Posts from "October 2012"


Politics and power: Americans who vote more use 7-10% less electricity

  • By Barry Fischer
  • October 31, 2012

Three out of four voters in this year’s election view energy as a “very important” issue in their evaluation of candidates.  As voters examine the candidates’ positions on energy, we decided to turn the tables — and instead shine the light on the energy profile of the voters themselves…

We wondered if any statistical relationship might exist between voting patterns and energy usage. So we cracked open Opower’s energy data warehouse (spanning 50 million US households), and linked it up with historical election data from the last seven years.

Our analysis uncovered a striking pattern that held resolutely across geographic regions: Americans who vote more also use less electricity.

Let’s dig into the data streams to assess what the voting-energy relationship looks like, and what may be driving it.

Higher voting frequency is associated with lower household electricity usage

Our analysis zeroed in on 137,000 households, whose electricity consumption data we were able to match with publicly available records indicating how many elections the utility account holder has voted in since 2004.

We sourced voting records and energy usage data from two states that exhibit distinct electoral identities – a western state and an eastern state. Then we matched up the voting/energy data streams in a way that fully protects the personally identifiable information of the households.

Between 2004 and 2010, voters in each state had the opportunity to participate in an average of about 10 elections (spanning primary, general, local, and special elections). As you can see below, people who voted in a greater number of those elections also consistently use less electricity.

Note that the data above represents only utility customers who lived in their respective households over the full 2004-10 period, so all of them faced a similar number of opportunities to cast votes. (And if you’re wondering why the eastern state exhibits higher average electricity usage in general — it’s simply because a larger fraction of households there use electricity as a heating source.)

The inverse relationship between voting frequency and home electricity usage becomes even more apparent when we categorize people as “rare voters,” “sometimes voters,” or “frequent voters.”

So what’s the explanation for frequent voters’ persistently lower household electricity usage?

Could it be that the frequent voters’ homes are simply smaller and therefore use less electricity? No: our property-level data indicates that homes in all three voting-frequency groups are similarly clustered around an average size of 1900-2000 square feet.

Or could it be that frequent voters are for some reason more likely to use gas furnaces or other non-electric heating systems, so naturally would consume less electricity per year? It’s not that either: we found that rare voters and frequent voters are equally as likely to own homes with electric heating.

Indeed, we found that the sizable difference in electricity consumption across types of voters is related not so much to the physical characteristics of their homes, as it is to the characteristics of the voters themselves.

We’ve identified two major factors at play. Let’s explore them in turn…

Older Americans: their voter turnout is high, their home electricity consumption is low

Year after year, the most frequent voters in America are the country’s older citizens, as we can see clearly from voter turnout in the 2008 Presidential election (which had the highest overall turnout since 1968), and the 2010 midterm election.

As people get older, they vote more frequently – likely because, as explained in a recent article by Money magazine, older Americans have an acutely personal stake in policy issues (e.g. Social Security and Medicare) and have more time to register to vote, evaluate candidates, and turn out to the polls.

This age/voting relationship is similarly evident in our own analysis. On average, the “frequent” voters in our two-state dataset had an average age of 60 – noticeably higher than the people voting “sometimes” (average age 55) or “infrequently” (average age 52).

(If these average ages seem high in general, it’s because our dataset is intentionally restricted to householders who lived in the same voting jurisdiction and residence from 2004 to 2010 – a tenure length that is more characteristic of higher age groups.)

And while growing older is associated with voting more often, it’s also associated with using less electricity.

The chart above exclusively displays electricity consumption (i.e. it excludes fuels like natural gas and heating oil that can also be important components of home energy usage). But the pattern is consistent with Department of Energy data indicating that older households also exhibit lower total energy consumption than their younger neighbors. For example, homes whose head of household is aged 65-74 consume 13% less total energy per year than homes whose householder is aged 45-54.

