MONTREAL -- Quebec Premier François Legault has insisted time and time again that there is no systemic racism in the province.

However, a Léger survey conducted for the Association for Canadian Studies finds that the majority of Quebecers -- and Canadians as a whole -- believe otherwise.

According to the survey, two in three Canadians (67 per cent) believe systemic racism is an accurate way of describing prejudice and discrimination in their province or region.

That number is highest in Atlantic Canada, with 77 per cent in agreeance. In Quebec, that percentage falls to 66 per cent.

The majority of younger people, aged 18 to 24 and 25 to 34, also agree with the statement -- 77 per cent in each age group.

That number falls as ages go up: 70 per cent for 35 to 44-year-olds, 63 per cent for 45 to 54-year-olds and 55 to 64-year-olds and 55 per cent for those aged 65 and older.

Women are more likely to agree with using the term "systemic racism" at 73 per cent, compared to men at 59 per cent.

In addition, people who identify as visible minorities are more likely to agree, at 76 per cent compared to 65 per cent for people who do not identify as a visible minority.

Those percentages are similar when comparing non-francophones (71 per cent) and francophones (65 per cent).

Across the federal political scale, supporters of the Liberal Party and NDP are most likely to agree (77 per cent and 86 per cent respectively), while followers of the Conservatives and the People's Party of Canada do not (51 per cent and 41 per cent respectively.)

The survey also finds that those who strongly disagree with using the term "systemic racism" are also most likely to reject the idea of making Truth and Reconciliation Day a statutory holiday, while supporting newcomer assimilation -- asking immigrants to give up their customs and traditions to "become more like us."

The survey gathered responses from 1,537 Canadians via web panel between Sept. 24 and 26.

There is usually no margin of error associated with a non-probability sample, but for comparative purposes, a probability sample of 1,537 respondents would have a margin of error of ±2.52%, 19 times out of 20.