Growth in the “New Economy”: U.S. Bandwidth Use and Pricing Across the 1990s

Douglas A. Galbi

Senior Economist[1]

Competitive Pricing Division

Common Carrier Bureau, FCC

July 9, 2000

** This paper (including a pdf version) is freely available from http://www.galbithink.org  **

Version 1.4

Summary

An acceleration in the growth of communications bandwidth in use and a rapid reduction in bandwidth prices have not accompanied the U.S. economy’s strong performance in the second half of the 1990s.  Overall U.S. bandwidth in use has grown robustly throughout the 1990s, but growth has not significantly accelerated in the second half of 1990s.  Average prices for U.S. bandwidth in use have fallen little in nominal terms in the second half of the 1990s.  Policy makers and policy analysts should recognize that institutional change, rather than more competitors of established types, appears to be key to dramatic improvements in bandwidth growth and prices.  Such a development could provide a significant additional impetus to aggregate growth and productivity.

The U.S. economy’s performance in the second half of the 1990s was much better than in the first half of the 1990s.  Output grew much faster: U.S. real non-farm business output grew at an average annual rate of 4.8% 1996-1999, compared to 2.8% 1991-1995.   Productivity growth also showed significant improvement between the first and second halves of the 1990s, with labor productivity growth rising to 2.6% per year 1996-1999 compared to 1.5% per year 1991-1995.[2]  Understanding these beneficial developments and their implications for the future is vitally important in formulating a wide range of economic and social policies.

The U.S. economy’s performance in the second half of the 1990s has stimulated belief that a “new economy”, based on digital information processing and communications, is emerging.  Empirical studies show that computer hardware, software, and communications equipment drove U.S. growth in the second half of the 1990s,[3] while the Internet and electronic commerce have attracted much public attention.[4]  The preface to a report entitled “Digital Economy 2000” notes: “…confidence has increased among experts and the American public that the new, proliferating forms of e-business and the extraordinary dynamism of the industries that produce information-technology products and services are harbingers of a new economic era.”[5]  Another report associates over a half trillion dollars of U.S.-based company revenue in 1999 with what it calls the Internet Economy.[6]  The European Union has conceptualized impending economic changes in terms of an Information Society, and it has launched an “eEurope” initiative to ensure that the European Union benefits fully from developments it considers “the most significant since the Industrial Revolution”.[7]

The U.S. growth experience in the 1990s and the perceptions of a new economy contrast sharply with trends in bandwidth use and pricing across the 1990s. Dramatic increases in bandwidth use and dramatic reductions in bandwidth prices that were predicted to occur in the second half the 1990s did not occur.  Over-all bandwidth use has grown robustly throughout the 1990s, but growth has not significantly accelerated in the second half of 1990s.  Reductions in bandwidth prices, in sharp contrast to reductions in computer prices, have not accelerated in the second half of the 1990s. With respect to communications bandwidth, the development of a new economy, and any associated impact on macroeconomic growth and productivity, appear to be still yet to come.

I. Bandwidth Expectations, Technology, and Use in the 1990s

Influential industry observers have long anticipated dramatic changes in communications bandwidth.  In an article published in 1994, George Gilder foresaw a “bandwidth tidal wave,” a “tsunami of gigabits.”[8]   Microsoft’s Bill Gates declared in October 1994, “We’ll have infinite bandwidth in a decade’s time.”  In the same year, Andy Grove, Chairman of Intel, stated, “If you are amazed by the fast drop in the cost of computing power over the last decade, just wait till you see what is happening to the cost of bandwidth.”  Long-time industry observer Jack Rickard, then editor of a highly regarded trade journal and known for debunking spin and hype, declared in 1996, “…bandwidth across the board will increase by at least one order of magnitude every two years.”  He predicted that in the year 2000 U.S. backbone bandwidth would be 5 Gbps, while home users would have bandwidth of 2.88 Mbps.[9]

