Why exit polls got UP election results wrong
Latest News »
One possible reason opinion and exit polls failed to anticipate the Bharatiya Janata Party (BJP)’s spectacular performance in the assembly election in Uttar Pradesh was that they couldn’t quite figure out how to treat the 2014 general election. Those election results showed the BJP secured over 40% of the popular vote in Uttar Pradesh on the way to winning 90% of the Lok Sabha seats from the state.
This was a massive improvement over the party’s performance in the 2012 assembly election, where it got a paltry 15% of votes, with less than 50 seats in the 403-member house.
One school of thought was that the performance in the general election was a blip, the result of the so-called Modi wave, and that the BJP wouldn’t be able to replicate the performance in these assembly elections.
ALSO READ | UP election results: The day after
The other school considered the 2014 performance a change of regime, and assumed a strong BJP as a “prior” while forecasting the results of the assembly elections this year.
While the role of Prime Minister Narendra Modi in the BJP’s victory in these elections is undeniable, the results also show that the BJP’s performance in 2014 was no blip, and that there has indeed been a “regime change” in the politics of the state.
While the BJP didn’t cross the 40% vote share mark this time, its 39.7% was still sufficient in a three-cornered contest to give it a three-fourths majority in the Uttar Pradesh assembly.
When a party is trying to come back from a position of weakness in a particular state or region, it would be expected to start with consolidating the areas it is already strong in before tackling the more challenging regions.
Based on the data, however, it appears that the BJP has not followed this strategy. Instead, the party went for a broad-based strengthening across the state, and scored victories even in constituencies where its vote share was in the low single digits in 2012.
Figure 1 plots the constituency-wise vote shares of the BJP, comparing 2017 to 2012. Apart from the fact that most points lie above the 45-degree line (indicating that the BJP’s vote share in most constituencies has improved), what also stands out is the rather flat nature of the plot.
More than a third of the BJP’s victories—in 107 constituencies—came in places where the party got less than 10% of the votes in the 2012 elections.
Not only did the BJP win a lot of seats, but it also won big.
As Figure 2 shows, the average (median) margin of victory for a BJP candidate was nearly 14 percentage points, the highest among all major parties (independents had a bigger margin, but that is skewed by Raghuraj Pratap Singh winning his seat by over 50 percentage points).
As the figure shows, the major party with the smallest average victory was the Bahujan Samaj Party (BSP), whose victories came at an average (median) margin of less than 3%. Incredibly, the party got the second highest vote share among all parties (after the BJP).
What hurt the party, though, was the consistency of its vote share.
In a first-past-the-post system, a consistent performance across constituencies rarely helps. A more spiky performance, on the other hand, can lead to victories in the constituencies with the spikes.
The problem for the BSP was that the 22.2% vote share the party received was spread too evenly across constituencies—it wasn’t able to garner enough votes in each constituency to win—giving it a paltry 19 seats.
It will be interesting to see in the future whether other parties that draw a significant portion of their strength from a particular caste/community similarly underperform.
Unlike what some commentators say, the BSP hasn’t lost its support. The problem with the party is that its core support base is spread too thinly through the state to give it many seats.
Finally, Bihar chief minister Nitish Kumar suggested that the reason the BJP had won big was that non-BJP parties in the state had not come together in a grand coalition to beat the BJP, unlike in Bihar in 2015.
While such analysis is usually not robust (since it ignores correlation effects, overlaps, etc.), we can see that if the SP, BSP and the Congress had managed to come together in a grand alliance, the coalition would have won a total of 270 out of 403 seats.
Significantly, even when up against this hypothetical grand alliance, the BJP would have won more than 110 seats in the assembly, which should be a good indicator of the strength of the party’s performance in these elections.