| In many Canadian cities, retail gasoline prices appear to cycle, rising by large amounts in one or two days followed by several days of small consecutive price decreases. While many empirical studies examine such markets, certain questions cannot b e properly answered without high frequency, station-specific price data for an entire market.;The second paper uses these same data to examine the implications o f an informal theory of competitive gasoline pricing, as advanced by industry and government. Consistent with this theory, stations do tend to set prices to match (or set a small positive or negative differential with) a small number of other stations, which are not necessarily the closest stations. Also, while retailers frequently respond to price changes within two hours, many take considerably longer to respond than is predicted by the theory. Finally, while price decreases do ripple across the market like falling dominos, increases appear to propagate based more on geographic location and source of price control than proximity to the leaders.;The third paper uses both these data and Guelph price data collected every 12 hours during the same 103 days from OntarioGasPrices.com to examine the sample selection biases that might exist in such Internet price data, as well as their implications for empirical research. It is found that the Internet data tend to accurately identify features of cycles that can be distinguished using company-operated, major brand station prices, while features that require individual independent station data or very high frequency data might not be well-identified.;Thus, the first paper in this thesis uses bi-hourly price data collected for 27 stations in Guelph, Ontario, eight tunes per day for 103 days to examine several basic predictions of the Edgeworth cycle theory. The results are largely consistent with this theory. However, most independent firms do not tend to undercut their rivals' prices, contrary to previous findings. Furthermore, the tuning, sizes and leaders of price increases appear to be very predictable, and a specific pattern of price movements has been detected on days when prices increase. These findings suggest that leading a price increase might not be as risky as one may expect. |
