Economists Suggest Tweaks to Proposed Method for Limiting USF Reimbursements

"Although well-suited to the problem at hand, there are some issues which when addressed, will make a more compelling case for using this methodology to identify high-cost carriers, set cost limits, and spur efficiency,” said Paroma Sanyal, an economist with the FCC Office of Strategic Planning. A “major concern” is with the model’s underlying assumptions that “the quantile of the sums equals the sum of the quantiles,” Sanyal wrote. In other words, by capping the individual cost component at the 90th percentile, the model implies that the total cost would be capped below the 90th percentile. That’s not necessarily so, she said. “Applying the quantile regression to the individual cost components may miss some high cost carriers, or mislabel others as high cost.” Such individual cost capping, which ignores complementary or substitutability between various cost components, “may discourage a company from overall cost-minimization if that means that after minimization, one of the cost categories will fall above the 90th percentile threshold, even though the total costs are lower,” she said.

Also of concern is “omitted variable bias, as several important factors that may explain loop cost have not been included in the regression,” Sanyal said. For example, bedrock in the construction area, soil type, the presence of roads and streams, amount of rainfall and number of frost-free days can all influence the cost of loop construction, she said. The Nebraska Rural Independent Companies’ capital expenditure study made a “fairly compelling case for including these variables,” the economist wrote.

Sanyal also suggested that loop length, or “subscribers per mile of loop plant,” may be a better predictor of costs than simple loop counts. “One could argue that there could be cases where carriers in very rural and sparsely populated areas are forced to build longer loops to serve customers, as opposed to carriers who serve relatively denser areas,” she said. “Arguably, the cost of the one long loop will be greater than the cost of a short loop, and thus using the number of loops as a covariate distorts the cost predictions on the long-loop carrier."

Media Bureau Chief Economist Tracy Waldon said the method “rests on a sound theoretical footing,” but the order “does not make a convincing argument that the existing explanatory variables are sufficient to adequately determine similarly situated study areas.” A “more convincing” presentation would begin with a detailed discussion of each of the cost categories, and factors likely to drive that cost, she said. “Cost categories for operating expenses may be heavily influenced by the prevailing wage, while the cost categories for capital expenditures may not be influenced by labor expenses at all."

"It’s bizarre,” said Mosaic Telecom CEO Rick Vergin of what he called the “arcane quantile regression formula.” The Wisconsin LEC found itself running afoul of the 90 percent cutoff, and has canceled projects for want of funding (CD March 9 p3). Mosaic’s exchanges butt up against another co-op whose costs per line are much higher on a per-line basis, but that co-op wasn’t as affected by the new rules because it was grouped with a different set of “similarly situated” companies, he said. “We're next-door neighbors."

In implementing the order, the bureau should pay careful attention to the methods used to select the explanatory variables, and specifically should distinguish between a variable’s statistical significance and its economic significance, Waldon said. “The process by which firms produce telecom services is fairly well-known,” she said. “Existing knowledge about that production process from engineering models and studies may provide the best guidance in regards to which factors are the most significant cost drivers."

Office of Management and Budget guidelines require that when a federal agency relies on scientific information in a rulemaking proceeding, it be subject to peer review “to enhance the quality and credibility” of the scientific information, according to the order.

"We are being responsive to all comments, whether they are from peer reviews or stakeholders, and want to be transparent about the information and input we receive,” a bureau spokesman said, adding that the bureau invites comments on the peer review findings. “Although the formal comment deadline has passed, ex partes are included in the record and we are considering all comments."

The bureau is considering the use of several additional data sets in the regression analysis, said Patrick Halley, legal counsel for the Wireline Bureau. Possible data sets include USDA soil surveys and “plant hardiness” maps, census data on road information, U.S. Fish and Wildlife Service on climate information and a Department of the Interior survey on topology information.