D-W statistic *Statistically significant (p <.05) 1.89 2.14 For the treatment period 1986 through 1999, when the high quality product is the electronic ballasts and the low quality product is the high power factor magnetic ballast, the model indicates that the demand for the higher quality product was highly elastic. In this period, a one percent decrease in price resulted in a 4 percent increase in demand. Since this price elasticity estimate is the combined result of market effects and public program effects, and since the comparison period unit elasticity is an estimate of the impact of market effects only, the public program effect can be viewed as approximately four times greater than the market effect. In other words, the results of the comparison period model suggest that about one-quarter of the estimated price elasticity in the treatment period is due to the product cycle effect, and about three-quarters of the elasticity estimate in the treatment period is the result of heightened price sensitivity due to the public program effect. Note that a t-test confirms that there is a statistically significant difference in the price elasticities between the two periods. In interpreting this finding it is important to understand that if the public interventions consisted of a single program in the treatment period, these results could represent the accomplishments of this program exclusively. However, since the treatment period consisted of many different programs directed at the electronic ballasts market from the Green Lights Partnership to utility DSM programs and federal and state-run programs-the results of this analysis must be interpreted as the effect of all public programs combined on the transformation of the targeted market. With respect to the exogenous variables, the model indicates that in the comparison period marginal percentage increases in electricity prices and the prime lending rate are associated with small percentage decreases in demand. However, a marginal percentage increase in the rate of change of the consumer price index -- an indicator of economy-wide price level or inflation expectations -- is associated with a small percentage increase in demand. In the treatment period, the model suggests that a marginal percentage increase in electricity prices is associated with a large percentage increase in demand, and a marginal percentage increase in the rate of change of the consumer price index is associated with a small percentage increase in demand. Finally, a marginal increase in the prime lending rate is associated with a small decrease in demand. 342 Share Capture Model The semi-log functional form is appropriate for the share capture model when it is expected, holding all else constant, that the impact of relative price on the market share of the high quality product increases as the relative price decreases. For this model, the independent variables remain in natural logarithm form and the dependent variable, S, is defined as the annual market share, in percent, of the higher quality product. As in the market demand model, relative or normalized price is taken to be endogenous, leading to the two-stage least squares model: where it is assumed, as above, that the error process is first-order autoregressive, and the terms v, and e, are not autocorrelated. The model variables are as defined above and the interpretation of the final model coefficients are modified such that: coefficient representing the absolute change in the market share of product x due to a (relative) percentage change in the relative price of product x. coefficient representing the absolute change in the market share of product x due to a (relative) percentage change in the national average annual retail cents per kWh in the commercial sector coefficient representing the absolute change in the market share of product x due to a (relative) percentage change in the prime lending rate coefficient representing the absolute change in the market share of product x due to a (relative) percentage change in the change in the consumer price index Exhibit 3 contains the findings for the share capture model for the comparison and treatment periods. The findings of the share capture model in the comparison period indicate that, at the margin, a ten percent decrease in the relative price of the high quality ballasts leads to an (absolute) increase in market share of about 1.4 percent. In the treatment period, a ten percent decrease in relative price leads to a 5.3 percent increase in the market share of the high quality product. The difference between these estimates is statistically significant. As with the market demand model, the findings indicate that impact of the market effect appears to be about onequarter the size of the public program effect. Again, in interpreting these findings it is important to understand that the public program effect represents the impacts of all public programs combined. The share capture model findings also indicate that in the comparison and treatment periods a marginal percentage increase in electricity prices is associated with a decrease in market share. However, a marginal percentage increase in the prime lending rate is associated, in each study period, with an increase in market share. At the margin, a percentage change in the rate of change of the consumer price index is associated with an increase in market share in the comparison period and a decrease in market share in the treatment period. 