The current study examines costs, measured by educational and general (E&G) spending, and cost efficiency at 252 public master's institutions in the United States over a nine-year (2004-2012) period. We use a multi-product quadratic cost function and results from a random-effects model with a first-order autoregressive (AR1) disturbance term to calculate economies of scale with regard to undergraduate enrollment, graduate enrollment, and research. We also employ a slightly modified version of Kumbhakar et al. ("J Prod Anal" 41(2):321-337, 2014) multi-step approach, involving the use of stochastic frontier analysis (SFA) and taking into account spatial interdependency, to decompose cost efficiency into long-term stable (persistent) and short-term (residual) efficiency. The key results of this study include evidence that: (1) regional clustering of costs exists; (2) there are economies of scale in undergraduate education and diseconomies of scale in graduate education; (3) relatively few institutions are cost inefficient; and (4) cost inefficiency tends to be long-term and persistent rather than short-term and residual. This research also identifies public master's institutions that are the most cost efficient. Our inquiry has implications for future research as it points towards specific institutions, which may be engaged in effective practices to keep costs low, for possible follow-up case studies. Going forward, the techniques used in this study could be applied to examine economies of scale and scope as well as cost efficiency among other types of higher education institutions, such as public or private research universities, baccalaureate institutions, and community colleges.