Developments in optical technology unquestionably have made massive increases in bandwidth possible.[10]  An October, 1998 Nortel technology newsletter reported that, since 1996, Nortel had shipped more than 1,500 10 Gbps transport systems.[11]  In a recent column the president of Lucent’s Optical Networking Group stated that over the past decade, optics’ price-performance ratio has improved 100-fold.[12]  To get a sense of the current state of the technology, consider these facts.  An optical transport system scheduled for volume shipments in the third quarter of 2000 uses dense wavelength division multiplexing (DWDM) to provide, without opto-electronic regeneration, a total of 560 Gbps across a single 3600 km optical fiber.[13]   Routers with a rich set of packet-forwarding functions are currently available that provide 160 Gbps throughput using a box that is half the size of the typical telecom equipment rack.[14]  An all-optical wavelength router, scheduled to be commercially available in December 2000, will handle 256 40 Gbps data channels.[15]

While technological developments have enabled vast increases in bandwidth, the existence of such technology does not necessarily imply its widespread deployment in wide-area communications networks.  New transmission technologies work most effectively over new fiber strands that have enhanced optical properties. Thus new fiber deployment indicates an expansion of potential capacity via both an increase in the gross volume of installed fiber and an improvement in its technological vintage.  Data indicate that growth in fiber miles deployed in the U.S. has been falling for over a decade.  Table 1 shows that fiber miles grew about 25% per year in the beginning of the 1990s, while they grew about 18% per year at the end of the 1990s.  The explanation for this trend is clear from Table 1: local incumbents, who currently account for about two-thirds of total fiber miles, have reduced their rate of deployment of new fiber miles.[16]  The trend in fiber miles deployed suggests that growth in potential capacity has not accelerated in the second half of the 1990s.

Table 1 U.S. Fiber Miles (in thousands)
	Local Incumbents		New Local Entrants		Long Distance		Total
Year	miles	grth.	Miles	grth.	miles	grth.	miles	grth.
1988	1,754				1,722		3,476
1989	2,255	29%			1,892	10%	4,147	19%
1990	3,181	41%	50		2,085	10%	5,315	28%
1991	4,389	38%	76	53%	2,203	6%	6,668	25%
1992	5,863	34%	116	52%	2,227	1%	8,206	23%
1993	7,508	28%	211	82%	2,291	3%	10,010	22%
1994	9,018	20%	365	73%	2,456	7%	11,839	18%
1995	10,698	19%	640	76%	2,585	5%	13,924	18%
1996	12,343	15%	1,306	104%	2,940	14%	16,588	19%
1997	14,017	14%	1,826	40%	3,419	16%	19,262	16%
1998	16,077	15%	3,038	66%	3,681	8%	22,796	18%
1999	17,885	11%	4,739	56%
Notes and sources: See Appendix B.

More significantly, available data for bandwidth in use do not show a rapid upsurge in the second half of the 1990s.  From 1989-1994, regional Bell Operating Company (RBOC) inter-office non-switched bandwidth sold grew about 33% per year, the same growth rate subsequently experienced 1995-1999.[17]  See Table 2.  Data on combined U.S. trans-Atlantic and trans-Pacific bandwidth, which includes bandwidth that all companies provide, shows growth of 47% per year and 79% per year in 1989-1994 and 1995-1999, respectively.   Much of the growth in the latter period, however, depends on the 1999 bandwidth estimate; growth for the period 1995-1998 is 51% per year.[18]

While network providers other than local incumbents have been active in supplying domestic bandwidth, new local providers are not large enough to dramatically affect the growth rates for domestic bandwidth in use.   New local network providers earned about 12% of total local leased line revenues in the U.S. in 1998.[19]  Surveys of end users and network build-out in major U.S. cities suggest that three or more major new companies are providing leased line services in large cities, and companies other than the incumbent local exchange company (LEC) provided about a third of channel termination facilities on a DS1 equivalent basis in those cities about 1998.[20]  However, even if the bandwidth in use from new local network providers grew from zero in 1995 to an amount equal to that of the RBOCs in 1999, the aggregate bandwidth growth rate from 1995 to 1999 would be only about 58% per year.[21]