A few key reasons that older households have lower energy consumption:

  • Empty nest: The kids have already moved out, so there are fewer electricity users roaming around the house.
  • Fixed income: Older households are more likely to live on a fixed-income, so may be more attentive to minimizing energy bills.
  • Fewer gadgets: Although older Americans do watch more TV than any other age group, they also spend considerably less time using other consumer electronics like computers and power-hungry gaming consoles.

As a variable, voter age drives most of the electricity consumption difference between frequent- and infrequent-voting households. And yet, our analysis shows that variation in age alone is not sufficient to fully explain why frequent voters use less electricity.  We identified another statistically significant dynamic that appears to transcend age…and implies a fascinatingly direct relationship between political participation and energy-efficient behavior.

Two sides of the same coin: civic engagement and energy efficient behavior

Our analysis of the relationship between voting frequency and electricity usage suggests that, beyond age, there is something special about politically engaged Americans that also leads them to consume less energy.

Consider, for example, two American householders — let’s call them Sam and Pat  – who are virtually identical: same age, same neighborhood, same income level, same home square footage, same kind of heating system, and same number of children.

There is, however, one key behavioral difference between Sam and Pat: Sam voted in ten elections during the 2004-10 time period, while Pat voted in none.

When we compare the yearly household electricity consumption of Sam and Pat, a stark difference appears: Sam, the frequent voter, on average consumes 659 fewer kilowatt-hours per year (~$80 lower annual electricity bill) than Pat.

This pattern is precisely the one that emerged from our analysis across 137,000 American homes. Using a straightforward statistical technique called linear regression (see Methodology), we were able to isolate the statistical relationship between voting frequency and household electricity usage, and arrived at a striking conclusion:

Holding all other factors fixed, each ballot that a voter cast between 2004-2010 is associated with an incremental 66 kilowatt-hour (~$8) reduction in the voter’s average annual household electricity usage.

This remarkable result prompted us to ask:

What is it about frequent voters in particular that puts downward pressure on their household electricity consumption?

We can propose a few hypotheses:

  • Patriotism? People who regularly cast ballots may do so in part from a sense of patriotism or national pride – which, in a nation that increasingly prizes the notion of energy security, can also motivate a commitment to energy efficiency.
  • More informed? It’s possible that frequent voters may simply be more informed or engaged when it comes to popular public issues like energy, and so are more attuned to adopting energy efficiency in their own lives.
  • Perceived importance of individual actions? Enthusiasm for voting and a passion for energy efficiency may originate from a common underlying belief – specifically, a belief in the power of individual actions (e.g. a single vote or one person’s habit of turning off the lights when leaving a room) to influence large-scale outcomes (e.g. election results or global environmental sustainability).

The causal reason for the inverse relationship between voting frequency and electricity usage isn’t clear-cut — and we’re curious to read your thoughts and hypotheses in the comments below. But the statistical phenomenon we’ve identified does make one thing clear…

No matter who you’re voting for on November 6th, your vote itself is a good sign for the future of energy efficiency in America.  Put simply, a more engaged citizenry is also a more energy-efficient one.

Special thanks to Ashley Sudney, Arjun Dasgupta, Tyler Curtis, Nathan Srinivas, Rachel Zuraw, David Moore, Will Pierce, Efrat Levush, Jillian Cairns, and Peter Kjeldgaard.

Follow @OpowerOutlier on Twitter

Methodology: Publicly-sourced voter data was matched with household electricity consumption data across 137,415 homes, corresponding to customer utility accounts that remained continuously active from January 2004 to August 2010. Electricity consumption data is based on monthly usage reads from January 2008 onwards.

The aggregation of householders into voting intensity categories (i.e. “infrequent”, “sometimes,” and “frequent”) reflects the fact that there is not a meaningful behavioral distinction between, say, a person who voted 3 times and a person who voted 4 times during the 2004-10 time period; but there is a fundamental difference between voting 3 times and voting 9 times. The underlying findings of our analysis are not sensitive to marginal changes in the definitional boundaries of the aggregated groups.

