In the Matter of	)
	)
Price Cap Performance Review	)	CC Docket No. 94-1
for Local Exchange Carriers	)
	)
Access Charge Reform	)	CC Docket No. 96-262
	)

By the Commission: Commissioners Quello, Ness, and Chong issuing separate statements.

Paragraph

I. Introduction 1

II. Background and Overview 2

A. Background 2

B. Overview of Revised Price Cap Plan 7

C. Price Cap Regulation and Access Reform 14

III. X-Factor Calculation Issues 16

A. Background 16

B. X-Factor Approaches 19

1. Methods for Estimating the X-Factor 19

2. Direct Approach 27

C. TFP Calculation Issues 29

1. Background 29

2. TFP Models Placed in the Current Record 35

3. Output Index Issues 39

a. Mathematical Construction of Output Indices 39

b. Number of Output Categories 44

c. Weighting of Output Categories 47

4. Input Index Issues 49

a. Capital 49

b. Labor 77

c. Materials 80

d. Weighting of Materials and Labor Indices 83

5. Summary 91

D. Other X-Factor Calculation Issues 95

1. Input Price Differential 95

2. Adjustment to X-Factor for Interstate-Only Activity 107

3. Effect of Universal Service and Other

Subsidy Programs on LEC TFP 117

4. Inclusion of Other Firms in Study 120

5. Consumer Productivity Dividend 122

6. Effects of Access Reform 128

E. Analysis and Prescription 133

IV. Price Cap Structure Issues 144

A. Overview 144

B. Sharing Obligations 147

C. Number of X-Factors 156

V. Updating the X-Factor 163

A. Background 163

B. Discussion 165

VI. Common Line Issues 168

A. Common Line Formula 168

B. Reliance on Forecasted Data 171

VII. Exogenous Cost Issues 173

VIII. Other Issues 177

A. Application of the New Price Cap Formula to Incumbent LEC PCIs 177

B. Video Dialtone Basket 182

C. Miscellaneous Issues 183

IX. Procedural Issues 191

A. Tariff Filing Requirements 191

B. Final Regulatory Flexibility Act Analysis 192

X. Ordering Clauses 196

Appendix A - List of Comments and Replies

Appendix B - Pleading Summaries

Appendix C - Rule Changes

Appendix D - Estimation of TFP Under FCC Rules

1. In this Order, we make significant revisions to our current price cap plan for regulating incumbent local exchange carriers (incumbent LECs) as part of our plan to construct a dynamic regulatory framework to further the new pro-competitive, deregulatory paradigm set out in the Telecommunications Act of 1996 (1996 Act).⁽¹⁾ In conjunction with the Access Reform First Report and Order⁽²⁾ and the Universal Service Order,⁽³⁾ this Order adopts reforms needed to set the stage for the progressive deregulation of incumbent LECs with the development of competition. We adopt a reasonable, challenging price cap plan that effectively requires price cap LECs to reduce inflation-adjusted prices for interstate access services by approximately 6.5 percent annually. This new price cap reflects a more reliable productivity estimate than in past Orders, one that is based on a careful analysis of the rate of growth of incumbent LEC total factor productivity (TFP) and the rate of change of LEC input prices. We also eliminate the sharing requirements of the current rules, which substantially undercut the efficiency incentives of price cap regulation and retained some of the cost-misallocation incentives inherent in rate-of-return regulation. These forward-looking reforms to our price cap plan for incumbent LECs will allow services to be more readily removed from price regulation as warranted by the development of a competitive marketplace.

A. Background

2. Price cap regulation seeks to replicate the beneficial incentives of competition in the provision of interstate access services,⁽⁴⁾ while striking a reasonable balance between the interests of ratepayers and stockholders. Price cap regulation is intended to encourage growth in productivity by permitting incumbent LECs that increase their productivity to earn higher profits,⁽⁵⁾ while at the same time ensuring that interstate access customers share in the benefits of productivity growth in the form of lower rates.⁽⁶⁾ The price cap formula was designed to ensure that "[b]oth carriers and customers will be better off" under price cap regulation.⁽⁷⁾

3. The Commission adopted LEC price cap regulation in 1990 because it found that rate-of-return regulation did not create adequate efficiency incentives for incumbent LECs, and required administratively burdensome cost allocation rules to enforce.⁽⁸⁾ Rather than adjusting prices to allow LECs the opportunity to earn a pre-determined return on interstate investment, price cap regulation directly regulates prices and allows earnings to vary. Under price cap regulation, the ceiling or maximum price a LEC can charge for interstate access services is adjusted annually by a measure of inflation minus an "X-Factor." A separate adjustment is made for "exogenous" cost changes, which are changes outside the carrier's control and not otherwise reflected in the price cap formula.⁽⁹⁾

4. In the 1990 LEC Price Cap Order, the Commission scheduled a review of the performance of the price cap plan, to begin in 1994, to determine whether any revisions or modifications to the plan would be necessary.⁽¹⁰⁾ In the first phase of that performance review, completed in 1995,⁽¹¹⁾ we made several revisions to the price cap plan.⁽¹²⁾ We also concluded, however, that we required a more complete record to resolve several important issues, including how the X-Factor should be calculated in the future,⁽¹³⁾ and whether it would be possible to develop a price cap plan that did not impose sharing obligations.⁽¹⁴⁾ Accordingly, we adopted an "interim plan" in the LEC Price Cap Performance Review and sought comment on additional issues in the Price Cap Fourth Further Notice.⁽¹⁵⁾

5. In that Notice, we sought comment on methods for developing an X-Factor, the appropriate number of X-Factor options, and whether we should represcribe the X-Factor periodically or adopt a method for recalculating the X-Factor annually. We requested comment on sharing, the price cap common line formula, and our exogenous cost rules. We tentatively concluded that the X-Factor should have three characteristics. First, it should provide a reliable measure of the extent to which changes in LECs' unit costs have been less than the change in level of inflation.⁽¹⁶⁾ Second, it should pass through ongoing unit cost reductions to consumers. Finally, the calculation of the X-Factor should be relatively simple and based on publicly available data.⁽¹⁷⁾

6. In the Access Reform Notice,⁽¹⁸⁾ we invited further comment on whether and how we should revise our LEC price cap plan as part of access reform. We sought comment, inter alia, on whether we should adopt a higher X-Factor based on the record developed in response to the Price Cap Fourth Further Notice or on similar, more recent economic studies.⁽¹⁹⁾

B. Overview of Revised Price Cap Plan

7. In this Order, we make significant changes to our interim price cap plan and adopt the revised plan as our permanent price cap regulatory regime for incumbent LECs. Incumbent LECs have distributed their interstate services among four groups of access services, called baskets.⁽²⁰⁾ A price cap index (PCI) limits the weighted average of rate increases for each basket to the rate of inflation minus an "X-Factor."

8. In the original and the interim price cap plans, the baseline X-Factor was based on the average of the short-term and long-term trends in rate reductions prior to our adoption of the original price cap plan in 1990, plus a consumer productivity dividend (CPD) of 0.5 percent. We selected the X-Factor and the CPD so that, at minimum, rates would decline more quickly than they had declined before 1990, and thus would ensure that the first benefits of price cap regulation would flow to access customers in the form of lower rates. In the LEC Price Cap Performance Review, we tentatively concluded that an analysis that directly measured the growth of LEC productivity and input prices would provide a better basis for prescribing an X-Factor.⁽²¹⁾ In the Price Cap Fourth Further Notice, we invited comment on the total factor productivity (TFP) methodology and other alternatives for calculating the X-Factor. We also tentatively concluded that we should base our X-Factor on a TFP-based measure of productivity and an input price differential.⁽²²⁾ We find below that the record supports prescribing a single X-Factor of 6.5 percent, based on our conclusions regarding a reasonable method of calculating LEC TFP and input prices, our findings regarding the input price differential, and our decision to retain the 0.5 percent CPD.

9. In its simplest form, total factor productivity is the ratio of a firm's (or industry's, or nation's) total output to its total input.⁽²³⁾ A firm can become more productive by producing greater output from the current level of inputs, by producing the same level of output from fewer inputs, or through a combination of both. In TFP calculations, output and input are represented by indices. The output index represents the quantities of goods or services produced, and the input index represents the quantities of capital, labor, and materials used in the production of those goods and services. TFP studies most often develop output and input price indices to adjust output and input quantities for the effects of inflation. The development of composite quantity and price indices, and the weighting of these indices in TFP calculations, raise important issues that we decide in Section III.C. of this Order. In addition to these TFP calculation issues, we also resolve issues about whether to adjust the X-Factor for the difference between LEC input prices and input prices for the national economy (an "input price differential"), and about whether to adjust for any difference between interstate and intrastate productivity growth.

10. Our interim price cap plan permits LECs to choose among three X-Factors, two of which include obligations to share certain earnings. Sharing requires incumbent LECs to "share" half or all earnings above specified rates of return with their access customers by lowering the maximum prices LECs may charge during the next year. We tentatively concluded in the LEC Price Cap Performance Review that we should move to a system of pure price caps, without sharing, because we found that sharing tends to blunt the efficiency incentives that we sought to create with price cap regulation.⁽²⁴⁾ We retained sharing in our interim plan, however, because we found that it served three beneficial functions: a "flow-through" function, a "matching" function, and a "backstop" function.⁽²⁵⁾ In the Price Cap Fourth Further Notice, we proposed to eliminate sharing if we found a way to replace these three beneficial functions or if we found these functions no longer necessary to the operation of our price cap regulatory regime.⁽²⁶⁾ The "backstop" and "flow-through" functions were necessary in part because we were not certain that the productivity targets established by our X-Factors were sufficiently challenging.

11. We conclude that, under the price cap plan we adopt today, the beneficial aspects of these functions are outweighed by the benefits of eliminating sharing. As explained in detail below, we consider the X-Factor we adopt today to be based on a much more reliable estimate of incumbent LEC potential productivity gains. Therefore, we have substantially more confidence that this X-Factor will flow through a reasonable portion of LEC productivity gains to access customers. We also find that, because we establish a price cap plan with only one X-Factor, a matching mechanism is no longer necessary. To guard against our new X-Factor requiring individual LECs to charge unreasonably low rates, we will retain our current low-end adjustment mechanism.

12. In the Price Cap Fourth Further Notice, we sought comment on updating the X-Factor annually using a moving average of TFP, or periodically during performance reviews. We decide, in light of the fundamental changes to the marketplace resulting from the new competitive paradigm of the 1996 Act, that the better course is to select a new generally applicable X-Factor, based on the current record, that will remain in place until we change it in a new performance review.

13. We also sought comment on how to revise the common line PCI formula and the exogenous cost rules should we decide to adopt a TFP-based X-Factor. In our companion Access Reform First Report and Order, we are revising the PCI formula for the common line basket to reflect our revisions to common line recovery, and we therefore decline to discuss common line issues further here. We also conclude that our decision to adopt a fixed X-Factor precludes the revision of the exogenous cost rules that we contemplated in the Price Cap Fourth Further Notice.

C. Price Cap Regulation and Access Reform

14. The rules we adopt in this Order are an essential part of access reform. They are necessary to promote, and plan for, the growth of competition envisioned by the Telecommunications Act of 1996. An X-Factor based on TFP and an input price differential provides, with the Consumer Productivity Dividend, a reasonable, challenging target for LEC access prices. Importantly, eliminating the sharing requirement will increase the incentive of incumbent LECs to become more productive and will enable us to deregulate competitive services while noncompetitive services remain under regulation. In addition, eliminating the sharing requirement will remove the incentives that incumbent LECs now have to misallocate costs from services not subject to sharing, such as those no longer subject to price cap regulation, to services that are subject to sharing. A price cap plan without sharing should greatly facilitate our overarching goal of deregulating services that face sufficient competition by making it easier to remove from regulation those services subject to competition.

