The Energy Cost of Covid-19

Energy burden was a major issue even before the pandemic. The last 17 months have only made it worse.

Kaushik S Mohan
Aug 5 · 5 min read

It has now been over a year and a half since the pandemic started shaping our daily lives with many of us working and learning remotely. Given this event, our individual energy consumption has shifted from our schools and workplaces to our homes. Data from the Energy Information Administration (EIA) reflects this change, with residential energy sales increasing by 1.5% in 2020 compared to 2019, while commercial and industrial sales reduced by 6.5% and 8.25% respectively, due to businesses downsizing and, in some cases, shutting down.

Even before the pandemic, a report by ACEEE found that low-income, Black, Hispanic, and Native American households all face dramatically higher energy burdens — spending a greater portion of their income on energy bills — than the average household. High energy burdens are correlated with greater risk for respiratory diseases, increased stress and economic hardship, and difficulty in moving out of poverty, issues that have only been exacerbated by the health and economic effects of the pandemic. Gaining a better understanding of the increase in energy burden this past year can provide insight into some of the indirect economic effects of the pandemic.

Unfortunately, the EIA’s high-level numbers don’t tell the complete story. Energy use can vary from one year to the next due to several factors, such as changes in population or simply weather conditions. A year with a hotter summer or a colder winter would see greater energy usage from increased air-conditioning and heating needs, respectively. Therefore, the year over year (YoY) change in energy use alone cannot help us determine the effect of the changing consumption patterns during the pandemic.

To tease out this effect, we would need to compare the actual energy usage since the start of the pandemic relative to the expected energy use with all other factors staying the same, i.e. what would energy consumption since March 2020 have looked like if consumption behavior from past years had continued. By building a predictive model which takes weather, population, and geography into account, we can use historical energy use data to reasonably project energy use for the last year and half without the effects of the pandemic.

Predicting energy usage

We start building the predictive model for energy consumption by using data on energy sales and number of consumers from the EIA for each state and for each month going back to 2008. To account for changes in population and different number of days for each month, we focus the rest of the analysis on average daily energy use per capita. From the below plot for New York state, we observe the annual seasonal variation: large spikes during the summer due to air conditioning and smaller peaks in the winter corresponding to heating. The summer peak in 2020 is much higher than the last three years, and we hope to understand if this increase is due to a possibly warmer summer in 2020 and from increased residential energy consumption during the pandemic.

Avg. daily residential energy consumption in NY State (2008–2020)

Using data from 2008 to 2019, we built a model that predicts monthly energy consumption factoring in trends over time, seasonal variations, how hot (cold) each summer (winter) was, and the number of weekends and holidays each month. Trends in energy use seemed to vary by each state and hence to account for these differences, we built a separate model for each one. The predictions from the model, along with the observed values between 2008 and 2019 for NY state, are shown below to validate model performance prior to forecasting for 2020. While we have tried to account for several different factors that may contribute to energy consumption, it is important to recognize that the prediction model is limited by the data available and several other factors may remain unaccounted for.

Actual vs. predicted residential energy consumption in NY State (2008–2019)

Estimating the impact of COVID on energy consumption

Using the predictions for each state from the start of 2020, we evaluate the percentage deviation from the actual consumption where a positive deviation indicates an increase in energy use during the pandemic. The below interactive plot shows this net deviation over the last two years with the mean values before and since the start of the pandemic marked by the dashed horizontal lines.

We observe that the predictions start deviating from March 2020, coinciding with the onset of the pandemic and first set of stay-at-home orders. Overall, we notice an average of 3.3% pandemic-related increase in residential energy consumption in the US with a peak of 7.85% coinciding with the most intensive pandemic-related restrictions in May. States in the Northeast region saw the highest average increase at over 7%, followed by states in the Midwest and Mountain regions which recorded 4–6% increases in residential energy use during the pandemic. Southern states show the least increase averaged over the course of the pandemic, although the peak change during this time was nearly 7% in several states. The average and peak COVID-related monthly increase in energy use for all the states is shown on the map below.

Statistics from the Office of Energy Efficiency and Renewable Energy indicate that the lowest third of earners in a state spend nearly 30% of their income on energy. Given the above observations of increasing residential energy consumption since the start of the pandemic, this figure could be as high as an additional 10% of income, even before taking into account the unprecedented unemployment rates and subsequent loss of income faced by this segment of the population during the pandemic.

For more information on the prediction model and the datasets used for this analysis, check out our GitHub repository.

Data Clinic

We help nonprofits have a greater impact through data and…

Data Clinic

As the data- and tech-for-good arm of the financial services company Two Sigma, Data Clinic provides pro bono data science and engineering support to nonprofits and engages in open source tooling and research that contribute to the broader Data and Tech for Good movement.

Kaushik S Mohan

Written by

Data Scientist @ Data Clinic

Data Clinic

As the data- and tech-for-good arm of the financial services company Two Sigma, Data Clinic provides pro bono data science and engineering support to nonprofits and engages in open source tooling and research that contribute to the broader Data and Tech for Good movement.