And this right here invalidates the article's conclusion, the simple number of iPhone tweets overwhelms the Android tweets. The map just shows who is more chatty, versus actual income disparity.

As a side note, I do find the map very interesting - it's just the article's conclusion that I'm not convinced of.

But they don't overwhelm the Android tweets everywhere, only in some places.

I agree that it is too big a leap to say that phone choice always indicates wealth, but I do think there is a connection.

No, it supports the article's conclusion. iPhones have far more tweets, yet are concentrated in high-income areas. Androids, with far less tweets, are nevertheless more widely and evenly spread in all areas - and dominant outside of wealthy areas.

The point of the map isn't to show income disparity, it's to show the smartphone brand of tweeters. The fact that it incidentally does show income disparity is what's interesting.

You should double-check the compositing. It looks like the iPhone layer is being drawn directly on top of the Android layer. So there is a good chance that if you reverse the composting order, the map will light up green.

The correct approach would be to decide a winner, or do some blending for every pixel on the map.