The effect is probably real, but the problem is the correlation coefficient probably varies over space and time.

E.g. you probably have a lot more Foursquare users in Silicon Valley than in Smalltown, USA.

Also, Foursquare probably has its popularity change over time. Right now it's still a little new and shiny, but its popularity might change over time.

If it was the only information available, it would be better than nothing. But your proposed method would be extremely noisy and unreliable.