Pump prices and oil prices: a tale of two directions.

Introduction

It has been argued that gasoline prices adjust more quickly to crude oil price increases than to price decreases. Empirical results supporting this contention can be found in Asplund et al. (2000), Bacon (1991), Borenstein et al. (1997), and Peltzman (2000). Several explanations are offered as to why gasoline prices may respond asymmetrically to changes in input cost. The two most common explanations alluded to in the literature are based on search theory and oligopolistic coordination theory. Search theory implies that consumers search more for better deals when prices increase than when prices decrease. (1) According to the theory of oligopolistic coordination, when input costs increase, sellers increase the price of the final good because otherwise the sellers' profit margins decrease. On the other hand, when input cost decreases, the sellers' profit margins do not decrease. Thus, a seller might want to sustain high prices until its competitors start cutting their prices. This implies that the sellers might exercise short-term market power and that prices would decrease slowly once such an implicit collusion starts to break down when competitors start cutting prices.

Existence of asymmetric responses to crude oil price changes has important implications, especially when oil prices are volatile. Specifically, it implies that an increase in crude oil price followed by an equivalent decrease in crude oil price would negatively affect consumer welfare. For example, Borenstein et al. (1997) reported that "... if a consumer uses 10 gallons of gasoline per week, a 5 cents per gallon increase in crude oil prices ... costs the consumer $1.30 more over the life of the adjustment than a 5 cents per gallon saves her" (p. 322). This would imply that oil price volatility is costly to consumers even when the average level of oil prices is unchanged.

The asymmetric price response hypothesis seems plausible based on anecdotal consumer observations. However, even with some empirical support for the hypothesis, it is not clear how the magnitude of the response may have changed in recent years. In addition, most of the previous empirical studies have used data prior to 2000. Since early 2000, gasoline prices increased in magnitude as well as in volatility. Furthermore, in the 1980s crude oil prices increased mostly as a response to supply shocks. Recent increases in crude oil prices, however, were driven both by an increase in demand and by supply shocks. Given that the factors giving rise to gasoline price increases and volatility are different from those in the 1980s the examination of price movement asymmetry appears warranted.

This paper is organized into five sections. The next section reviews the literature concerning behavior of gasoline prices and asymmetric response models. The structure of the empirical model of asymmetric price response used in the paper is presented in the section entitled Asymmetric Price Response Model. The following section discusses the data and presents the estimation results. The last section explores findings of the study and their implications.

Literature Review

Borenstein et al. (1997) and Peltzman (2000) presented strong evidence that input prices might have an asymmetric effect on prices. Borenstein et al. (1997) empirically investigated gasoline price changes and found that gasoline prices respond more quickly when crude oil prices increase. The authors argued that the source of asymmetry can be attributed to asymmetric adjustment lags in inventory accumulation or to retailers' short-term market power. Short-term retailer market power implies that prices are sticky downward to input price changes. This creates an asymmetric response to oil price increases and decreases. The empirical models in Borenstein et al. (1997) study were estimated using the data for retail gasoline prices which may have had some limitations when testing the asymmetric price response hypothesis. Specifically, retail gasoline prices used were the average of unleaded regular self-service gasoline prices in 33 U.S. cities east of the Rocky Mountains collected semimonthly. Other than the complications with the retail price data discussed by the authors (Borenstein et al. 1997, p.310), in the light of new findings, additional complications with the data may have biased Borenstein et al.'s findings. That is, if the responses of retail gasoline prices to changes in the price of crude oil depend on geographic markets (i.e., were not uniform across 33 U.S. cities), then averaging of prices across these markets might bias the results of the study. An alternative estimation method that allows for differing price adjustment mechanisms in different markets would be a panel data estimation of the price response model. This estimation method, however, requires city-level data. As discussed in the fourth section of the paper, our study uses the data that avoids difficulties or complications existing in the paper by Borenstein et al. (1997).

