A commonly applied policy to India's ongoing depletion of groundwater is feeder separation. Introduced in Gujarat as the Jyotigram Yojana (JGY) scheme, it provides a separate and rationed electricity supply to farmers and an unrationed power supply to non-agricultural users. JGY is claimed to increase groundwater storage. By using Gujarat district-level data from 1996 to 2011 and by separately applying difference-in-differences and Bayesian regressions, we find that groundwater storage has continued to decrease with JGY. We contend that our empirical results show that JGY has been implemented without adequate consideration of (1) a publication bias whereby researchers have a greater likelihood of having their results published if they are statistically significant and show a positive outcome and (2) a 'barrier' effect such that communicating evidence across science and policy divides means that evidence may not be accepted, even when true, and this limits policy advice and options.