It is not just the defeated candidates who will be the losers this Sunday: some opinion pollsters are likely to be joining them. Throughout the campaign, a number of polls have placed the Democratic Party ahead of The Frontier's Emily Lau Wai-hing in the key battlefield of New Territories East - the seat where the outcome remains in most doubt. An early poll conducted by Asian Commercial Research for this newspaper even predicted Ms Lau would fail to secure a seat. Another, by the Baptist University's Hong Kong Transition Project, suggested the Democrats would take three of the five seats in the constituency, a result which far exceeds the party's own expectations. Highly respected polls by the University of Hong Kong's Social Sciences Research Centre have also put the Democrats ahead of The Frontier, by a narrower margin. But there have also been several other opinion polls which have shown quite the reverse. Surveys conducted by a team at Chinese University and published in both Ming Pao and Apple Daily, as well as the Democrats' own private polls, have placed Ms Lau's ticket in the lead by a wide margin, a finding most analysts in the district would endorse. One such survey even suggested she and running mate Cyd Ho Sau-lan will take two of the top three seats in the constituency. Clearly, both sets of poll results cannot be right. Nor is this likely to be the only constituency in which the polls are likely to be proved wrong. In New Territories West, surveys have repeatedly recorded less than two per cent support for the Heung Yee Kuk list led by vice-chairman Daniel Lam Wai-keung. But such findings are hard to reconcile with the situation on the ground, where the other parties now view this rural alliance as a serious threat which is certain to poll respectably well and may even take one seat in the constituency. Like almost everyone else, the pollsters have been badly affected by the switch to a proportional-representation list system. In previous elections, they were pretty good at predicting the winners in the then much smaller constituencies, so masking their relative inaccuracy in estimating the percentage of the vote that different parties received. But, under the new system, that is what is crucial in determining who will win. The introduction of huge new constituencies also poses particular problems for the pollsters. With 28 candidates in New Territories West alone, many have given up reciting the complete list when asking respondents how they plan to vote, although the University of Hong Kong team have commendably continued to do so. But taking shortcuts can prove disastrous. One poll, which only gave respondents a list of parties to choose from, vastly underestimated the support for Ms Lau and Christine Loh Kung-wai, whose support comes from their personal popularity rather than party affiliations. Some anomalies can be explained by special factors. The kuk's rural supporters tend to be older and so less willing to answer pollster's questions. Nor can survey findings properly reflect the mobilisation skills of this group, who are renowned for persuading a particularly high percentage of their supporters to turn out to vote. Such mobilisation skills also mean the Democratic Alliance for the Betterment of Hong Kong's support is often under-represented in polls. But the widely varying survey findings in New Territories East are more difficult to explain. Irrespective of who wins in that constituency on Sunday, some pollsters are going to have a great deal of explaining to do. Just as the 1992 British general election marked a turning point for British pollsters, who were forced to re-examine their methodology after badly mispredicting the result, so this election may prove a similar watershed for Hong Kong pollsters.