With the increased availability of longitudinal data, dynamic panel data models have become commonplace. Moreover, the properties of various estimators of such models are well known. However, we show that these estimators breakdown when the data are irregularly spaced along the time dimension. Unfortunately, this is an increasingly frequent occurrence as many longitudinal surveys are collected at non-uniform intervals and no solution is currently available when time-varying covariates are included in the model. In this paper, we propose several new estimators for dynamic panel data models when data are irregularly spaced and compare their finite sample performance to the naïve application of existing estimators. We illustrate the practical importance of this issue by turning to two applications on early childhood development.