15. In the Access Reform Notice, we invited comment on increasing the X-Factor, either on the basis of the record submitted in response to the Price Cap Fourth Further Notice, or on more recent economic studies.⁽²⁷⁾ In response to the Access Reform Notice, a number of parties have argued that, in light of the 1996 Act, we should move forward to reform our current price cap plan.⁽²⁸⁾ In this Order, we consider all the comments filed in response to both the Price Cap Fourth Further Notice and the Access Reform Notice pertaining to calculation of the X-Factor and other price cap structure issues.⁽²⁹⁾

A. Background

16. Under price cap regulation, the weighted average of the prices for the services in a given price cap basket, or the actual price index (API), must be less than or equal to the price cap index (PCI). An incumbent LEC's PCIs are adjusted annually pursuant to formulae set forth in our rules.⁽³⁰⁾ The PCI formula consists of an inflation measure, in this case the Gross Domestic Product Price Index (GDP-PI),⁽³¹⁾ minus the X-Factor, plus or minus any permitted exogenous cost changes.

17. In the Price Cap Fourth Further Notice, we proposed to adopt a total factor productivity (TFP) method for deriving the productivity component of the X-Factor, as advocated by USTA, but also sought comment on several other possible X-Factor calculation methods and invited parties to propose additional methods. For instance, we sought comment on AT&T's Historical Revenue Method, which would explicitly set the X-Factor to produce an industry-average rate of return of 11.25 percent.⁽³²⁾ In addition, we considered the Historical Price Method, which would set the X-Factor based on updated versions of the two studies relied upon in the LEC Price Cap Order. The first, the Spavins-Lande study, compared prices for LEC services to price levels for the U.S. national economy between 1929 and 1989; the second, the Frentrup-Uretsky study, examined the trend in LEC prices for switched access between 1984 and 1990.⁽³³⁾ Additionally, we sought comment on combining elements of the Historical Revenue Method and the Historical Price Method, or retaining the interim price cap plan on a long-term basis.

18. In the next section of this Order, we find that the record provides compelling evidence in favor of adopting the TFP methodology. In Section III.C., we address the issues raised by TFP calculations. In Section III.D., we consider X-Factor calculation issues other than those raised by use of TFP, such as the input price differential. Finally, in Section III.E., we find that an X-Factor prescription of 6.5 percent, including a CPD of 0.5 percent, is a reasonable one.

B. X-Factor Approaches

1. Methods for Estimating the X-Factor

19. In the Price Cap Fourth Further Notice, we tentatively concluded that we should base our X-Factor on a TFP-based measure of productivity and an input price differential.⁽³⁴⁾ In line with a majority of the commenters, including Ad Hoc, AT&T, and USTA, who support TFP in some form, we base our X-Factor prescription on productivity growth and input price differential, derived on the basis of the TFP methodology.⁽³⁵⁾ For the reasons discussed below, we conclude that TFP measures productivity growth more accurately than the method we adopted in the LEC Price Cap Order and the LEC Price Cap Performance Review, and more accurately than any other method proposed in the record before us. In the LEC Price Cap Performance Review, we noted that we were forced to reject TFP-based productivity studies because they were not specific to the telephone industry, or because they were based on non-public information.⁽³⁶⁾ Pacific notes that the California Public Service Commission has based its intrastate price cap plan on a TFP model. Pacific cites a recent California Public Utilities Commission (California PUC) opinion finding that TFP lies between 1.8 percent and 2.6 percent.⁽³⁷⁾ We now have before us TFP studies that are specific to the telephone industry and rely on publicly available data. Finally, we note that the Bureau of Labor Statistics (BLS) uses TFP to measure productivity growth in the national economy.⁽³⁸⁾

20. Several parties oppose the use of TFP because they maintain that the X-Factor resulting from this method is lower than the X-Factors in the interim plan.⁽³⁹⁾ We interpret these arguments as opposing USTA's method of calculating TFP, not as objections to the principle of basing the X-Factor on TFP generally. Similarly, ICA opposes TFP because it anticipates that any TFP-based approach will inevitably raise data availability problems.⁽⁴⁰⁾ We find that the record demonstrates that publicly available data can now provide an adequate basis for TFP analysis. We address TFP calculation issues below.

21. We have considered but do not rely on alternatives to our TFP approach. In the Price Cap Fourth Further Notice, we sought comment on alternative methods of calculating TFP, including an econometric estimation method.⁽⁴¹⁾ The only parties commenting in the record on the econometric estimation method opposed it. USTA and NYNEX assert that an econometric estimation of productivity growth sophisticated enough to be economically meaningful would not meet the goal we established in the Price Cap Fourth Further Notice of being relatively simple.⁽⁴²⁾ No party to this proceeding has placed an econometric TFP model in the record. Therefore, we have no basis at this time on which to adopt an econometric estimation of productivity growth to measure TFP.

22. We also decline to adopt the Historical Revenue Method discussed in the Price Cap Fourth Further Notice and supported by GSA and TRA.⁽⁴³⁾ The Historical Revenue Method would set the X-Factor prospectively at the level that would have, in retrospect, produced an industry-wide average rate of return of 11.25 percent under price cap regulation.⁽⁴⁴⁾ Adopting the Historical Revenue Method on a moving-average basis, as GSA recommends, would create substantially similar incentives to those under rate-of-return regulation, because the X-Factor would be explicitly linked to earnings. The Historical Revenue Approach also would re-create many of the administrative burdens of rate-of-return regulation, including a substantial reliance on accurate demand and cost forecasts. In addition, in the Price Cap Fourth Further Notice, we expressed concerns that the Historical Revenue Approach might not provide sufficient incentives for productivity growth, to the extent that increases in industry-wide earnings would increase the X-Factor.⁽⁴⁵⁾ No one has adequately responded to this concern. GSA recommends using a moving average to update an X-Factor developed pursuant to the Historical Revenue Method.⁽⁴⁶⁾ For the reasons set out below, however, we decline to adopt a moving average. TRA supports the Historical Revenue Method because it believes that it would help reduce rates to economic cost levels,⁽⁴⁷⁾ but presents no reasons why a "historical" revenue method better achieves that end than a TFP methodology. In addition, in our companion Access Reform First Report and Order, we reject proposals to adopt prescriptive measures at this time to drive access rates to economic cost-based levels.⁽⁴⁸⁾

23. We also decline to continue using the Historical Price Method developed in the LEC Price Cap Order. None of the commenters supports this approach.⁽⁴⁹⁾ Furthermore, the Historical Price Method bases the X-Factor on historical trends in prices of telecommunications prices relative to the economy as a whole, and thus uses price changes as a surrogate for productivity growth. We find that TFP is a more accurate measure of LEC productivity because it is based on incumbent LECs' actual outputs and inputs.

24. We also reject MCI's alternative to our TFP approach. MCI asserts that LECs electing the 5.3 percent X-Factor, which entails no obligation to share, must have believed that their unit costs (productivity growth plus decrease in input prices) would decrease by at least 8.54 percent. MCI claims that, otherwise, these incumbent LECs would have earned greater profits by selecting a lower X-Factor, notwithstanding the accompanying sharing obligations. Therefore, MCI recommends a fixed X-Factor of at least 8.54 percent.⁽⁵⁰⁾ In response, USTA criticizes MCI's calculations, in part because MCI implicitly assumes that all price cap LECs earned an 11.25 percent rate of return at the time of their 1995 annual access filings. According to USTA, correcting this error results in an X-Factor of 2.85 percent.⁽⁵¹⁾ In reply, MCI filed an ex parte statement agreeing with USTA's methodological point, but arguing that USTA erred in basing its analysis on a 13.78 percent return, the incumbent LECs' rate of return in 1994.⁽⁵²⁾ According to MCI, the price cap LECs' 1994 rates of return are not the correct starting point because the LECs' expected earnings were depressed by two exogenous cost decreases required in the LEC Price Cap Performance Review in 1995.⁽⁵³⁾ MCI contends that, after adjusting the LECs' rates of return to remove the effects of these two exogenous cost decreases, its alternative X-Factor approach produces an X-Factor of 7.9 percent.

25. We conclude that MCI's method is inherently ill-suited for prescribing an X-Factor, regardless of whether MCI's calculation can be perfected. Fundamentally, MCI's alternative does not estimate expected productivity growth, but instead derives an X-Factor based on LEC X-Factor choices that depend critically on the LECs' earnings for a single tariff year. It would not be reasonable to base a long-term X-Factor prescription, as MCI suggests, on short-term LEC expectations. Furthermore, the results of MCI's alternative methodology rely heavily on LEC interstate earnings. For example, LECs choosing the 4.0 percent X-Factor under the interim plan are required to share half of their earnings in excess of 12.25 percent, and all of their interstate earnings in excess of 13.25 percent. As a LEC's sharing obligations increase, its gains from increases in productivity decrease. Thus, if an incumbent LEC expects its interstate earnings to exceed 12.25 percent, and also anticipates that it will increase its productivity, it is more likely to choose the no-sharing 5.3 percent X-Factor than a LEC that expects the same increases in productivity, but forecasts that its interstate rate of return will be 11.25 percent. As we have said consistently in our discussions of price cap regulation over the years,⁽⁵⁴⁾ we achieve beneficial incentives by placing less rather than more importance on LEC interstate earnings. For these reasons, we reject that alternative as a means for prescribing an X-Factor.

26. US West suggests setting the X-Factor equal to the GDP-PI, and thereby freezing the PCIs at their current levels as a means of simplifying the price cap plan.⁽⁵⁵⁾ We reject US West's proposal, because it would not provide access customers with any benefits from productivity growth, and so would not strike a reasonable balance between stockholders and ratepayers.⁽⁵⁶⁾

2. Direct Approach

27. In the Price Cap Fourth Further Notice, we invited comment on replacing the PCI formula completely with a formula based on what we called the "direct approach."⁽⁵⁷⁾ Under the direct approach, the PCI would change by the percentage change in LEC input prices minus the percentage change in LEC TFP. The direct approach eliminates the GDP-PI (or any other measure of economy-wide inflation), nation-wide TFP indices, and nation-wide input price indices needed to calculate the X-Factor in our current PCI formula.

28. We decide not to modify our PCI formula so that the X-Factor can be calculated under the "direct approach," as suggested by Sprint and GTE, among other parties. First, for reasons discussed in Section V. below, we adopt in this Order a fixed X-Factor until the next scheduled performance review. Adopting a direct approach without also adopting a moving average-based method of updating the X-Factor on an annual basis would result in a PCI formula that reduces PCIs by a certain percentage every year. By definition, a direct approach without a moving average would require prices to decrease by the same nominal percentage regardless of whether the national economy is experiencing high or low inflation. Under a direct approach, with the PCI formula updated only in periodic performance reviews, there is no possible mechanism to incorporate an unexpected increase or decrease in inflation that occurs between performance reviews. Retaining a PCI formula that reflects changes in overall prices is more consistent with our decision to prescribe a fixed X-Factor rather than updating the X-Factor on a moving average basis. Second, we agree with AT&T that the direct approach does not simplify the PCI formula nearly as much as Sprint claims, because the approach eliminates only non-controversial terms from the PCI formula, or terms that can be based on publicly available data.

