The Personal Responsibility and Work Reconciliation (PRWORA) Act of 1997 marked a significant change in US welfare policy. Under the terms of this legislation, welfare recipients are limited to 24 consecutive months of monetary benefits, not to exceed 60 months over an individual’s lifetime. The primary intent of PRWORA is to force people off of welfare roles and into the work force. However, it may also have created a second effect; namely that wel-fare recipients treat the 60 months of welfare benefits as a stockpile of wealth. If so, recipients might strategically move on and off of the welfare roles in order to increase the length of time before one exhausts his or her total lifetime benefits. This leads to a high number of “welfare spells”, each relatively short in duration. We present an empirical analysis using data from three counties in Washington State to test whether (and how) an individual’s ability to maxim-ize welfare spells varies by their potential employment opportunities and their socio-economic characteristics. We find that socio-economic characteristics such as race and gender, family structure and educational attainment all significantly influence welfare spells. Additionally, welfare spells differed significantly by county, indicating that local labor market conditions specific to those counties are also important in decisions to “bank” welfare benefits.