POLLING has become a thriving business in the territory, thanks to elections. A number of people have grown famous, and probably rich too, for the polls they are commissioned to take by the media in connection with our elections. With the Legco elections only a few weeks away, results of these surveys appear in the media every day. They tell the public how much support each candidate has, and which candidates are likely to win. One does not need to be an expert statistician to see some of the technical problems of these polls. The surveys are done with a sample size of at most a few hundred. In one widely-published survey, responses from 50-odd people were used to determine the voter support percentage for each candidate in a geographical constituency where there are over 100,000 registered electors. The sampling method is seldom, if ever, explained. The respondents are invariably 'interviewed' by telephone. There is no way for a polling agent to acquire the telephone numbers of registered voters, which means the agent cannot possibly start with a comprehensive register of electors to obtain a random sample. In 1991 a researcher in one of the tertiary institutions predicted a 70 per cent turnout rate for Hong Kong's first direct election for the Legislative Council. His judgment was based on an opinion survey carried out shortly before the election. When polling day came, fewer than 40 per cent of electors cast their vote. Perhaps accuracy is totally irrelevant in the polling business, for the academic whose optimistic survey results proved to be wide of the mark in 1991 is now back in the media confidently analysing next month's election, working on surveys commissioned by a number of newspapers. Whether accurate or not, these survey-based predictions may influence voter behaviour. The Boundary and Election Commission (BEC) is concerned about the impact of publishing exit poll results on the decision of electors on election day. In its Guidelines on Election-related Activities the BEC says: 'Any announcement of results of exit polls or predictions, particularly in relation to individual candidates, during the polling hours may affect voter behaviour and have an impact on election results. 'The BEC, therefore, appeals to the media and organisations concerned to refrain from announcing the results of exit polls or making specific remarks or predictions on individual candidates' performance until after the close of polls.' The BEC, however, seems to see no problem in the publication of polling results before and up to election day. It makes no attempt to discourage 'the media and organisations concerned' from reporting surveys on electors' preference, 'making specific remarks or predictions on individual candidates' performance'. The people who read the published survey results with the greatest interest and care will no doubt be the candidates themselves and their campaign chiefs. Political parties fielding candidates in various constituencies, if they believe in the poll results, will make use of them to modify and improve their campaigning strategy. A party can, for example, pull resources away from a constituency where a sure win is predicted for its candidate, and reinforce its campaign elsewhere for a weaker contestant. The influence of the poll results on the individual elector is more difficult to assess. If an elector has made up his mind to support a particular candidate, and is told that his favourite candidate is either leading or falling behind the opponent by a very wide margin, he may feel his vote has little significance, and lose the motivation to exercise it. On the other hand, he may also notice the large proportion of undecided respondents, and believe that his vote is important because a swing in the balance can still happen. It is not unlikely that survey results published before polling day may have similar effects on voter behaviour as exit poll reports. It would, of course, be impossible for the BEC to ban election-related surveys or the publication of their results, but the commission should at least be aware, and be ready to admit, that there are ways of 'unfair interference with the election process by unduly influencing electors' that cannot be prevented.