The results of any new readership study are inevitably accompanied with a chorus of protests. While many of these protests may simply be a case of sour grapes, many others make one wonder. It is difficult to fathom, for example, the consistent decline in readership of virtually all dailies in Pune and Bangalore over the past 2-3 years despite the obvious population boom in these cities and the rapidly increasing circulation levels of several newspapers. Similarly, it can be difficult for a newspaper marketer to understand why the readership of the different editions of his own brand should show such poor correlation with his actual circulation figures.
While IRS (or NRS, if it ever rises from the ashes) must take a fundamental re-look at all aspects of the research methodology involved – from sampling to the interview process to the method for projection – there is one area that I believe needs particularly urgent attention. And that is to do with the fundamental sampling methodology used.
The IRS currently draws up its sample entirely from the electoral rolls. By restricting the study only to those homes where the member(s) have gotten around to registering themselves in the electoral rolls, new migrants into the city tend to get hugely under-represented. Especially educated white collar migrants, who are notorious for their impatience in going through the fairly arduous process of registration. An independent study conducted by us suggests that in a fast growing city like Bangalore or Pune, as many as 35 per cent of SEC A readers are not registered in the rolls. Any readership research that does not represent this affluent and mobile set of readers is to my mind fundamentally flawed. This segment is likely to be a heavier consumer of media as well as of most categories of consumer products, and therefore a prime target for marketers of all hues. Hence, their under-representation is a serious lacuna.
The above sampling method also introduces a strong bias at the brand level. Different cities in India tend to have different historically-entrenched newspaper brands. Because of the strong factor of loyalty in newspapers, older residents in any city are more likely to be readers of these historical leaders. On the other hand, new residents are likely to have a very different set of preferences – in favour of a newspaper they may have been reading in their earlier place of domicile; or of a newer/more ‘happening’ brand in the new location, which they are likely to adopt with more ease than an older resident. Such readers have a lower chance of being picked up in the sample, and as a result a growing and relatively new brand in the city is likely to be underrepresented in the study.
But is there an alternative? Yes, a good one. It’s called Cluster Sampling. This method depends on the Electoral Rolls only to the extent of selection of starting addresses; thereafter, by randomly selecting a cluster of homes around the starting address, the method does not distinguish between registered voters and others while selecting respondents and therefore does not introduce the bias discussed above.
Cluster Sampling was, in fact, the methodology employed by readership studies till a few years ago. But it was apparently discontinued with - in favour of a 100 per cent reliance on the electoral rolls - with the laudable objective of reducing discretion at the field level, and therefore the possibility of error through misjudgment or willful deceit. However, this objective needs to be dealt with through stricter controls and back checks at the field level to make fieldwork quality more tamper-proof.
The problem cannot be addressed by migrating to a methodology, which introduces a whole new bias.
(Rahul Kansal is Chief Marketing Officer at Bennett Coleman & Company Ltd.)