Skip to content

Do Energy Efficiency Investments Deliver at the Right Time?

Energy-efficient air conditioners save electricity during the most valuable hours of the year.

(Today’s post is co-authored with Judson Boomhower, who recently received his Ph.D. at Berkeley where he was a graduate student researcher at the Energy Institute and is now a post-doc at Stanford).

Along with everyone else in Berkeley, we’ve enjoyed watching the home-team Golden State Warriors pull out comeback after miraculous comeback on their way to the NBA Finals. Has anyone else watched Steph Curry and Klay Thompson catch fire at just the right time and thought, “this team could really teach us something about energy efficiency policy?”

Photos (1 and 2) by Keith Allison, Creative Commons License BY-SA 2.0

Crunch time in electricity markets are those few highest demand hours each year when generation is operating at full capacity. During these ultra-peak hours there is little ability to further increase supply so demand reductions are extremely valuable.

This feature of electricity markets is well known, yet most analyses of energy-efficiency policies completely ignore timing. For example, when the Department of Energy considers new energy-efficiency standards, they focus on total energy savings without regard to when these savings occur. With a few notable exceptions, mostly from here in California, there is surprisingly little attention both by policymakers and in the academic literature to how the value of energy-efficiency varies over time.

We take on this issue in a new Energy Institute working paper, available here. Our evidence comes from Southern California Edison’s residential air conditioner program. We use anonymized hourly smart-meter data from 9,700 rebate recipients to estimate how electricity savings vary across months-of-the-year and hours-of-the-day. As the figure below shows, electricity savings tend to occur between June and September, and between about 3pm and 9pm.

Electricity Savings

As a side note to duck chart aficionados, this savings profile differs somewhat from engineering models, which predict more savings earlier in the afternoon and in non-summer months. As more solar generation comes online, there is growing concern about meeting the steep evening ramp. Our estimates suggest that air conditioning investments deliver more savings than expected during these evening hours, and thus could become more valuable as renewables penetration increases.

These savings are highly correlated with the value of electricity. The figure below shows the value of electricity by hour-of-day in California for February and August, in dollars per megawatt-hour. We include wholesale electricity prices and the “resource adequacy” payments that generators receive to make sure they will be available when demand is high. The different data series in each panel show different methods for allocating resource adequacy contract prices to high load hours. For example, with “Top Hour” we assign the entire capacity value to the highest load hour-of-day in each month.

Wholesale Electricity Prices and Capacity ValuesCAISOfeb.PNGCAISOaug.PNG

Regardless of exactly how we allocate resource adequacy payments, these figures make clear that summer afternoons are crunch time in California electricity markets. Unlike natural gas, electricity cannot be cost-effectively stored even for short periods so during these ultra-peak periods there is nothing preventing wholesale prices and capacity values from rising sky high.

And this is exactly when air conditioning investments yield their largest electricity savings. Efficient air conditioners don’t save electricity in the middle of the night or during the winter, but electricity is less valuable at these times anyway. Overall, we estimate that accounting for timing increases the value of air conditioner investments by 50% relative to a naive calculation that ignores timing.

How does this compare to other energy-efficiency investments?  So glad you asked. We next brought in engineering-based savings profiles from the E3 calculator for a whole variety of energy-efficiency investments and calculated the timing premium for California and for several other major U.S. markets. The table below shows the results.

Timing Premiums for Energy-Efficiency Investments


Overall, there is a remarkably wide range of value across investments. Residential air conditioning has a 35%+ average premium across markets. The premium is similar whether we use our econometric estimates (first row), or the engineering estimates (second row), reflecting the fact that, despite some interesting differences, both sets of estimates indicate large savings during high-value summer peak hours.

Other investments also gain value when timing is considered. Non-residential heating and cooling investments enjoy a 20-30% timing premium, reflecting the relatively high value of electricity during the day when these investments yield savings. This is particularly true in CAISO and ERCOT, but also true in NYISO.

Refrigerators and freezer investments have the lowest timing premium. This makes sense because savings from these investments are only weakly correlated with system load. Lighting also does surprisingly poorly, reflecting that LEDs save electricity mostly during the winter and at night, when electricity tends to be less valuable.

