What's common on to Electricity consumption - Voting behaviour - Car Offers - Policy Advocacy?
Long before, I worked on an RCT evaluating an intervention aiming to reduce electricity consumption. The intervention was simple. Every week, households would receive letters with data about their consumption previous week and the average consumption of their households in the previous week. The hypothesis was that everyone wants to save electricity but households may not know if their consumption is appropriate or they are overusing. If people are given a benchmark (neighbour's consumption) to check if they are using appropriately, they would adjust their consumption.
Since letters had to be sent every week, we needed a location where we can collect data digitally. It happened to be a large apartment complex. It also happened that the apartments had a generator for backup. There were two prices for electricity. One price for the "usual electricity". The price of generator electricity was higher.
The results showed that people significantly reduced their consumption. Telling people that they are over-consuming as compared to their neighbours, worked. The interesting aspect is that the electricity consumption reduced only for "usual electricity". It didn't change for "diesel generated electricity". It happened because diesel electricity was high. So, people already made a conscious effort of saving it. Letters didn't help as they were already saving diesel electricity. Letters helped saving usual electricity, as they weren't sure of "whether they were using appropriate amount".
In discussion with my boss, I had then mentioned that this pattern can be linked to voting behaviour. The model is same as above. In the above case, there is a product (generator), about which people are sure. So, external influence didn't matter. It mattered only when people were unsure, usual electricity. Similarly, there are two types of voters: strong supporters and others. Strong supporters would vote for a party irrespective of the situation. So, they wouldn't be necessarily affected by campaigning. The other supporters are malleable.
Two situations arise. When a party is extremely unpopular, then everyone has decided NOT to vote for it. In a two party system, opposition gains. In a not so clear context, a party can increase their votes by focusing on the malleable people.
Convincing the malleable people is a tough task. Local leaders come into play here. Since parties can't reach out to everyone, they target these local leaders, who act as a lever to convince others. I later developed a more comprehensive framework including other variables but this was the crux: convinced voters and neutral voters (fence sitters).
After few years, I was excited to know that Abhijit Banerjee and others are working on a similar area, called social networks. They were getting into the depth of this idea on the modalities of influencing people. Essentially, the spread of information through local leaders (nodes) depends on the "connectivity of the person" and the person's "credibility". Further, one doesn't need to focus on everyone in the community. There may exist an optimum of nodes for a given population size, through which information can be effectively disseminated to the rest. They are figuring out these things.
A bit on the influence of "credibility of person". During the "intolerance debate", some argued that there is actually no data supporting raise in intolerance in the country. They had used crime data for this analysis. I had then written in News Laundry that crime data is not a good proxy for measuring perception of intolerance for two reasons. One, crime is an extreme measure of intolerance. Crime data hence doesn't capture verbal and other forms. Two, the perception of intolerance is magnified when certain key leaders of the society engage in it and get away with it. It's similar to the social networks argument: The messenger of information matters.
Coming back to our central argument of malleable people and non-malleable people, some time later, I learned that Malcolm Gladwell discussed a similar idea in his book Tipping Point.
Famous economist textbook author Hal Varian had long back worked on a similar model of consumer and price behaviour. His model was on informed consumers and uninformed consumers. His idea was based on offer prices of cars. His argument was that only uninformed consumers would be affected by such advertisements and so on.
Central to all these ideas is the presence of two sections of people: "malleable" and "non-malleable". Malleable people are swayed by others (neighbours etc). Again, not all neighbours are same. Credibility of messenger and connectivity etc. also matter.
Cass Sunstein has a new paper on people's support to policy. His finding is that if people are not well informed of a policy, being told that majority supports it, will increase their odds of supporting the policy. This is similar to our model. Informed people don't get influenced but uninformed people get swayed by information of neighbours.
Simple idea but wide application!