Data/Methods: I started web-scraping daily prices and stations locations of 140,000 gasoline retailers from a network of web pages[2]. I have been collecting this information since July 2017, and I have over 30 million price observations in my dataset. These stations are representative of the universe of gasoline retailers in the US.
My spatial model explains the preferences of retailers for the low tax side of a state border as well as discontinuous prices of gasoline close to the border. Together these provides insights into the incidence of taxation on retailers and consumers, and the distortionary impacts of tax notches. The methodology is a refinement of the model presented in Serrato and Zidar (2016), with consumer preferences represented by a Random Utility Model over observable brand choices and unobservable establishment features.
Results/Expected Results: I find that gasoline retailers bunch at state borders on the low tax side. Also, I find fewer gasoline retailers on the high tax side of the borders. Moreover, I find that small state tax differences generate statistically equal mean prices on both the low and the high tax side of the border. However, for substantial tax differences, I find a discrete jump in the gasoline price at the border. Finally, I find that for every cent of the gasoline tax, the consumers at the state border pay, on average, less than one extra cent on gasoline.
[1] See for example Bartik (1985), Coughlin et al. (1991), Holmes (1998), Chirinko and Wilson (2008),Chetty et al. (2009), Chiou and Muehlegger (2014)
[2] The names of the local domains are: www.newyorkgasprices.com, www.losangelesgasprices.com, etc.