Research Plan: 2015 Upstream and Residential Downstream Lighting Program Impact Evaluation (ED_I_Ltg_4) California Public Utilities Commission Date: July 22, 2016 LEGAL NOTICE This document was prepared as an account of work sponsored by the California Public Utilities Commission. It does not necessarily represent the views of the Commission or any of its employees except to the extent, if any, that it has formally been approved by the Commission at a public meeting. For information regarding any such action, communicate directly with the Commission at 505 Van Ness Avenue, San Francisco, California 94102. Neither the Commission nor the State of California, nor any officer, employee, or any of its contractors or subcontractors makes any warranty, express or implied, or assumes any legal liability whatsoever for the contents of this document. DNV GL - Energy – www.dnvgl.com/energy 1.1.1 Table of contents 1.1.1 1.1.2 1.1.3 Table of contents List of figures List of tables 2 2 2 1 INTRODUCTION ............................................................................................................ 4 1.1 Background 1.2 Objectives 4 1.3 2015 overview 5 1.4 Measure groups 7 1.5 Key research questions 9 2 STUDY METHODS ....................................................................................................... 12 2.1 Overview 12 2.2 2.2.1 2.2.2 2.2.3 2.2.4 2.2.5 Data sources Consumer surveys Supplier interviews Retail lamp stock inventories and shopper intercept surveys Lamp Choice Model Program tracking data and other secondary data sources 12 13 15 18 19 21 2.3 2.3.1 2.3.2 2.3.3 2.3.4 Approach Measure quantity adjustments Gross Savings Net savings NTG analyses 22 22 23 28 37 3 REFERENCES.............................................................................................................. 38 4 1.1.2 List of figures Figure 1. Conventional gross and net savings overview ...................................................................... 28 1.1.3 List of tables Table 1. Quantity and share of upstream and residential downstream lighting measures by delivery mechanism, sector and IOU, 2015 .................................................................................................... 5 Table 2. Quantity and share of residential and nonresidential upstream lighting measures by measure group and IOU, 2015 ............................................................................................................................... 6 Table 3. Quantity and share of residential downstream lighting measures by measure group and IOU, 2015 7 Table 4. Uncertain measures to be prioritized - 2015 upstream and residential downstream lighting program impact evaluation ........................................................................................................................... 8 Table 5. Quantity and share of upstream residential lighting measures by measure group and IOU, 2015 (high-priority measure groups highlighted) ......................................................................................... 9 Table 6. Research questions and associated data sources - 2015 upstream residential lighting impact evaluation ................................................................................................................................... 13 Table 7. 2015 California actual market shares for lamps in the A-lamp replacement category ................. 16 Table 8. Replacement lamp categories and included technologies......................................................... 17 Table 9. 2015 California “counterfactual” market shares for lamps in the A-lamp replacement category . 17 Table 11. Number of lamps addressed by intercepted lamp purchasers by measure group, 2013—2016 ... 19 Table 12. Ex post share of residential versus nonresidential upstream lighting measures by IOU, 2015 ..... 23 Table 13. Overview of 2015 upstream and residential downstream impact evaluation inputs and data sources ................................................................................................................................................. 24 DNV GL - Energy – www.dnvgl.com/energy Table 14. Residential lighting HOU estimates by upstream lighting measure group and IOU ..................... 25 Table 15. Residential lighting peak CF by measure group and IOU ....................................................... 25 Table 16. CFL HVAC interactive effects factors by IOU ........................................................................ 27 Table 17. Net savings calculation example for A-lamp replacements (≤ 30 W) ...................................... 31 Table 18. Illustration of net savings calculation for A-lamp replacements (≤ 30 W) relative to traditional incandescent/halogen lamps .......................................................................................................... 33 Table 19. Illustration of net savings calculation for A-lamp replacements (≤ 30 W) – all positive and negative savings allocated to efficient technologies ............................................................................ 34 Table 20. Illustration of net savings calculation for A-lamp replacements (≤ 30 W) using different baseline technologies ................................................................................................................................ 36 DNV GL - Energy – www.dnvgl.com/energy 1 INTRODUCTION 1.1 Background This document presents the research plan for an impact evaluation of the California’s investor-owned utilities’ (IOU) 2015 upstream and residential downstream lighting energy-efficiency programs. The January 2016 Energy Division (ED) & Program Administrator Energy Efficiency Evaluation, Measurement and Verification (EM&V) Plan, (Version 6) refers to this study as Lighting-16: 2015 Lighting Impact Evaluations and Efficiency Savings and Performance Incentive (ESPI) evaluations. 1 This research plan addresses nonresidential upstream and all residential lighting measures included in the IOUs’ 2015 lighting programs. 2 DNV GL will closely coordinate this work with other relevant studies. This plan presents the overarching goals and objectives of the 2015 Upstream and Residential Downstream Lighting Program Impact Evaluation included in work order (WO) ED_I_Ltg_4 (2013-2015 Lighting Impact Evaluation and Market Research Studies). The plan also presents details regarding the scope, budget, and timeline associated with this effort. This draft version of the plan currently represents collaboration between the California Public Utilities Commission (CPUC) ED and DNV GL staff along with input from the CPUC ED’s ex ante team and California’s electric and dual-fuel IOUs (PG&E, SCE, and SDG&E). Section 3 of this document provides complete references for the reports and other sources cited herein. 1.2 Objectives The January 2016 EM&V Plan states that “the goal of this study is to collect impact evaluation data on selected lighting measures for the 2015 IOU lighting programs (and provide results as rapidly as possible).” Objectives include: 1. Verify the quantity of IOU-discounted upstream lighting products that were shipped, sold and installed by residential and nonresidential customers within the PG&E, SCE and SDG&E service territories during the 2015 program period; 2. Estimate the gross energy and peak coincident demand impacts from these measures; and 3. Determine an appropriate estimate of net energy and demand impacts. The impact evaluation will verify and validate the IOU-reported energy savings and peak demand reduction claims for upstream and residential downstream lighting measures and programs through a combination of pass-through approaches (relying on information gathered in previous evaluation cycles) and primary data collection. The evaluation will provide feedback to the IOUs regarding program performance, report on the outcomes of long-term market transformation efforts, and provide results that will support future improvements to IOU program design. To support these efforts, the evaluation team will prepare and present timely and accurate reporting to the CPUC and/or IOUs. 1 2 CPUC ED, 2016. We refer to this document herein as the “January 2016 EM&V Plan” or “the Plan.” Work order ED_I_Com_1 (Non-Residential Downstream Impact Evaluation) will address nonresidential, non-upstream lighting impact evaluation issues. Another work order (TBD) will address nonresidential, non-upstream lighting market research issues. DNV GL - Energy – www.dnvgl.com/energy 1.3 2015 overview The tables below summarize the IOUs’ 2015 upstream (residential and nonresidential) and residential downstream lighting program activities. As shown in Table 1, they provided incentives for nearly 15.7 million lighting units via these three combinations of delivery mechanism and market sector in 2015. At 90% of the total, residential upstream measures composed the largest share of total relevant lighting units, followed by nonresidential upstream and residential downstream (at 6% and 4%, respectively). SCE provided incentives for nearly three times as many units as either of the other IOUs at nearly 10.3 million (65% of total units compared to 22% for PG&E and 13% for SDG&E). Table 1. Quantity and share of upstream and residential downstream lighting measures by delivery mechanism, sector and IOU, 2015 Measure Quantity Delivery Mechanism/Sector Upstream – residential Upstream – nonresidential Downstream – residential Total Percent of Column PG&E SCE SDG&E Total PG&E SCE SDG&E Total 3,133,197 9,182,908 1,804,935 14,121,040 91% 90% 92% 90% 208,751 579,401 132,517 920,669 6% 6% 7% 6% 98,312 496,518 32,041 626,871 3% 5% 2% 4% 1,969,493 15,668,580 100% 100% 100% 100% 3,440,260 10,258,827 Percent of Total 22% 65% 13% Source: 2015 IOU program tracking data (preliminary analyses) DNV GL - Energy – www.dnvgl.com/energy 100% Table 2 shows the quantity of residential and nonresidential upstream lighting measures for which the IOUs provided incentives in 2015 as well as the share comprised by each measure group by IOU. Again, indoor LED A-lamps comprised the largest share of total measures across IOUs at 25%, followed by indoor CFL reflector lamps (17%), indoor LED reflector lamps (17%) and indoor high-wattage CFLs (> 30 W; 17%). Together, these four measure groups comprised approximately three-quarters of all residential upstream lighting measures in 2015 (76%). Table 2. Quantity and share of residential and nonresidential upstream lighting measures by measure group and IOU, 2015 Measure Group Measure Quantity Percent of Column (Measure Quantity) SDG& Grand PG&E SCE E Total PG&E SCE SDG&E Grand Total Indoor CFL > 30 Watts - 2,315,789 199,135 2,514,924 0% 24% 10% 17% Indoor CFL 3-way - 356,432 24,093 380,525 0% 4% 1% 3% 91,306 1,126,146 265,251 1,482,703 3% 12% 14% 10% 499,902 98,892 29,773 628,568 15% 1% 2% 4% Indoor CFL fixture - 8,973 351 9,324 0% <1% <1% <1% Indoor CFL globe - 140,577 6,243 146,820 0% 1% <1% <1% Indoor CFL reflector - 2,540,155 88,564 2,628,719 0% 26% 5% 17% 665,456 - 213,596 879,052 20% 0% 11% 6% 1,381,811 1,548,699 808,637 3,739,147 41% 16% 42% 25% 694,575 1,626,451 299,753 2,620,779 21% 17% 15% 17% - 195 5 200 0% <1% <1% <1% 8,898 - 2,051 10,949 <1% 0% <1% <1% 3,341,948 9,762,309 1,937,452 15,041,709 100% 100% 100% 100% Indoor CFL A-lamp Indoor CFL basic Indoor LED fixture Indoor LED A-lamp Indoor LED reflector Outdoor HID Refrig. case LED