28, pp. 14-19] and Hay [21, pp. 531-45].

19
Following Eckstein and Fromm [12, p. 1166] and using their notation, target rate of return pricing may be expressed as:
P=πKXN+ULCN+UMCN

Where π is the target rate of return, K is the firm’s capital stock, XN is the standard output, ULCN is standard unit labor cost (the current wage divided by a weighted moving average of current and past output per man-hour), and UMCN is standard unit material cost (a weighted moving average of current and past unit material cost). The firm’s decision parameters are the definition of the standard rate of output and the target rate of return. Similarly, full cost pricing can be expressed as P = (1 + λ) (ULCN + UMCN) where λ is the markup factor. Since standard costs change less markedly and less frequently than do actual costs, these formulas call for relatively few price changes—i.e., prices respond only to those cost changes deemed “permanent” by producers. Also, it should be noted that these formulas relate price to average rather than to marginal costs.

20

Although no explicit account is taken of demand changes in target rate of return or full cost pricing, demand factors may intrude into the pricing decision in various ways. For example, prices can be set as a markup over standard costs, but the markup may be reduced in periods of weak demand and increased in periods of excess demand. In fact, Schultze and Tryon [46] have proposed a pricing model which incorporates this assumption. Also, in periods of weak demand, there may be some retreat from list prices (see Stigler and Kindahl [47]). Finally, demand factors may influence the terms of sale (other than price), such as transportation charges, waiting times, product quality, or the existence of extras.

21

Nadiri and Rosen [36, p. 464] found that the typical response pattern to a change in output in U.S. manufacturing was for the firm to adjust first the utilization rate of capital, then the utilization rate of labor, then the stock of labor, and finally the stock of capital.

22

In a study for Canadian manufacturing industries, Wilson [49, pp. 37-43] found that the relationship between employment changes and output changes became closer when one moved from monthly to quarterly to annual data (and from lower to higher levels of aggregation). Wilson’s results can be interpreted as lending support to a concept of permanent employment. In fact, the concept of “permanent” versus “transitory” variables, so fruitfully employed in consumption and investment theory, would seem to have wide applicability to both price and employment theory.

23

The user cost of employment includes hiring, training, and related costs discounted at the relevant rate of interest. The price of hours per man is the wage rate plus any overtime pay. The user cost of capital includes not only the price of capital goods but also the cost of capital and depreciation charges. The price of services per unit of capital is the depreciation rate. See Nadiri and Rosen [36, p. 460].

24

The assumption that a job applicant’s skill level is an important factor in determining the duration of his successful job search is consistent with recent microeconomic research on the Phillips curve. See Mortensen [33, pp. 847-62]. However, if a worker’s skills are highly specialized, his job mobility may be low.

25

If the coverage of desired price decreases is very wide, a general rather than a selective policy must be used. However, even in this case, there can be some merit in selectively assigning anti-inflationary policy instruments to the proper industry groups. A selective policy can therefore refer to either (1) a policy where only a limited number of industries are selected for policy action, as in “bottleneck” or “problem-sector” inflations, or (2) a policy where almost all industries are selected for policy action, but where different policy instruments are used to obtain price decreases in the different industry groups or sectors. Most of the emphasis in this paper is on the second interpretation of selective policy—i.e., on a balanced anti-inflationary policy where policy instruments are assigned according to their comparative advantage.

26

On the other hand, it should be recognized that price decreases in final goods industries may be more significant than price decreases in intermediate goods industries, because the latter may not be passed on to consumers.

27

Klein and Evans [26, pp. 11, 36] also found that wage changes in U.S. manufacturing had a significant spillover effect on wage changes in the nonmanufacturing sector. Pierson [43, p. 459] found that the wage changes in highly unionized U.S. manufacturing industries had a significant influence on wage changes in less unionized industries.

