Essays in Empirical Industrial Organization

Abstract

This dissertation studies pricing behavior in the retail gasoline market. The main research questions are: How does traffic congestion affect market power of gas stations? How does traffic congestion affect equilibrium price dispersion between gas stations? Is gasoline price cycle consistent with Edgeworth Cycle in terms of how their shapes respond to aggregate demand elasticity? These questions are explored in three separate chapters respectively using unique datasets comprised of station-level gasoline price data and direction-specific road-level traffic data from metropolitan Sydney and the rest of New South Wales.

Evidence from chapter 2 suggests that traffic congestion, through its impact on spatial friction for consumers, dampens the intensity of price competition between gas stations. By exploiting a panel of 61 gasoline stations on 13 roads in Metropolitan Sydney, it is found that the margins of regular gasoline increased with travel delay in local traffic. Specifically, retail margins of regular gasoline increased by 0.32 cents per liter (4%) when travel delay in local traffic has increased unexpectedly by 1 minute per kilometer. Unique to this paper is the high-frequency nature of its data which allows me to examine how fast gasoline companies are responding to spatial frictions at the hourly resolution. Analysis based on a dynamic model suggests that this response is “instantaneous” as margins rise as early as the same hour when a shock in traffic congestion is observed.

In chapter 3, traffic congestion is found to have an impact on equilibrium price dispersion between gas stations. Motivated by empirical evidence that the majority of consumers search for cheaper fuel while they drive, I exploit variation in traffic delay to identify the effect of consumer search cost on price dispersion in the equilibrium retail gasoline market. I find that the relationship between price dispersion of regular gasoline and search cost is indeed non-monotonic (inverse U-shaped). This finding is consistent with the consumer search model by a consumer search model presented in the chapter which predicts no dispersion at the extremes: market prices converge to marginal cost when search cost approaches zero and to the monopoly price when search cost approaches the upper bound. I find that at the sample average level of traffic delay (0.387 Min/KM) in New South Wales, pricing for regular gasoline is more consistent with competitive pricing, but becomes more monopolistic once traffic delay rises above 1.20 Min/KM.

Finally in chapter 4, I establish new empirical evidence which suggests that gasoline price cycles are consistent with Edgeworth Cycles. Using daily station-level price data for regular gasoline over 2 years, I find that a higher price than a reference-price at the start of a gasoline price cycle has an positive effect on cycle length and a negative effect on undercutting aggressiveness. Based on established empirical evidence that gasoline demand is reference-dependent and the predicted pricing response to aggregate demand elasticity under the Edgeworth Cycle equilibrium, I infer from these results that the shape of a gasoline price cycle depends on aggregate demand elasticity the same way the shape of a Edgeworth Cycle does.

Insights from this dissertation can inform public agencies who are concerned with addressing the issue of traffic congestion and regulators who are concerned with competition and pricing behavior in the retail gasoline industry.