lighting Total Source: 2015 IOU program tracking data (preliminary analyses) DNV GL - Energy – www.dnvgl.com/energy Table 3 shows the total quantities of residential downstream lighting measures in the IOUs’ 2015 programs by measure group and IOU as well as the share of total units comprised by each measure group. Forty percent of the IOUs residential downstream lighting units in 2015 were indoor LED A-lamps. All other measure groups accounted for much smaller shares of the residential downstream lighting sales; at the top of that list are indoor basic CFLs and indoor CFL fixtures (at 13% and 12%, respectively). Table 3. Quantity and share of residential downstream lighting measures by measure group and IOU, 2015 Measure Group Percent of Column (Measure Quantity) Measure Quantity PG&E SCE SDG&E Grand Total PG&E SCE SDG&E Grand Total - 2 - 2 0% <1% 0% <1% 75,139 1,336 3,047 79,522 76% 0% 10% 13% Indoor CFL fixture 2,768 71,378 1,783 75,929 3% 14% 6% 12% Indoor CFL reflector 1,651 72 - 1,723 2% <1% 0% <1% 1 53,548 - 53,549 <1% 11% 0% 9% 641 53,235 795 54,671 1% 11% 2% 9% 11,776 218,024 20,141 249,941 12% 44% 63% 40% - - 1,738 1,738 0% 0% 5% <1% 1,752 23,020 - 24,772 2% 5% 0% 4% 801 22,264 1,147 24,212 1% 4% 4% 4% 2,315 1,254 - 3,569 2% <1% 0% 1% Outdoor CFL fixture 956 33,911 2,820 37,687 1% 7% 9% 6% Outdoor CFL reflector 303 26 - 329 <1% <1% 0% <1% Outdoor LED fixture 209 393 570 1,172 0% 0% 2% 0% - 18,055 - 18,055 0% 4% 0% 3% 98,312 496,518 32,041 626,871 100% 100% 100% 100% Indoor CFL > 30 Watts Indoor CFL basic Indoor controls* Indoor LED fixture Indoor LED A-lamp Indoor LED night light Indoor LED reflector Indoor other Outdoor CFL basic Outdoor LED other Grand Total * Complete measure group name is “Indoor Controls Wall or Ceiling Mounted Occupancy Sensor.” Source: 2015 IOU program tracking data (preliminary analyses) 1.4 Measure groups As described above, the IOUs provided incentives for nearly 15.7 million lighting units via their upstream and residential downstream lighting programs in 2015. Because upstream lighting measures comprise more than 96% of these measures delivered via the upstream mechanism and/or to residential IOU customers (see Table 1), the 2015 evaluation will focus on upstream measures. As such, the evaluation team will “pass through” all of the ex ante savings estimates associated with residential downstream units in the 2015 impact evaluation. Similarly, because upstream residential measures alone comprise 90% of all units delivered via the upstream mechanism and/or to residential IOU customers, the 2015 impact evaluation will focus exclusively on residential upstream measures. The evaluation team will also “pass through” all of the ex ante savings estimates associated with nonresidential upstream units in the 2015 impact evaluation— however, as we will discuss in Section 2.3.1.2 below, we will assess how the IOUs allocate upstream lighting DNV GL - Energy – www.dnvgl.com/energy units between the residential and nonresidential sectors and update this parameter as part of the evaluation as necessary. For residential upstream measures, the January 2016 EM&V plan notes that “the Energy Savings Performance Incentive [ESPI] is a key driver in schedules for prioritizing and delivering impact studies for uncertain measures as noted earlier. 3” The EM&V Plan identifies lists of measures to be prioritized in 2015 impact evaluations as a result of the uncertainty associated with specific impact evaluation parameters for these measures. The list of uncertain lighting measures relevant to this evaluation includes high-wattage CFLs, LED lamps, LED downlight fixtures, and plug-in LED night lights. Table 4 provides more details from the ESPI measure list. Table 4. Uncertain measures to be prioritized - 2015 upstream and residential downstream lighting program impact evaluation Measures IOUs Areas of Uncertainty Screw-in CFLs of all types with PG&E, Update of the gross baseline assumptions to account for the type and wattage of the lamp being replaced. Previous net-to-gross (NTG) wattage values greater than SCE, 30 Watts SDG&E studies have not focused on high wattage CFL lamps. Screw-in LED lamps including downlight replacement kits Update baseline assumptions and net savings including replaced lamp PG&E, for early retirement versus standard practice for normal replacement SCE, and replace-on-burnout; market move to LED technology and rapidly SDG&E changing products and pricing requires verification of assumptions. LED night lights SCE Assumptions of replaced equipment and use uncertain. Source: CPUC ED, 2016. Table 44 (2015 Uncertain Measures to be Prioritized). Pages 181—182. The evaluation team also examined the upstream residential lighting measure groups included in the IOUs’ 2015 program tracking data to help develop priorities for the 2015 impact evaluation (Table 5). Because the “Lighting indoor LED fixture” measure group includes LED downlight replacement kits and other measures, and because LED downlight replacement kits are the specific measures of interest for the uncertain measures list, we have disaggregated the “Lighting indoor LED fixture” measure group into two subgroups in the table to distinguish between downlights and other fixture types. 4 As shown, the IOUs’ 2015 residential upstream lighting programs included relatively large quantities of measures for three of the ESPI measure groups—high-wattage CFLs (> 30 W) and LED lamps (which includes LED A-lamps and LED reflector lamps). LED downlight replacement kits comprised only 6% of total upstream residential lighting program units across IOUs but approximately 20% of PG&E’s total units. And although LED night lights appear on the 2015 ESPI list, none of the IOUs provided incentives for these 3 4 CPUC ED, 2016. Page 6. For the purposes of concise display in Table 5, we disaggregated the “Lighting indoor LED fixture” measure group into its component measures and then grouped these into two subgroups: “LED recessed downlight” and “LED interior fixture – 28W.” The “LED recessed downlight” subgroup includes the following measures: LED RECESSED DOWNLIGHT <10 W LED; LED RECESSED DOWNLIGHT >=10 W TO 12 W LED; LED RECESSED DOWNLIGHT >12 W TO 25 W LED; LED RECESSED DOWNLIGHT 10 WATT; LED RECESSED DOWNLIGHT 11 WATT; LED RECESSED DOWNLIGHT 12 WATT; LED RECESSED DOWNLIGHT 15 WATT; LED RECESSED DOWNLIGHT 16 WATT; LED RECESSED DOWNLIGHT 21 WATT; LED RECESSED DOWNLIGHT/RETROFIT 9 WATT; COMMERCIAL-LED RECESSED DOWNLIGHT 12 WATT; COMMERCIAL-LED RECESSED DOWNLIGHT 16 WATT. The “LED interior fixture – 28W” subgroup includes only LED INTERIOR FIXTURE 28 WATT. DNV GL - Energy – www.dnvgl.com/energy measures through their upstream residential lighting programs in 2015. As such, DNV GL does not plan to conduct specific impact evaluation activities to address LED night lights. In addition to these, three other measure groups comprise at least 10% of total units for at least one of the IOUs, including CFL reflector lamps, CFL A-lamps, and basic spiral CFLs. Because of the relatively large quantities of these measures in the 2015 portfolio, we consider these additional measure groups as high priorities for the 2015 impact evaluation in addition to the ESPI measure groups of high-wattage CFLs, LED A-lamps, LED reflector lamps, and LED downlight replacement kits. These 7 measure groups comprise approximately 96% of all residential upstream lighting units included in the IOUs’ 2015 programs and will be the focus of the 2015 residential upstream lighting impact evaluation. Table 5. Quantity and share of upstream residential lighting measures by measure group and IOU, 2015 (high-priority measure groups highlighted) Measure Quantity Measure Group Percent of Column (Measure Quantity) SDG& Grand PG&E SCE E Total PG&E SCE SDG&E Grand Total 1,298,902 1,457,390 751,725 3,508,017 41% 16% 42% 25% - 2,389,085 84,362 2,473,447 0% 26% 5% 18% 653,235 1,531,653 261,222 2,446,110 21% 17% 14% 17% - 2,177,284 191,316 2,368,600 0% 24% 11% 17% 85,828 1,058,807 249,352 1,393,987 3% 12% 14% 10% LED recessed downlight*† 625,324 - 210,514 835,837 20% 0% 12% 6% Indoor CFL basic 469,908 93,042 24,280 587,230 15% 1% 1% 4% Indoor CFL 3-way - 335,070 24,093 359,163 0% 4% 1% 3% Indoor CFL globe - 140,577 5,979 146,556 0% 2% <1% 1% Outdoor LED fixture - - 2,051 2,051 0% 0% <1% <1% LED interior fixture – 28W† - - 37 37 0% 0% <1% <1% Outdoor CFL fixture - - 5 5 0% 0% <1% <1% 9,182,908 1,804,93 5 14,121,040 100% 100% 100% 100% Indoor LED A-lamp Indoor CFL reflector Indoor LED reflector* Indoor CFL > 30 Watts Indoor CFL A-lamp* Grand Total 3,133,197 Source: 2015 IOU program tracking data (preliminary analyses) * 2015 uncertain measure (ESPI) † Included in the “Lighting indoor LED fixture” measure group in previous tables. 1.5 Key research questions The January 2016 EM&V Plan lists one key research question for the 2015 impact evaluation: 1. What are the ex post savings results? DNV GL staff thus worked with the stakeholders listed in Section 1.1 to identify additional relevant research questions for the evaluation. These include: 2. What is the appropriate baseline for residential upstream LED lamps? 3. What is the appropriate baseline for residential upstream CFLs? DNV GL - Energy – www.dnvgl.com/energy 4. What is the freeridership level for residential upstream LED lamps? 5. What is the freeridership level for residential upstream CFLs? 6. In the absence of updated metering data, what are the best estimates of hours of use (HOU) by IOU? 7. Is there any evidence that consumer satisfaction differs with LED lamps that meet the California Quality LED Specification 5 versus LED lamps that do not meet the spec? Question 1 (What are the ex post savings results?) addresses the key research question in the January 2016 EM&V Plan as well as the 2015 uncertain measure list’s requirement to “update … net savings” for screw-in LED lamps for PG&E, SCE, and SDG&E (as well as updating net savings for other measures). 6 Question 2 (What is the appropriate baseline for residential upstream LED lamps?) addresses the 2015 uncertain measure list’s requirement to “update baseline assumptions … including replaced lamp for early retirement versus standard practice for normal replacement and replace-on-burnout” for PG&E, SCE, and SDG&E. To clarify, however, the evaluation team’s interpretation of the phrase “replaced lamp for early retirement versus standard practice for normal replacement and replace-on-burnout” is that the uncertainty lies in the extent to which “normal replacement” activities (or “standard practice”) involve “early retirement” or “replace-on-burnout.” As such, the evaluation team will focus evaluation efforts (in part) on understanding the share of installed LED lamps that replace functioning lamps (early retirement) versus lamps that have stopped working (replace-on-burnout). The evaluation will also address the baseline technology mix for LED lamps. Question 3 (What is the appropriate baseline for residential upstream CFLs?) addresses the 2015 uncertain measure list’s requirement to “update … the gross baseline assumptions to account for the type and wattage of the lamp being replaced” for “screw-in CFLs of all types with wattage values greater than 30 Watts” for PG&E, SCE, and SDG&E. Question 4 (What is the freeridership level for residential upstream LED lamps?) and Question 5 (What is the freeridership level for upstream CFLs?) address concerns raised by program administrators and other stakeholders during discussions regarding results from the Impact Evaluation of 2013-14 Upstream and Residential Downstream Lighting Program 7 and discussions of study priorities for the 2015 impact evaluation. Question 6 (In the absence of updated metering data, what are the best estimates of HOU by IOU?) and Question 7 (Is there any evidence that consumer satisfaction differs with LED lamps that meet the California Quality LED Specification versus LED lamps that do not meet the spec?) arose from the same discussions and IOU comments on the 2013-14 upstream and residential downstream lighting program impact evaluation report. • 5 6 7 Question 6 (regarding HOU estimates) will address (at least in part) concerns from SDG&E regarding the differences in average hours of use for replacement lamps in SDG&E’s service territory versus the other IOUs’ service territories. We expect a more definitive answer to this question from the CEC, 2014. CPUC ED, 2016. Table 44 (2015 Uncertain Measures to be Prioritized). Pages 181—182. The same source serves as the basis for all remaining quotations in this section of the research plan (Section 1.5). DNV GL, 2016. DNV GL - Energy – www.dnvgl.com/energy forthcoming in-home California lighting inventory and metering study. 