India's National Magazine
From the publishers of THE HINDU
Vol. 15 :: No. 06 :: Mar. 21 - Apr. 3, 1998
Of polls and predictions
The electoral outcome in Tamil Nadu.
THE outcome at the national level of elections to the 12th Lok Sabha has confirmed the prediction of a hung Parliament made by practically all the pre-poll opinion surveys. However, beneath this apparent confirmation, lie some State, level poll outcomes that have been quite different from those predicted by most pollsters. By far the most stunning surprise has been the electoral verdict in respect of the 39 Lok Sabha seats from Tamil Nadu. Practically all the serious pre-election public opinion surveys, including the Frontline-Apt Research Group poll, predicted a sweep for the DMK-TMC front. While not all the pollsters provided details of the estimated vote shares of the rival political fronts nor of the methodology by which vote share forecasts were translated into seat predictions, practically all of them concurred in predicting that the DMK-TMC front would get anywhere between 33 and 36 seats - and that the AIADMK-BJP combine would get no more than six seats. As it turned out, the outcome was almost the reverse of these forecasts. The AIADMK-BJP combine which includes the PMK, the MDMK, the Janata Party and the Rajiv Congress, has won 30 Lok Sabha seats, while the DMK-TMC-CPI front has picked up only nine seats.
THIS raises two questions. First, why did all the serious polls - both the pre-election public opinion surveys and the exit polls - get it so completely wrong? Second, and obviously far more important, why did the DMK-TMC front lose so badly?
The question of why the actual electoral outcome in Tamil Nadu turned out to be so different from that forecast by pre-election public opinion surveys needs to be addressed seriously. Otherwise there is every possibility that "popular opinion", reinforced by those with a vested interest in undermining serious efforts to assess public opinion on issues, might come to dismiss scientific opinion polls as irrelevant at best and downright "motivated" at worst. As was noted in an earlier article (Frontline, March 20, 1998, pages 127-129), carefully designed opinion surveys do have far more of an objective basis than impressionistic reports by journalists.
Yet this does not imply that scientific opinion surveys cannot go wrong. There are, in fact, some very specific reasons why electoral forecasts made on the basis of such surveys can go wrong. In sampling, one usually assumes that the characteristics of a population from which one has drawn a sample are stable. In the case of voting intentions, this may not be a valid assumption, especially when one or more events that occur between the time of the survey and the time of the actual voting impact directly on the voting decision. In the case of the latest Lok Sabha election, the Coimbatore bomb blasts and the events of the week that followed could definitely be regarded as such a set of events with regard to voting decisions in Tamil Nadu.
TAKE the case of the Frontline-Apt Research Group pre-poll opinion survey in Tamil Nadu. An examination of the results in the four Lok Sabha constituencies where the Frontline-Apt Research Group pre-poll survey was conducted with those for the State as a whole shows that these four constituencies taken together had more or less the same distribution of vote shares among the two major fronts as the State as a whole. In that sense, the proposition that the four sample constituencies may be taken to be reasonably representative of voter behaviour in the State is not altogether without merit. Further, the modified cube law applied to the actual vote shares (and not to expressed voting intentions prior to the election) does give a seat forecast which is close to the actual outcome.
On the assumption of a somewhat bipolar contest, the seat forecasts based on actual vote shares of the AIADMK-BJP combine and the DMK-TMC front work out to 26 and 13 respectively for the two fronts. Allowing for tripolar contests in Nilgiris, Madurai, Pudukkottai and Tenkasi, these forecasts can be modified to 30 and nine for the AIADMK front and the DMK front respectively, which is in fact the actual outcome. This suggests that the formula used in the Frontline-Apt Research Group survey to convert vote shares into seat shares was empirically robust.
Thirdly, a number of cross checks on the sample of voters drawn in the Frontline-Apt Research Group pre-election opinion survey appeared to confirm the representative character of the sample and the consistency of the expressed voting intention with responses to other related questions on voter approval of the DMK's record of governance, the DMK-TMC alliance, the AIADMK-BJP alliance and so on. The fact that the electoral verdict is so drastically different from the forecast by the Frontline-Apt Research Group pre-election survey should therefore not be read as demonstrating either the impossibility or the irrelevance of scientific opinion surveys. What is indeed warranted is an analytical political explanation.
