Eccentric infrared photorefraction is an attractive tool for measuring refractive errors of young children and uncooperative subjects, for it allows quick and non-invasive acquisition of data from both eyes simultaneously over a reasonably large dioptric range. Accuracy of refraction in this technique depends on calibration of luminance slope formed across the pupil into diopters (defocus calibration factor). Commercial photorefractors, like the PowerRef 3™ used in this study, employ an universal defocus calibration factor from one population (Caucasian) to convert raw data of all populations. This study reports significantly larger defocus calibration factors of PowerRef 3™ in 132 East Asian, African and Indian eyes, relative to the machine's default calibration (p < 0.001). The calibration slope of 50 Indian eyes was over-estimated by 64 ± 11% (mean ± 95%CI), vis-à-vis, retinoscopy (p < 0.001). The error reduced to ~6-7% upon rescaling the data using a calibration factor specific for Indian eyes or to that individual (p > 0.9, relative to no over-estimation). Our results therefore strongly suggest the use of an ethnicity- or individual-specific defocus calibration factor for accurate estimation of refraction using photorefraction. Inaccurate refraction estimates due to calibration errors will otherwise severely undermine the advantages of this technique.