8 As such, efforts to address this question as part of the 2015 upstream and residential downstream lighting impact evaluation will be limited. • The same is true with regard to Question 7 (regarding the California Quality LED Specification): the in-home inventory and metering study will provide a more concrete opportunity to investigate satisfaction with LED lamps in use in IOU customer households. That study will thus provide a more appropriate vehicle for addressing the IOUs’ concerns regarding the possible market transformational effects of their incentives for LED lamps that meet the spec. The 2015 upstream and residential downstream lighting impact evaluation will provide high-level insights regarding IOU customer satisfaction with LED lamps in general (as we will describe further in Section 2.2.2 below). Note that there is no specific research question to address “assumptions of replaced equipment and use” for LED night lights for SCE as mentioned in the 2015 uncertain measure list. SCE did not provide incentives for LED night lights through its 2015 upstream residential lighting program (nor did PG&E or SDG&E). 8 DNV GL is currently engaged in the scoping phase (Phase 1) of this study and expects to deliver a finalized research plan for the study in early 2017. The January 2016 EM&V Plan provides more detail on the Phase 1 study (CPUC ED, 2016). DNV GL - Energy – www.dnvgl.com/energy 2 STUDY METHODS 2.1 Overview The 2015 impact evaluation will build on the methods from 2010-12 and 2013-14 upstream lighting impact evaluations. The proposed approach for 2015 will still result in estimates of gross savings, net savings, and NTG ratios but will attempt to calculate net savings at the market level. The approach relies upon marketlevel estimates of energy consumption by technology and measure group (including technologies such as incandescent and halogen lamps, which are not included in the IOUs’ programs). Based on market-level estimates of net savings along with estimates of gross savings, we propose to calculate NTG ratios for each of the evaluated upstream lighting measure groups. Below we describe the impact evaluation data sources (Section 02.3) and overall approach (Section 2.3). 2.2 Data sources The 2015 impact evaluation relies upon seven data sources. Table 6 below shows the data sources aligned with the research questions described above. We provide more details on these sources and the overall study approach below. DNV GL - Energy – www.dnvgl.com/energy Table 6. Research questions and associated data sources - 2015 upstream residential lighting impact evaluation Retail Lamp Stock Inventories Shopper Intercept Surveys Lamp Choice Model Program Tracking Data Secondary Data Sources 1. What are the ex post savings results? Supplier Interviews Research Question Consumer Surveys Data Sources X X X X X X X 2. What is the appropriate baseline for residential upstream LED lamps? X X X 3. What is the appropriate baseline for residential upstream CFLs? X X X 4. What is the freeridership level for residential upstream LED lamps? X X X X X 5. What is the freeridership level for residential upstream CFLs? X X X X X 6. In the absence of updated metering data, what are the best estimates of HOU by IOU? X 7. Is there any evidence that consumer satisfaction differs with LED lamps that meet the California Quality LED Specification versus LED lamps that do not meet the spec? X X X X* * Supplier interviews will not focus on consumer satisfaction with LED lamps explicitly but instead focus on suppliers’ perspectives regarding the influence of the CEC specification on their LED lamp sales – e.g., whether they would sell lamps that meet the CEC spec in absence of the program. 2.2.1 Consumer surveys DNV GL will conduct surveys with residential electric customers of PG&E, SCE, and SDG&E to obtain several key inputs to the 2015 upstream and residential downstream lighting program impact evaluation. These include: 1. Installation rate. The consumer surveys will obtain details regarding the installation rates for CFLs and LED lamps. The surveys will do so at the technology level (i.e., one installation rate for CFLs and another for LED lamps). As part of this discussion, we will address the ESPI priority related to the share of LED lamps that replace functioning lamps (early retirement) versus LED lamps that replace lamps that have stopped working (replace-on-burnout). 2. Satisfaction with LED lamps. While the energy savings associated with LED lamps that meet the California Quality LED Specification do not differ from the savings for LED lamps that do not meet the spec, the IOUs have suggested that the superior quality of the lamps that meet the spec will have market transformational consequences in California. The theory is that the higher quality lamps will result in high consumer satisfaction with LED lamps, ultimately leading to repeat purchases. DNV GL - Energy – www.dnvgl.com/energy Conversely, LED lamps that do not meet the spec are of lower quality, which could result in consumer dissatisfaction that could reduce or eliminate their future purchases of LED lamps. As described above in Section 1.5, the forthcoming in-home inventory and metering study will provide a more concrete opportunity to investigate satisfaction with LED lamps in use in IOU customer households, and thus provide a better vehicle to address the IOUs’ concerns than the 2015 impact evaluation. However, the consumer survey will briefly address satisfaction with LED lamps in general. While we will be unable to distinguish during the survey whether the LED lamps purchased by consumers do or do not meet the CEC spec, consistent and widespread satisfaction with LED lamps may suggest that the specification does little to affect consumer satisfaction. Conversely, if survey results suggest that consumer satisfaction with LED lamps is inconsistent, this may indicate that there are differences in satisfaction with LED lamps that could be attributable (at least in part) to the CEC spec. Such a result could demonstrate the need for more focused consumer research on this topic. 3. HOU assessment. Results from the 2010 California residential lighting metering study suggest lower lighting HOU for SDG&E than for PG&E or SCE. 9 Using the consumer survey, we will test the hypothesis that average daily HOU are as different among the IOUs as suggested in the 2010 metering study, and in the direction suggested by the study (i.e., that average daily HOU for SDG&E is approximately one-third lower than the other IOUs’ estimates). Specifically, we will test whether consumer-reported HOU are as different as indicated by the metering study. This test will not confirm actual HOU, nor will we use consumer-reported values as an alternative HOU estimate. However, the test will offer an indicator to either support or refute the differences across IOUs from the metering study. To construct the test, we will estimate the average daily HOU for three high-use rooms (e.g., kitchen, living room, and master bedroom) for each IOU using 2010 metering study results. This will establish the magnitude and directionality of differences in average daily HOU by IOU for the purposes of this assessment. During the consumer surveys, we will ask respondents to estimate the average daily HOU for lamps in each of these three rooms. Once we complete the consumer surveys, we will average the HOU estimates from each respondent across the three rooms and then aggregate by IOU. The test will have one of 2 outcomes: a) Null hypothesis is not rejected: customer-reported HOU do not provide strong evidence that the difference among IOUs is less than that found in the metering study. In this case, we will continue to use IOU-specific HOU values based on the 2010 metering study. b) Null hypothesis is rejected: customer-reported HOU provide strong evidence that the difference among IOUs is less than that found in the metering study. In this case, we will develop statewide estimates of HOU based on the 2010 metering study and use these statewide estimates for SDG&E in the 2015 impact evaluation. For PG&E and SCE, we will use the IOUspecific HOU estimates based on the 2010 metering study. As stated in Section 1.5 above, we do not intend this assessment to represent the best and final estimate of average daily HOU by IOU. Instead, the goal is to expend a reasonable level of effort in the near-term to reassess the validity of the 2010 estimates in advance of updated (and more precise) estimates from the forthcoming in-home lighting inventory and metering study. 4. 9 Baseline technology mix and wattage for high-wattage CFLs. During the 2016 consumer surveys, DNV GL will attempt to identify purchasers of high-wattage CFLs (> 30 W). We will ask these consumers to identify the technology and wattage of the lamp replaced with high-wattage CFLs. DNV GL recognizes that it will be challenging to obtain meaningful point estimates of lamp wattage over the phone. While we would likely obtain more reliable information from face-to-face interactions with consumers, no such interactions are planned in support of the 2015 impact evaluation. As such, the consumer survey represents the best-available opportunity to attempt to address this issue. Nonetheless, it may be necessary to rely on secondary data sources (or a KEMA, Inc., 2010, Volume 1. See Table 56: Recommended Average Residential Daily HOU by IOU with Confidence Intervals (Appendix B, page 112) which suggests average daily HOU estimates of 1.3 hours/day for SDG&E and 1.9 hours/day for both PG&E and SCE. DNV GL - Energy – www.dnvgl.com/energy combination of secondary sources and other supporting information) to estimate baseline wattage for high-wattage CFLs for the 2015 impact evaluation. DNV GL notes that the forthcoming in-home lighting inventory and metering study provides an ideal opportunity to address baseline technology and wattage with consumers for installed high-wattage CFLs and have incorporated this objective into preliminary plans for that study. The inventory and metering effort will provide more concrete estimates of baseline wattage for high-wattage CFLs. 5. Inputs into LCM. Section 2.2.4 below describes the LCM in detail as well as the expected inputs to and outputs from the model for this study. From the consumer surveys, we will obtain: a. Distribution of lamp purchases by retail channel. One challenge in using the LCM in previous impact evaluations is that the model relies, in part, on results from the in-store shopper intercept surveys. The intercept surveys are, by necessity, based on a convenience sampling approach. To improve the LCM’s ability to represent the distribution of lamp purchases by retail channel within the purchaser population, we will include questions in the consumer surveys to address recent purchase locations (retail channels). b. Customer demographics together with recent lamp purchase information. As described in the detailed LCM discussion, we will use the consumer survey respondents, rather that intercept survey respondents, to represent the universe of lamp purchase decisions. We will first review the 2015 consumer telephone survey results to determine whether they provide a sufficient level of detail to serve as LCM inputs. If so, will use the 2015 consumer telephone survey results regarding the distribution of lamp purchases by retail channel to ensure that the LCM reflects this distribution rather than the distribution of intercepted purchasers. If not, we will collect information in the 2016 consumer survey to serve this purpose. 