Pressure to reinstate voting rights for some prisoners is growing, with those calling for change accusing the Government of dragging its heels.
It comes as a slim majority of New Zealanders believe some prisoners should be allowed to vote, the latest 1 NEWS Colmar Brunton poll found.
Prisoners are not allowed to vote in New Zealand, after a 2010 policy change removed the right for any prisoner with less than three years of their sentence to vote.
The poll found 26 per cent of people believed all prisoners should vote and 28 per cent wanted just prisoners serving sentences with three years or less to be able to vote - pulling total support for sentences three years or less to 53 per cent.
Forty-four per cent were against any prisoner voting.
Those who were more likely than average to believe all prisoners should vote were Green Party supporters, men aged 18-34, Asian New Zealanders, Māori and Labour supporters.
Those who were more likely than average to think only prisoners with sentences of three years or less should vote were people aged 18-29.
The groups that were more likely to be against any prisoners voting were National supporters, people aged over 55 and New Zealand Europeans.
Advocate Arthur Taylor said politicians "seem to have some idea that this will lose them votes - it won't lose them votes at all".
Justice Minister Andrew Little said there was "certainly a majority support for re-instating the situation as it was pre-2010".
He called the 1 NEWS Colmar Brunton poll an indicator and another factor Cabinet "ought to take into account".
A report on the issue will go to Cabinet in the next few weeks.
NZ First's longstanding position was that they do not believe prisoners should have the right to vote.
"I suppose the most fundamental question - why do you want to make the laws if you're breaking the laws?" leader Winston Peters said.
In November last year, the Supreme Court upheld the High Court's finding the blanket ban was inconsistent with the Bill of Rights.
Between October 5 to 9, 1008 eligible voters were polled by landline (502) and mobile phone (506). The maximum sampling error is approximately ±3.1%-points at the 95% confidence level.