How to compare frequency of word use over time between British and American English?

On Google's Ngram viewer you can set the corpus to be American English or British English, and get a graph for each. You can then compare the y-axis values, being careful to note that Google autoscales it.

For example: American English and British English.

You can also download the datasets of each corpus if you'd like to do your own data processing.

• one solution is to compare frequencies in COCA (AmE) to those in BNC (British). You'll have to account for the size of the corpus as the 'denominator' though.

• Google NGrams allows specifying as tags the corpus American or British. For example

  appropriation:eng_us_2012, appropriation:eng_gb_2012

will graph 'appropriation' over time for their American corpus against their British corpus. This isn't terribly recent (there's lots more new functionality there too) but it is slightly more recent than the time of the original question. All the usual caveats about using NGrams still apply (OCR, punctuation, grammar, polysemy, limited text, only written, etc)