Expenditure forecast bias associated with centrally budgeted expenditures

There is a vast literature on government budget forecasting that shows that in state and local jurisdictions in the United States (and in some other, mostly regional and local, jurisdictions) revenue forecasting is biased to underestimation (Williams & Calabrese, 2016). This practice is not widely observed where jurisdictions expect to continue spending despite the absence of revenue, generally through the use of debt or transfers. Revenue underestimation is sometimes labeled prudent, conservative, or pessimistic in that it reduces the risk of having funds expire before the fiscal year expires. Thus, this practice is risk averse or it reflects the use of an asymmetric loss function, that is, a preference for one type if error (underestimation) over the other (overestimation). Another possible explanation is that underestimating revenue suppresses the response to political pressure to start more programs which risks tax increases in subsequent budget cycles while simultaneously suppressing response to political pressure to reduce taxes because the current level is barely adequate to cover current programs.

Revenue is only one half of the budget formulation. The other half is expenditure. A natural expectation is that the same risk averse behavior with revenue forecasting would be found in the estimation of expenditures. In this mirror image of underestimated revenue, one should see overestimated expenditures.

However, there is a problem. Budgeting involves two-stages. The first stage, budgeting, includes forecasts, estimates and plans. The second stage, appropriating, assigns the expected revenue it to intended uses. An appropriation law authorizes departments and agencies to spend money. Thus, Williams and Calabrese (2016) suggest that budgets will not include overestimated expenditures because the resultant appropriation would effectively grant excess discretion.

Yet, governments can over-forecast some expenditures, those that are appropriated to centrally controlled accounts. This category may include such items as pension or health contributions, funds for pay increases, and costs of the central operation of government. This study examines the budget-to-actual variance of centrally controlled expenditures as compared with the variance of those for a sample of non-central expenditures. The expectation is that centrally controlled budget-to-actual variance will leave larger surpluses than non-central expenditures, which would comprise significant evidence of previously unexamined risk-averse forecast bias and evidence of concern to suppress agency discretion.

Sources:

Williams, D. W., & Calabrese, T. (2016). The Status of Budget Forecasting. Journal of Public and Nonprofit Affairs, 2(2), 127-160. doi:http://dx.doi.org/10.20899/jpna.2.2.127-160