This paper discusses the empirical strategies employed in the literature to evaluate the effect of institutional features by exploiting cross-country institutional variability with international assessments and focusing on the effect of early tracking on learning inequalities. To control for country-level confounding factors, Hanushek and Woessmann (2006) proposed a simple two-step difference-in-difference strategy using assessments administered at different age/grades. Other scholars extended this approach to analyze the effect of early tracking on learning differentials across social groups, using individual level models pooling together the data from different countries and assessments. However, since test scores delivered by international assessments are not vertically scaled, strategies based on individual-level models may deliver severely biased results. Instead, the scaling problem does not affect the two-step approach.