What has been driving consolidation of the cellular industry in recent years? Now that traffic volumes are increasing rapidly, the cost of expanding capacity has become a large portion of expenditures for cellular carriers. This paper develops an engineering-economic model of the principal factors that determine both cost and capacity of cellular infrastructure. It then uses analysis of that model to show that there are strong economies of scale when cellular capacity is rapidly expanding, assuming that each carrier makes design and resource decisions that minimize cost for any given capacity. This occurs in part because a carrier with more spectrum benefits more from every new cell tower, and a carrier with more towers benefits more from every new MHz of spectrum. While it is technically possible to expand capacity by increasing either towers or spectrum holdings, we find that the cost-effective approach for carriers is to increase both types of assets at a similar rate, which contradicts the publicly-stated assumptions of some spectrum regulatory agencies. This makes access to spectrum important for all carriers. However, our model shows that large carriers should be willing to pay more for spectrum in auctions and other markets. In the absence of countervailing policies, the big carriers will get bigger, in terms of spectrum holdings, towers, capacity, and ultimately market share. For policymakers, this economy of scale creates a trade-off between two important objectives: reducing the cost of cellular capacity, and increasing competition. This paper derives the Pareto optimal division of spectrum with respect to these two competing objectives, and shows that any Pareto optimal assignment will split the spectrum fairly evenly (although not exactly equally) among competing carriers. (More specifically, in any Pareto optimal division with k competitors, k-1 would have the same amount of spectrum, and the kth would have less.) This is not simply a method of ensuring that there are many competitors; spectrum should be divided fairly evenly regardless of whether the number of competitors is large or small. A large disparity in spectrum holdings among competitors may yield poor results with respect to both objectives, i.e. the lower cost-effectiveness of a larger number of carriers, and the lower competitive pressure of a smaller number of carriers. One effective way to achieve a division of spectrum that is close to Pareto optimal is a spectrum cap, provided that this cap is set at a level consistent with other regulations and policy objectives. We show that in some counties, such as the U.K., this is not the case today. Thus, there is reason to change current policy
