Asymmetric price adjustment: cross-industry evidence.

• Author: Gwin, Carl R.

1. Introduction

   A consumer need only go to the gasoline station or the grocery store to ponder the question: Why do retail prices always seem to go up so fast when the price of crude oil or farm products go up, but they are so slow to come down when oil or farm prices go down? This supposed phenomenon is referred to alternatively as asymmetric price adjustment, asymmetric price transmission, and price transmission asymmetry (PTA). The question to consider is as follows: Do firms raise prices faster when their costs go up than they cut prices when their costs go down? The macroeconomic version of this question is the following: Are prices more sticky (slow to change) downward than upward? An assumption of downward price stickiness has been central to Keynesian macroeconomics. Certification of this assumption, and thus support for the ideas of a convex aggregate supply curve and the Phillips curve, has important implications for monetary and fiscal policy. The clear route to examining downward price stickiness is microeconomic price theory. However, price theory from textbook economics cannot easily answer our questions. Profit maximization based upon the optimization criteria that marginal benefit should be equal to marginal cost implies that marginal revenue should adjust to marginal cost immediately, so that price should symmetrically adjust in response to increases or decreases in cost.

   However, evidence has begun to emerge that this standard story may fail to apply in at least some industries. Asymmetric price adjustment has been widely documented in gasoline and agricultural markets in a number of countries. More general studies of an economy are few and limited in scope. In one of the broadest studies currently available, Peltzman (2000) finds some evidence of asymmetric price adjustment in eight combinations of 15 2-digit Standard Industrial Classification (SIC) subsectors. Peltzman's work is constrained by the need to match a measure of industry price to some measure of industry cost. Peltzman uses a producer price index (PPI) for an input that made up a significant portion of the production of an output to proxy cost for the "matching" output PPI or consumer price index (CPI). Peltzman's matching technique limited his study to 15 SIC subsectors characterized by relatively simple and clear production processes.

   This paper employs an actual measure of industry cost to investigate asymmetric price adjustment in a wide cross section of the U.S. economy with the aim of answering two very important questions. First, is asymmetric price adjustment a real phenomenon in the overall U.S. economy? Second, in what sectors of the economy is asymmetric price adjustment most prevalent? Can the answers to these questions help us to understand why asymmetric price adjustment exists? An understanding of this relationship is important because there are significant implications for social welfare. The inflated prices and reduced output, even if temporary in nature, that follow from the failure of the market to adjust prices in a timely fashion result in a deadweight loss to society. Additionally, consumer perceptions that they are being taken advantage of by "greedy" corporations that are quick to raise price but slow to lower price fuel the debate about how well or how badly our market system really works.

   Theoretical work that attempts to explain the asymmetry between changes in prices and changes in costs have been based on menu costs, market power, inventory management, markups over the business cycle, and search costs.

   Ball and Mankiw (1994) build on the menu cost literature by noting that positive trend inflation automatically reduces a firm's relative price between price adjustments. In their model, it then follows that shocks that increase a firm's desired price result in relatively larger adjustments to price than do shocks that decrease the desired price.

   Borenstein, Cameron, and Gilbert hypothesize an explanation for asymmetric price transmission in the gasoline market: "Prices are sticky downward because when input prices fall the old output price offers a natural focal point for oligopolistic sellers" (1997, p. 325). Oligopolistic firms maintain prices unless their market share declines. A decline in market share indicates that some firm has cheated. Firms necessarily raise price when costs go up but delay price cuts until there is some signal that some other firms have cut their price.

   Reagan (1982) develops a model of a monopolist who carries inventory and faces uncertain demand. If actual demand is lower than expected, the firm holds inventory in the expectation that demand will eventually increase. Holding inventory eases downward pressure on price. If actual demand is higher than expected, the firm draws down inventory to zero, and price must rise for the market to clear. According to Borenstein, Cameron, and Gilbert, "Production lags and finite inventories of gasoline imply that negative shocks to the future optimal gasoline consumption path can be accommodated more quickly than positive shocks" (1997, p. 327). Thus, asymmetry between the short-run costs of decreasing versus increasing inventories leads to short-run prices responding more to excess demand than to excess supply.