TO say this is not to argue that the methodology used in the Frontline-Apt Research Group poll was perfect. Perhaps, a simple random sample of a larger number of Lok Sabha constituencies would have captured the complex political heterogeneity of different regions within the State, such as the strength of the PMK in the north and the Dalit-Thevar caste polarisation in the south. Nevertheless, it seems reasonable to argue that a cataclysmic event like the Coimbatore blasts could have significantly changed voter preferences.
While the Frontline-Apt Research Group pre-poll public opinion survey was the only serious pre-poll survey carried out with respect to Tamil Nadu, there were at least three exit polls: the Frontline-Apt Research Group exit poll, the exit poll carried out by Development Research Services (DRS) for Doordarshan and the exit poll done by ORG-MARG for the television channel TVI. It is interesting to note that their predictions were substantially similar. While the ORG-MARG poll gave the AIADMK-BJP combine three Lok Sabha seats, the DRS and Apt surveys forecast four seats for the combine. Thus all three exit polls predicted a sweep for the DMK-TMC front. The exit polls thus seemed to confirm categorically the Frontline-Apt Research Group pre-poll survey forecast as also that made by A.C. Nielsen for Outlook before the elections. In fact, the apparent confirmation of pre-poll forecasts by the exit poll findings made the actual electoral outcome in Tamil Nadu appear all the more surprising. However, exit polls are inherently unreliable, given the way they are conducted.
Basically, exit polls are done in a completely non-random manner. There is no sampling frame (that is, a list of the units in the population) from which a sample is chosen. Only the more articulate among the voters respond to exit poll inquiries, and this biases the sample considerably. Exit polls are conducted at or near polling stations as voters emerge after voting, which makes for a very inappropriate interviewer situation. Non-responding voters are usually arbitrarily replaced with responsive ones. For these and other reasons, exit polls are untrustworthy. Only in a massive wave would exit polls also turn out right.
This last point provides an important clue to an analytically based political explanation of the electoral verdict in Tamil Nadu. Although the AIADMK-BJP-PMK-MDMK combine won 30 out of 39 Lok Sabha seats, its vote share at 48.53 per cent of votes polled was only 5.83 percentage points higher than that of the DMK-TMC-CPI front at 42.70 per cent. Clearly, we do not have a "wave" of any kind here. However, the impression of a wave arises not merely because the AIADMK front has won more than three quarters of the seats, but also because the result is so much at variance with both the 1996 poll outcome and the pre-election opinion surveys as well as exit polls of 1998.
IT is important to recognise that the 1996 election was a very unusual election. The 54.86 per cent vote share of the DMK-TMC front in that election was far higher than would be warranted by "normal" political arithmetic, and represented a large protest wave against the then ruling alliance of the AIADMK and the Congress. The vote share of the DMK-TMC-CPI front in the 1998 elections at 42.70 per cent is certainly much closer to normal in terms of political arithmetic. It is significant that the Frontline-Apt Research Group pre-election public opinion survey estimated the vote share of the DMK-TMC front at 42.46 per cent, remarkably close to the final outcome. The Coimbatore blasts and the almost daily haul of explosives reported in the media between February 14 and February 22 seem to have decisively influenced the nearly 10 per cent of undecided voters and a large proportion of supporters of the Congress(I)-AIADMK(T) front (which had 13 per cent of the expressed vote support in the Frontline-Apt Research Group pre-poll survey but secured only 5.88 per cent of the votes in the elections) to switch to the side of the AIADMK-BJP front. A section of those who favoured the DMK-TMC in the survey as well as the 4.36 per cent who did not support any of the three fronts may also have shifted their allegiance, but this may be of marginal relevance. In any event, the blasts and their aftermath have clearly had a significant negative impact for the DMK-TMC front.
THE Coimbatore carnage appears to have affected the ruling party and its allies in the State in at least three ways. The ruling party has been perceived by a section of the voters as being 'soft' on minority fundamentalism. The blasts have also promoted communal polarisation and consolidated a section of voters behind the BJP-AIADMK front. Thirdly, the week-long haul of explosives and sporadic clashes together with the blasts have led a section of voters to take a very dim view of the ruling party's handling of the law and order situation.
