Rockets and Feathers Revisited: Asymmetric Retail Gasoline Pricing in the Era of Market Transparency.

• Author: Asane-Otoo, Emmanuel

1. INTRODUCTION

   This paper revisits the debate on the asymmetric response of retail gasoline prices to crude oil price changes. Retail fuel pricing, in general, remains an area of significant interest for motorists, the media, and regulatory authorities in many countries. This is partly due to the widespread and persistent public perception that oil companies are quick to adjust retail prices and profit margins in response to input cost increases rather than decreases--a behavior characterized as the rockets and feathers phenomenon (Bacon, 1991). This pricing pattern leads to consumers' welfare losses since they do not benefit from price changes possible under symmetric adjustment conditions.

   Consequently, the retail segment of the fuel market, in particular, has been the subject of regulatory and antitrust scrutiny in many countries, in some cases resulting in charges, convictions, and hefty fines. (1) For Germany, the Federal Cartel Office (FCO) conducted an inquiry in 2008 in response to consumer concerns and found a dominant oligopoly that consists of five firms-BP (Aral), ConocoPhillips (Jet), ExxonMobil (Esso), Shell, and Total. The oligopoly possesses not only a nationwide network of stations but also has significant access to refinery capacity that further amplifies their collective dominance and market power (German Federal Cartel Office, 2011). Among others, this finding instigated the implementation of a price transparency regulation in 2013 that permits consumers to access real-time station-level prices.

   Besides the public interest and scrutiny by antitrust agencies, asymmetric retail pricing has also been the subject of intensive research (see Eckert, 2013; Periguero-Garia, 2013). However, the empirical evidence is inconclusive. The diverse findings can partly be attributed to temporal and spatial aggregation of price data (Bachmeier and Griffin, 2003; Faber, 2015). On the one hand, an aggregation that yields low-frequency price data neither adequately reflects short-run input cost changes nor the frequency of price decisions at the station level. Since station-level prices change several times within a given day, the frequency of price adjustment to input cost shocks or daily price volatilities cannot be detected using weekly or monthly data. On the other hand, an aggregation across stations that results in average prices at either the national, regional, or city level ignores station-specific characteristics and obvious heterogeneity, such as differences in pricing strategy, local competition, and demographic features of the market. These forms of aggregation could compromise the validity of estimations since time series of heterogeneous stations might exhibit dynamics that differ distinctly from cross-sectionally aggregated time-series data (e.g., Granger, 1980; Pesaran and Smith, 1995; Pesaran and Chudik, 2014).

   This paper examines asymmetric cost pass-through in the retail gasoline market using pooled-panel asymmetric error correction models based on a unique panel of daily E5 gasoline (i.e., with 5% bioethanol) price data that spans the vast majority of stations in Germany. The panel data from January 1, 2014, to December 31, 2018, for 12,804 stations allow us to include geographically diverse stations in rural and urban areas. Therefore, our analysis provides a complete representation of retail market competition and goes beyond the representative agent assumption implicit in most empirical studies. The spatial and temporal dimensions of the data also permit us to investigate the effect of aggregation on the type of asymmetry. Moreover, the analysis accounts for the price effects of spatial or local competition and demand-side fluctuations caused by the occurrence of holidays and changes in local weather conditions.

   Our analysis points to a pervasive rockets and feathers pattern in the German gasoline market--both across brands and stations in areas of different population density. We also find that temporal data aggregation matters, as this form of aggregation yields inaccurate inferences. The results suggest that it is insufficient to abstract from determinants other than crude oil prices--as done in previous studies--when examining the rockets and feathers pattern. Given that other demand-side factors and local weather conditions are inherent in the pricing decisions, a precise model specification that accounts for these drivers is required to explain the observed price changes.

   The remainder of the paper is organized as follows. Section 2 offers a concise summary of the recent literature on the retail gasoline market in Germany, while section 3 provides a brief description of the retail market and the data used for the analysis. Section 4 outlines the empirical strategy, section 5 presents the results and discussion, and section 6 concludes.

