The Bureau of Labor Statistics began conducting the Current Employment Statistics Survey well before probability sampling was accepted as a standard for government surveys. The sample for this survey was recently (1997) converted from a quota sample to a probability sample. The resources for each state's probability sample were fixed to the size of its quota sample. Total resources for data collection are limited to the sum of all states' resources. Since the inception of the probability redesign, some states have experienced a rapid employment increase, causing state sample needs to change. We explored different methods for revising the resources available to each state while keeping the total number of resources fixed. One method involved equalizing relative standard errors (RSEs) between all states. We also reviewed several bounded optimum allocations to produce an estimate of national employment with the smallest sampling error under the given conditions. This paper reviews the methods we investigated and their effects on the national and state RSEs.