A re‐examination of Wagner's Law using US total state and local expenditure and its sub‐categories

Purpose – This paper seeks to examine the validity of Wagner's Law using annual data (1957‐2006) for the US state‐local government (SLG) real expenditure and eight of its sub‐categories. Design/methodology/approach – The co‐integration tests of Johansen and the bounds testing approach to co‐integration proposed by Pesaran et al. were carried out to determine whether a long‐term equilibrium relationship existed between real per capita GDP (pcgdp) and the expenditure variables scaled by real GDP. The income elasticity coefficients of the expenditure variables were then estimated. The direction of causality was tested in the context of error‐correction models (ECM) and the Toda‐Yamamoto approach, which allows for estimating level relationships without pre‐testing for unit roots. Findings – Most SLG expenditure variables were found to be non‐stationary and income‐elastic. However, with the exception of total expenditure (te), insurance trust benefits (ins) and social services and income maintenance (ssim), no other non‐stationary expenditure variable was co‐integrated with pcgdp and error‐corrected over time. The ECM results suggested that te, ins and ssim were driven by pcgdp, consistent with a Wagnerian causal ordering. The Toda‐Yamamoto approach, however, indicated that in these and a few other cases the causal effect was bidirectional. Originality/value – This paper provides a fairly comprehensive test of Wagner's Law at the US sub‐national government level with an emphasis on the concepts of co‐integration and (long‐run) causality in the income‐expenditure nexus. Its findings underscore the importance of using disaggregated expenditure measures to test Wagner's Law, as they suggest that some, but not all, rapidly growing and non‐stationary expenditure sub‐categories were decoupled from pcgdp in the long run.