### UPA 3 in 2019 - What are the odds?

A statistical analysis of election data.

### Snapshots

• No party will get a majority – numbers indicate alliance rule.
• Click here to view the simulated results for different vote share changes.
• Third parties will play a major role in the next government.

Elections in India always attract widespread public and media interest, and this has only grown with the advent of social media . There is no dearth of quantitative and qualitative analyses on the outcome of elections, alliance arithmetic and winners. The first past the post system makes any calculation significantly complex. Predicting outcomes is not just about forecasting how many votes a party will get – the more important (and complex) task is working out the votes to seats conversions.

### How are predictions made?

The swing of votes between the major political parties plays an important role in determining electoral outcomes. We have used a systematic approach of data analysis, whereby we use the 2014 election vote shares as the base vote share and the subsequent assembly elections’ changes (from 2014-18), to forecast the swing for 2019. The analysis is done at the parliamentary constituency level.

Our analysis looks at vote share primarily and not at the absolute numbers. Though we have not used voter turnout in our analysis, we would like to showcase the interesting patterns in the turnout in the past two elections.Here’s a look at voter turnout in the two previous general elections:

### Voter Turnout in Lok Sabha Elections

How to read the plot
• Select the ‘Election year’ from the drop-down menu at the top right corner.
• X axis shows the percentage of eligible voters who casted their vote, and Y axis (height of the bar) shows the number of constituencies.
• Each bar corresponds to a 2% range in voter turnout, e.g: the last bar in the 2009 plot represents turnouts between 90% and 92%.
Distribution of voter turnout across all constituencies
Notable observations
• In 2009, turnout distribution was bimodal but in 2014, the turnout increased and the distribution converged.
• The lowest turnout in both years was in the 24% to 26% range.
• The most common voter turnouts (tallest bars) in were 44% - 46% in 2009 and, 58% - 60% and 62% - 64% in 2014.

Converting the votes to seats is difficult and this is where most opinion polls go wrong. For instance, in 2014 the BJP won a comfortable majority with less than 31% of the vote while the Congress (INC), despite suffering a major defeat, got a 19.5% vote share which is higher than the BJP’s 18.8% from 2009. Not a single opinion poll predicted such an outcome. Vote share vs seat share for the two major parties:

### Vote share vs seat share

How to read the plot
• Use the drop-down menu to select the year.
• In each pair of bars, the first shows the percentage of votes that the party received and the second bar shows the percentage of seats they won.
• Hover the pointer over each bar to see more details.
Vote share vs seat share for different parties
Notable observations
• 2009: BJP won 116 seats with 18.8% vote share.
2014: INC won 44 seats with 19.3% vote share.
• 2009: both BJP and Congress had higher seat shares than their corresponding vote shares.
• 2014: BJP won a disproportionately high number of seats with just 31% vote Congress won much fewer seats than their 19.3% vote share would suggest.

The tools that we deploy enable specific predictions to be made for various scenarios, including unexpected situations such as these. For instance, we predict that for smaller increases in vote shares Congress will win many more seats than the BJP. How do we make this bold assertion?

### Data for the analysis

For this analysis, the past data we have used is from 2009 and 2014 General elections. We have also appropriately aggregated the data from the assembly elections held between 2014 and 2018 to extrapolate General election (Parliamentary Constituency level) results. Though assembly elections could have local issues as the focus, they have proved to be reliable indicators of the trends in the subsequent Parliamentary elections in the past. Note: the combined data from the assembly elections is called ‘2014-2018’.

To begin with, we analyse the victory margins in different constituencies to understand how closely fought elections were.

Victory margin is the difference in vote share between the winner and the closest rival.

Consider an example – a constituency in which a total of 1,00,000 votes were polled. Here, if winning candidate A received 40,000 votes and runner-up B won 30,000 votes.the victory margin is 10,000 votes which is 10% of all votes polled. Thus a lower margin indicates a close contest. Next, we look at the distribution of victory margins:

### Winning margins in all constituncies

How to read the plot
• This histogram shows how many seats recorded a particular victory margin.
• X axis shows the percentage victory margin, and the Y axis - number of constituencies which recorded victory margins within the corresponding X percentage range.
• Here, each bar represents a 1% range of the victory margin.
Distribution of victory margins across all constituencies
Notable observations
• In election year 2009, more than one third of seats were won by margins of less than 5%.
• The first bar (from the left) shows that 37 constituencies were won by a margin of less than 1%.
• In 2014, 4 constituencies had victory margins above 45%, with the highest being 50%.

In 2009, a majority of the seats were won with lower margins compared to the 2014 general elections. This is clearly visible in the form of the left skewed nature of the 2009 plot. But in 2014, the distribution evened out as the number of seats with higher margins increased. This is indicative of BJP’s emphatic 2014 victory. In 2014-2018, the margins are closer to the 2014 figures than to 2009. So we can expect the 2019 results to have fewer close contests than 2009. To get the complete picture, we must also study the magnitude of losses.

Loss margin of a party in a given constituency is the vote share difference between their candidate and the winning candidate.

### Party-wise victory and loss margins:

How to read the plot
• The cumulative distribution plots below show the spread of victory and loss margins in previous elections.
• Use the check-boxes to select which margins you want to see, e.g: select ‘BJP WIN’ to see the victory margin curve for BJP. Use the drop-down to select year.
• Hover over any point on the curve to see the corresponding X (margin %) and Y (% of seats) values, at that point.
• The X value gives a percentage victory margin, and the Y value shows what percentage of seats recorded a victory margin that was less than the current X value.
Cumulative distribution of victory & loss margins in all constituencies
Notable observations
• The loss margins of BJP in 2009 and Congress in 2014 were similar.
• 2014 BJP victory margins were much better than 2009 Congress victory margins.
• 2014 Congress loss margins were less severe than 2009 BJP loss margins.

