Measuring Broadband’s Economic Impact*

1. Measuring Broadband’s Economic Impact* Sharon E. Gillett, William Lehr, Carlos Osorio Massachusetts Institute of Technology Communications Futures Program (CFP) Marvin Sirbu Carnegie Mellon University Engineering and Public Policy March 2006 Research supported by the Economic Development Administration of the U.S. Department of Commerce, with matching funds provided by MIT ITC and CFP industry sponsors Papers available at http://cfp.mit.edu/groups/broadband/measuring_bb_pp.html

2. Why Broadband’s Economic Impact Matters? • To governments and non-profits • Competitiveness issue: Local, regional / state, national • Funding, involvement in broadband • Spillover benefits / positive externalities • To firms • Investment decisions, revenue opportunities • Government relations • To citizens • Economic implications of broadband public policies • Brand X, network neutrality, municipal broadband, video franchising, interconnection, spectrum, etc.

3. Many Reasons to Expect Economic Impacts… • Firms • Increased productivity via information & communications technology (ICT) • Forman, Goldfarb and Greenstein (2005); Bresnahan, Brynjolfsson, and Hitt (2002); Brynjolfsson and Hitt (1996); Autor, Levy and Murnane (2003) • “ICT users” – use ready-made applications such as email and web to raise quality, lower cost of gathering market intelligence, communicating with customers and suppliers • “ICT enhancers” – develop, customize, and integrate more complex “e-business” applications e.g. automated supply chain management, online sales into remote markets • Workers: ICT complements knowledge workers, substitutes for more routine jobs • Overall effect on jobs ambiguous • Individuals (residential users) • Work at home / self-employment / remote employment • Educational opportunities (workforce quality) • Civic participation, social capital, quality of life • More efficient job hunting, household management (e.g. online bill pay), shopping etc. • Overall effect on local economy ambiguous

4. …But Impacts Likely Hard to Measure • Solow’s “IT Productivity Paradox” • Poor measures of inputs (e.g. rapid depreciation, lack of usage metrics) • Inputs work in conjunction with each other and organizational / institutional changes • Outputs hardest to measure in service sector, where impacts most likely • Impact happens over time • Particular to broadband • BB as an input: Local variability in quality, price, adoption • But FCC data only lets us observe availability (poor proxy) at the local level • Have statewide penetration, but statewide data aggregation masks effects • Limited output data available at local level (not GDP) • Different types of impact at different time scales • BB changes internal business processes, entry/exit decisions, prices, worker skills, etc. • Local impact (temporary advantage) vs. national (competitive necessity) • Need good micro data to resolve!

5. Progression of BB Impact Studies 2001-2 2003 2005 • 1G: Prospective, hypothetical • Crandall & Jackson (Verizon): BB to add $500b to GDP by 2006 • Pociask (New Millenium Research Council): BB to create 1.2m jobs • Ferguson (Brookings): Lack of BB to lower productivity growth by 1% annually • 2G: Case studies, individual communities • Kelley: Cedar Falls, Iowa (muni bb since 1997) improved vs. neighboring Waterloo • Strategic Networks: S. Dundas, Ontario (muni fiber since 2000) grew sales, jobs, tax revenues • 3G: Controlled, statistical, larger geographic scope • Ford & Koutsky (Applied Economic Studies): Retail sales grew in Lake County, Florida (muni bb since 2001) vs. 10 control counties • This study: U.S. national scope, compares 2002 economic indicators by zip code, based on FCC report of BB availability by 1999

6. Data Sources

7. Key Findings • National data supports the conclusion that broadband positively affects economic activity • Even after controlling for community-level factors known to influence BB availability and economic outcomes • Controls: urban, income, education, growth in previous period • Communities where mass-market BB was available by December 1999 experienced more rapid growth by 2002 in: • Jobs (employment) • Number of businesses (overall) • Businesses in IT-intensive sectors • Property values higher in 2000 where BB available by 1999 • Higher market rates for rental housing in 2000 • Rents reported more accurately than home values in Census data

8. Estimated Magnitude of Impacts

9. How Strongly is Causality Supported? • Does BB cause or follow economic activity? • Impacts stronger than expected • Be aware of methodological limitations • Three controlled regression approaches tested here • State-level • BB metric = penetration (FCC data on residential & small biz lines in service) • Weak results due to excessive geographic aggregation • Zip code level • BB metric = availability (FCC data on # of bb providers) • Two methods tested: ordinary controls & matched sample (“hidden” controls) • Leads to estimated range of results • Issues • How independent (exogenous) are control variables from economic growth? • Accuracy of BB availability data (esp. in rural areas), and relation to penetration / use • Short time scale, limited data (small universe of potential methods, for now) • Future improvements always possible!

10. Potential Next Steps • Further refinements to methodology • Pursue data for instrumental variables as alternative approach to causality issue • More micro-level look within firms • Further exploit enterprise panel data set(s) • Case study research on BB impacts • Incorporate additional data of economic impacts/effects • Later years of business census • Next household census -> voting behavior, eGov metrics, self-employment • Richer FCC BB metrics as of 2005 • Stimulate better data about BB use • Penetration and QoS of BB data? (Generation of technology) • FCC? States? Private (e.g. Pew)?

11. Conclusions • Results consistent with view that broadband access does enhance economic growth and performance, and that the assumed (and oft-touted) economic impacts of broadband are real and measurable • Analysis suggests stronger-than-expected impacts • Data limitations warrant cautiously optimistic interpretation • While we have not proven that BB causes economic impact, neither have we proven that it doesn’t, and data clearly show association • Support will be stronger over time as data improves and multiple studies reinforce • Policy makers promoting BB are doing the right thing • How much BB is enough? • As BB becomes more uniformly available, will be competitive necessity rather than local advantage. Suggests shift of policy emphasis toward how BB used. Need for balanced portfolio of economic development strategies: availability AND adoption/usage. • Can you improve the data available to assess the impacts of broadband policy?