In an Alternate Reality Where Common Sense Doesn’t Exist... Regression Analysis is Still Flawed

It is beginning to appear that the only thing predictable about the FCC’s quantile regression analysis (QRA) methodology for slashing HCLS for RLECs is that RLECs will keep fighting it until the FCC admits the methodology is indeed “fatally flawed” and makes appropriate modifications. Since we can’t bank on the FCC having a sudden change of heart no matter how much evidence of QRA-related anxiety and harm RLECs provide, we can only reiterate the arguments against this methodology with hopes that the powers that be take note. The Rural Broadband Alliance filed reply comments to the USF/ICC Transformation Order FNPRM on February 17, 2012, which fit the predictable pattern of resistance to QRA from the RLEC industry. The central argument presented by RBA is that the regression methodology is unreflective of the real world of RLEC operating and capital expenses; and therefore the FCC should either discard the methodology or somehow ensure a “fail safe” alternative path for RLECs to receive predictable and sufficient support based on actual, real world costs.

RBA reiterates many of the “macro” arguments made in earlier comments (which JSICA discussed in great detail earlier this month) but also adds several “micro”-level arguments to the mix, which proved very useful to visualize the impacts of QRA. RBA argued that individual companies facing significant financial harm by QRA are not “anomalies or exceptions.” RBA insists that the FCC ignored repeated warnings, alternative recommendations, and even long-standing legal and policy principles by adopting QRA. Furthermore, QRA is a defective model, which “fails to consider whether the incurred costs to provide universal service are just, reasonable and necessary and ‘used and useful” in accordance with the Commission’s established standards… [and] does not take into account the real world operational context of a carrier’s investment or the location-specific conditions upon which a carrier makes prudent expense decisions.”  RBA cites several examples of why QRA is a flawed model, including the inappropriate data set, poor predictive capabilities, inadequate correlation, inaccuracy, unpredictable annual recalculations, and arbitrary expense limits that “bear no relation to a fact-based consideration of whether expenses targeted by the model are prudent.”

Imagine for a second an alternate reality where QRA is not fatally flawed… Could it work if the data set was appropriate, or if the correlations were stronger? What if the expense limit was set at the ninety-fifth percentile, or the FCC planned to recalculate every five years instead of every year? Would it be OK then? According to RBA—probably not, simply due to the nature of models. RBA explains that even if the methodology wasn’t “pervasively flawed,” it would still be impossible to predict the costs of providing universal service “under any and all circumstances in each and every instance.” To remedy this problem, RBA recommends the FCC include an alternative process “based on actual costs available to address any instance where the model and resulting benchmarks are not adequate to ensure the provision of universal service.”

RBA provides some insightful—and scary—comments about the investment outlook for the rural telecom industry as a result of QRA. For example, “The Order has made it impossible for small rural telecom businesses serving rural communities to plan their investments and operations to serve their communities;” and “Investment has been chilled and job cutbacks are being planned.” It is fairly easy to imagine an RLEC approaching a lender for financing for a fiber build and saying “Well, we just don’t have any clue what our federal investment recovery will be two years from now because the rules make it impossible to predict,” and a lender responding with “Sorry, no cash for you.” In what alternate reality does a lender approve financing for a company that cannot produce reliable investment recovery data for more than one year?

Yet, the FCC contends that QRA, and the other USF/ICC reforms, are intended to insert more predictability into an already-fairly unpredictable HCLS methodology. According to RBA, the fact that the current HCLS rules are unpredictable is not a valid reason to implement QRA: “The Commission’s defense of the lack of predictability of its proposed mechanism essentially says, ‘So what? You rural carriers already have to live with an unpredictable HCLS mechanism as a result of the cap we imposed!’”

Overall, unpredictability aside, lenders will also be deterred by the fact that a single carrier’s investment and expense recovery will be contingent upon other, unknown carriers’ spending. In our alternate QRA reality, should a company be armed with the business plans of all of its similarly-situated companies when it approaches a lender? According to RBA, “It is impossible for any carrier, financial advisor or financial institution to predict what the aggregate rural carrier expense decisions will be under the proposed model or how the changes in expenses encouraged by the model in any one year will impact the limitations developed by the model in future years.”

Nothing illustrates how disassociated QRA is from the real world like real world examples, and the RBA provides two provocative case studies in the reply comments. The case studies illustrate how companies who make reasonable network investments based on unique real world geographic and market realities could face tremendous financial punishment simply for making seemingly prudent engineering decisions. In the cases of Ellijay Telephone Company (Georgia) and Cordova Telephone Cooperative (Alaska), preliminary analysis of QRA caps actually shows that these two companies would be safe from significant HCLS cuts if they had invested more money in technologies not well suited for their geographic areas.

In the first case study, Ellijay Telephone Company is shown to have “prudently invested in a distributed network architecture of host-concentrators which allows the company to provide the shorter loop lengths necessary for robust broadband deployment,” which was the most cost-effective engineering decision given Ellijay’s service area. This engineering decision will result in a higher Category 4.13 (COE 4.13) Subscriber Transmission Equipment costs relative to Category 1 Exchange Cable and Wire Facility (CFW 1) costs, and “as a result of this prudent investment made on the basis of a rational engineering decision, however, the quantile regression analysis punished Ellijay’s prudent use of resources,” to the tune of a 51% reduction in HCLS. The real kicker is that Ellijay is shown to be 13.8% below the ninetieth percentile in total spending, and even if the company had spent $12.5m more in CFW 1 it would have been safe. RBA observes, “As a reward for Ellijay’s efficiency in deploying broadband, the Commission…would deprive Ellijay of 51% of its High Cost Loop Support.” In Ellijay’s alternate reality where QRA is an appropriate indicator of reasonable expenses, the company would have made inefficient investments in possibly unsuitable technology.

Cordova’s story is similar. Due to the Alaska company’s extremely remote location, it has chosen to use microwave transmitters and towers, which were “cost effective… and in some situations, the only viable deployment option.” Like Ellijay, Cordova’s COE 4.13 Is higher than “similarly situated companies,” even though its CWF 1 is way under—“If CTC had spent $10m in undersea cable facilities to provide telecommunications services to remote areas (as opposed to $3m in microwave transmitters), the company’s High Cost Loop Support would not be limited by the regression analysis model. Instead, CTC’s prudent and efficient investment in microwave transmitters is deemed ‘excessive’ by the quantile regression analysis and results in a 20% reduction in HCLS.” Furthermore, Cordova’s operating costs are actually 30.1% less than comparable companies, but Cordova will have its HCLS clipped and become ineligible for redistributed funding. Once again, QRA “fails to account for real world realities… and punishes CTC for its needed use of microwave transmitters and towers.”

RBA’s “micro” examples of QRA applied to real world situations shows a very disturbing potential consequence of this methodology: companies who have invested considerable time and effort to make engineering decisions based specifically on their service areas will essentially be told by the government that they made the wrong decisions and should be penalized. In the cases above, the companies could have made more expensive, less efficient and less appropriate engineering decisions and not been penalized… and the FCC thinks the new rules will reduce waste, fraud, abuse, and gaming?

QRA appears to exist in an alternate reality where long-standing legal, policy, engineering, economic, and business commonsense simply does not exist. Meanwhile, in the real world, RLECs face a vexing reality where rational investment decisions suddenly become excessive and future investment decisions must be based on the spending of other, completely unrelated companies. Is this a reality RLECs want to operate in?