2. Forecasts

In the forecast portion of their semiannual NRUF Reports, carriers submit their estimates of the quantity of numbers they will need during each of the next five calendar years including the year the report is submitted.⁹ The forecasts for pooling areas such as the 909 area code are done at the level of the rate center, and show the carrier's forecasted need either in thousand-blocks (pooling carriers), or in whole prefixes (non-pooling carriers).

2.1 AUDIT FINDINGS

NEED FOR POOLING BY CELLULAR CARRIERS OR AREA CODE CHANGE

The crucial factor determining when the 909 area code will run out of numbers is when cellular carriers will begin number pooling. In February of 1999, the FCC extended the deadline until November 24, 2002 for cellular carriers to support the technology that allows number pooling. The demand for phone numbers by cellular carriers appears likely to exhaust the supply of whole prefixes in the 909 area code between 12 and 18 months from now unless cellular carriers begin participating in the number pool in November 2002, as scheduled. Since the aggregate number utilization rate of the cellular carriers in the 909 area code is high (around 76%), and their subscriber growth rates are substantial, when cellular carriers begin pooling, they will probably exert heavy demands on the pool for thousand-blocks. Nonetheless, the numbering needs of all carriers, including cellular carriers, can probably be met from the 985 thousand-blocks now in the number pool for about two to three years if the cellular carriers begin pooling in November of 2002. On the other hand, if the FCC grants the cellular carriers a further extension before they are required to pool, the CPUC will need to act swiftly to split or overlay the 909 area code.

Thus, the Commission should move forward in adopting an area code change plan for the 909 area code within the next six months in order to allow adequate time for carriers to implement an area code change, should it be necessary. As the remaining prefixes are allocated in the lottery, the Telecommunications Division should periodically re-designate some of the 23 prefixes identified as reserved for the 909 number pool to the lottery, to meet non-pooling carriers' numbering needs. There are currently 13 prefixes available for the 909 lottery. Shifting 14 prefixes back from the pool to the lottery would leave a more than adequate 9 prefixes available for the 909 pool and would raise the lottery supply to 27 prefixes. Although wireless carriers forecasted numbering needs for 54 prefixes for the year 2002, all carriers returned 11 unneeded prefixes in 2000 and 10 unneeded prefixes, so far, in 2001. Taking into account returned codes, staff estimates approximately 24 to 30 codes will be needed over the next 12 to 18 months to meet carrier requests through the lottery and emergency code request processes.

FORECAST DETAILS:

VERIZON WIRELESS

Staff examined the forecast portion of the NRUF Report of Verizon Wireless in some detail. Verizon Wireless forecasted a rate of growth of --% in assigned numbers in the Los Angeles Local Access and Transport Area (LATA), which includes several area codes in addition to the 909 area code, from year-end 2000 to year-end 2005. This forecasted growth rate closely matches Verizon Wireless's actual growth in assigned numbers in the Los Angeles LATA during 1999, but ----------------- the ---% growth rate Verizon Wireless experienced in the Los Angeles LATA during 2000. Verizon Wireless experienced about --% growth in assigned numbers in the 909 area code during 1999, and --% growth during 2000. Considering this ------ of the growth rate from 1999 to 2000 and the further slowing of the California economy during 2001, staff feels that --% growth in Verizon Wireless's assigned numbers during 2001 through 2003 is a more likely outcome. Staff tested this scenario, as described at the end of the following paragraph.

To forecast its need for prefixes in the six rate centers in the 909 area code in which it holds prefixes, Verizon Wireless first multiplied its forecasted --% growth rate by its year-end 2000 actual assigned numbers in the Los Angeles LATA, and increased the resulting forecasted assigned numbers by --% for each of the remaining years of the forecast. Verizon Wireless derived its expected LATA-wide annual net gains in assigned numbers by subtracting each year-end figure from the prior year-end figure. Verizon Wireless then spread these annual net gain figures to its rate centers in the 909 area code using its actual net gains in assigned numbers during 2000, by rate center, as a proportion of its total net gains in assigned numbers in the Los Angeles LATA. Verizon Wireless divided the resulting figures by 10,000 to convert forecasted numbers to equivalent prefixes, rounding appropriately or conservatively. Staff found no errors in Verizon Wireless's forecast calculations, and found this method of rendering its forecast to the rate center level to be reasonable.

