The Broadband Bonus: Accounting for Broadband Internet’s Impact on U.S. GDP

1. The Broadband Bonus: Accounting for Broadband Internet’s Impact on U.S. GDP By Shane Greenstein and Ryan McDevitt Northwestern University

2. Prologue • Thanks for giving me opportunity to present. • Ryan and I wrote this academic study. • Did not write with intent to help with a stimulus bill. • Some of the findings and concepts speak to your concerns. Will give overview. Develop implications. • Favorite quote for this paper: • “That, sir, is the good of counting. It brings every thing to a certainty, which before floated in the mind indefinitely.” – Boswell’s Life of Johnson –

3. Outline • Motivation • A primer on measuring value creation • Internet Deployment History • Data • Benchmarks • Implications

4. Motivation: What is the value created by new good? • Potentially big economic change as new technology replaces old. • In 2001: 45m hh use dial up, 10m use broadband. • In 2006: 34m hh use dial-up, 47m use broadband. • How much economic value created through replacement of dial-up by broadband? • Internet access is a big industry (39B in GDP in 2006). Merits attention for its own sake. • Questions about extent of economic gains from deployment of new technology.

5. What the paper aspires to do • Provide benchmark estimates for policy • Relentlessly quantitative. Assemble best public data. • Only examine households. Not business. Only US. • Two traditional measures in economics. • Revenue growth  GDP growth or producer surplus. • Buyer willingness to pay (WTP)  consumer surplus. • Compare revenue & consumer surplus w/broadband to what would happened w/o broadband. • Robert Fogel: contribution to growth is contribution above what would have occurred in absence of new. • Compare with a world with only dial-up.

6. Bottom line: Estimates for households, 1999-2006 • Broadband’s contribution. • “New revenue created” or “New consumer surplus created” or equivalently expressed as price index. • Use of broadband in households accounts for approx $20 -$22B in new revenue, but that is not the same as created value. • Approx $15B of newly created value. • Approx $8.3B to $10.5B is new revenue for firms. • Approx $6.7B to $4.8B is consumer surplus, which is not measured as part of GDP. • Equivalent to approx 1.6% to 2.2% price decline, earlier than measured by official price indices.

7. Contributions to the lit on broadband’s economic impact • Positive contribution to economic growth. • By the norms of GDP accounting at the BEA, that is a big number for a new technology. • Like every other major innovation in history; Better off with it than without, but not a revolution. • These estimates are much lower than others. By an order of magnitude. Why? • Popular forecasts are not grounded in – or calibrated against – historical data, as ours are. • We strictly employ traditional economic methods and Fogel’s conceptualization of the issues.

8. Outline • Motivation • A primer on measuring value creation • Internet Deployment History • Data • Benchmarks • Implications

9. Theory of the measurement of the economic value from new goods • The relevant question for new goods focuses on additional benefits beyond the old. • Broadband brings benefit above and beyond dialup. • Not a new idea. Famous illustration: Robert Fogel illustrated on contribution of US railroads above and beyond canals – a Nobel Prize. • Generally divide benefits between aggregate gains/losses to all users and producers. • We will follow convention and not worry about which vendor gains/loses, nor which user gains/loses.

10. Why focus on WTP & producer gains? Value is internalized. • Willingness to Pay determines economic value. • Shaped by factors considered by parties involved in a transaction. Anything that shapes anticipated costs of providing dial-up and perceived benefits from upgrade to broadband. • Illustrations • Suppliers: sale of second lines, access revenue, dial-up ISP revenue (for telephone companies with ISPs). • Users: savings on second line; savings on commute time; health benefits & entertainment benefits. Also savings on phone bill (e.g., if user moves to VoIP).

