State differences in the provision of AFDC: Concept definition and causal structure

The literature of comparative state studies on the determinants of public policy has focused nearly exclusively on the question of the relative effects of socio-economic and political factors without development of any more general model. This paper moves beyond the economics-politics debate in the study of redistributive policy by focusing on the specification of the dependent variable public policy in empirical research. Public policy is disaggregated into three separate, but related factors: (1) policy intent, (2) policy specification and (3) policy output. The effects of this respecification of policy as a research construct are examined using state differences in the provision of Aid of Families with Dependent Children (AFDC). As hypothesized, this complex measurement model provides a better goodness-of-fit to the data than a single factor policy model which makes no distinction between input and output policy indicators. The significance of this finding is demonstrated by showing that the causal structure exhibited by the economic and political antecedents of public policy will vary systematically if any one of the three policy factors is used alone as the dependent variable. Lastly, several, general models of the determinants of public policy which incorporate not only the familiar socio-economic and political factors but also public demand and the response of the target population are developed through a heuristic, exploratory search process. These data-driven models further emphasize the need to differentiate between policy measures expressed in terms of legislative appropriations and administratively determined policy indicators which describe the level of benefits and access and between these aspects and policy outcomes such as average monthly family benefit. While state political processes shape legislatively determined policy intent, the socio-economic environment and public demand affect policy specification. And both policy intention and policy specification affect policy outcome. Multivariate, linear structural equation models using latent variables are used throughout to express all measurement and causal modes.