When staff checked Verizon Wireless's calculations, staff identified forecasted prefix needs in the 909 area code of --, --, and -- prefixes by Verizon Wireless during 2001, 2002, and 2003 respectively. Verizon Wireless rounded down some of its rate center forecasts for 2001 to reflect its recent receipt of prefixes in specific rate centers; the forecast it actually submitted in its June 2001 NRUF Report was for --, --, and -- prefixes in the 909 area code. Using the --% growth rate that staff projects during these three years, Verizon Wireless's resulting prefix need would ------ to --, --, and -- prefixes during these three years. Verizon Wireless has received only -- prefixes in the 909 area code so far during 2001.

OTHER CELLULAR CARRIERS

The June 2001 forecasts of all five cellular carriers in the 909 area code total 67, 49, and 49 prefixes during 2001, 2002, and 2003 respectively. One carrier's 2001 forecast for the 909 area code, which it acknowledges to be a "rosiest-case" scenario, will probably be revised downward shortly by about 40 prefixes, bringing the combined area forecast to about 27 prefixes for the last half of 2001. Because cellular carriers tend to need more additional prefixes during the second half of the year, this forecast figure compares fairly well with the 36 prefixes issued during the first 11 months of 2001. Carriers' built-in bias towards over-forecasting and staff's expectation that cellular carriers' demand for prefixes during the next two years will be similar to 2001 demand leads staff to expect cellular carriers to need about 36 prefixes during 2002 and about 40 during 2003.

ARCH WIRELESS

Arch Wireless -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------. Arch's average utilization rate of --% in the 909 area code supports this view. Similarly, the five paging companies who did submit forecasts for the 909 area code with their June 2001 NRUF Reports only forecasted an aggregate need for --- prefixes in 2001, --- in 2002, and ---- in 2003, not substantially changing the outlook for the life of the prefix supply in the 909 area code. These paging companies have received a combined total of one prefix so far during 2001.

WIRELINE CARRIERS

The forecasts of numbering needs of the wireline carriers is of less urgent concern than the forecasts of wireless carriers' needs because the 909 number pool is meeting the needs of wireline carriers almost entirely from blocks donated by other carriers, and because growth of wireline numbers in service has been substantially lower than wireless growth for at least several years. Nonetheless, staff examined the forecasting methods of the wireline carriers audited in the 909 area.

VERIZON CALIFORNIA INC.

Verizon California Inc. (Verizon) bases its NRUF Report forecasts on a database of 13 months of its monthly historical number inventory by area code and prefix. From the most recent 13 months of actual inventory, it calculates 12 months of actual monthly number usage by wire center. Verizon "smoothes" the actual data by discarding all the negative monthly usage figures and up to two "spikes" that exceed average monthly usage by 120% or more, and estimates its future monthly usage for each wire center as the average of the remaining monthly usage data points. From this resulting average monthly usage, Verizon also estimates its six-month inventory needs by wire center. From these 12 months of usage data per wire center, Verizon simulates its available number supply for each wire center on a quarterly basis for the 4 ½ or 5 year period covered by its forecast, adding thousand-blocks when its inventory level drops to its expected quarterly usage, and recalculating its available numbers to reflect each quarter's usage and block additions. Staff finds Verizon's overall approach to producing its semiannual NRUF forecasts to be systematic and well documented.

Staff's only disagreement in theory with Verizon's approach is that Verizon estimates its block needs for each wire center rather than aggregating up to the rate center level for those rate centers in which it has more than one switch. The FCC said in its Second NRO Order¹⁰ that "All applicants for growth numbering resources shall achieve a 60%¹¹ utilization threshold, calculated in accordance with paragraph (g)(3)(ii) of this section, for the rate center in which they are requesting growth numbering resources" (emphasis added). Earlier in the same Order¹², the FCC stated "We are concerned that allowing carriers to receive additional numbering resources when they have not reached the overall rate center utilization threshold will increase the likelihood that numbering resources will become stranded in underutilized switches." Later in the same paragraph the Order states that carriers that do not meet the utilization threshold for a given rate center but have a demonstrable need for additional numbering resources "may continue to seek waivers from the Commission to obtain additional numbering resources." Thus, Verizon's forecasting method seems to assume that, whenever Verizon's utilization in a rate center fails to meet required criteria for receiving numbers even though one switch in the rate center meets the criteria, every request for waiver to obtain additional numbers for the switch would be granted. It is not clear that this is a valid assumption. While this assumption may increase Verizon's forecasted block needs only slightly, Verizon should not assume that it will be able to justify, under the FCC's and CPUC's number allocation rules, all of the block needs it forecasts using this assumption.

While Verizon's forecasting method appears rational and systematic in theory, in practice it did not predict Verizon's actual numbering needs during 2001very accurately. Verizon's December 2000 NRUF Report forecasted a need for -- thousand-blocks in the 909 area code during 2001. Verizon has taken only 4 thousand-blocks from the 909 number pool so far during 2001.

