Technology Policy Institute White Paper Misses the Big Picture
About a week ago a press release was issued jointly by NECA, NTCA, OPASTCO, the Rural Alliance and the Western Telecommunications Alliance, which angrily refuted the conclusions drawn in a recent white paper published by the Technology Policy Institute (TPI).
After reading the press release, and then the white paper itself, I was still unclear as to how the math was being done—by either side! Much of the argument that author Scott Wallsten put forth in the TPI white paper is virtually indecipherable. But the tersely worded press release issued by industry advocates also gave little support for its claims, other than “the FCC said so.”
So, being the numbers geek that I am, I went back to the source and downloaded the entire spreadsheet of NECA cost data for 2010 and did my own number crunching.
I can tell you straight off that while I am not necessarily an advocate for the existing high cost funding mechanism—more on that below—my results came in far closer to the stats reported by NECA et al than to the conclusions drawn by TPI. I believe the methodology used by Wallsten in his analysis, and therefore many of his conclusions, were faulty, or at the very least, got lost in the minutia and missed the big picture. That said, Wallsten did make a few points that I consider valid.
First, let’s go through the key points made by the TPI white paper. Wallsten states that “A large literature documents the inefficiency and ineffectiveness of these subsidies, raising the question of where the money goes.” He goes on to say that his analysis “reveals that of each dollar distributed to recipient firms, about $0.59 goes to ‘general and administrative expenses’—overhead such as planning, government relations, and personnel—rather than to making telephone service more affordable.” His conclusion is that the USF method for subsidizing service in high-cost areas should be radically overhauled, and the author contends that as the FCC and legislators are considering how to shift support from voice service to broadband, that now is the time to make that overhaul.
The author gives a great deal of background on the history and evolution of the USF throughout his paper, but it’s clear from the start that Wallsten is not a believer in the program and that bias comes out in his evaluation of the statistics. Without citing any specific source, he says, “Economics research overwhelmingly finds that despite its good intentions the existing high-cost program is ineffective, inefficient and inequitable…This hat-trick of policy ineptitude suggests two questions: First why would any reasonable person consider basing support for broadband on such a poor foundation? Second, where has the money gone?” He goes on to snidely say that the first question is rhetorical, “given the nature of the policy process,” but that his statistical analysis seeks to answer the second question.
Wallsten notes that $39b in high-cost support has been paid out by the Federal government since 1998, including around $6b to wired and wireless CETCs. That leaves $33b paid to the incumbent LECs—and, Wallsten wants to know where it went. His conclusion, however, that the money went to firms with bloated overhead for fat salaries and such, just doesn’t hold water.
The paper suggests that “because subsidies are a function of the firm’s costs, the firm has little incentive to reduce its costs.” Wallsten cites a 2006 study (Hazlett) that concluded that rural carriers tend to have higher corporate overhead rates than non-rural carriers. The biggest issue I have with this study, as well as with Wallsten’s paper, is that all of the analyses are done based on per-line data. Wallsten concedes that “It is conceivable that overhead costs per line are inherently higher for smaller firms than for larger until some minimum scale is achieved,” but then he goes on to conduct his own regression analysis based on per line data, arguing that he has adjusted for firm size and therefore drawn a valid conclusion. (Not!)
Specifically, Wallsten examines the relationship between high-cost subsidies and general and administrative expenses, both on a per-line basis. He gives the detailed definition of what can be included in overhead expenses for the purpose of calculating high-cost support and suggests that “the two should be unrelated if universal service funds were being used to subsidize the high costs of building and operating infrastructure. That is, while it is possible that these per-line overhead costs might be higher for smaller firms, they should not be related to the magnitude of universal service funds received when controlling for firm size.”
The first argument against Wallsten’s conclusions comes right from his own graph. Figure 7, pictured below, shows how the total high-cost payment per line has risen as the number of lines has declined, as overall high-cost support payments have remained relatively constant. But if you look at the plots for general and administrative expenses per line, and the associated benefits, you see that neither line rises as sharply as the high-cost payment line…given that the denominator—total loops—is a constant in the analysis, the implication is that overall G&A and benefits expenditures have actually declined over the period graphed.
Next, Wallsten goes into his regression analysis, which seeks to find a causal relationship between high-cost payment per line and G&A per line. Without his base set of data the table shown in the paper is useless, but I believe his conclusion—“The estimates suggest that each dollar of payments from the high-cost fund is associated with an increase of about $0.59 in a recipient ILEC’s general and administrative expenses. Of that $0.59, about $0.24 goes to benefits paid by the ILEC related to those expenses.”—is relatively easy to refute.
