SENTELLE, Circuit Judge:

Section 276 of the Telecommunications Act of 1996, that comprehensively amended the Communications Act of 1934, see Telecommunications Act of 1996, Pub.L. No. 104-104, 110 Stat. 56 ("1996 Act"), concerns payphone services. It requires the Federal Communications Commission ("FCC" or "Commission") to promulgate regulations to "establish a per call compensation plan to ensure that all payphone service providers are fairly compensated for each and every completed intrastate and interstate call using their payphone." 47 U.S.C. § 276(b)(1)(A) (Supp. III 1997). Petitioners representing various interests of the payphone industry seek review of the FCC's third attempt at a sustainable per-call fee plan to fulfill its § 276 obligations. We hold that the FCC's order withstands scrutiny under the Administrative Procedure Act. See 5 U.S.C. § 706 (1994).

I. Background

This case is before us for the third time. In two previous orders, the FCC has attempted to develop and justify a percall fee for coinless calls from payphones. See In re Implementation of the Pay Telephone Reclassification and Compensation Provisions of the Telecommunications Act of 1996, 11 F.C.C.R. 20541 (1996) ("First Order"); In re Implementation of the Pay Telephone Reclassification and Compensation Provisions of the Telecommunications Act of 1996, 13 F.C.C.R. 1778 (1997) ("Second Order"). Acting on previous petitions for review, we have twice remanded the Commission's determinations for a lack of reasoned decisionmaking. See Illinois Pub. Telecomms. Ass'n v. FCC, 117 F.3d 555, 558 (D.C. Cir. 1997) ("Payphones I"); MCI Telecomms. Corp. v. FCC, 143 F.3d 606, 607 (D.C. Cir. 1998) ("Payphones II"). Today we consider petitions challenging the FCC's third order on the subject. See In re Implementation of the Pay Telephone Reclassification and Compensation Provisions of the Telecommunications Act of 1996, 14 F.C.C.R. 2545 (1999) ("Third Order").

Historically, only local phone service providers (local exchange carriers or "LECs") provided payphone services. The development of so-called "smart" payphones in the mid 1980s allowed independent payphone service providers ("PSPs") to compete with the LECs. PSPs obtained their revenues from either coin calls or from contracts with interexchange carriers ("IXCs" or operations services providers, "OSPs") for collect calls and calling card calls. See Payphones I, 117 F.3d at 558-59.

Before the 1996 Act was passed, PSPs were largely uncompensated for a third type of payphone call: "dial around" coinless calls, where the caller uses a long distance carrier other than the payphone's presubscribed carrier. "Dial around" coinless calls include toll-free calls to long distance providers (such as 1-800-CALL-ATT), and the 10-10-XXX type of calls. See id. at 559. PSPs are prohibited from blocking these dial around calls. See Telephone Operator Consumer Services Improvement Act of 1990, Pub.L. No. 101-435, 104 Stat. 986 (codified at 47 U.S.C. § 226 (1994)). In § 276 of the 1996 Act Congress addressed the problem of uncompensated calls by requiring the FCC to "establish a per call compensation plan to ensure that all payphone service providers are fairly compensated for each and every completed intrastate and interstate call using their payphone." 47 U.S.C. § 276(b)(1)(A) (Supp. III 1997). The statute directs the Commission to prescribe regulations "[i]n order to promote competition among payphone service providers and promote the widespread deployment of payphone services to the benefit of the general public" to meet this end. Id. § 276(b)(1).

The FCC decided that the best way to ensure fair competition was to allow the market to set the price for each call. See First Order, 11 F.C.C.R. 20541 p 70. But because no market has previously existed for dial around coinless calls, the Commission first adopted a market-based surrogate — the price of a local coin call at a typical deregulated payphone of $.35. In imposing this rate, the FCC simply said that the "cost[s] of originating the various types of payphone calls are similar." Id.

Various parties sought review of this part of the Commission's decision, as well as several other portions of the First Order. See Payphones I, 117 F.3d at 563-64. We remanded the coinless call rate determination because the Commission had ignored record evidence that the costs of coin calls and coinless calls are not similar. See id.; see also Illinois Pub. Telecomms. Ass'n v. FCC, 123 F.3d 693, 694 (D.C. Cir. 1997). For example, numerous IXCs had noted that coin calls cost more than coinless calls because of the typical costs of using coin mechanisms in payphones. We concluded that "[t]he FCC's ipse dixit conclusion, coupled with its failure to respond to contrary arguments resting on solid data, epitomizes arbitrary and capricious decisionmaking." Payphones I, 117 F.3d at 564 (citing Motor Vehicle Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 46-57 (1983)).

On remand, the FCC attempted to develop an actual market-based rate for coinless calls. See Second Order, 13 F.C.C.R. 1778 p 29. The Commission used the deregulated coin market rate as a starting point ($.35), and subtracted $.066 per call as representing the difference between coin and coinless calls, resulting in a per call rate of $.284. See id. p 41-42. On appeal, we again found error in the agency's decisionmaking. See Payphones II, 143 F.3d at 608-09. We faulted the Commission's failure to explain why the coinless market rate could be found by simply subtracting costs from coin call rates: "If costs and rates depend on different factors, as they sometimes do, then this procedure would resemble subtracting apples from oranges." Id. at 608. We noted that although the Commission "may have depended on the premise that the market rate for coin calls generally reflects the costs of those calls," it had failed to articulate its assumptions and connect them to its reasoning. Id. We remanded for further proceedings. See id. at 609.

The Commission went back to the drawing board one more time. On February 9, 1999, the FCC issued its Third Order, which we now review. The FCC switched from the "topdown methodology" of the Second Order to a "bottom-up" method, meaning that it started from zero and added up the costs of coinless calls to develop a coinless call rate. See Third Order, 14 F.C.C.R. 2545 p 13. The resulting new rate is $.24.

Briefly put, the Commission first determined the "joint and common" costs of a payphone; that is, the monthly capital expense of a payphone, using the cost of a typical payphone and accoutrements. The FCC did not include the cost of a coin mechanism in this figure because it determined that that cost is only necessary for coin calls, but did include amounts as joint and common costs for monthly line charge costs, maintenance costs, overhead costs (known as Sales, General, and Administrative Costs or "SGA"), and coding digit costs. Total monthly costs per payphone came to $101.29.

To translate total monthly costs into a per call rate, the FCC divided that figure by the average number of calls received by a marginal payphone. A marginal payphone is one that gathers revenue to meet its costs (including an assumption that the payphone does not pay location rent to the owner of the premises because of its marginal status) but is not otherwise profitable. Relying on data submitted by the Regional Bell Operating Companies Coalition ("RBOC Coalition"), the FCC came up with a figure of 439 calls per month. This number represents the midpoint between 414, where the data showed that a premises owner would not need to subsidize a payphone in order to keep it, and 464, where the data showed that location rents would be typically required by premises owners. The Commission declined to rely on other data which used call volumes from an average payphone because it would cause many payphones with below-average call volume to become unprofitable.

This yielded a per call figure of $.231 ($101.29 divided by 439, rounded to the nearest one-thousandth). The FCC adjusted the figure upwards $.009 to cover the interest associated with having to wait for payment from IXCs, for a grand total of $.24. The FCC declined to add additional amounts to the dial around rate for bad debts and collection costs associated only with dial around calls.

Two groups of petitioners again seek review of the FCC's determination, raising multiple issues. The first, representing the interests of PSPs, claims that the final rate is too low. The other, representing the interests of IXCs, claims, not surprisingly, that the final rate is too high. Each interest group has also filed briefs intervening in the petitions of the other.

II. Analysis

Although the petitions from the First Order were more wide-ranging, the area of dispute has now narrowed to the coinless call rate. PSPs and IXCs raise a number of objections to the Commission's order on that subject. Although we have given attention to each, only three are sufficiently weighty to warrant separate discussion in this opinion: (1) the FCC's failure to include a bad debt figure in the coinless call rate, (2) the FCC's failure to include a separate figure to account for collection costs associated with coinless calls, and (3) the decision to use data based on marginal rather than average payphones. In considering those three objections, along with those which we do not separately discuss herein, we apply the standard of review drawn from the Administrative Procedure Act and uphold the Commission's determinations unless they are "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law." 5 U.S.C. § 706(2)(A) (1994); see, e.g., Achernar Broad. Co. v. FCC, 62 F.3d 1441, 1445 (D.C. Cir. 1995). Each of the decisions questioned by petitioners herein survives review under that standard.