6. Insights into channel shift. The consumer survey will include questions to address whether shoppers typically choose from among the lamps available to them in a preferred retail channel or if they will move to a different retail channel to purchase a preferred type of lamp. The supplier interviews will also get their perspectives as to which retail channels and which lamp technologies are most likely to be experiencing channel shift. As part of the consumer survey task, we will also investigate: • Which mode or modes may be most appropriate (for example, a telephone survey, online survey, etc.) • Whether we need to rake survey results to balance demographics • Whether it is necessary to address these objectives in one or more consumer survey efforts (to minimize the length of individual surveys). 2.2.2 Supplier interviews DNV GL will conduct approximately 40 interviews with lamp supplier representatives with the primary goals of assessing the 2015 program’s influence on lamp sales and estimating market-level sales of replacement lamps to California consumers in 2015. The most efficient respondent population for the supplier interviews is likely the manufacturers’ representatives (because there are many more retailers than manufacturers) and retail buyers for large retail chains. However, there may be occasions when interviews with retail buyers for smaller retail chains or even store managers may be appropriate. For example, if our interviews with lighting manufacturers or retail buyers leave certain retail channels underrepresented, we may want to add some interviews with smaller retail buyers or store managers from these underrepresented channels. We refer to all of these representatives collectively as “suppliers” herein. DNV GL will contact these DNV GL - Energy – www.dnvgl.com/energy representatives as appropriate using one general interview guide that we will modify to suit the specific respondent characteristics. 10 Starting from scratch (rather than modifying interview guides used in prior upstream lighting program evaluations), DNV GL will create an interview guide to address: • Share of 2015 lamp sales represented by upstream lighting program sales. In advance of each interview, we will prepare a summary of the quantity of lamps for which each supplier received 2015 upstream lighting program incentives, by technology and measure group. During the interview, we will ask the representative to confirm the quantity and to estimate the percentage of total lamp sales for that measure group represented by the program-discounted quantity. • Total 2015 lamp sales. For each replacement lamp category, we will ask each supplier representative to estimate the proportion of his or her company’s 2015 California sales represented by each common alternative lamp type within that replacement lamp category. For example, for the A-lamp replacement category, we will ask respondents to estimate the proportion of their 2015 California sales represented by each lamp type shown in Table 7. Since we will know the actual quantities of their lamp sales through the upstream program from the program tracking data, we will be able to estimate their sales volumes for the other lamp types not discounted by the program based on their relative market shares. Table 7. 2015 California actual market shares for lamps in the A-lamp replacement category A-Lamp Replacements (≤ 30 W) % of Supplier’s 2015 Lamp Sales in California CFL A-lamps/basic spiral CFLs through the upstream program __% CFL A-lamps/basic spiral CFLs not through the upstream program __% LED A-lamps through the upstream program __% LED A-lamps not through the upstream program __% EISA-compliant incandescent/halogen A-lamps __% Traditional incandescent/halogen A-lamps __% Total 2015 California sales of A-lamp replacements 100% We will repeat this exercise for two other replacement lamp categories (reflector replacements and high-wattage replacements) although in these cases, the lamp technologies in each category will be more limited than for A-lamps. Table 8 shows all three relevant replacement lamp categories and their included technologies. 10 For example, if a small independent retailer received a large volume of lamps through the program, it may be most effective to speak with the store manager regarding these lamps (possibly in addition to the manufacturer’s representative[s]). If a major retail chain received a large volume of lamps through the program, it may be most effective to speak with the corporate-level retail buyer regarding these lamps (again, possibly in addition to the manufacturer’s representative[s]). DNV GL - Energy – www.dnvgl.com/energy Table 8. Replacement lamp categories and included technologies Technology • Replacement Lamp Category A-lamp Reflector High-Wattage X CFL spiral X CFL (same shape as category name) X X LED X X EISA-compliant incandescent/halogen X Traditional incandescent/halogen X X X X Estimated 2015 lamp sales in absence of the upstream program. Next, we will ask the supplier representative to provide similar market share estimates for the counterfactual scenarios— in other words, what their market shares would have been in absence of the program. Table 9 shows the specific line items for which we will request market share estimates using the example of the Alamp replacement category. Table 9. 2015 California “counterfactual” market shares for lamps in the A-lamp replacement category A-Lamp Replacements (≤ 30 W) % of Supplier’s 2015 Lamp Sales in California without Upstream Program CFL A-lamps/basic spiral CFLs __% LED A-lamps __% EISA-compliant incandescent/halogen A-lamps __% Traditional incandescent/halogen A-lamps __% Total 2015 California sales of A-lamp replacements 100% It is possible that the absence of the program would have led not only to a shifting of the relative sales shares of each lamp type but also to a change in the total volume of 2015 California sales. To account for this, we will ask the interviewees whether there would be any change in the total volume of their 2015 California sales for lamps in a given replacement lamp category (e.g., A-lamps, reflectors, high-wattage lamps) in the absence of the program. If so, we will obtain details regarding the directionality (larger or smaller) and the magnitude (% change). It is also possible that some suppliers or retail buyers will be unwilling to estimate their relative sales shares with and without the program. In such cases, we will give them the option of providing a more generic estimate for the 2015 California market as a whole. • Insights into channel shift. We will also ask the supplier representatives whether they would expect sales to shift among channels for any lamp types in absence of the program. For example, they may report that the absence of the program might lead to a significant shift in sales in LED Alamps from the discount channel to the home improvement channel due to discount stores not stocking these lamps at full price. In such cases, we would ask the suppliers or retail buyers to try to estimate the direction and magnitude of these channel shift effects. • LED lamps that do not meet the CEC spec. DNV GL also recognizes the importance of addressing LED lamps that do and do not meet the California Quality LED Lamp Specification in the supplier interviews. This issue is particularly important given that during previous interviews, some suppliers suggest that they would not have sold any LED lamps that met the spec in absence of the upstream lighting program. We will ask supplier representatives whether they sold any LED lamps that met the CEC spec without program incentives. We will also ask them whether they would have sold any LED DNV GL - Energy – www.dnvgl.com/energy lamps that met the spec in 2015 if the program did not exist. Finally, we will ask them to describe any other influences of the CEC specification on their LED lamp sales in California. The supplier sample will include major lamp manufacturers whether or not the manufacturers received incentives from the IOUs through the 2015 upstream lighting program, and whether or not the manufacturers produced lamps that were eligible for program incentives. In other words, if there are major manufacturers that produce only (for example) traditional incandescent lamps or program-eligible lamps that did not receive IOU discounts in 2015, we will attempt to interview them to obtain estimates of 2015 lamp sales. The ultimate goal is to best represent market-level sales of replacement lamps through brickand-mortar retail stores to California consumers. 2.2.3 Retail lamp stock inventories and shopper intercept surveys In late 2015 through early 2016, DNV GL staff conducted stock inventories of lamps for sale in California retail stores throughout PG&E, SCE, and SDG&E service territories. We have conducted similar inventories roughly every 6 to 12 months since 2011. During the last several phases of stock inventories, we conducted concurrent shopper intercept surveys with lamp purchasers. The stock inventories gathered detailed information regarding all residential replacement lamps stocked in the stores other than linear fluorescent lamps, and the shopper intercept surveys focused on shopper purchasing decisions and installation intentions for the newly-purchased lamps. 