28

This assumes that selective demand reductions can induce wage moderation in the key-wage-bargain industries. It may well be, however, that it would be less costly to reduce the rate of wage increase in these industries by using a non-demand-changing policy, such as wage-price guideposts.

29

The 20 manufacturing industries included 19 two-digit Standard Industrial Classification industries (SIC 20-36, 38, 39) and 1 three-digit SIC industry (SIC 371). Since the two-digit SIC industry groupings are quite broad and cover a rather heterogeneous group of products, the industry price analyses may be subject to aggregation problems. However, if lower levels of aggregation are sought to obtain more homogeneous industry groups, the data and computation requirements increase rapidly; for example, at the three-digit SIC industry level, there are about 200 manufacturing industries. A list of the industries which comprise each two-digit industry can be found in the 1967 Census of Manufactures [52, pp. 28-43].

30

The use of annual data did, however, offer some compensation for the purposes of this study. In particular, annual data may give a more revealing picture than quarterly data of the influence of demand changes on price changes. If no influence of demand changes on price changes is found with quarterly data, it is difficult to know whether demand changes are not significant or whether quarterly changes in demand are too short and too erratic to lead to price changes. Annual demand changes should have a more permanent nature.

31

Schultze and Tryon [46, pp. 288-311] and Eckstein and Fromm [12, pp. 1171-76] both found that prices in U.S. manufacturing were best explained by equations which contained the two factors of cost and excess demand.

32

Standard unit labor cost, however, did not exactly follow this method of construction (see Appendix I).

33
The relationship of price changes to changes in variable costs can be simply illustrated. Following Eckstein and Fromm [12, p. 1162] and using their notation, assume the firm maximizes its short-run gross returns to capital (Z).
Z=PXWL(X)PMMX(1)
where P is price, X is output, W is the wage rate, L(X) is total manhours, PM is material prices, and M is material inputs per unit of output. For profit maximization, let us differentiate equation (1) with respect to X and set this expression equal to zero. This gives:
P=W(dL/dX)+PMM(2)
Omitting the second-order interaction terms, an approximation for price changes is then:
ΔP=WΔ(L/X)+(L/X)ΔW+PMΔM+MΔPM(3)

The change in price is then the sum of changes in unit labor cost—the first two terms in equation (3)—and changes in unit materials cost—the last two terms inequation (3).

34

Equations (la) and (lb) are identical to the industry price change regressions estimated by Houthakker [23] for these industries for the 1947-67 period.

35

The price studies of Schultze and Tryon [46, pp. 290-91] and Eckstein and Fromm [12, pp. 1171-76], both based on quarterly data, found that changes in standard unit labor costs have a greater influence on price changes than do changes in actual unit labor costs.

36

The demand-change proxies are intended to serve as approximations for the state of excess demand in the industry. In the classic competitive market, the rate of price change (ΔPP) is a function of the level of excess demand (dXS), where d, X, and S are short-run demand, short-run supply, and shipments, respectively. Inventory changes and unfilled order changes are disequilibrium factors which are used to measure d − X; i.e., when demand is greater than supply, inventories will be drawn down and unfilled orders will increase. The capacity utilization rate is also a proxy for excess demand, since it reflects the relationship between the industry’s rate of actual production and its rate of full capacity production.

37

In order to identify reliably those industries in which demand changes affect price changes through changes in costs, it would be necessary to estimate a separate equation for each of the cost variables (cost of materials, wages, and productivity) in each industry. In the cost equation, it would be desirable to include not only excess demand conditions in that industry but also supply and demand conditions in related industries and in the economy as a whole. For example, the change in the cost of materials in the motor vehicle industry depends not only on changes in demand in that industry but also on demand changes in other industries that purchase the same materials. The estimation of such cost equations was considered to be beyond the scope of the present study. Some information, although imperfect, on the effect of demand changes on costs can be gleaned from the variance-covariance matrix of the estimated coefficients in the price equation. That is, if demand changes have a significant effect on cost changes, then there might well be high multicollinearity between the cost variables and the demand-change proxy. Of the six industries in the cost responsive group, there was evidence of high multicollinearity between the cost variables and the demand-change proxy in only two industries: motor vehicles and chemicals.