C. TFP Calculation Issues

1. Background

29. In the LEC Price Cap Performance Review, we noted that changes in a firm's costs of producing a unit of output are the product of both changes in the quantity of resources used, i.e., changes in productivity, and changes in the prices paid for those resources, i.e., changes in input prices.⁽⁵⁸⁾ We tentatively concluded that the X-Factor should include both a measure of productivity growth and a measure of input price changes.⁽⁵⁹⁾ In this Section, we consider methods to estimate changes in productivity. In Section D. below, we consider methods to estimate changes in LEC input prices.

30. In general, TFP models measure productivity as the ratio of an index of the outputs of a firm (or industry, or nation) to an index of its inputs over a given period of time.⁽⁶⁰⁾ The growth in productivity is simply the amount by which this ratio changes over time. In these calculations, every effort is made to isolate the real change in productivity from the effects of simple price changes. This is why, in a subsequent section, we consider separately the matter of changes in input prices.

31. A LEC's outputs are the services it provides, and the output index represents the quantities of services provided. For purposes of constructing the output index, quantities of services can be measured directly, based on such measures as minutes of use or number of access lines, or indirectly, by dividing revenues by an index of output prices. Output indices can be developed to represent changes in the quantity of each individual LEC service over time, or services can be aggregated into one or more categories. The categories are weighted, either on the basis of costs or revenues, to make the output index.

32. LEC inputs consist of three major factors of production: labor, materials, and capital services (services provided by plant and equipment). As explained further below, TFP analysis assumes capital services are a fixed proportion of the capital stock.⁽⁶¹⁾ TFP theory and practice estimates the growth in capital services using the assumption that the level of capital services is some fixed proportion of the capital stock available at the beginning of the year. Capital services can be measured as changes in the level of capital stock. Although these factors can be disaggregated further, all the parties presenting TFP models limited themselves to these three input factors. The growth rate of total input index is determined by the growth rates of the capital, labor, and materials input indices, and by their relative by the relative weight given each input index. As discussed below, measuring the growth rate of capital input is a particularly complicated procedure, requiring, among other things, a determination of capital stock and the flow of capital services from capital stock.

33. We have reviewed the TFP models submitted by Ad Hoc, AT&T, and USTA in response to the Price Cap Fourth Further Notice, the comments received in response to the Access Reform Notice, the numerous ex parte filings in both dockets providing additional or updated data or critiques, and the various estimates of TFP and input price differentials. On the basis of our review, we have determined the most reasonable method of performing each step of a TFP calculation. We discuss our conclusions on each of these TFP calculation issues below. We find that no study in the record embodies all the best TFP calculation practices. We then calculate TFP using the most reasonable parts of each TFP study as it was presented by the record. As explained in detail below, we rely primarily, but not exclusively, on the results of that analysis for our X-Factor prescription.⁽⁶²⁾

34. In Section 2., we summarize the results of USTA's, AT&T's, and Ad Hoc's models. In Section 3., we address output index issues. We address issues regarding the capital, labor, and materials input indices in Section 4. Subsequently, in Section D, we analyze other X-Factor calculation issues, such as how to calculate the input price differential, whether to adjust for claimed differences in interstate and intrastate productivity growth, whether to include a CPD, and whether to make adjustments at this time for the access charge reforms we adopt in the Access Reform First Report and Order. In Section E. below, we prescribe an X-Factor of 6.5 percent, based on our analysis of these issues.

2. TFP Models Placed in Current Record

35. USTA has submitted its simplified TFP model. That model is a revision of its original TFP model,⁽⁶³⁾ which was addressed in our LEC Price Cap Performance Review.⁽⁶⁴⁾ USTA supports updating the X-Factor annually on the basis of a five-year moving average. For the nine LECs included in its original TFP study, USTA claims its simplified TFP model results in average difference between LEC and U.S. national productivity growth of 2.9 percent from 1988 to 1993, 3.1 percent from 1989 to 1994,⁽⁶⁵⁾ and 2.7 percent from 1990-95.⁽⁶⁶⁾ USTA asserts that the input price differential is zero, and makes no adjustment for a consumer productivity dividend.⁽⁶⁷⁾

36. AT&T maintains that its TFP-based model corrects errors in USTA's original TFP model.⁽⁶⁸⁾ In response to the Price Cap Fourth Further Notice, AT&T recommends a baseline X-Factor of 7.8 percent, based on estimates of interstate-only TFP and an input price differential, and including a Consumer Productivity Dividend (CPD).⁽⁶⁹⁾ We discuss AT&T's interstate TFP adjustment in Section D.2. below. Later, in its 1997 pleadings, AT&T updated its study with 1995 data, and found an interstate-only TFP-based X-Factor of 9.0 percent from 1985 to 1995, including a CPD.⁽⁷⁰⁾

37. Ad Hoc also adjusts USTA's original TFP model to correct for alleged methodological errors. Specifically, Ad Hoc recommends adjusting TFP to estimate interstate-only productivity, and including an input price differential in the X-Factor. Ad Hoc proposes an X-Factor of 9.4 percent, which is composed of an estimated TFP growth of 6.0 percent for interstate services, and an input price differential of 3.4 percent.⁽⁷¹⁾ Ad Hoc states that adopting all its recommendations except its interstate/intrastate adjustment results in an X-Factor of 6.6 percent.⁽⁷²⁾

38. Ad Hoc submitted its models in the proprietary format of a commercial software program to which we do not have access. The format makes it quite difficult for us to validate its results or to compare them with those of other models in a manner similar to that shown in Section III.E. below. To the extent that Ad Hoc reveals its intermediate results, its input price index appears to suffer some of the same infirmities as USTA's original model, and to exhibit erratic fluctuations. Furthermore, as discussed further below, we find that the revisions Ad Hoc does make to USTA's original TFP model do not improve the model. Specifically, Ad Hoc makes an interstate-only TFP adjustment, recommends making a hedonic adjustment, and does not weight the capital input index on a residual earnings basis. Therefore, we do not give any weight to Ad Hoc's X-Factor estimates. We discuss AT&T's and USTA's models below in greater detail, and we resolve TFP calculation issues on the basis of that analysis.

3. Output Index Issues

a. Mathematical Construction of Output Indices

39. Background. In the Price Cap Fourth Further Notice, we invited parties to recommend appropriate methods for calculating output price indices for TFP studies.⁽⁷³⁾ As noted earlier, output quantities can be measured directly based on such measures as minutes of use or number of access lines, or indirectly, by deriving quantities by dividing output revenues by a price index. In the Price Cap Fourth Further Notice, we identified various potentially relevant mathematical techniques for constructing indices: the Laspeyres Price Index, the Chained Laspeyres Index, the Paasche Price Index, and the Fisher Ideal Index.⁽⁷⁴⁾

40. Discussion. USTA and AT&T both use physical output measurements for certain access service categories.⁽⁷⁵⁾ While AT&T's TFP study measures all output directly⁽⁷⁶⁾ using the Fisher Ideal Index method,⁽⁷⁷⁾ USTA advocates indirect measures for certain outputs. For example, USTA uses deflated revenue to measure special access output, arguing that using special access line counts is too simplistic.⁽⁷⁸⁾ When it measures output indirectly, USTA divides total revenues by output price indices that are based on an approximation of a chain-linked Paasche method, and then creates output quantity indices using the Tornquist index method.⁽⁷⁹⁾ USTA also contends that using physical measures of output in its local service and toll service categories is inaccurate because it treats each local call identically, and does not capture differences such as the time of day of toll calls, or the effects of vertical services. USTA claims this causes AT&T's study to overstate TFP growth by 0.9 percent.⁽⁸⁰⁾

41. We find that, although both methods can be reasonable for calculating TFP growth in most contexts, use of physical output measures is better suited to calculating TFP for purposes of prescribing an X-Factor. Use of physical output measures simplifies the analysis, and USTA has not shown that that method yields results less accurate than use of deflated revenues. Specifically, USTA has not explained why a toll call made during the day should count more than a night or weekend call for purposes of determining output in a TFP study. Furthermore, we disagree with USTA's contention that using physical measures overstates TFP growth because they do not adequately reflect vertical services. We expect that the quantities of vertical services will increase faster than the inputs used to provide those services in the future, because the price cap LECs have only relatively recently deployed the SS7 facilities necessary to provide vertical services widely in their networks. Thus, increased output of vertical services reasonably could occur as a result of such recent investment rather than directly requiring further inputs through new investment. To the extent that new investment does occur, we believe it likely would result in further or additional increases in output beyond the output increases generated by the prior investment. At the same time, since the LECs have begun marketing vertical services only relatively recently, demand for these services is likely to grow. Thus, physical measures of services should produce conservative measures of productivity and productivity growth.

42. In its 1997 comments, USTA claims that AT&T overstates output growth because it measures common line output by minutes of use rather than number of access lines.⁽⁸¹⁾ USTA also criticizes AT&T's model because it derives common line minutes of use for the period from 1984 to 1985 on the basis of an extrapolation of data for the period from 1986 to 1992.⁽⁸²⁾ AT&T replies that its extrapolation is necessary in order to create a consistent series from divestiture to the present, because common line data were not recorded separately from switched access before 1988.⁽⁸³⁾ We find that where both line and minute data are available, converting all common line output to a per-minute basis is not desirable. Therefore, in our staff analysis, we measure end user common line growth on a per-line basis, and carrier common line growth on a per-minute basis. For the period before 1988, switched access minutes provide a reasonable surrogate for carrier common line minutes. Thus, in our staff analysis in Appendix D, we measure output quantities directly on the basis of switched access lines,⁽⁸⁴⁾ special access lines, and switched access minutes of use.

43. As a technical matter, our review of the relevant economic literature indicates that the Fisher Ideal Index is superior to the approximated Paasche chain index and Tornquist Index used by USTA for the construction of deflated revenue quantity indices.⁽⁸⁵⁾ For example, Diewert states that the Fisher Ideal Index is the only index that satisfies twenty well-defined mathematical tests.⁽⁸⁶⁾ We therefore use the Fisher Ideal Index form in our analysis.

b. Number of Output Categories

44. Background. In the Price Cap Fourth Further Notice, we noted that USTA developed output indices for seven categories in its original TFP study. We sought comment generally on whether USTA's output categorization was reasonable, or whether any of USTA's categories should be combined or subdivided.⁽⁸⁷⁾

45. Discussion. Both USTA and AT&T base their output categories on ARMIS 43-02 reporting groups. USTA uses seven categories, while AT&T uses three. We include three output categories in our analysis of the record: local, intrastate toll plus intrastate access, and interstate access. We find that this categorization is sufficiently disaggregated to provide an accurate measure of output growth, and is easy to implement because we have collected data in ARMIS on this basis.