11 DNV GL field staff conducted surveys in chain and independent retail stores, including stores that participated in the IOUs’ 2015 upstream lighting programs as well as non-participating stores. Field staff spent a minimum of four hours in each store completing the lamp stock inventories and attempting to intercept shoppers. We completed 207 lamp stock inventories purchaser 433 intercept surveys during the winter 2015-16 period. Field staff completed purchaser intercept surveys opportunistically—that is, with individuals who were shopping during the time periods during which we conducted intercept surveys in specific stores. As such, results from the intercept surveys may not represent the broader population of shoppers purchasing replacement lamps at various stores throughout the year. Nonetheless, given the range in timeframes and store types in which we conducted these surveys, results provide general indications of shopper preferences, price sensitivity, lamp installation intentions, and so on. Additionally, the 2015 impact evaluation will leverage consumer survey results to better reflect the distribution of lamp sales by retail channel in our analyses. To support the 2015 impact evaluation, DNV GL will leverage the retail lamp stock inventory results in two ways: 1. To support the LCM. The LCM reflects the lamp prices and availability that DNV GL staff observed in retail stores during the retail stock inventories. We will update the LCM to ensure that it represents the mix of lamp stock found on retail shelves during the most recent phase of stock inventory research conducted by DNV GL (winter 2015-16). 11 Field researchers also conducted shopper intercept surveys with respondents who were not purchasing lamps (non-purchaser shopper intercept surveys), but the impact evaluation focuses on surveys with lamp purchasers only because these surveys included detailed questions regarding lamp replacement intentions. DNV GL - Energy – www.dnvgl.com/energy 2. To support the supplier interviews. We use the retail stock inventory results to estimate the share of lamp stock by technology and measure groups for specific retail chains and manufacturers to support the supplier interview effort described above. We will also use the purchaser intercept survey results in two ways: 1. To support estimates of baseline technology mix. During the purchaser intercept surveys, we asked respondents to identify the lamp technologies they intended to replace with their new lamps (for up to two lamps per respondent). Error! Reference source not found. shows the number of lamps in each measure group addressed by purchaser intercept surveys during the summer 2013, winter 2014-15, and winter 2015-16 data collection periods. As shown, we have robust data from the winter 2014-15 and winter 2015-16 periods (which include the program period) for many measure groups, but if include one additional period (summer 2013), we have robust data on which to base these analyses for six of the seven groups. The exception is high-wattage CFLs, for which the purchaser intercept surveys addressed only 13 lamps during the relevant data collection periods. For high-wattage CFLs, we will thus attempt to obtain details regarding the baseline technology mix from the 2016 consumer surveys. Table 10. Number of lamps addressed by intercepted lamp purchasers by measure group, 2013—2016 Measure Group Quantity of Lamps Addressed by Purchaser Intercept Surveys Summer 2013 Winter 2014-15 Winter 2015-16 Overall 180 181 101 462 CFL A-lamp ≤ 30 W 22 12 13 47 CFL reflector ≤ 30 W 31 21 9 61 4 4 5 13 LED A-lamp (all wattages) 54 94 190 338 LED reflector (all wattages) 57 93 81 231 348 405 399* 1,152 Basic spiral CFL ≤ 30 W High-wattage CFL (> 30 W) Total Lamps Note that the total lamps shown here is smaller than the total quantity of intercepted purchasers during the winter 2015-16 period because some survey respondents purchased lamp types not shown in this table (e.g., traditional incandescent lamps). 2. To support the LCM. The intercept surveys ask lamp purchasers to rank a set of hypothetical lamp technologies as most-to-least likely to buy given specific price points. The LCM uses these data to estimate model coefficients, as described in Section 2.2.4. This evaluation will update model coefficients using 2015-16 purchaser intercept survey data. 2.2.4 Lamp Choice Model DNV GL developed the LCM for the 2010-12 impact evaluation to quantify consumer responses to upstream lighting incentives. The underlying program theory is that providing discounts for a CFL or LED lamp makes that CFL or LED lamp a more attractive choice than the alternatives. The program’s effects include providing lower-priced lamps in retail stores than would be available without the program and enabling specific retail DNV GL - Energy – www.dnvgl.com/energy stores (such as those in the discount channel) to stock lamps that otherwise would not meet their price point requirements. Discrete choice models are the analytical framework designed to address these types of effects. These models combine the relevant information about each possible choice— for example, the lamp price and consumer characteristics—and assign a probability to each of the choices. To estimate attribution, we use the model to estimate the market share for each lamp type with and without the program in place. Our approach is as follows: 1. Re-estimate the LCM. We will re-estimate the LCM with 2015-16 purchaser intercept survey data. Re-estimation ensures that the model reflects consumer price sensitivities regarding the different lamp technologies available in brick-and-mortar retail stores during the 2015 program period. 2. Estimate market shares under three scenarios by channel. We will estimate market shares using a simulation-based approach. The simulation takes two inputs. The first input is a representation of consumers. We will use the respondents in the 2016 consumer survey to form this input. Unlike the intercept survey data, the consumer survey data are a representative sample of consumers in the lamp market. The second input is a representation of available lamp choices based on retail lamp stock inventory data. We will run the simulation against three scenarios: • With-program scenario. This scenario reflects the lamp prices and availability that DNV GL observed in retail stores during the retail lamp stock inventories conducted in winter 201516. This scenario results in an estimate of the share of program lamp sales for each modelled technology. • No-discount scenario. This scenario reflects the lamp prices that consumers would have seen in California retail stores in 2015 in the absence of IOU discounts. DNV GL estimated price differences based on clearly-labeled IOU discounts in the stores or by matching lamps to program tracking data. This scenario results in a counterfactual estimate of market shares that would have occurred if only prices on program-discounted lamps changed due no program activity. This scenario represents the first of two “no program” scenarios. • Constrained scenario. In addition to the price effects described in the no-discount scenario, the constrained scenario reflects stocking changes that would have occurred in absence of the program. During our supplier interviews, when a respondent indicated that his or her company would not have sold a lamp through a specific retail channel in the absence of the program, we consider those lamps to be program-reliant. (For example, if a supplier representative told us he or she would not have sold LED A-lamps to discount stores without upstream lighting program incentives, we would consider the presence of these lamps in discount stores to be program-reliant.) This scenario results in a counterfactual estimate of market shares if program-reliant lamps were not in stores and if the IOUs did not discount lamps. This scenario represents the second “no program” scenario. In prior evaluations, we have used these results to directly estimate a NTG ratio that represents the relative percent changes in a given technology’s market shift. However, this approach leads to some challenges in the presence of competing efficient technologies. For the 2015 evaluation, the team proposes to use the LCM to estimate the overall difference in market shares with and without the upstream lighting program considering all technologies jointly. The market share differences together with market volume and unit energy consumption (UEC) will yield market-level net savings. We will combine the market shares generated from the LCM with the market shares estimated by supplier interviews to produce the best estimate for program impacts on replacement lamp market share, and thus net savings (see Section 2.3.3 for greater detail). DNV GL - Energy – www.dnvgl.com/energy The evaluation team recognizes that this approach depends on a robust estimate of total market volume for each set of competing technologies. In the event that a market volume sales estimate cannot be produced with sufficient rigor in the time and budget available for this evaluation, the evaluation team will leverage the vetted and proven methodology implemented in the 2013-14 upstream lighting impact evaluation. This methodology uses supplier and consumer data to estimate NTG ratios, and applies those ratios to gross savings, which produces net savings. Nevertheless, a market-level interpretation of net savings is the most comprehensive method to evaluate upstream program performance, and will be the preferred approach for this evaluation. 2.2.5 Program tracking data and other secondary data sources The 2015 impact evaluation will also rely on program tracking data as well as the result of other EM&V efforts (including the California Lighting and Appliance Saturation Study [CLASS] 12). 2.2.5.1 Program tracking data Each of the IOUs uploads program tracking data onto a centralized server to be downloaded by the evaluation team. The team will analyze, clean, re-categorize, reformat, and merge these separate datasets into one program tracking database. 13 The tracking data will provide details regarding the quantity of lighting measures shipped as well as details regarding the manufacturers and retailers involved in the 2015 programs, the latter of which will provide the sample frame for the in-depth lighting supplier interviews. It will also produce estimates of average discounted wattage per measure group, and will be used to pass through ex-ante results for specific parameters not addressed in the 2015 evaluation. 2.2.5.2 Other EM&V reports We will mine data from other EM&V studies to support the overall evaluation efforts described herein for work order ED_I_Ltg_4. These data sources may include: 12 13 • Impact Evaluation of 2013-14 California Upstream and Residential Downstream Lighting Programs (DNV GL, 2016). This study included all lighting measures associated with upstream delivery mechanisms and all downstream lighting measures targeted at the residential sector. The impact evaluation focused on seven measures that collectively accounted for over 90% percent of each IOUs’ ex ante net savings from upstream and residential downstream measures. These measures included basic spiral CFLs, CFL A-lamps, CFL globes, CFL reflectors, LED A-lamps and LED reflectors. Several of the impact evaluation parameters and methodologies used in the 2013-14 program cycle will be “passed through” and utilized in the current evaluation. • Impact Evaluation of the 2010-12 Residential/Advanced/Upstream Lighting Programs (DNV GL, 2014b). As with the 2013-14 impact evaluation, this study included all lighting measures associated with upstream delivery mechanisms and all downstream lighting measures targeted at the residential sector. This evaluation focused heavily on four measure groups that each individually made up at least 1 percent of each IOU’s portfolio-level kWh or kW claims: CFL A-lamp, CFL basic, CFL globes, and CFL reflectors. Many methodologies that were used in the 2013-14 evaluation and that will continue to be used in this evaluation were developed under WO28. • CLASS (DNV GL, 2014a). The CLASS study updates and augments saturation and efficiency characteristics from previous CLASS studies conducted in 2005 and 2000 for use in understanding future energy savings potential and past accomplishments in the residential sector. The 2012 CLASS DNV GL, 2014a. Note that we have completed some preliminary analyses of these data in support of this research plan. DNV GL - Energy – www.dnvgl.com/energy study included onsite observations on a sample of 1,987 single-family, multi-family and mobile home residences with individually-metered electric accounts across the service territories of PG&E, SCE and SDG&E. The 2015 impact evaluation will rely on CLASS to update the delta Watts, HOU, and peak coincidence factors for CFLs and LED lamps (see Section 2.3.2 for more detail). 