I receive variants of the above question from prospective job candidates all the time. My totally off the cuff response is usually "aim for 5 articles", based on three years sitting on our (uOttawa) hiring committee and another 8 years of being attentive to the hiring process (and the cvs of candidates). But I have never actually measured it.
In my collection of data for my article on common law prof public engagement (see my prior blog posts on this topic), I have pulled together the publicly available info on Canada's 600 full, associate and assistant common law profs. I have collected data on the publication tempo of these profs, where available (the amount of information varies from website to website). I shall report on that in due course.
But in this, the fourth in my series on "the Canadian common law prof", I can report on a subset of those data. Ten assistant professors post their complete cvs, and from those cvs one can determine the number of publications each had as of the year they were hired. (Admittedly, this is a very small sample size, but it's better than holding the finger up to the wind. I could include other ranks who post cvs, and back out their pubs according to the date of their appointment, but that would take some time and all this is collateral to my article. Moreover, those data would be "stale" reflecting the (probably less competitive) hiring market of bygone years, rather than the more recent markets in which today's Assistant Profs were hired.)
The results: an average of 4.64 journal articles (a median of 3) for the 10 individuals, and two professors had 1 book each when hired (so I suppose that's an average of 0.2 per prof, an obviously meaningless number).
So I suppose I haven't been that far off in telling candidates to shoot for 5 journal articles.
Of course, there are lots of other qualities that go into being competitive in the job market -- but pubs are undoubtedly important in my experience.
Thanks for those providing feedback and asking questions about my dataset on twitter @cforcese. This is turning into a fun article to write (although a gruelling exercise in data collection) and it may, gasp, even be useful.