11. What is not in WTP and producer revenue: Growth externalities • Growth externalities: not considered by parties during the transaction. • Suppliers: Benefit to Cisco from selling more Wi-Fi equipment to users. Benefit to Amazon from additional sales b/c broadband users experience more satisfying service. Benefit to Google from more ad sales b/c users stay on line longer. • Users: Unanticipated slowness that one neighbor’s use imposes on another in a cable architecture, or benefits that one person’s participation in a p2p network confers on another (as long as there is no membership fee).

12. Measurement strategy: measure the internalized costs/benefits. • Suppliers: Examine revenue for new good and compare against old. Then subtract. • No multiplier effects (to get to total impact on GDP). • Users: Examine willingness-to-pay (WTP) for broadband instead of dial-up for all users. • Consumer surplus for broadband less consumer surplus for dial-up, sold to same set of users. • Or add up WTP for all users, but it must measure WTP to change from dial-up to broadband. • No attempt to measure growth externalities in this study. Open Question.

13. Outline • Motivation • A Primer on Measuring Value Creation • Internet Deployment History • Data • Benchmarks • Implications

14. Overview of the general story • Dial-up technology diffused first • Starts growing quickly in ‘95 (growth of WWW). • Broadband came a few years later. • Limited availability in 1990s (cable first), but widens in ‘00 (especially DSL). Many regulatory issues. • Reliability/service improves over time. • Complementary services spring up (e.g., music, Web 2.0, etc.). • Internet access a large market by 2006 • Broadband almost nothing in 99, but becomes substantial part of revenue by 2006. $39B revenue, increasing % goes to broadband each yr.

15. Diffusion to households (NTIA)

16. Growth of broadband lines (FCC)

17. Revenue for households and businesses (Census)

18. Price index (CPI)

19. The CPI for Internet access is misleading. • No measureable big change in broadband prices over entire period. • Hard to tell even if there has been a little change. • Stable until fall ‘06, drops 23% in 3 months. • Informed speculation: AOL dominates late 90s and early 00s and their prices do not decline. • Most other prices do not decline much, if at all. • Timing in fall ’06 resulted from AOL switch to advertising-supported service. A big % of the index declines b/c price falls by 100% for over 20% of the users.

20. Outline • Motivation • A primer on measuring value creation • Internet Deployment History • Data • Benchmarks • Implications

21. Sources for data • Adoption: NTIA for 97–03, Pew 04-06. • Second lines: FCC report only “total use”. • Assume 1/3 of dial-up households use second line. • New users and “converts/switchers” • All BB users experienced w/dial-up in early years. • Our benchmark: 100% converts in ‘99-’01, then 81% converts in ‘02-’06. Get sense of bounds with simulated aggressive & unaggressive conversion. • Prices: Dial-up at $20. Vary BB price b/w $40 and $36. Cost of second line $20. • Common average.

22. Users and second lines

23. Outline • Motivation • A primer on measuring value creation • Internet Deployment History • Data • Benchmarks • Conclusion

24. New revenue created

25. Summary of key findings • Summary: 59% to 54% of broadband revenue is replacement of dial-up & second lines. • Revenue in 99-06 is $10.6B in 2006 if price = $40. That is 46% of $22B for households. • $8.3B when P = $36, which is 41% of $20.3B. • Aggressive conversion (too high)  $2.3B lower, while unaggressive (too low)  $0.9 higher. • Not an estimate of profitability. • Can see cable is big grower, dial-up ISPs biggest loser. Telco gains small b/c also lose second line. • Revenue levels consistent w/cost estimates for upgrade (e.g. $150-$400 per household).

26. Creation of consumer surplus • Use Savage and Waldman estimates of WTP among dial-up & broadband users in ‘02. • In same range of other estimates. • WTP assumes prior dial-up use… adoption occurs if WTP exceeds conversion costs. • We take conservative approach. • Average WTP… not including most inelastic demanders (though we suspect this will not matter). • Also conservative: new users get no surplus. • Note: We include cost of second line. • Treat most recent adopters as getting little surplus.

