I've read the post, and I've skimmed the paper on which it is based. It's an interesting premise: if you know which census tract a person lives in, you can identify them down to about 1:1500; if you also know which census tract they work in, you can up your percentages to about 1:20.
But I'm missing the threat. What's an example of where this particular knowledge (feel free to assume for sake of argument that the pair is actually unique) leads to either personal disaster or a chilling effect? I presume it exists, but I'm not seeing how it compares to something like showing your ID (with address) to use a credit card.
Part of the point of the paper is to show, yet again, that data that could be considered sufficiently anonymized by stripping IP addresses names etc. is not anonymous enough.
Your credit card information identifies you, but you generally don't give your credit card information to any location based service you might want to use.
Additionally, there's still this thing called cash, or visiting places that don't require exchanging money.
A location profile could expose information like your doctors, what restaurants/bars/clubs you like, where you shop, who your friends are, what political or religious places/events you attend... do I need to go on?
But I'm missing the threat. What's an example of where this particular knowledge (feel free to assume for sake of argument that the pair is actually unique) leads to either personal disaster or a chilling effect? I presume it exists, but I'm not seeing how it compares to something like showing your ID (with address) to use a credit card.