In a retrospective look, George Ford of the Phoenix Center finds, “no economic payoff from the 15 Mbps speed difference.” He’s comparing 10 Mbps versus 25 Mbps broadband speeds in 2013-2015 in the U.S. Better broadband is great, but economic claims range from well-paid gibberish to essentially unproven. (My gut is there is an effect, but it’s too small to prove. Those “1.3% of GDP” type figures are absolutely unsupportable and embarrassing when quoted.)
Anyone with common sense can see this is on target. Few websites run over a megabit or three. 10 megabits is enough for 2 HD videos, plenty of surfing/homework, and five music channels. Higher speeds are great for pirating music, playing games, watching 3-5 HD TV (or sharp 4K TV,) and folks like Jennie, who does video professionally. But Jennie’s business wouldn’t suffer greatly if occasionally her uploads run overnight.
Sometimes, research proves common sense inadequate. But statisticians in evidence-based medicine have learned that unusual findings need to be looked at carefully. Always ask, “Is there a plausible mechanism that explains this finding?” If not, everything needs to be doublechecked and verified.
I do not believe it plausible that watching more HD TV has a significant economic effect.
Currently, we are seeing incredibly egregious claims about 5G, especially in mid-band. It is only 15-50% faster than good 4G LTE. I can’t see how raising wireless speeds from, say 200 megabits to 300 megabits, is going to have an important economic effect. 5G NR Only 25% to 50% Faster, Not Truly a New Generation is one of the most important articles I’ve written, especially because the primary source works for T-Mobile.
George also found, “The results of an earlier and earlier and frequently cited study by Crandall, Lehr, and Litan (2007) are probably spurious.” I know them all to be competent, but even MIT scholars can suffer from confirmation bias. My statistics teacher would never have let Crandall’s work pass.
In a separate paper about Sprint-T-Mobile, he writes, “the merger will put upward pressure on wholesale prices even after accounting for merger efficiencies.” It’s easy to infer from that that wholesale prices are likely to go up. With costs coming down 40% per year across mobile, at least as measured in price per bit, almost all prices will come down. Sprint-TMO will result in prices higher than they would be otherwise, but they will likely fall in absolute terms.
If you haven’t read John Ioannidis’ seminal paper, Why Most Published Research Findings Are False – PLOS, do so before you believe almost any paper in Telecom Policy.
Especially in D.C., far too many are worthless.
Here’s the abstract and conclusion of the paper.
Abstract: In this BULLETIN, I aim to quantify the relationship between higher broadband speeds (10 Mbps versus 25 Mbps) and the growth rates in important economic outcomes in U.S. counties including jobs, personal income, and labor earnings. Doing so exposes the potential for severe selection bias in studies of broadband’s economic impact, which is addressed in this study using Coarsened Exact Matching. Once balanced, the data reveal no economic payoff from the 15 Mbps speed difference between the years 2013 and 2015. I also revisit the Crandall, Lehr and Litan (2007) study on broadband’s effect on employment to evaluate the possible impacts of selection bias, and conclude that the positive benefits of broadband reported in that particular study are likely spurious. The selection bias problem may infect other studies on the economic impacts of broadband Internet services.
Conclusion Do counties with mostly 25 Mbps broadband connections fare better economically than counties with mostly 10 Mbps broadband connections? I find no evidence of such an effect here, at least with respect to the growth in jobs, personal income, or labor earnings between 2013 and 2015. This analysis is, of course, limited to those time periods and the outcomes considered, and is dependent on the particular empirical model employed. While these are limitations, the empirical analysis attempts to answer a specific policy question of some importance, and upon doing so, is perhaps the sole evidence available on the presently important topic. What my statistical analysis does find evidence of is profound differences in the economic character of counties with different broadband speeds. Selection bias appears to be a serious problem in quantifying broadband’s economic impact. Broadband (and higher speed broadband) is not randomly distributed across geography, but rather is deployed in areas where the ratio of demand to costs is favorable, complicating the task of discovering broadband’s influence on economic outcomes. Not only does my own analysis of broadband speeds demonstrate these empirical difficulties, but application of standard techniques reveals that the results of an earlier and frequently cited study by Crandall, Lehr, and Litan (2007) are probably spurious. In future research on broadband’s impact on economic outcomes, whether at a macro or micro level, statistical techniques must be used to address both problems of confounding influences and the potentially large differences in the economic and demographic characteristics across geographic markets