Neighbourly insights: Pollsters ask voters to predict neighbours' choices

Instead of asking voters about their own preferences, pollsters are asking them to predict who their neighbours will vote for-adding a new twist to political polls

The US elections have thrown a spotlight on a controversial polling method called the “Neighbour Effect.” A French national who calls himself “Theo” made a significant fortune betting on Donald Trump.

 

Theo put the equivalent of at least $30 million in bets on a Trump victory on the crypto currency betting site Polymarket. He may have received winnings of somewhere between $48 million and $84 million, according to various analysts. Theo bet on an Electoral College victory, Trump winning in four key swing states, and a win of the popular vote for the Republican.

 

He claims to have placed these bets based on polls he commissioned using a reputed US pollster. What is really interesting is the way those polls were designed — instead of asking voters about their own preferences, they were asked to predict who their neighbours would vote for.

 

According to Theo, the neighbour polls showed strong preferences for Trump in states where standard opinion polls were tied within very narrow margins. One caveat: It is impossible to judge the veracity of Theo’s statements, as he has refused to share any details about the polls, or even the name of the pollster he hired.

 

But this type of polling is not unknown. It teases out information by looking at revealed preferences rather than stated preferences. We do it all the time as individuals.

 

For example, suppose you are trying to think of a gift for somebody you don’t know well. You might ask people close to that individual for suggestions on what would be suitable. You will tend not to ask the recipient directly, as politeness and social conventions may prevent that person from telling you what they really want.

 

In an opinion poll involving politics, voters may shy away from revealing their true intentions, or refuse to respond. “Shy voters” are a well-known phenomenon. However, they may be more honest when discussing their neighbour’s likely preferences. And crucially, they may reveal their own in their responses, if such a poll is cleverly designed.

 

In many cases, the respondent may also not consciously know their own preferences. Amazon and Netflix have discovered, for instance, that what people actually buy, or watch, often varies significantly from stated preferences, or wish lists. Zomato or Spotify know more about your culinary and musical preferences than you do. In politics, this can translate into an undecided voter making a snap decision right in the voting booth.

 

However, leveraging a “neighbour effect” to get useful answers can also backfire. For one, it presupposes neighbours know each other. This is unlikely in a modern urban environment, or a gated community. You may have no idea who your neighbours are, if you live in a typical 30-storey building. Neighbour effects would have to be gauged by accessing and data mining RWA whatsapp groups.

 

In a rural community or an old-style urban environment, people know each other well, but they may be afraid of standing out. Indian villages are often ghettoised by caste and community, and to a lesser extent, so are cities. This effect holds true even in America, with its Little Indias and Little Cubas.

 

People who belong to a certain community and live in an area with many members of the same community will tend to avoid revealing their preferences if they run counter to the norm. I know, for example, Sikhs who smoke and Jains who are non-vegetarian. However, they won’t indulge in these activities publicly or reveal these preferences to members of their respective communities.

 

Personally speaking, I often endure music I dislike in social gatherings. I will even use Spotify to play music I dislike out of politeness, if I’m entertaining folks with those preferences. Without controls to figure out what I listen to when alone, you will make the wrong guesses about my musical tastes, or assume the account was used by different people.

 

These are pitfalls that make it hard for pollsters to use neighbour effects. Do such factors get amplified or normalised when dealing with large samples?

 

That’s why the design, framing, controls and crosstabs of whatever Theo commissioned would be fascinating. Neighbour effects could be a sort of tiebreaker when conventional polls are tight. But they need to be used with care.