Panel data have several advantages over cross-sectional data, for example in controlling for omitted variable bias through the inclusion of time-invariant fixed effects, yet there is a lack of genuine panel data in many countries whereas cross-sections are often available. As an alternative to repeated cross-sectional studies, Deaton (1985) suggested to build pseudo-panels from cross-sectional data. This is done through grouping observations in different cohorts based on time-invariant characteristics. The mean of each variable within cohorts is then treated as a single observation and it allows to track cohorts over time (instead of individuals). This presentation will discuss the interest of such method in empirical studies and show how to implement it through an illustrative application in R and Stata.