38

It would also be useful to have an ordinal ranking of price responsiveness to demand changes within the cost-plus-demand responsive group. However, given the differences from one industry to another in the best price regression and given the high costs of attempting further to normalize these regressions with a very small number of observations, it seemed most practical to settle for a simple classification rather than a ranking at this time. As the data for each of the demand-change proxies become available for more industries, it will be possible to develop an ordinal ranking of price responsiveness to demand changes.

39

It should be noted that there are a number of sources of simultaneous equation bias in these regressions. The most serious bias is that produced by the feedback from price changes to the demand-change proxies. When such simultaneity exists, it leads to biased and inconsistent coefficient estimates in the price equation. See Section IV for a discussion of this simultaneity problem.

40

This overlapping occurs because the output series at the two-digit SIC industry level sometimes includes its own basic raw material input as final product. To obtain a rough estimate of the upward bias on the coefficients for cost of materials, these coefficients were compared with each industry’s percentage of materials input to total product. These percentages are taken from the transactions table of the economy’s 1958 input-output table, as reported in a recent paper by Eckstein and Wyss [14, p. 26]. The upward bias on the coefficient for cost of materials was found most serious for tobacco, paper, nonelectrical machinery, motor vehicles, and instruments. In other industries, the bias did not seem serious. Data problems are discussed further in Section IV.

41

Yordon [50, pp. 287-94] also found that cost factors had a much greater impact on industry price changes than demand factors had in both concentrated and nonconcentrated U.S. manufacturing industries.

42

The beta coefficients indicate the relative contribution made by each of the independent variables to estimating the dependent variable; i.e., it indicates the relative importance of each independent variable in a multiple regression. The beta coefficient (Bi) corresponding to the regression coefficient (ai) is defined as Bi=aiσxσy where σx and σy represent the standard deviations of that independent variable and the dependent variable, respectively. See Ferber and Verdoorn [15, pp. 98-100].

43

The only exception was the inventory variable, Δ(INV/S), which took on a positive sign in one case, that of instruments. One possible explanation for the positive sign is that it reflects the increase in inventory holding costs, which rise when the stock of inventories increases.

44

This suggests that producers regard changes in wages and the cost of materials as more permanent than year-to-year changes in labor productivity.

45

The best price equation for each of these industries was corrected for autocorrelation by using the nonlinear autoregressive estimation package written by Robert Hall for the Econometric Programming Language of Data Resources, Inc. The estimates of the parameters in the autoregressive transformation of the regression are maximum likelihood estimates. The autoregressive corrections left the composition of the two indusry price groups unaffected. The autoregressively corrected price equations are available from the author upon request.

46
The appropriate formula for measuring the multicollinearity between two estimated coefficients, â1 and â2 in a multiple regression is
R¯2a^1a^2=1[[var(a^1)var(a^2)cov(a^1a^2)2]/var(a^1)var(a^2)]

The closer is R2â1â2 to unity, the more serious is the multicollinearity. The computed R2 between the coefficients of ΔlnCM and ΔlnCU in the regression for apparel was 0.64. The computed R2 between the coefficients of ΔlnCM and ΔlnCU in the regression for leather was 0.46.

47

Four or five alternative demand-change proxies were tried in the price regressions for this industry, but none proved to be significant.

48

A good discussion of the scale decision and the adjustment decision in the demand for labor can be found in Brechling [4]. The effects of relative factor prices on the demand for labor are well explained in Nadiri [35]. In conventional geometric analysis, the scale decision is represented by the position of the labor-capital isoquant, i. e., the isoquant shifts outward from the origin for a positive change in output and inward for a negative change in output. Similarly, the relative factor price decision is represented by the labor-capital price line. In equilibrium, the quantities of labor and capital demanded are jointly determined at the point where the labor-capital isoquant is tangent to the labor-capital price line.

