Research teaser: CNZ x Cramton Cramton Associates x University of Cologne x University of Maryland
Centre for Net Zero • 23 November 2021
Centre for Net Zero • 23 November 2021
When asked to picture the energy industry, many of us might think of a series of giant pylons and tangled cables. The reality is that energy’s infrastructure is changing rapidly, with much less reliance on a single, centralised system.
People — and their low carbon technologies — are already becoming increasingly important actors in a low carbon future. Why? Because they can provide the grid with the flexibility it needs to deliver reliable, clean energy. We know that the increasing electrification of heat and transport will push peak demand up (which usually occurs between 4pm-7pm), whilst renewable energy remains inherently intermittent; we can’t guarantee sunshine or windy weather.
If supply lacks flexibility, what about demand? The concept of ‘price responsive demand’ is gaining mainstream attention, whereby customers respond to a price signal to alter, or shift, the amount of energy they draw from the grid, at a specific time. Dynamic tariffs are a key demand response tool — which expose customers to wholesale market rates — and financially reward them for shifting consumption from periods where energy is scarce to periods where it is abundant. In this way, households are encouraged to turn down or shift when they use energy-hungry appliances via these price signals, making them a critical gateway for households to participate in electricity markets.
While the data shows that households are generally responsive to dynamic pricing, it is more challenging to define this exact relationship between price and consumption, because numerous factors influence how much energy we use in the home aside from price.
Over the past few months, Centre for Net Zero has collaborated with eminent American economist Professor Peter Cramton and his market design research team at the University of Cologne and the University of Maryland to explore the price elasticity of demand of UK households. We asked ourselves: what impact does a change in the half-hourly electricity price make on aggregate energy consumption? Where a price shock is experienced by a household (e.g. peak time), what does it mean for adjacent periods of the day, when electricity prices are cheaper again?
We were interested in comparative household behaviour across two types of pricing contracts: Fixed and Dynamic (wholesale-linked). The Octopus Energy dynamic plan reflects day-ahead auction prices for electricity, though a cap at 35p/kWh protects customers from surge pricing. As negative wholesale prices can occur, customers can also receive money during ‘plunge’ pricing.
To support this research, we compiled an anonymised sample of a year of half-hourly electricity smart meter readings from ~15,000 UK households. Each household produces up to 48 readings every day (one per half-hour: 09:00, 09:30, etc). We also compiled supplementary data around factors such as weather conditions, energy performance and low carbon technology (LCT) ownership (e.g. ownership of a battery or an EV).
Complexities in these datasets include the fact that households are able to change houses, LCTs and tariffs across our sample period (for example, meaning they could belong to multiple tariffs over the year), and smart meter readings are inherently subject to missingness — due to complex smart metering communications infrastructure across households in the UK.
To establish the price responsiveness of households, we chose to use a fixed effects model to evaluate the impact of price signals against the electricity consumption of households on dynamic tariffs, while controlling for periodicity and trends.
A key factor necessary to control for is seasonality — i.e. electricity usage systematically varies by half-hour of day, day of week and month of year. By accounting for the mean consumption of households at a given ‘time segment’ (e.g. 09:00 on a Monday in December), every consumption delta reading is controlled for its characteristic or ‘typical’ value.
Given that other factors like weather can further influence consumption, the model also includes a term to account for the energy consumption of households on fixed tariffs, which can be thought of as a control or ‘baseline’ consumption in a given weather environment. In this way, the analysis isolates the part of household behaviour which is the response to price signals, and not time, weather, or other factors.
Households modify their electricity consumption in response to price, and are particularly motivated/able to do so when they own certain LCTs
As expected, the findings confirm that households do exhibit price elasticity: as prices increase, demand decreases. A one-percent increase in price results in a 0.26% decrease in demand. Price elasticity is substantially influenced by various LCT ownership, which appear to facilitate households in responding to price — either through their provision of stored energy or, in the case of EVs, because they need lots of energy to charge — thereby motivating people to do so when electricity is cheaper. We are cognisant of the fact that households on dynamic tariffs have self-selected to pay flexible prices, and therefore their behaviour will likely reflect a willingness to take part in price movements.
Households may modestly redistribute their demand to cheaper periods of the day
The analysis also explored how households might be redistributing their consumption across other parts of the day, either side of a price shock, when prices are cheaper. The results suggest that customers may modestly increase their demand in adjacent periods to a price shock, though these results are generally not statistically significant.
Why might households alter demand adjacent to a price shock? One interpretation could be that households sometimes use more energy than is strictly required to meet their comfort needs, and price signals are helpful as a reminder to forgo such energy usage in that general part of the day. Another might be that a proportion of households directly optimise their consumption to save money — even though such optimisation is a manual process today, with the vast majority of home appliances lacking an ability to optimise for dynamic rates. Potential savings may be particularly significant for low-income households, who spend a larger proportion of income on electricity.
Household responsiveness varies by time-of-day and season
The findings also indicate that price elasticity may vary by time-of-day and season. One rationale for this might be that households find it harder to adjust their electricity usage during working hours than in the evenings or at night, and that households have more ‘headroom’ to change consumption during more extreme weather seasons (when electrical heating/cooling may be in use).
We’re looking forward to publishing our findings in detail shortly — so keep your eyes peeled for updates on our CNZ channels.