Inference on Stock Performance under Monte ... - 3rd SAICON 2011
Inference on Stock Performance under Monte ... - 3rd SAICON 2011
Inference on Stock Performance under Monte ... - 3rd SAICON 2011
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Christoffersen and Jacobs (2004) compares a range of financial models al<strong>on</strong>g a<br />
different dimensi<strong>on</strong>, using opti<strong>on</strong> prices and returns <strong>under</strong> the risk-neutral as well as the<br />
physical probability measure. By evaluating an objective functi<strong>on</strong> based <strong>on</strong> opti<strong>on</strong> prices<br />
they judge the relative performance of various models. They find that opti<strong>on</strong>-based<br />
objective functi<strong>on</strong> favors a relatively parsim<strong>on</strong>ious model. Their analysis favors a model<br />
that, besides volatility clustering, <strong>on</strong>ly allows for a standard leverage effect, when<br />
evaluated out-of-sample through simulati<strong>on</strong> process.<br />
Gabriel (2003) compares the relative performance of several tests for the null<br />
hypothesis of co-integrati<strong>on</strong>, in terms of power and size in restricted samples, generally<br />
carried out by using MSC for a range of likely data-generating processes. Authors also<br />
examine the shock <strong>on</strong> size and power of selecting various procedures to estimate the<br />
variance of the error terms. Their study found that the parametrically tuned tests are the<br />
most well-balanced <strong>on</strong>e characteristically as they display relatively smaller distorti<strong>on</strong>s in<br />
the l<strong>on</strong>g run situati<strong>on</strong>.<br />
Chou (1988) investigated the issues of volatility persistence and the changing risk<br />
premium in the stock market. Data cannot reject the n<strong>on</strong>-stati<strong>on</strong>ary volatility process<br />
specificati<strong>on</strong> because the persistence of shocks to the stock return was so high. The<br />
parameter estimates and the n<strong>on</strong>- stati<strong>on</strong>ary test are both robust to changes in the frequency<br />
of data measurements.<br />
Francq and Zakoian (2000) detected a sequential correlati<strong>on</strong> in the squared<br />
regressi<strong>on</strong> errors. This can be awkward because such autocorrelati<strong>on</strong> structures are<br />
compani<strong>on</strong>able with severe misspecificati<strong>on</strong>s. Standard (quasi-) maximum likelihood<br />
procedures can be inc<strong>on</strong>sistent if the c<strong>on</strong>diti<strong>on</strong>al first two moments are unspecified. To<br />
assuage these troubles of potential misspecificati<strong>on</strong>, they deem weak representati<strong>on</strong>s<br />
characterized by the squared error terms. The weak representati<strong>on</strong> eliminates the need for<br />
correct requirement of the first two unc<strong>on</strong>firmed moments. However, using c<strong>on</strong>fidence<br />
intervals based <strong>on</strong> str<strong>on</strong>g assumpti<strong>on</strong>s can be ambiguous and the need of models with<br />
simulati<strong>on</strong> like M<strong>on</strong>te Carlo’s arrives.