Ex. ____ (BJ-Rebuttal Testimony) Docket No. UT-980311(a) Witness: Ben Johnson Rebuttal Testimony of Ben Johnson Docket No. UT-980311(a) 1 ATTORNEY GENERAL OF WASHINGTON Public Counsel Section 900 - Fourth Avenue, Suite 2000 Seattle, WA 98164-1012 (206) 464-7744 BEFORE THE WASHINGTON UTILITIES AND TRANSPORTATION COMMISSION In the Matter of Determining Costs ) For Universal Service ) DOCKET NO. UT-980311(a) ) Rebuttal Testimony of BEN JOHNSON, PH.D. On Behalf of Public Counsel August 21, 1998 Q. Would you please state your name and address? A. Ben Johnson, 1234 Timberlane Road, Tallahassee, Florida 32312. Q. Are you the same Ben Johnson that earlier provided direct and response testimony in this proceeding? A. Yes, I am. Q. What is the scope of this rebuttal testimony? A. I am responding to certain statements made by the following witnesses: Mr. James D. Dunbar, Jr. for Sprint, Dr. Gregory M. Duncan for GTE-Northwest, and Dr. Curt Huttsell, Dr. Kevin Duffy-Deno, Mr. Peter B. Copeland, and Mr. David L. Teitzl for U S WEST. The fact that I do not respond to other statements made by these and other witnesses should not necessarily be construed as agreement with such unchallenged statements. In addition, I am taking this opportunity to present some brief comments concerning the recently filed multi-line studies. Q. Let’s begin with Mr. Dunbar. In his response testimony for Sprint, Mr. James D. Dunbar, Jr., accuses you of claiming that BCPM sponsors have used more resources on attacking other models than in improving their own. Is this an accurate representation of your statement? A. No, it is not. The passage to which Mr. Dunbar refers is reproduced in full below. Rather than concentrating their resources on improving the accuracy and capabilities of their models, the sponsors of both the BCPM and HAI models have expended vast sums on trying to persuade regulators to reject the other model, and to rely instead upon their own cost estimates. These parties have built-in incentives to gloss over demonstrated problems with their own models, and to present partial or ad hoc solutions as if they were total cures. Fortunately, the models themselves continue to improve, and there is every reason to believe that better, more accurate cost estimates can be developed if the Commission insists upon further improvements. (Johnson Direct, p. 20). It should be obvious that in these remarks I am chiding the model sponsors for diverting resources from the important task of improving their model. Whatever the exact balance, it is obvious they have spent a very sizable fraction of their resources on the advocacy war that has been waging before the FCC and various state regulatory bodies for the past two years. I don’t know whether more resources have been spent on advocacy than on modeling, but I suspect this might be the case, particularly when one considers the combined efforts of all the different ILECs that are advocating BCPM, and when one considers resources spent on selling BCPM as well as resources expended on criticizing the alternatives. This is particularly apparent when reviewing the thousands upon thousands of pages of comments and studies that the incumbent LECs have filed at the FCC criticizing various aspects of HAI and/or arguing the superiority of BCPM. As I clearly state, all of the models -- and that includes BCPM -- are continually being improved. But I have watched Hatfield (HAI) and BCPM proponents flail away at each other in venue after venue and docket after docket. These attacks and counterattacks obviously use substantial resources, and to the extent they do, they dilute (the opposite of concentrate) the resources available to the modelers for more positive and productive activities. By comparing the specific amounts expended by Sprint on technical critiques of HAI relative to the amounts spent on improving BCPM, Mr. Dunbar is ignoring an important component of the overall equation: the vast resources that have been expended on trying to persuade regulators that BCPM is good and HAI is not. Q. Mr. Dunbar also refers to your testimony on pages 3 and 5 of his response testimony, in a passage in which he explains that most network costs, in his chosen example, are common to both single-line service and multi-line service. Is this true? A. In his example it is true. Other examples could be adduced in which costs will differ depending upon whether or not both categories of service are being provided. Mr. Dunbar goes on to say that if a pole is required by both groups of subscribers (single-line and multi-line): Both share in the responsibility of that pole being placed. It is only appropriate for both to carry some burden of the cost. That CANNOT be achieved with the netting process Dr. Johnson espouses. (Dunbar Responsive Testimony, p. 5.) These comments are true as far as they go, but they miss my point. The cost of the pole doesn’t increase if single-line service is added to a network already providing multi-line service. Neither does the cost of the pole increase if multi-line service is added to a network already providing