700 people is a good sample size if they are a truly random representative sample of your population. In real life, polling error tends to vary far more than 1/sqrt(n) because of systemic biases in how you select participants. Depending on how the survey was conducted, it could intrinsically favor certain demographics.