Commenters Provide Dozens of Arguments Against QR, None in Favor
During the FCC’s year-plus long campaign for USF/ICC reform, the Commissioners continually mentioned how byzantine, complex, convoluted, anti-consumer, loophole-filled and backwards the old USF and ICC rules were. And then the FCC came up with a Quantile Regression Analysis (QR) methodology for HCLS, which even the so-called "father of qunatile regression," Dr. Roger Koenker, believes is applied incorrectly. Now we have a method for capping HCLS that is not only all of the above, but also punitive, retroactive, inaccurate, utterly unpredictable and all-around nefarious—unless by some chance the methodology and inputs are completely revamped, or QR gets tossed in the appeals cases. Since we can’t assume either a revamp or an appeals victory, it is best to try and comprehend how QR will impact your company and why it is so precarious.
Last month, JSI Capital Advisors looked extensively at several prominent reasons why the 11.25% rate of return for RLECs should not be disrupted until the dust settles on the myriad cuts and caps adopted in the reforms (The ILEC Advisor: Saving Rate of Return is Saving RLEC Financial Integrity). Unfortunately, the QR objections are not quite as easy to condense—while there were about a half dozen solid reasons why RoR should be left alone for the time being, I’ve counted over two dozen reasons why, as most commenters claim, QR is “fatally flawed.” Derrick Owens, vp of government affairs at the Western Telecommunications Alliance commented to JSICA, “We are troubled by the Commission’s intent to impose its quantile regression model on RLECs, particularly the desire to apply it retroactively on investment that has already been made. This model will have a negative effect on companies’ ability to plan invest and deploy broadband in high cost rural areas. While there are many aspects of the Order and further rulemaking that concern the rural associations, this is one that the Commission must alter.”
For an extremely detailed, high-level analysis of at least 21 reasons why QR is fatally flawed, look to the Nebraska Rural Independent Companies (NRIC) FNPRM comments submitted on January 18, 2012. Meanwhile, analysis of a sampling of prominent arguments from the Nebraska companies and other rural stakeholders are described below.
Garbage In = Garbage Out
The variables chosen by the FCC are perplexing and downright lazy. The FCC has insisted that variables be derived from readily available data, but then the FCC makes an illogical presumption that the chosen readily available data is also accurate, sufficient and relevant—apparently without considering various data sets that are both readily available and accurate. The Rural Associations commented, “Independent variables should be included in models only to the extent that they produce statistically significant results,” which many of the FCC’s variables do not accomplish. Furthermore, the decision to apply QR to 11 of the 26 HCLS algorithm lines creates additional problems, and “would undermine investment decisions made prudently on the basis of overall cost analysis.” NRIC also adds that the input data, the methodology itself, and the design of the caps will together “discourage carriers from investing in the expanded broadband facilities that the Commission seeks.”
NRIC provides many examples of how the FCC’s input data is flawed, and insists that the FCC “must make substantial improvements in its regression models to include more than ‘readily available’ Census data,” for example:
- The number of census blocks in a study area might be correlated with density, but “a small correlation between two variables does not make one a ‘proxy’ for the other.” The FCC has determined that census blocks are a proxy for density rather than using an explicit density variable.
- Soil data is not included; even though soil data is widely available (see map below), a significant factor in construction costs, and has been used in other FCC models. Likewise the FCC’s proxies for terrain—land area and percentage of water—are not necessarily reasonable or accurate for all study areas.
- The frost index is not included, but frozen ground has a considerable impact on costs, the construction season, and outside plant maintenance. NRIC brings up the fact that the FCC used no climate data—without an explanation—which “further calls into question any real-world confidence in the methodological approach the Commission has chosen to pursue.”
The most glaring input flaw is the FCC’s geographical mapping data for study area boundaries, derived from the 2010 TeleAtlas database. According to the Rural Associations, “Significant errors occurred in more than 90% of the study areas where the data are currently available.” NRIC points out that the FCC data produced 459 exchange boundaries for the state of Nebraska, but there are actually 514 unique exchange boundaries based on more reliable Nebraska PSC data:
Generally speaking, the rampant data inaccuracies inspire zero confidence in the FCC’s ability to create a reasonable, predictable and accurate mechanism for encouraging efficient investment and cost cutting. The Rural Associations argue, “Unless extraordinary efforts are undertaken to correct these inaccuracies, the analysis conducted by the Commission will incorporate substantial errors from the start and result in caps with no valid statistical tether to their intended purpose.”
