Longitudinal administrative data generated by public education systems in the United States have become increasingly available and have tremendous potential to inform policy research. While it is tempting to take the accuracy of administrative data for granted, there are reasons why researchers should approach these data with a degree of caution. We illustrate this point using detailed teacher compensation records from Washington State to examine the issue of salary spiking in public pensions, where end-of-career increases in salary are used to boost the value of retirement benefits.