“The sum of anecdotes is not data.” Roger Brinner (economist).
As someone who took pride in my role overseeing the production of the 12th CMRS Competition Report, I read with great interest the 14th Mobile Wireless Competition Report (“14th Report”). Although I was impressed with many elements of the 14th Report, I was also perplexed by some of its conclusions. Perhaps the most perplexing aspect of the report is its discussion of spectrum, in which it relies on a number of anecdotes masquerading as data to reach a new “conclusion.”
The new element of the 14th Report in respect to spectrum is its discussion of the differences between spectrum below 1 GHz and spectrum above 1 GHz. For the first time ever the FCC concludes that “providers whose spectrum assets include a greater amount of spectrum below 1 GHz spectrum may possess certain competitive advantages for providing robust coverage when compared to licensees whose portfolio is exclusively or primarily comprised of higher frequency spectrum.” (14th Report at paragraph 283.) But the 14th Report doesn’t include any actual data supporting this “conclusion.” The 14th Report instead relies on a series of anecdotes to justify its new position, and doesn’t attempt to quantify the extent of any such competitive advantage (assuming one actually exists) at all. Without some data demonstrating that differences in frequency actually yield significant competitive advantages, the “conclusion” is really just a hypothesis.
So, upon what facts does the FCC rely to support its hypothesis? First, the FCC notes that “spectrum resources in different frequency bands can have widely disparate technical characteristics.” (14th Report at paragraph 268.) While that’s true in the abstract, it doesn’t tell us anything about competition. The question is whether any such differences in technical characteristics have a significant effect on competition.
The 14th Report recites the views of “several commenters” that “have noted the advantages of lower frequency spectrum for coverage and penetration.” (14th Report at paragraph 269.) The 14th Report then reasons that “a licensee that exclusively or primarily holds spectrum in a higher frequency range generally must construct more cell sites (at additional cost) than a licensee with primary holdings at a lower frequency in order to provide equivalent service coverage, particularly in rural areas.” (See 14th Report at paragraph 270.) But the 14th Report doesn’t cite any authority for this proposition or include any data regarding actual differences in cell densities and deployment costs based on spectrum propagation. Because cell density is affected by numerous factors other than propagation characteristics, a data-driven conclusion regarding the impact of spectrum on deployment costs would need to analyze the real-world impact of propagation characteristics and other factors affecting cell density at typical power levels in a variety of geographic terrain and population densities. The 14th Report skips this essential step as it hurries to its preferred end result.
In fairness to the FCC, the 14th Report does recognize in a footnote that cell density in urban areas may be driven by the need for capacity (i.e., larger cell sizes, or lower cell density, adversely affects capacity). (14th Report at footnote 735.) And the 14th Report appears to recognize that cell density can be a function of available bandwidth as well. (See 14th Report at paragraph 272.) But it doesn’t actually analyze the impact of these factors on cell density. If capacity needs dictate similar cell density across both higher and lower frequency bands, then differences in propagation characteristics would have no competitive impact whatsoever, let alone a significant one.
The question is whether any such differences in technical characteristics have a significant effect on competition.
Worse, the FCC fails to analyze or even recognize relevant data altogether. For example, for spectrum propagation to have a competitive impact in rural areas (where physics would dictate such a result is most likely), additional competitive entry in those areas would need to be economically feasible. In other words, access to frequencies with better propagation characteristics would have no impact on competition in rural markets where additional entry is uneconomic at any frequency. Indeed, the FCC just relied on a finding in its 2010 Roaming Order that “in some areas of the country with very low population densities, it is simply uneconomic for several carriers to build out.” (2010 Roaming Order at paragraph 23.) Given its recognition of this fact in the 2010 Roaming Order, it is surprising that the FCC doesn’t raise it in relation to its spectrum hypothesis.
Roaming is yet another unrecognized and unanalyzed factor undermining the FCC’s hypothesis that propagation characteristics are competitively significant. One of the FCC’s justifications for eliminating the “home roaming exclusion” (see my posts here and here) was that “it may be significantly more costly to build out when the carrier only has access to higher spectrum frequencies where propagation characteristics are less advantageous.” (2010 Roaming Order at paragraph 23.) If the FCC is correct in its assumption that elimination of the “home roaming exclusion” will remedy this alleged difference in build out costs, then differences in propagation characteristics would no longer be competitively significant. Although the FCC does discuss roaming in the 14th Report, it makes no effort to relate the impact of roaming obligations on its conclusions regarding lower frequency spectrum. Instead, the 14th Report erroneously treats these two topics as if they were unrelated.