single-line service. Economies of scope exist which cause the incremental cost of either service to be less than its stand-alone cost. The phenomenon we are discussing is not unique to the incremental cost of single-line service, nor does it result from the particular calculation method I have proposed (subtracting the total stand-alone cost of one service from the total cost of providing both services). Rather, the problem is inherent to the incremental cost concept, particularly with regard to the Total Service Long Run Incremental Cost (TSLRIC) of services that are produced in common with other services. It is for precisely this reason that the TSLRIC of custom calling is so low. A typical TSLRIC study shows a cost of custom calling that is mere pennies per month, yet the price is typically several dollars per month. The incremental cost is extremely low because nearly all of the switching and other costs of providing custom calling service are left out of a TSLRIC study, since these costs would be incurred whether or not custom calling service is provided. Similarly, the TSLRIC of access and toll service tends to be extremely low for the same reason. Pole costs, for example, are “netted out” of the TSLRIC of toll and access cost, thereby contributing to the pattern of very low costs that typically emerges from a TSLRIC study for toll or access service. By definition, the incremental cost of an individual service contains none of the joint and common costs that are shared with other services. I am not asking, “How do you allocate shared costs to the individual services?” I am asking, “What is the TSLRIC of single-line service?” For many years, this Commission has no doubt heard parties debate the strengths and weakness of TSLRIC studies, and some parties have argued that they provide valuable information which should be considered in the regulatory process, despite the fact that TSLRIC costs tend to exclude joint and common costs. I am not trying to resolve the joint and common cost recovery issue; I am simply suggesting that the record will be more balanced, and the Commission’s decision making process more complete, if it has an opportunity to compare the relatively low incremental cost of single-line service with the relatively high stand-alone cost of this service, particularly since the latter type of cost data was previously filed by some of the parties. Q. Dr. Curt Huttsell, testifying for USWC, attacks your procedure for indirectly measuring the incremental cost of single-line service. Do you wish to respond? A. Yes. Dr. Huttsell’s first criticism is stated as follows: Dr. Johnson’s definition of [a] network providing multiline [sic] service includes all customers’ second and subsequent lines plus special access lines. Such a network is so hypothetical as to be purely imaginary. It presumes the existence of another network providing only one line to every customer. The companies owning the two different networks would have to enter into extraordinarily complex and costly interconnection arrangements in order to facilitate traffic interchange. (Huttsell Rebuttal Testimony, p. 6, emphasis added.) There are several components of this objection, none of which are valid. First, I have not presumed “the existence of another network providing only one line to every customer” nor is it necessary to make such an assumption. The approach I am advocating does not require a scenario in which “two different networks” are simultaneously in existence and thus, it does not require “ extraordinarily complex and costly interconnection arrangements in order to facilitate traffic interchange” between these two networks. Instead, one simply compares two alternative scenarios. Each of these scenarios can include just one network. All that is necessary is that the two scenarios be comparable with the difference between the scenarios being the focus of the comparison. Certainly, one of these scenarios is purely hypothetical. No one would actually build a network to serve exclusively multi-line customers, nor would anyone build a network to serve exclusively single-line customers. However, it is useful to visualize such a network, in order to understand how costs might incrementally differ between the two hypothetical scenarios. Modeling of incremental costs is not constrained by the specific combinations of goods and services that are actually provided by firms operating in the real world. Nor is the fact that one of the scenarios is a seemingly unrealistic ”hypothetical” of any importance. To the contrary, when dealing with incremental cost calculations, it may be necessary to compare a realistic scenarios with an unrealistic one, in order to isolate the incremental costs of a specific aspect of the realistic scenario. To determine the incremental cost of custom calling, for example, one effectively compares a scenario in which custom calling is provided with one in which it is not provided. Yet today no one in his right mind would actually build a telephone network that does not provide custom calling service. The incremental cost of adding custom calling service is small, relative to the revenues that will be generated by this service; hence, a rational firm would always build a network that included custom calling capabilities. Nevertheless, in order to precisely estimate the incremental cost of custom calling, one compares two scenarios: an unrealistic scenario where the service is not offered, and a more realistic scenario where the service is offered. The difference is the incremental cost of custom calling. Similarly, given the appropriate input data, one can model a network to serve any definable subset of the total customer base -- shut-ins, or the hearing impaired, or senior citizens, or pizza parlors, or Internet Service Providers, just to name a few possibilities. The hypothetical network that excludes this subset of customers would be purely imaginary, of course, and would probably never be built. Still, it is helpful to visualize a hypothetical “stand-alone” network that serves everyone but this narrow group of customers, in order to isolate the incremental cost of the group in question. There is no reason to assume that two different networks would actually be created, one serving the pizza parlors and the other serving everyone else, nor do we need to concern ourselves with how the various customers would talk with each other. We just want to know how much more it costs to add pizza parlors to a network that would otherwise serve everyone else. To find out, we analyze a hypothetical network without pizza parlors, and compare the total cost of that imaginary network to the total cost of a network that includes both pizza parlors and other customers, thereby isolating the extra cost of serving this narrow group of customers. As I explained in my direct testimony, the function of such comparative modeling (e.g., an all-line network vs. a multi-line network) is to isolate the TSLRIC of the service that is excluded from the smaller network. Q. Dr. Huttsell also criticizes “Dr. Johnson’s notion of a network providing both single and multi-line service” (Id., p. 6), claiming that if the network is limited to local service, Ait cannot be observed in reality,” and ignores economies of scope. (Id., p. 7). Is this true? A. Perhaps, but is really irrelevant. By definition, stand-alone estimates are hypothetical and ignore economies of scope. That is what makes them useful in isolating incremental costs since incremental cost tends to be less than average cost due to the impact of economies of scope. I feel compelled to point out that the modeling approach and procedures that Dr. Huttsell is attacking are not original with me but are familiar tools from the economists’ tool kit. Mathematically, there are certain relationships between stand-alone, incremental and total cost which make it convenient to calculate incremental cost by comparing stand-alone and total cost. However, the concepts and procedures I am discussing were not invented by me, nor are they in any way unique to local telephone service, multi-line service, or single-line service. The same procedures and attributes can be demonstrated and applied in many other situations and industries as well. Furthermore, the specific studies he objects to were prescribed in the second of the Commission’s guidelines in its Notice of Prehearing Conference of May 4, 1998, in this proceeding, which was issued before I filed my testimony and to which I refer in that testimony. By implying that these are my inventions, he is attacking the Commission’s conclusions and requirements, as set forth in the Notice, without appearing to do so. As noted in my response testimony, the ALJ has issued an order that largely rejects U S WEST’s arguments that a multi-line study would be misleading and irrelevant. (Second Supplemental Order, Prehearing Conference Order, Docket UT-980311(a), July 7, 1998). Q. Dr. Huttsell also claims that the two proxy models in this case are not suited to estimating total or stand-alone costs, and thus Ayour” indirect estimating procedure is impractical. Is this a valid criticism? A. No. While there are weaknesses in the models which may preclude developing precise estimates of stand-alone or incremental costs, these problems are not so great as to preclude gaining any insight into these different types of cost. By far the bulk of network costs are loop, trunking, and switch costs that can certainly be modeled for specific customer groups. In fact, Sprint has provided BCPM results that attempt to present stand alone costs of single-line or multi-line service. I have used the Sprint figures to derive reasonable estimates of the incremental costs (TSLRIC) of single-line and multi-line service. I have prepared a five-page schedule that shows these stand-alone and incremental cost estimates, along with average cost estimates for a network that includes both single-line and multi-line service. Q. Would you briefly describe your schedules, and explain how you derived incremental cost estimates from Sprint’s figures? A. Yes. Schedule 1 reproduces Sprint’s BCPM estimates of monthly total costs (including both operating expenses and capital costs) for a network that includes all lines, both single and multiple, in each of Sprint’s 30 wire centers. As shown on that schedule, the studies include 87,324 total lines with an estimated total monthly cost of $4,763,339.80, or $54.55 per month per line. The single-line stand-alone study results on schedule 2 reflect Sprint’s estimated cost for a network in each wire center that would serve only single-line customers. Sprint estimated there were 55,632 such lines, and it estimated the total monthly cost of such a network to be $4,029,111.14. This equates to a stand-alone cost of $72.42 per month per line. Q. Sprint’s per-line cost for stand-alone single-line service is substantially higher than the average per-line cost for a network that includes all types of lines. Is this an expected result? A. Yes. As I stated in my direct testimony, in a stand-alone study the stand-alone cost of providing single-line service to customers would involve modeling a network which does not include multiple lines, thereby estimating the cost of providing a single-line to every customer without the economies of scale and scope that arise when multiple lines are provided over the same network. (Johnson Direct Testimony, p. 6.) Thus the substantial difference that can be seen by comparing schedules 1 and 2 reflects the absence of these economies of scale and scope in the stand-alone single-line study. Q. Is the cost of providing multi-line service on a stand-alone basis also higher than the average per-line cost for a network that includes all types of lines? A. Yes. The results of Sprint’s stand-alone multi-line study are shown on schedule 3. Using BCPM, Sprint has estimated the cost per line of a network providing only multi-line service, on a stand-alone basis. A limited network of this type would have a cost per line that is almost twice as high (at $106.32) as the per-line cost of a network providing both single and multi-line service. As I explained in my direct testimony, stand-alone costs tend to be higher than TELRIC [or average costs], because the specified group, class, or area bears the entire burden of the cost of the network infrastructure that is needed to serve that area, class, or group. Joint and common costs, which are spread or allocated across multiple groups or areas in a TELRIC study, are borne entirely by the specific group, class, or area that is the focus of the stand-alone study. (Id.) Q. Sprint did not provide any incremental cost studies. In Schedule 4 of your exhibit you develop an incremental cost study for multi-line service. Would you please explain the procedure you used to develop these cost estimates? A. Yes. For any of the 30 wire centers in the Sprint studies, one can take the total monthly cost shown in the first column of schedule 1, and compare it to the corresponding total monthly cost in the first column of schedule 2. The difference between these two figures is the incremental amount of cost that is incurred as one moves from a scenario in which the network provides only single-line service to a scenario in which the network also provides multi-line service. This incremental change in cost is attributable to the provisioning of multi-line service. Arithmetically, to derive the TSLRIC of multi-line service, one subtracts the total monthly cost shown in the first column of schedule 2, from the corresponding total monthly cost in the first column of schedule 1. The result of this subtraction process is the total incremental cost for that wire center of adding multi-line service to a single-line network. This process is shown on page 1 of schedule 4. To develop incremental cost per line, a few additional steps are necessary. You subtract the number of lines included in the single-line study for that wire center from the total number of lines in the “all line” study for that wire center. The result of this subtraction is the incremental number of lines associated with the incremental total cost that was just developed. This incremental group of lines is the set of lines that are associated with all services other than single-line service (i.e. those lines associated with multi-line service). This subtraction process is shown on page 2 of schedule 4. Finally, dividing the resulting number of incremental lines into the incremental total cost of multi-line service gives you the per-line monthly incremental cost of multi-line service, shown on page 3 of schedule 4. For example, total network monthly cost in the Granger wire center is estimated at $97,803.29 (schedule 1), and the monthly stand-alone cost of single-line service in that wire center is estimated at $91,105.28. So, $97,803.29 minus $91,105.28 equals $6,698.01, which is the estimated monthly incremental cost of providing multi-line service in the wire center (schedule 4, p. 1). There is a total of 1,499 lines in the Granger wire center (schedule 1) and there are 1,219 single-line customers (schedule 2). 1,499 minus 1,219 equals 280 lines in the wire center not associated with single-line service (schedule 4, p. 2). Thus, the monthly incremental cost of multi-line service per-line in the Granger wire center is $6,698.01 divided by 280 ‘ $23.93 (schedule 4, p. 3). Q. In Schedule 5 of your exhibit you develop an incremental cost study for single-line service. Did you use a similar procedure to develop these cost estimates? A. Yes. I used essentially the same process to derive the incremental single-line cost figures shown on page 3 of schedule 5. I subtracted the stand-alone multi-line cost figures from schedule 3 from the corresponding all-line cost figures from schedule 1 and computed the results on a per-line basis, as shown on pages 1 and 2 of schedule 5. Q. There are some discrepancies in the line counts shown in the various Sprint studies. Would you please discuss these inconsistencies? A. Yes. In reviewing the three studies supplied by Sprint, we noticed that the line totals are not internally consistent for some of the wire centers. Logically, if one adds the number of lines in the single-line study to the number of lines in the multi-line study, the total should exactly match the number of lines in the study which includes both single and multi-line service. However, in 28 out of 30 wire centers, if you add the number of single-lines listed on schedule 2 to the number of multiple lines on schedule 3, they do not precisely add up to the total number of lines for that wire center, as shown on schedule 1. For example, whereas Sprint has estimated the total of all lines for Granger to be 1,499, it estimated the number of single-lines to be 1,219 and the number of multiple lines to be 317. Together these categories add to 1,536, indicating a discrepancy of 37 lines. In fact, except for the Bickleton and Willard wire centers, every wire center line count is similarly exceeded in the totaling of the two stand-alone line numbers. The overage averages a bit more than 6%. I do not know the reason for this discrepancy: perhaps it results from some type of rounding error in the BCPM line estimation procedure, or perhaps a minor subcategory of lines has been inadvertently included in both the single and multi-line service categories. While I am not certain where the discrepancy in Sprint’s figures comes from, it appears to be a relatively minor problem which is unlikely to create any substantial bias in the results. In any case, the discrepancies are not so great as to invalidate the studies for the purpose of illustrating differences between the cost per line developed on an average, stand-alone and incremental basis. Q. The incremental costs per line of multi-line and single-line service, across all 30 wire centers, average just $23.17 and $22.19, respectively. Are these figures surprising? A. Not at all. The results shown in Schedules 1 through 5 are quite logical. The cost per line for a stand alone network is the highest, because of the absence of economies of scale and scope. The costs per line of a combined network (similar to TELRIC) fall in the middle. The incremental cost per line (TSLRIC) is the lowest, because the benefits of economies of scope are focused on whichever service is deemed to be the incremental service for purposes of the study in question. In fact, I anticipated this general pattern in my direct testimony: per-unit incremental cost will also tend to be less than per-unit total cost (e.g., TELRIC), since various fixed costs of production will be excluded from the incremental cost calculations. Similarly, incremental costs tend to be less than stand-alone costs. (Id., p. 7.) Q. That brings us back to Dr. Huttsell’s criticisms. He objects that if incremental cost is measured indirectly, as in the studies you have just discussed, “the incremental cost of single-line service would exclude the costs that an ILEC like U S WEST incurs standing ready to provide local exchange service on demand.” (Id., p. 12.) Is this true? A. Of course certain costs are excluded from an incremental study, but this has nothing to do with whether the incremental costs are measured directly or indirectly. By definition incremental cost will exclude those costs that would still be incurred even if the service in question were not provided. Thus, to the extent some joint and common costs are necessary to provide service to one group of customers, and these costs don’t need to be increased or duplicated in order to also provide service to a second group of customers, the incremental costs of serving the latter group will be relatively low. That is an inherent feature of incremental costs. As I said in responding to Mr. Dunbar’s protest along the same lines, it is important not to confuse the issues of cost level (or cost definition) with cost recovery. But Dr. Huttsell goes on to protest that The financial viability of ILECs like US WEST would be endangered by raising all of the revenue necessary to cover the costs of an obligation to serve from multi-line customers alone. (Id.) Nowhere have I proposed anything of the sort. Dr. Huttsell makes a huge and totally unwarranted logical leap from the determination of incremental cost to the assumption that revenue