They address this in the "Limitations" section (don't put a lot of stock in the absolute values - look at relative differences within the sample): One issue to

They address this in the "Limitations" section (don't put a lot of stock in the absolute values - look at relative differences within the sample): One issue to keep in mind with the Sapien Labs dataset is that the participants in each country are not a random or representative sample of the people in that country. Such studies would be extremely expensive to run, and now that so few people agree to phone solicitations or even answer their phones, it is unclear how representative such surveys can be. Those who agree to be interviewed, or who are motivated by money to participate, are not representative of the broader population. For this Sapien Labs report, participants came to the site on their own, or from online advertisements paid for by Sapien Labs, for the purpose of getting a detailed report on their wellbeing. So, the means reported for any country should not be treated like direct measures of the true means. However, samples such as these are still very useful for examining differences within the sample, such as those between men and women, or between those who got a smartphone early and those who got one late. And the much larger size of the Sapien Labs dataset, compared to Gallup and other survey organizations, allows for many additional analyses.