Learning to coordinate: A study in retail gasoline David Byrne 1 Nicolas de Roos 2 1 Department of Economics The University of Melbourne 2 School of Economics The University of Sydney 2 nd ATE Symposium December, 2014 Byrne, de Roos Learning to coordinate 1 / 45
Brand average prices, Perth 2013 Byrne, de Roos Learning to coordinate 2 / 45
Goals 1. Document evolution of pricing strategies extensive and exhaustive sample of prices regular price variation is arguably exogenous to demand and costs sample is rich in market turmoil 2. Test between competing frameworks for repeated interaction unique policy setting provides precise information on timing of play Byrne, de Roos Learning to coordinate 3 / 45
Outline Market and policy setting 1 Evolution of price strategies 2 Cycle calibration Standardisation of price dynamics Price leadership Price wars Edgeworth cycles 3 Conclusion 4 Byrne, de Roos Learning to coordinate Market and policy setting 4 / 45
The retail petrol market in Perth Market structure dominated by four (later two) vertically integrated oil majors entry of Coles and Woolworths in 2004 Fuelwatch retailers upload prices for tomorrow by 2pm prices become public at 2:30pm prices must be fixed for 24 hours from 6am the next morning imposes a discrete time simultaneous move environment provides a census of prices data metro Perth daily station-level retail prices, 2001 - 2013 Perth daily wholesale prices, 2003 - 2013 Byrne, de Roos Learning to coordinate Market and policy setting 5 / 45
Monthly brand average prices Byrne, de Roos Learning to coordinate Market and policy setting 6 / 45
Outline Market and policy setting 1 Evolution of price strategies 2 Cycle calibration Standardisation of price dynamics Price leadership Price wars Edgeworth cycles 3 Conclusion 4 Byrne, de Roos Learning to coordinate Evolution of price strategies 7 / 45
Cycle calibration the retail petrol market is subject to substantial cost variation to maintain retail margins, the cycle must be adjusted calibration mechanisms the pace of undercutting the timing of the cycle the magnitude of the price rise the calibration mechanism has adjusted over Cycle Episodes Byrne, de Roos Learning to coordinate Evolution of price strategies 8 / 45
Cycle calibration: Episode 1 (2001-2003), BP and Caltex Byrne, de Roos Learning to coordinate Evolution of price strategies 9 / 45
Cycle calibration: Episode 2 (2006-2007), BP and Caltex Byrne, de Roos Learning to coordinate Evolution of price strategies 10 / 45
Cycle calibration: Episode 3 (2009-2013), BP and Caltex Byrne, de Roos Learning to coordinate Evolution of price strategies 11 / 45
Standardisation cycle calibration has been coordinated across rival brands we disaggregate the distribution of price changes by brand and time we focus on coordination since 2009 additional examples are in the paper Byrne, de Roos Learning to coordinate Evolution of price strategies 12 / 45
Standardisation: BP, 2009 Byrne, de Roos Learning to coordinate Evolution of price strategies 13 / 45
Standardisation: BP, 2010 Byrne, de Roos Learning to coordinate Evolution of price strategies 14 / 45
Standardisation: BP, 2011 Byrne, de Roos Learning to coordinate Evolution of price strategies 15 / 45
Standardisation: Caltex, 2009 Byrne, de Roos Learning to coordinate Evolution of price strategies 16 / 45
Standardisation: Caltex, 2010 Byrne, de Roos Learning to coordinate Evolution of price strategies 17 / 45
Standardisation: Caltex, 2011 Byrne, de Roos Learning to coordinate Evolution of price strategies 18 / 45
Standardisation: Coles, 2009 Byrne, de Roos Learning to coordinate Evolution of price strategies 19 / 45
Standardisation: Coles, 2010 Byrne, de Roos Learning to coordinate Evolution of price strategies 20 / 45
Standardisation: Woolworths, 2009 Byrne, de Roos Learning to coordinate Evolution of price strategies 21 / 45
Standardisation: Woolworths, 2010 Byrne, de Roos Learning to coordinate Evolution of price strategies 22 / 45
Standardisation: Gull, 2009 Byrne, de Roos Learning to coordinate Evolution of price strategies 23 / 45
Standardisation: Gull, 2010 Byrne, de Roos Learning to coordinate Evolution of price strategies 24 / 45
Standardisation: Gull, 2012 Byrne, de Roos Learning to coordinate Evolution of price strategies 25 / 45
Standardisation: Gull, 2013 Byrne, de Roos Learning to coordinate Evolution of price strategies 26 / 45
