Please check the theory presented below the calculator, it explains why such a low number is enough for a confidence level of 95% that the real value is within ±5% of the measured/surveyed value.
Yes, if you had a list of all 8,000,000,000 people in the world and went to visit 385 truly randomly selected from that list and didn't leave until you got an answer, yes the theory says that would be a reasonably trustworthy sample.
But that theory is useless in practice, when the efficacy of the survey is entirely determined by the sampling bias. Here we have a survey of a few thousand people who had internet access and were on whatever websites promoted this survey and chose to click through the survey and had their votes counted by whatever "spam filtering" logic was used, and multiplied by whatever "population correction factor" was selected, and they're trying to somehow pass it off as "85% of People".
> The samples in Argentina, Australia, Belgium, Canada, France, Germany, Great Britain, Hungary, Italy, Japan, the Netherlands, Poland, the Republic of Korea, Spain, Sweden, and the United States can be taken as representative of these countries' general adult population under the age of 75.
> Due to the fact that the vast majority of the data was collected via Ipsos' online panels (India being the only exception, in which 1,800 were interviewed face-to-face and 400 were interviewed online), participation tends to consist of those who have access to the necessary technology. The samples in Brazil, Chile, Colombia, Indonesia, Ireland, Malaysia, Mexico, Morocco, Nigeria, Peru, Singapore, South Africa, Thailand, Turkey and Uganda are more urban, more educated and/or more affluent than the general population. While not nationally representative, the survey results for these countries provide a useful and unique indication of the direction of public opinion.
>
> Weighting has been employed to balance demographics and ensure that the sample's composition reflects that of the adult population according to the most recent census data.
>
> The precision of Ipsos online polls are calculated using a credibility interval with a poll of 1,000 accurate to +/- 3•5 percentage points and of 500 accurate to +/- 5.0 percentage points. For more information on the Ipsos use of credibility intervals, please visit the Ipsos website.
The calculator also assume random selection. Don't trust statistics math too much. Even in acccounting where only arithmetic is being used, with legal auditing done yearly or quarterly basis, fraud is incredibly common (working in corporates for decades, they even termed it as massaging numbees by financial engineers). This kind of surveys without any audit and legal framework backing it, at 385 sampling is just inviting fraud. At least 2 to 3Mil needed and back by local election boards or local enforcement agencies. If this is done even in communist countries like China or Vietnam will still be way more legit than done in California or Oslo.
This calculator says you need a 385 sample size for a billion (or infinite) population: https://www.calculator.net/sample-size-calculator.html?type=...
Please check the theory presented below the calculator, it explains why such a low number is enough for a confidence level of 95% that the real value is within ±5% of the measured/surveyed value.