4. Green Lights® Partnership Accomplishments Quantifying the climate protection impacts of the Green Lights Partnership requires merging the comparison and treatment period model findings, and then combining them with historical data on DSM electronic ballasts rebates. Once this is done, various engineering algorithms and accompanying inputs are enlisted to convert electronic ballasts shipments to energy savings and carbon equivalents. As noted earlier, throughout this evaluation all national fluorescent ballasts shipments are assumed to be sold to end users and installed within the United States in the same calendar year in which they were reported to be shipped. In estimating the Green Lights Partnership impacts, the market demand and share capture model findings are used to separate the total number of shipped electronic ballasts in the treatment period into those that are due to product cycle or market effects and then those that are attributable to the public program effect. Secondly, all of the units attributable to the public program effect are separated into those whose purchases were connected to DSM rebates versus those whose purchases were independent of rebates. For this evaluation, the former category of purchases signifies a temporary change in the ballasts market which, as some sales evidence already suggests, may disappear once rebates are no longer available. On the other hand, the latter category of purchases signifies permanent change, or market transformation, due to public programs. The penultimate analysis for completing this impact evaluation requires that the effects of substituting electronic for magnetic ballasts be converted into annual energy savings at the site or end user level. Finally, the Green Lights Partnership climate protection accomplishments are derived by converting annual site energy savings into source energy savings and national carbon equivalents. 4.1 Falling-Price and Public Program Effects From the market demand and share capture model findings in the comparison and treatment periods, different estimates of public program impacts can be generated. First, the ratios of the coefficients of the price variable are calculated. This ratio is interpretable as the percentage of the price effect in the treatment period that is due to market effects. Multiplying these ratio by the number of shipped electronic ballasts units, Units, yields the falling-price effect. This is the number of units that would have been shipped, in the treatment period, in the absence of public programs. The equations for calculating the falling-price effect over the 1986 to 1999 treatment period are: This calculation is performed for each year in the treatment period. Once the electronic fluorescent ballasts units resulting from the falling-price effect are determined, the remaining units of shipped electronic ballasts can be attributed to the public program effects as: Exhibit 4 contains, by year, the total number of units shipped of electronic fluorescent ballasts in the treatment period and the fraction of the total shipments that are attributable to the fallingprice and public program effects. The 14 years of shipments data indicate that manufacturers shipped a total of over 258 million units of electronic ballasts from 1986 through 1999. Of this amount, the market demand models indicates that approximately 24 percent of the electronic ballasts would have been shipped due to the effects of decreasing relative prices over this time period. Approximately 76 percent of the remaining total shipments are attributable to the influences of all public programs from 1986 through 1999. The share capture model findings agree with these estimates within a few percentage points. Exhibit 4: Annual Falling-Price and Public Program Effects The number of units of electronic ballasts that are attributable to the Green Lights Partnership and the market-transformative effects of other public programs is derived from the electronic ballast shipments that the statistical models attribute to all public programs. The market transformation share of this amount is acquired by subtracting from the total public program effect the number of electronic ballast units that were given rebates through electric utility DSM programs. The number of units given rebates between 1991 and 1999 is taken from a study of DSM program activity from 1992 through 1997 (LBNL 2000). Although two different sets of estimates are provided in this study, it is likely that the most reliable set is derived from the sample that consists of six electric utilities. Together, they accounted for about 17 percent of national utility spending on energy efficiency over the study period. The second set of estimates comes from a sample of three additional utilities whose rebate data were not as complete as the data for the other six utilities. The rebate estimates from this sample are significantly lower than those of the first sample. No estimates are available of the number of electronic fluorescent ballasts given rebates in 1991, 1998 and 1999. For the purposes of this study, these estimates are constructed by applying the percentage change in national DSM program expenditures for these years to the adjacent year's published rebate estimates. Also, as no rebate data are available for 1986 through 1990, the public program effect for these years is taken to be the result of DSM rebate programs exclusively. This is a reasonable attribution given that most programs designed to promote market transformation in the national electronic fluorescent ballasts market, such as the Green Lights Partnership, were not launched until the 1990s. Exhibit 5 contains the estimates of the impact of the Green Lights Partnership and other public programs on the transformation of the ballasts market, derived by subtracting the estimated number of rebated electronic ballasts from the public program effect. |