What is the probability that the various opinion polls, most of which give the BJP-led NDA a big lead in the Lok Sabha polls, will go seriously wrong on 16 May?
If the results of the last two elections are any guide, the chances are near 100 percent that they will be wrong. The two main opinion polls that have made their final calculations, NDTV-Hansa and Lokniti-CSDS-CNN-IBN, give the NDA 275 and 234-246 seats respectively. Since the latter poll excludes the Telugu Desam – an alliance stitched together after the surveys were completed – the chances are that its total for the NDA will also rise to around 250-260. That’s just a hop, step and jump away from a majority. (For numbers from earlier opinion polls, do read this BusinessLine article).
However, if these projections are wrong, majorly wrong, the NDA cannot be counting its chickens just yet. The Congress party has been repeatedly pointing out that opinion polls have been wrong in the past, and so they will be this time.
However, this is no reason for the Congress to celebrate either. Getting the final results horribly wrong and getting the incipient trends wrong are two different things. Where the opinion polls have failed in the past is in projecting much higher seats from emerging trends, especially late trends. They usually do not fail to report the beginnings of a trend. Since seat counts change dramatically on 3-5 percent swings, late swings are particularly hard to detect.
The second area they fail to gauge is the impact of regional alliances.
In 2004, for example, the final opinion polls gave the NDA a clear victory. Star News-CVoter gave the alliance 267-279. India Today-ORG 282, and NDTV-IE-Nielsen 287-307.
The final figures were 185 for the BJP and allies and 275 for Congress and allies. What a fall. The exact opposite of the final poll projections.
Of the three polls, it is Star News that got the BJP and allies’ declining trend in 2004, by giving it a lower score than the others. The India Today-ORG poll also caught the slippage. These figures were significantly below the 300-plus totals that were being projected when the first opinion polls were done in early 2004.
The opinion polls also caught the Congress’ late surge of 3-4 percent, and the BJP’s slippage, but failed to calculate its magnitude in terms of seat gains and losses.
In 2009, too, the pollsters got it badly wrong. Most polls saw the Congress slipping, and the BJP rising. While Star News-Nielsen gave the Congress plus allies 203, TheWeek-CVoter gave Congress 234, CNN-IBN 185-205, and India TV-CVoter 189-201.
The result: 262 to Congress+, and 159 for BJP-plus. The pollsters were defeated by the Indian voter.
However, look closer, and the numbers do not show a huge victory for the Congress, though it actually increased its seat-count from 145 in 2004 to 206 five years later. But Congress plus allies fell marginally from 275 to 262. Put another way, the Congress gained at the expense of the BJP, and so did some regional parties. The pollsters got the Congress+ slippage right, but not the split between Congress and its allies.
The pollsters were right to see some surge for the Congress towards the end, but failed to see the BJP’s fall.
One can hypothesise from 2004 and 2009 that pollsters get it wrong on two counts: failure to catch a rising or falling trend early enough to project where a mini-wave will lead; and two, a failure to see who will lose if one party can gain.
If we see the Indian political landscape as having three broad players – two national parties, and at least one major regional player in each major state – the net result of the elections will depend on both the relative rise or fall of the main parties, and the relative rise or fall of the national party vis-à-vis regional allies. Right or wrong alliances can make all the difference. Also, the relative performance of a national party in areas traditionally dominated by regional parties can unexpectedly impact the overall results.
So what does this leave us with in terms of projects for 2014?
The trend is crystal clear: the mood is against the Congress and in favour of the BJP wherever these two parties are in direct contention. Even where the BJP is contending with regional parties, the trend is towards the BJP or BJP-led alliances, but not in all states. Tamil Nadu and Punjab, for example, could be exceptions. Uttar Pradesh and Maharashtra are iffy, given the emergence of third and fourth players, including regional players.
So what do the failures of the pollsters in the past tell us for this time?
My own guess, no doubt coloured by my biases, is that the trend towards the NDA may accentuate as polling day nears. This is one reason why early trends caught by the pollsters in 2004 and 2009 failed to show up closer to the final figure. The chances are the surge towards NDA is being underestimated.
So, my conclusion is that the NDA may actually outperform the pollsters. One cannot rule out 300-plus for the NDA in the final analysis. Surjit Bhalla, writing in The Indian Express, projects 290 for the BJP plus allies. The only way this can be wrong is if the BJP has got its regional alliances wrong. That is the unknown X factor in these polls.
Updated Date: Apr 16, 2014 14:54 PM