† It should be noted that this labor demand function requires that output be exogenously determined. If simultaneity is present between output and employment, then single equation ordinary least squares estimates will be biased and inconsistent.

49

Another feature of the log-linear form is that it incorporates the assumption of declining marginal products of factor inputs, as in the Cobb-Douglas production function, whereas the simple linear form would imply constant marginal products of factor inputs.

50

The short-run elasticity of employment with respect to output was used for the rankings rather than the long-run equilibrium elasticity—i.e., B1 in equation (5) rather than B1α. The short-run elasticities were used because: (1) with annual data, the short-run employment response is itself fairly long (covering a one-year period); and (2) when all employees rather than total man-hours (all employees times hours paid per employee) is used as the dependent variable, the estimated adjustment coefficients (the α’s) are often implausibly small. It should be recognized that a ranking of industries by their long-run price and employment responses to demand changes may differ markedly from the ranking by short-run responses.

51

The industry employment demand regressions were also estimated with total man-hours as the dependent variable. As expected (since hours worked per employee will partially absorb the difference between desired and actual employment), the employments with respect to output elasticities were higher, and the speeds of adjustment of actual to desired employment were faster, for the man-hour regressions. The industry composition of the two employment groups was identical for the two dependent variables; however, the ranking of industries within each group differed slightly for the two dependent variables. Despite the somewhat better statistical quality of the man-hour regressions, the rankings were based on the employee regressions because policy targets are usually specified in terms of employees. The results of the man-hour regressions are available from the author upon request.

52

Waud [48] also found that relative factor prices had an important influence on the demand for labor in U.S. manufacturing industries.

53

When account is taken of changes in hours worked per employee, as well as for changes in the number of employees, the speed of adjustment of actual to desired employment naturally increases. In fact, the man-hour regressions indicated that over 70 per cent of the total adjustment occurs within one year for 18 of the 20 industries.

54

In cases where α > 1 (petroleum and chemicals) or where α < 0 (tobacco), some mis-specification is probably present. In the man-hour regressions for these industries, the adjustment coefficients assumed more plausible values.

55

The extent of autocorrelation in these regressions is difficult to evaluate, because they contain a lagged dependent variable which biases the Durbin-Watson statistic when ordinary least squares estimation is employed. See Griliches [18] and Nerlove and Wallis [37].

56

In Appendix II, it is demonstrated that interindustry differences in employment behavior are related to interindustry differences in the costs of changing employment. More specifically, it is shown that the costs of changing employment are significantly higher for the low employment sensitivity industries than for the high employment sensitivity industries.

57

Interindustry price linkages are also important in selecting good target industries. However, a proper analysis of industry price interdependence by means of the economy’s input-output relationships far exceeds the scope of this study; therefore, only interindustry wage linkages are examined.

58

Although only manufacturing industries have been included in this study, it should not be assumed that all key-wage-bargain industries belong to this sector. In fact, the construction industry has probably been a leading sector in the recent inflation. See the 1971 Economic Report of the President [58, pp. 57-60, 81-82].

59

For an analysis of the effect of steel wages and auto wages on wages in other industries, see McGuire and Rapping [30].

60

The procedure used in this study for identifying good target industries can be implemented with any given policy objective or with any given weighting pattern for the target characteristics. If, for example, the policy objective is to decrease unemployment with the lowest possible increase in prices, then the good target industries are those with relatively low price response and relatively high employment response to demand changes. The good target industries in expansionary periods would then be the opposites of the good target industries in contractionary periods.

61

An estimator is inconsistent if its bias persists for infinitely large samples.

62

The feedback from price changes to demand changes will be greater as the industry’s price elasticity of demand at the old price is higher.

