Friday 10 August 2018
New election prediction method could reshape the pre-election cycle
A new system of predicting election results, developed at The University of Otago, is proving so accurate developers
hope it will change future political discussion and journalism.
The value-based method of predicting election results developed by Dr Mathew Parackal, Dr Damien Mather and Dr David
Holdsworth – Senior Lecturers at Otago’s Department of Marketing – featured recently in The International Journal of Market Research.
Dr Mather says the method offers advantages to traditional polling, which are often merely a point-in-time snapshot of
how the various parties are performing.
“Along with more stable political poll predictions, this system provides better insights into how parties address their
electorate’s values around and leading up to the elections. For example, voters who scored one on a scale ranging from
zero to two for the value of respect (e.g. child safety, equality and injustices) were substantially less likely to vote
National than any other party (by a factor of over six times)”
The system was crafted ahead of the 2014 General Election using a sample from a consumer panel involving 858 people.
The methodology used open-ended questions that asked for the issues that mattered in the election to the respondents,
and things that made them happy. The responses were analysed using a natural language processing technique to identify
the values relevant to the election.
The terms used in the respondents’ answers were given a weighting and then organised into groups based on societal
value-topics derived from The Laswell and Kaplan Societal Value Framework, named after Harold Laswell and Abraham Kaplan
(pioneers of political science and philosophy in the twentieth century).
Regarding the 2014 General Election, the mean absolute deviation (MAD) from the actual vote was 1.8 per cent, much less
than the margin of error of 3.5 per cent expected from sampling error alone. The methodology was successful in
predicting the outcome for the minor parties with good accuracy, for example, the prediction for the then newly formed
Internet-Mana was only out by about 0.5 per cent.
“Our data showed that the Internet Party-Mana Movement party collaboration was tracking to fail. The values attributed
to these parties revealed their ideologies would not gel, much like iron and clay,” Dr Parackal says.
“The framing-balanced, value-based predictions exhibited reasonable stability, considering they were made six weeks
before Election Day. Thus the values relevant to the voters, and a good prediction of the voting behaviour, became
evident ahead of the official campaign period, which started four weeks before Election Day in New Zealand,” adds Dr
Drs Mather and Parackal believe the value-based prediction method shows promise for improving the quality of political
journalism and public engagement in the period of election campaigns. Particularly this methodology will assist in
focusing public debate more on values that are influential on citizens’ voting decisions. The system could also be used
for Local Government Elections.
The University of Otago team behind the methodology hope to put it into further practice in the lead-in to the next
General Election, in 2020.