The Dutch justice department could benefit greatly from the personal digital data from social media since psychological attributes are known to be strong predictors for recidivism, and assessing the risk of recidivism (with the RISc) in the Netherlands is a very cost- and labour intensive process with low prediction power. The research question "How can personal digital data extracted from social media, as an alternative for existing ways to measure personality attributes, efficiently and accurately contribute to determination of the recidivism probability of an individual?" is raised to give the Dutch justice department recommendations on how the assessment of the risk of recidivism can be improved based on the predictability of psychological characteristics from social media data. We performed a meta-analysis to explore (1) the strength of the predictability of social media data of the Big Five personality traits, and (2) how potential moderators influence the accuracy of the prediction. Main findings were the point estimates of the random effects model (Agreeableness 0.26; Extraversion 0.36; Conscientiousness 0.27; Openness 0.30; Neuroticism 0.31 all with ) and the highest significant values from the moderator analysis for Agreeableness , Extraversion , Openness , and Neuroticism for the moderator 'Activity', and for Conscientiousness for the moderator 'Social Media Platform'. This study gives new insights which will help the Dutch justice department make the assessment of recidivism (1) relatively effortless, (2) cheaper, (3) more accurate, and (4) without cognitive bias.