If one changes to a voluntary form, given the absence of benchmarks, first, you have no way of confirming the accuracy, and one of the real dilemmas is that if you have bias, sample self-selection bias, increasing the number of people you talk to does nothing to solve those bias problems.
I'll give you a trivial example, having run a survey research company. If I did a national survey of 1,500 people and did it in English only, it would mean that in places such as Chicoutimi or Beauce or Lac Saint-Jean the overwhelming majority of potential respondents could not answer and there would be a significant problem. If I doubled the sample size to 3,000, it would in no way change the bias that existed. Those unilingual francophones would not be included among the respondents and the responses would be highly biased.