Chain drift is the difference between the rate of change calculated by chaining an index over a multi period interval and that obtained using endpoints only. More generally, different linking intervals yield different estimates. In this paper, procedures are suggested for evaluating the severity of chain drift and choosing the most accurate estimate of multi period change when chain drift is significant. Applying these procedures to real aggregates in the national income and product accounts of the United States, it is shown that, in most cases, chain drift does not impair the accuracy of the estimates. Occasionally, however, the effects of chain drift are large.