Table 2 Bandwidth in Use
	U.S.Trans-Atlantic & Trans-Pacific		U.S. incumbent (RBOC) inter-office
Year	Gbps	Yr-to-Yr Growth	Gbps	Yr-to-Yr Growth
1989	2.1		330
1990	2.1	0%	475	44%
1991	3.1	53%	683	44%
1992	6.5	108%	835	22%
1993	9.4	44%	1,014	21%
1994	14.3	51%	1,380	36%
1995	20.6	44%	2,362	71%
1996	32.2	56%	2,693	14%
1997	36.0	12%	3,461	29%
1998	70.9	97%	5,150	49%
1999	210.2	197%	7,407	44%
2000	868.3	313%
Notes and sources: See Appendix B.

Growth of bandwidth in use for Internet traffic has been dramatic since 1995, but Internet bandwidth is only a small part of total bandwidth in use.  A careful study suggests that Internet backbone traffic grew 1000% per year in 1995 and 1996, but the growth rate fell to 100% per year in 1997 and 1998.  Total Internet backbone bandwidth in mid-1998 was probably about 110 Gbps.[22]  As Table 2 shows, total Internet bandwidth is only about 2.1% of total RBOC interoffice bandwidth in use in 1998.[23]  Thus the growth of total bandwidth in use is thus far not greatly affected by the rapid growth of Internet bandwidth.

The evidence and analysis above contrasts not only with expectations of industry leaders in the early 1990s, but also with some current perceptions of a bandwidth explosion.  Citing a statistic that has become more general and more authoritative over time, Reed Hundt, who served as chairman of the U.S. Federal Communications Commission (FCC) from 1993 to 1997, stated in his recent book, You Say You Want a Revolution, “In 1999 data traffic was doubling every 90 days, as connected personal computers spread across the globe.”[24]  Doubling every 90 days implies an annual growth rate of 1,663% per year.  An early citation of a similar statistic was the U.S. Department of Commerce’s April 1998 report, The Emerging Digital Economy, which stated, citing an Inktomi 1997 White Paper, “UUNET, one of the largest Internet backbone providers, estimates that Internet traffic doubles every 100 days.”[25]  John Sidgmore, Chairman of UUNet and Vice-Chairman of Worldcom, recently stated that UUNet’s traffic in the U.S. has been growing 750% per year consistently since 1995.[26]

Such claims, which have expanded in generality and authority with repetition and distance from the source, have provided unwarranted support for the view that a communications bandwidth revolution has helped to create a “new economy” in the U.S. in the second half of the 1990s.[27]   Recent evidence indicates that UUNet’s traffic growth claims are based on a rather peculiar measure of traffic.[28]  Consideration of the more comprehensive and better documented evidence put forward in this section suggests a much different picture of bandwidth growth. The preponderance of evidence indicates that there has not been a dramatic upsurge in total U.S. bandwidth in use in the second half of the 1990s.

II. Trends in Bandwidth Prices Across the 1990s

A rapid decline in computer prices is an important fact closely associated with the development of a new economy.   The decline in U.S. computer prices accelerated to 28% per year 1995-1998 from 15% per year in 1990-1995.[29]  Performance-price trends for computers are popularly described more abstractly in terms of Moore’s Law: the performance-price ratio for processors doubles every 18 months.[30]  In the second half of the 1990s, the period for doubling the performance-price ratio appears to have shortened to 12 months.[31]  These developments indicate high productivity growth in the production of information technology and encourage investment in information technology capital throughout the economy. 

