Do download reports reliably measure journal usage? Trusting the fox to count your hens?

Download rates of academic journals have joined citation rates as commonly-used indicators of the value of journal subscriptions.  While citation rates reflect worldwide influence, the value that a single library places on access to a journal is probably more accurately measured by the rate at which it is downloaded by local users.  If local download rates accurately measure local usage, there is a strong case for employing download rates to compare the cost-effectiveness of journals. We examine download data for more than five thousand journals subscribed to by the ten universities in the University of California system. We find that controlling for measured journal characteristics - citation rates, number of articles, and year of download - download rates, as captured by the ratio of downloads to citations, differs substantially between academic disciplines. This suggests that discipline specific adjustments to download rates are needed to construct a reliable tool for estimating local usage. Even after adding academic disciplines to the variables we control for, we find that there remain substantial ``publisher effects'', with some publishers recording significantly more downloads than would be predicted by the characteristics of their journals. While the usage tool can be modified to incorporate the publisher effect, this raises the question of what causes such substantial differences across publishers once journal and discipline characteristics are accounted for