True, but it is probably easier to create an honest poll than an honest election.
You could sample 10,000 names from taxpayer rolls, driver’s licenses, census rolls, etc.. Then have 3 local law enforcement officers contact each person with an official paper questionnaire, verify the person’s identity, and then ensure that the person has privacy in which to fill out and seal the questionnaire before they return it to polling central.
This would be sufficiently accurate, lots cheaper than elections, less open to fraud, and would suffer from the bias that elections only obtain the opinion of the segment of population that votes.
So many polls use far fewer than this, but they claim they're okay because they "intelligently" create the sample size, i.e. make sure that the male/female ratio, racial percentages, etc. match the overall demographic.