A Vector-Autoregression Analysis of State-Government Expenditure
使用向量自回归模型分析美国州政府六类支出间的动态关系，发现支出变化更多受居民收入约束而非预算固定，未发现明显的挤出效应。
The level and the composition of stategovernment expenditure have changed considerably in recent years. Real per capita welfare spending (including Medicaid) has doubled, and corrections spending has risen by 70 percent, while spending on elementary and secondary education and higher education have risen by only 31 percent and 24 percent, respectively. Concerns have been raised that increased spending on Medicaid and corrections have crowded out spending in other areas. Most past work on state spending has focused either on aggregate spending or on spending in one or two categories where crowding-out has been thought to occur. Some studies have examined the determinants of several state-expenditure categories (see e.g., Jane Leuthold, 1988; Richard F. Dye and Therese J. McGuire, 1992). In this paper we examine interactions among categories that encompass the entire state budget for indications of crowding-out among budget categories. To observe patterns of state financial practice that a structural model of disaggregated state spending should explain, we estimate a vector autoregression (VAR) to examine the dynamic interactions among spending categories. This VAR does not allow us to draw causal conclusions about why state spending patterns change, but it does allow us to observe whether changes in one spending category are correlated with changes in other categories, either contemporaneously or over time. Our VAR includes six categories of spending as well as federal aid and state income. The aid and income variables allow us to observe the impact on state spending of changes in available resources. The absence of crowding-out suggests that spending on particular categories is constrained by state residents' resources rather than by a fixed state budget; states shift income from the private sector into the public sector in response to increased demand for public-sector spending. Past studies of interactions among spending in various budget categories find limited, if any, evidence of crowding-out. James Fossett and James Wyckoff (1996) find that growth in Medicaid spending has resulted in lower education spending but that this effect is small relative to the effect of lower own-source revenue stemming from the recession. Michael L. Marlow and Alden S. Shiers (1996) find no evidence that higher corrections spending crowds out education spending. The hypothesis that high spending on municipal services crowds out local spending on schools is tested in Harvey Brazer and McCarty (1987). That study finds that any correlation between municipal and school spending is positive, not negative. Personal income of state citizens, rather than the 2-6 percent of income allocated to municipal spending, is the effective budget constraint. George von Furstenberg et al. (1986) estimate a VAR model of federal revenue and six categories of expenditure. They find that impulses in one spending category do not normally crowd out spending in other categories over time.