I work on an empirical study, where I have 20 annual observations, which is less than the required number.
Now if I want to convert my data from annual to quarter by using eviews software. will it be acceptable?
Dear Abdul,
One problem would be how to deal with seasonal variation that is inherent in quarterly data. Another problem is how would make one quarter variation different from another. Firms usually target quarterly returns, see if you can measure variations on that basis. All such unknowns would be shared in the quarterly residual and can violate some regression assumptions.
In principle you can do it (i.e. transform an annual time series into a quarterly one) by assuming fixed values within each quarter. In some cases this can make sense but I would always suggest to adopt temporal disaggregation methods.
You will be doing that at the risk of losing vital information like seasonal variations which would have been eminent in real quarterly data
Like the others, I cannot recommend this mainly if the reason is to increase the number of observations. Let us make clear that the improvement will be fictive. The subject is well treated, see e.g. https://www.youtube.com/watch?v=9v8n9oZTo7w. Before trying, have a look at that paper: Chan, Wai-Sum, Disaggregation of annual time-series data to quarterly figures: A comparative study, Journal of Forecasting Vol. 12, N° 8, (Dec 1993): 677. If you have a quarterly proxy variable, then I think it is possible. here is the answer I gave to a similar question in 2020: "Yes there are methods to convert yearly data to quarterly data provided you have a related quarterly series. A typical old example was GDP for which value added tax (VAT) could be used. I entered your request in Google Scholar and found lots of papers. I will just mention Ginsburgh, Victor A. (1973), A Further Note on the Derivation of Quarterly Figures Consistent with Annual Data, Journal of the Royal Statistical Society. Series C (Applied Statistics), Vol. 22, No. 3 (1973), pp. 368-374. Looking at papers citing it you will find more recent papers. Now, don't expect too much from this. I know that nowadays VAR models are built routinely with a mixture of quarterly, monthly, and even daily data. Good luck"
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs