Reversing the Keynesian Asymmetry

Abstract

The assumption that nominal price adjustment is costly for firms (there are “menu costs”) has generated a stream of important theoretical papers over the last decade or so. Insofar as this literature generates asymmetric adjustments, it provides a theoretical underpinning for the (old) Keynesian assumption that nominal prices are more flexible upward than downward. Yet, the empirical evidence, while confirming that asymmetries exist, does not indicate the dominance of any particular form of asymmetry (see Dennis W. Carlton, 1986; Alan S. Blinder, 1991). In this paper we argue that the gap between theory and practice may be the result of the focus of menu-cost models on specific forms of market structure. Existing menu-cost models are based on the assumption of relatively uncompetitive market structures— monopoly, oligopoly, or monopolistic competition with a fixed number of firms. We widen the scope of the analysis by examining what we call a quasi-competitive industry and demonstrate that it displays a pattern of adjustment quite different from that found in other models. The Keynesian asymmetry is reversed, with nominal price being more flexible downward than upward. We suggest therefore that a relationship exists between market structure and the pattern of nominal price adjustment. Since there is presumably a variety of market structures, this may help explain the inconclusive empirical evidence. We model the most competitive market configuration compatible with menu costs: Bertrand oligopoly in a dynamic setting with free entry. It is assumed that (a) an incumbent in one period can continue to sell at its existing nominal price in the next period without incurring any additional menu cost, whereas an entrant would have to incur a menu cost; and (b) among the firms willing to sell at the lowest price in any given period, one is chosen randomly to sell the product. These simplifications enable us to abstract from matters—such as determining the identity of active firms—extraneous to our main concern of establishing a clean connection between market structure and the pattern of price (in)flexibility. To set a benchmark and to obtain a simple solution by backward induction, we begin by assuming a two-period time horizon. Then we extend the analysis to the case where the incumbent faces an ever-recurring threat of entry, that is, with an infinite horizon. This is the scenario we call “quasi-competitive.” Comparing these two extreme cases yields an intuitively appealing relationship between competitiveness and the pattern of nominal price adjustment. * Bennett: Department of Economics and Finance, Brunel University, Uxbridge, Middlesex, United Kingdom, UB8 3PH; La Manna: Department of Economics, University of St. Andrews, St. Andrews, Scotland, United Kingdom, KY16 9AL. This work developed out of earlier discussions with Subhashish Gupta. We are also grateful to V. Bhaskar, Huw Dixon, Elisabetta Iossa, Jonathan Thomas, and anonymous referees for very helpful comments. 1 For surveys, see N. Gregory Mankiw and David Romer (1991), Torben M. Andersen (1994), and Huw David Dixon and Neil Rankin (1994). 2 Models providing some support for the Keynesian asymmetry include Daniel Tsiddon (1993) and Laurence Ball and Mankiw (1994). However, Robert J. Barro (1972) and Mankiw (1985), among others, produce two-directional stickiness. The Tsiddon and Ball-Mankiw models are based on the assumption of positive trend inflation, whereby a firm that wishes to reduce its relative price finds that it can do so costlessly merely by keeping its nominal price unchanged. Conversely, a firm that wishes to raise its relative price finds that inflation widens the gap between its desired and actual nominal price, thereby providing a strong incentive to incur a menu cost and raise its nominal price. 3 A similar pattern of greater downward flexibility is found in kinked-demand-curve models, but there it occurs essentially by assumption; see, for example, Jean Tirole (1988 pp. 243–44). 4 A Bertrand duopoly model that yields the Keynesian asymmetry can be found in Per Svejstrup Hansen et al. (1996). Although their model and ours are not directly comparable (insofar as they consider real shocks and produce asymmetric price adjustments in the absence of menu costs), we conjecture that the opposite asymmetry generated by our model is due to our key assumption of free entry. 5 To establish the basic message of the paper it is not necessary to examine the more complicated case of a Tperiod model, where ` . T . 2.