There is consistent evidence for a positive association between education and a wide range of cognitive abilities. In particular, spatial abilities have been shown to be strongly related to academic achievement. However, studying this association is complex as both education and spatial abilities are modulated by multivariate sociodemographic factors, likely to vary across countries. Most previous studies relied on small sample sizes or were restricted to a limited number of countries, thus were unable to control for these covariates. To overcome these limitations, we used a spatial navigation task embedded in a mobile video game. We quantified the wayfinding ability of 397,162 people across 38 countries and showed that on average, education level was positively associated with wayfinding ability. This difference was stronger in older participants and increased with task difficulty. However, the effect of education was different across countries, from near-zero and non-significant in India (Bayes' factor = 0.08, Hedge's = -0.03, 95% CI = [-0.15, 0.08]), to modest and significant in Romania (Bayes' factor = 345.44, Hedge's = 0.15, 95% CI = [0.08, 0.22]). We did not find any relationship between the education effect size of countries and economic indicators such as GDP per capita. Using the 1972 reform increasing the minimum school leaving age in the UK as a natural experiment, we used a regression discontinuity design to show that education has a causal effect on wayfinding ability.