Has Greater Competition Restrained U.S. Inflation?

John V. Duca [*]

David D. VanHoose [+]

This paper shows how increased goods market competition affects the behavior of inflation in a multisector economy. By raising the price elasticity of demand, increased goods market competition theoretically lowers inflation and makes the aggregate price level less sensitive to aggregate demand shocks. We find that proxies for the aggregate degree of goods market competition are statistically and economically significant in short-run Phillips curve models of core inflation. Evidence indicates that heightened goods market competition has flattened the slope of the short-run, expectations-augmented Phillips curve and slightly lowered the nonaccelerating inflation rate of unemployment (NAIRU).

1. Introduction

The U.S. economic expansion of the mid-1990s has been characterized by low CPI inflation despite declines in the unemployment rate to levels previously associated with accelerating inflation. This combination of low unemployment and inflation has sparked debate over whether the economy can operate at higher full-information levels of production and employment without fueling a pickup in inflation. [1] One common argument is that heightened competition has lowered the nonaccelerating inflation rate of unemployment (NAIRU). Some attribute this change to international trade, [2] but past empirical work has found either no role or a limited role for trade variables (e.g., Tootell 1994; Fuhrer 1995). [3] This evidence is consistent with the fact that commodities and nonfood commodities comprise only 43% and 26% of the CPI, respectively, and that services are not (yet) highly tradable across U.S. borders.

Certainly, increased international trade may have plausibly raised the price elasticity of demand faced by a typical U.S. firm, so the recent emphasis that some observers have placed on international factors is not misplaced. Nevertheless, Milton Friedman (1968) originally pointed out that the NAIRU and its effect on inflation depend on the complete set of microeconomic relations in the goods and labor markets. Since the 1970s, purely domestic elements, such as the deregulation of several major industries, surely have also played a role. Undoubtedly, a number of factors together may have changed the competitive structure of U.S. goods markets.

In this paper, we seek to determine whether an increased degree of overall competition in U.S. goods markets may have lowered the NAIRU. In addition, we attempt to infer the extent to which altered aggregate competition may have reduced the slope of the short-run Phillips curve for the United States. In the Appendix, we provide an explicit theoretical model that shows how greater goods market competition can reduce the NAIRU and flatten the trade-off between inflation and unemployment. The basic channels giving rise to these effects are easy to understand at an intuitive level, however. With respect to NAIRU, an increase in the aggregate degree of competition in goods markets moves a monopolistically competitive economy closer to the perfectly competitive ideal, so that wedges between actual employment and the long-run employment level are smaller under perfect competition. As a result, increased aggregate competition reduces the NAIRU. At the same time, greater competition raises the price elasticity of dem and for the products sold by individual firms. Consequently, the marginal revenue product of labor at each firm becomes more elastic, making firm-level employment decisions more sensitive to price-level variations.

These basic effects carry through irrespective of the source of increased competition. Consequently, we focus on examining the empirical relationships between a measure of the aggregate degree of competition in U.S. markets and the inflation-unemployment trade-off and whether aggregate measures of competition help explain the changing behavior of inflation. Our overall index of competition, displayed in Figure 1, is adjusted for cyclical and energy-price-related movements in an aggregate markup variable for the nonfinancial corporate sector (see section 1 for details). This index indicates that the overall degree of competition in the United States has tended to be higher, not so much since the mid-1970s when import penetration jumped, but rather since the early 1980s shortly following the deregulation of many domestic nonfinancial industries (e.g., trucking, airlines, and telephone) and the spread of technological innovations that have reduced the market power of some industries (e.g., the rise of personal computers and the fall of IBM, and telecommunications in general). The view that U.S. goods markets are now generally more competitive is consistent with Federal Reserve Beige Book reports of the mid-1990s, in which retailers and producers reported facing greater competitive pressures that reduced their ability to pass along increased costs to their customers or to raise profit margins in the face of higher industry-wide demand (see Duca 1998).

Our study departs from two strands of recent research that document the time variation in the NAIRU. One approach uses more modern statistical techniques, such as Staiger, Stock, and Watson (1997), who show how variable the NAIRU is and track changes in it over time. Nevertheless, this approach does not explain why changes in the NAIRU have occurred and how to track the economic factors behind them. Another approach simply attributes time-varying changes in the NAIRU to temporary supply shocks, such as Gordon (1997). A drawback of this approach is that it may miss changes in the underlying structure of the U.S. economy. Our study departs from both recent strands in the NAIRU literature by assessing whether changes in the underlying, aggregate degree of product market competition have significantly affected the inflation process within a NAIRU-like framework.

Section 2 begins by presenting our empirical specification for evaluating the relationships between competition and NAIRU, measuring the overall degree of goods market competition, and assessing how competition affects the NAIRU and the short-run Phillips curve trade-off. Section 3 then tests whether this index of competition is statistically and economically significant when incorporated into short-run Phillips curve models of inflation. Section 4 concludes.

2. The NAIRU and the Inflation-Unemployment Trade-Off

To test whether increased competition has reduced the short-run sensitivity of inflation to unemployment rate variations and reduced the NAIRU, we incorporate measures of goods market competition into a benchmark Phillips curve model by jointly estimating regressions of inflation and profits. We first discuss these equations separately before presenting the two-equation system and estimation results.

