The usual suspects

What the experts all forgot to a man (and woman) was that a similar drama was played out a year ago with the Uttar Pradesh election results. I had said yesterday that the forecasters then had failed to give adequate weightage to the BSP's late surge. This time, they compounded that error by not factoring it into their current analysis. What else do they need to do?

 

To begin with, they must understand that pre-election and exit polls are fact-finding, not hypothesis testing, exercises. Simple, straight-forward questions, without any baggage of value-laden presumptions, are the only ones needed. We cannot be seeking explanations for actions even before we know what they are. We also must remember what Nobel Laureate Daniel Kahneman has said: "People often act without knowing why they do what they do."

From what one saw of the ever-so long pre-poll investigations in Gujarat, anyone with a questionnaire or a microphone in hand believed they had a divine right to interrogate their respondents with virtually "have-you-stopped-beating-your-wife" type of questions. Rocket science is not required to predict responses to this kind of browbeating.

Second, in the name of disaggregation, analysts often impose their own perceptions in place of those of the voters. Caste- or region-wise groupings do not always work, and efforts at interpreting results in these straitjacketed patterns could well be misleading. The presumed Patel and Koli revolts in Gujarat never happened, nor were there any significant regional differences in voting, bar a couple of districts. The repeated tagging of Mr Yeddyurappa as a Lingayat and not a state leader was similarly not quite correct. We must not bend voter behaviour to suit our pet theories; rather, we must accept ground reality first and invent theories later.

Third, the task of translating vote shares into seats is daunting, especially in multi-cornered elections. Dr Prannoy Roy had come up with his novel multiplier in the 1980s, which effectively translated swings into seats. Unfortunately, the rise of local parties and considerations now make it nearly inapplicable. Simple statistical exercises or computer models will be of little help.

What is needed is on-the-ground observation helping analyse the swings in a local-specific manner. Knowledgeable persons (who must first put aside their own political preferences and values) need to be deployed to make subjective judgments on, say, a district-by-district basis, in addition to surveyors or reporters. One hopes that errors of commission and omission will balance out and the resulting forecast will be more accurate.

One never-failing forecast is that representatives of various parties will question unfavourable survey findings and claim they have information to the contrary (e.g. Ms Natarajan, the KPCC president, Mr H D Kumaraswami yesterday). They routinely appear on all TV specials, seeming to learn nothing from each such outing. Their political future depends on their steadfast suspension of disbelief. But should experts also not learn from their past mistakes? The loss of professional credibility is bound to catch up with them sooner rather than later.