The other anecdotes relied upon by the FCC to support its hypothesis are simply irrelevant. The 14th Report suggests that lower frequency spectrum offers competitive advantages because it has been valued more highly at auction – and uses Auctions 66 (AWS-1) and 73 (700 MHz) as evidence. There is a threshold question here – which is whether different auction valuations are evidence of competitive advantage or instead can be explained by other factors (such as the state of the credit markets at the time of the auction). But one needn’t even answer the threshold question where Auctions 66 and 73 are concerned. I was responsible for both auctions at the FCC, and it is likely that the price difference between the two auctions resulted from differences in regulation and auction structure rather than propagation characteristics of the spectrum. AWS-1 auction prices were lower because more spectrum blocks were available (thus reducing competition for any one block), the spectrum was generally unencumbered by regulation, and AWS-1 bidders weren’t subjected to anonymous bidding. The 700 MHz band, in contrast, was subject to anonymous bidding and only had two blocks that were particularly well suited to broadband service as a technical and practical matter, and one of those blocks was heavily encumbered by open access regulation. Indeed, the best explanation for certain bidders’ behavior in the 700 MHz auction is that those bidders rejected the C Block open access regulations, which drove the lower band B Block to disproportionately high MHz/pop levels in major markets. For example, Metro PCS paid almost as much for Boston alone in the unencumbered B Block as Verizon paid for the entire Northeast REAG in the wider but generally unwanted C Block – and MetroPCS pulled its potentially winning bid for the Northeast C Block early on in the auction. The 700 MHz auction doesn’t reflect accurate “market” prices because the spectrum wasn’t offered on the same market-oriented basis as the AWS-1 spectrum. Thus, pricing differentials in auctions 66 and 77 are irrelevant to an analysis of cell densities. (This subject is worthy of a much more in-depth treatment than I give it here, but even this brief treatment is more complete than that offered in the 14th Report. I also note that, even if these price differences were relevant to cell densities, such evidence is indirect, and thus less probative than actual evidence of real-world cell densities and deployment costs.)
The FCC also cites a propagation model developed by the National Institute of Standards and Technology (“NIST”) to provide telecommunications designers working in public safety communications with channel propagation models to use in simulation and testing. But the proposition for which the FCC cites NIST – “it required nine cells at 2.4 GHz and four cells at 1.9 GHz to span 100 meters squared, it was projected to require only one cell at 700 MHz” – wasn’t derived from the NIST propagation model; rather it was reiterated (inaccurately) by NIST from an unsubstantiated March 2007 article in Gigaom. The inaccuracy is that the Gigaom article doesn’t compare coverage using a metric of 100 meters squared; it compares coverage using 1,000 square miles (which is likely not a typical, market-based cell size). This “data” in the FCC report is thus unsubstantiated and just plain wrong (and wildly so). And even if it were accurate, the anecdote in the Gigaom article doesn’t attempt to take into account any of the factors affecting cell density discussed above – i.e., terrain, population density, capacity, available bandwidth, and economic entry. (This is an example of why agencies are required to subject technical reports upon which they rely in rulemaking proceedings to a peer review process.)
Even assuming that propagation characteristics do have an effect on competition, the 14th Report is devoid of any analysis regarding the extent of the effect. There are data, however, supporting a hypothesis that any such competitive advantage likely isn’t significant. First, the 14th Report finds that at least 95.8 percent of the population is served by at least three mobile voice providers, (14th Report at page 7) and “[t]he four facilities-based providers that analyst reports typically describe as nationwide all have mobile wireless networks that cover in excess of 86 percent of the U.S. population in large proportions of the western, mid-western, and eastern United States.” (14th Report at paragraph 27.) Such significant coverage – despite the differences in spectrum holdings below 1 GHz – would tend to indicate that propagation characteristics are not playing a significant role in the market, because sufficient low frequency spectrum has been made available with enough diversity of license holders to cover the vast majority of the population with at least three different providers. And, in the worst case scenario, the fourth competitor lacks facilities covering only about 12% percent of pops covered by three providers, absent roaming coverage. When coupled with the FCC’s finding in the 2010 Roaming Order that additional competitive entry would be uneconomic in areas with very low population densities, which would presumably narrow the 12% gap, there is evidence that disparities in spectrum holdings below 1 GHz are having little impact on competition (if any at all).
TAGS:
data roaming mobile spectrum wireless wireless competition