their debt management practices - other aims conflict with the ability to deliver fiscal insurance.
By focusing on the case of full commitment we also abstract from the issues analysed in Lucas
and Stokey (1983) of how debt management may be needed to resolve time inconsistency issues.
Further reputational issues may mean that governments face a trade off between tax smoothing
considerations and cost minimisation. Once again this may lead to governments achieving limited
fiscal insurance due to the presence of other aims - in this case cost minimisation. A similar
problem arises from our choice of a standard representative agent Real Business cycle model in our
simulations. If in reality the price of risky assets does not reflect agents risk aversion then debt
management will once again require a trade off between cost minimisation and tax smoothing - an
issue our simulations abstract from.
The results of our simulations are shown in Tables 1a-c - the first two tables are for the case
of no capital accumulation and i.i.d and persistent shocks respectively and Table 1c is for the case
with capital and persistent shocks. We show results for both complete and incomplete markets and
quote the average value of each test statistic as well as the 25th and 75th quartile values from the
distribution of the simulated test statistics.
i) The persistence measures Ψ∣,'l and Φ2fe show across all three simulations an excellent ability
to discriminate between complete and incomplete markets. Not only are the mean value of the
statistics distinct between complete and incomplete markets but the quartile values also suggest
that sampling error is not an issue. Ψ∣j∙ takes low values across all three complete market simula-
tions with values of around zero. However the value of Ψ∣j∙ in the incomplete market simulations
depends on the details of the model. Understandably the persistence of debt is a function of the
persistence of shocks in the economy and the persistence of the transmission mechanisms at work.
The normalization that is involved in Ψ-2∕.∙ reduces this issue substantially and leads to less variation
in estimates across the incomplete market scenarios.
ii) Implementing the impact measure, Im, requires identifying fiscal shocks. We do so using
a Cholesky decomposition on a trivariate VAR ordered as the primary deficit∕GDP, GDP growth
and debt∕GDP. Although we report results for this ordering and VAR we also tested our results
were robust to the ordering and the inclusion as a fourth variable the holding period return. The
16