I am working on different bond, equities, mutual fund listed in the Stock Exchanges in Bangladesh. By reviewing various literature, VAR model is found useful for forecasting the price/trend of stocks. ARCH models are found very regularly in the volatility of share price research.
I want to learn and apply various techniques/models/procedures in my research to anticipate the trend of economic variables, such as stock price, GDP, Inflation, Money supply etc.
VAR, ARCH, GRACH, ARIMA etc., are few time series models that caan be used depending on the research objectives.
There is huge difference between stock price, and other variables you mentioned. Stock returns are almost unpredictable, whereas other variables are quite predictable. Probable to most common and easiest way to predict them is the VAR model.
If you want just to model the time behavior, Time Series Models (for price returns) are the way to go (ARMA, GBM, GARCH, etc) since each of these models are good to capture specific data patterns (volatility, clustering, seasonality, dependence on past returns, etc).
VaR and CVaR are risk measures, giving you the probability (or conditional probability)for a given data (extreme, usually) event to happen. I suppose you are calling VAR to a method based in the Markowitz MPT approach.
You can also go for econometric models where the price (or price returns, usually) is the dependent variable and macro economic variables can be tested to become the independent (explaining) variables. For example, the spot crude oil price can be, to some extent, explained by the actual world consumption, the actual world production, the transportation costs, the taxes policies, the geopolitical balances, etc.
I study about forecasting stock price ,too.I realized that artificial neural network and nowadays extreme learning machine methods are popular.Can we say these methods are econometric?And econometrics methods are old methods to forecast stock price.Because mostly studies show artificial neural network and extreme learning machine are better than ARMA,VAR about forecasting performance.
