We analyze the effect of OARE on (1) scientific output and (2) scientific input as a measure of accessibility in five developing countries. We apply difference-in-difference-in-differences estimation using a balanced panel with 249,000 observations derived from 36,202 journal articles published by authors affiliated with 2,490 research institutions. Our approach allows us to explore effects across scientific fields, i.e. OARE vs. non-OARE fields, within institutions and before and after OARE registration. Variation in online access to scientific literature is exogenous at the level of scientific fields. We provide evidence for a positive marginal effect of online access via OARE on publication output by 29.6% with confidence interval (18.5%, 40.6%) using the most conservative specification. This adds up to 2.07 additional articles due to the OARE program for an average institution publishing 7.0 articles over the observation period. Moreover, we find that OARE membership eases the access to scientific content for researchers in developing countries, leading to an increase in the number of references by 8.4% with confidence interval (5.6%, 11.2%) and the number of OARE references by 14.5% with confidence interval (7.5%, 21.5%). Our results suggest that productive institutions benefit more from OARE and that the least productive institutions barely benefit from registration.