Tax compliance nudges are used increasingly by governments because of their perceived cost-effectiveness in raising tax revenue. We collect about a thousand treatment effect estimates from 45 randomized controlled trials, and synthesize this rapidly growing literature using meta-analytical methods. We show that interventions pointing to elements of individual tax morale are on average ineffective in curbing tax evasion (when evaluated against a control group of taxpayers receiving neutral communication). In contrast, deterrence nudges – interventions emphasizing traditional determinants of compliance such as audit probabilities and penalty rates – increase compliance. However, their effects are modest in magnitude increasing the probability of compliance by 1.5-2.5 percentage points more than non-deterrence nudges. Our additional results suggest that nudges i) work better on sub-samples of late payers and when delivered in-person, ii) are less effective in the long-run and in lower-income countries, and iii) are somewhat inflated by selective reporting of results.