Error with importing annual data from Datastream online database
Posted: Tue Aug 10, 2021 12:43 am
Hi,
I'm in the process of trying to download large quantities of annual data from Datastream into eviews using the "fetch" command however I seem to have encountered a problem. When I import the data it appears to have been offset by one year. For example, if I try to import GDP forecasts from the IMF (for example, the code BDWOGDP.C) the observation for 1980 appears as the 1979 observation so the end product is a series running from 1979-2025 as opposed to a series running from 1980-2026 (which I know to be the true timespan of the data).
When I spoke to Datastream, they told me that the issue was that annual series display in their database as the 30th June for each year and so eviews was probably rounding them incorrectly, though I'm unconvinced this is the case and it doesn't explain why this issue hasn't occurred before (I am updating a model that I run every quarter).
Of course, I could write some code to offset the variables by one year but there are a lot of variables to do this for and, as this appears to be a relatively new problem, I'm concerned that it will revert in future and that the code would need to be rewritten to account for this.
I'm currently working with eviews 11 but I know of a colleague running the latest version of eviews 12 who also has this problem (and, similarly, it was not a problem for him before).
If anyone has any ideas on this, they would be very gratefully appreciated.
Thanks!
I'm in the process of trying to download large quantities of annual data from Datastream into eviews using the "fetch" command however I seem to have encountered a problem. When I import the data it appears to have been offset by one year. For example, if I try to import GDP forecasts from the IMF (for example, the code BDWOGDP.C) the observation for 1980 appears as the 1979 observation so the end product is a series running from 1979-2025 as opposed to a series running from 1980-2026 (which I know to be the true timespan of the data).
When I spoke to Datastream, they told me that the issue was that annual series display in their database as the 30th June for each year and so eviews was probably rounding them incorrectly, though I'm unconvinced this is the case and it doesn't explain why this issue hasn't occurred before (I am updating a model that I run every quarter).
Of course, I could write some code to offset the variables by one year but there are a lot of variables to do this for and, as this appears to be a relatively new problem, I'm concerned that it will revert in future and that the code would need to be rewritten to account for this.
I'm currently working with eviews 11 but I know of a colleague running the latest version of eviews 12 who also has this problem (and, similarly, it was not a problem for him before).
If anyone has any ideas on this, they would be very gratefully appreciated.
Thanks!