Why it’s difficult to have a Nate Silver in India4 min read . Updated: 12 Nov 2013, 10:58 AM IST
It is extremely difficult for an Indian Nate Silver to be able to call the 2014 elections with any degree of reliability
A question that has come up with unfailing regularity on social media forums recently is: “Who is India’s Nate Silver?" For the uninitiated, Silver is an American election analyst who shot into the limelight in 2008 when he correctly called the results of the presidential elections in 49 of the 50 states. He did one better in the 2012 US presidential elections when he correctly predicted the results in all 50 states.
What is interesting about Silver’s work is that it was completely quantitative. He used publicly available data from various opinion polls, developed a system to weight the opinion of different pollsters, came up with a rating system for election experts and then put all this together to come up with his predictions.
With the rise in data-oriented analysis of elections in India (including this series), there is a belief that an Indian Nate Silver is lurking somewhere, and that if the American elections can be called using quantitative methods, there is no reason that India’s elections cannot be similarly forecast.
However, there are several hurdles an Indian election analyst faces that Silver didn’t have to get past. This has to do with both the quality and quantity of opinion polling data, and our electoral system.
Number of polls
In the US, from the time the nominations were announced (on 6 September 2012) till the country went to polls in early November, there were at least 105 (yes, you read that right—105) opinion polls conducted and results published at the national level. Apart from this, there were a large number of opinion polls at the state level. This volume of opinion polling is simply absent in India.
Quality of polls
As we touched upon in the previous edition of Election Metrics, the US has trade bodies for pollsters such as the American Association for Public Opinion Research or the National Council on Public Polls, which issue guidelines on the dissemination of opinion poll results. As a result, each poll is forced to reveal its sample size, methodology and margin of error. Such information is absent in India.
The US political system
In the presidential elections in the US, each state elects an electoral college, which then elects the President. To simplify, each American state is by itself a constituency. Most large states had their own opinion polls, so in effect we had opinion polls at the constituency level, which allowed the likes of Silver to call elections at the constituency level. While the electoral college in the US has 538 members, there are effectively only 50 constituencies, since all members of the college from a particular state vote the same way. In India, with 543 independent constituencies, this granularity in polling is impossible.
Despite the presence of smaller parties and independents, the election in the US has fundamentally been a bilateral contest for a really long time now. It is much easier to predict a bilateral than a multilateral contest. Also of importance is that it is the same two political parties that have a presence in all fifty states of the US. In India, there is a school of thought that says that a general election is just a collection of 35 state elections, and that makes forecasting harder.
Also, in India, given the size of a constituency, it is realistic to expect an independent or a member of a small party to win a seat. Opinion polls usually end up ignoring such candidates. This problem doesn’t exist in the US.
Ease of opinion polling
This flows from the bilateral nature and the presidential structure of the US political system. It is realistically possible in the US to conduct a credible opinion poll with just 1,000 respondents (assuming they are sampled well). This reduces the cost (both monetary and time) of an opinion poll and results in a large number of such polls. The multi-party, state-specific system in India means one needs a much larger sample, which results in less polls, thus giving less information.
Vote to seat conundrum
The presidential structure in the US means that there is no issue of converting vote share to seats. In India, on the other hand, this is the biggest problem for most pollsters. To take a recent example, in the elections in Karnataka in May this year, most exit polls concurred on the overall vote shares of the major parties, but differed widely in their seat predictions. As it turned out, their prediction of vote share was broadly right while hardly anyone got the seat prediction correct.
Given the above constraints, it would be extremely difficult for an Indian Nate Silver, even if one exists, to be able to call the 2014 general elections with any degree of reliability. While predicting the total number of seats per party should theoretically be simpler than getting every constituency right, given the fragmentation in our politics, it is unlikely someone can get that consistently right.