APPENDIX E DESCRIPTION OF PROPOSED METHODOLOGY FOR ESTIMATING SWITCHING COSTS 1. Switch Cost Data. The depreciation rate reports filed by LECs contain information on Bell Operating Companies' (BOCs') digital switches that were reported as installed between 1983 and 1995 in the states specified, with certain exceptions. To increase the reliability of analysis using these data, the proposed methodology removes the following switches from the data set: (1) switches for which there were no lines of capacity, such as those functioning solely as tandem switches; (2) switches with fewer than 1,000 lines of capacity; and (3) switches that were deemed to be "outliers" because of unusually high or low per-line costs. The following procedure was used to identify outliers: (A) if there was a gap of 20 percent or more between the per-line cost of a switch and the next lower, or higher, cost switch, that switch [and any others with lower, or higher, per-line cost] were excluded; (B) a low-priced switch that failed test A nevertheless was retained in the data set if a switch with that per-line cost would have passed test A in a previous year; (C) a high-priced switch that failed test A was retained in the data set if a switch with that per-line cost would have passed test A in a subsequent year. In addition, a small number of switches associated with apparent inconsistencies in the studies were not included in the set. In particular, for several locations in California, switches that were at the same location, but had different capacities, types, and year of installation, were reported as having the same per- line costs. These anomalies were judged to be the results of averaging by the respondent, and the switches in these locations were excluded from the data set. 2. In addition to the removal of the above outliers, the sample was restricted to the period following the divestiture of AT&T, and to those switch types that could clearly be identified as either host or remote switches. These included the DMS-100, DMS-100 remote, DMS-10, 5ESS, 5ESS remote, and EWSD switch types. In total these restrictions removed about 600 observations from a data set of nearly 3,600 observations. Thus, after exclusions, the data set compiled by the Commission in conjunction with Gabel and Kennedy and the Bureau of Economic Analysis (BEA) of the Department of Commerce consisted of 3,023 switches. In order to estimate the costs associated with the purchase and installation of new switches, and exclude the costs associated with upgrading switches, we propose to remove those switches installed more than three years prior to the reporting of their associated book-value costs. The three-year restriction resulted in the removal of 2,102 observations. The depreciation data included in the data set selected by the Commission includes the remaining 921 observations. 3. The reports made to RUS by rural telephone companies contain information on the 181 digital switches installed in 1995 and 1996. To increase the reliability of analysis using these data, we propose to remove the following observations from the data set: (1) observations containing information on switching equipment classified as upgrades to existing equipment and (2) observations containing information on switches reported as having no attached lines. These exclusions result in the removal of 42 observations. The RUS data included in the data set we select includes the remaining 139 observations. 4. Combined, the data set we propose includes 1,060 observations, 921 from the depreciation information and 139 from the RUS information. The RUS information includes a variable identifying switches as either hosts or remotes. The depreciation information does not. Therefore, an additional variable uniquely identifying switches as host switches or remote switches was added to the data set. Where data classifications were deemed unreasonable, we propose to reclassify the switch types. For example, switches identified as DMS-100 and 5ESS switches which terminated less than 2,000 customers and cost in the neighborhood of $500,000 were reclassified as remote switches. These classifications identified approximately 56% of the switches included in the combined data set as remotes. 5. Regression Formulation. The proposed regression employed is of the form: Cost =a1 + a2*Lines + a3*Host + a4*(1/Time) + a5*Lines*(1/Time) + a6*Host*(1/Time) + e where time takes on the value of 1 in 1985, 2 in 1986...15 in 1999. Regression results, including estimated coefficient values (in 1997 dollars), are: Cost = 96,610 - 13.09*Lines - 299,800*Host + 1,262,000*(1/Time) + 1,399*Lines*(1/Time) + 8,288,000*Host*(1/Time) (91,160) (35.40) (521,700) (838,300) (298.5) (3,949,000) Robust (heteroscedasticity adjusted) standard errors in parenthesis. Regression R-squared = 0.78. Proposed estimates, identified using the regression equation, for the fixed cost of host and remote switches and for the per-line cost of all switches (in 1997 dollars) are, respectively: Host Fixed Cost = a 1+ a3+ a4*(1/Time) + a6*(1/Time) Remote Fixed Cost = a1+ a4*(1/Time) Per-line Cost = a2 + a5*(1/Time) In estimating switch costs for 1999, the regression results (with time defined as 15) were converted into 1999 values using actual inflation between 1997 and 1998 and projected inflation between 1998 and 1999. Estimates for 1999, in 1999 dollars, identified using the regression equation, for the fixed cost of host and remote switches and for the per-line cost of all switches are, respectively: Host Fixed cost =(1+inflation1998)*(1+inflation1999)*(a 1+ a3 + a4*(1/15) + a6*(1/15)) Remote Fixed Cost =(1+inflation1998)*(1+inflation1999)*(a1+ a4*(1/15)) Per-line Cost = (1+inflation1998)*(1+inflation1999)*(a2 + a5*(1/15)) The inflation rate for 1998 is measured by the gross-domestic-product chain-type price index as published monthly by the Bureau of Economic Analysis of the U.S. Department of Commerce in the Survey of Current Business. The projected inflation rate for 1999 is reported in The Economic and Budget Outlook: Fiscal Years 2000-2009, published by the Congressional Budget Office. Inserting these inflation rates, the fixed cost of a host switch, the fixed cost of a remote switch, and the per-line cost for host or remote switches (in 1999 dollars) are, respectively: Host Fixed cost = (1.01)*(1.021)*(a 1+ a3 + a4*(1/15) + a6*(1/15)) Remote Fixed Cost = (1.01)*(1.021)*(a1+ a4*(1/15)) Per-line Cost = (1.01)*(1.021)*(a2 + a5*(1/15)) Inserting the coefficients from the regression analysis, the fixed cost of a host switch, the fixed cost of a remote switch, and the per-line cost for host or remote switches (in 1999 dollars) are, respectively: Host Fixed cost = (1.01)*(1.021)*(96,610 - 299,800 + 1,262,000*(1/15) + 8,288,000*(1/15)) = 447,000 Remote Fixed Cost = (1.01)*(1.021)*(96,610 + 1,262,000*(1/15)) = 186,400 Per-line Cost = (1.01)*(1.021)*(-13.09 + 1,399*(1/15)) = 83