Hello,
I want to examine the following topic:
The Causality between Macroeconomic variables and U.S. Stock Market (Returns) – A Vector Autoregressive model approach
First I do not know whether it makes sense to use level data for the macroeconomic variables and - per definition - changes of stock markets (stationary I(0)) in one model? Thus I thought using the stock market index (which has the same order of integration) as such as target variable just because it has the same order of integration - I fear that I need to apply a ARDL model in this case?!
Secondly, in case using stock market index as suitable variable, I would build a VAR modell in differences since there is no cointegration and no stationarity at levels? Is such a model meaningful? I mean one would lose maybe too much information such as the trend. Or would you prefer an ARDL with stock market returns as I(0) and the rest as I(1)? I mean I do not want to manipulate the data to fit my model but the other way around - OF COURSE!
In the VAR model I use the logs of S&P500 index as target variable, and the logs of the interest rates, the core CPI, the Industrial Production Index, the Money supply and exchange rate.
Thank you very much!!
Hi Philipp Mackert ,
It's generally a good idea to use variables that are of the same order of integration in a Vector Autoregressive (VAR) model. If the stock market index is stationary, but the macroeconomic variables are not, you may want to consider differencing them to make them stationary before including them in the VAR model. In this case, you can use a VAR model in differences to estimate the relationships between the variables.
A VAR in differences can be a meaningful model, as it captures the short-term dynamics and fluctuations in the variables. However, it may not capture long-term trends or relationships between the variables, so you would lose some information. An autoregressive distributed lag (ARDL) model, which allows for different orders of integration among the variables, may be a better choice. Using stock market returns as I(0) and the rest as I(1) in ARDL model is also a way to capture the long-term relationships and it's suitable for your purpose since you want to preserve the order of integration information.
Finally, using logs in your model is a good practice, as it helps to stabilize the variance of the variables, which can make the model's estimates more reliable.
Keep in mind that model selection is not a straightforward task and you might need to try multiple models and compare the results, in order to ensure that you are choosing the one that best fits your data and research question.
hello,
since the variables doesn't the same ordre of intergation an ARDL is more suitable, using log is also a good tip and it helps to limit the variance of the variables, but i prefer to keep them in nature since there is a way or a model wich can resolve the problem
- Badges
- Science topic
- Similar topics
- Thesis Research