Quick Question About Checking For Seasonality.

[Mitchell Grant]

Hello All,

In my current project, I am working with a financial time-series process which, upon plotting it over time, appears to follow an eight to ten year seasonal trend. Now, I am doubtful of this as the spikes in my data correlate almost perfectly with the exogenously occurring 1990, 2001, 2008, and 2020 recessions. However, I still want to verify that this is the case.

To this point, I have converted my monthly data to quarterly, plotted an ACF with 80 lags (equivalent of 20 years), and I don't really see any significant spike at lags 40 (equivalent of 10 years) or lags 80. Therefore, I am fairly confident that my data does not follow a seasonal pattern but I was just wondering if there is any other way of double-checking?

I thought of running a simply OLS regression of the type: reg y L4.y L20.y L40.y L60.y L80.y. This way I can see if there is any explanatory power of my y variable 10, 20 years ago on my y variable today?

Then again, I don't even know if that makes sense.

Could someone more versed than me in time-series modelling give me some advice?

Thanks,

Mitchell

[Anton Belgrave]

Hi Mitchell
You can check out the unobserved components model in Stata. The command is ucm. It allows for a breakdown in time series to seasonal, cyclical and trend component. Makes sure you distinguish between seasonal and cyclical trends though. The seasonal - really cyclical - dummy variable you seem to be trying will probably not work well unless the cycles are strong and deterministic, but this is my guess. The link from Boston Colleges Christopher Baum has a good outline in the lower half of the pdf of ucm. http://fmwww.bc.edu/ec-c/s2013/823/E...n11.slides.pdf. Of course you can look up other time series decomposition in Stata for help as well.

[Joro Kolev]

I do not think that what you are describing is seasonality, it is more like some sort of business cycle.

In any case, the traditional way to test for seasonality and to generate and include dummies.

E.g., if you are running a regression on quarterly data and you suspect quarterly seasonality, you include in your regression dummies for quarter 2,3,4, that is, three regressors where each of them is equal to 1 if it is the said quarter, and 0 otherwise.