According to the basic NAIRU framework,

[[pi].sub.t] = [beta]([U.sub.t] - [[U.sup.*].sub.1]) + [gamma][E.sub.t-1][[pi].sub.t], (1)

where [[pi].sub.t] [equivalent to] inflation rate, [U.sub.t] [equivalent to] unemployment rate, and [[U.sup.*].sub.t] [equivalent to] NAIRU, which is unobserved or latent. If one estimates

[[pi].sub.t] = [alpha] + [beta][U.sub.t] + [gamma][E.sub.t-1][[pi].sub.t], (2)

then substituting Equation 2 into Equation 1 implies that [[U.sup.*].sub.t] = -([alpha]/[beta]). If a rise in an index of overall competition ([[epsilon].sub.t], discussed in greater detail below) affects the NAIRU ([[U.sup.*].sub.t]) and the slope of the short-run Phillips curve, then we can rewrite Equations 1 and 2, respectively, as:

[[pi].sub.t] = ([[psi].sub.0] + [[psi].sub.1][[epsilon].sub.t])[[U.sub.t] - ([[micro].sub.0] + [[micro].sub.t][[epsilon].sub.t])] + [[gamma]E.sub.t-1][[pi].sub.t], (3)

and

[[pi].sub.t] = [alpha] + [beta][U.sub.t] + [gamma][E.sub.t-1][[pi].sub.t] + [tau][[epsilon].sub.t][U.sub.t] + [phi][[epsilon].sub.t], (4)

where ([[micro].sub.0] + [[micro].sub.1][[epsilon].sub.1] substitutes for the latent variable [[U.sup.*].sub.t] in Equation 1. Note that the cross-product quadratic term ([[psi].sub.1][[micro].sub.1][[[epsilon].sup.2].sub.t] from Equation 3 is absent from the empirical specification in Equation 4 because it is a second-order effect that proved to be statistically insignificant in other tests, likely because [[[epsilon].sup.2].sub.t] is multicollinear with [[epsilon].sub.t]. The empirically derivable NAIRU in Equation 4 equals -([alpha] + [phi][[epsilon].sub.t])/([beta] + [tau][[epsilon].sub.t]). The slope of the short-run Phillips curve in Equation 3 equals ([[psi].sub.] + [[psi].sub.1][[epsilon].sub.1]), and the empirically derivable slope in Equation 4 equals ([beta] + [tau][[epsilon].sub.t]).

As noted earlier, our measure of aggregate goods market competition, [epsilon], has noticeably trended up since the late 1970s amid waves of deregulation in the trucking, airline, and communications industries (Figure 1). If greater competition flattens the negative slope of the short-run Phillips curve ([beta] [less than] 0), then the interactive coefficient should be positive ([tau] [greater than] 0). If a rise in competition also lowers the NAIRU, then the overall effect of the noninteractive and interactive coefficients should reduce the value of (-[alpha] - [phi][[epsilon].sub.t]/([[beta].sub.1] + [tau][[epsilon].sub.t]), the expression used to infer the time-varying NAIRU.

A Basic Empirical Model of the Inflation--Unemployment Trade-Off

To implement Equation 4, we use Fuhrer's (1995) empirical model as a benchmark, both because of its stability over time and because of its relative parsimony:

[[pi].sub.t] = [alpha] + [[[sigma].sup.12].sub.i=1] [[delta].sub.i][[pi].sub.t-i] + [[beta].sub.1][U.sub.t-1] + [[beta].sub.2][[delta]U.sub.t-1] + [[gamma].sub.1][ENERGY.sub.t], (5)

where [[pi].sub.t] is core inflation (CPI excluding food and energy) and ENERGY [equivalent to] real energy price inflation (PPI energy prices deflated by the PPI). A lag in [U.sub.t] is used to avoid simultaneity, [4] and [delta][U.sub.t-1] captures the "speed effect" of unemployment rate changes. [5]

Measuring Overall Goods Market Competition

Following Duca and VanHoose (1998a, b), we develop an index of overall goods market competition. To do this, we measure the aggregate price elasticity of demand via a cyclically adjusted series of the average markup of price over marginal cost. In contrast with studies of whether markups are cyclical (e.g., Ramey 1991; Rotemberg and Woodford 1991; Basu and Fernald 1997), we focus on long-term movements in the markup. Although returns to scale vary across industries, we assume constant returns to scale in the aggregate, consistent with Basu and Fernald, and derive the markup using profits data.

Below, we consider a more specific version of a general profit function, given by

[psi] = constant + [a.sub.1]S + [a.sub.2]X, (6)

where [psi] denotes after-tax profits, S is a vector of structural determinants of profits, and X is a vector of variables reflecting cyclical and other short-run factors. From Equation 6, profit share is adjusted for short-run factors ([[psi].sup.a] = [psi] - [a.sub.2]X), which implies a cyclically adjusted markup of [[psi].sup.a] = 1 - 1/[[micro].sup.a], with [[micro].sup.a] [equivalent to] price/marginal cost. The standard model of imperfect competition implies [[micro].sup.a] = 1 - 1/([epsilon] - 1) and [epsilon] = 1/[[psi].sup.a], where [epsilon] [equivalent to] /price elasticity of demand/.

To implement Equation 6, we use nominal corporate sector GDP to proxy for sales and include two measures of fixed costs and several variables that reflect short-run factors. [6] In principle, fixed costs can be measured by the ratios of consumption of fixed capital (FC) and net interest to nonfinancial corporate GDP (NI). However, because the relative use of debt and equity has shifted and it is difficult to disentangle equity investment from stock price shifts, NI reflects swings in leverage, interest rates, and inventory financing. Because not adjusting for NI would give a distorted picture of fixed costs and the markup, we subtract off the estimated effect of NI to measure [epsilon]. [7]

To cyclically adjust the markup for profits (after inventory valuation [IVA] and capital-consumption adjustments [CCAdj]) we include the t to t -- 3 lags of real GDP growth ([delta]y), and year-over-year GDP growth lagged four quarters (YOY[delta]y). [8] One complication in estimating GDP effects is that profits for the nonfinancial corporate sector (which includes the energy industry, for which only complete annual NIPA data are available) fell less in the oil-induced recessions of the 1970s when oil profits surged than in other recessions. To better estimate GDP effects, we include lags of the real retail price of energy (CPI energy/CPI) that parallel those of the GDP terms. Because our inflation model estimates core inflation, netting out energy effects from profits yields a more relevant measure of competition.

