High temperatures drive up peak electricity demand which drives up investment in costly power lines and power plants. Looking ahead, it will be important to better understand the relationship between hot weather and energy demand as system operators gear up for climate change.
The relationship between heat and electricity demand can also help us look back in time to assess some demand-side investments we’ve already made. A new working paper uses this relationship to estimate energy savings from California’s building energy codes.
If you’re a regular reader of this blog, you might be having a déjà vu moment. Haven’t we tackled the “Do California building codes actually save energy?” question before? Do we really need to go down this path again? The answer is yes and yes. This paper shows how ‘big data’ can shed new light on this old (but still relevant) question.
Do Building Energy Codes Save Energy?
If you are looking for a landmark energy efficiency program to evaluate, California’s Title 24 is a good place to start. It boldly went where no building code had gone before, and it has provided an energy efficiency model/inspiration for the nation and the world.
The original Title 24 code (circa 1977) for new buildings was designed to reduce the energy required to heat in the winter and cool in the summer. Given the size of the projected energy savings (significant for electricity, even larger for natural gas), we might expect to see an impact on energy consumption when we look at energy use today in homes built just before and just after these codes took effect. But annual electricity consumption at California homes built in the early 1980s is slightly higher, on average, as compared to homes built just before Title 24 was implemented.
Of course, there are all sorts of problems with this simple before/after comparison. New homes changed in many ways between the 1970s and 1980s (e.g. size and location, air conditioning penetration, the popularity of shag carpet). And some pre-1977 homes have presumably been updated.
Constructing a what-would-have-happened-without-Title-24 benchmark that accounts for all the other time-varying factors that determine energy consumption is not easy. For one, it’s hard, if not impossible, to find homes that are truly identical but for the Title 24 “treatment”. Second, even if you manage to find comparable houses, it can be hard to detect the effect of building codes amidst noisy electricity consumption.
The signal in the noise
One way to cut down on the electricity consumption noise is to focus on the energy uses that were targeted by these building codes: heating in the winter and cooling in the summer. If Title 24 is delivering real energy savings, we should see energy consumption responding differently to changes in outdoor temperatures.
Howard Chong was the first to use this clever strategy using electricity data from Riverside, California. He found that newer homes subject to more stringent building codes use more electricity in hot weather. However, he has fairly limited information about these homes and the people who live in them. So his estimates could be picking up the effects of other house characteristics that are changing over time.
Cue Arik Levinson and his provocative 2016 paper, which brings more detailed data from over 14,000 California households to this building code question. He looks at how monthly energy consumption responds to temperature after removing the effects of neighborhood characteristics (e.g. average income, education) and home characteristics (e.g. house size, air conditioning, number of rooms). The graph below shows his estimates of weather sensitivity of electricity use for houses built in different eras (relative to a pre-1940 home).
The markers summarize how electricity consumption increases per 10 cooling degree days (CDD). Pre-1940s houses are the baseline category.
Like Howard, Arik estimates that electricity use in homes constructed after the Title 24 code increases faster when temperatures rise as compared to the pre-code homes. In other words, he finds no evidence that Title 24 building codes reduced electricity consumption.
You might have thought Arik’s important paper would be the last word on Title 24 energy savings. But an econometrician’s work is never done, especially as data quality continues to improve. When our UC Davis colleagues Aaron Smith, Kevin Novan, and Tianxia Zhou got their hands on hourly electricity smart meter data from homes in Sacramento, they were eager to take another look at this temperature response/home vintage relationship. They released their working paper last week (presented at our recent POWER conference).
Moving from monthly billing data to hourly smart meter data is like increasing the magnification power on your statistical microscope. You can see patterns in the data you could not see before. With high-frequency smart meter data, these guys can flexibly estimate the relationship between temperature and electricity consumption for each individual home and look much more precisely at how these relationships vary with home age and vintage.
The graph below helps to illustrate the value added by higher resolution data. The black squares estimate the electricity response to an increase in summer temperatures relative to a 1977 home using the SMUD smart grid data. To construct comparable estimates from prior work, the authors go back to Arik’s monthly data, isolate homes that can be exactly matched to the construction year, and re-estimate impacts using monthly data from across California (in blue). The bars measure the precision (or imprecision) of these estimates.
One clear takeaway is that high-frequency smart meter data can allow us to see an effect that could not be detected with ye olde fashioned monthly data. Estimates using smart meter data are much more precise.
A second takeaway is that past failure to detect an effect of these building codes on electricity consumption could be chalked up to data limitations (e.g. noisy data, measurement error). In other words, it could be that these electricity savings have been there all along, we just couldn’t see them.
Estimated impacts on energy consumption may look tiny in the graph…but they add up! The authors estimate that a house built just after 1978 uses 8-13% less electricity for cooling than a similar house built just before 1978. The good news is that these are significant reductions, particularly when you consider that cooling drives peak consumption in California. The not so good news is that these reductions fall far short of engineering estimates (which project ~40% cooling reduction for a Sacramento house).
The authors argue that their estimated savings are in the right engineering ballpark IF we assume 42 percent of new homes would have reached Title 24 standards and another 27 percent would have had added some insulation even if Title 24 had not been imposed. The steps the authors must take to square their estimates with savings projections highlight the nuanced relationship between engineering estimates and real-world policy impacts. Engineering estimates evaluate differences in projected energy use between buildings with and without particular efficiency measures. Policy advocates often take those estimates and present them as the energy savings that would be realized with a policy mandating these measures. This is not the right interpretation if some of the targeted measures would be adopted without the policy and/or if the policy cannot be perfectly enforced.
How many economists does it take to measure energy savings?
California’s energy efficiency policies serve as a model for other jurisdictions so it’s really important to separate the success stories from the not-what-we’d-hoped. This new evidence suggests that California’s original Title 24 building codes delivered more savings than prior research had found (although not as much as the program architects had originally envisioned).
Is this the last stop on the do-building-codes-save-energy line of inquiry? I hope not. More than half of U.S. households are now equipped with smart meters monitoring electricity consumption as temperatures rise and fall. As these new data accumulate, there’s room for more research that plumbs the depths of this important question.