A. Bad Debt

As we noted above, the Commission declined to add any amount to the coinless call fee for bad debts associated with the collection of coinless call fees. The PSPs, before the FCC, advanced arguments based on their own alleged bad debt experience, and now argue that the FCC should have been able to calculate some amount for inclusion in the coinless call rate based on that evidence. Cross petitioners argued before the Commission, and here, that the debts on which the proffered evidence was based were either the result of PSP negligence in collection, or do not genuinely represent bad debt losses at all, but only unresolved billing disputes. The Commission concluded that it had insufficient information about the levels of bad debt to enable it to rationally calculate an appropriate figure for inclusion.

Specifically, the Commission found that the data regarding uncollected per-call compensation was not reliable enough to predict accurately future levels of bad debt. See Third Order, 14 F.C.C.R. 2545 p 162. The Commission noted that it could not determine what percentage of uncollected per-call compensation was the result of PSP billing errors (i.e., not charging the correct IXC), as opposed to deadbeat carriers (i.e., the appropriate party is billed but refuses to pay). The Commission further noted that providing an improperly computed allowance for uncollectibles could result in double recovery if the PSP ultimately collected from the delinquent carrier. That is, the PSP would collect once from the IXC and once from the consumer (through the bad debt cost element included in the higher compensation amount). Finally, the Commission determined that a bad debt allowance was unnecessary because the agency had ensured in the Third Order that PSPs will receive interest on late payments for as long as such payments are overdue. In short, with insufficient information, the Commission found "that it would be unwise to establish a cost element for bad debt at this time." Id.

The PSP petitioners argue that the Commission was required to include some estimate of bad debt in its calculation and that the failure to do so "effectively determin[es] that dial-around uncollectibles would be zero." (The PSPs rely on some of the same data that the Commission deemed not sufficient to allow a rational decision.) We disagree.

Perhaps the FCC could have formulated some best-guess figure for bad debt, but we cannot require an agency to enter precise predictive judgments on all questions as to which neither its staff nor interested commenters have been able to supply certainty. "Where existing methodology or research in a new area of regulation is deficient, the agency necessarily enjoys broad discretion to attempt to formulate a solution to the best of its ability on the basis of available information." Industrial Union Dep't, AFL-CIO v. Hodgson, 499 F.2d 467, 474-75 n. 18 (D.C. Cir. 1974) (citing Permian Basin Area Rate Cases, 390 U.S. 747, 811 (1968)); see also FCC v. National Citizens Comm. for Broad., 436 U.S. 775, 813-14 (1978). That is exactly the situation the FCC faced here. The agency was presented with bad debt data culled from a relatively short historical period, while knowing that some of the factors affecting that data may change in the future. Any figure that it might have chosen to represent bad debt would likely be challenged on that and other similar evidentiary bases. We conclude that it was prudent and reasonable for the Commission to decide that, on balance, the existing bad debt data was not reliable enough to warrant any educated guess as to future bad debt percentages. It may not have been the only decision it could have made, but it was a reasonable one under the circumstances.

In upholding the reasonableness of the Commission's exclusion of the bad debt element from coinless call cost, we are mindful of the nature of the debt involved. As intervenor long distance carriers remind us, the "[f]ailure to pay the required compensation is a violation of FCC rules for which the carrier is subject to damages as well as fines and penalties." See 47 U.S.C. § 206-08, 501-03 (1994). The plight of the allegedly uncompensated payphone service provider does not equate to that of a merchant pursuing deadbeat customers in the marketplace. Furthermore, for any harm that may be done to the PSPs, they are not left without remedy. After noting that it was "unable to generate a sufficient record on this question for issuing this Order," the FCC invited the parties to file petitions for clarification on the bad debt issue. Third Order, 14 F.C.C.R. 2545 p 162. The RBOC Coalition has made such a filing; the Commission has received that petition; sought and received comments; and, is considering the issue. See Common Carrier Bureau Seeks Comment on the RBOC/GTE/SNET Payphone Coalition Petition for Clarification Regarding Carrier Responsibility for Payphone Compensation Payment, CC Docket No. 96-128, DA 99-730 (1999), available at 1999 WL 335783.

B. Collection Costs

The Commission's calculation of the joint and common costs of a payphone include a figure representing "Sales, General, and Administrative (SGA) costs." Third Order, 14 F.C.C.R. 2545 p 178. SGA includes "overhead costs, such as legal fees, administrative costs, salaries, and management costs." Id. The FCC reasoned that as the proportion of coin calls changes as compared to coinless calls, more employees in a payphone company would likely take on duties related to the busier type of call traffic, but that the overall overhead costs should remain the same. The Commission considered data on the subject filed with it before the issuance of the Second Order and data provided by the RBOC Coalition in the present proceeding. Based on its review of the evidence, the Commission determined that a reasonable estimate of SGA costs on a per-phone-per-month basis was $19.62. See id. p 178-79.