46. USTA, in effect, holds that both we and AT&T should have retained miscellaneous services as a fourth category. The three output categories that both we and AT&T use include the services in six of USTA's seven output categories, but exclude those in USTA's miscellaneous services category. USTA claims that, by excluding these miscellaneous services, AT&T's model overestimates TFP growth by 0.4 percent from 1988-94, and 0.5 percent from 1989-94, because miscellaneous services output has grown more slowly than other LEC outputs.⁽⁸⁸⁾ This apparently slower growth, however, is a direct result of USTA's use of GDP-PI when it calculates output quantities by deflating revenues by a price index. USTA used the GDP-PI because it did not have a specific measure of miscellaneous service prices. Because GDP-PI rose substantially over the period while the prices of LEC services other than miscellaneous services fell sharply, it is obvious that miscellaneous output estimated in this manner would grow more slowly. It is not at all obvious, however, that GDP-PI is an appropriate price index for miscellaneous services. Furthermore, examining the major components of this category reveals that it is a collection of highly diverse activities. Many of these, such as White and Yellow Pages operations,⁽⁸⁹⁾ are at best ancillary to telecommunications services. We also note that the composition of this category varies widely from year to year. Because of these characteristics, we do not believe it is feasible to construct a valid quantity measure for this category. Accordingly, we exclude USTA's miscellaneous services category from our analysis. Moreover, because most of the services in this category appear to be produced using a separate production function from that used to produce telecommunications services, it is not unreasonable to exclude miscellaneous services. For these reasons, we exclude the miscellaneous services output category completely from our output index.

c. Weighting of Output Categories

47. Background. Regardless of whether output quantity growth rates are based on physical measures or deflated revenues, TFP studies with more than one output category must adopt some weighting scheme to combine the categories into a single index. In the Price Cap Fourth Further Notice, we sought comment on the proper weights for aggregating output quantity categories. We observed that USTA's original TFP study used revenue weights for the output index, and we found that this weighting implicitly assumes that the revenue of a service is a reasonable measure of its value. We questioned whether it is reasonable to make this assumption in an industry where incumbent LECs face different levels of competition for their services, and rates diverge to varying degrees from the costs of producing those services. Therefore, we sought comment on alternative weighting schemes for output categories.⁽⁹⁰⁾

48. Discussion. We conclude that, despite the doubts we expressed in the Price Cap Fourth Further Notice,⁽⁹¹⁾ revenue weights are the best weighting method available. In its comments in response to the Price Cap Fourth Further Notice, AT&T recommends weighting the output indices on a marginal cost basis, arguing that revenue weights will not approximate more economically meaningful marginal cost weights until competition has developed further.⁽⁹²⁾ Neither AT&T nor any other party in this proceeding, however, has provided estimates of marginal cost weights. Instead, AT&T uses booked costs as a surrogate for marginal cost weights. BellSouth asserts that using fully distributed costs, such as booked costs, as a surrogate for marginal costs would be unreasonable except in cases where there are no economies of scale, and therefore booked cost weights are inappropriate for calculating LEC TFP.⁽⁹³⁾ In its TFP model, the Interstate Commerce Commission (ICC) concluded that use of revenue weights was unlikely to bias its output index seriously over time.⁽⁹⁴⁾ Finally, we note that AT&T has switched its recommendation from cost-based weights to revenue weights.⁽⁹⁵⁾ Accordingly, we agree with the parties that revenue weights are the most reasonable basis of aggregating output indices.

4. Input Index Issues

a. Capital

(1) Background

49. The capital input index measures the amount of capital services used by the LEC to produce output. "Capital services" represent the contribution capital makes to the production of output. Capital input quantities generally assume that the capital services in a time period are proportional to the stock of capital available in that period. Capital input quantities are constructed for a number of asset categories of plant and equipment.⁽⁹⁶⁾ The development of the aggregate capital input index requires three determinations: (1) the capital stock for each asset category, (2) the capital input quantities from these capital stocks, and (3) the relative weight that each asset category should have in the final aggregate capital input index.

50. Typically, the "perpetual inventory method" is used to develop a constant dollar capital stock. The nominal dollar level of capital stock in the first period, called the benchmark capital stock, is generally derived by adjusting gross booked investment, either by subtracting the associated accumulated depreciation and amortization reserves,

or multiplying by a ratio of market to book value of investment derived from another source. The capital stock for the next period is derived by reducing the first period's capital stock for depreciation, and increasing it by the second period's plant additions that have been deflated by an asset price index.⁽⁹⁷⁾ We discuss this process in detail in subsections (2) and (3) below.

51. Once we have calculated constant dollar capital stocks, we need to measure the capital services that these stocks generated. In the Price Cap Fourth Further Notice, we sought comment on two measures. One measure assumes capital services are a constant proportion of capital stock, and that the growth of capital services is measured by the growth in capital stock. A second measure focuses on "capital consumption," i.e., changes in the level of efficiency in the capital stock over time.⁽⁹⁸⁾ We discuss this issue further in subsection (5).

52. The aggregate capital input quantity is a weighted average of the input quantity of all the capital input categories.⁽⁹⁹⁾ The weights are based on the price, or "rental value" of the capital services provided by each asset category, or in other words, an estimate of what the rental value of those assets would be in an competitive market, if one existed. We stated in the Price Cap Fourth Further Notice that this "implicit rental price" includes the rate of return, the depreciation rate, and tax rates.⁽¹⁰⁰⁾ Below in subsection (2), we decide to have only one capital input index. Nonetheless, issues relating to weighting asset categories are still relevant because the method used to develop weights for aggregating asset categories into a single index are also used to aggregate capital, labor, and materials into the final, single input index. We discuss the weighting of the capital input index relative to the labor and materials indices in subsection (6).

(2) Capital Stock

53. Background. The capital input index for a TFP study requires the calculation of capital stock -- the real (or constant dollar) value of LEC net investment. In the Price Cap Fourth Further Notice, we invited comment on several issues related to the calculation of capital stock. We asked generally whether the perpetual inventory model in USTA's original model was the best method to derive capital stock quantity indices, and if not, what other method would be preferable.⁽¹⁰¹⁾ In particular, we asked whether the benchmark capital stock, i.e., the capital stock level in the first period for the study, should be based on the original cost or current replacement cost of assets.⁽¹⁰²⁾ We also noted that USTA used proprietary telephone plant indices (TPIs) to deflate plant additions to constant dollars, and asked several questions regarding the sources and reliability of USTA's TPIs.⁽¹⁰³⁾

54. In the Price Cap Fourth Further Notice, we also noted that USTA's original model aggregated capital into six asset categories, and then developed a depreciation rate for each category to use in calculating the implicit rental price of capital stocks. We asked whether USTA's six classes were the most appropriate classification scheme, noting that the Commission prescribes depreciation rates for 30 asset categories.⁽¹⁰⁴⁾

55. Discussion. Both USTA and AT&T agree that the perpetual inventory model is a theoretically correct and practical method of constructing capital stocks. Therefore, we have decided to use the Perpetual Inventory Model for calculating capital stocks in our analysis.

56. Both USTA and AT&T use BEA asset price indices to deflate their capital stock additions to constant dollars. USTA, AT&T, and Ad Hoc agree that BEA asset price indices avoid the proprietary issues raised by TPIs based on incumbent LEC data. BEA asset price indices measure the movement of asset prices in the U.S. economy. Although BEA asset price indices do not measure precisely the prices of LEC assets, BEA's indices are sufficiently disaggregated that they can be used to develop a surrogate for LEC capital asset prices. Therefore, we have decided to use BEA asset indices.

57. AT&T uses USTA's original six asset categories, but USTA's simplified TFP model reduced the number of asset categories to three. Although USTA and AT&T use different numbers of asset categories, they have not criticized each other's choices, and no one else has criticized either model on the basis of number of asset categories. In our staff analysis, we have used one asset category, and one depreciation rate, because further disaggregation does not appear to provide a more accurate measure of TFP growth, and one asset category simplifies the calculation.

(3) Adjustments to Capital Stock

58. Background. In the Price Cap Fourth Further Notice, we treated as separate issues measurement of the accumulated depreciation used in the perpetual inventory model used to calculate the benchmark capital stock, and the depreciation rates in the implicit rental price. Upon review of the record, we find that these issues are interrelated, and consider them together here. For example, USTA emphasizes the need for the starting value of capital in the perpetual inventory equation to be consistent with the depreciation assumptions used elsewhere in the study.⁽¹⁰⁵⁾

59. In the Price Cap Fourth Further Notice, we observed that the implicit rental price calculation in USTA's original study relied on depreciation rates it characterizes as "economic" depreciation rates, developed by an economist named Dale Jorgenson.⁽¹⁰⁶⁾ We questioned whether it was reasonable for carriers to use depreciation rates in TFP calculations that differ from the Commission's prescribed depreciation rates.⁽¹⁰⁷⁾ In our discussion of benchmark capital stock adjustments, we noted that the perpetual inventory model in USTA's original study multiplied the replacement cost of capital by "economic stock adjustment factors," and sought comment on economic stock adjustment factors.⁽¹⁰⁸⁾

60. Discussion. Ad Hoc and AT&T contend that we should use the depreciation rates prescribed by the Commission, and these parties use those rates in their studies.⁽¹⁰⁹⁾ They criticize Jorgenson's "economic" depreciation analysis on which USTA relied in its original TFP study, as well as in its simplified study. Ad Hoc and AT&T state that Jorgenson's analysis was based on a 1981 article by Hulten and Wykoff,⁽¹¹⁰⁾ which in turn was based on data ending in 1971, and examined depreciation of business assets for the economy as a whole rather than of telecommunications assets specifically.⁽¹¹¹⁾ USTA explains that it adopted only the depreciation method developed in the 1981 article, and substituted the most recent BEA data on telecommunications equipment lifetimes to develop depreciation rates.⁽¹¹²⁾

61. Some commenters argue that the depreciation rates should be those prescribed by the Commission.⁽¹¹³⁾ Ad Hoc maintains that the prescribed rates are designed to reflect the actual rate of plant retirement.⁽¹¹⁴⁾ MCI asserts that the prescribed rates in fact adequately reflect the economic life of plant and equipment.⁽¹¹⁵⁾ MCI includes a study of depreciation rates to support its conclusions.⁽¹¹⁶⁾ In particular, MCI asserts that the study shows that depreciation reserve deficiencies are not excessively high at this time.⁽¹¹⁷⁾ A number of LECs criticize MCI's study.⁽¹¹⁸⁾

62. We conclude that USTA has not shown that the depreciation rates it developed for its TFP calculations are in fact "economic" depreciation rates, or are reasonable for use in a LEC TFP study. First, although USTA states that it has updated the depreciation rates from the 1981 Hulten-Wykoff article with more recent BEA data, USTA has not shown that the depreciation rates it has developed are applicable to LEC equipment. Ad Hoc notes that the depreciation rates in the USTA study are lower than either the prescribed depreciation rates or the rates advocated by LECs in depreciation represcription proceedings, and argues that underestimating depreciation artificially reduces TFP and the X-Factor.⁽¹¹⁹⁾ USTA has not explained why it used depreciation rates lower than our prescribed rates,⁽¹²⁰⁾ when in other comments its members advocate higher depreciation rates.⁽¹²¹⁾

63. In our analysis, we have decided to use our prescribed depreciation rates. We find that it would not be reasonable, based on this record, to prescribe a set of depreciation rates for TFP calculations that differs from the depreciation rates currently in place for determining operating expenses. First, there is no sound basis in the record in this proceeding for determining whether and to what extent our depreciation rates differ from economic depreciation rates. Second, developing an additional distinct set of depreciation rates would clearly increase administrative burdens, and the record before us does not reveal any countervailing benefits that would justify this additional burden.⁽¹²²⁾ Third, under our recently established streamlined procedures for determining LEC depreciation rates, incumbent LECs have considerable influence and some discretion in setting their specific depreciation rates.⁽¹²³⁾ Commenters in this proceeding have not persuaded us that the depreciation rates we have currently prescribed do not reflect the LECs' depreciation costs.