Peltzman (2000) studied the effect of a change in input prices by investigating the price changes of 77 consumer and 165 producer goods. He found that prices tend to increase faster but they decrease slower. In addition, more input price volatility implies less asymmetry. Peltzman also found that the immediate price response to a positive change in input cost is two times greater than the price response to a negative change in input cost. The author further argued that an increase in the number of intermediaries results in an increase in price response asymmetry.

Bakytzhanova (2005) investigated the fuel market in Moscow. Her findings confirmed the existence of asymmetric response to input prices. Also, she concluded that the effect of asymmetric response can be amplified due to an increase in market concentration. Borenstein et al. (2004) studied retail gasoline prices in California and suggested that consistently high gasoline price premiums in the state can be attributed both to the existence of market power and to real scarcity.

Asplund et al. (2000) used daily data on gasoline price changes in Sweden from 1980 to 1996. The authors found that gasoline prices gradually shift to long-term equilibrium levels and that prices are sticky downward when input prices decrease in the short-term. In addition, the authors found that gasoline prices respond quicker to exchange rate changes than to spot market crude oil price changes.

Radchenko (2005) analyzed the effect of oil price volatility on the asymmetric response of gasoline prices to crude oil price changes. He found that as volatility of crude oil prices increases, the degree of response asymmetry decreases. Radchenko argued that this finding is not consistent with standard search theory explanation of asymmetric price response but is consistent with oligopolistic coordination theory. He also found that retailers responded faster to decreases in oil prices when price volatility increased.

There are several studies that cast doubt on the existence of asymmetric price response in gasoline markets. For example, Bachmeier and Griffin (2003) studied daily gasoline and crude oil prices but did not find any evidence for asymmetric price response. In addition, Godby et al. (2000) examined Canadian gasoline markets using a threshold regression model. The authors' findings did not lend support to the asymmetric price response hypothesis. These studies concluded that the existence of asymmetric price response might be dependent on the day the prices are observed or on the specifics of the market studied.

Asymmetric Price Response Model

A simple long-term relationship between crude oil prices and gasoline prices can be described by Gas = [[phi].sub.0] + [[phi].sub.1] Oil + [epsilon], where Gas is the price of gasoline, Oil is the price of crude oil, and [epsilon] is an i.i.d, error term normally distributed. (2) Prices here denote either nominal prices or natural logarithm of those prices. Using natural logarithm of prices implies that the retailer profit margins vary proportionally as input prices vary. On the other hand, using the level data is more appropriate when there is a linear relationship between nominal oil and gasoline prices.

The above model denotes a long-term relationship. In the short-term, gasoline prices will adjust over time to input price fluctuations. We denote a change in crude oil and gasoline prices by [DELTA][Oil.sub.t] = [Oil.sub.t] - [Oil.sub.t-1] and [DELTA][Gas.sub.t] = [Gas.sub.t] - [Gas.sub.t-1], respectively. Then, assuming it takes n periods for the adjustment to take place, total adjustment of gasoline prices to an initial change in oil price, [DELTA][Oil.sub.t], is given below:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)

Equation 1 assumes a symmetric response to an oil price increase and decrease. To incorporate asymmetric response, we allow [beta] to vary depending on whether a change in oil price is positive or negative. Let [DELTA][Oil.sup.+.sub.t] max{[DELTA][Oil.sub.t], 0} and [DELTA][Oil.sup.-.sub.t] = min{[DELTA][Oil.sub.t], 0}. A simple asymmetric gasoline price response model to a change in crude oil price can be expressed as:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (2)

where [epsilon] is an i.i.d, error term.

The model represented by Eq. 2 is quite general because no restrictions are placed on parameters [beta]. Thus, the adjustment path can take various forms and does not have to be monotonic. That is, the model does not assume that the prices would increase or decrease systematically. Therefore, it is possible for the gasoline prices to rise and or to fall in the adjustment path.

This empirical model, however, does not take into account seasonal variations and time trends. To allow for measurement of these |'actors, one can add dummy variables for time and week. Let TIME and WEEK denote a time dummy associated with a particular year and a week dummy associated with a particular week, respectively. Then, the modified empirical model is given by:

Print Get the Mag Weekly Updates