Polling Accuracy in a Multiparty Election

In: International journal of public opinion research, Band 26, Heft 1, S. 113-124

Abstract

How accurate are political opinion polls in multiparty elections? Many countries have proportional representation systems with low vote thresholds for parliamentary representation. Germany has such a system, as do Israel, Poland, the Netherlands, New Zealand, Norway, Portugal, the Republic of Macedonia, Romania, and South Africa. Minor parties are a viable voting choice in these countries, resulting in multiparty elections. Despite this, there is little work on polling accuracy for multiparty elections. Political polling usually has three purposes: to forecast the outcome of an election; to understand voter behavior; and to inform campaign strategy (Hillygus, 2011). In multiparty elections, these imperatives remain, but with complications. The forecasting problem includes whether small parties will gain representation at all. Party supporters may engage in strategic voting to strengthen coalition partners (Meffert & Gschwend, 2011). If a party lies below the proportional representation threshold, supporters may change their vote to avoid wasting their ballot (Holtz-Bacha, 2012). Campaign strategy may take account of postelection coalitions; strategic voting may be encouraged to help a coalition partner above the threshold. Alternatively, campaigning against a minor party may drive it below the threshold, forcing it out of parliament, allowing successful parties to hold more seats. Therefore, to extend work on poll accuracy to multiparty elections, both empirical comparisons and methodological innovations are required. As an empirical base, the present work uses data from New Zealand, a developed western democracy that last held a general election in November 2011. New Zealand has a system of mixed member proportional representation with a threshold of 5% of the vote, resulting in multiparty elections. The present study demonstrates that New Zealand polls have comparable accuracy to those of other developed countries, making it a suitable context for the study of multiparty elections. It then extends formal accuracy analysis to minor parties, including very small parties. Finally, it develops a new method, using odds, to translate poll outcomes into forecasts that a party will cross the proportional representation threshold. Adapted from the source document.