The Commission included this SGA figure in calculating the coinless call cost but did not include in the coinless call rate any additional amount to account for the marginal costs of billing and collection of coinless fees. See id. p 163-64. The FCC reasoned that it had insufficient information with which to determine the variance of administrative costs which occur from a rise in coinless calls relative to coin calls. See id. p 164. It stated that "it [is] fair to assume that the amount that coin-related SGA positions contribute to SGA expenses approximate the same expense that billing and collection positions contribute to SGA." Id.

The PSPs claim that record evidence showed considerable actual expenses in the collection process. In their view, SGA costs cannot be counted as covering these expenses because coinless call collection costs are properly viewed as an incremental expense of coinless calls, not a joint and common cost of payphones.

We again disagree. It is plausible to reason, as the FCC did, that the percentage of SGA overhead costs which can be traced to coinless call business will increase in the future if the market embraces coinless calls. Before the advent of dial around call compensation, overhead necessarily constituted costs attributable only to the prior forms of payphone compensation. As the payphone service market shifts between coin calls and coinless calls, it is reasonable to expect that the relative portion of overhead attributable to separate underlying elements of expense will change with it. This does not mean that either the Commission or the regulated entities should expect to undertake a perennial and constant adjustment of cost allocation based upon that moving target. The use in accounting of the concept of "overhead" presupposes that some details of costs will be submerged in that greater item of calculation. If this were not the case, and if the PSP's argument were accepted and taken to its logical extreme, we would be forced to conclude that virtually every dollar characterized as overhead should be treated by the Commission as either a cost of coin calls or coinless calls. But the collective concept of overhead prevents us and the Commission from having to determine that because a data input employee of a PSP spends ten percent of the time at her computer on coinless call matters and ninety percent on coin calls, the cost of her mousepad should be divided on a one-to-nine basis between those expense categories rather than classified as overhead. The FCC reasonably did not go down that detailed a path, and therefore did not act arbitrarily, capriciously, or contrary to law in deciding that the collection costs of dial around compensation are fairly represented by the SGA portion of joint and common costs.

C. Marginal Payphone Methodology

The FCC based its calculations on the number of calls from a marginal payphone — a payphone that breaks even — to ensure fair compensation under § 276(b)(1). The Commission wanted to ensure the "widespread deployment of payphones" as required by the statute, and declined to use average payphone call volume because that would render below average payphones unprofitable. Third Order, 14 F.C.C.R. 2545 p 141.

To determine the number of calls a marginal payphone receives, the FCC requested that the RBOC Coalition provide two figures: (1) the number of calls placed at a phone that does not pay rent, and (2) the number of calls made from a location that begins to pay rent. The two numbers reported back were 414 and 464, with a midpoint of 439 which the FCC adopted.

The IXCs fault the FCC for relying on the RBOC Coalition data. They claim that the data cannot be used because the RBOC Coalition did not explain their underlying methodology for developing the data. In City of New Orleans v. SEC, 969 F.2d 1163 (D.C. Cir. 1992), we found error in an agency's reliance on estimates which had "no explanation or underlying support." Id. at 1167. However, that is not the case here. The RBOC Coalition did explain how it developed the data, and noted certain difficulties it had in doing so. For example, it pointed out that average revenue depends in part on factors other than call volume, such as the mix of types of calls and the maintenance expense of specific locations. It explained its attempt to determine the average daily revenue needed to decide to place a new payphone and the average revenue needed to begin paying commissions on such a phone, and then determined what mix of calls will produce that revenue. The RBOC Coalition also explained that the final numbers were a weighted average of numbers submitted by members of the Coalition. While the data submitted by the RBOC Coalition could be subjected to various challenges, we cannot say that it was unreasonable or arbitrary for the FCC in the exercise of its expertise to rely upon it. See Madison Gas and Elec. Co. v. SEC, 168 F.3d 1337, 1344 (D.C. Cir. 1999).

III. Conclusion

In summary, we conclude that petitioners have not established that any portion of the FCC's rate calculation for coinless calls is arbitrary, capricious, or otherwise contrary to law. The errors which required us to remand on two prior occasions have been rectified. The petitions for review are therefore

Denied.