It's not time dependent. If it were time dependent then time would be a covariate, as it's not a covariate you can't say that it's time dependent. Again the time period doesn't matter because he's not using it to conclude that this is the opinion of people now, tomorrow, two weeks ago, at the end of the sample period, at the beginning of the sample period, or anything else BUT the opinion of people during the sample period. It doesn't matter if somebody he sampled changed their mind as he could have randomly selected that person again or somebody else who changed their opinion too. Randomization is necessary, not because we might luck on the perfect sample but because it's required to make rigorour mathematical inference on the data.