VALUE AT RISK FOR TURKISH BANKS:
STANDARD MODEL OR TIME-SERIES MODEL?
Oral ERDOŠAN (Istanbul Bilgi University)
Harald SCHMIDBAUER (Istanbul Bilgi University)
The present study focuses on the traditional and alternative, time-series based,
models to correctly assess value at risk of banks in Turkey. The traditional
approach to determine value at risk contained in on- and off-balance-sheet data
is to apply a multiple of the standard deviation of certain time-series observations
to the values in question. The underlying theoretical concept makes the implicit
assumption of a constant standard deviation of certain price changes. The temporal
structure of the time-series is thus neglected. In the present study, we investigate
an approach based on stochastic time-series models, such as ARIMA and GARCH.
These models are able to account for the temporal structure of price changes
in terms of (conditional) expectations, (conditional) standard deviations, and
serial correlations. The major advantage of this approach is that short-term
predictions, as required for assessing value at risk, can be made in the light
of recent information, as provided on and off balance-sheets, thus making the
forecast more elastic, accurate, and up-to-date. Restricting our interest to
three risks (foreign exchange, stock market, and interest rate changes), we
compare different approaches (standard vs. time-series based) to computing value
at risk, using the most recent available bank balance-sheet data for banks in
Turkey.