64. To incorporate the effects of accumulated depreciation on its benchmark capital stock level, USTA states that, in its simplified TFP model, it multiplies gross book values by "economic stock adjustment factors" derived by dividing BEA market value measures by BEA original cost measures for certain asset classes.⁽¹²⁴⁾ For the same reasons we find above that the Commission's prescribed depreciation rates are better suited than USTA's depreciation rates for our TFP analysis, we are not using USTA's economic adjustment factors to adjust the benchmark capital stock level for the effects of depreciation. Instead, we have decided to base the benchmark capital stock calculations in our analysis on net book costs: gross book costs minus the accumulated depreciation reserves associated with our prescribed depreciation rates.

65. We note that we are making only limited findings in this Order regarding depreciation: (1) TFP calculations for purposes of determining an X-Factor at this time should use the same depreciation rates as those the incumbent LECs are required to use to determine their operating expenses, and (2) USTA has failed to show that the depreciation rates used in its simplified TFP model measure depreciation better than the Commission's depreciation rates. We reach no decision in this Order on the possible use of "economic" depreciation methods in general. In the Access Reform Notice, we sought comment on whether some portion of the incumbent price cap LECs' "residual" or "legacy" costs might be the result of underdepreciation.⁽¹²⁵⁾ We plan to address this issue in conjunction with the other residual cost issues we raised in the Access Reform Notice. Nor are we suggesting that we plan to continue exercising our Section 220(b) prescription authority indefinitely. The 1996 Act amended Section 220(b) of the Communications Act, so that we are no longer required to prescribe depreciation rates. The telecommunications industry is evolving, and this evolution may well require us to revise our prescription methods, or possibly discontinue depreciation rate prescriptions altogether. If we do revise the price cap LECs' depreciation rates substantially, or if we permit them to develop their own depreciation rates, we will determine the effect of the revised depreciation rates on TFP and the X-Factor in our next performance review.

(4) Hedonic Adjustments

66. Background. Both AT&T, initially, and Ad Hoc apply "hedonic" adjustments to their capital asset price indices, i.e., adjustments to reflect that new equipment differs from the old in technology as well as in price. AT&T and Ad Hoc argue that capital input prices must be adjusted for technological improvements to avoid understating the change in the effective level of real capital stocks. AT&T states that, to the extent that succeeding generations of capital equipment are more productive, a hedonic adjustment increases the computed level of capital stock, increases the flow of capital services, and, holding output constant, decreases measured TFP. AT&T also states, however, that a hedonic adjustment would decrease the price of capital input, thus increasing the input price differential. AT&T therefore argues that its computed X-Factor is not greatly affected by its hedonic adjustment.⁽¹²⁶⁾ By contrast, Ad Hoc asserts that a hedonic adjustment would increase the X-Factor, rather than merely result in offsetting changes in TFP and the input price differential. Ad Hoc makes no recommendation at this time, however, as to how to adjust for technological improvements, but asserts that, if this adjustment caused a 10 percent annual decrease in the price indices for the capital input asset categories that include computers, the X-Factor would increase by about 0.4 percent.⁽¹²⁷⁾

67. Discussion. We find nothing in the record to suggest that our TFP calculation would be more accurate with a hedonic adjustment. AT&T observes that its hedonic TFP adjustment results in an offsetting adjustment to its input price differential, leaving its X-Factor recommendation unchanged.⁽¹²⁸⁾ In addition, neither AT&T nor Ad Hoc have shown that their hedonic adjustments accurately measure the effects of technological improvements. The hedonic adjustment to the price per unit of capital proposed by AT&T in its TFP model is incompletely documented, and the details on all the components of the hedonic adjustment are not clear and replicable. Ad Hoc's 10 percent per year adjustment to certain asset price indices is not supported, but stated as an assumption. Based on the record before us, there is no need to include a hedonic adjustment.

(5) Deriving the Level of Capital Services from Capital Stock

68. Background. We invited comment on whether capital services should be measured by "capital consumption," i.e., the loss of efficiency in the capital over time, or by the level of capital stock. We noted that basing capital services on the level of capital stock assumes that the level of capital services is proportional to the level of the capital stock, and that the factor of proportionality does not vary over time. Alternatively, we sought comment on whether capital services could or should be based on some combination of the amount of capital consumption and the change in the level of capital stock.⁽¹²⁹⁾

69. Discussion. Our review of the economic literature on TFP and the pleadings of AT&T and USTA support the view that capital services (the quantity of capital services input) should be measured as proportional to the level of capital stock, and that capital consumption (such as depreciation expense) should be included in the measure of the cost (price) of the capital stock.⁽¹³⁰⁾ Further, the parties argue that capital services do not decline over the useful life of a unit of the capital stock. A piece of capital equipment with a ten-year life does not provide 10 units of capital services in its first year and only 3 units in its eighth year.⁽¹³¹⁾ All the TFP studies submitted in the record of this proceeding measure the change in capital services as the change in the level of capital stock.

(6) Implicit Rental Price

70. Background. The weight given to the capital services input when it is aggregated with labor and materials inputs is based on the capital cost, which is the product of the implicit rental prices of the total capital stocks for the asset categories. The implicit rental price represents the hypothetical price of renting the LECs' capital stock in a competitive market, if such a market existed.⁽¹³²⁾ In the Price Cap Fourth Further Notice, we observed that the implicit rental price in USTA's original TFP model is based on the rate of return, the depreciation rate, certain tax rates, and its TPIs.⁽¹³³⁾ In addition to asking specific questions regarding the rate of return, depreciation, and taxes, we sought comment on whether USTA's method of calculating the implicit rental price is reasonable. We also asked whether data would be available on a timely basis to make these calculations in the future, and about alternatives to USTA's method.⁽¹³⁴⁾

71. We also asked questions regarding the rate of return component of USTA's implicit rental price. We observed that USTA's original TFP model used Moody's Yield on Public Utility Bonds as the rate of return, and questioned whether it would not be more reasonable to include the cost of equity as well as the cost of debt in the rate of return.⁽¹³⁵⁾ We also noted that we have determined the LECs' rate of return in our past rate-of-return represcription orders, and questioned whether it would be reasonable to allow LECs to use any other rate of return. We also sought comment on how often, and by what method, the rate of return should be updated for purposes of TFP calculations. Finally, we invited comment on whether a represcription of the rate of return applicable to carriers subject to rate-of-return regulation should also be incorporated into TFP calculations.⁽¹³⁶⁾

72. Discussion. USTA estimates the rate of return in its implicit rental price calculation by deriving a nationally averaged return on capital from the National Income and Product Accounts. AT&T claims that USTA's implicit rental price introduces unreasonable distortions because it does not reflect price cap LECs' actual payments to capital. AT&T bases its weight for the capital input, or the "cost of capital" in terms of TFP calculations, on LEC revenues less the costs of labor and materials.⁽¹³⁷⁾ We find that AT&T's residual earnings method is a more accurate estimate of the contribution of capital to the production of output than USTA's method of measuring rate of return, because AT&T's method measures the actual flow of funds to capital. In other words, the residual earnings method reflects actual payments to capital. We have decided to use AT&T's approach in our analysis of the record, with the minor modifications discussed below.

73. AT&T cites several economic articles supporting the use of residual earnings as the cost of capital in TFP calculations.⁽¹³⁸⁾ For example, to correct for the potential distortion in the measurement of TFP growth, Berndt and Fuss propose two measures of implicit rental prices as alternatives to the equation proposed by USTA, one of which is similar to the implicit rental price proposed by AT&T.⁽¹³⁹⁾ Dhrymes calculates an implicit rental price in a similar manner.⁽¹⁴⁰⁾ Additionally, AT&T states that Christensen, USTA's consultant, has used a similar construction in a TFP study Christensen presented to the Public Service Commission of North Dakota on behalf of US West.⁽¹⁴¹⁾

74. USTA and a number of LECs assert that AT&T's weighting of the capital input index replicates the incentives of rate-of-return regulation because it results in limiting carriers to a particular rate of return.⁽¹⁴²⁾ We disagree. Under rate-of-return regulation, increases in a LEC's earnings lead directly to reductions in that LEC's rates. Under AT&T's capital weighting method, an increase in a LEC's earnings will increase the weight placed on its capital input index relative to its labor and materials indices. This would increase TFP and the X-Factor only to the extent that capital is growing less quickly than labor and materials. Also, the X-Factor is based on an industry average, and an increase in a particular LEC's TFP has only a limited effect on the industry average.

75. In our TFP calculation, we follow AT&T's proposal with modifications. The estimated implicit rental price is measured in terms of gross returns to capital divided by the capital stock. The weight used for aggregating capital services into the overall input quantity index is the share of gross payments to capital in total payments to all factors.

76. As a result of our decision to rely on AT&T's rather than USTA's implicit rental price, we need not determine whether a rate of return based on National Income and Product Accounts, Moody's bond indices, or the Commission's prescribed rate of return would be the most reasonable measure of the rate of return to incorporate into an implicit rental price calculation. We also do not need to address AT&T's contentions regarding USTA's treatment of depreciation or taxes in its calculation of the implicit rental price. Depreciation rates are relevant to AT&T's treatment of capital stock, however, and accordingly, we considered depreciation issues above.

b. Labor

77. Background. Labor is the second of the three factors of the TFP input index. In the Price Cap Fourth Further Notice, we noted that USTA's original TFP study used two categories of labor: management and non-management. We asked whether labor should be further disaggregated to account for different levels of education and vocational experience in the work force.⁽¹⁴³⁾ We also asked about adjustments for carrier "outsourcing," i.e., replacing the services of workers employed by carriers with services provided by outside firms.⁽¹⁴⁴⁾

78. Discussion. In USTA's simplified TFP model, there is one category of labor, and the quantity of labor is measured as the number of employees. AT&T's TFP calculations are based on two categories, full-time and part-time employees. AT&T measures the quantity of labor as number of employees, with part-time employees counted as a fraction of a full-time employee. No one has suggested a more disaggregated labor input index. In our analysis of the record, we base the rate of growth of labor on total number of employees, to be consistent with our current collections of ARMIS data.

79. We agree with USTA that, when outsourcing occurs, the decrease in labor input growth is offset by an increase in expenses for services, and is reflected in the materials index.⁽¹⁴⁵⁾ Because materials expenses are inputs to the TFP calculation, no additional adjustment for outsourcing is needed.

c. Materials

80. Background. The original USTA TFP study derived materials quantities indirectly. USTA calculated materials expenses by subtracting depreciation and amortization expense, and employee wages, salaries, and benefits, from total operating expenses, and then deflated (or divided) this residual expense by the GDP-PI to construct a materials input index. AT&T's TFP study calculated materials expense by subtracting total labor compensation and the change in the depreciation reserve from total operating expense. AT&T deflated this residual expense by a materials price index. In the Price Cap Fourth Further Notice, we sought comment on whether it would be preferable or possible to construct a LEC-specific price index for deflating materials expense instead of relying on GDP-PI for that purpose. We stated that our objective was to measure TFP accurately with data that are verifiable and publicly available. In this section, we address only materials price and quantities index issues. We will address materials index weighting issues below.

81. Discussion. All the parties use the residual expense method of measuring materials. USTA uses the GDP-PI as the materials price index to deflate residual expense to derive materials quantities in its simplified TFP model. We find that USTA has not shown that use of GDP-PI accurately measures the prices of LEC materials and, therefore, TFP, because it does not reflect price changes in the narrow range of inputs used by LECs. This significantly affects measured TFP, and it disguises a significant portion of the input price differential.

