The macro/finance group at NIPFP has been maintaining a website titled Tracking and researching the Indian business cycle at http://www.mayin.org/cycle.in. At present, this has a dataset with seasonally adjusted monthly and quarterly data. This is a big step forward for tracking Indian macroeconomics, since month-on-month changes show what happened in the most recent data while year-on-year changes are (on average) 6 months late by virtue of averaging the latest 12 values of the month-on-month change.
On this website, for each series, there is a technical note showing the work done in seasonal adjustment. There is an RSS feed and an email update. There are downloadable .png / .pdf files which can be used directly, and csv files which can go into your software. There is a report.pdf which is a self-contained document which can go into a printer or an e-book reader, and there is a web page which gives you the picture in a browser. Most important: the website is updated every monday.
With many series, Radhika Pandey (who leads the work) has been finding that the null hypothesis of no seasonality cannot be rejected. Earlier, these series were being kept out of this web page. But it's increasingly clear that the more useful strategy for these series is: to have a technical note for these series as well, so that external users can see the steps that led to this conclusion, and to then proceed with the use of non-seasonally adjusted (NSA) data for the purpose of computing month-on-month changes. Radhika is increasingly moving in this direction.
Ajay, have you guys fixed the Diwali seasonality issue where the month of Diwali keeps moving every year?
ReplyDeleteWhat wuold also be interesting is to show similar graphs for NSA data right below and see if there's substantial different (IIP for instance is a huge difference, WPI maybe not so much)
Yes, we do deal with Diwali seasonality.
ReplyDelete