In a previous research report (Cantwell 1988), we obtained variance formulae for the generalized composite estimator applied to surveys using a balanced one-level rotation plan. The Current Population Survey satisfies the requirements of that design. The Survey of Income and Program Participation and the National Crime Survey, however, operate under “multi-level” designs. In each month one of p different groups is interviewed. Respondents then answer questions referring to the previous p months. Currently neither the SIPP nor NCS uses composite estimation to measure characteristics of interest. We derive simple expressions for the variance of the generalized composite estimator of level, change in level, and average level over time under longitudinal multi-level designs. The results apply to a wide range of survey designs.