I guess the main reason is that data dumps are sometimes hard to sift through, can become inconsistent (if there are multiple copies) and because from my experience, if you provide an API for something, someone will use it. On the flip side: I've found that often people have always asked "why isn't there an API for that?"

And of course, you can easily get a data dump from this by just cURLing all the resources and storing them :)

You don't really want to have to download potentially gigabytes of data to find a few sets of results!

"potentially gigabytes of data" is now a microSD card smaller than my thumbnail, or alternately the cheapest instance on AWS.

Local caching will be the way to go.

All resources will have a "date_updated" or similar attribute which you can check against.

Ahh, if only HTTP had defined a standard header for checking if the requested content has been modified, and sending a specific status code if that's the case...