82. The record contains a materials price index created by AT&T based on a subset of categories of national input/output expenditures prepared by the U.S. Bureau of Labor Statistics (BLS) that is more narrowly focused on materials purchases of communications industries than the economy-wide GDP-PI. We have replicated the index using the same BLS data that AT&T used in an ex parte filing received on April 11, 1996.⁽¹⁴⁶⁾ AT&T's materials price index is a Tornquist index calculation, where the logarithmic percentage changes are replaced by arithmetic percentage changes. Because AT&T's materials price index is more narrowly focused on communications services than GDP-PI, we use AT&T's materials price index.

d. Weighting of Materials and Labor Indices

83. All the models placed in the record base the weight of the materials index in the final input index on materials expense. Since all the models determine materials expense as the residual expense left after labor compensation and depreciation are subtracted from total operating expense, both the labor and the materials shares of total inputs are affected by the specification of labor and depreciation expense.

84. USTA notes that AT&T's materials input index weight is calculated residually on the basis of total operating expenses minus total labor compensation and the change in depreciation reserves. USTA claims that AT&T's treatment of both labor expense and materials expense is flawed, and that those calculations distort the weights placed on the materials and labor input indices. USTA further claims that distorting the weights placed on the materials and labor input indices results in distorting the capital input index as well.

85. First, USTA claims that AT&T erred in subtracting total labor compensation from total operating expense. USTA claims that the proper measure of current period labor expense is wages, salaries and benefits. According to USTA, total labor compensation includes labor costs that are capitalized rather than expensed in the year in which they are incurred. Each year a portion of previously capitalized labor expense enters the current year total operating expense as part of depreciation expense. USTA claims that total labor compensation results in some double counting of labor expense,⁽¹⁴⁷⁾ and thus improperly shifts weight from the materials expense index to the labor input index.

86. Second, USTA claims that AT&T improperly calculated materials expense because it used the change in depreciation reserves instead of recorded depreciation and amortization expense. The increase in depreciation reserves may be less than depreciation and amortization expense because plant retirements draw down the reserve. This issue is different from the depreciation rate issue discussed above. Here, the issue is not to determine the proper rate of depreciation, but to determine materials expense by subtracting the depreciation (and labor) expense components of operating expense from total operating expense. USTA claims that changes in depreciation reserves understate depreciation expense, and, thus, overstate materials expense and place too great a weight on the material input index.

87. USTA claims that these errors result in an understatement of 0.2 percent in TFP for the period from 1988 to 1994, and an understatement of 0.3 percent for the period from 1989 to 1994.⁽¹⁴⁸⁾ USTA also admits, however, that these errors would have offsetting effects on the calculation of the input price differential in AT&T's model, and, consequently, no overall effect on an X-Factor that includes an input price differential.⁽¹⁴⁹⁾ In its 1997 reply, AT&T states that it has switched to using depreciation and amortization expense, rather than changes in depreciation reserves,⁽¹⁵⁰⁾ for this calculation.

88. Both USTA's and AT&T's models double count some labor costs by basing labor quantities on the number of employees. This double-counting occurs because capitalized labor expense is reflected in capital stock as well as labor. USTA has not solved this problem by basing labor expense on wages, salaries, and benefits rather than total compensation, because capitalized labor remains fully reflected in capital stock. Instead, USTA's approach merely changes the relative weights placed on the labor, materials, and capital input indices. We have decided in our staff analysis to weight the labor input index in our analysis on total compensation rather than wages, salaries, and benefits.

89. In summary, we base the weight placed on the materials input on Total Operating Expense, less total labor compensation, as AT&T recommends, and depreciation/amortization expense, as USTA recommends.

90. In the Price Cap Fourth Further Notice, we were "particularly concerned" about whether to adjust labor costs for other post-employment benefits (OPEBs) given that we had first permitted price cap LECs to make an exogenous cost increase to reflect these costs, and then later required those LECs to make an exogenous cost decrease.⁽¹⁵¹⁾ We decide that no special adjustment of the labor input index is needed to reflect our changing regulatory treatment of OPEBs. The only relevant OPEB issue for purposes of TFP is whether amortizing OPEB expenses over longer or shorter periods can have any effect on the labor index, and thus TFP. We find that it does not because LECs record OPEB costs in their books at their present value, regardless of the amortization period we require. As a result, recording OPEB costs now has no greater or lesser effect on the labor input index than recording those costs in the future.

5. Summary

91. Total factor productivity (TFP) is the relationship between the output of goods and services to inputs of basic factors of production -- capital, labor, and materials. A TFP study attempts to quantify this ratio of output to inputs and measure the improvement in the ratio over time. The following outlines the staff TFP analysis, which is presented in detail in Appendix D.

92. We measured the change in the quantity of output using the change in physical measures such as access lines, messages, and minutes. Output quantities are then converted to index numbers and combined using their relative shares of total revenues as weights.

93. For inputs, the quantity of labor is measured directly, using the reported number of employees. We create the labor quantity index by taking a ratio of number of employees in a year to the number of employees in the base year, 1985. We measure capital services as a constant proportion of the capital stock. Thus, the change in capital services is proportional to the change in the capital stock. We have no direct measure of the quantity of materials consumed in the production of any period's output. Instead, we calculate materials expense by subtracting from total operating expense the operating expenses attributable to labor, and depreciation and amortization expense. To convert materials expense into a quantity, we deflate materials expense by a price index specifically created to measure changes in materials prices. To combine these inputs into a single index of inputs, we need to calculate weights (or factor shares) that represent the relative contributions of the inputs in the production process. We assume the contribution of each input is proportional to the payments to that factor of production. The weight for each factor is its share of total factor payments. For labor, this is total employee compensation. For materials, we use a number we have already calculated -- total material expense. The payment to capital is equal to gross return to capital, which is the difference between total revenue and the sum of materials and labor expense.

94. Estimating the change in total factor productivity allows us to develop an input price index that measures the change in the unit cost of purchasing basic resources. The labor and capital prices are transformed into indices, and the three input price indices are combined using the factor shares calculated above.

D. Other X-Factor Calculation Issues

1. Input Price Differential

95. Background. In the LEC Price Cap Performance Review, we noted that changes in a firm's costs of producing a unit of output are the product of both changes in the quantity of resources used, i.e., changes in productivity, and changes in the prices paid for those resources, i.e., changes in input prices.⁽¹⁵²⁾ We tentatively concluded that the X-Factor should include both a measure of productivity growth and a measure of input price changes.⁽¹⁵³⁾ Specifically, we found that, as a theoretical matter, because LEC unit costs are also affected by the prices they pay for inputs, an input price differential should be included in the X-Factor.⁽¹⁵⁴⁾ In general, any TFP study generates an estimate of the change in input prices over the study period, in the price indices used to calculate the input indices. "Input price differential" refers, in the present context, to the difference between the rate at which input prices change in the economy in general and the rate at which LEC input prices change. Thus, when USTA claims that the long-term input price differential is zero, it is saying that the prices LECs pay for the resources they use in producing telecommunications services change at about the general rate of inflation. An input price differential of 2 percent, on the other hand, would mean that the prices LECs pay for the resources they use rise more slowly than the general rate of inflation. A higher input price differential produces a higher X-Factor.

96. Based on data USTA supplied in its comments filed in this proceeding prior to the LEC Price Cap Performance Review, and in ex parte statements filed in January and February 1995, we tentatively concluded in the Price Cap Fourth Further Notice that the input price differential was about 2.7 percent for the period from 1984 to 1990.⁽¹⁵⁵⁾ We found that USTA's conclusion that the long-term input price differential is zero was theoretically unsound, and unsupported by USTA's data.⁽¹⁵⁶⁾ In the Price Cap Fourth Further Notice, we also sought comment on whether the input price differential should be based on a long-term trend as USTA suggested, or on a shorter period, such as the period used for the TFP analysis, as Ad Hoc suggested. We invited comment on the data that should be used to calculate the input price differential.⁽¹⁵⁷⁾

97. Discussion. USTA and other parties agree that changes in LEC input prices should be reflected in the X-Factor if productivity is measured using a TFP method, because TFP adjusts input and output prices to "real" or constant dollar terms to measure "real" productivity. USTA advocates a long-run analysis of input prices, and asserts that, in the long run, there is no statistically significant difference between LEC input price changes and economy-wide input price changes. Other parties contend the relevant period is roughly from 1984 to the present. AT&T estimates that the input price differential was 2.54 percent per year from 1985 to 1994, using BLS data rather than the data in the Christensen Study sponsored by USTA.⁽¹⁵⁸⁾ AT&T also estimates that the input price differential between 1985 and 1995 was 2.35 percent.⁽¹⁵⁹⁾ Ad Hoc claims that the input price differential from 1984 to 1993 is 2.1 percent based on USTA's data, or 3.4 percent based on USTA's data corrected for certain errors alleged by Ad Hoc.⁽¹⁶⁰⁾ Sprint compares its price indices for capital, labor, and materials to its economy-wide input price index, and finds that the five-year moving averages for the period from 1985 to 1993 range from 0.84 percent to 1.64 percent.⁽¹⁶¹⁾

98. On the basis of the record before us in this proceeding, we conclude, for the reasons discussed below, that short-term data should be used to select an input price differential for use in prescribing a TFP-based X-Factor. All the TFP models in the record include price indices for capital, labor, and materials, and the weights needed to calculate an average input price index. All parties used TFP models that determined an X-Factor by estimating productivity and input prices simultaneously, because both the inputs and outputs must be measured in real, or inflation-adjusted, terms. Therefore, any estimate of TFP includes an estimate of an input price differential. If we adopted a methodology that used one set of assumptions and data to measure LEC input prices for use in calculating TFP, and a different set for measuring the input price differential, the calculations would be inconsistent. We see no reason to calculate TFP using one set of data and assumptions, and then calculate the input price differential using a different set of data and assumptions. Therefore, we do not estimate the input price differential separately from TFP, and we will not make independent prescriptions of the productivity and input price components of the X-Factor. Instead, we will focus directly on selecting the appropriate combined X-Factor. Accordingly, in the table in Section III.E. below, we display X-Factor estimates which are combined TFP and input price differentials, rather than separate forecasts of TFP and input price differentials.

99. The LECs make four arguments in favor of setting the input price differential equal to zero: (1) the input price differential should be based on long-term studies; (2) short-term studies do not show a positive input price differential, but rather a temporary effect of divestiture; (3) it is not reasonable to estimate input price changes on the basis of the price indices in TFP calculations; and (4) including an input price differential might make the X-Factor volatile in a moving average-based price cap plan. For the reasons discussed below, we find none of these arguments persuasive.

100. We give no weight to USTA's estimate of the long-term trend. Both the Christensen Study and the NERA Study submitted by USTA, and discussed in Appendix F of the LEC Price Cap Performance Review, base their conclusions on four different TFP studies, each covering different periods of time. Each of these studies was conducted using disparate and inconsistent techniques. For example, different methods of measuring materials input prices, and different depreciation rates, were used to develop capital input prices for different portions of the study period. In addition, the data in the Christensen Study could support a conclusion that the input price differential is either zero or 2.6 percent.⁽¹⁶²⁾ Although the LECs focus their attention on the fact that zero is within the range of possible input price differentials supported by USTA's studies,⁽¹⁶³⁾ none adequately addresses the fact that the data support a wide range of other possible outcomes. Because neither the Christensen Study nor the NERA Study is based on a consistent set of data or methodology throughout the period covered by either study, we find that their conclusions about the long-term trend of LEC input prices are not supported.

