Research Information System
9th Multinational Energy and Value Conference, Brussels, Belgium, 11 - 12 May 2023, (Unpublished)
There
are several tests to detect structural change at unknown change points. The
Andrews Sup F test (1993) is the most powerful, but it requires the assumption
of homoskedasticity. Ahmed et al. (2017) introduced the Sup MZ test, which
relaxes this assumption and tests for changes in both the coefficients of
regression and variance simultaneously. In this study, we propose a model
update procedure that uses the Sup MZ test to detect structural changes at
unknown change points. We apply this procedure to model the weekly returns of
the Istanbul Stock Exchange's common stock index (BIST 100) for a 21-year
period (2003-2023). Our model consists simply a mean plus noise, with
occasional jumps in the level of mean or variance at unknown times. The goal is
to detect these jumps and update the model accordingly. We also suggest a
trading rule that uses the forecasts from our procedure and compare it to the
buy-and-hold strategy.