Price leadership Several price leadership mechanisms have been employed mixed leadership (Episode 1) BP leadership (Episodes 2 and 3) no leadership (Episode 3) Byrne, de Roos Learning to coordinate Evolution of price strategies 27 / 45
Price leadership: Episode 1 Byrne, de Roos Learning to coordinate Evolution of price strategies 28 / 45
Price leadership: Episode 2 Byrne, de Roos Learning to coordinate Evolution of price strategies 29 / 45
Price leadership: Episode 3 Byrne, de Roos Learning to coordinate Evolution of price strategies 30 / 45
Price wars Cycle disruptions March-September 2004: price war following entry of Coles and Woolworths 2005: cycle break due to Hurricane Katrina 2008: cycle break due to global crude oil price shock Nov. 2009 - Feb. 2010: price war Conflict between BP and Caltex may have played a role in 2009 price war Caltex attempted to reignite cycles almost alone in 2005 BP became sole price leader from 2006 to 2011 Caltex followed BP with a two-day lag during 2009-2010 Byrne, de Roos Learning to coordinate Evolution of price strategies 31 / 45
Price wars: brand average prices, 2009 Byrne, de Roos Learning to coordinate Evolution of price strategies 32 / 45
Price wars: brand average prices, 2009-10 Byrne, de Roos Learning to coordinate Evolution of price strategies 33 / 45
Price wars: brand average prices, 2010 Byrne, de Roos Learning to coordinate Evolution of price strategies 34 / 45
Outline Market and policy setting 1 Evolution of price strategies 2 Cycle calibration Standardisation of price dynamics Price leadership Price wars Edgeworth cycles 3 Conclusion 4 Byrne, de Roos Learning to coordinate Edgeworth cycles 35 / 45
Price commitment (Maskin and Tirole, 1988) Setup (usually) 2 firms homogeneous products price setting alternating moves Markov strategies discrete price grid Equilibria focal price equilibrium Edgeworth cycle equilibrium Example Byrne, de Roos Learning to coordinate Edgeworth cycles 36 / 45
Price commitment vs repeated games Price commitment (Maskin and Tirole, 1988) firms change price infrequently coordination resolved by war of attrition (or a consistent price leader) coordination mechanism unclear Repeated games consistent with simultaneous moves rewards and punishment can aid coordination price dispersion could aid obfuscation (de Roos and Smirnov, 2013) Tests the timing of price changes coordination and price wars Byrne, de Roos Learning to coordinate Edgeworth cycles 37 / 45
A simple timing test Under the price commitment model with many firms, Edgeworth cycles require price commitment for several periods we should observe infrequent price changes Under Fuelwatch we observe the precise timing of price changes test the proportion of periods in which each station adjust prices For each station j , test H j H j 0 : θ 0 = 1 / 2 , 1 : θ 0 > 1 / 2 . (1) Byrne, de Roos Learning to coordinate Edgeworth cycles 38 / 45
A simple timing test Byrne, de Roos Learning to coordinate Edgeworth cycles 39 / 45
A simple timing test Byrne, de Roos Learning to coordinate Edgeworth cycles 40 / 45
Outline Market and policy setting 1 Evolution of price strategies 2 Cycle calibration Standardisation of price dynamics Price leadership Price wars Edgeworth cycles 3 Conclusion 4 Byrne, de Roos Learning to coordinate Conclusion 41 / 45
Conclusions: evolution of pricing strategies Price dynamics regular asymmetric price cycles coordinated by price leadership Evolution qualitative adjustments to price leadership and cycle duration standardisation of price dynamics across rival firms results in highly regimented price cycles with higher retail margins Coordination protracted price wars due to entry, cost shocks, (likely) internal conflict temporary price rises used to signal end of price wars Byrne, de Roos Learning to coordinate Conclusion 42 / 45
Conclusions: theory of Edgeworth cycles Firms adjust prices on most days conflicts with the Maskin and Tirole (1988) price commitment model consistent with standard repeated games settings Coordination the central role of coordination suggests Markov strategies are unlikely Standardisation the standardisation of price cycles suggests a desire for transparency Price dispersion consistent with obfuscation (de Roos and Smirnov, 2013) may achieve both transparency and obfuscation Byrne, de Roos Learning to coordinate Conclusion 43 / 45
Price commitment: an Edgeworth cycle equilibrium p t t Return Byrne, de Roos Learning to coordinate Conclusion 44 / 45
Recommend
More recommend
