Exit polls pose several layers of complexity in
India. The sheer size of the electorate pushes pollsters to increase their sample sizes, stretching resources of all players. Competition among media organisations doesn't allow
polling agencies to pool resources, which could improve
forecasting outcomes. Polling is also less reliable in multi-contests, a hallmark of Indian politics. It becomes more difficult to predict seats from vote shares when swings for and against aren't restricted to two options, like in US or British elections. A fallout of
multi-party democracy is pre-poll alliances that cloud intertemporal comparisons.
Randomising samples among a heterogenous electorate needs extra care with mathematical modelling. These are the theoretical complications involved in our exit polls. The practical ones arise from the nature of India's politics. As a
developing nation,
policy decisions are usually without precedent, and pollsters usually have a higher rate of failure with referenda. It's trickier to capture the mood of sections of the electorate, say, pre- and post-imposition of the GST. These are nationwide policy changes with a range of geographical, economic and political effects. Local policy changes with similarly divergent impact sub-locally require fairly intricate calculations to arrive at apples-to-apples comparison. A more settled politics yields steadier
political forecasting, a scenario beyond the reach of most developing economies.
Exit polls here need to be more transparent about their particular shortcomings. More
transparency is required over methodologies and statistical modelling. Self-regulation is an obvious route. But if it fails, sectoral regulators may have to step in with standards of disclosure. So far, rules applying to exit polls are designed to reduce their impact on voter behaviour. More rules are needed to counteract their influence on investors. Burden of
regulation should be spread wider. The key, however, is to make polling a more rigorous enterprise. India-specific forecasting risks have to be identified and addressed.