2.3 Approach As described in Section 1.2 above, the 2015 impact evaluation has three main objectives: 1. Verify the quantity of IOU-discounted upstream lighting products that were shipped, sold and installed by residential and nonresidential customers within the PG&E, SCE and SDG&E service territories during the 2015 program period 2. Estimate the gross energy and peak coincident demand impacts from these measures 3. Determine an appropriate estimate of net energy and demand impacts As such, the evaluation approach includes three key components: 1. Adjust rebated measure quantities 2. Develop gross savings inputs 3. Develop net savings estimate We explore these components in more detail below. As part of this evaluation, we will attempt to estimate the program’s net impacts relative to changes in the California market for residential replacement lamps (as noted in Section 2.1 above). We will then produce a NTG ratio for each measure group by dividing the net savings associated with the measure group by the gross savings associated with the measure group. In the event that collected data are unable to produce a rigorous estimate of total market volume, the evaluation will instead rely on the vetted approach used in the 2013-14 evaluation, and apply a NTG ratio to the gross savings estimate. 14 2.3.1 Measure quantity adjustments In recent evaluations of the upstream and residential downstream lighting programs, evaluators have applied three adjustments to the quantity of rebated measures claimed by the IOUs as having been sold to their residential and nonresidential customers during the program period. As with the 2013-14 impact evaluation, 15 the 2015 evaluation will maintain the values associated with each of these adjustments, including: 1. Quantity of IOU-discounted products shipped by participating manufacturers to retailers as determined through the verification of a sample of program invoices/applications 14 15 2. Percent of IOU-discounted products purchased by residential versus nonresidential customers 3. Percent of IOU-discounted products purchased by non-IOU customers (i.e., leakage) Note that the net savings methodology from the 2013-14 evaluation requires a gross savings estimate without efficient-to-efficient lamp replacements. To accommodate the mixed baseline, the methodology estimates net savings by applying a NTG ratio to the legacy gross savings estimate. It then calculates the alternative, mixed-baseline net-to-gross ratio by holding net savings constant, and dividing it by the mixedbaseline gross savings. See the 2013-14 impact evaluation for more details and an example of that calculation (DNV GL, 2016). DNV GL, 2016. DNV GL - Energy – www.dnvgl.com/energy We provide more detail regarding measure quantity adjustments below. 2.3.1.1 Invoice verification The 2013-14 residential and upstream lighting impact evaluation relied upon the same invoice verification rate as the 2010-12 evaluation. 16 For the 2010-12 evaluation, evaluators verified the quantity of IOUdiscounted units shipped by participating manufacturers to retailers based on their review of a sample of program invoices and applications. The evaluation estimated an ultimate verification rate of 100% for all IOUs and retail channels. As such, we will apply the 100% verification rate in the 2015 impact evaluation. 2.3.1.2 Residential versus nonresidential To estimate the portion of upstream CFLs that are installed in nonresidential applications, the 2013-14 and 2010-12 evaluations relied on the results of two onsite survey studies conducted during the 2010-12 period—the CLASS 17 and the Commercial Market Share Tracking Study. 18 These efforts yielded the residential versus nonresidential shares of total upstream lighting program measures shown in Table 12. We will apply these estimates in the 2015 impact evaluation. Table 11. Ex post share of residential versus nonresidential upstream lighting measures by IOU, 2015 Share of Upstream Measures IOU Nonresidential Residential PG&E 7% 93% SCE 6% 94% SDG&E 6% 94% Overall 7% 93% Source: DNV GL, 2016. 2.3.1.3 Leakage Leakage is defined as the quantity of program-discounted upstream lamps that “leak” out of the collective IOU service territories—in other words, IOU-discounted lamps that are ultimately installed outside of the IOUs’ service territories. Due to the lack of strong data supporting leakage, no adjustment to quantity was applied in the impact evaluations of the IOUs’ 2013-14 or 2010-12 programs. We will therefore apply the same 0% leakage rate in the 2015 impact evaluation. 2.3.2 Gross Savings This section describes the methodology for analyzing gross impacts. There are six elements to these analyses: 16 17 18 DNV GL, 2014b. DNV GL, 2014a. Itron, Inc., 2014. DNV GL - Energy – www.dnvgl.com/energy 1. Average daily HOU 2. Average percent of installed measures operating at peak (coincidence factor, or CF) 3. Wattage displaced by IOU-discounted products (delta Watts) 4. HVAC interactive effects (IE) 5. Unit energy savings (UES) estimates (kWh/year and peak kW) 6. Installation rate Table 13 provides an overview of these inputs and their associated data sources for the 2015 impact evaluation by market sector (residential and nonresidential) and program delivery mechanism (upstream and downstream). Table 12. Overview of 2015 upstream and residential downstream impact evaluation inputs and data sources Market Sector and Program Delivery Mechanism Evaluation Input Residential Upstream HOU 2010 metering study; 2016 consumer survey (to test difference in 2010 HOU results for SDG&E) Peak coincidence factor 2010 metering study Delta Watts a. Baseline technology mix b. Baseline wattage c. Replacement technology mix and wattage a. 2016 consumer survey for high-wattage CFLs (> 30 W); Purchaser intercept surveys for all other measures b. 2016 consumer survey for high-wattage CFLs (> 30 W); 2012 CLASS inventory for other measures; all with assumed baseline technology replacement ratios from purchaser intercept surveys c. 2015 program tracking data HVAC interactive effects Pass through from ex ante Installation rate 2016 consumer survey Residential Downstream Nonresidential Upstream Pass through from ex ante* * With exception of residential/nonresidential split. 2.3.2.1 HOU Estimates of the average daily HOU for residential lighting in this evaluation will use the same approach as in the 2013-14 impact evaluation with the exception of a 2016 consumer survey exercise to test the validity of IOU-specific estimates for SDG&E (as detailed in Section 2.2.1 above). As explained, should the consumer survey exercise conclude that the customer-reported HOU do not provide strong evidence that the difference among IOUs is less than that found in the metering study, we will apply IOU-specific HOU values based on the 2010 metering study (as shown in Table 14). If the customer-reported HOU do provide strong evidence that the difference among IOUs is less than that found in the metering study, we will develop statewide estimates of HOU based on the 2010 metering study and use these statewide estimates for SDG&E in the 2015 impact evaluation. In this case, we will still use the IOU-specific HOU estimates based on the 2010 metering study for PG&E and SCE. DNV GL - Energy – www.dnvgl.com/energy Table 13. Residential lighting HOU estimates by upstream lighting measure group and IOU Evaluated Upstream Lighting Measure Group PG&E SCE HOU 90% CI MSB CFL basic spiral ≤ 30 W 1.6 MSB CFL A-lamp ≤ 30 W MSB CFL reflector ≤ 30 W SDG&E HOU 90% CI ±0.1 1.9 1.5 ±0.2 1.7 MSB CFL high-wattage (> 30 W) Overall HOU 90% CI HOU 90% CI ±0.2 1.4 ±0.2 1.7 ±0.1 1.9 ±0.2 1.3 ±0.3 1.6 ±0.2 ±0.3 1.9 ±0.2 1.2 ±0.4 1.7 ±0.2 * * * * * * 1.9 ±0.2 LED A-lamp, all wattages * * * * * * 2.1 ±0.2 LED reflector, all wattages * * * * * * 2.1 ±0.2 * Sample sizes were too small to produce IOU-specific estimates. We will apply the overall estimates in calculating impacts. Please refer to DNV GL, 2014b for more details regarding metering study sample sizes. 2.3.2.2 Peak coincidence factor Peak CF represents the average percent of time that a lamp is used during the peak period. The peak periods vary by climate zone. Similar to the HOU estimates, the estimates for CF will be passed through from the 2013-14 impact evaluation. We derived these estimates based on the logger data collected for the 2010 metering study 19 and applied to in-home lamp inventory data collected during CLASS. 20 Table 15 shows the peak CF values that we will pass through in the 2015 impact evaluation. Table 14. Residential lighting peak CF by measure group and IOU Measure Group PG&E SCE SDG&E Overall Peak CF 90% CI Peak CF 90% CI Peak CF 90% CI Peak CF 90% CI MSB CFL basic spiral ≤ 30 W 0.05 ±0.01 0.07 ±0.01 0.04 ±0.02 0.06 ±0.01 MSB CFL A-lamp ≤ 30 W 0.05 ±0.02 0.06 ±0.02 0.04 ±0.02 0.05 ±0.01 MSB CFL reflector ≤ 30 W 0.05 ±0.02 0.06 ±0.02 0.04 ±0.03 0.06 ±0.02 MSB CFL high-wattage (> 30 W) * * * * * * 0.06 ±0.01 LED A-lamp, all wattages * * * * * * 0.06 ±0.02 LED reflector, all wattages * * * * * * 0.06 ±0.02 * Sample sizes were too small to produce IOU-specific estimates. We will apply the overall estimates in calculating impacts. Please refer to DNV GL, 2014b for more details regarding metering study sample sizes. 2.3.2.3 Delta Watts For each evaluated measure group, the delta Watts calculation consists of replacement lamp wattage, baseline technology mix, and baseline wattage. We will calculate delta Watts as the difference between a measure group’s technology-weighted baseline wattage and its average replacement lamp wattage. We provide more details on these inputs below. Replacement lamp wattage 19 20 DNV GL, 2014b. DNV GL, 2014a. DNV GL - Energy – www.dnvgl.com/energy We will calculate the average replacement lamp wattage for each measure group based on the 2015 program tracking data. Baseline technology mix As described in Section 2.2.3 above, the 2015 impact evaluation will rely upon estimates of baseline technology mix by measure group from in-store purchaser intercept survey data. Baseline wattage We will use the baseline technology mix (described above) and the average baseline lamp wattage from the CLASS in-home lamp inventory data to estimate baseline wattage all measure groups except high-wattage CFLs (> 30 W). For high-wattage CFLs, we will use 2016 consumer survey data to estimate baseline wattage. For all measure groups, we will estimate baseline wattage based on the average lamp wattage for each baseline technology weighted by the proportion of all baseline measures represented by each technology. Because we collected the CLASS data prior to extensive adoption of EISA-compliant general purpose lamps, we will perform a fairly simple calculation to account for an increased adoption in EISA-compliant lamps. We will consider the midpoint between the average CLASS incandescent wattage and the average 2015-16 Shelf inventory EISA wattage as the incandescent/EISA baseline wattage. The forthcoming in-home lamp inventory and metering study will obtain more accurate and detailed information regarding EISA-compliant lamp adoption among the California IOUs’ electric customers. 