It is well known that Japanese stock price and exchange rate(USD/JPY) is correlation, So I built a VAR model in python about stock data(Tokyo stock price index) and exchange rate(USD/JPY) and I fitted it by AIC.

As a result, a certain year(2017) was calculated that order is zero, only constant value is displayed.

That means current stock price can explain only constant value, not need past price or exchange rate.

That calculation result is some times occurred when economic data analysis or my calculation is wrong?

data17['TOPIX'] = np.log['TOPIX'].diff()
data17['EXCH']  = np.log['EXCH'].diff()

result = VAR(data17).fit(maxlags=20,ic='aic')
Results for equation TOPIX
           coefficient       std. error           t-stat            prob
const         0.000937         0.000568            1.650           0.099

Results for equation EXCH
           coefficient       std. error           t-stat            prob
const        -0.000159         0.000409           -0.388           0.698
  • 1
    $\begingroup$ Well without seeing your code and model specification it’s very hard to say if that’s due to some mistake in code or just simply because using AIC constant model has the best fit. If you are using classic VAR you should also remember that variables should stationarity (and your variables are definitely not) otherwise you will get some roots that are on the unit circle and model will be biased (although that’s not necessary what drives this result). There are so many things that can go wrong here that without seeing model diagnostics and specification you can’t tell... $\endgroup$ – 1muflon1 Dec 1 '19 at 11:06
  • $\begingroup$ Thank you for your kind comment. $\endgroup$ – Eiji Dec 8 '19 at 7:19
  • $\begingroup$ I fixed my question. I took TOPIX data from Wall Street Journal and exchange rate data from investing.com. TOPIX data is Japanese Standard Time, and exchange rate is american Eastern Standard Time I think, so i took a data below and match the standard time. 1)TOPIX is open time data. 2)Exchange rate is close time data. (There are 3 or 4 hours time lag) $\endgroup$ – Eiji Dec 8 '19 at 7:28
  • $\begingroup$ investing.com/currencies/usd-jpy $\endgroup$ – Eiji Dec 8 '19 at 7:34
  • $\begingroup$ quotes.wsj.com/index/JP/TOKYO%20EXCHANGE%20(TOPIX)/… $\endgroup$ – Eiji Dec 8 '19 at 7:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.