4. CARE and ULTS Penetration Rates - Methodological Issues

As described above, Energy Division held workshops to examine the methodologies used by the energy and telecommunication companies to calculate the penetration rates for their low-income rate discount programs, CARE and ULTS, respectively. Attachment 7 presents excerpts from the working report summarizing these methodologies, including their similarities and differences.

The most striking revelation from the workshops is that telephone service affordability studies required by the Commission do not produce penetration rates that reflect the number of customers participating in the ULTS rate discount program, relative to the number that are estimated to be eligible. Instead, Verizon and Pacific Bell calculate penetration rates that are designed to monitor universal service goals by monitoring how many households have basic phone service. In other words, the penetration rates (in the 94-98% range) that have been submitted to the Commission by telecommunications utilities in the past, and compared with CARE penetration rates by some parties in our low-income assistance proceedings,²¹ reflect the percentage of the population that have phones. Energy utilities, on the other hand, calculate penetration rates that reflect CARE program participation relative to eligibility. The workshops clearly revealed this important definitional difference.

For the purpose of these workshops, Verizon and Pacific Bell arranged to have their consultants apply the survey results from the most recent affordability study to develop a ULTS penetration rate that would be comparable to what the energy utilities calculate. The results indicate that approximately 70% of customers that have phones and are eligible for ULTS actually participate in the program. However, as discussed in the Workshop Report, this statistic may overestimate actual ULTS penetration rates somewhat because the telecommunication utilities do not conduct random verification of their ULTS enrollees (as do the energy utilities).²²

Based on the information presented in workshops, Energy Division recommends improvements to the methods used by PG&E, SDG&E, SCE and SoCal for calculating penetration rates.²³ In particular, Energy Division recommends that these utilities order the special tabulations of 2000 Census data when they are available (Fall 2002) to update demographic information on the joint relationship between household size and income. An example of this joint relationship would be the number of households with 3 members and household income in the $15,000 to $19,000 range. Because CARE income eligibility is based on both household income and size, the utilities must have this information in order to estimate the eligible CARE population.

As described in Attachment 7, the primary source of joint household size and income distribution information is the 1990 Census Public Use Microdata Sample (PUMS) data. The utilities update the 1990 PUMS data in their calculations of penetration rates by using vendor data to update household income and size information individually.²⁴ Then they statistically match these updates to the 1990 Census joint distribution data. By definition, this approach is fundamentally tied to the joint distribution data collected by the 1990 Census.

Per our reporting requirements manual, the utilities are required to update their CARE penetration rates on an annual basis, by June 1 of each year. In response to Energy Division's Workshop Report, the utilities argue that they should not be required to change the manner in which they update the 1990 Census data or be required to order special tabulations of the 2000 Census data. Rather, they prefer to wait to utilize the 2000 Census information on the joint distribution of household income and size data when the 2000 Census PUMS tables are published during the later half of 2003. They argue that the effort required to make such a change to their calculations of penetration rates at this time would be significant and potentially duplicative of the efforts to be undertaken during Phase 2 of the Needs Assessment Study.

We do not find these arguments persuasive. At whatever point we make the transition to using 2000 Census data as the basis for the joint distribution of household income and size, there will be costs and effort involved. Energy Division estimates that the expense to procure these special tabulations is relatively modest ($800 to $1,000 per utility), which none of the utilities refute in their comments. What they do argue is that the special tabulations lack the flexibility and corresponding tables that are published later with the 2000 Census PUMS data, thus increasing the effort and costs of updating the penetration rates. Although the comments are not clear on this point, it appears that the utilities are requesting to wait until their June 1, 2004 annual update to modify the joint distribution of household income and size based on 2000 Census data, i.e., by waiting until the PUMS tables are available later in 2003.

We do not find it acceptable to continue for two more updates (June 1, 2002 and June 1, 2003) relying on the joint distribution data from the 1990 Census as a primary source, as the utilities' comments suggest. Moreover, the utilities will need to move ahead in using these special tabulations for the Needs Assessment Study, given the current timetable for initiating our Phase 2 assessment.²⁵ Accordingly, we direct the utilities to update their penetration rates for the June 1, 2003 report using special tabulations from the 2000 Census.

Energy Division also recommends that PG&E, SCE, SDG&E and SoCal proceed to complete certain sensitivity tests that were described in the technical workshop materials as currently underway. Specifically, these sensitivity tests involve income smoothing alternatives, variations in small area weighting and an analysis of whether differences between program and Census definitions of household incomes influence eligibility estimates significantly.²⁶ In the interest of time and consistency, Energy Division recommends that the utilities retain their current vendor, AGS, for updating Census data at this time. For similar reasons, Energy Division recommends that the utilities' current methods for classifying rural and urban areas remain unchanged.

In its comments on the workshop report, SCE argues that the utilities should be permitted to reconsider the use of AGS data as the source for current year data on household size and income distributions, and pursue the option of returning to Claritas as the source for future updates. SCE also contends that the current method for classifying rural and urban areas could be improved by using the Goldsmith method, rather than the Rural Health Council method, and that the utilities should be authorized to make this change. Finally, SCE recommends that the utilities expand their efforts to validate estimates against independent data sources, beyond those already undertaken to date.

We will adopt Energy Division's recommendations at this time. As discussed above, the utilities will be updating their penetration rate estimates with 2000 Census data for their June 1, 2003 annual report. To introduce additional methodological refinements between now and then could divert limited resources from this updating task, which we consider to be of highest priority. As part of the PY2003 or subsequent program planning cycles, the utilities may propose additional sensitivity tests or validation activities for our consideration. In doing so, the utilities should present a proposed budget for each of the recommended activities.

Moreover, nothing in this record indicates that changes in vendors would be critical to the accuracy of the utilities' current estimates. As Energy Division points out, AGS and Claritas are both well-respected firms and a lot of time and effort went into deciding on AGS as the vendor for updating Census data. With respect to the benefits of the Goldsmith method over the Rural Health Council method, this methodological issue is also being considered in Phase 1 of the Needs Assessment Study, as is SCE's recommendation to evaluate how master-metered households may upwardly bias estimates of household eligibility.²⁷ We will address these recommendations when we address all other Phase 1 issues. Other longer-term refinements to the methodology used by PG&E, SCE, SoCal and SDG&E are being considered in the Needs Assessment Study, such as the development of estimates of willingness to participate in CARE. We may direct PG&E, SDG&E, SCE and SoCal to incorporate further changes into their methodology as we consider the Needs Assessment Study reports and recommendations.

Energy Division recommends no changes to the current methodologies utilized by Avista and Southwest to calculate CARE penetration rates. As described in Attachment 7, Avista uses a simplified method of applying Census information to estimate its CARE eligible population. In particular, Avista based its calculation of the eligible population within its service territory on the relationship between El Dorado County and California poverty statistics. This assumption (11%) is likely to overestimate the number of eligible population (thereby underestimating actual program penetration) since it does not reflect the unique characteristics of many of the South Lake Tahoe residences, e.g., as recreational or seasonal homes.

We believe that the manner in which Avista estimates its eligible population needs to be improved. To this end, we direct Energy Division to ensure that Phase 2 of the Needs Assessment Study is designed to obtain income and household size data specific to Avista's service territory for the purpose of estimating the number of CARE eligible homes. This may be done by using current Census data, by arranging for the study consultants to conduct an independent survey, or a combination of both. The utilities currently funding the study will absorb the cost of this additional task, which should be relatively small.

Southwest uses an outside company to survey their universe of customers for economic and demographic data. Their estimates are based on cross-tabulations of income and household size and are checked against current Census data for reasonableness. We concur with Energy Division that this approach appears reasonable.

With regard to the telecommunication utilities, the Workshop Report notes several areas where improvement in the calculation of ULTS penetration rates and eligibility verification is needed. Energy Division has referred these recommendations to the Assigned Commissioner in the ULTS Proceeding, R.98-09-005, for further consideration.

Based on the record in this proceeding, we find that the CARE penetration rate methodologies currently utilized by PG&E, SDG&E, SCE, SoCal, Avista and Southwest to be reasonable, subject to the improvements noted above.