Joint evaluation of the directional accuracy of federal budget forecasts

Abstract

Most research focused on deficit, revenue and outlay in budget forecasts has addressed these issues separately. In this study, we investigate changes in budget forecasts from the Congressional Budget Office, applying a recently developed market-timing test. We find that the combined forecasts of revenue and outlay are useful with a horizon of 8 months in predicting an increase/decrease, and useful with a horizon of up to 20 months in predicting an acceleration/deceleration in the deficit.

Keywords:

• government forecasts
• market-timing test
• directional analysis
• forecast evaluation
• nonparametric methods

JEL Classification:

• C53
• H62

Notes

¹ See, for example, Fritsche et al. (2013), Leitch and Tanner (1991) and Schnader and Stekler (1990).

² The CBO plays an important role in the federal budget process. See, for instance, White (2009) for further discussion. Most reports have focused on the Office of Management and Budget’s forecasts (e.g. Campbell and Ghysels, 1995).

³ Not reported in this article: the budget forecasts from 2-year-ahead to 5-year-ahead are examined; however, We find evidence that none of those forecasts are useful.

⁴ Note that deficit is represented as a negative value. Therefore, a negative value on an increase/decrease and acceleration/deceleration of deficit imply an increase and acceleration, respectively, in contrast to revenue and outlay.

⁵ See Sinclair et al. (2010) for details and remarks on the tests statistics in the 4 × 4 case.

⁶ See Chou and Chu (2011) for remarks.

⁷ In the 4 × 4 case for the FE test, a hypergeometric distribution is used to directly calculate the probability of independence. Thus, only p-values are reported. Note that p-values in the test cannot be calculated because at least one column or row in the contingency tables contains only zeros.