2.3.2.4 HVAC interactive effects The Database for Energy Efficient Resources (DEER) includes savings factors for kWh, kW and therms savings for indoor CFLs and LED lamps. The DEER team develops HVAC interactive effects (HVAC IE) estimates for internal load changing measures such as interior lighting and appliances and other plug loads measures. These effects can alter the "direct" electric and gas impacts for those measures due to resulting changes in heating and cooling HVAC system energy use. As such, these savings factors are applied to the direct impacts as a multiplier for both kWh and kW and factor of therm/kWh for therm impacts. Table 9 shows the multipliers used for the upstream CFL measures evaluated in the 2013-14 upstream and residential downstream impact evaluation report, which we obtained from DEER 2011. 21 DEER does not currently distinguish among different CFL measures (e.g., basic spiral CFLs versus CFL A-lamps), therefore we apply the same HVAC IE factors to all evaluated CFL and LED lamp measure groups. We show the factors in Table 9 by IOU and building type. The residential and commercial factors are the weighted average of all existing building types for residential or commercial buildings with a given IOU. 21 DNV GL, 2016. DNV GL - Energy – www.dnvgl.com/energy Table 15. CFL HVAC interactive effects factors by IOU Building Type CFL HVAC Interactive Effect Adjustment PG&E SCE SDG&E kWh 1.02 1.07 1.03 kW 1.33 1.40 1.23 Therms -0.025 -0.019 -0.018 kWh 1.06 1.12 1.12 kW 1.21 1.24 1.23 Therms -0.0061 -0.0032 -0.0028 Residential Commercial IOU Source: DEER 2011. 2.3.2.5 Installation rate In the 2015 impact evaluation (as in the 2013-14 evaluation), the installation rate is the percentage of all upstream lamps that that will ultimately be installed. DNV GL will develop updated installation rates for CFLs and LED lamps in the 2016 consumer survey (described in Section 2.2.1 above). 2.3.2.6 Unit energy savings For measures discounted through the IOUs’ 2015 upstream lighting programs, we will compute unit energy savings (UES) using Equation 1 below and peak demand reductions using Equation 2 below. Equation 1. Unit energy savings 1 ππππβ 365 ππππππππ ππππβ οΏ½ = Δπππππππππ π πΏπΏ [ππ ] ∗ π»π»π»π»πππΏπΏ [β] ∗ πππππππΏπΏ οΏ½ ∗ ∗ πΌπΌπΈπΈπΏπΏ [ππππβ] 1000 ππβ 1 π¦π¦π¦π¦π¦π¦π¦π¦ π¦π¦π¦π¦π¦π¦π¦π¦ Where: ΔWattsL = average displaced (delta) wattage for IOU-discounted lamp measure group, L, in Watts (W) HOUL = annual average HOU for IOU-discounted lamp measure group, L, in hours (h) IEL = DEER HVAC interactive effects in kilowatt-hours (kWh) DNV GL - Energy – www.dnvgl.com/energy Equation 2. Peak demand reduction 1 ππππ ππππ οΏ½ = Δπππππππππ π πΏπΏ [ππ ] ∗ πΆπΆπΉπΉπΏπΏ ∗ πππππππΏπΏ οΏ½ ∗ πΌπΌπΈπΈπΏπΏ [ππππ ] 1000 ππ π¦π¦π¦π¦π¦π¦π¦π¦ Where: ΔWL = average displaced (delta) wattage for IOU-discounted lamp measure group, L, in Watts (W) CFL = average percent on at peak for IOU-discounted lamp measure group, L IEL = DEER HVAC interactive effects in kiloWatts (kW) 2.3.2.7 Estimate gross savings We will calculate gross savings as energy savings, peak demand reduction, and increased natural gas usage (based on interactive effects). Figure 1 illustrates the components used to calculate gross savings. Figure 1. Conventional gross and net savings overview 2.3.3 Net savings We will build our net savings analysis on methods used in the previous two cycles, with some improvements based on our experiences with that work. As in past evaluation cycles, we rely on supply-side and demandside estimates of program influence and integrate the results based on the strengths and limitations of each approach. Our approach to estimating net savings involves the following steps: 1. Market share estimation. We will determine the market share of each competing technology with and without the program in place separately for 3 categories of replacement lamps: a. A-lamp replacements (≤ 30 W): CFL A-lamps/basic spiral CFLs; LED A-lamps; EISAcompliant incandescent/halogen A-lamps; traditional incandescent/halogen A-lamps b. Reflector lamp replacements (≤ 30 W): CFL reflector lamps, LED reflector lamps, incandescent/halogen reflector lamps c. High-wattage lamps: high-wattage CFLs (> 30W), equivalent EISA-compliant incandescent/halogen A-lamps; equivalent traditional incandescent/halogen A-lamps 2. Market volume estimation. We will then determine the total 2015 sales in California for lamps within each set of competing technologies. 3. Net savings calculation. We will determine the difference in total consumption for each set of competing technologies with and without the program in place. This difference is based on the difference in market shares, the total market sales, and the unit energy consumption (UEC) for each measure. We will calculate the UEC based on the gross parameters defined in Section 2.3.2, DNV GL - Energy – www.dnvgl.com/energy except that instead of applying delta Watts, we will use 2015-16 retail lamp stock inventory data to calculate and apply the average wattage of each lamp technology on the 2015 market. We will allocate the total net savings for each replacement lamp category to the efficient technologies. NTG calculation. Finally, we will determine the NTG ratios for each measure group as the ratio 4. of net savings to gross savings at the measure group level. We describe these steps in more detail below (including the basic methods and planned improvements). 2.3.3.1 Market share estimation The following section describes the two perspectives we will use to estimate lighting market shares in the presence and in the absence of the program. Demand-side estimate: LCM This method determines market share by estimating and applying a set of lamp choice models (described in more detail in Section 2.2.4). For the 2015 impact evaluation, we will use the winter 2015-16 purchaser intercept survey data to fit the models. We will use similar model forms in the 2015 impact evaluation as in the 2013-14 evaluation. Once we have estimated the LCM model coefficients, we will apply the model to a simulated population of lighting purchases to determine the market shares of competing technologies with and without the program in place. The relative change in market share for each program-supported technology is the basis for the net savings estimate. The 2013-14 evaluation encountered two challenges in applying the LCM: 1. Prior to their 2013-14 programs, the IOUs offered CFL A-lamps ≤ 30 W and LED A-lamps ≤ 30 W in limited quantities, with a larger share of upstream incentives directed toward basic spiral CFLs ≤ 30 W. In 2013-14, the programs offered all three of these competing technologies in varying combinations. As a result, the effect of the program on one technology was dependent on which of the other technologies were simultaneously present, with or without program discounts. We thus had to segment our analyses based on the combination of technologies present which led to a complex analysis and thin data for some channel-segment combinations. In addition, assessing the effect of the program separately for each technology required interpretation of how to account for the effect increasing one efficient technology at the expense of another. In the 2015 program, however, there are few cases in which discounted technologies compete within the same replacement lamp category (but there is still the potential for competition between discounted energy-efficient lamps and energy-efficient lamps not discounted in the same stores). 2. We simulated the universe of lighting purchase choices based on purchaser intercept respondents. If we did not have enough respondents for a particular replacement lamp category in a particular channel, the LCM was unable to estimate a NTG ratio for that channel and the relevant measure group(s). In addition, we did not have a basis for weighting the simulated responses within a channel and segment to the population level. To address these challenges, the 2015 impact evaluation analyses include the following changes. DNV GL - Energy – www.dnvgl.com/energy 1. We will simulate the universe of lighting purchase choices based on consumer survey results rather than on in-store purchaser intercept results. The consumer survey results represent a broader crosssection of lighting purchasers, and will not be restricted to customers present in particular stores on particular days. Use of this population reduces concerns as to how representative the simulated universe is, and also will make more cases available to represent particular choices. 2. We will account for the effect of competing efficient technologies by considering the overall change in market shares across all competing technologies. Supply-side estimate: supplier interviews As part of the supplier interviews, we will ask each respondent to estimate the market share of each technology within a replacement lamp category, with and without the program in place (described in more detail in Section 2.2.2 above). These market share estimates will not be channel-specific. Consolidated market share estimates For each replacement lamp category, we will obtain the proportion of lamp purchases made in each retail channel from consumer survey responses. We will weight the channel-specific market shares by the channel proportions to determine the overall market shares based on the demand-side approach. To determine the ultimate with-program and without-program market share estimates, we will take a weighted average of the supplier-reported market share estimates with and without the program and those based on the LCM. 2.3.3.2 Market volume estimation We will estimate total market volume for each replacement lamp category by two means: 1. Stock turnover analysis of 2012 CLASS data 2. Supplier survey responses Stock turnover analysis We will estimate the total volume of purchases for each replacement lamp category by assuming steady state total stock. For each technology, we will have: • The total number of lamps in sockets during 2012 (from CLASS 22) • The rated lifetime hours of use (from lamp package information/manufacturers’ data) • Annual hours of use (from prior HOU analyses) Average unit life in years is the ratio of lifetime hours of use to annual hours of use. We will explore alternative sources of assumptions for lifetime hours of use, or for average unit life. 22 DNV GL, 2014a. DNV GL - Energy – www.dnvgl.com/energy We will combine these elements by dividing the number of lamps in sockets by the average lifetime to estimate the number of lamp replacements occurring each year. While some of the replacements will come from in-home storage, the steady-state assumption is that whatever is withdrawn from storage will be replaced into storage, so that the number of replacements required is equal to the number of purchases. This assumption is not strictly true, but is a reasonable high-level assumption. Summing the annual replacements over technologies within a replacement lamp category gives the total market sales for that category. Supplier interview responses As discussed in more detail earlier in the plan, we will also obtain information from lamp suppliers that will allow us to generate a second estimate of total market size. The main approach will involve aggregating each supplier’s estimates of his or her company’s sales (both program and non-program sales as well as efficient and inefficient sales) to come up with an estimate of total market volume for lamps in each replacement lamp category. Consolidated estimate We will assess the quality of supplier interview responses and the uncertainties associated with the stock turnover analysis and develop a best estimate of overall market volume. 