Gujarat election results: Why exit polls predicted the winner, but missed the margin of victory | opinion | top | Hindustan Times
  • Thursday, Jul 19, 2018
  •   °C  
Today in New Delhi, India
Jul 19, 2018-Thursday
-°C
New Delhi
  • Humidity
    -
  • Wind
    -

Gujarat election results: Why exit polls predicted the winner, but missed the margin of victory

From a strategic perspective, it was an extremely interesting election as the challenger Congress took on the incumbent BJP head-on in the latter’s strongholds, and managed to do fairly well there, gaining about 30 seats

opinion Updated: Dec 19, 2017 08:27 IST
Gujarat election results,Gujarat poll results,Gujarat results
A BJP supporter holds a placard as he celebrates outside the party headquarters in New Delhi on December 18.(AFP Photo)

For a change, the pollsters got it right, at least in terms of direction. Exit polls conducted by various agencies after the recent assembly elections in Gujarat were unanimous in their conclusion that the Bharatiya Janata Party would win the elections with a comfortable majority. The BJP is indeed headed towards victory, though with a much smaller margin.

The reason the polls failed to forecast the appropriate margin of victory is a familiar one — they stumbled while converting votes into seats. Today’s Chanakya, famous for calling the 2014 general elections right, for example, had predicted that the BJP would get 49% (+/- 3% points) of the vote, while the Indian National Congress would get 39% (+/- 3% points) of the vote share. While these numbers are bang on (at the time of filing, the BJP has 49% of the votes while INC has 41%), the agency got the seat distribution horribly wrong (predicting that the BJP would get 135 +/- 11 seats).

Now, converting vote shares into seats is not an easy process, as I’d argued in a piece written for Mint in 2014.

Traditional models for conversion involve converting vote share forecasts into ‘vote swings’, and then forecasting the seats based on that. While it works well in most cases, the problem with such measures is that they assume that party strongholds will broadly remain constant, and it’s only the marginal seats that will tip over. Instead, what we have seen in Gujarat is a kind of crossover of party strongholds, as illustrated in Figure 1 (below).

Thanks in part to the Patidar agitations (for OBC reservations) in 2016, and the consequent alignment of Hardik Patel who led the agitations with the Congress, the Saurashtra region, which used to be a stronghold of the BJP, has tipped in favour of the Congress. Instead, the BJP has gained seats in North Gujarat, an area that used to be a Congress stronghold.

The story of the two parties’ strategies in the elections, and the reasons pollsters got it wrong, can be illustrated if we look at the constituency-wise lead for the BJP over the Congress in terms of vote share, and compare numbers across the 2012 and 2017 elections (Figure 2, below).

The first thing that stands out from Figure 2 is that in both 2012 and 2017, there were a few constituencies in which the BJP won really big (top right of the graph). On the other hand, the Congress’ biggest victories in both elections (bottom left) have come with a far smaller margin, and the party’s biggest wins of 2017 are not in the same constituencies where it scored big wins in 2012.

These two facts put together imply that despite the BJP having a large margin in terms of overall vote share (nearly 8 percentage points), the margin in terms of seats is much smaller.

The other interesting fact from Figure 2 is the number of points in the 2nd and 4th quadrants of the graph (top left and bottom right respectively), which show constituencies that changed hands.

Nearly 30 constituencies that the BJP won in 2012 tipped over towards the Congress in 2017. This includes constituencies such as Jamalpur and Kalavad, which the BJP had a lead over the Congress in 2012, but in which the Congress led by a whopping 20 percentage points in 2017; and Jamjodhpur, which the BJP had a 20 percentage point lead in 2012 and where it trailed the Congress in 2017.

Normally, a swing of 30 constituencies in favour of the Opposition would be enough to tilt the election, but the BJP struck back in what used to be the Congress strongholds, gaining a lead in 17 constituencies that the Congress had led in 2012. While most of these constituencies were led by the BJP by small margins, they made a crucial difference in terms of the BJP holding on to a simple majority in the assembly.

From a strategic perspective, it was an extremely interesting election as the challenger Congress took on the incumbent BJP head-on in the latter’s strongholds, and managed to do fairly well there, gaining about 30 seats. The BJP on its part, instead of fighting to retain these strongholds where the Congress mounted a big campaign, focussed on winning seats in the Congress’ strongholds and won just enough to keep its majority.

The only downside of this fascinating electoral story, if one might call it that, was that this “constituency exchange” resulted in poll forecasts getting badly tripped up. Nevertheless, with most pollsters getting the direction of their forecasts right, they can go home happy!

Karthik Shashidhar is author of Between the Buyer and the Seller, a book on market design and liquidity. He has been writing Election Metrics (a data-driven take on elections) for Mint since 2013

The views expressed are personal