   An upward-sloping marginal cost curve implies that marginal costs are procyclical; costs increase as output increases during business booms. If markups are countercyclical, then prices are acyclical. Price rigidity results because an increase (decrease) in cost is offset by a decrease (increase) in markup. Markups may be countercyclical due to procyclical elasticity of demand, as in Blinder et al. (1998), or if price wars in oligopoly are more frequent during booms, as suggested by Rotemberg and Saloner (1986). Rotemberg and Saloner (1986) establish their result with the idea that the gain from cheating (deviating from the collusive price) to a firm in an oligopoly is the largest when demand is high during a boom. Haltiwanger and Harrington (1991) extend the research of Rotemberg and Saloner (1986) by considering a firm's expectation of future demand. They note that the discounted loss from cheating is lowest when demand is expected to fall in a recession, because a recession period is the toughest time for coordination in an oligopoly. Therefore, markups are procyclical instead of countercyclical. This can lead to asymmetric price adjustment.

   Borenstein, Cameron, and Gilbert provide a third and final alternative explanation of asymmetric price adjustment in the gasoline market. Specifically, "Volatile crude oil prices create a signal-extraction problem for consumers that lowers the expected payoff from search and makes retail outlets less competitive" (1997, p. 329). The temporary increase in the market power of the gasoline retailer accelerates the transmission of increases in crude oil prices while moderating the transmission of decreases.

   Empirical work on asymmetric price adjustment is relatively abundant but generally focuses on specific subsectors of the economy. Much of the literature examines gasoline and agricultural industries, but there are some papers with a slightly broader focus. Frey and Manera (2007) provided a rather extensive survey of empirical models and findings in their study of asymmetric price transmission.

   Evidence of asymmetric price adjustment at a more general level is found by Buckle and Carlson (2000) and the aforementioned Peltzman (2000). Buckle and Carlson (2000) test the prediction of Ball and Mankiw (1994) that asymmetric price adjustment should be more pronounced with increases in general inflation. Using an ordered probit model to analyze data from a survey of firms in New Zealand soliciting information about the direction of changes in price and cost, Buckle and Carlson find evidence that inflation does lead to added asymmetric price adjustment. The paper by Peltzman appears to be the only paper that has studied asymmetric price adjustment using actual measures of industry prices and goods across a spectrum of the economy. Peltzman also tests whether inventory management, menu costs, or market power could explain asymmetric price adjustment and found no evidence supporting any of these explanations.

   Section 2 presents the data and methodology, section 3 presents the results, and section 4 concludes.

2. Data and Methodology

   The Data

   The ideal dependent variable would be the actual price of each firm's product, and the ideal independent variable would be the product's true cost. Time series of individual product prices are not available, so this paper follows Peltzman (2000) by using the PPI from the U.S. Bureau of Labor Statistics (BLS) as a proxy for industry price. Industry cost is constructed from costs available from Standard & Poor's (S&P) Compustat financial information database. Adequate coverage of quarterly revenue (R) and cost of goods sold (VC) for the period 1st quarter 1966 to 4th quarter 2006 is available from the Compustat database for over 10,000 publicly traded U.S. firms. In 2006, financial data were collected on 841 6-digit North American Industrial Classification System (NAICS) industries. The data "cost of goods sold" is defined by S&P as 'This item represents all expenses that are directly related to the cost of merchandise purchased or the cost of goods manufactured that are withdrawn from finished goods inventory and sold to customers." This data definition fits well with the economics characterization of variable cost.

   The lack of availability of data for privately held companies is not a significant limitation for many of the 24 two-digit NAICS sectors of the U.S. economy classified by the U.S. Census Bureau. A comparison of 2002 sales from Compustat to the 2002 Economic Census in Table 1 shows that most, if not all, sales are accounted for by publicly held companies in Mining (sector 21), Utilities (22), Manufacturing (31-33), Retail Trade (45)...