These distributions also show that the BJP’s victory margins were similar to those of other parties in 2009. They won about 80% of their seats with margins of 15% or less. But their 2014 wins were significantly different - their margins increased dramatically, clearly indicating a favourable ‘wave’ that was widely attributed to Prime Minister Narendara Modi. Moving on to the INC, the magnitudes of their 2009 victories were almost the same as all other parties (Others). This explains the all-round low margins in 2009 that we observed in the skewed histogram.

In 2014, the Congress lead by Rahul Gandhi, saw their margins change for the worse - they won close to 70% of their seats with a margin of less than 10%. Considering that they won only 44 seats, the number of seats in which Congress candidates won by a comfortable margin barely crosses a dozen. But the Others improved on their 2009 performance and had better margins when compared to 2014. This shows that the Congress lost a significant number of votes not just to the BJP but also to other parties. In 2014-2018, other parties increased their victory margins further even as the BJP’s margins have reduced. This suggests that the BJP’s 2019 victory margins will not be as commanding as their 2014 highs, and that parties other than the BJP and INC could very well have better victory margins than both the main contenders.

Moving on to loss margins, the Congress and the BJP had very different numbers in 2009 - BJP’s losses were much more severe. Half of their defeats involved margins over 25%. This explains the Congress victory despite the BJP having similar victory margins. Whereas in 2014, Congress had more severe defeats, with margins similar to BJP’s 2009 numbers, reflecting the BJP victory. But, the loss margins of the 2 parties were much closer here - despite losing many more seats, the Congress’ 2014 losses were slightly better than the BJP’s 2009 loss margins. The difference between 2009 Congress losses and 2014 BJP losses is much more striking. In 2009, 44% of Congress defeats involved single digit margins – ie, they suffered several close losses. On the other hand, in 2014 the BJP was handed more than 80% of their defeats by double digit margins. This means that the BJP could not have won many more seats in 2014 even if their vote share had increased by 10%, touching the 40% mark. This is further reinforced by the fact that when it comes to close contests, BJP candidates have lost more than they have won. So the BJP’s impressive victory appears to have been a peak, leaving little room for improvement.

Finally, focusing on 2014-2018, both BJP and Congress suffered more severe losses. This could be the result of the increased victory margins of other parties, and leads us to believe that the two main parties will suffer more high-margin defeats in 2019. Note: Loss margins for other parties have not been included since the large numbers of candidates from smaller parties and independents will heavily skew the figures.Finally, focusing on 2014-2018, both BJP and Congress suffered more severe losses. This could be the result of the increased victory margins of other parties, and leads us to believe that the two main parties will suffer more high-margin defeats in 2019. Note: Loss margins for other parties have not been included since the large numbers of candidates from smaller parties and independents will heavily skew the figures.

### Importance of Margins

Why focus on margins? Can they really help predict the 2019 results? To illustrate how estimating vote share and margins can help, let’s see what would have happened in the previous two cycles if the vote shares had been different.

## Effect of changes in vote share

How to read the plot
• Select ‘INC’ from the dropdown to see how many seats Congress would have won in 2014, if their vote share had increased or decreased by a fraction of what they gained between 2009 and 2014. Similarly, you can see the changes for BJP.
• Actual vote shares of INC and BJP is shown by default for comparison.
• If vote share is increased, victory margin will increase – this causes the curve to move right.
• Increase in vote share also gives more wins. Hover over each curve to view additional details.
Change in seats for a simulated change in vote share

##### Total seats
Notable observations
• BJP's vote share gain (swing) in 2014 was 12.5%, giving them 31.3% vote share. If this gain had increased by 25%, that is, if BJP's gain was 15.63%, they would have got a vote share of 34.4%. This would have given them 297 seats.
• 5% vote share increase for Congress would give them a 67.9% increase in seats.
• The more dramatic shifts in the BJP curves reflect the prevalance of low margin contests in 2009.
• The relatively small shifts in the INC curves reflect the high margin BJP wins in 2014.

For 2014, if we simulate a 5% increase in vote share for the Congress, they would have won 39 additional seats (for a total of 83 instead of 44), out of which 16 would have been gained from the BJP (dropping them from 282 to 266). This is a conservative estimate as it looks at the Congress independently, without considering alliances and other factors. What’s more, since the UPA did not include regional heavyweights from most states (INC contested 464 seats), many constituencies had multipolar vote splits. Thus, a 5% increase would have resulted in significant gains for the Congress, even with a fractured vote. This reinforces our earlier finding that Congress can gain a significant number of seats with a small increase in vote share, whereas the BJP is vulnerable in close contests.

However, if BJP vote share is decreased by 5% instead, the results are very different: BJP would have gotten only 219 seats, while the INC is shown to win only 66 seats. The values are expected to vary due to the presence of other parties. However, the magnitude of the difference (282 vs 219 for the BJP) makes it clear that other parties play a significant role.

### What’s the verdict?

Considering recent trends from assembly elections, Congress vote share could increase by up to 8%. If they manage a 7.5% gain, they will win 109 seats in the Lok Sabha. But as discussed above, these numbers are highly dependent on the performance of other parties. In any case, they will end up with at least a 100% increase in seat share. Trends in margins and vote shares also indicate that the BJP will find it hard to win close contests, and that their vote share will decrease. Meanwhile, regional parties have been gaining ground and they will post bigger wins.

BJP will likely be the single largest party but they won’t win a majority - their numbers will be significantly lower compared to 2014. Congress will return as the Opposition party, but their number of wins will be strongly influenced by the alliances they make. Other parties will gain votes at the expense of both the BJP and the Congress, and will play an important role in the next 5 years no matter which alliance forms the government.