I identify two significant factors that inhibit the Internet retailers’ freedom to locate in the U.S.: sales taxes and transportation cost. The role of transportation costs in retailers’ location choices has been widely emphasized in the economics literature. Even in the world of the Internet, it is still true that most categories of retail goods must be transported to consumers physically. We, therefore, cannot overlook the importance of transportation costs in Internet retailers’ location decisions. Sales taxes are also a factor in determining the Internet retailers’ locations. It is common for Internet retailers not to collect sales tax for out-of-state purchases and that has led to a debate over taxes and the Internet, especially among state policymakers. The loophole in sales tax collection works as a crucial factor in consumers’ purchase decisions. Goolsbee (2000) investigates how the sales tax issue influences consumers’ purchase decisions. He finds that, controlling for observable characteristics, people living in high sales tax locations are significantly more likely to buy online. In contrast, I consider the loophole in tax collection as an important factor for Internet retailers’ decisions as well as for consumers’. My study explores Internet retailers’ location decisions, by providing and testing a model where the sales tax loophole in conjunction with transportation cost plays a key role in location choices.

 I develop a simple model of one retailer and two states to illustrate the joint role of sales taxes and transportation costs in a model of locational choice. For simplicity, the two states are assumed to be points; the Internet retailer’s strategy is to choose one of them. The model includes such parameters as demand size, wholesale price, transportation cost, and sales tax rates. Given the loophole in sales tax collection and the assumptions above, consumers making a purchase from a retailer located in their own state would have to pay sales tax but no transportation cost, while consumers from the other state would not have to pay sales tax but do pay transportation cost.

Two specific cases are considered under the model. In the first case, controlling for sales tax rates, I assume that the demand sizes of those states are different. Under the assumption, two variables are found to be the determinants of location choice: the state sales tax rate and the ratio of transportation cost to wholesale price. For instance, when the sales tax rate is greater than the ratio of transportation cost to wholesale price, locating in the state with lower demand is optimal for the Internet retailer. Conversely, when the sales tax rate is less than the ratio of transportation cost to wholesale price, the Internet retailer’s optimal strategy is to choose a state with higher demand. In the second case, controlling for demand conditions, I deal with two states with different sales tax rates. In this case, locating in the state with lower sales tax rate turns out to be optimal for the Internet retailer. The outcome in this case is consistent with the finding by Goolsbee (2000) in that the Internet retailer’s strategy is to attract consumers living in the state with the higher sales tax rate. My model sheds insights into the Internet retailers’ optimal location strategy in the presence of the sales tax loophole.

For the empirical analysis of location choice, I apply a conditional logit model to retailer location data classified by product. The data also include state characteristics such as population, per capita income, land area, a proxy variable measuring average transportation cost, the sales tax rate, and a California dummy. Among the retail product categories investigated are apparel, home and garden, books, electronics, and computers.

In the first part of the empirical analysis,  I study computer retailers, comparing the location decisions of Internet retailers with those of local retailers. The comparative analysis uses the data on the locations of local computer superstores and computer Internet retailers. The empirical results confirm our intuition that the location pattern of Internet retailers is different from that of local retailers.

The second part of the empirical analysis compares the determinants of Internet retailer location for each product category. The results show that the determinants of Internet retailer location vary substantially across product categories. As my theoretical model suggests, the coefficients of demand size variables are significant in most product categories. The coefficient of the sales tax rate is significantly negative in some product categories although insignificant in the majority of product categories. The coefficients for transportation cost are positively significant or insignificant, against our intuition that the Internet retailers would find a location where they can lower transportation cost. Bipolar clustering of Internet retailers into east coast and west coast seems to be closely related to this result. I attempt to explain the seemingly counterintuitive result on transportation costs, borrowing some results from game theory.

 The geographic concentration of firms has been an economic phenomenon not accounted for adequately. I use land area and the California dummy to investigate the geographic concentration of the Internet retailers in the U.S. Land area is commonly used to examine the dartboard theory of industrial location. Bartik (1985), for example, found that the land elasticity of new branch plants in manufacturing industry was approximately one. Upon using a population variable instead of the land area variable, I find some evidence supportive of the dartboard theory of random locations. The California dummy is significant and the probability of new births of computer Internet retailers in California is estimated to be 0.32, successfully capturing geographic concentration of computer Internet retailers and co-agglomeration of those with the computer industry in Silicon Valley.

In work currently in progress, I study the possible effects of Internet sales taxes on state tax authorities, consumers, Internet retailers, and local retailers. The magnitudes in the paper by Goolsbee (2000) suggest that applying sales taxes to Internet commerce might reduce the number of online buyers up to 24 percent. The effect of the imposition of sales taxes will be differential to consumers across states and should be closely related to the locational pattern of Internet retailers and the degree of possible relocations after the taxation. In a structural framework, I’m estimating the consequences of taxation on consumer welfare and retailer profits using detailed consumer and retailer level data.