PACIFIC BELL

Pacific Bell (Pacific) bases its NRUF Report forecasts on six months of its monthly historical number usage by wire center, which it extracts from a number inventory database. From this point forward, Pacific's calculations for its five-year NRUF forecast are done with pencil, paper, and calculator. Pacific estimates its future monthly usage for each wire center as the average of the most recent six months of usage. Dividing its available numbers by the average monthly usage per wire center, Pacific calculates months-to-exhaust to determine when it will need its first additional block. It appears that Pacific generally (but not consistently) subtracts six months from the months-to-exhaust figure in order to reflect requesting a block soon enough to keep a six-month inventory. Staff found one rate center in which Pacific mistakenly divided its available number supply by six months of usage rather than the monthly average usage, yielding a number of months-to-exhaust that was six times smaller (and 69 months sooner) than the correct figure. Pacific has agreed to file a revised NRUF forecast to correct this error.

Pacific introduces a more systematic error into its NRUF forecasting in its estimation of how many blocks it will need after its first additional block. For pooling areas, Pacific divides the 1,000 numbers in a block by the wire center's monthly average usage, yielding "months between block requests". Unless this division happens to yield a whole number, problems arise. For example, if the months between block requests is 1.5, the analyst is faced with the dilemma of when to put tallies on a monthly tally sheet for additional block requests. In one case for its June 2001 forecast, Pacific rounded down the resulting -- months between block requests to --, thus forecasting -- block requests per year when -- would have sufficed. In another wire center, Pacific rounded down -- months between block requests to --, forecasting -- block requests per year when -- blocks per year would have sufficed. Carriers can't request fractional blocks. A better approach than either of these "block-at-a-time" or "year-at-a-time" methods would be to multiply the monthly number usage by the number of months in the entire forecast period¹³, then divide by 1,000, yielding the total blocks needed during the forecast period. Rounding up that one number to whole blocks and spreading the resulting number of blocks appropriately among the calendar years included in the forecast would essentially eliminate the serious rounding errors that Pacific's method entails.

Comparing Pacific's December 2000 NRUF forecast with its June 2001 NRUF forecast shows disturbingly large fluctuations between forecasts. The following observations exclude the Riverside rate center, in which an identified error changed the later forecast drastically. Focusing solely on the 9 other rate centers in which Pacific operates in 909, and comparing forecasted total blocks for years 2 through 5 of these forecasts, which cover the same years (2002 through 2005), - of the - rate centers showed infinite increases (i.e., increased from forecasting 0 blocks to forecasting some positive number of blocks), while - of the - rate centers showed 100% decreases (from some positive number of blocks to zero blocks). Put differently to better emphasize the magnitude of fluctuations, - rate centers' forecasts increased by -- blocks or more in the later forecast compared to the earlier forecast, while - rate centers' forecasts decreased by -- blocks or more in the later forecast compared to the earlier forecast. In other words, - out of the - rate centers did not show reasonable consistency between the two forecasts. Such large fluctuations between two forecasts rendered six months apart makes the reliability of the overall forecast as well as the individual rate center forecasts suspect. For pooling carriers, which take blocks tied to specific rate centers, the specific rate center from which a carrier will need blocks is significant. Staff suspects that the extreme variability of Pacific's forecasts from one forecast to the next is caused in large part by Pacific's using only six months of usage history to forecast its next five years of numbering needs. This may magnify the effects of any extraordinarily large positive or negative monthly number usage data points. For this reason, staff recommends that Pacific base its estimates of monthly usage on 12 months (or as many as 24 months, if it chooses) of historical monthly usage instead of just 6 months, in order to reduce the wide swings in its forecasts from one forecast to the next.

All of the mistakes, inconsistencies, and volatility of Pacific's NRUF forecasts point to the need for Pacific to develop better tools (e.g., an electronic spreadsheet) for its forecasting personnel to use in order to minimize errors and increase consistency. Staff finds Pacific's overall approach to producing its semiannual NRUF forecasts to be rather unsystematic, not well documented, and not generally reliable.

Staff has the same disagreement with Pacific as with Verizon on the issue of forecasting by wire center rather than by rate center, as described in the section on Verizon, above. While this assumption may increase Pacific's forecasted block needs only slightly, Pacific should not assume that it will be able to justify, under the FCC's and CPUC's number allocation rules, all of the block needs it forecasts using this assumption.

Pacific's forecasting method did not predict Pacific's actual numbering needs during 2001very accurately. Pacific's December 2000 NRUF Report forecasted a need for -- thousand-blocks in the 909 area code during 2001. Pacific has not taken any thousand-blocks from the 909 number pool so far during 2001.