To avoid misspecification bias affecting the GDP and oil coefficients, the profits equation we estimate below includes a real exchange rate term. This variable, RER, tracks the mid-1980s decline in profits stemming from the dollar's huge appreciation and profit recovery arising from the dollar's subsequent decline. Levels are used because RER is dominated by a hump in the mid- 1980s that depressed profits even as the dollar was falling off its 1985 highs. To conserve degrees of freedom and simplify joint maximum likelihood estimation, only the t - 4 lag of RER, which was the only significant lag, was used. Because the real exchange rate can plausibly affect the pricing power of firms, its estimated effect is not subtracted from profits when constructing [epsilon].

A Joint Estimation of Competition and the Inflation-Unemployment Trade-Off

To test the implications of Equation 4 using the specific model in Equation 5, we imbed Fuhrer's model into a full-information, maximum-likelihood approach similar to that of Mishkin (1978). [9] We jointly estimate the following profit and inflation equations over the intervals 1959: Q1-1997:Q4 and 1960:Q2-1997:Q4, respectively:

[[psi].sub.t] = [[upsilon].sub.0] + [[upsilon].sub.1][G.sub.t] + [[upsilon].sub.2][NI.sub.t] + [[upsilon].sub.3][RER.sub.t-4] + [[upsilon].sub.4][FC.sub.t], and (7)

[[pi].sub.t] = [alpha] + [[[sigma].sup.12].sub.i=1] [[delta].sub.i][[pi].sub.t-i] + [[beta].sub.t][U.sub.t-1] + [[beta].sub.2][delta][U.sub.t-1] + [[gamma].sub.1][ENERGY.sub.t] + [[omega].sub.1][NIXON.sub.t] + [[omega].sub.2][NIXOFF.sub.t] + [[MATHEMATICAL EXPRESSIONS NOT REPRODUCIBLE IN ASCII].sub.1]{1/0.5[[([psi] - [[upsilon].sub.1]G - [[upsilon].sub.1]G - [[upsilon].sub.2]NI).sub.t-1] + [([psi] - [[upsilon].sub.1]G - [[upsilon].sub.2]NI).sub.t-2]]} + [[MATHEMATICAL EXPRESSIONS NOT REPRODUCIBLE IN ASCII].sub.2]([[U.sup.*].sub.t-1]{1/0.5[([[psi] - [[upsilon].sub.1]G - [[upsilon].sub.2]NI).sub.t-1] + ([[psi] - [[upsilon].sub.1]G - [[upsilon].sub.2]NI).sub.t-2]}), (8)

where Equation 8 imbeds a jointly estimated value of the index of goods market competition equal to

[[epsilon].sub.t] = 1/0.5[([[psi] - [[upsilon].sub.1]G - [[upsilon].sub.2]NI).sub.t] + [([[psi] - [[upsilon].sub.1]G - [[upsilon].sub.2]NI).sub.t-1]] (9)

Because profits are a residual claim and thus are noisy, we base [epsilon] on a two-quarter moving average. In light of Equation 9, the inflation Equation 8 incorporates the direct and Phillipscurve slope effects of [epsilon], with the coefficient [[MATHEMATICAL EXPRESSIONS NOT REPRODUCIBLE IN ASCII].sub.1] multiplying a term equivalent to [epsilon], and the coefficient [[MATHEMATICAL EXPRESSIONS NOT REPRODUCIBLE IN ASCII].sub.2] multiplying a term equivalent to the interactive variable ([epsilon]U). [10]

Because of the unusual Nixon wage-price controls episode, in Equation 8, we follow Fuhrer by including Gordon's (1977) dummies for the effects on inflation of the imposition (NIXON = 1 over 1971:3--1972:3) and lifting (NIXOFF = 1 over 1974:2--1975:1) of price controls. In estimating this equation, the sum of coefficients on lagged inflation ([[sigma].sub.i][[alpha].sub.i]) is constrained to equal unity for the model to have sensible long-run properties (i.e., expected inflation has a one-to-one effect on inflation when the unemployment rate is equal to the NAIRU). To see if the model's overpredictions of inflation in the early 1990s owed to changes in how people form inflation expectations, in other runs we replaced lags of inflation with the average, year-ahead expectation of inflation from the University of Michigan's survey. This substitution, however, yielded a poorer fit, perhaps because the survey question covered overall rather than core CPI inflation.

3. The Empirical Results

Results for two empirical models are reported in Table 1. One is a "conventional" model (model 1) that implicitly assumes that the degree of goods market competition has remained constant over time. The other "competition" model (model 2) jointly estimates a modified version of that inflation equation along with a regression of corporate profits from which a simultaneously estimated degree of product market competition is derived. Not surprisingly, the profits results indicate that GDP growth and real oil prices boost nonfinancial corporate profits, whereas higher net interest, capital stock depreciation, and real exchange rates tend to reduce profits.

