This paper provides a framework for an empirical analysis of the scope for cost minimization in public debt management. It assumes that a debt manager aims at minimizing the expected cost of government�s debt portfolio for a given level of short term interest rate and subject to a number of risk and market impact constraints. The analysis is applied to the UK government debt over the period April 1985 to March 2000, by simulating �real time� interest costs of alternative portfolios constructed using monthly forecasts of return spreads based on recursive modelling (RM) procedure recently developed by Pesaran and Timmermann (1995, 2000), which limits the extent of data snooping. Statistically significant evidence of predictability of return spreads are provided before the introduction of reforms of the UK debt management system in 1995, although there seems to be little evidence of predictability once the post reform sample is included. Nevertheless, there appears to have been some scope for a small reduction in interest costs over the 1985-2000 period even if portfolio shares and their monthly changes are constrained to lie within historically observed upper