Better Data, New Conclusions

(This is a response to Low-Income Reality Check by Trevor R. Roycroft.)

Experience with time-of-use pricing programs shows that a large majority of low-income customers will benefit from dynamic prices. In fact, not making such prices available to these customers might be harmful. In the most efficient system, all customers will face the same prices—and policy makers can provide direct relief to ease the burden for low-income customers.

The impact of dynamic pricing on low-income consumers of electricity is of great import in regulatory policy. In our article, we used empirical information from several jurisdictions to arrive at some basic conclusions on A) whether or not low-income customers were likely to respond to dynamic pricing and B) even if they didn’t respond, would low-income consumers be made worse off by dynamic pricing.

We found that low-income consumers did respond to dynamic pricing, but in most cases their rate of price responsiveness was lower than that for non-low-income consumers. We also found that at least for one utility service area, a majority of low-income consumers would be better off if they were moved from flat rate pricing to dynamic pricing. In that particular service area, we had information from a load research sample that included low-income customers. The low-income customers were defined as those whose incomes were a certain multiple of the federal poverty guidelines.

In his article, Dr. Roycroft seemingly reaches conclusions that contradict our findings. But he doesn’t base his conclusions on new data, just on reinterpreting our analysis.

Data Integrity

The Roycroft article alleges that we changed our findings across different studies. But one must change one’s position as new data become available. In preparing the report for the Institute for Electric Efficiency (IEE) on which our Fortnightly article is based, we did some additional work on the primary data behind two of the dynamic pricing experiments.

Specifically, the BGE and CL&P pilots didn’t have designated low-income test cells in their design, because the primary purpose of those pilots was to determine the demand response impacts for the average customer, and not low-income customers specifically. For that reason, we had to estimate new regressions models to estimate the low-income consumer impacts separately from the average customer impacts reported earlier in the BGE and CL&P pilot reports. However, since information on income status wasn’t available on every customer, we re-estimated the regression models on the limited sample of customers who had reported their income status. That limitation is acknowledged in our Fortnightly article and in the IEE whitepaper.

But a bigger issue that Roycroft identifies is that we don’t use a standardized definition of low income in our various studies. Although that’s true, it’s because the studies themselves don’t use a standardized definition of low income. It wouldn’t be possible to apply a single definition across the studies, without going back to the customers and resurveying their income status—a step that wasn’t feasible for the purposes of our report.

As a practical matter, there’s no single definition of low income across the nation. Our analysis used the definitions that were used in each individual study.

Take the case of California’s statewide pricing pilot (SPP) where one of the definitions of “low income” is customers who are on the California Alternate Rates for Energy (CARE) rate, and another definition is customers earning less than $40,000. People on the CARE rate get a subsidized electricity rate which, given the tiered structure of rates in California, can range from a low of 20 percent to a high of 72 percent. To be eligible for the CARE rate, consumers simply have to declare that they meet its low-income guidelines. For households of one or two individuals, the income guideline is $31,300.

Not all low-income consumers who meet these income criteria are on the CARE rate; and some consumers who don’t meet the guidelines are probably on the CARE rate. CARE customers get a very hefty price discount that renders them insensitive to price fluctuations, and just relying on their price responsiveness would be misleading. And so we report both sets of price responsiveness estimates.

In Figure 1 of his article, Roycroft shows that low-income customers were found to be much less price-responsive than were other customers in the SPP. Our results show that when compared to average customers in that pilot, CARE customers exhibit a reduction of 3 percent in critical peak demand versus a reduction of 13 percent for the average customer. But when “low-income” customers (defined as income below $40,000) are compared to average customers, the former have a lower but similar response as the latter (11 percent vs. 13 percent).

For the PG&E Smart Rate program, the results cited in our report show that CARE customers reduced critical peak demand in the year 2008 by 11 percent vs. 17 percent for average customers; in 2009, the corresponding numbers were 8 percent and 15 percent.

These results are consistent with the results in Roycroft’s Figure 1, as are the Pepco results that we report. So the discrepancies aren’t as great as the Roycroft narrative might suggest.

Standing by the Study

While we appreciate Roycroft’s effort in bringing out once again the importance of the issue, we aren’t convinced that he has presented any new evidence that contradicts the conclusions in our article. We stand by them.

The debate on whether dynamic pricing will make low-income customers better off or worse off will continue. Our position remains that all customers should face the same prices and if, for social reasons, the government wants to protect the economic well-being of some subset of low-income households, it might consider offering them “energy stamps” similar to the kind of assistance that is offered through a food stamps type of program. (See Ahmad Faruqui, “Residential dynamic pricing and ‘energy stamps,’” Regulation, Winter 2010-2011, Volume 33, No. 4, pp. 4-5.) However, we are convinced that a large majority of low-income customers will benefit from dynamic prices and that not making such prices available to these customers may in fact be harmful.