We are the 10 Percent
It’s almost like some backwards take on the waning Occupy movement—the FCC has deemed that an arbitrary 10% of RLECs are unfairly gaining too much from the USF system and their “wealth” needs to be redistributed to the bottom 90%. However, instead of living in constant fear that you may be penalized for earning an arbitrarily-defined “excessive” income, RLECs will live in constant fear that they are spending too much on broadband infrastructure for rural Americans—an infrastructure that this very government has demanded and insisted is necessary to “win the future”—and will be penalized for doing so. What’s worse is that preliminary studies by the rural associations indicate that 283 of 720 total study areas will likely have support clipped: “While each quantile model is designed to limit data associated with 10 percent of study areas, different study areas are affected by each model differently, resulting in 41 percent of study areas being limited by one or more models.”
Considering the defective input variables and flawed methodology, the ninetieth percentile cap is way too stringent, according to NRCI: “The harm would fall randomly, depriving carriers of not only sufficient but also predictable USF payments required as a precursor for investment.” NRIC recommends that the FCC instead adopt a ninety-fifth or ninety-eight percentile cap, at least until the QR inputs and methodology are made more relevant and predictable.
In addition to the overly-severe ninetieth percentile cap, “When carriers lose universal service support due to the new caps, the amount of support lost will be ‘redistributed to those carriers whose unseparated loop cost is not limited by the operation of the benchmark methodology.’” Even if a carrier’s investment and expense costs come in well below the cut-off on 10 of the 11 algorithm lines, this carrier’s support will be capped and the shaved-off support will be redistributed to other carriers, who may or may not be operating more efficiently. To say that this particular aspect of HCLS capping is vindictive is overly generous.
The Only Thing that’s Predictable is Unpredictability
Ironically, the FCC intended for QR to inject more predictability in the HCLS mechanism, but the result is the exact opposite of predictability. John Kuykendall, vice president at John Staurulakis, Inc. explained, “In my opinion, the most threatening aspect of the regression analysis for RLECs is the randomness that permeates the entire analysis which makes it impossible for RLECs to know how the analysis will impact not only past but future investments.” Kuykendall explained the challenges of comparing companies to “similarly situated” cost companies without actually knowing which companies are included in the subgroups: “Companies have no idea of who they are being compared to, so [they] have no knowledge of whether the investments they are making would make them an ‘outlier’ or not. Even companies that have taken such routine actions such as replacing existing copper with fiber have been found to be outliers in certain categories. We have also found that undergoing certain types of construction may make a company an outlier in one or two categories but way under in others.”
NRIC explains that RLECs need to be able to reasonably predict their levels of USF support for at least 5-10 years, but with QR, a carrier cannot know what level of investment or expense spending will trigger the cap; the trigger is largely dependent on what other carriers are spending and investing; and there is a high probability that today’s reasonable investments will suddenly become excessive. The proposed annual QR recalculation “will create extraordinary financial uncertainty”—actual and perceived. Furthermore, this level of unpredictability is “in stark contrast to the rules for price cap carriers,” who “have advanced knowledge of future support levels for a period of five years or more.” Not allowing RLECs at least as much predictability about future support levels not only violates Section 254 but seriously threatens future investment in rural broadband.
QR could be a lethal injection for many RLECs investment strategies, and the penalties for exceeding the FCC’s arbitrary definition of reasonable investment are quite severe. According to NRIC, “A carrier that becomes subject to one cap in just one category, even by a miniscule margin, will suffer substantial financial loss—even if the carrier is under the other ten caps.” For example, exceeding one cap by $0.01 could reduce the RLEC’s support by at least $2.76 per line per month, and “Even a carrier that imposes a freeze on spending cannot be assured that it will not become subject to a cap and, if it is not currently subject to a cap, might someday fall off the financial cliff.”
NRIC also describes a “destructive spiral,” where caps affect spending which then affects future caps: “This feedback cycle can occur repeatedly as carriers adjust their spending over time. In short, there is a substantial risk that the caps will become more demanding over time, demanding ever lower levels of spending. At some point, carriers will simply no longer remain viable.” Adding insult to injury, lenders will become ever more reluctant to issue loans to RLECs, because lenders “seek as much certainty as possible with respect to future revenue streams…even highly trained and informed analysts will find it difficult to predict collective carriers’ behavior, and therefore will not be able to accurately predict the level of future investment and expense caps.”
Basically, QR punishes RLECs financially by limiting incentives to make any investments; by making capital more difficult to obtain; and by pitting RLECs against other, unknown RLECs in a reverse-incentive “race to the middle.” How does anyone benefit from this arrangement, exactly? Interestingly, there are virtually no comments in response to the FNPRM that support the use of the FCC's QR methodology.
What do you think of QR? Is it a lost cause, or can it be improved? Is there a strong possibility that the Appeals Court will conclude QR is a violation of Section 254? Share your thoughts on JSICA’s LinkedIn USF Forum.