indexed there seems to be little systematic connection with the degree of Hscal insurance achieved.
It is of course possible that our inability to Hnd a relationship between our measures of Hscal
insurance and debt structure is due to the poor power of the former, although our simulation
results in Section 4 suggest otherwise. With such limited data multicollinearity may be another
reason for weak signiHcance levels although not for the overall explanatory power of the regressions.
Our empirical results point to two facts - the Hrst is that governments have achieved limited Hscal
insurance and the second is that there is little link between variations in acheived Hscal insurance
and the structure of debt issued. The Hrst point could be due to many reasons - for instance
that governments wish to smooth taxes but the existence of incomplete markets and the lack of
contingent debt available means Hscal insurance is hard to achieve. However other possibilities also
exist. For instance, that governments may focus on cost minimisation rather than tax smoothing or
that concerns over time inconsistency and inflation control pin down the optimal debt composition.
Our empirical analysis is unable to determine the relevance of these other aims - all we can conclude
is that potentially pursuing these aims comes at a cost in terms of Hscal insurance.
To better understand our second Hnding - a weak relationship between debt structure and
achieved Hscal insurance - it is useful to reconsider the results of Angeletos (2002) and Buera and
Nicolini (2004). These authors show how even in the case where governments issue risk free bonds
of different maturities it is still possible to achieve the level of Hscal insurance attained by the
complete market outcome by exploiting shifts in the term structure of interest rates. Assume that
Hscal policy follows a two state Markov process and that the government can issue a short and a
long bond. Let the net present value of future primary surpluses be denoted zh when the Markov
process for the deHcit is in a high state and zl when the low state is realized. Bond prices are
denoted p for j = L,H and z=l,2 where 1 denotes a short bond and 2 a long bond. The complete
market outcome is achieved by setting
zh = pf b1 + pf b2
zl = pfbι + p2b2∙
Without loss of generality let N denote the total number of bonds issued by the government
i.e. bɪ + b2 and let bɪ = φb% where φ can be either negative or positive. Using these assumptions
22