Low birth weight is associated with adverse health outcomes. If birth weight records are not available, studies may use recalled birth weight. It is unclear whether this is reliable. We performed a systematic review and meta-analysis of studies comparing recalled with recorded birth weights. We followed the Meta-Analyses of Observational Studies in Epidemiology (MOOSE) statement and Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. We searched MEDLINE, EMBASE and Cumulative Index to Nursing and Allied Health Literature (CINAHL) to May 2015. We included studies that reported recalled birth weight and recorded birth weight. We excluded studies investigating a clinical population. Two reviewers independently reviewed citations, extracted data, assessed risk of bias. Data were pooled in a random effects meta-analysis for correlation and mean difference. In total, 40 studies were eligible for qualitative synthesis (n=78,997 births from 78,196 parents). Agreement between recalled and recorded birth weight was high: pooled estimate of correlation in 23 samples from 19 studies (n=7406) was 0.90 [95% confidence interval (CI) 0.87–0.93]. The difference between recalled and recorded birth weight in 29 samples from 26 studies (n=29,293) was small [range −86–129 g; random effects estimate 1.4 g (95% CI −4.0–6.9 g)]. Studies were heterogeneous, with no evidence for an effect of time since birth, person reporting, recall bias, or birth order. In post-hoc subgroup analysis, recall was higher than recorded birth weight by 80 g (95% CI 57–103 g) in low and middle income countries. In conclusion, there is high agreement between recalled and recorded birth weight. If birth weight is recalled, it is suitable for use in epidemiological studies, at least in high income countries.