Stressing correlations

Similar arguments apply to correlations as for volatilities. Stressing correlations is
equivalent to undertaking a stress test directly on a portfolio. Consider a portfolio of
Eurobonds hedged with government bonds. Under normal market conditions these
two assets are highly correlated, with correlations typically lying between 0.8 and
0.95. A VaR stress test might involve stressing the correlation between these two
asset classes, by setting the correlations to zero. A more direct way of undertaking
this stress test is to change the price of one of the assets whilst holding the price of
the second asset constant. Again, the question that must be asked is; what is
intended by stressing correlations?
Applying scenarios to VaR input parameters may give some useful insights and an
interesting perspective on where the sources of risk are in a portfolio and equally,
where the sources of diversification are in a portfolio. Nonetheless, stressing a VaR
calculation, by altering volatilities and correlations, is not the most effective or
efficient way of performing stress tests.
Stressing volatilities and correlations in a VaR calculation will establish what the
underlying risk in a portfolio would become if volatilities and correlations were to change to the levels input. Bank management can then be asked whether they would
be happy with regular losses of the level given by the new VaR.