The Economics of Quality Investment in Mobile Telecommunications
Total Page：16
File Type：pdf, Size：1020Kb
The Economics of Quality Investment in Mobile Telecommunications Patrick Kainin Sun Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Graduate School of Arts and Sciences COLUMBIA UNIVERSITY 2015 © 2015 Patrick Kainin Sun All Rights Reserved ABSTRACT The Economics of Quality Investment in Mobile Telecommunications Patrick Kainin Sun This dissertation studies the U.S. mobile telecommunications industry, with a particular emphasis on the incentive to maintain antenna facilities, or base stations, to produce better signal quality. It combines insights from economic analysis to draw inferences from unique datasets for the state of Connecticut. Chapter 1 gives a broad overview of the industry and highlights the apparent importance of signal quality as a driver of demand. Publicly available information reveals that plan features, phone selection, and pricing seem to be less determinant of overall quality relative to the quality of the call network. Reduced form evidence from proprietary data on demand and base station location data from Connecticut confirm that signal quality is important and that base stations are important to signal quality. However, the analysis reveals base station numbers themselves are not the only thing important for signal quality but likely interact with other carrier characteristics such as spectrum, transmission technology, and network management. Given the importance of base stations, Chapter 2 asks what are the competitive incentives to provide them and how would these incentive change in proposed mergers between two of the four largest firms in this industry. To answer this question, I use proprietary demand data and base station locations to estimate a structural model of supply and demand in this industry. The analysis improves on the analysis in Chapter 1 by incorporating time- varying preference heterogeneity and by controlling for the endogeneity of quality through an instrumental variables strategy. I use a measure of land use regulation stringency from data on Connecticut zoning codes as instruments for the costliness of construction. Instrumenting matters, as it increases baseline estimates of quality sensitivity by a factor of 2. Overall, I find base stations to have important competitive implications, as they represent a significant proportion of costs. Findings also indicate base stations are strategic substitutes: if one firm increases the number of their base stations, their rivals will have a lower incentive to maintain base stations. Simulating mergers between AT&T and T-Mobile and Sprint and T-Mobile, I find these mergers to be generally consumer welfare reducing without efficiencies. In particular, and consistent with previous literature, I find the mergers induce increased differentiation between merging partners, so much so that one partner's quality is severely degraded. However, the natural efficiency of being able to use a single network instead of two can make the mergers welfare-improving if resulting price increases are not too high. This result implies that merger reviews in industries with networks should investigate the scope of network integration as potentially important efficiency. Chapter 3 expands on the instrumental variable in Chapter 2 and explores how exactly land use regulation impacts the incentives of firms to invest across different jurisdictions. Through more costly requirements, land use regulations can discourage firm investment. But land use regulation can also encourage investment through reducing legal ambiguity and thus the risk of investment hold-up through legal technicalities. To test these ideas and to control for unobserved location heterogeneity, I use a border discontinuity approach that looks at sites placed near town borders and compares the relative stringency of regulation be- tween bordering towns. To measure the stringency of regulation and to separate out different regulations, I use both researcher-coded measures from manual inspection of the regulations and measures derived from the computational linguistic technique of Latent Dirichlet Allo- cation topic modeling. I confirm that regulations that impose more requirements on things like landscaping, electric power and signal strength, and impose time limits for approvals are associated with fewer sites. I also confirm that regulations that improve clarity, particularly the acknowledgment of federal standards and regulations, are associated with more sites. This has the important policy implication that the Federal Telecommunications Act of 1996 may have helped reduce investment risk. A simple counterfactual shows that regulation has only a modest effect on the reallocation