This paper identifies two problems in Wallis's (1998) analysis of New Deal spending, addresses those problems, and contributes new findings. One problem is the interpretation of 1/population as an apolitical variable. The other is that, although Wallis discusses land area as a criterion in spending formulas, his econometric analysis omits land area. Controlling for land area eliminates Wallis's estimated effects of 1/population. After discussing those problems, this paper presents an empirical relationship sought but not uncovered by previous authors: A few land and income variables can predict the vast majority of the cross-state variation in per capita spending. (C) 2001 Academic Press.