A second key factor in the change of mind of a sizeable section of voters as the voting days approached would appear to be a concern with stability at the Centre, which would also have possibly been intensified by the Coimbatore blasts. This concern, together with the perception that the other national party, the Congress(I), had no chance of winning in Tamil Nadu, would have influenced a significant section of voters who had expressed a preference for the Congress(I) front in the pre-poll survey to shift to the BJP-led front. There is thus a considerable element of truth in the argument that the vote for the AIADMK-BJP combine represents a concern for stability rather than support for the AIADMK. Voters in Tamil Nadu have been known to differentiate between elections to the Assembly and elections to the Lok Sabha - the 1980 and 1989 elections to Parliament versus the corresponding Assembly elections being cases in point.
Besides the immediate impact of the Coimbatore blasts and the concern over stability, several other factors worked against the DMK-TMC front. Particularly important in this context was the failure of the DMK-TMC leadership to handle the issue of electoral seat adjustments in a mature way. Their attitude to the CPI(M), which had stood steadfastly by the DMK over the Jain Commission report (unlike the TMC which vacillated and took its time to decide its stance), cost them dear in terms of active cadre support as well as popular image, besides specifically contributing to defeat in the Madurai Lok Sabha constituency. The State secretary of the CPI(M), N. Sankariah, had warned the ruling front against complacency and had stated clearly that the electoral contest would be closely fought. The CPI(M) had in fact tried to get the DMK front to accommodate the PMK as well, but had been rebuffed. In retrospect, these two failures directly cost the DMK-TMC front five Lok Sabha seats: Madurai, Chidambaram, Dharmapuri, Vandavasi and Vellore. It is another matter that the CPI(M), being a disciplined and cadre-based party, helped in a big way in the victory of the TMC in the Nagercoil Lok Sabha constituency.
Another factor at work was the alienation of the ruling front from Dalits as a result of its somewhat inept handling of caste clashes in southern Tamil Nadu. In two Lok Sabha seats - Tenkasi and Sivakasi - the Puthiya Thamilagam, led by Dr. K. Krishnasamy, and having a significant following among Dalits of the southern districts, polled more than 100,000 votes. In the resultant tripolar contest in Tenkasi, TMC candidate M. Arunachalam lost to the AIADMK candidate.
TWO further points need to be made. The lacklustre performance of the DMK Government, and the vigorous implementation by the U.F. at the Centre and the DMK in the State of policies of liberalisation and globalisation that have had a negative impact on the lives of working people, have been factors working against the DMK-TMC front and eroding their popular support over the last 20 months. That the most vociferous spokespersons of these policies - P. Chidambaram and Murasoli Maran - won does not in any way negate this argument. The second point here is the general anti-incumbency factor, to which the impact of corruption in local bodies may be added. While this may not by itself have been significant across the entire State, such discontent would have made a difference in marginal constituencies. In at least four constituencies - Chidambaram, Dindigul, Tiruchirapalli and Tirunelveli - the DMK front lost by margins of less than 20,000 votes or less than 3 percentage points.
The electoral outcome in Tamil Nadu holds important lessons for the ruling DMK-TMC combine. It needs to engage in a serious process of self-critical introspection and must address people's concerns about law and order, economic policies and terrorism and fundamentalism. It must also specifically reflect on the reasons for Dalit alienation, and on the perceptions of its allies concerning its approach to them. Finally, corruption is by no means a non-issue. By the same token, the ruling front has no reason to feel despondent. While the AIADMK-BJP camp may be temporarily jubilant, the electoral outcome does not imply that people have ceased to be concerned about corruption, ostentation and political arrogance. People have been particularly influenced by a major cataclysmic event, concern for stability and the fact that they are voting for the Lok Sabha. Too much should not be read into the verdict in terms of voter preferences regarding State governance.
A final word on opinion surveys. Exit polls should be heavily discounted, but scientific public opinion surveys need to be taken seriously even if their forecasts prove quite wrong on occasion. There is of course room for improvement in their methodology, design and actual conduct of field work.