Turning to inflation, the level and first difference of the unemployment rate have negative and statistically significant signs in models 1 and 2. Consistent with the theoretical framework presented in the Appendix, greater goods market competition flattens the negative slope of the short-run Phillips curve, as implied by the highly significant and positive coefficient on the interactive [epsilon] U term. Also consistent with theory is a negative and highly significant coefficient on the noninteractive [epsilon] lag, which together with the interactive term implies that the NAIRU is declining in the degree of product market competition. The NAIRU implied by model 2 was equal to 5.60% in 1997:Q4, as compared with the 6.04% estimate for NAIRU derived from model 1. The Wald statistic on the two [epsilon] terms implies that they are jointly significant, in line with the higher corrected [R.sup.2] of model 2. Using results from model 2, Figure 2 shows, in a comparison of 1979:Q4 with 1997:Q4, how a greater degree of product market competition has flattened the short-run Phillips curve and reduced the NAIRU [11, 12]

One concern with model 2 is that the coefficients may only be significant because they capture an effect of changes in the real exchange rate. To address this issue, we added the level, log-level, first difference, and log first difference of the real exchange rate to the inflation equations in models 1 and 2, using either a t -- 1 or t -- 4 lag to avoid simultaneity bias, since the RER variable uses CPI data. With respect to model 1, the real exchange rate term that added the most information was the t -- 1 lag of [delta] log(RER). In this case, this term was only marginally significant (see Table 2). In every modification of model 2, the [epsilon] and [epsilon] U terms remained statistically significant with the anticipated signs, but the real exchange rate term was always insignificant. These findings imply that it is difficult to identify a significant, direct effect of the real exchange rate on U.S. inflation in reduced-form, short-run Phillips curve models.

There are several plausible reasons for this. First, much of the direct effect of exchange rates may be reflected in the lagged inflation terms. Second, adding such terms to short-run Phillips curve models may not be adequate if pass-through of exchange rate changes to goods prices is complex and changes over time, as suggested by the literature on exchange rate passthroughs. [13] Third, the bulk of the indirect effect of exchange rates by altering the degree of economic slack is likely captured by the unemployment terms.

These results, however, do not rule out the possibility of indirectly capturing exchange rate effects on inflation by incorporating exchange rate movements in the overall degree of product market competition, as the [epsilon] terms do in model 2. In other regressions not shown, we tested [epsilon] terms that netted out the effect of real exchange rates. Although the [epsilon] terms remained significant with the hypothesized signs, the models yielded slightly lower corrected [R.sup.2] values and less sensible values for the NAIRU. In general, the findings support the approach of controlling for the varying effects of exchange rates on inflation by indirectly gauging these effects on the overall degree of competition in product markets, which reflects both foreign and domestic factors.

4. Conclusion

This paper assesses whether increased goods market competition has reduced inflation and the NAIRU in the 1990s. Our empirical findings provide some support for the view that the NAIRU is lower and that the short-run Phillips curve is shallower because goods markets are now more competitive. With respect to what affects the overall degree of goods market competition, the results favor a more general approach that does not focus on any one source (foreign or domestic) influencing the aggregate degree of pricing power in the United States. From a broader perspective, these findings imply that freer markets not only boost an economy's sustainable output path, but also reduce the vulnerability of the price level to swings in capacity pressures.

Nevertheless, the reduced-form nature of short-run Phillips curve models implies that the empirical results should be interpreted with some caution because these models reflect the confluence of several factors, such as supply shocks, the formation of inflation expectations, and government interventions (e.g., price controls). With these caveats in mind, the empirical findings support the view that the greater competition may be restraining inflation and that this working hypothesis is worthy of further study.

(*.) Senior Economist and Assistant Vice President, Federal Reserve Bank of Dallas, Box 655906, Dallas, TX 75265, USA.

(+.) Professor of Economics and Young Boozer, Jr., Faculty Fellow, College of Commerce and Business Administration, Box 870224, University of Alabama, Tuscaloosa, AL 35487-0224, USA; E-mail [e-mail]; corresponding author.

We thank Michelle Burchfiel and Justin Marion for providing research assistance, Jeffrey Marquardt for suggesting this line of research to us, and James Cover and an anonymous referee for extending helpful comments. The views expressed are those of the authors and are not necessarily those of the Federal Reserve Bank of Dallas and the Federal Reserve System. Any remaining errors are our own.

Received May 1997; accepted June 1999.

(1.) Several noteworthy quotations from this debate follow.

The markets and the Fed react to any appearance of acceleration with higher interest rates and the economy [GDP growth] then falls back to 2.2% or lower. Rohatyn (1996, A21)

We don't see a connection between the numbers out there and what we feel is going on in our business. There is absolutely no inflation. There's no pricing power at all. John F. Welch, Jr., CEO General Electric (see Stevenson 1996, C1)

...there is now a debate, a serious debate in the country, about whether there is a maximum growth rate we can have over any period of years without inflation. President Clinton (see Purdum 1996, C25)

The fast-growing gang ... contend that the world economy has changed fundamentally....Low unemployment won't trigger higher prices as it once did, they say; instead, global competition and corporate restructuring will keep prices down. ... Some say unemployment could drop as low as 5% without triggering inflation. Wilke (1996, Al)

NAIRU is to economics what the Nehru jacket is to fashion: Outdated. Dana Mead, CEO, Tenneco, Inc. (as quoted in Stevenson 1996, C15)

(2.) "There are a lot of people, including a lot of Republican executives in the manufacturing sector, who believe that global competition will keep down inflation ... and permit higher growth rates in the next 10 years than the last 25." President Clinton (see Purdum 1996, C25)

(3.) This is consistent with the tendency for forecasters to overpredict inflation during the most recent expansion.

(4.) To control for the 1994 time series break in the household survey of employment, 0.2 percentage points is added to [U.sub.t] before 1994. Compared to Fuhrer (1995), ENERGY replaces real oil prices to account for broader energy prices, but this difference did not affect the qualitative results.