Specific average consumption levels for each voting intensity category, from infrequent to frequent, are as follows: Eastern State (11,911 kWh; 11,774 kWh; 11,051 kWh), Western State (10,313 kWh; 10,131 kWh; 9,332 kWh). Confidence intervals (at the 95% significance level) span approximately +/- 50 kWh for the Eastern state, +/- 140 kWh for the Western state.

Multivariate linear regression analyses were run for each regional dataset (neast = 121,847; nwest = 15,298), regressing annual household electricity usage against a number of key household characteristics, householder traits, and a variable representing the number of total votes that a householder cast during the 2004-10 timeframe. In both sets of model results, the regression coefficients for householder age and total votes cast were strongly negative and statistically significant at the 99% confidence level. The regressions exhibited F-statistics that substantially exceed the critical values required to reject null hypotheses: (15, 91991) = 1740; (15, 10357) = 226. Both regressions as a whole were significant at the 99% level (p < 0.001).

Data Privacy: All data analyzed here are anonymous and treated in strict adherence to Opower’s Data Principles.

Author’s note: The analysis and commentary presented above solely reflect the views of the author(s) and do not reflect the views of Opower’s utility partners.

Read More

Turning weather forecasts into power outage forecasts

  • By Barry Fischer
  • October 30, 2012

Is it possible to examine a hurricane forecast and figure out in advance how many customers will lose power because of the storm? An innovative engineering professor at Johns Hopkins University is demonstrating that it can be done with reasonable accuracy.

The professor, Seth Guikema, has developed models that take in 120 climatic, topographic, and demographic variables (including wind gust speeds, soil moisture, and population density). The models then crunch together all the data to project the likelihood of power outages and generate an outage risk map like the one below, produced ahead of Hurricane Sandy. The brighter colors show which locations are more likely to suffer from blackouts.

Source: Atlantic Cities

During Hurricane Irene in August 2011, Guikema’s models predicted the outage impact in some states within an impressive 10% margin.

The Johns Hopkins research team issued a conservative pre-Sandy forecast suggesting that 10 million utility customers would face power outages this week. As of this morning, the preliminary outage toll has reportedly reached more than 7.5 million customers across 15 states and Washington DC.

Read More
Our Thinking

Nick Heiner, Come On Down!

  • By Hillery Brown
  • October 30, 2012

Nick Heiner just joined the Web Platform team as an Associate Software Engineer. He will be working on both the front and back end, developing additional functionality allowing consumers to make intelligent decisions based on their energy usage data.

Nick graduated from Cornell in May 2012 with a degree in Information Science and a minor in Game Design. His current favorite language is F#, although he also likes C#, Python, and long walks on the beach. For his game design coursework, he developed HollowWood, a game involving zombies, ghosts, and a dark forest, and Corpionage, a Facebook game that’s like Farmville but you can raid your friends. This summer, he implemented the card game Dominion in F#. He had a lot of fun TA-ing for several intermediate programming courses.

In prior summers, Nick has interned at Microsoft and Google, working on Outlook 2013 and Adwords, respectively.

1. What made you want to join Opower?

When I first interviewed at Opower, I was struck by how laid back and fun the culture was. I haven’t done many interviews where I get to ride a scooter around the office. When I was trying to decide which of my job offers to accept, Opower allowed me to speak to engineering lead Michael Parker, CEO Dan Yates, and investor Hadi Partovi. I was flattered that people who were that busy were willing to make time to talk to a kid fresh out of college, and having direct access to them provided valuable information for my decision.
2. What has been your favorite thing about Opower so far?

Smart people, fun problems, free beer, 16gb new Macbook Pro, etc. Arlington and DC are nice areas – there is a lot to do without being as imposing as NYC. I can learn about cool new technologies as part of my job, as opposed to school where it mostly had to happen on the side.

3. What are some of your hobbies?

I enjoy fencing, gaming, biking, running, eating, and of course I have an ever-growing list of coding side projects I’d like to work on. (In particular, I like F# and Windows Phone apps.) I fenced for 4 years on the team at Cornell, and am proud to join fellow alums Mike Fotinatos and Matt Herndon here at Opower.

Read More
Older Posts