101. We agree with the parties who argue that consistency requires us to use data from the same period to determine both TFP growth and input price differential.⁽¹⁶⁴⁾ Furthermore, our objective here is to prescribe an X-Factor that will set a reasonably aggressive productivity goal for LECs for the near future until completion of the next performance review. Given all the changes that have occurred in telecommunications during the 44 years covered by the long-term input price studies that have been placed on the record here,⁽¹⁶⁵⁾ we find that data from a recent, shorter period of time provide a more reliable basis for estimating input price trends for the near future than the longer term data.

102. Some incumbent LECs contend that any input price differential revealed by an analysis of the data from 1985 to 1994 is a temporary effect of divestiture. According to these commenters, the input price differential appears in 1984, returns to zero in 1989 or 1990, and is likely to continue to be zero in the future. USTA, on the other hand, claims that the input price differential is not related to divestiture at all, and that the input price differential started to increase in 1980 and began declining in 1990.⁽¹⁶⁶⁾ USTA also contends that the difference in input price differential in the Christensen Study before and after 1984 is a result of the different methodologies used to generate the pre- and post-1984 data series.⁽¹⁶⁷⁾ We conclude that the input price differential is not a temporary effect of divestiture. LEC input prices have grown at a different rate from input prices in the economy as a whole for all the years analyzed in our study. Furthermore, no party making this argument provides any theoretical argument to explain why the input price differential was exclusively a result of divestiture, and therefore could not ever recur. Therefore, we are not persuaded by this record that the observed LEC input price differential was merely a temporary effect of divestiture, or is unlikely to continue.

103. AT&T argues that LEC input prices for capital and materials in USTA's simplified TFP model are closely related to GDP-PI, and thus artificially reduce the input price differential.⁽¹⁶⁸⁾ USTA adopts GDP-PI as its materials input price index for LECs, and bases its capital input price indices for LECs on National Income and Product Account data. Thus, USTA's TFP study simply assumes away much of the difference between LEC input price growth and U.S. input price changes by basing most of its input price information on data directly related to GDP-PI and U.S. input price growth. Using GDP-PI to measure input prices is unreasonable because GDP-PI measures output prices, i.e., the prices of final goods and services, rather than input prices, the prices of intermediate goods and services. Therefore, we base our analysis of the input price differential on the input price indices we use in our analysis of the record.

104. A number of LECs assert that the design of USTA's original TFP model precludes any derivation of a meaningful estimate of LEC input price changes. These parties argue further that the Commission erred in Appendix F of the LEC Price Cap Performance Review in concluding that the price indices in USTA's TFP study can be used to produce reliable results regarding the input price differential for our purposes.⁽¹⁶⁹⁾ Ad Hoc argues that the Commission's input price differential results are not unreliable simply because USTA did not intend its TFP study to be used to derive the input price differential.⁽¹⁷⁰⁾ We agree with Ad Hoc on this issue. The LECs have not explained why we should assume that the price indices used for their TFP calculations do not reflect their input prices for purposes of calculating the input price differential.

105. Several parties assert that the X-Factor should represent a prediction of the LECs' achievable future productivity growth, and that including the input price differential in the X-Factor would make it too volatile to have any predictive power, and could cause rate churn.⁽¹⁷¹⁾ As we explain further in Section V. below, we have decided to adopt a fixed X-Factor, which will preclude any volatility in the input price differential from being reflected in the X-Factor. Finally, we reject Southwestern Bell's assertion that the past input price differential should not be relevant for setting a future X-Factor.⁽¹⁷²⁾ Changes in input prices affect incumbent LECs' unit costs, and so should be reflected in the X-Factor. We have no more reliable basis for predicting future input price changes than past input price changes.

106. In the LEC Price Cap Performance Review, we defined the input price differential as the difference between the rate of change in LEC input prices and economy-wide input price changes, rather than the difference between LEC input prices and GDP-PI.⁽¹⁷³⁾ We estimate LEC input prices on the basis of the price indices we use to calculate TFP, and we have chosen to use the BLS Non-Farm Business Sector Input Price Index as our measure of economy-wide input price changes, as AT&T used.⁽¹⁷⁴⁾ We have chosen the BLS Non-Farm Business Sector Input Price Index for economy-wide input prices because this is the broadest index of the prices of non-farm input goods and services available. It is also produced in conjunction with, and is therefore consistent with, our measure of productivity growth for the economy as a whole. We did not choose GDP-PI because the input price differential measures the difference between LEC input prices and input prices in the economy in general, and GDP-PI is a measure of price changes for final goods and services. The most recent published data in these series is for 1994. We estimate the 1995 changes using the average of the five most recent years.

2. Adjustment to X-Factor for Interstate-Only Activity

a. Background

107. USTA's original TFP study was based on total company data. AT&T claimed that the LECs' interstate access services have grown faster than LEC output overall, so that interstate productivity growth was greater than total company productivity growth. Thus, according to AT&T, reliance on total company data in measuring TFP tends to understate the LECs' interstate access productivity growth.⁽¹⁷⁵⁾ We noted that interstate and intrastate services are usually provided over common facilities, and questioned whether it would be possible to develop separate production functions for interstate and intrastate services.⁽¹⁷⁶⁾

108. In the Price Cap Fourth Further Notice, we invited comment on several issues related to this subject, including whether consideration of total company TFP data might exceed our jurisdiction. We also sought comment on whether there was any way to develop "economically meaningful" separate production functions for the purposes of calculating interstate TFP, or if not, whether there was any adjustment that could be made to total company TFP to account for any existing differences between interstate and intrastate productivity growth.⁽¹⁷⁷⁾ Finally, we asked whether basing the X-Factor on total company TFP would require us to revise our ARMIS or Form 492 reporting requirements.⁽¹⁷⁸⁾

b. Discussion

109. We stated in the LEC Price Cap Performance Review that we would consider making an adjustment to account for differences in interstate and intrastate productivity growth if including intrastate data created a "systematic downward bias" in the X-Factor.⁽¹⁷⁹⁾ We also stated that we would prefer to address any such bias "directly," rather than by attempting to construct an interstate factor based on regulatory accounting and other regulatory requirements that may not fully reflect economic costs.⁽¹⁸⁰⁾

110. We find that the record before us does not allow us to quantify the extent, if any, to which interstate productivity growth may differ significantly from total company productivity growth. AT&T argues that interstate productivity growth is greater than intrastate growth because there are greater economies of scale for interstate services.⁽¹⁸¹⁾ CCTA assumes that interstate productivity growth is greater because some state public service commissions have retained rate-of-return regulation.⁽¹⁸²⁾ On the other hand, BellSouth asserts that interstate services are more capital-intensive than intrastate services, and that capital inputs have grown faster than labor or materials inputs. On this basis, BellSouth infers that interstate productivity may have grown more slowly than intrastate productivity.⁽¹⁸³⁾ Neither CCTA nor BellSouth has provided any empirical data to substantiate either the effects they describe or their significance. AT&T and Ad Hoc calculate interstate TFP by measuring the growth in interstate outputs, but assume that interstate inputs grow at the same rate as intrastate inputs. USTA argues that it would be more reasonable to assume that interstate inputs grow at the same rate as interstate outputs. None of these parties, however, provides a factual or theoretical explanation as to why its assumptions might be correct. Accordingly, we find no basis in the record for making an adjustment to the X-Factor to account for any differences between interstate and total company productivity.

111. Arguing that interstate productivity growth is systematically greater than intrastate productivity growth, Ad Hoc and API assert that basing the X-Factor on total company TFP might give LECs a windfall unless the states also adopt regulations based on total company data.⁽¹⁸⁴⁾ Ad Hoc also asserts that we should require an interstate TFP adjustment because some LECs have advocated making some intrastate TFP adjustment before state public service commissions.⁽¹⁸⁵⁾ Unsupported claims of a potential LEC windfall do not by themselves convince us that there is any factual basis for concluding that there is a systematic difference between interstate and total company productivity. Ad Hoc's claims that some LECs have supported intrastate TFP adjustments in some state jurisdiction does not show that there is a nation-wide difference between interstate TFP and total company TFP significant enough to warrant making some adjustment to our LEC industry-wide X-Factor.

112. Legal Considerations. AT&T and others make various arguments that using total company data to calculate TFP violates Section 2(b) of the Communications Act or the requirements of Smith v. Illinois Bell.⁽¹⁸⁶⁾ Because we have determined above that the record does not demonstrate any systematic bias in using total company productivity growth, we need not reach this legal issue at this time.

c. TFP Adjustment for Differences in Regulated and Nonregulated Productivity Growth

113. Background. We also solicited comment in the Price Cap Fourth Further Notice on whether we should measure TFP on any less-than-total-company basis other than interstate-only, such as the TFP for regulated services.⁽¹⁸⁷⁾ We also asked whether we should exclude the productivity growth associated with certain specific regulated services or groups of services. The example we used in the Price Cap Fourth Further Notice was video dialtone services. We noted that nonregulated services might not share joint and common costs with regulated services to the same extent as interstate and intrastate services.⁽¹⁸⁸⁾

114. Discussion. Ad Hoc claims that the initial investment required to begin providing certain nonregulated services or video services could increase capital inputs, and thus decrease measured TFP growth.⁽¹⁸⁹⁾ If we adopted a moving-average methodology, Ad Hoc's assertion might warrant closer analysis. We are instead prescribing an X-Factor based on data from 1986 to 1995. We find that nonregulated investment during this time period was too small, relative to total regulated investment, to have a significant effect on our TFP calculations. We therefore make no adjustment to the X-Factor or to TFP to account for the effects of nonregulated activities.

115. In its 1997 reply, AT&T asserts that USTA has recognized the legitimacy of making a regulated/non-regulated adjustment by doing so in its TFP analysis.⁽¹⁹⁰⁾ AT&T does not specifically identify the adjustment that it maintains USTA has made to account for differences in regulated and non-regulated productivity, but it appears to be in USTA's miscellaneous services output index. As we discuss above, USTA's miscellaneous services output index contains several anomalous results, including negative growth in some years. As a result, we have excluded that output category completely from our output index.

d. Reporting

116. We sought comment on whether basing the X-Factor on total company TFP would require us to expand our ARMIS or Form 492 reporting requirements to collect total company data.⁽¹⁹¹⁾ Below, we decline to adopt a price cap plan in which LECs would be required to recalculate the X-Factor annually on the basis of a prescribed method. Instead, we prescribe an X-Factor that will remain in effect at least until the next performance review. Accordingly, we conclude that we need not expand our reporting requirements at this time.

3. Effect of Universal Service and Other Subsidy Programs on LEC TFP

117. Background. In the Price Cap Fourth Further Notice, we noted that there were a number of universal service or other subsidy programs at both the federal and state levels, and asked to what extent such programs affect or should affect LECs' productivity calculations.⁽¹⁹²⁾

118. Discussion. A number of commenters argue that total company TFP captures the effects of any universal service fund or subsidy programs, and thus no special adjustments are needed.⁽¹⁹³⁾ BellSouth contends that changes in universal service funding requirements are treated exogenously, and supports continuing this treatment.⁽¹⁹⁴⁾ CCTA supports considering universal service fund revisions in the Universal Service Order proceeding rather than here.⁽¹⁹⁵⁾

119. We have no reason to believe that replacing the implicit subsidies in incumbent LECs' current rates with explicit subsidies, as required to meet the 1996 Act's universal service provisions, will affect productivity significantly. The implicit subsidies were designed to promote universal service, and have been generally successful.⁽¹⁹⁶⁾ We expect subscribership levels to remain high under our new universal service rules. Thus, there should not be any dramatic increases or decreases in incumbent LEC outputs, and so there should be little effect on TFP. Accordingly, we will not take any further action on this issue here.