(5.) This model is equivalent to Fuhrer (1995), who includes [U.sub.t-2] instead of [[delta]U.sub.t-1] and infers the speed effect from -[[beta].sub.2]. The version used here, rather than including two lags of the level of unemployment, is preferable because extending it to test for competition effects involves adding only one interactive ([epsilon]U) term and does not indirectly test for an effect of [epsilon] on the [delta]U "speed effect," which theory does not necessarily predict and which proved to be insignificant in tests of the model.

(6.) Using revised NIPA data, various time-trend terms are insignificant.

(7.) For example, the profit-to-output ratio would fall if firms increased their leverage for a few years and then rebound if firms de-levered because firms would be making relatively greater payments to debt capital holders than to equity capital holders. Without correcting for this swing, the cyclically adjusted profit-to-output ratio would have a U-shape, as it did during the late 1980s rise and the early 1990s fall in leverage.

(8.) The [[delta]y.sub.1-1] control for short-run growth effects and the four-quarter lag of YOY[delta]y controls for slightly longer lagged effects.

(9.) We estimate this system using a SAS procedure similar to that of Mishkin (1983).

(10.) We did not include dummies for the wage-price control terms in both equations because such dummies were insignificant in the profit equation. This likely reflects that multicollinearity makes it difficult to identify separate effects of the same variable in a two-equation system. Indeed, such dummies have much higher t-statistics and are marginally significant in a single-equation estimation of profit share. Such difficulties posed by multicollinearity coupled with a need to obtain convergence led us to avoid including the same variables in both equations.

(11.) The models estimated use an overall unemployment rate, even though the changes in demographics, particularly in the relative importance of younger, less employable adults, can plausibly alter the NAIRU. Other models were estimated using the unemployment rate for those 20 years and older. The qualitative results with respect to the significance of the competition variables are similar to those in Tables 1 and 2, but are quantitatively somewhat larger.

(12.) The model is difficult to evaluate over short samples because it has 35 coefficients and four unemployment variables in two equations that are jointly estimated using maximum likelihood. Over 1960:Q2-1991:Ql, which drops the most recent expansion, we obtain similar results, but with competition terms significant at the 95%, instead of 99%, confidence level. In samples dropping the 1980s and 1990s, the competition and unemployment rate terms were insignificant, likely reflecting two factors. First, in short samples, it is hard to identify the model's many coefficients, especially those on the four unemployment terms. (This may also explain why the model did not converge in short samples such as 1983-1997.) Second, it is easier to identify the effect of the degree of product market competition when the 1980s or 1980s-1990s are included because these two periods are similar to the 1960s in having-relative to the 1970s-a higher average degree of product market competition, a lower average inflation rate, an d a greater tendency to post inflation rates below those predicted by standard Phillips curve models.

(13.) The effect of exchange rates could change over time because of changes in the degree of openness, the importance of traded versus nontraded goods for inflation, trading partners, and exchange rate pass-through behavior. The latter could stem from hysteresis (Baldwin 1988; Baldwin and Krugman 1989) or from changes in nontariff barriers (Branson 1989), Most studies find evidence that the pass-through of exchange rates to U.S. import prices fell in the 1980s (e.g., Piggot and Reinhart 1985; Mann 1986; Baldwin 1988). Other studies have found either mixed evidence supporting this view (e.g., Hooper and Mann 1989) or evidence against it (e.g., Melick 1990).

References

Baldwin, Richard E. 1988. Hysteresis in import prices: The beachhead effect. American Economic Review 78:773-85.

Baldwin, Richard E., and Paul Krugman. 1989. Persistent trade effects of large exchange rate shocks. Quarterly Journal of Economics 104:635-54.

Ball, Laurence. 1988. Is equilibrium indexation efficient? Quarterly Journal of Economics 103:299-311.

Basu, Susanto, and John G. Fernald. 1997. Returns to scale in U.S. production: Estimates and implications. Journal of Political Economy 105:249-83.

Branson, William H. 1989. Comments on "Exchange rate pass-through in the 1980s: The case of U.S. imports of manufactures." Brookings Papers on Economic Activity 1:330-3.

Duca, John V. 1987. The spillover effects of nominal wage rigidity in a multisector economy. Journal of Money, Credit, and Banking 19:117-21.

Duca, John V. 1998. How increased product market competition may be reshaping America's labor markets. Federal Reserve Bank of Dallas Economic Review Fourth Quarter:2-16.

Duca, John V., and David D. VanHoose. 1991. Optimal wage indexation in a multisector economy. International Economic Review 32:859-67.

Duca, John V., and David D. VanHoose. 1998a. The rise of goods market competition and the fall of wage inflexibility: Endogenous wage contracting in a multisector economy. Federal Reserve Bank of Dallas Working Paper No. 98-05.

Duca, John V., and David D. VanHoose. 1998b. The rise of goods market competition and the decline in wage indexation. Journal of Macroeconomics 20:579-98.

Friedman, Milton, 1968. The role of monetary policy. American Economic Review 58:1-17.

Fuhrer, Jeffrey. 1995. The Phillips curve is alive and well. Federal Reserve Bank of Boston New England Economic Review March/April:41-56.

Gordon, Robert J. 1977. Can the inflation of the 1970s be explained? Brookings Papers on Economic Activity 1:253-76.

Gordon, Robert J. 1997. The time-varying NAIRU and its implications for economic policy. Journal of Economic Perspectives 11:11-32.

Hooper, Peter, and Catherine L. Mann. 1989. Exchange rate pass-through in the 1980s: The case of U.S. imports of manufactures. Brookings Papers on Economic Activity 1:297-325.