We hope our paper will help move the energy efficiency discussion away from total savings and toward total value. To do this will require more rigorous ex post analyses of energy savings based on real market data. It will also require integrating these savings estimates with prices from wholesale and capacity markets, rebalancing the energy efficiency portfolio toward investments that save energy in more valuable hours.

Of course, these premiums are not everything.  In evaluating energy-efficiency policies it is still important to evaluate all the costs and benefits. The numbers above don’t say anything about how much these different types of programs cost, or about how large ex post savings are relative to ex ante estimates, or about how many participants are inframarginal (i.e., “free-riders” in the energy efficiency literature). We’ve discussed these issues in previous blog posts here, here, and here. But our paper makes a strong case that, when calculating benefits, it is important to account for timing.

More generally, our paper highlights the power of smart-meter data. The econometric analysis we performed for residential air conditioning would have been impossible just a few years ago, but today more than 40% of U.S. residential electricity customers have smart meters, up from less than 2% in 2007. We are just scratching the surface of what is now possible using this flood of new data and its potential to facilitate smarter, more evidence-based energy-efficiency policies that are better integrated with market priorities.

For more see, “Do Energy Efficiency Investments Deliver at the Right Time?” American Economic Journal: Applied Economics, 2020, 12(1), 115-139.

Keep up with Energy Institute blogs, research, and events on Twitter @energyathaas.

Suggested citation: Davis, Lucas. “Do Energy Efficiency Investments Deliver at the Right Time?” Energy Institute Blog, UC Berkeley, June 6, 2016,

Lucas Davis View All

Lucas Davis is the Jeffrey A. Jacobs Distinguished Professor in Business and Technology at the Haas School of Business at the University of California, Berkeley. He is a Faculty Affiliate at the Energy Institute at Haas, a coeditor at the American Economic Journal: Economic Policy, and a Research Associate at the National Bureau of Economic Research. He received a BA from Amherst College and a PhD in Economics from the University of Wisconsin. His research focuses on energy and environmental markets, and in particular, on electricity and natural gas regulation, pricing in competitive and non-competitive markets, and the economic and business impacts of environmental policy.

23 thoughts on “Do Energy Efficiency Investments Deliver at the Right Time? Leave a comment

  1. What about Jevon’s paradox? What will stop those with a more efficient air conditioner from devouring the energy savings by dropping the thermostat an extra degree or two during summer?

    • Your observation is correct; energy efficiency causes the effective price of the energy product (e.g., cooling) to drop so the owner uses more of the product (i.e., the “snap-back effect”). However, Jevon’s paradox requires price elasticity to equal or exceed 1.0, which generally doesn’t happen.

      • Jevon’s paradox relates to income elasticity, and very often it is greater than one. For example, using international survey results two decades ago, I found the income elasticity for transportation costs was over 2.0. This is one reason why VMT increased over 60% per capita in the U.S. while fuel economy improved.

  2. Interesting paper – the timing premiums for various EE measures was v. useful. Couple questions:
    – Why is the timing premium for res. lighting appear greater than refrigerators/freezers? Seems counterintuitive.
    – Is EE currently compensated in any capacity markets according to its timing value?

    • I would suspects it is because the refrigerator or freezer runs all night but residential lighting is only on during dawn or night hours when residents are (mostly) at home and nor sleeping. The overnight hours when people sleep tend to have the lowest wholesale market prices. The result is that the average avoided cost of fridges/freezers is lower than the average avoided cost of residential lighting.

  3. To build on Mr. Borlick’s point, occasional peak prices can be so high that customers may also be willing to reduce load temporarily. (Eat dinner outside after 7pm and delay cooling down the house, metaphorically or literally.) But this works much better if customers “see” the peak prices without them being filtered through a time-averaging rate structure.
    Another implication of the paper is that to encourage customers to buy more efficient AC, they should have the option of paying real time prices. That removes most of the wedge between social and private benefits.

      • Many on this blog (including some of the posters) believe that if we can just compel consumers to pay attention to electricity prices more, and those prices reflected spot market activity, that all consumers would magically begin using energy more efficiently and all subsidies would end. That ignores one reality that consumers have only limited attention and may only notice electricity prices in the extreme (which does not necessarily lead to rationale choice), and that the existing institutions of regulation inhibit contractual relationships that could compel efficient choice across a wider spectrum. We should spend less time ranting about the need for retail real time pricing, and more time about allowing entry of more players into the retail market, many of whom are better equipped to translate real time prices into understandable and manageable messages for consumers.

