State Highway Capital Expenditure and the Economic Cycle

Abstract

The conventional wisdom on state and local budgeting suggests that policy makers are quick to cut capital spending during fiscal stress and that budgetary politics may create a bias against long-term capital investment. This study analyzes changes in highway capital spending of U.S. state governments between 1988 and 2000, and shows a contrary result. During economic recessions, states on average slow their capital spending growth, but do not impose significant cuts, and, during economic booms, they are quick to capitalize on the opportunity and raise capital spending significantly. The article draws from the results of the study to discuss implications for the current fiscal crisis and for future research on capital budgeting.

Keywords:

• state budgeting
• highway expenditure
• economic cycle

Notes

2. Lemov, P. Deficit Deluge. Governing. www.governing.com/5deficit.html (accessed May 2002). Please also see the National Council of State Legislators. New National Survey Reports: State Budgets Fall $17.5 Billion Short. www.ncsl.org/programs/press/2002/pr021122.htm (accessed November 2002).

3. The National Association of State Budget Officers. (2002) Update budget shortfalls: Strategies for closing spending.

4. Levine, C., Rubin, I., & Wolohojian, G., eds. (1981). The politics of retrenchment. Newbury Park: Sage. See also, U.S. General Accounting Office. (1991). Distressed communities: Capital investments were postponed in Texas as local economies weakened. Washington, D.C: U.S. General Accounting Office; Wolman, H. (1983). Understanding local government responses to fiscal pressure. Journal of Public Policy, 3 (August), 245–264.

9. All spending and economic data are expressed in constant dollars to control for the effect of price changes. Since changes in the price levels of different “baskets” of goods and services vary significantly over time, different price deflators were used in the analysis. To measure the real changes of capital spending and federal grants for highways, the analysis uses the chain-type price deflator of government investment in structure (1996 = 100), published by the Bureau of Economic Analysis. To measure the real changes of the state economy, the analysis used a different deflator—the implicit price deflator for the gross domestic product (1996 = 100), also published by the Bureau of Economic Analysis. To reflect the real changes in the consumption power over time, the median household income was deflated by the consumer price index for urban residents (CPI-U). Since the original CPI index from the Bureau of Labor Statistics uses 1982–1984 as the base year (i.e., 1982–1984 = 100), the deflator was converted to the 1996 base to ensure consistency in the data. All conversions from current dollars to constant dollars were calculated by the following formula:

10. Bergstrom, T. & Goodman, R. (1973). Private demands for public goods. American Economic Review, 63(3) 280–296. See also, Inman, R. (1978). Testing political economy's “as if” proposition: Is the median income voter really decisive? Public Choice, 33(4): 45–65.

12. Bradford, D., & Oates; W. (1971). Towards a predictive theory of intergovernmental grants. American Economic Review, 61(2), 440–448. See also Gramlich, E. (1977). Intergovernmental grants: A review of the empirical literature. In Oates, W. E., ed. The Political economy of fiscal federalism; Lexington: Lexington.

14. One year of data (year 1987) was lost in the first-stage analysis due to the calculation of spending changes. So there are 12 dummy variables for years 1989–2000, respectively, included in the first-stage of analysis. No dummy variable for 1988 was used to prevent perfect collinearity in the analysis. In the second-stage analysis, 11 dummy variables for years 1990–2000, respectively, were included. An additional year of data was lost due to the fixes for the auto-regressive problems. (Please refer to endnotes 15 and 16 for more details.)

15. The first-order serial correlation of the residuals for each state was analyzed by Durbin-Watson (DW) tests for each state, respectively. Please see Savin, N. E., & White, K. (1977). The Durbin-Watson Test for serial correlation with extreme sample sizes or many regressors. Econometrica, 45(8), 1989–1996. The data in this article, however, present a small problem in using traditional D-W statistics tables because by pooling cross-sectional data together in the regression analysis, I can allow the number of regressors (k) to be greater than the number of years for each state (n), but the lower and upper limits of the DW statistics (DL and DU) are not available in these traditional tables that are not designed for panel data. Hence, I have to limit the number of regressors to be 7 (omitting the dummy variables) to obtain DL and DU. With k = 7 and n = 12 (the first year is omitted to use the lagged residual to calculate the DW statistics), only two states (Delaware and Oregon) show a first-order serial correlation at the 5 percent level. The autocorrelation problems of these states were later corrected by the second stage analysis using the special procedure specified in the article.

16. The procedure to correct for the autocorrelation problem is specified as follows:3

For more details of the procedure, please see Pindyck, R. and Rubinfeld, D. Econometric Models and Economic Forecasts, 4th edition. McGraw-Hill: Boston, 1998.

18. NASBO, 1999, op cit. See also, the National Association of State Budget Officers. Budgeting Processes in the States. The National Association of State Budget Officers: Washington, D.C., 2002.