Mann, Catherine L. 1986. Prices, profit margins, and exchange rates, Federal Reserve Bulletin 72:366-79.

Melick, William R. 1990. Estimating pass-through: Structure and stability. International Finance Discussion Paper No. 387, Board of Governors of the Federal Reserve System, September.

Mishkin, Frederic S. 1978. Efficient-markets theory: Implications for monetary policy. Brookings Papers on Economic Activity 3:707-52.

Mishkin, Frederic S. 1983. A rational expectations approach to macroeconometrics: Testing policy ineffectiveness and efficient-markets models. Chicago: University of Chicago Press.

Piggot, Charles, and Vincent Reinhart. 1985. The strong dollar and U.S. inflation. Federal Reserve Bank of New York Quarterly Review 10:23-9.

Purdum, Todd S. 1996. Clinton sees global competition as a way of curtailing inflation. New York Times, 17 February, p. C25.

Ramey, Valerie. 1991. Comments on "Markups and the business cycle." In NBER macroeconomics annual 1991, edited by Olivier Blanchard and Stanley Fischer. Cambridge, MA: MIT Press, pp. 134-9.

Rohatyn, Felix G. 1996. Recipe for growth. Wall Street Journal, 11 April, p. A21.

Rotemberg, Julio, and Michael Woodford. 1991. Markups and the business cycle. In NBER macroeconomics annual 1991, edited by Olivier Blanchard and Stanley Fischer. Cambridge, MA: MIT Press, pp. 63-129.

Staiger, Douglas, James H. Stock, and Mark W. Watson. 1997. The NAIRU, unemployment, and monetary policy. Journal of Economic Perspectives 11:33-50.

Stevenson, Richard W. 1996. It's heresy at Fed but critics say: Step on the gas. New York Times, 7 June, pp. C1 and C15.

Tootell, Geoffrey. 1994. Restructuring, the NAIRU, and the Phillips curve. Federal Reserve Bank of Boston New England Economic Review September/October:31-44.

Wilke, John. 1996. War on inflation curbs world-wide growth too much, some say. Wall Street Journal, 10 April, p. Al.

         Core CPI Inflation and Profit Share, 1960: Q2-1997:Q4 [a]
                                    Inflation Models [b]
Variable                            Model 1              Model 2
Constant                              0.0050 [**]          0.0735 [**]
                                     (3.53)               (3.17)
[U.sub.t-1]                          -0.0008 [**]         -0.0108 [**]
                                    (-3.74)              (-3.06)
[delta][U.sub.t-1]                   -0.0042 [**]         -0.0034 [**]
                                    (-4.44)              (-2.84)
[NIXON.sub.t]                        -0.0027 [+]          -0.0025
                                    (-1.78)              (-1.49)
[NIXOFF.sub.t]                        0.0094 [**]          0.0088 [**]
                                     (4.68)               (5.29)
[ENERGY.sub.t]                        0.0035               0.0034
                                     (1.29)               (0.91)
[[epsilon].sub.t-1]                                       -0.0070 [**]
                                                         (-2.82)
[([epsilon]*U).sub.t-1]                                    0.0010 [**]
                                                          (2.69)
NAIRU, 97:Q4                          6.04%                5.60%
Wald: [epsilon] . U = [epsilon] = 0                        8.32
q (24)                               18.79                17.50
RMSE [c]                              0.00314              0.00307
[R.sup.2]                             0.8020               0.8104
                                    Profit Model Jointly
                                    Estimated With Model 2
Variable                            Variable                Model 2
Constant                            Constant                  0.1717 [**]
                                                            (10.52)
[U.sub.t-1]                         [delta][y.sub.t]          0.1622 [**]
                                                             (4.83)
[delta][U.sub.t-1]                  [delta][y.sub.t-1]        0.1307 [**]
                                                             (2.92)
[NIXON.sub.t]                       [delta][y.sub.t-2]        0.0886 [*]
                                                             (2.01)
[NIXOFF.sub.t]                      [delta][y.sub.t-3]        0.1007 [**]
                                                             (3.09)
[ENERGY.sub.t]                      YOY[delta][y.sub.t-4]     0.0472 [+]
                                                             (1.91)
[[epsilon].sub.t-1]                 [delta][oil.sub.t]        0.0293 [*]
                                                             (2.14)
[([epsilon]*U).sub.t-1]             [delta][oil.sub.t-1]      0.0170
                                                             (1.36)
                                    [delta][oil.sub.t-2]      0.0524 [**]
                                                             (3.04)
                                    [delta][oil.sub.t-3]      0.0251 [+]
                                                             (1.67)
                                    YOY[delta][oil.sub.t-4]   0.0202 [+]
                                                             (1.66)
                                    [FC.sub.t]               -0.4222 [**]
                                                            (-3.20)
                                    [NI.sub.t]               -1.2625 [**]
                                                            (-6.70)
                                    [RER.sub.t-4]            -0.003 [**]
                                                            (-3.64)
NAIRU, 97:Q4                        p (1)                     0.9446 [**]
Wald: [epsilon] . U = [epsilon] = 0                         (26.06)
q (24)                              Q (24)                   26.11
RMSE [c]                            RMSE                      0.00296
[R.sup.2]                           [R.sup.2]                 0.9712

(a.)[[epsilon].sub.t] is jointly estimated with the profits model, but lagged [epsilon] is used in the inflation equation. * (**, +) denotes significance at the 5% (1%, 10%) level.

(b.)Sums of lagged [pi] coefficients not reported; they are constrained to equal 1.

(c.)RMSE, root mean square error.

          Core CPI Inflation and Profit Share, 1960:Q2-1997:Q4 [a]
                                    Inflation Models [b]
Variable                            Model 1              Model 2
Constant                              0.0049 [**]          0.0707 [**]
                                     (3.49)               (3.14)
[U.sub.t-1]                          -0.008 [**]          -0.01O3 [**]
                                    (-3.71)              (-3.03)
[delta][U.sub.t-1]                   -0.0043 [**]         -0.0035 [**]
                                    (-4.57)              (-2.94)
[NIXON.sub.t]                        -0.0031 [*]          -0.0027
                                    (-2.04)              (-1.63)
[NIXOFF.sub.t]                        0.0096 [**]          0.0089 [**]
                                     (4.82)               (5.56)
[ENERGY.sub.t]                        0.0040               0.0037
                                     (1.50)               (1.03)
[[epsilon].sub.t-1]                                       -0.0067 [**]
                                                         (-2.77)
([epsilon] * U).sub.t-1]                                   0.0010 [**]
                                                          (2.64)
[delta]log[(RER).sub.t-1]            -0.0136 [+]          -0.0104
                                    (-1.67)              (-0.99)
NAIRU, 97:Q4                          6.01%                5.58%
Wald: [epsilon] * U = [epsilon] = 0                        8.05 [*]
q (24)                               18.72                16.96
RMSE [c]                              0.00312              0.00307
[R.sup.2]                             0.8046               0.8115
                                    Profit Model Jointly
                                    Estimated With Model 2
Variable                            Variable                Model 2
Constant                            Constant                  0.1717 [**]
                                                            (10.50)
[U.sub.t-1]                         [delta][y.sub.t]          0.1630 [**]
                                                             (4.73)
[delta][U.sub.t-1]                  [delta][y.sub.t-1]        0.1321[**]
                                                             (2.95)
[NIXON.sub.t]                       [delta][y.sub.t-2]        0.0887 [+]
                                                             (1.94)
[NIXOFF.sub.t]                      [delta][y.sub.t-3]        0.1009 [**]
                                                             (3.04)
[ENERGY.sub.t]                      YOY[delta][y.sub.t-4]     0.0463 [+]
                                                             (1.88)
[[epsilon].sub.t-1]                 [delta][oil.sub.t]        0.0291 [*]
                                                             (2.14)
([epsilon] * U).sub.t-1]            [delta][oil.sub.t-1]      0.0168
                                                             (1.34)
[delta]log[(RER).sub.t-1]           [delta][oil.sub.t-2]      0.0523 [**]
                                                             (3.09)
                                    [delta][oil.sub.t-3]      0.0247
                                                             (1.66)
                                    YOY[delta][oil.sub.t-4]   0.0198
                                                             (1.64)
                                    [FC.sub.t]               -0.4288 [**]
                                                            (-3.26)
                                    [NI.sub.t]               -1.2384 [**]
                                                            (-6.13)
                                                             -0.0003 [**]
                                    [RER.sub.t-4]           (-3.55)
NAIRU, 97:Q4                        p (1)                     0.9452 [**]
Wald: [epsilon] * U = [epsilon] = 0                         (25.82)
q (24)                              q (24)                   26.25
RMSE [c]                            RMSE                      0.00296
[R.sup.2]                           [R.sub.2]                 0.9730

(a.)[[epsilon].sub.t] is jointly estimated with the profits model, but lagged [epsilon] is used in the inflation equation. (*.)((**.),(+.))denotes significance at the 5% (1%, 10%) level.

(b.)Sums of lagged [pi] coefficients not reported; they are constrained to equal 1.

(c.)RMSE, root mean square error.

Appendix: A Model of Goods Market Competition in a Multisector Economy

This appendix outlines a simple theoretical model based on Duca and VanHoose (1998a, b), who in turn draw from the work of Duca (1987), Ball (1988), and Duca and VanHoose (1991). Consider a model based on monopolistic competition among atomistic firms in a multisector economy. An exogenous fraction, [omega], of the sectors have nominal wage contracts. In the remaining portion of sectors, 1 - [omega], nominal wages are determined by market forces. The sectors, indexed j, are distributed uniformly along a unit interval. Production by a firm in sector j is

[y.sub.j] = [[alpha]l.sub.j] + [theta], (A1)

where [y.sub.j] is the log of firm j's output, [l.sub.j] is the log of employment at firm j, and [theta] is a common productivity shock whose mean value is zero and finite variance is [[[sigma].sup.2].sub.[theta]]. The demand for the output of firm j as a share of aggregate output is

[y.sub.j] - y = -[epsilon]([p.sub.j] - p), (A2)

where y [equivalent to] [[[integral of].sup.1].sub.0] [y.sub.j] dj is aggregate output, p [equivalent to] [[[integral of].sup.1].sub.0] [p.sub.j] dj is the aggregate price level, and [epsilon] [greater than] 1 is the elasticity of demand for firm j's output. Aggregate demand is given by:

y = m + v - p, (A3)

where v has a zero mean and variance [[[sigma].sup.2].sub.v]. Both [theta] and v are assumed to be independently distrubuted.