  4. While I have only perused this paper I agree with its conclusions. The value of both generation and demand reduction critically depends on the time of day when it delivers its megawatts or megawatts. Levelized cost comparisons are largely useless when applied to any resources other than caseload generation.

    For example, the value of solar energy production is almost entirely driven by when it is delivered, i.e., during hours of high demand (and therefore high avoided energy costs). On the other hand, this principle works against onshore wind generation, with is largely produced at night when avoided energy costs are low, and at times even negative.

    Regarding the capacity value of energy efficiency, one needs to keep in mind that generation capacity is a artificial product created by imposing low regulatory caps on energy prices. This causes a “missing money” problem for generators which must be made up through residual side payments called “capacity payments.” These payments are not determined by the cost of building peaking plants but rather represent the remainder of what is needed to justify investment in these plants. ERCOT is the only wholesale electricity market in the US that does not provide capacity payments. Instead it relies on very high energy prices during times of supply shortage to eliminate the missing money problem. In this regard, ERCOT is miles ahead of the CAISO – and all of the other ISOs – who are constantly struggling to patch up the flaws in their capacity markets.

    Regardless of the shortcomings in our ISO capacity markets, this paper correctly states that energy efficiency measures that reduce demand during hours of high demand are avoiding costs that exceed the marginal cost of energy production in those hours, except possibly for ERCOT. With this minor knit-pick, the paper makes a valuable contribution.

  5. If this article leads to policies that push for wholesale replacement of existing air conditioners, then policymakers are clearly reading only the parts that fit their agenda. In fact, a) there are cheaper alternatives, and b) it typically makes little economic sense for homeowners to replace air conditioners that function without enormous subsidies. If the CPUC absolutely has to subsidize improvements, they should start with better attic ventilation and whole house fans before engaging in wholesale air conditioning replacements.

  6. Any ‘power engineers’ out there listening.
    Water and charge are analogous; pressure and flow rate are analogs of voltage and current.
    When water availability drops the pressure and flow rate can be reduced with no-minimal impact on users and equipment. When ‘charge available’ drops, voltage-current-frequency might change, but if we design equipment that can still work efficiently and effectively the whole infrastructure that exists to maintain frequency and voltage will then become unnecessary — and ‘some’ economist will have to find new problems.

    • It’s true that the critical short-run constraints in a power grid are to maintain voltage and frequency. These operate on a time scale of seconds. But those are proxies for system stress. The underlying stressor is from power balance equations, and the integral of power, energy. Equipment that can run at reduced voltage does not help with this,i.e. the energy balance equations on a 5 minute time scale won’t be any different. So the fundamental issue discussed here will be exactly the same, especially since the AC timing issue they are talking about is on a time scale of 10 minutes or longer.

  7. Several years ago, I wrote the large majority of a report from a public utility commission of a state to that state’s legislature. This point about the timing of the savings was something that I mentioned as a means to reform the existing EE laws. My reasoning was that the preface of the legislation stated the state’s goals were to reduce the need for future generation. I argued that the current legislation incented lighting and other programs that focused on off-peak usage. That these incentives made the law inconsistent or suboptimal at achieving the intended goals of the legislation. It’s nice to see other economist finally having the data to confirm my theories from back then.

    If I understand your Timing Premiums table, the data reflects engineering estimates from E3 modeling. Given that your SCE data shows that engineering models misestimate the time intervals for AC savings, how much weight should we give the Timing Premiums table?

    • Thanks for the comment. There are some interesting differences between our econometric estimates and the E3 engineering estimates but, overall, the implied timing premium is very similar. The table above shows both; 35% premium for our econometric estimates versus 38% premium based on the E3 estimates.

  8. The more I read about the experiments in generation and distribution occurring in California, the more enamored I become with federalism. Thank goodness that California has the power and authority to perform this experimentation on something so critical to everyday survival. You go California! May all the other states watch and learn by your example.

%d bloggers like this: