Starving the Best Revisited

In this article, I revisit the evidence in favor of the fiscal illusion hypothesis, as presented in Young (2009) for the period 1959 Q3–2007Q4. According to Young, the tax and spend variables are integrated of order one and share a long-term linear common trend (cointegration). He attempts to depict asymmetric behavior by incorporating interaction dummies in the standard expenditure regression model. Subsequently, he correctly states that the employed interaction dummies (eight dummies in a univariate model, four of which appear to be redundant) are incapable of capturing the essence of nonlinear dynamic adjustments. Eventually, following the findings presented by Ewing et al. (2006), Young identifies residual thresholds and reestimates the regression model using the TAR (threshold autoregression) and M-TAR (momentum TAR) models. He concludes that while TAR and M-TAR estimate different speed of adjustment, the findings are not favorable to starving the beast. His findings reject the type of tax cuts that are not contingent on spending cuts of the same magnitude. This article empirically demonstrates that the findings presented by Young are heavily plagued by the lack of an “attractor” (i.e., tendency toward budgetary long-run equilibrium) in the residual of the model he estimated within the utilized asymmetric framework. Indeed, the findings reported in Table 4 of his article do not include any evidence of an attractor — nor do they indicate asymmetry or the lack of residual autocorrelation. Consequently, his findings in Table 4 invalidate the results presented in Table 5 and overturn the main contention of his article