On October 26, a 19th century bridge over the Machhu river in the town of Morbi in Gujarat — it was reconstructed with the money of the local people — was opened to the public in a grand show. Five days later, the bridge collapsed, killing an estimated 135 people and injuring 180 others. It was obvious that there was mischief and dubious dealings in the awarding of the contract to an inexperienced private firm to renovate the bridge. The Bharatiya Janata Party (BJP) was (and is) in power in the State and in the local administration; the local legislator was also from the BJP.
Falling into a trap
A little over a month later, in December, there was a State election where voters in Morbi cast their votes. Nearly 50,000 more people in Morbi voted for the BJP than they had in the election in 2017 or the 2020 by-election. That is, soon after the bridge collapsed and where many people lost their lives, the people of Morbi seem to have rewarded the BJP by voting for the party in greater numbers than they had previously. The BJP’s vote share in Morbi was a whopping 60% in 2022 vis-à-vis just 45% in 2017. In fact, even in the polling areas right near the collapsed bridge, more voters voted for the BJP in 2022 than they had in 2017. Can we then infer that the BJP won Morbi constituency only because a bridge collapsed, and many lost their lives?
It is obviously a silly and facetious, but a deliberately provocative question. Such a stupid question insults our basic intelligence and assaults our moral senses. If any, the real question to ask is: why did voters in Morbi not punish the BJP for this tragedy, especially when it happened just a few weeks before the election?
Yet, if one were to look at it dispassionately and in an abstract manner, an event x happened (i.e., the bridge collapse) and then a seemingly related event y happened (the BJP’s victory). It is tempting and intuitive for the human mind to correlate the two and conclude that x led to y.
This happens all the time in our public commentary.
In the rush to explain electoral outcomes, media commentators and even so-called experts such as pollsters and political scientists fall into this trap constantly. If a political party makes a promise or carries out a specific campaign (x) for an election and then that party wins the election (y), commentators immediately conclude that the party won the election because of its promise (x led to y). Why is it that the “x led to y” rationale seems so absurd in the case of the collapse of the bridge but is entirely plausible in the case of an election promise or a campaign, even though the underlying logic is exactly the same? Because, context matters.
Just as it is illogical to assert that the two events — the collapse of a bridge due to the BJP’s misgovernance and the subsequent victory of the BJP in that constituency — reflect a causal relationship, it is equally inappropriate to assert, without rigorous evidence, that a particular campaign or a promise or an event led to an electoral victory or loss.
Pollsters who carry out surveys to determine what influences people’s voting choices, often ask people what their major issues are, and then which way they plan to vote. They then correlate the two simplistically to draw/make strong conclusions. For example, a survey may show that people want lower taxes. If a political party promises lower taxes and wins an election, it does not automatically mean that the party won because it promised lower taxes that the people wanted, although it seems very intuitive. The more rigorous way to analyse this is to ask this: would that party not have won the election had it not promised lower taxes?
The philosopher Karl Popper famously argued that unless a theory can pass the test of ‘falsifiability’ it cannot be regarded as the absolute truth. To put it simply, he said claiming all swans are white should also conclusively mean that there is not a single black swan. The economists Abhijit Banerjee, Esther Duflo and Michael Kremer won the Nobel prize (2019) for developing a methodology called randomised control trials (RCT) that designs experiments to conclusively infer that x led to y in policy experiments such as providing midday meals in schools will lead to improved educational outcomes. A simplified version of such RCTs can be adapted for electoral surveys to draw more robust inferences.
In our example, two sets of survey questions can be designed, with one set telling voters clearly that a political party has promised lower taxes and another set in which the party does not make such a promise. Voters can be randomly given one of these two questionnaires for a survey of equal sample sizes. If there is no difference in vote shares for the party across these two survey sets, then it will be clear that a promise of lower taxes has no influence on people’s voting intent, and if there is a significant difference, then it makes sense for the party to promise lower taxes.
This is a more rigorous way of establishing causal links between the two than the current simplistic and grossly misleading manner of imputing causation through what pollsters call ‘cross-tab’ analysis. This is one major reason why the same election promise or tactic seems to work for just one party or in one State or one area or with one section of people while it does not in others, puzzling political parties and leaders.
For example, the Aam Aadmi Party promised many of the same things that the Congress party did in Himachal Pradesh in the run-up to the elections in November, but it did not seem to work for them; nor did some of the same promises work for the Congress party across all districts of Himachal Pradesh. Misinterpreting causal links for electoral outcomes through simplistic and non-rigorous narratives and developing a misleading narrative for an electoral outcome is often more dangerous than having no explanation at all.
The last word
It is easy to dismiss all this as some inane academic discussion of scientific methods (by intellectuals far removed from the ground) that are irrelevant to the real world. Learning from electoral victories or failures is extremely important to democracies as they shape and influence political and policy ideas for the nation. For example, one may wrongly infer that religious polarisation or promising zero taxes, foe example, wins elections, prompting all political parties to indulge in it in their quest for electoral victories, thereby causing irreparable damage to the nation. Attributing causes to electoral outcomes in a democracy form a serious exercise that needs to be carried out with extreme rigour and care by experts. It is too important and dangerous to reduce it to flippant theatrics for television or lazy research.
Praveen Chakravarty is a political economist and Chairman, Data Analytics department of the Congress party