Price response, information, and asymmetry of price dispersion

ABSTRACT

In this study, we examine how differently gasoline prices in 25 regions of Seoul, Korea respond to asymmetric information between retailers and consumers. We estimate the region-specific likelihood that retailers engage in price undercutting under asymmetric information and investigate inter-regional differences. We find that in response to increases in wholesale price, regions with a high likelihood of price undercutting experience intensified gas station price competition while dispersions of price and markups tend to decrease more in response to cost shocks. Understanding the geographical dispersion of retailers’ price responses to information frictions and search intensity is crucial to lowering information barriers across regions and redistributing profit among market participants.

KEYWORDS:

• Retail gasoline price dispersion
• Information asymmetry
• Consumer search
• Price Competition

JEL CLASSIFICATION:

• L13
• L41
• L81

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here.

Notes

¹ When costs rise, isolated gas stations are constrained by the fact that customers do not yet know the cost has increased.

² Farrel and Klemperer (2007), Dube, Hitsch, and Rossi (2009) stated that switching costs are either anti-competitive or pre-competitive. First, they increase the market power of sellers with locked-in customers, which in turn, leads firms to increase prices. Second, they increase competition for new customers, which leads firms to decrease prices. Cabral (2016) stated that if markets are very competitive, then switching costs make them even more competitive, whereas if markets are not very competitive, then switching costs makes them even less competitive.

³ Chandra and Tappata (2011) defined the rank reversals between stations i and j as the proportion of days to total days in which $p_{jt}>p_{it}$. Instead, we calculate the likelihood of retailers’ undercutting cross-station prices as the proportion of days to total days in which the price of gas station $\mathrm{i}$ is lower than that of competing station $\mathrm{j}$ in the region $\mathrm{k}$ under the condition that seven days prior, station i’s price was greater than or equal to that of station j.

⁴ The gasoline tax rate had been fixed for 10 years before 14 October 2018.

⁵ The previous version, Choi (2016), estimated Beta across regions controlling for brand and region fixed effects only.

⁶ In the extant literature, there are examples of using estimated beta-coefficients for next stage regression analysis. For example, Perez-Gonzales and Yun (2013) use the estimated beta as an instrumental variable for second stage regression. They measure an energy company’s risk exposure to weather change (beta) as an instrumental variable for the energy company’s hedging decision.

⁷ We measure the distance between gas stations once in 2012 by using the path finding algorithm offered by the Naver portal website (https://map.naver.com/). The actual distance is approximated by using the Euclidean metric. We chose the gas stations based on samples from that time; the annual turnover rate of gas stations by new entry and exit is about 5%.

⁸ On 22 July 2018, 1 dollar is exchanged into 1,135KRW. 1 gallon = 3.785 liter.

⁹ Studies showing similar outcomes to Newman and Kenworthy (1999) include Handy (1996), Kahn (2000), Giuliano and Dhiraj (2003), Bento et al. (2005), and Grazi, van Den Bergh, and van Ommeren (2008).

¹⁰ The analysis provides a robustness check for second stage analysis. Surprisingly, the estimated search cost is less correlated with beta because the two measures are differently constructed. The beta is computed by imposing different weights on all possible pairs of retailers’ price rankings, depending on the distance between retailers while the search cost is estimated by comparing retailer prices sequentially.

¹¹ Consistent with the estimation of Beta, we estimated search costs across regions while Yilmazkuday (2017c) estimated search costs across zip codes in the U.S. Yilmazkuday (2017c) investigated the relationship between expected numbers of searches and regional characteristics.