The true number of hits for the “true number may be higher” are lower than reported.
Mind Hacks (via Cheap Talk and Marginal Revolution) points out news reports often stress when stating an estimated value that the “true number may be higher,” but infrequently that the “true number may be lower.” The primary evidence cited is a comparison of Google search results for the two phrases, with “higher” shown at 20,300 hits and “lower” shown at a mere 3 hits. (I found about 43,000 hits for higher and about 5,300 hits for “lower”, so Google may be targeting results a bit based on search history.)
Ah, but what about the careful journalist who reports that the “true number may be higher or lower“? Searching for “true number may be higher” will inadvertently capture these quotes as well, biasing the results.
However, the “true number may be higher or lower” only produced 412 hits in my Google search, so the bias is small relative to the numbers above. Perhaps there are few careful journalists, or many careful journalists who are afflicted by less discriminating copy editors.
No one, at least as indicated by my Google search results, says the “true number may be lower or higher“. (Of course, sometime soon Google will index this page, and “lower or higher” will gain an entry in that vast catalog.)