2.3.3.3 Net savings calculation Combining markets shares, market volume, and UEC yields the total annual energy consumption of each technology for the with-program or without-program scenarios. Summing over technologies yields the total consumption for the replacement lamp category. Subtracting the with-program total from the withoutprogram total yields the total net savings. Table 17 illustrates the calculation for lamps in the A-lamp replacement category assuming a total market volume of 100 million lamps. Table 16. Net savings calculation example for A-lamp replacements (≤ 30 W) Market Share Technology Adjusted UEC Energy Consumption Without Program (A) With Program (B) Savings (A-B) Without Program With Program CFL A-lamp 25% 20% 62 1,547,095,680 1,237,676,544 309,419,136 CFL spiral 40% 36% 43 1,705,772,160 1,535,194,944 170,577,216 LED A-lamp 10% 12% 17 173,354,054 208,024,865 (34,670,811) 5% 9% 8 37,855,382 68,139,688 (30,284,306) 20% 23% 17 346,708,109 398,714,325 (52,006,216) 3,810,785,386 3,447,750,367 EISA inc/hal A-lamp Traditional inc/hal A-lamp Net Savings This calculation provides overall savings, but does not explicitly allocate savings to each of the efficient technologies. To obtain that allocation, we first re-calculate the net savings starting with the savings for each technology relative to the least efficient technology (the traditional incandescent/halogen A-lamp). DNV GL - Energy – www.dnvgl.com/energy Table 18 illustrates this calculation. The UES for each technology is the difference between the incandescent UEC and the technology’s UEC. The resulting net savings is the same as for the initial calculation. DNV GL - Energy – www.dnvgl.com/energy Table 17. Illustration of net savings calculation for A-lamp replacements (≤ 30 W) relative to traditional incandescent/halogen lamps Technology Market Share Without Program With Program Traditional inc/hal A-lamp 25% EISA inc/hal A-lamp Relative to Traditional Incandescent/Halogen UES Total Savings 20% - - 40% 36% 19 (76,958,093) CFL A-lamp 10% 12% 45 89,096,844 Basic spiral CFL 20% 23% 45 133,645,265 5% 9% 54 217,251,003 LED A-lamp Net Savings 363,035,019 The calculation shown above still has negative savings for EISA-compliant lamps, as well as positive savings for the efficient technologies. We will allocate the negative savings from EISA-compliant lamps to the other technologies in proportion to the net incremental number of units of each technology due to the program, as illustrated in Table 19. DNV GL - Energy – www.dnvgl.com/energy Table 18. Illustration of net savings calculation for A-lamp replacements (≤ 30 W) – all positive and negative savings allocated to efficient technologies Market Shares Technology Without With Program Program Traditional inc/hal A-lamp EISA inc/hal A-lamp CFL A-lamp Basic spiral CFL LED A-lamp Net Savings DNV GL - Energy – www.dnvgl.com/energy 25% 40% 10% 20% 5% 20% 36% 12% 23% 9% Relative to Traditional Incandescent/Halogen UES 19 45 45 54 Total Savings (76,958,093) 89,096,844 133,645,265 217,251,003 363,035,019 Number of Incremental Lamps Efficient Tech % of Net Shift Negative Savings from EISA Lamps Reallocated to Efficient Lamps Proportional to Incremental Lamps (5,000,000) (4,000,000) 2,000,000 3,000,000 4,000,000 22% 33% 44% (17,101,798) (25,652,698) (34,203,597) Efficient Technology Net Savings (with EISACompliant Reallocated) 71,995,045 107,992,568 183,047,406 363,035,019 These analyses provide net savings with estimates for each measure group calculated relative to traditional incandescent/halogen lamps and EISA-compliant incandescent/halogen lamps only. Since LED lamps may displace CFLs as well as both traditional and EISA-compliant incandescent/halogen lamps, we will calculate LED lamp savings relative to a baseline that includes a proportion of CFLs. This calculation involves multiplying the net incremental LED lamps by the UES determined with a mixed baseline of CFLs, traditional incandescent/halogen lamps, and EISA-compliant incandescent/halogen lamps. While this calculation yields a reasonable representation of net savings for LED lamps, the total net savings across all efficient technologies is still the total indicated in Table 17, Table 18, and Table 19. Reducing the net savings allocated to LED lamps means allocating more of the total net savings to CFLs. The savings allocated to LED lamps in Table 19 is the number of incremental LED lamps due to the program times a UES based on a baseline that is the weighted average of traditional and EISA-compliant incandescent/halogen lamp UECs, with weights proportional to the total of each lamp displaced by the program. Implicitly, this assumes that each LED lamp displaces traditional and EISA-compliant incandescent/halogen lamps in these same proportions. Calculating LED lamp savings instead using a UES that includes some CFLs in the baseline implicitly says that each LED lamp displaces some CFLs, and fewer traditional and EISA-compliant incandescent/halogen lamps. If the LED lamps are displacing some CFLs, but the net incremental CFLs is as indicated in Table 19, there must be additional CFLs attributable to the program that are balanced out by those displaced by LED lamps, and these additional CFLs displace the traditional and EISA-compliant incandescent/halogen lamps that were previously attributed to the LED lamps. This is the source of the savings re-allocated to CFLs when the mixed baseline of CFLs, traditional incandescent/halogen lamps, and EISA-compliant incandescent/halogen lamps is used for LED lamps. Table 20 illustrates this reallocation. In this case, we assume that the mix of lamps displaced by LED lamps includes 15% CFL A-lamps, 25% basic spiral CFLs, traditional/EISA-compliant incandescent and halogen lamps in the same proportions as indicated in Table 19. The first two columns of Table 20 show the savings relative to a blended traditional/EISA-compliant incandescent and halogen baseline. Relative to this baseline, savings for traditional and EISA-compliant incandescent and halogen lamps are 0. Savings for the efficient technologies are the same as those in Table 19. The next block of the table shows the LED lamp savings recalculated relative to the mixed baseline of CFLs, traditional incandescent/halogen lamps, and EISAcompliant incandescent/halogen lamps with the remainder of the original LED lamp savings re-allocated to the CFLs. The final block in Table 20 shows the total net savings for each technology with this re-allocation and the equivalent incremental lamps implied. DNV GL - Energy – www.dnvgl.com/energy Table 19. Illustration of net savings calculation for A-lamp replacements (≤ 30 W) using different baseline technologies Savings Relative to Traditional/EISA Mix Savings Relative to Mixed LED Baseline UES UES Technology Total Total Savings with Different Baseline Savings Reallocated from LED Lamps to CFLs Baseline UES Savings Total Equivalent Incremental Lamps Traditional inc/hal A-lamp (9) 42,754,496 Traditional inc/hal EISA inc/hal (9) 42,754,496 (5,000,000) EISA inc/hal A-lamp 11 42,754,496) Traditional inc/hal EISA inc/hal 11 (42,754,496) (4,000,000) CFL A-lamp 36 71,995,045 21,598,514 Traditional inc/hal EISA inc/hal 36 93,593,559 2,600,000 Basic spiral CFL 36 107,992,568 35,997,523 Traditional inc/hal EISA inc/hal 36 143,990,090 4,000,000 LED A-lamp 46 183,047,406 (57,596,036) Traditional inc/hal EISA inc/hal CFL 31.4 125,451,370 4,000,000 363,035,019 1,600,000 Net Savings DNV GL - Energy – www.dnvgl.com/energy 363,035,019 31.4 125,451,370 - The adjustment illustrated in Table 20 assumes that the baseline for CFLs is based on the net shift in nonefficient lamps, while the baseline for LED lamps is based on the replacement rates obtained from the consumer surveys. We will perform a similar set of adjustments to provide net savings for each of the technologies relative to baselines defined by the replacement rate mix, Adjustment all efficient technologies in this way will require some iteration to reconcile to the total net savings. 2.3.4 NTG analyses In prior evaluations, we used the LCM and the supplier interviews to estimate NTG ratios for each measure group and multiplied the NTG ratio for each measure group by gross savings to estimate net savings. In this evaluation, we start from market-level net savings using the approach described in Section 2.3.3., then determine a NTG ratio for each measure group and IOU by dividing the net savings for each measure group and IOU by the respective gross savings. This approach will not yield channel-specific NTG ratios; however, the LCM will produce channel-specific program and non-program market shares, which will provide valuable insights regarding the impacts of the program at the channel level. DNV GL - Energy – www.dnvgl.com/energy 3 REFERENCES CEC, 2014. Voluntary California Quality Light-Emitting Diode (LED) Lamp Specification: A Voluntary Minimum Specification for “California Quality” LED Lamps. CEC-400-2015-001. December, 2014. CPUC ED, 2016. 2013-2016 Energy Division & Program Administrator Energy Efficiency Evaluation, Measurement and Verification Plan - Version 6. January 14, 2016. DNV GL, 2016. Impact Evaluation of 2013-14 Upstream and Residential Downstream Lighting Programs. Prepared for the CPUC ED. April 1, 2016. _____, 2014a. Final Report - WO21: Residential On-site Study: California Lighting and Appliance Saturation Study (CLASS 2012). Prepared for the CPUC ED. November 24, 2014. _____, 2014b. California Upstream and Residential Lighting Impact Evaluation Work Order 28 (WO28) Final Report. Prepared for the CPUC ED. August 4, 2014. KEMA, Inc., 2010. Final Evaluation Report: Upstream Lighting Program, Volumes 1 and 2. Supported by The Cadmus Group, Inc.; Itron, Inc.; PA Consulting Group; and Jai J. Mitchell Analytics. Prepared for the CPUC ED. February 8, 2010. Itron, Inc., 2014. California Commercial Market Share Tracking Study. Prepared for the CPUC ED. November 13, 2014. DNV GL - Energy – www.dnvgl.com/energy ABOUT DNV GL Driven by our purpose of safeguarding life, property and the environment, DNV GL enables organizations to advance the safety and sustainability of their business. We provide classification and technical assurance along with software and independent expert advisory services to the maritime, oil and gas, and energy industries. We also provide certification services to customers across a wide range of industries. Operating in more than 100 countries, our 16,000 professionals are dedicated to helping our customers make the world safer, smarter greener. DNV GL - Energy – www.dnvgl.com/energy