Thirteen months and what do you get? As far as Delhi is concerned, three exciting and unpredictable elections with the decision of the third scheduled for February 10. The excitement and unpredictability is all because of the presence of the new and different political party—the Aam Aadmi Party (AAP). Every opinion poll in December 2013 had the AAP barely making its presence felt. It ended up second, with 28 out of 70 seats, and even formed the government before unceremoniously kicking itself out. In the Lok Sabha elections, the AAP’s expectations were high, not only for Delhi, but also the rest of India. We had some expert commentators even believing that the AAP would be part of the national government. They ended up with 4 Lok Sabha (LS) seats nationally (all in Punjab), and none in Delhi.
An assembly segment break-up of the Delhi LS election reveals that the BJP won 60 of the 70 assembly segments, and the AAP won 10. Opinion polls in January, to date, show a huge bounceback for the AAP—a 5% increase in their vote-share compared to December 2013—of 35%versus 30%. The BJP also shows a near identical increase in vote-share, from 33% to 39%. In terms of seats, the prediction is for a near-dead-heat between the two, with the BJP marginally ahead, 35 to 31. Even a dead cat bounces so why should we grudge Congress this chance of spoiling the BJP party. Given forecast errors, another hung Assembly in February 2015 cannot be ruled out.
Only brave souls, and statisticians, can even attempt to forecast this election. It has been further jumbled up with the AAP defections, and with the announcement of Kiran Bedi as the chief ministerial candidate of the BJP. Does that mean that an uncertain election is now certain for the BJP ? As we await opinion polls, exit polls, and the final result, herewith some statistical calculations—from a brave forecaster!
The election outcome is forecast based on two separate statistical methods. The first is to look at history for parallel events. The first cut of history is to look at all assembly elections post 1991 which followed a national election within 12 months. The second cut of the data is to look at only those parties who obtained at least 20% of the vote in both the LS and the subsequent assembly election. The results are revealing. Only 9 out of 26 elections show an increase in vote-share for any party in the assembly as compared to the previous LS election; and just three show an increase in vote-share of more than 1.3%! One of these outliers was in Delhi itself; in 1998, the Congress increased its vote-share by 5.1% over the Lok Sabha election earlier the same year. The other two outliers were in Punjab and Haryana, for the Akali Dal (8.9% increase in 1997) and INLD (4.3 percentage point increase in 2005). Average for all the 26 elections is minus 3%.
Opinion polls are forecasting a 2% gain for the AAP over its LS share. The opinion poll loss for the BJP, at 8%, is nearly three times the average loss in the 26 comparable elections since 1991. As they say in statistics-land, the odds are heavily against both the AAP gaining 2 percentage points and the BJP losing such a large vote-share. But this is a special election in that—a third party, the Congress, is in danger of falling off the map. So its vote-share will likely go to the two parties, the BJP and the AAP, with the latter likely to get more. But it takes a stretch of imagination, and calculation, to say that the net loss for the BJP over the LS election would be 8% and a net gain for the AAP of 2%.
The second method is to predict “reasonable” swings in vote-shares for the February 2015 elections over the observed December 2013 Delhi polls. Five different experiments were conducted with the AAP median gain in vote-share of 2 percentage points over the 2013 assembly polls. For the BJP, the median “experiment” vote-share is 40%, 7 percentage points less than the Lok Sabha election. For the Congress, the median vote-share change over a disastrous LS election is a gain of 5 percentage points.
Details of all the five experiments are provided and the results converge on the following. First, it seems highly unlikely that Congress would get much more than 5 seats, a fall of 3 seats from their already low 2013 total. The AAP is unlikely to gain much more than 21 seats—four of the five reasonable experiments converge on 21 seats, and only one scenario yields 28 seats for AAP. The BJP should easily win this election with a median prediction of 44 seats. A hung assembly or an AAP victory would be a Black Swan or a “man bites dog” event.
As I said at the beginning, the motivation for this analysis was to have a bit of fun, and to see how likely were the opinion poll predictions of a close contest. No matter how one slices the data, the initial set of opinion polls are likely to be way off the mark. If the AAP succeeds in getting around 31 seats, and BJP only a few seats more, I will be off the mark. Let the fun; sorry, games; sorry, elections begin.
The author is chairman, Oxus Investments, and a senior advisor to Zyfin, a leading financial information company. Twitter: @surjitbhalla