The performance-price ratio for leading-edge optical communications technology also is improving rapidly.  Table 3 shows the investment cost per Mbps of capacity on transatlantic optical transport systems going into operation between 1988 and 2000.  The investment cost per Mbps of capacity fell by about a factor of 10 from 1989-1995 and from 1995-2000.  Note that there does not appear to have been a significant difference in the reduction in investment cost per Mbps from the first half of the 1990s to the second half.  Note also that investment cost per Mbps is not the same as the end-user price per Mbps.  The latter, and not the former, is much more relevant to bandwidth use.

Table 3 Optical Technology in Trans-Atlantic Cables
Year	Cable	Cost ($ mil)	Capacity (Gbps)	Unit Cost ($ mil/Gbps)
1988	TAT 8	360	0.5	744.0
1989	PTAT 1	400	1.1	367.4
1992	TAT 9	406	1.0	419.6
1992	TAT 10	300	1.5	206.7
1993	TAT 11	280	1.5	192.9
1995	TAT 12/TAT-13	756	23.2	32.6
1997	Gemini	520	23.2	22.4
1998	Atlantic Crossing (AC-1)	850	61.9	13.7
2000	TAT-14	1500	495.5	3.0
2001	Level 3	600	247.7	2.4
2001	Hibernia	630	123.9	5.1
Source: See Appendix B.

Bandwidth prices do not appear to have fallen significantly for most bandwidth users in the second half of the 1990s.  Table 4 shows average nominal prices per Mbps, on a distance-standardized basis, for four types of inter-office circuits that local exchange carriers offer.[32]  Prices decreased significantly from 1990-1994 but only slightly from 1995-2000.[33]  Circuits that connect customers to telephony company offices (“channel terminations” or “circuit tails”) show even less of a price decline.  See Appendix Table A2.  Given the large differences in bandwidth and price per Mbps among the different types of circuits in Table 4, shifts among circuit types could have a major affect on average prices for bandwidth.  But whether constructed using bandwidth weights or revenue weights, an aggregate index covering VG, DDS (56Kbps), DS1, and DS3 circuits shows only a slight decline in the second half of the 1990s.[34]  Such a price trend is much different than the sharp and accelerating declines in computer prices across the 1990s.

Table 4 U.S. Local Inter-Office Circuit Prices (dollars per month per Mbps)
Year	VG (64, 128Kbps)	DDS (56Kbps)	DS1 (1.5Mbps)	DS3 (44.7Mbps)	Over-all (bw wtd)	Over-all (rev wtd)
1990	159	2,514	191	16.17	161	347
1991	156	2,421	174	19.03	122	361
1992	140	2,092	177	18.07	107	352
1993	121	1,399	136	16.21	79	265
1994	134	971	123	17.95	69	202
1995	140	1,171	114	17.15	66	276
1996	144	900	116	16.53	63	219
1997	138	846	118	16.46	65	233
1998	149	925	114	16.73	63	231
1999	147	942	113	17.49	60	219
2000	143	878	112	17.17	56	185
Notes and source:  See Appendix B.

While local exchange carriers offer a variety of circuits and services and a large number of rate elements, those discussed above have been the most important on a revenue basis.  Throughout the 1990s VG, DDS, DS1, and DS3 circuits have accounted for over 80% of RBOC leased-line revenue.  See Table 5.  In the beginning of the 1990s the largest component of the “other” category was telegraph service; by the end of the 1990s the largest component of the “other” category was various types of SONET services.  Making an interstate leased line operational typically includes a variety of installation charges, cross connect charges, and other charges that a local exchange carrier might establish.  However, in revenue terms such charges have not been large; monthly rate elements for channel terminations, inter-office links, and inter-office mileage account for about 85% of RBOC leased line revenue.  Thus price trends for these rate elements are a good indication of overall price trends for local exchange carrier leased lines.

Table 5 RBOC Leased Line Revenue in 1989 and 1999
	Revenue  Shares
Circuit Type	1989	1999
VG	52.9%	3.5%
DDS	8.1%	11.6%
DS1	35.1%	47.3%
DS3	1.7%	19.0%
Other	2.2%	18.7%
	Revenue ($ millions)