4. Inclusion of Other Firms in Study

120. Background. In the first phase of this proceeding, Ad Hoc argued that basing the X-Factor on industry-wide moving average data might encourage excessive network investment, and thus might lead to "gold-plating" incentives similar to those created by rate-of-return regulation. Therefore, Ad Hoc recommended including data from other telecommunications service providers in the TFP calculations.⁽¹⁹⁷⁾ We invited comment on Ad Hoc's proposal, and requested parties to discuss whether the data necessary to perform an expanded TFP study would be available annually in a timely manner.⁽¹⁹⁸⁾

121. Discussion. Below, we decline to adopt a methodology for the X-Factor on an industry-wide moving average. Therefore, we conclude that there is no need at this time to include data from other industries to address the concern raised by Ad Hoc. At this time, we also need not address NYNEX's, GTE's, and US West's arguments against inclusion of such data.

5. Consumer Productivity Dividend

122. Background. In the LEC Price Cap Order, we added 0.5 percentage points to the X-Factor to ensure that the first benefits of the price cap plan are flowed through to access customers. We called this addition the consumer productivity dividend (CPD).⁽¹⁹⁹⁾ In the Price Cap Fourth Further Notice, we invited parties to discuss whether we should retain the CPD in the long-term price cap plan, in order to, for example, reflect anticipated productivity growth resulting from the elimination of sharing.⁽²⁰⁰⁾ We also sought comment on whether the CPD should remain at 0.5 percent or be set at some other value.⁽²⁰¹⁾

123. Discussion. Consistent with our practice in both AT&T and LEC price cap regulation, we retain a 0.5 percent Consumer Productivity Dividend in our revised price cap plan. We decide below to adopt a single fixed X-Factor in our revised price cap plan, based on LEC industry-wide data. The CPD will act as a mechanism to ensure that price cap LECs flow-through a reasonable portion of the benefits of productivity growth to ratepayers. The importance of this purpose in our revised price cap plan is enhanced because we are eliminating the current sharing requirements and we are not adopting a moving average method of updating the X-Factor.⁽²⁰²⁾

124. Parties arguing in favor of eliminating the CPD are not persuasive. Several incumbent LECs maintain that it is arbitrary and capricious to transfer any productivity gains to access customers. In a competitive market, however, competitors will continuously provide firms with incentives to lower their unit costs more quickly than they have in the past so that they can lower their prices and win customers from their competitors. By this mechanism, a competitive market passes cost reductions on to customers in the form of lower prices. By requiring incumbent LECs to transfer at least part of their productivity gains to access customers, the CPD tends to replicate the results of a competitive market. Therefore, we find that it is reasonable to use a CPD to require incumbent LECs to transfer some portion of their unit cost reductions to their customers. USTA asserts that the price cap plan properly balances shareholder and ratepayer interests without the CPD,⁽²⁰³⁾ but does not explain why we should not continue our established practice.

125. Some contend that the CPD was adopted because of uncertainty regarding the X-Factors in the original price cap plan, and our experience under price cap regulation should have alleviated this uncertainty. We disagree that the passage of time by itself has eliminated the need for a CPD. The CPD remains necessary to require LECs to transfer some portion to their unit cost reductions to their access customers. Also, the CPD was, in a sense, an expression of certainty that LECs would respond to the incentives provided by the price caps plan by becoming more productive, and that there would be productivity gains that could be shared between ratepayers and shareholders. The passage of time has not altered the need to strike this balance between ratepayer and shareholder interests.

126. BellSouth and GTE argue that there was no principled basis for selecting 0.5 percent as the CPD. We explained in the LEC Price Cap Order that setting the CPD at 0.5 percent would ensure that access customers share a portion of the productivity benefits of price cap regulation.⁽²⁰⁴⁾ Although GTE broadly asserts that including a 0.5 percent CPD would cause the X-Factor to be excessive, we believe that a 0.5 percent CPD, with the elimination of sharing, continues to be necessary to ensure that access customers receive benefits.

127. We are mindful that, while some incumbent LECs have achieved high earnings under price caps, others have not always done so. We therefore retain the low-end adjustment mechanism for LECs with substantially below-average earnings. The low-end adjustment mechanism permits incumbent price cap LECs with rates of return less than 10.25 percent to increase their PCIs to a level that would enable them to earn 10.25 percent.⁽²⁰⁵⁾

6. Effects of Access Reform

128. In the Access Reform Notice, we invited comment on the potential effects of access reform on TFP.⁽²⁰⁶⁾ Some parties argue that replacing the per-minute carrier common line charge with a per line charge will depress measured TFP because access lines have historically grown more slowly than access minutes.⁽²⁰⁷⁾ USTA argues that if either competition or regulatory action reduces the price-marginal cost margin on rapidly growing services, measured TFP will fall. USTA concedes it has no direct evidence of the expected magnitude of this effect and makes no specific prediction of the size of the reduction in TFP growth.⁽²⁰⁸⁾ USTA estimates, however, that its access reform proposal, holding everything else constant, would reduce measured TFP growth for the period from 1990 to 1995 by 0.4 percent by changing the revenue weights of per-line and per-minute common line services.⁽²⁰⁹⁾ USTA claims support for its assertion that measured TFP growth will be affected by restructuring the collection of common line costs from two articles from the literature of economics.⁽²¹⁰⁾ On the other hand, AT&T anticipates that access reform would increase productivity growth, because reducing rates to cost-based levels would stimulate demand.⁽²¹¹⁾

129. We find that USTA has not sufficiently considered the effect that moving prices towards marginal cost will have on LEC efficiency. Under our current access rate structure rules, before the revisions adopted in our companion Access Reform First Report and Order, incumbent LECs are often unable to offer access services at rates that reflect the manner they incur costs and therefore are faced with artificially depressed demand. The implicit cross-subsidies in our current access rate structure rules have resulted in increased demand for certain services and decreased demand for others. When demand for services is distorted in this fashion, incumbent LECs must provide those services at levels that do not enable them to minimize their per-unit costs. When prices reflect marginal costs, however, consumers increase their purchases of services previously priced above marginal cost, and reduce their purchases of services previously priced below marginal cost. The net result of such a change in rate structure will allow LECs to minimize the per-unit cost of producing their total output. Based on the current record, we find that access reform will have at most a very modest effect on the revenue weights used to aggregate output and that this effect will be offset at least in part by changes on the input side of the TFP equation as LECs adjust inputs to produce a more efficient mix of outputs. Thus, it would be speculative to attempt to adjust our TFP estimates now.

130. The articles cited by USTA are consistent with this analysis. They provide support only for the proposition that, if everything else is held constant, adjusting the weights of each category of LEC outputs for the margin between price and marginal cost reduces measured output, measured TFP, and TFP growth.⁽²¹²⁾

131. Some parties contend that measured TFP will decrease under competition because incumbent LEC output will fall as new entrants successfully compete for existing customers. USTA asserts that a one percent reduction in LEC output growth will reduce LEC TFP growth by 0.3 to 0.5 percent. We are not persuaded that we should reduce our baseline productivity estimates we are using here to set an X-Factor that will apply to all incumbent price cap LECs and all their access services. We are not deciding what, if any, changes to the X-Factor we should make with the lowering of barriers to competitive entry or the development of competition.⁽²¹³⁾

132. In summary, we find that the parties have not shown it reasonable to reduce the measured TFP growth of incumbent LECs in light of the overall effect of the rate restructuring adopted in the Access Reform First Report and Order.

E. Analysis and Prescription

133. Above, we have examined several individual issues regarding TFP calculation, determination of the input price differential, and other X-Factor calculation issues. On the basis of the record in this proceeding, we have determined the best available methods to perform each of the calculations necessary to conduct a TFP study, and we have developed a reasonable prediction of the future input price differential. We recognize that the results of any study are reliable only to the extent that the data used in the study is taken from a consistent series, and that the methods used in the study are internally consistent. We conclude that our staff analysis relies on consistent data sources and methods, and that our input price differential findings are based on consistent and reliable data.

134. For reasons discussed in Section V. below, we have decided not to adopt a moving average mechanism to update the X-Factor. In the Price Cap Fourth Further Notice, we sought comment on the best time period for studies used to calculate a fixed X-Factor.⁽²¹⁴⁾ Ad Hoc contends that we should use all the data since 1984, arguing that the divestiture of the Bell System in 1984 creates a "break" in the data, and that comparing data from before and after that time could yield anomalous results.⁽²¹⁵⁾ AT&T also uses post-divestiture for its TFP study. USTA recommends basing the X-Factor on a five-year moving average, and includes post-1988 data in its TFP study. USTA also contends, however, that the relevant period for the input price differential is from 1948 to the present. No other party commented on this issue. As discussed below, we base our analysis on data from 1986 to 1995.

135. USTA criticizes AT&T's model because it includes data only from the Bell Operating Companies (BOCs), while USTA's model includes data from GTE, Sprint, SNET, and Lincoln.⁽²¹⁶⁾ USTA also finds, however, that including non-BOC data results in only a 0.1 percent difference in the X-Factor for the period from 1988 to 1994, and no difference from 1989 to 1994.⁽²¹⁷⁾ In our analysis of the record, we rely only on BOC data, as AT&T does.

136. Parties have presented a wide range of X-Factor recommendations in our two proceedings. On the basis of its model, USTA proposes X-Factors ranging from 2.7 to 3.1.⁽²¹⁸⁾ At the other extreme, AT&T and Ad Hoc propose X-Factors between 8.0 and 10.0, in part on the basis of adjusting TFP for interstate productivity.⁽²¹⁹⁾ As discussed above,⁽²²⁰⁾ MCI proposes an X-Factor of 8.5 percent based on a non-TFP methodology. Recently, a number of parties filing a joint ex parte statement have advocated an X-Factor of at least 7.5 percent, based largely on MCI's and Ad Hoc's recommendations.⁽²²¹⁾

137. The table in this Section presents the yearly X-Factor estimates (TFP plus any input price differential) submitted by USTA and AT&T, and the results of our analysis of the best methods and data available in the record of this proceeding, as well as various multi-year averages of total company productivity derived from the AT&T model and our own analysis. In its model, Ad Hoc does not present comparable yearly estimates, but only average estimates. We find that, for the 1985-95 period, the average annual growth in TFP estimated by USTA's simplified TFP model are about 0.2 percent less than our estimates. Based on more recent periods, the differences are somewhat greater. As discussed above, however, USTA has not provided any reliable estimate of the input price differential. For that reason, we cannot give any weight to its X-Factor estimates. Also as discussed above, Ad Hoc's model relies heavily on methodologies USTA employed in its original TFP model reviewed in the LEC Price Cap Performance Review and discussed in the Price Cap Fourth Further Notice. Ad Hoc's adjustments to the USTA original model do not adequately address the problems we found with that model, so we also give no weight to Ad Hoc's X-Factor estimates. We also place no weight on the joint ex parte statement's recommendation, which relies, without further analysis, on the MCI, Ad Hoc, and AT&T interstate-only proposals.⁽²²²⁾ Our analysis does incorporate a number of the methods advocated by AT&T, but AT&T's estimate of the X-Factor relies as well on methods that do not provide the best estimates of productivity from this record. Thus, we will accord some weight to AT&T's estimates of the X-Factor, but will rely primarily on our own analysis, which is a synthesis of the most persuasive treatment of TFP suggested by the record. The results of our analysis are displayed in the table below.

SUMMARY OF X-FACTORS
YEAR	FCC	AT&T	USTA
1986	-0.5%	0.2%	N/A
1987	5.0%	4.1%	N/A
1988