My key takes on this article are: 1) Gov needed more $$$. They could either reform/progress the state and let the economy generate freely the extra $$$ in 2-5-10 years OR impose a new tax "on those slackers who spend all day in social media". They chose the latter as they can create a narrative to convince the masses on the validity of this new tax. Edit/addition: And the fact that reforming the state requires change…
For well designed, stratified and representative surveys 2700 is actually pretty good sample size and will get you to +/- couple of percent points accuracy. There's a whole well studied sciencee behind "design of experiments" that goes into pretty much any R&D or market research effort. Nothing will save you however if you don't design and administer the research well.
If you don't believe this and know a bit of programming, simply whip up a Monte Carlo program to repeatedly randomly sample 2700 without replacement from a dataset of 10 million a few thousand times, measure and histogram your estimtes and see the range where you end up most of the time. I actually do this when teaching introductory stats to new analysts and market researchers because seeing is the first step to understanding :-)