Do prices respond asymmetrically to cost changes?

Abstract

This article investigates price asymmetry by using an almost complete sample of Taiwanese industries to see whether output prices respond faster to cost increases than they do to decreases. In contrast with the earlier literature, the evidence shows that positive price asymmetry cannot be applied to all industries, especially those industries with rapid productivity growth. The evidence also indicates that firms exhibiting fast (slow) productivity growth see their prices react less (more) to cost increases than to cost decreases.

Notes

¹ See Bacon (1991), Reilly and Witt (1998) and Borenstein et al. (1997).

² See Kinnucan and Forker (1987), Hann (1990), Bernard and Willett (1996), Aguir and Connor (1997) and Aguir and Santana (2002).

³ See Neumark and Sharpe (1992) and Jackson (1997).

⁴ See Peltzman (2000).

⁵ See Cubbin and Geroski (1987), Goddard and Wilson (1999) and Bentzen et al. (2005).

⁶ The consumer price increases 1.80% annually.

⁷ For example, Borenstein et al. (1997) analyse the responses of gasoline prices to oil price changes. Peltzman (2000) selects only those products in which a single input accounts for over 20% of the output's value, using the IPI for that input to explain the OPI.

⁸ See Peltzman (2000, p. 477) for more details.

⁹As mentioned previously, a stable long-run relationship is required to estimate the ECM. The Augmented Dickey–Fuller (ADF) unit-root tests on P and C indicate that both series are I(1). Moreover, the hypothesis that the residuals of Equation 1 are not stationary is rejected. Thus, P and C are cointegrated and Equation 1 represents a stable long-run relationship.

¹⁰ In his sample, 29 out of 32 industries exhibit positive price asymmetry.

¹¹ Please refer to Peltzman (2000, p. 490) for the calculation of PERSISTENCE.

¹² See Martin (1993, p. 532).

¹³ In the L–R model, some conditioning variables used to explain economic growth are fixed and used in each regression. However, in the literature of price asymmetry the number of variables that are correlated with the price asymmetry is much less than the number in the growth literature such that I have to relax the test procedure by no longer applying the set of fixed variables. (Actually, I have only seven potentially explanatory variables proposed by the earlier literature.)

¹⁴ The subset of conditioning variables is taken from the full set of seven potentially explanatory variables except the variable of interest under consideration. Thus, in each regression I have one variable of interest and three conditioning variables. Since the number of conditioning variables might be arbitrary, I have experimented with including two or four conditioning variables. However, this hardly changes the empirical evidence.

¹⁵ Note that each conditioning set of variables corresponds to a conditional mean of β.

¹⁶ The mean value of the coefficients on PRODUCTIVITY is −8.01, and the standard variation is 0.85.

¹⁷ Note that the dependent variable is positive price asymmetry.

¹⁸ See White (1980).

¹⁹ Labour productivity growth can be decomposed into two components, changes in the capital-labour ratio (also called capital deepening) and changes in multifactor productivity, which captures the effect of a number of factors including technical change, measurement error, economies of scale and capacity utilization.

²⁰ See Hitt and Snir (1999).

²¹ The only indicator that is not significantly different from zero is ECOMMERCE. In a way, this is not too surprising since ECOMMERCE is likely to be the poorest indicator to proxy for productivity growth. The reason for this is that using e-commerce technology to market products is less relevant to the production process and a firm's pricing policy.