2. RELATED LITERATURE

   Competition and price-setting behavior in the retail fuel market have been the focus of extensive empirical research. The literature on this subject can be categorized broadly into those that examine the determinants of retail prices, on the one hand, and those investigating price dynamics, on the other hand (see Eckert, 2013, for a comprehensive review of the different strands of the empirical literature). Empirical literature that focuses on price determinants for Germany includes those that evaluate the impact of regulatory changes on retail prices (Dewenter et al., 2017; Eibelshauser and Wilhelm, 2017), the impact of station-specific characteristics (Haucap et al., 2017), duopoly price rivalry (LeSage et al., 2017), differential cost pass-through among major brands (Kihm et al., 2016), and the impact of local market structure (Haucap et al., 2016).

   A strand of the price dynamics literature explores different aspects of recurring Edgeworth cycles (Maskin and Tirole, 1988). For Germany, Siekmann (2017) draws on a large-scale data set similar to ours to investigate recurring Edgeworth cycles and finds evidence of intra-day cycles across municipalities with cycle asymmetry and intensity more pronounced in concentrated markets. Other recent studies that employ station-level price data to investigate asymmetric price cycles in the German context include Eibelshauser and Wilhelm (2018), Wilhelm (2019), and de Haas (2019).

   Another type of price dynamics that has received considerable attention in the literature is the asymmetric response of retail prices to upstream price changes. Regarding the underlying causes, market power exploitation or tacit collusion among retailers has gained traction as a plausible driver of asymmetric pass-through of input cost changes to retail prices. Earlier studies in this research line, such as Borenstein et al. (1997), motivate the rockets and feathers pattern with a stylized version of the trigger price model of oligopolistic coordination (Green and Porter, 1984). (2) Although rigorous theory underlying tacit collusion as a profit-maximizing strategy for retailers is limited, empirical evidence lends credence to this hypothesis (Verlinda, 2008; Lewis, 2011).

   However, in a perfectly competitive market, firms earn zero profits, and input cost changes are transmitted to consumers symmetrically. This market outcome changes when consumers have imperfect information about prices, and a significant proportion of consumers have positive search costs. In this case, firms can extract information rent from consumers. A general result from the theoretical search-based literature is that firms' asymmetric price response to input cost changes emerges naturally due to consumer search behavior. However, the models offer different mechanisms through which consumers' search efforts relate to asymmetric pricing by firms.

   For example, Lewis (2011) argues that for consumers with adaptive expectations of prices, rising input costs reduce the expected price distribution and cause consumers to intensify search more than they otherwise would. In essence, consumers search more actively when prices are increasing than decreasing, resulting in asymmetric price response by firms. This assertion is consistent with Cabral and Gilbukh (2020) finding that consumers search more when prices are high or increasing. Empirical evidence by Hastings and Shapiro (2013) has validated these results by showing the increased sensitivity of consumers to price increases.

   In contrast, Yang and Ye (2008) and Tappata (2009) suggest that consumers' searching activities can result in asymmetric pricing if the incentive to search for better prices is high when input costs are low. Although consumers have imperfect knowledge of firms' input costs, they learn whether the input costs are high or low through search activities and purchasing decisions. Accordingly, retail prices are expected to be more dispersed at low input costs, and consumers with positive search costs anticipate higher gains from increased search activities. However, at high input costs, price dispersion decreases as firms have less flexibility in setting prices, and the benefit from search and search intensity declines. At this level, if an unexpected negative cost shock occurs, firms may have less incentive to adjust retail prices to reflect the cost changes due to the reduced search intensity. In this case, asymmetric search intensity leads to consumers being less knowledgeable about input cost decreases and enables firms to extract information rent in the short-run.

   Despite the increased traction of market power and search-based theories, the empirical findings are inconclusive (Periguero-Garia, 2013). For the German retail gasoline market, studies based on weekly (e.g., Kristoufek and Lunackova, 2015; Asane-Otoo and Schneider, 2015) and monthly data (e.g., Bagnai and Ospica, 2016) follow a similar trend in terms of the findings being inconclusive. Frondel et al. (2020), however, employs daily E10 gasoline (i.e., with 10% bioethanol) prices from January 2014 to...