As I noted above, I think that there are two major flaws in this analysis. First, results are driven by per-line statistics rather than overall figures for support and expenses. The author does acknowledge this possibility when describing the counterintuitive results of his analysis, but then calls it unsatisfactory. It’s satisfactory! The numbers don’t tell the whole story when you only look at them on a per line basis. Second, his analysis ignores plant specific expense and depreciation expense! Considering the USF money is designed to subsidize plant specific expense, I would think you might want to start there!
Which is what I did, and then I also took a look at the overall picture. A note on methodology: where Wallsten studied the data for roughly 1,400 study areas over several years, my necessarily more simplistic approach looked at just the more than 1,000 study areas which use the cost-basis for their funding calculations, and I only looked at 2010 data (for calendar 2009).
First, the big picture. Of the 1,028 study areas included, only 699 of them received high-cost funding, leaving 329 which collected no such funding. Those that received funding collected $933m in annual support. The total sample reported more than 62m loops, of which those collecting support served about 11.1m loops. All in, therefore, the average annual cost per loop was $15; for those receiving support, the average was $84 per loop.
Now we come to the part where we find that “missing” $33b that Wallsten was trying to find. The more than 1,000 firms I examined had an aggregate gross telecom plant in service of nearly $272b, the net plant investment in the aggregate was $55.4b. Looking just at those who received support, gross plant was nearly $60b while the net investment was just more than $16b. That implies about $43b in depreciation. I know from our recent analysis of ILEC reinvestment that public ILECs are reinvesting less than 100% of depreciation into new capital expenditures. In fact, they reinvested about 72% in the twelve month period ended September 30, 2010 and slightly less in years prior to that. Lo and behold, rounding to an estimated 75% reinvestment rate over the past several years, and multiplying by that $43b in depreciation and you get just more than $32b invested in capex—and that’s over fewer years than all the way back to 1998. So is Wallsten not finding the $33b of "missing" support simply because he is not considering the cost of the network?
That’s an admittedly over-simplified calculation, but the point is that depreciation is a real cost, even when it overstates new cash investment.
Next I looked at aggregate expenses, and broke them out by corporate expenses (including the benefits Wallsten is so fired up about), plant expense and depreciation expense. I found that the overall group had more than $32b in expenses, against which they collected $933m, or less than 3%. Looking just at the companies that received funding, more than $7.5b in total expenses were incurred, implying that the support covered roughly 12.4% of total expenses.
If you look only at cash expenses (excluding D&A) for those receiving support, approximately 22.1% was covered by high-cost support payments. That means that nearly 78% of cash expenses had to be supported by internally generated funds, in addition to that pesky capex line.
Valid Points Raised in the Wallsten Analysis
As I said right from the start, I didn’t set out to prove the perfection of the existing high-cost subsidy program with this analysis, and despite the many flaws I see in his white paper, Wallsten did raise some valid points and/or statistics worth noting.
First of all, he cited a 2000 study (Rosston and Wimmer) which estimated that elimination of the high-cost fund entirely would result in telephone penetration falling by just one-half of one percent—and that was before cord cutting became a dominant trend. Given the multitude of options voice (and even broadband) customers now have in many markets, it’s not a given that the system is necessary anymore to fulfill its original mission: to ensure that all Americans have access to affordable service. While I don’t think yanking the rug out from under the incumbents is fair, it must be acknowledged that their role has changed and the methods for funding high-cost markets should be reassessed.
Second, the paper points out that CETCs have been allowed to collect subsidies based on the costs incurred by the incumbent rather than their own cost basis. No argument here!. This has never made sense, nor does the notion that more than one competitor per study area should receive government subsidies to ensure “universal” service.
A more competitive process which would allow the most efficient operator (which may be a function of the technology deployed) to “bid” for the right to the subsidy for a given market might achieve the desired result of reasonably priced service to all Americans, at a lower cost to those users. Remember "Name That Tune"? "I can serve this market with this many support dollars"..."I can serve this market with even fewer support dollars"..."Serve that market!" Important here is the need to make sure that regardless of the new system ultimately fashioned, the existing recipient companies are provided with a transition process that can allow them to succeed. Also, what happens when nobody wants to serve a particular market?
This won’t be popular among many readers, but then, neither were Blair Levin’s comments last month at the NTCA convention in Dallas…times are getting tough for ILECs and the days of the comfortable, government protected monopoly are gone. But that doesn’t mean that regulators should be using a faulty study as support material while they work to figure out a new system!