I actually wrote an entire article about this: How reliable is the StackOverflow Developer Survey?
I don’t think the developer survey is statistically valid, at least, not entirely. For one thing, it’s a self-selecting poll. Those are the least reliable.
I would rather see a random poll of 50,000 developers across the entire IT industry. Not just a bunch of zealous web developers who feel it’s their mission to spread the web disease. (Web development is nowhere near representative of the whole industry.)
I also question the way they ask poll questions. They seem to have an arbitrary cut-off at 25 programming languages. Did Smalltalk appear in their question one year and not the next? Why?
In other words, I would take the StackOverflow results with a huge grain of salt. Unfortunately, too many people in our industry take them as gospel.