Taking antilogs of Equations Al, A2, and A3 and combining the resulting expressions with the profit function [[pi].sub.j] = [P.sub.j][Y.sub.j] - [W.sub.j][L.sub.j], yields the labor demand function for a firm in sector j (with the intercept term suppressed for the moment):

[[l.sup.d].sub.j] = -[epsilon]([w.sub.j] - p) + (m + v - p) + ([epsilon] - 1)[theta]/[alpha] + [epsilon] - [alpha][epsilon]. (A4)

where [w.sub.j] is the log of the normal wage at a firm in sector j. Each firm acts as a perfect competitor in its labor market, in which, following Ball (1988), it faces a pool of immobile workers whose labor supply schedule is given [[l.sup.s].sub.j] = [lambda]([w.sub.j] - p), where [lambda] [greater than] 0. For both groups of sectors, the full-information, market-clearing wage is

[[w.sup.*].sub.j] - p = (m + v - p) + ([epsilon] - 1)[theta]/[lambda]([alpha] + [epsilon] - [alpha][epsilon]) + [epsilon]. (A5)

This is the wage actually paid by a firm in sector j if its sector is among the share 1 - [omega] of sectors that do not use contracts. For such a firm, denoted "nc," substitution of Equation A5 into Equation A4 and the result into Equation A1 yields a noncontracting firm's output:

[[y.sup.nc].sub.j] = [alpha][lambda](m + v - p) + [epsilon](1 + [lambda])[theta]/[lambda]([alpha] + [epsilon] - [alpha][epsilon]) + [epsilon]. (A6)

At a contracting firm, however, nominal wage contracts are set to satisfy [w.sub.j] = [Ew.sub.j] = 0. Using this in Equations A4 and A1 yields the output of a firm with a wage contract:

[[y.sup.c].sub.j] = [alpha][epsilon]p + [alpha](m + v - p) + [epsilon][theta]/[alpha] + [epsilon] - [alpha][epsilon]. (A7)

Because [[y.sup.c].sub.j] = [y.sup.c] for all j [epsilon] [0, [omega]] and [[y.sup.nc].sub.j] = [y.sup.c] for all j [epsilon] ([omega], 1], it follows that y = [omega][y.sup.c] + (1 - [omega])[y.sup.nc]. Using Equations A6 and A7 yields aggregate output as a function of the economy-wide price level and the disturbances:

y = [omega][[alpha]([epsilon] - 1)p + [alpha](m + v) + [epsilon][theta]]/([alpha] + [epsilon] - [alpha][epsilon]) + (1 - [omega])[[alpha][lambda](m + v - p) + [epsilon](1 - [lambda])[theta]]/[lambda]([alpha] + [epsilon] - [alpha][epsilon]) + [epsilon]. (A8)

Differentiating Equation A8 yields an expression for the price sensitivity of aggregate output:

[delta]y/[delta]p = [lambda]([alpha] + [epsilon] - [alpha][epsilon])[alpha]([omega][epsilon] - 1) + [omega][alpha][epsilon]([epsilon] - 1)/[[lambda]([alpha] + [epsilon] - [alpha][epsilon]) + [epsilon]]([alpha] + [epsilon] - [alpha][epsilon]). (A9)

This price sensitivity expression governs the short-run-output-inflation tradeoff and depends on the elasticity of demand. Differentiating Equation A9 with respect to [epsilon] yields

[delta]([delta]y/[delta]p)/[delta][epsilon] = [lambda][alpha]([alpha] + [epsilon] - [alpha][epsilon])(2[epsilon][omega] + {([alpha] + [epsilon] - [alpha][epsilon])(1 - [omega]) + [lambda][1 - [alpha](1 - [omega])]}) + [omega][alpha][[epsilon].sup.2]/[[[lambda]([alpha] + [epsilon] - [alpha][epsilon]) + [epsilon]].sup.2][([alpha] + [epsilon] - [alpha][[epsilon]).sup.2]. (A10)

which is unambiguously positive. When monopolistically competitive firms face more elastic demand schedules, their prices are lower at each level of output demanded, and a given price change induces a greater change in the quantity of output demanded as each firm experiences a movement along the demand schedule that it faces. Thus, an increase in the

elasticity of demand induced by a rise in goods market competition leads to greater employment and output adjustments in response to a given price change, implying that greater goods market competition will yield an improved short-run-output-inflation tradeoff, as reflected by a flattening of the slope of the Phillips curve.

To this point, we have abstracted from the constants in the expressions for labor demand and the full-information wage, which would appear as [epsilon] ln[[alpha] ([epsilon] - 1)] in the numerators of Equation A4 and A5. Taking into account these constants and contemplating a long-run equilibrium in which the log of the money stock is normalized at zero, the v and [theta] shocks are equal to zero, and wages are equal to their full-information values, we have that [p.sub.j] = p = E(p), y = -p in Equation A3, and equilibrium output is

y = [lambda][alpha] ln[[alpha]([epsilon] - 1)/[epsilon]]/[epsilon][1 + [lambda](1 - [alpha])] (A11)

Differentiating Equation All ultimately reveals that

[delta]y/[delta][epsilon] = [[lambda][alpha]y/[[epsilon].sup.2][1 + [lambda](1 - [alpha])]][1/([epsilon] - 1) - ln[[alpha]([epsilon] - 1)/[epsilon]]]. (A12)

Under the maintained assumptions that 0 [less than] [alpha] [less than] 1 and [epsilon] [greater than] 1, the second bracketed term is positive, since [epsilon] - 1 [greater than] 0 and since [alpha]([epsilon] - 1)/[epsilon] [less than] 1, so that ln[[alpha]([epsilon] - 1)/[epsilon]] [less than] 0. As a result, the long-run output effect of an increase in goods market competition, [delta]y/[delta][epsilon], is positive under very plausible parameter values. Because y = -p under a constant, normalized money supply and velocity, [delta]p/[delta][epsilon] = -([delta]y/[delta][epsilon]) [less than] 0. As the demand and marginal revenue schedules of firms become less steeply sloped, firms supply more output and demand more labor at a given price level, implying that prices are lower at each level of output. If measured unemployment rates reflect some voluntary unemployment, then by boosting equilibrium employment, a rise in goods market competition may induce a decline in estimates of the unemployment rate associa ted with nonaccelerating inflation or NAIRU.