will be set equal to incremental cost. I am not suggesting that revenue recovery be based purely on incremental cost, without considering the need to adequately recover the firm’s joint and common costs. Nor would I suggest that all of the joint costs of two or more services should be recovered from just one of those services (although this approach has sometimes been advocated by others). Q. Do any of the witnesses for LECs recognize that the incremental cost of a specific service does not include joint and common costs? A. Yes. Dr. Gregory M. Duncan, testifying for GTE-Northwest, does understand the principle and concedes that my approach is consistent with Aan economic definition of incremental cost.” (Duncan Response Testimony, p. 32.) Q. Dr. Duncan goes on to criticize your suggestion that pricing based on incremental cost could result in a smaller USF funding requirement. (Id., p. 34.) Are you in fact proposing that the fund not recover joint or common costs? A. This is certainly an option open to the Commission and the Legislature, but I am not specifically advocating it. If other revenue sources were adequate to cover the joint and common costs in question, it is conceivable that a universal service funding mechanism could be established that excludes recovery of these costs. However, as Dr. Duncan notes, I am fully aware of the need for an equitable recovery of all such costs, and I said as much in my direct testimony: Of course, care should be used in interpreting TSLRIC results: joint and common costs tend to be excluded from TSLRIC estimates, yet these costs must also be considered in the pricing process, if total costs are be recovered. (Johnson Direct Testimony, p. 10.) There is no question that joint and common costs must be adequately considered in the regulatory process. The 1996 Telecommunications Act states as follows: The Commission, with respect to interstate services, and the States, with respect to intrastate services, shall establish any necessary cost allocation rules, accounting safeguards, and guidelines to ensure that services included in the definition of universal service bear no more than a reasonable share of the joint and common costs of facilities used to provide those services. (§ 254(f), emphasis added.) The implication is that something less than a reasonable share may be acceptable, though I would not necessarily advocate that. I went on in my testimony to explain why the Commission should have access to incremental cost results: By reviewing a series of different types of cost studies, the Commission will have a full range of economic cost estimates available for consideration in developing its report to the Legislature. As a policy matter, the Commission may determine that some approaches to costing are more appropriate than others; its analysis of these policy issues should not be limited to those options that the ILECs prefer, nor should it eliminate policy options merely because it is difficult to develop the corresponding cost estimates. (Id., p. 11.) Q. How might incremental cost data be used in establishing USF funding levels? A. There may be several possibilities, but one option would be to use incremental cost in setting the support level for services that otherwise not receive any USF support at all. For instance, a lesser degree of support might be provided for multi-line service, (including second and subsequent lines as well as special access lines), based upon the difference between the incremental cost and the revenue benchmark. Since the incremental cost tends to be lower than the average cost, this would reduce the number of wire centers where support for multi-line service would be indicated, and it would reduce the amount of support provided for multi-line services in the remaining wire centers. If the Commission decides that all lines in high-cost area should be supported, it may still want to distinguish between first and subsequent lines in some way, and incremental cost studies provide basis for establishing a lesser degree of support for such lines. Q. Please turn to the statements of Dr. Deno-Duffy. On page 3 of his responsive testimony he says that A[t]he address-geocode success rate is sufficiently low in the rural, low-density areas so that any model must estimate the majority of customer locations in these areas.” Do you agree? A. Only in part. It is true that the geocoding sources used by the modelers in this proceeding have low success rates in the most rural areas. However, that is a problem that can potentially be overcome; it is a weakness in the data sources, and not a fundamental flaw in the geocoding concept. Any customer location can be geocoded, if one is willing to spend some money to gather the data. For example, in many rural areas considerable effort is being expended to create enhanced 911 data bases. As a result, local government agencies and others are naming streets and assigning addresses, among other tasks, so that emergency agencies can precisely locate the source of distress calls. Once the 911 data base has been assembled, it would take rela