27. Consumer surplus created

28. Summary of key findings • CS approx $6.7B to $4.8B in 2006. • 44% or 32% of approx $15B total value created. • Aggressive conversion (too high) reduces total surplus by $0.8, while unaggressive (too low) increases $0.6, assuming $40 price, so the estimates are much more sensitive to assumption about pricing than conversion… • Cautionary notes: • No adjust for inelastic demanders or AOL’s pricing. • This data is from ’02. Would recent data from users w/recent experience show greater unwillingness to give up Broadband?

29. Weighted average of equivalent price decline • Give every convert the same WTP • Treat it reservation value. In year of conversion from dial-up to broadband, think of equivalent decline from this reservation to actual price. • Calculate weighted average for population. • Only converts experience price decline (e.g., = 0.76); others experience no price decline (= 1). • All other users get no price change except converts (dial-up users, existing broadband users, new broadband users experience same price). • Includes savings on second line.

30. Weighted average of price declines

31. Summary of key findings • Equivalent price decline = 1.6% to 2.2% per yr. • Explains diffusion in spite of no measured price decline in CPS. Slightly larger in more recent years. • Timing of benefit missing from official price index. • Normally savings in one category of good (phones) does not play a role in BLS price adjustment in another (internet access). • Second line accounts for 30% to 40% of benefit, depends on price (respectively $36 or $40). • Yes, we understand that norm, but present silence everywhere is misleading for policy.

32. Outline • Motivation • A primer on measuring value creation • Internet Deployment History • Data • Benchmarks • Implications

33. Summary • Use of broadband in households accounts for approx $20 -$22B in revenue in 2006. • Approx $15B of newly created, partly measured. • Approx $8.3B to $10.5B is new revenue, while $6.7B to $4.8B is consumer surplus. • CS equivalent to .6% to 2.2% price decline per year in eight yrs. Earlier gain than commonly recognized. • We have focused primarily on upgrade. • Interpret these as internalized economic gains from diffusion of new good. • All in all, economic gains are positive & big for a new technology, but not outsized.

34. Implications for next steps • Rural broadband expensive. • A decade of private-led build-out. Cable co. & local telcos have upgraded everywhere cost-viable. • Nothing except high cost locations left. Billion dollars to cover uncovered will not cover many high cost households. • Next generation of upgrade just starting to happen. Wireless applications get excitement. • Next five years: Either 3G/4G or WiMax. • Next 18 months? 3G or experiments w/W-Max. • If policy accelerates this deployment, will this be technologically neutral?

35. Growth externalities in the last decade? The big open question. • Development of broadband has generated some growth externalities at home. • Web2.0 (Yahoo, Facebook, Flickr); P2P (Bit-Torrent) • Music downloading (iTunes); Efficient search (Google). Would some of this had happened with dial-up anyway? • Some of the growth externalities that have not been as large as optimists forecast… • VoIP (Skype has not grown as forecast). • Virtual life (Second life, Xbox online are niche uses). • What will another million homes generate?

36. Growth externalities in the next decade? • Some gains “local”, but hard to measure. Most visible gains accrue to national businesses, such as Google, Amazon, E-Bay, advertising supported Web. • VCs focused on next wave: A viable mass market wireless access mode. • New apps, new apps, new apps. • Some gains “local” but hard to measure/see. • Zillow, real estate sites. • Loss of revenue at local newspapers.

37. Measuring the deployment of IT • Dial-up to broadband price measurement issues symptomatic of bigger issue. • One BLS price index for all Internet access not very useful in the face of multi-generations. Why not break out dial-up/broadband/wireless/etc. • Who is responsible & for what purpose? FCC or BLS or NTIA or Census have different needs. • Missing stats. Availability. Prices. Bandwidth. Upgrade patterns. • Pew has best public data about HH use of Internet. • US only country in world to rely on a private foundation (Pew) for basic Internet stats. Why?

38. Gracias! • Thanks for your attention!