impact measure also performs well in discriminating between the complete and incomplete market
case. As shown in Marcet and Scott (2004) in the case of persistent shocks the market value of
debt falls in response to adverse shocks and so Im < 0 whereas under incomplete markets debt rises
so Im > 0. This is confirmed in our simulation results in Tables Ib and c. Table la shows the case
of i.i.d shocks but although the impact effect is not negative it is essentially zero and still offers a
tool to discriminate between complete and incomplete markets.
iii) The correlation statistic pω*jβ* does discriminate successfully between complete and incom-
plete markets within the context of each model but the values of the correlation coefficients depends
strongly on the underlying features of the model. This variation across the model simulations re-
duces its value as a measure of the absolute quality of debt management although its ability to
discriminate between complete and incomplete markets shows its use as a relative measure. There
are also signs that the distribution of the correlation coefficient is quite diffuse under incomplete
markets which again makes inference difficult.
v) The standard deviation measures σω*,∕σω* and ɑʌ^v*∕σω* show a mixed performance.
Overall the indicator σω*.∕ σω* does manage within each simulation to distinguish complete markets
from the incomplete market outcome. However, the differences are often small between complete
and incomplete markets and once again the absolute value of the test statistic varies considerably
across the three models meaning it is unlikely to be a useful statistic when applied to OECD rather
than simulated data. The indicator based on the volatility of debt to deficits, aʌ^^* ∕σω*, also
performs unreliably. In the case of complete markets and persistent shocks the statistic clearly
discriminates between the market settings - however in this case the volatility of debt is huge under
complete markets as debt falls sharply in value in response to an adverse shock. However in the
case of i.i.d shocks the market value of debt is more volatile under incomplete markets. Therefore
without knowing the persistence of the shocks aʌ^j-* ∕σω* cannot be used to distinguish between
complete and incomplete markets.
vi) Our final proposed indicator of debt management was Φj - based on the estimated IRF
of debt to the cumulative IRF for the primary deficit using the same VAR as for Im. Across all
simulations this statistic provides a reliable test for complete versus incomplete markets. This
17