63

For example, Darling and Lovell [5, pp. 131-42] found no role for price changes in their study of the determinants of inventory changes in U.S. manufacturing. Similarly, Zarnowitz [51, p. 391], in his study of price changes and unfilled order changes in U.S. manufacturing industries, concluded that if there was any feedback from price changes to unfilled order changes it would be small. A priori, it would also seem that the effect of price changes on the capacity utilization rate would be small.

64

Interindustry price effects are represented in the price equations by the cost of materials variable.

65

Salant and Vaccara [44, pp. 80-95] have shown that for some industries the employment losses in its principal supplying industries (the indirect employment effect) can be substantially greater than the employment losses in that industry itself (the primary employment effect).

66

After this study was virtually completed, I learned that Otto Eckstein and David Wyss of Harvard University have constructed a quarterly cost of materials index for two-digit SIC manufacturing industries which excludes intraindustry flows. Eckstein and Wyss also seem to have obtained or constructed quarterly data for prices, average hourly earnings, capacity utilization, real shipments, and profits to equity ratios, for almost all two-digit SIC manufacturing industries. If these data are reliable, they should add considerably to an understanding of industry price behavior. See Eckstein and Wyss [14, pp. 5-24].

67

Since comparative advantage is a relative rather than absolute measure, it would be necessary also to know the effectiveness of nondemand-changing policy instruments in different industries in order to determine the best assignment of policy instruments. In addition, it would also be useful to know in what industries monetary policy or fiscal policy was more effective as a demand-changing instrument. Presumably, however, there would be a point at which the marginal costs of increasing the degree of specialization in policy implementation exceeded the marginal benefits.

68

Unfortunately, it was not possible to check these implied trade-offs by regressing industry price changes on the industry unemployment rate because unemployment rate data are available on a two-digit SIC industry basis for only a few industries (and even then, only from 1957 on). Also, it is questionable whether industry unemployment is really a meaningful concept. In addition, the procedure of estimating separate price change and employment equations is consistent with the hypothesis that price changes and unemployment are jointly determined by the level of aggregate demand. This hypothesis seems sounder than assuming that the unemployment rate is exogenous and that it explains price changes or wage changes.

69

This is not to say that some of these policies, such as manpower programs to reduce job search time and turnover rates, would not be desirable as longer-term structural policies to improve the aggregate trade-off between inflation and unemployment.

70

For an analysis of the differences between industry transactions prices and industry list prices, see Stigler and Kindahl [47, pp. 56-70].

71

Food and tobacco were excluded from Table 5 because these were the only two industries in which the estimated elasticity of employment with respect to output was statistically insignificant.

72

The probability levels are for a two-tail test with 16 degrees of freedom, except for the unionization variable, where there are only 14 degrees of freedom. The computed t values were 2.80 for compensation per man-hour, 2.39 for the nonproduction-worker variable, 1.53 for general educational development, 1.48 for specific vocational preparation, and 1.23 for unionization.

73

Another important factor that could lead to interindustry differences in the elasticity of employment with respect to output is interindusry differences in the dispersion or fluctuation of output. That is, ceteris paribus, an industry whose output fluctuates widely would be expected to have a lower elasticity of employment with respect to output than one whose output shows less fluctuation, since relatively large output fluctuations make it difficult for the industry to assess its employment needs. In order to test this hypothesis on the two industry employment groups, the coefficient of variation (the standard deviation divided by the mean) of each industry’s output over the period 1954-67 was calculated. These industry coefficients were then averaged to obtain an average coefficient of variation of output for each of the two industry groups. In accord with the output-dispersion hypothesis, the low employment sensitivity group did display a higher dispersion of output than the high employment sensitivity group; the average coefficient of variation of output for the former group was 0.205, while that for the latter was 0.168. Again using the t test, the null hypothesis that the group means were equal was rejected at the 99 per cent probability level, the computed t value being 5.09 for 16 degrees of freedom.

  • Collapse
  • Expand
IMF Staff papers: Volume 19 No. 2
Author:
International Monetary Fund. Research Dept.