Nary a month goes by without a news story or academic publication that highlights the differences between liberals and conservatives. We drink different beer, we drive different cars, we give to different charities, we give our babies different names, we report different levels of happiness ☺.
And just weeks ago, researchers at Duke and the University of Pennsylvania reported that liberals and conservatives even act differently when shopping for light bulbs: conservative-leaning Americans, it appears, are substantially less likely to buy bulbs that carry a “Protect the Environment” label.
It seems that liberals and conservatives are just different. Or are they?
We’re excited to share one way in which liberals and conservatives are more similar than they might think: they both save significant amounts of electricity when presented with personalized analysis of their energy consumption.
Consider, for example, one of Opower’s longstanding behavioral energy efficiency programs in the Midwest. Using publicly available data, we identified the party affiliation of individuals across the region, then matched those data to the household electricity usage of approximately 100,000 utility customers participating in our programs.
As shown in the table below, we found a nicely balanced sample, with roughly one quarter of customers positioned on each side of the political spectrum, and the remaining half not identifying with any party. While there are some modest pre-existing differences across liberals (they live in slightly older houses), conservatives (they live in slightly bigger houses), and independents (they are slightly younger), the groups really aren’t all that different:
Notwithstanding some variation in structural housing characteristics like square footage and the prevalence of electric-heating systems, energy usage itself is quite similar between liberals and conservatives – with conservatives annually consuming just 3% more kilowatt-hours of electricity than liberals.
So, how did the political cohorts respond upon receiving personalized analysis of their electricity usage?
Almost exactly the same.
After 2.5 years of receiving simple energy insights (including a comparison of how they compare to the average home in their neighborhood), conservatives in the Opower program have reduced their electricity usage by 2.2%, while liberals have saved just a fraction less. The difference is statistically and practically negligible (see Methodology). Both groups have saved slightly more than their politically unaffiliated counterparts (1.8%), though they too have posted strong savings results. Over the life of the program, the average customer – party-affiliated or otherwise — has reduced their consumption by more than 1,000 kwh and saved about $120.
Does this finding hold across the country?
Like all good scientists, we try to replicate all our experiments to ensure that our findings aren’t spurious correlations. So we replicated the analysis above by examining data from three other regions — one served by a large investor-owned company in the Northeast, and two others served by municipal utilities in the Mountain West and Southeast.
Our findings were largely confirmed: when it comes to making behavioral changes to save energy at home, liberals and conservatives just aren’t that different. In two of the regions (Northeast and Southeast), conservatives tended to save slightly more than liberals. On the other hand, in our Mountain West example, liberals have saved more; interestingly, though, the conservatives there have reduced their electric bill by an impressive 2.3%, thus saving more energy than any group from all the other programs we analyzed.
What have other researchers found?
In 2011, UCLA economists Dora Costa and Matt Kahn wrote a working paper detailing a very similar analysis using data from the Opower program in Sacramento. They found that, on average, liberals saved 2.4% on their electric bills while conservatives saved 1.7%. This fits in nicely with our internal analysis: both liberals and conservatives save energy upon participating in behavioral efficiency programs, although the exact levels vary by location.
The Costa and Kahn working paper also issued a related hypothesis that piqued our interest. Namely, their paper asserts that liberals and conservatives may respond differently when told that they already use less energy than average. In particular, they suggest that when a conservative-leaning household discovers they use less than their neighbors to begin with, that this discovery may “backfire” – causing the household to feel they’re off the hook and to actually increase their energy consumption. This conservative-specific “backfire effect,” the authors posit, may stem from a combination of psychological and ideological factors associated with the conservative ethos.
We investigated the existence of the “backfire” phenomenon using the dataset described above, but arrived at a very different conclusion. We found that giving personalized feedback to energy customers who are already efficient is far from counter-productive: receiving a steady flow of analysis and energy savings advice clearly helps them to nudge their usage even further downwards. That’s true for liberals and conservatives across the country, as shown in the bar chart below.
So yet again we find that liberals and conservatives are more the same than they are different. Both tend to save energy in response to personalized feedback about their usage, even when they start out as relatively more efficient than their neighbors.
Bipartisan behavior on saving energy — how can it be?
Given the volume of research that tells us the manifold ways in which liberals and conservatives are different, especially with respect to environmental behaviors, how can we reconcile our findings with the all the other evidence? Here are a couple explanations:
- Behavioral energy efficiency transcends demographics: Political affiliation is just one of many characteristics that we’ve researched. We’ve combed through hundreds of data elements trying to understand which customers save the most energy in response to receiving personalized insights about their usage: income, age, family size, gender, religion, housing data, geography, retail purchase patterns… you name it, we’ve tried it. We always arrive at the same conclusion: behavioral energy efficiency is a universal resource.
- Good design principles work across party lines: In some ways, this universality isn’t an accident. In fact, it’s entirely consistent with effective design principles, which are aimed at delivering a powerful, relevant experience to everyone. Drawing upon proven behavioral science approaches and a deep understanding of user experience, we’ve tailored our product to appeal to all utility customers, regardless of where they stand on any particular issue.
Of course none of this means that liberals and conservatives will stop quarreling anytime soon. But it does suggest that when they’re done for the day, they will turn off the lights before leaving the room.
Tyler Curtis is Senior Director of Advanced Analytics at Opower. Special thanks to Ashley Sudney for graphics.
Follow @OpowerOutlier on Twitter
Bonus: For those of you who’d prefer to focus on energy-related differences rather than similarities in the political world, check out our November 2012 Outlier analysis examining how voter turnout correlates with home electricity usage.
Methodology: Designation of Liberal includes registered members of: Democratic Party, Green Party, Socialist Party. Designation of Conservative includes registered members of: Republican Party, Constitution Party, Libertarian Party.
Savings results are based on large-scale randomized control experiments that measure the difference in usage between test group households (e.g. receives personalized energy information) and control group households (statistically equivalent group, and of the same political affiliation, not receiving any intervention). For example, electric usage of liberal test group households is compared, ex post, to that of liberal control group households. In charts, household count of “n” refers to size of test group.
Results are calculated using panel data models with billing data through end of April 2013. All models include household month fixed effects. Results were not sensitive to other model specifications that include other regression variables such as income, baseline usage, gender, home size, and other household characteristics. Savings percentages are central estimates of 95% confidence intervals spanning +/- 0.2% to +/- 0.6%, depending on sample size, rendering the savings difference across party affiliations statistically negligible.
Data Privacy: All data analyzed here are treated anonymously and 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.