ABSTRACT
The chapter by Barrell, Hurst and Mitchell constitutes an excellent analytical exercise. It was a pleasure to read it. The chapter is a rigorous attempt at gauging the degree of uncertainty associated with the measurement of the cyclically adjusted budget (CAB) balance (and of the underlying output gap) in major EU member economies. Apart from the solution of analytical and computational challenges, my interest in this work is primarily from the standpoint of the policy maker. When reading the chapter I was reminded of the book by former US Treasury Secretary
Rubin (Rubin and Weisberg 2003) about his approach to decision making during his years in government and in investment banking. The essential point was that he took policy decisions or business decisions always with a view to assessing and weighing in real time the uncertainties surrounding those decisions – without the benefit of any formal gauge. In particular, the conduct of fiscal policy for stabilization purposes presupposes real-time estimates of the output gap. At the supranational EU level, output gap estimates are necessary for surveillance to ascertain whether a member country’s underlying fiscal position is in line with the requirements of the Stability and Growth Pact. Beyond the realm of fiscal policy, reliable output gap estimates are also a key ingredient in the formulation of monetary policy, not only by central banks that follow a Taylor rule, but even by those that target inflation alone. Yet the reliability of such estimates is viewed with widespread scepticism – as expressed for example recently by Rogoff (2006, p. 320). Therefore, in this broad context, the experiment reported in this chapter is most welcome from the perspective of multiple users. The chapter contributes to a clearer understanding of the cyclically adjusted budgetary
position in real time. It makes a quantitative assessment of the uncertainty surrounding point estimates of the output gaps, by comparing real-time estimates with full-sample measures of the gap. It applies a model-based approach, drawing on unobserved components. The resulting output gap measures and the respective uncertainty bands are then translated into CAB measures on the basis of stylized budgetary sensitivity parameters. By this measure, the authors find that in Italy, France, Germany and to a lesser extent the UK, the general government accounts have not been close to balance or in surplus position in recent years. On the other hand, for Spain, the Netherlands and Belgium fiscal balance is obtained. These results can be examined on the basis of four criteria: robustness, sensitivity, versatility and usefulness. On robustness, in principle, the calculations seem reasonable, although it is surprising that
the uncertainty bands are rather wide for most countries. Less surprising is that the realtime bands are larger than the full-sample bands. However, on this criterion, the jury is still out until further testing with alternative methodologies (as done to some extent for the
United Kingdom). Furthermore, the capacity of this approach to capture structural shifts is questionable, and more important, there remains an end-point problem. A possible solution of this problem may lie in the application of the trend-cycle filter developed by Mohr (2005) that combines an HP filter with an unobserved-components model-based approach. A closely related criterion is the sensitivity to alternative budgetary parameters. Obviously,
the entire computational chain cannot be stronger than the weakest link. In this regard, the sensitivity test conducted on the budgetary impact of various GDP components is persuasive. And, of course, future application of this approach could be fine-tuned to specific countries and specific periods. As regards versatility, the results for Spain and the Netherlands give rise to some scepti-
cism. It would be worth probing deeper whether the narrow uncertainty bands for Spain are attributable to some peculiar structural characteristic, model specification or virtual absence of statistical error. In the case of the Netherlands the estimates are counter-intuitive in that the real-time bands are somewhat narrower than the full-sample bands. Perhaps this is due to unusually large revisions of the underlying data. Research on the uncertainty of real-time estimates of the output gap should be useful for
policy-making purposes. A retrospective assessment of the cyclically adjusted budgetary position has considerable value added for fiscal policy analysis.2 Specifically, this should provide more reliable information for national policy decisions, as well as for surveillance by EU authorities under the Pact, especially in assessing performance under the excessive deficit procedure. However, a pending question, to be tackled by future researchers, concerns the forward-
looking application of the estimated uncertainty around real-time output gaps. In other words, this would involve inquiring into the extent to and manner in which past estimates of uncertainty bands can be extrapolated to the period ahead with a view to better capturing, and possibly to counteracting, the impact of cyclical fluctuations around trend output. Beyond fiscal policy, as anticipated in my opening remarks, this line of research is potentially useful for monetary policy as well. Estimates of the uncertainty in real-time measures of the output gap should be helpful for the preparation of increasingly accurate growth forecasts, including through fan charts that presuppose a reliable point estimate for the current time period.3
