Exit and Upgrading in Response to Entry: The Case of Gasoline Retailing

Abstract

Spatial competition models have established the importance of localized competition in determining competitive outcomes. However, few empirical studies attempt to determine to what extent actual local market conditions affect strategic decisions. This paper uses data provided by the acquisition of the Vancouver area Super‐Save chain of retail gasoline stations by ARCO to study the role of geographic space in competition, and the spatial response of the major competitors in the market to entry. The possibility of both accommodating and aggressive capacity responses by the major incumbent firms to entry are considered. While the empirical results show that proximity to ARCO increased the probability that a station shuts down, proximity to ARCO can explain only a limited amount of shutdown after ARCO’s entry. There is no evidence that incumbent firms used station locations and capacity changes to respond aggressively to ARCO’s entry with a spatial predation strategy.

Key Words:

• Gasoline Retailing
• Spatial Competition
• Predation
• Accommodation

JEL Classifications:

• L13
• L41
• L81

Notes

1. There have also been spatial studies of localized competition in a characteristics space framework. See, for example, Shaw (1982) and Stavins (1995).

2. Other studies examining the effects of local competition on gasoline pricing include Ning and Haining (2003), Barron et al. (2000, 2004), and Eckert and West (2004). Slade (1992) examined gasoline station pricing along a five‐mile strip in Vancouver, BC. While she was studying localized competition, she was not focused on the role of space and location in determining competitive outcomes.

3. See ‘Low Gasoline Prices in the Greater Vancouver Market are the Result of Vigorous Competition,’ available on the Competition Bureau’s website, http://www.competition.ic.gc.ca.

4. The Conference Board (2001: 46) states that the price war was probably responsible for the closure of several gas stations and the exit of some independent gas station operators from the Vancouver market.

5. The relatively small number of non‐major stations in the market precludes a detailed analysis of the reponse by non‐major stations.

6. A third possibility is that some proportion of the station shutdowns occurs for reasons that have little to do with ARCO’s entry. This might be the case if station closures occur that are not in the neighbourhood of an ARCO station.

7. The economic theory of accommodating versus entry‐deterring responses is discussed in Fudenberg and Tirole (1984) and Tirole (1988, Ch. 8).

8. Von Hohenbalken and West (1984) have found evidence consistent with this strategy being used in the supermarket industry.

9. See the Conference Board of Canada (2001) and the Restrictive Trade Practices Commission (1986: 309).

10. See Octane Magazine, various issues.

11. Station counts for 1988 to 1996 were obtained from Kent Marketing Ltd. Note that because the reported station counts in this section do not include Langley and Maple Ridge, they will be smaller than those reported in the Introduction and later in the paper.

12. The complete text of the press release, ‘ARCO to Enter British Columbia Gasoline Market, Acquire Vancouver Area Super‐Save Gas Outlets’, is available at http://www.bp.com/centres/press/p_r_detail_p.asp?id=440.

13. See Schreiner, John (21 December 1998), ‘Drivers Winning Vancouver Gas War: Market Share Battle: Battle between Chevron and ARCO More than Skirmish,’ National Post. We do not have station‐specific price data for Vancouver gasoline stations for dates just prior to and after the acquisition of Super‐Save stations by ARCO, so we cannot examine the immediate station‐specific price impacts of its entry.

14. See Eckert and West (2004) for a description of the price cycle in Vancouver from July to December 1999. The Conference Board (2001) has also produced a figure, Chart 13, that plots daily retail gasoline prices in Vancouver from January 1998 to December 1999. The chart shows pricing behaviour changing from rigid and uniform from January to August 1998, to price cycles from September 1998 to December 1999.

15. The fact that Hastings (2004) found that stations competing with a Thrifty station had a significant increase in price after the conversion of the station to ARCO, while ARCO’s entry in Vancouver is associated with aggressive price cuts, might be explained by the fact that ARCO was an incumbent firm in San Diego and Los Angeles. It increased its market shares with the acquisition of Thrifty. ARCO’s entry in Vancouver by acquisition of Super‐Save was new entry. It could be viewed as test market entry by incumbents and hence deserving of an aggressive response.

16. See Constantinau, Bruce (16 June 2001), ‘Arco to End Foray into Vancouver Market, Focus on Larger Centers,’ Vancouver Sun; Schreiner (1998), supra note 13.

17. Four of the former ARCO station addresses are now shown as 7‐11 and Mac’s convenience stores, and they may or may not have gasoline pumps on the site.

18. Note that Table 2 contains simply the breakdown of stations that shut down over the sample period according to their 1998 brand names. This table does not take into account new stations or stations that remained open but switched brand name

19. Kent Marketing divides convenience stores into different size categories. In this paper, the largest size category is used.

20. A Pearson test of association between whether a major brand station with no upgrades from 1992 to 1997 shuts down in the sample period and whether the station is within 1km of an ARCO station has a χ² ₍₁₎ test statistic of 5.44. Therefore, we can reject the null hypothesis of no association in favour of the alternative of some association at the 5% significance level.

21. A Pearson test of independence does not allow us to reject the null hypothesis of no association at any reasonable significance level.

22. Several other proximity measures were used, with similar results. Proximity measure considered included whether the station was within 1km, 2km, or 4km of an ARCO station and on the same road, and whether an ARCO station was one of the three closest.

23. The relationship between characteristics and contracts has been established empirically in several papers. Slade (1998), for example, found a Spearman rank correlation coefficient between a station’s contract type and whether the station had service bays of 0.71. A station is more likely to be company operated or a commission dealer if it is a gasbar without service bays.

24. A possibility is that the local competition variables are highly correlated, creating collinearity concerns. However, correlation coefficients among these variables are low, and regressions of each variable on the other competition variables reveal no collinearity problems.

25. As further informal indication that spatial correlation is not a serious concern, linear probability models were estimated, allowing for both spatial correlation in the errors and spatial lags of the dependent variable. Different weighting matrices were used. In none of these models was the spatial correlation or the coefficient on the spatial lag statistically significant at a 5% level.

Additional information

Notes on contributors

Andrew Eckert

We thank the editor H.E. Frech III, two anonymous referees, and participants at the third annual International Industrial Organization Conference in Atlanta, Georgia, for their helpful comments. We also thank Benjamin Atkinson for research assistance and the Social Sciences and Humanities Research Council of Canada for financial support.