of facilities across towns, holding the number of facilities fixed. Overall, these chapters clarify the role and costs of an important kind of quality provision in a major industry. They contribute significant insight for policy in both merger review and land use regulation. The second chapter is the first paper to treat signal quality as an endogenous characteristic in a study of the wireless telecommunications industry. The last chapter introduces to the economics literature topic modeling in the analysis of regulatory effects and statutory clarity as a regulatory concern. Table of Contents List of Figures ii List of Tables iii Acknowledgments iv 1 The Mobile Telecommunications Industry in Connecticut 1 1.1 Introduction . .2 1.2 Related Literature . .3 1.3 Industry Background . .4 1.3.1 Industry History . .4 1.3.2 Available Features and Services . .8 1.3.3 Pricing . .9 1.3.4 Handset Selection . 12 1.3.5 Signal Quality . 14 1.3.6 Base Stations and Sites . 14 1.3.7 Transmission Technology, Spectrum, and Network Management . 19 1.4 Connecticut Data . 22 1.4.1 Base Station Data . 22 1.4.2 Zoning Codes . 23 i 1.4.3 Proprietary Demand Data . 25 1.5 State-Wide Demand Analysis . 28 1.6 Sites and Base Stations . 36 1.7 Evidence on the Relationship between Signal Quality and Base Stations . 46 1.7.1 Data on Signal Quality and Base Stations . 47 1.7.2 Log Base Station Density . 51 1.7.3 Demand for Signal Quality and Base Stations . 53 1.8 Conclusion . 65 2 Quality Competition in Mobile Telecommunications: Evidence from Con- necticut 67 2.1 Introduction . 68 2.2 Competitive Effects of Quality . 73 2.3 The Industry Model . 83 2.3.1 Demand . 84 2.3.2 Supply . 89 2.3.3 Caveats . 92 2.4 Data . 94 2.5 Estimation and Results . 96 2.5.1 Endogeneity of Quality . 96 2.5.2 Demand Estimation Procedure . 97 2.5.3 Results: Individual Identified Demand Parameters . 100 2.5.4 Results: Quality Sensitivity Parameters and Brand-Year Effects . 104 2.5.5 Supply Side Estimation and Results . 113 2.6 Counterfactuals: Mergers of a Small Carrier . 118 2.6.1 Discontinue All Products from Purchased Firm (*) . 123 ii 2.6.2 Retain Products from Purchased Firm with Separate Networks (**) . 124 2.6.3 Retain Products from Purchased Firm with Fully Integrated Networks (***) . 132 2.7 Conclusion . 135 3 The Costs and Clarifying Effects of Regulation for Business Investment: Evidence from Cell Siting 137 3.1 Introduction . 138 3.2 The Impact of Land Use Regulation . 141 3.2.1 Burdens . 141 3.2.2 Clarification . 142 3.2.3 Causes of Inefficient Regulation Stringency . 146 3.3 A Model of Site Location Choice Under Regulatory Opportunism and NIM- BYism . 148 3.3.1 Setup of the General Model . 150 3.3.2 The Value of Zoning Codes: The One Town Case . 153 3.3.3 Zoning Codes in Equilibrium: Solution of the Multiple Town Case . 158 3.4 Econometric Specification . 163 3.4.1 Conditional Logit . 166 3.4.2 Potential Omitted Variable Bias . 168 3.4.3 Border Discontinuity . 169 3.5 Data . 172 3.6 Measuring Stringency of Regulation . 175 3.6.1 The Model of LDA . 181 3.6.2 Estimation of LDA . 183 3.6.3 Estimated Topics . 187 iii 3.7 Estimation . 194 3.8 Results . 198 3.8.1 Manually Coded Results . 199 3.8.2 Topic Focus Results . 202 3.8.3 Topic Presence Results . 206 3.8.4 Combination of Regulation Measures Results . 211 3.8.5 Fifty Topics Results . 217 3.9 Discussion . 218 3.10 Conclusion . 226 Bibliography 228 Appendices 244 Appendix A Appendix for Chapter 1 245 A.1 A Simple Model of Quality and Base Station Density . 245 Appendix B Appendix for Chapter 2 249 B.1 Comparative Statics of the Example with More than 2 Firms . 249 B.2 General Comparative Statics of the Merger Scenarios . 251 B.3 Comparative Statics Under Multi-Product Logit Demand Model . 254 B.4 Additional Tables and Figures . 256 Appendix C Appendix for Chapter 3 259 C.1 LDA Algorithm Details . 259 C.1.1 Text Transformation . 259 C.1.2 Estimating Algorithm . 263 C.1.3 Town-Level Topic Distributions . 264 iv C.2 Additional Tables . 266 C.3 Results with 50 Topics . 270 v List of Figures 1.1 Consolidation in U.S. Telecom . .5 1.2 Average Revenue Per User by Carrier and Plan Type . 11 1.3 Average Handset Characteristics over the Sample Period . 12 1.4 UBS Estimated Total Sites by Firm . 18 1.5 From page 53 of 17th Annual Wireless Competition Report (2014). 20 1.6 Estimated Market Shares over the Sample Period . 30 1.7 Estimated Carrier Prepaid Shares over the Sample Period . 31 1.8 Estimated Carrier Shares by Income Brackets . 32 1.9 Nielsen Estimated Sample Market Shares by Five-Year Age Brackets . 34 1.10 Site Ownership Fraction over Time . 38 1.11 Base Stations over Time . 40 1.12 2000 PUMAs - Subdivided into Block Groups Colored by 2010 Population Density . 41 1.13 Log Base Station Density on Quality Rating . ..