Asymmetric price adjustments in US gasoline markets: impacts of spatial dependence on the ‘rockets and feathers’ hypothesis

ABSTRACT

Gasoline retail prices show sometimes wild, asymmetric fluctuations over time. We explore the impact of spatial dependence on gasoline retail price formation by using for the first time an asymmetric spatial error correction model (ASpECM). We find evidence that the generally assumed symmetric price pattern is fully reversed when we account for spatial spillover effects, indicating that retail prices adjust more rapidly in an upward than in a downward direction. This finding suggests that empirical studies that ignore the role of spatial dependence and local competition may miss an important element of the nature of the gasoline price adjustment mechanism.

KEYWORDS:

• asymmetric spatial error correction model (ASpECM)
• asymmetric gasoline price adjustment
• localized competition
• spatial dependence

JEL:

• C23
• L13
• Q2
• Q41

ACKNOWLEDGEMENTS

The authors thank the Associate Editor and two anonymous reviewers for their constructive comments and suggestions, which substantially enhanced the merit of the paper. They also thank George Deltas, Thanasis Stengos and Nicholas Apergis for their fruitful comments and suggestions on an earlier version of the paper. The authors also thank the Organizational Committee of the 3rd Conference of the Society for Economic Measurement (SEM), which was held in Thessaloniki, Greece, 6–8 July 2016. Special acknowledgements should be given to seminar participants of the conference and Professor Stephen Spear for their constructive comments. All remaining errors are the authors’ alone. The usual disclaimer applies.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

ORCID

Konstantinos Eleftheriou http://orcid.org/0000-0003-4423-8451

Michael L. Polemis http://orcid.org/0000-0003-2223-0547

Notes

1. Anas and Hiramatsu (2012) develop a spatial computable general equilibrium model (SCGEM) to investigate how certain gasoline demand drivers respond to a gas price increase in Chicago city.

2. Moreover, while the use of OLS is not justified in a non-dynamic spatial setting (Anselin, 1988; LeSage & Pace, 2009), it seems to be supported in the case of non-stationary spatial panel data (Beenstock & Felsenstein, 2010, and the literature cited therein).

3. It is clear from the extensive literature on spatial econometrics that spatial dependence leads to biased inference, unless it is explicitly accounted for in the empirical specification (e.g., Elhorst, 2014).

4. The standard Im–Pesaran–Shin (IPS) test (Im, Pesaran, & Shin, 2003) is not robust to the existence of cross-sectional dependence. Moreover, the use of the cross-sectionally augmented Im–Pesaran–Shin (CIPS) and truncated CIPS tests (Pesaran, 2007), which account for cross-sectional dependence, is not suitable due to the nature of the data set (N = 7, T = 1457).

5. The results are sensitive to the selection of the lag structure of the model. Persyn and Westerlund (2008) point out that this sensitivity might occur in small data sets.

6. We have also estimated the two models using time dummies to control for seasonal effects. The estimation results are qualitatively similar in both cases.

7. The authors thank an anonymous referee for suggesting this robustness test.

8. The term ${\delta }_0^{\ast -}$ measures the impact of the spatially weighted average decrease in the wholesale price of the neighbouring stations on the change in the retail price of the gasoline station under consideration. The term ${\delta }_0^{\ast +}$ can be defined in a similar way.

9. Since wholesale prices are expressed in rack prices, their decrease may be due to the price discriminative behaviour of the upstream supplier (Varian, 1989).

10. Non-searchers can be thought as brand/gasoline station-loyal customers (Borenstein, 1991).

11. On the contrary, Lewis and Marvel (2011) and Lewis (2011) argue that search intensity decreases when prices are falling. However, Cabral and Gilbukh (2015) claim that these models fit better to markets of ongoing services (cable television, business to business) rather than repeated ‘anonymous’ purchases markets (gasoline).

12. The analysis for positive cost shocks follows a similar rationale.

13. This finding is also evident in other empirical studies (Polemis, 2012; Grasso & Manera, 2007).