# Forecasting variance decomposition vs historical decomposition **URL:** \ **Category:** Identification/Sensitivity\ **Created:** [7 January 2019 14:53 UTC](https://forum.dynare.org/t/forecasting-variance-decomposition-vs-historical-decomposition/13051 "2019-01-07T14:53:08Z")\ **Posts on this page:** 10\ **Page:** 1
**Author:** ![peter](https://forum.dynare.org/letter_avatar_proxy/v4/letter/p/278dde/32.png) [@peter](https://forum.dynare.org/u/peter)\ **Post date:** [7 January 2019 14:53 UTC](https://forum.dynare.org/t/forecasting-variance-decomposition-vs-historical-decomposition/13051/1 "2019-01-07T14:53:08Z")
Hallo everyone I have one question related to forecasting variance and historical decompositions in DSGE model One of the different points btw these two methods is that forecasting variance decomposition is a out-of-sample based technique, whereas historical decomposition is a in-sample technique. So my question is that are the finding from this two different approaches exactly identical to each other? For example, in terms of forecasting variance decomposition, I find that technology is main driver of output fluctuation, then this finding would be identical to that by using historical decomposition? I am looking forward to your answer soon Best ---
**Author:** ![jpfeifer](https://forum.dynare.org/user_avatar/forum.dynare.org/jpfeifer/32/5044_2.png) [@jpfeifer](https://forum.dynare.org/u/jpfeifer)\ **Post date:** [8 January 2019 16:57 UTC](https://forum.dynare.org/t/forecasting-variance-decomposition-vs-historical-decomposition/13051/2 "2019-01-08T16:57:44Z")
The variance decomposition is a theoretical and importantly asymptotic concept. It relies on shocks being orthogonal, i.e. uncorrelated. In contrast, in finite samples even theoretically uncorrelated shocks can show a correlation. For that reason, the historical shock decomposition can yield a different picture. ---
**Author:** ![peter](https://forum.dynare.org/letter_avatar_proxy/v4/letter/p/278dde/32.png) [@peter](https://forum.dynare.org/u/peter)\ **Post date:** [12 January 2019 11:37 UTC](https://forum.dynare.org/t/forecasting-variance-decomposition-vs-historical-decomposition/13051/3 "2019-01-12T11:37:57Z")
Dear Prof. Pfeifer Thank you so much for your interesting explanations So I can say that the finding from forecasting and historical decomposition might be either identical or different. My understanding is correct? Best wishes ---
**Author:** ![jpfeifer](https://forum.dynare.org/user_avatar/forum.dynare.org/jpfeifer/32/5044_2.png) [@jpfeifer](https://forum.dynare.org/u/jpfeifer)\ **Post date:** [13 January 2019 09:01 UTC](https://forum.dynare.org/t/forecasting-variance-decomposition-vs-historical-decomposition/13051/4 "2019-01-13T09:01:29Z")
As I said above: in finite samples that is true. ---
**Author:** ![htroug](https://forum.dynare.org/letter_avatar_proxy/v4/letter/h/90ced4/32.png) [@htroug](https://forum.dynare.org/u/htroug)\ **Post date:** [7 June 2020 02:23 UTC](https://forum.dynare.org/t/forecasting-variance-decomposition-vs-historical-decomposition/13051/5 "2020-06-07T02:23:55Z")
Dear Prof. Pfeifer, in the above mentioned case, which of the above methods should the researcher adopt if their results deviate from each other? ---
**Author:** ![jpfeifer](https://forum.dynare.org/user_avatar/forum.dynare.org/jpfeifer/32/5044_2.png) [@jpfeifer](https://forum.dynare.org/u/jpfeifer)\ **Post date:** [7 June 2020 16:35 UTC](https://forum.dynare.org/t/forecasting-variance-decomposition-vs-historical-decomposition/13051/6 "2020-06-07T16:35:32Z")
I am not sure what you mean. Please elaborate. ---
**Author:** ![htroug](https://forum.dynare.org/letter_avatar_proxy/v4/letter/h/90ced4/32.png) [@htroug](https://forum.dynare.org/u/htroug)\ **Post date:** [7 June 2020 17:05 UTC](https://forum.dynare.org/t/forecasting-variance-decomposition-vs-historical-decomposition/13051/7 "2020-06-07T17:05:20Z")
What I mean is if the results of the two methods are not the same, which one should we adopt for the analysis? ---
**Author:** ![jpfeifer](https://forum.dynare.org/user_avatar/forum.dynare.org/jpfeifer/32/5044_2.png) [@jpfeifer](https://forum.dynare.org/u/jpfeifer)\ **Post date:** [7 June 2020 17:09 UTC](https://forum.dynare.org/t/forecasting-variance-decomposition-vs-historical-decomposition/13051/8 "2020-06-07T17:09:35Z")
That depends if you want to say something about the sample at hand or the population properties. ---
**Author:** ![htroug](https://forum.dynare.org/letter_avatar_proxy/v4/letter/h/90ced4/32.png) [@htroug](https://forum.dynare.org/u/htroug)\ **Post date:** [7 June 2020 17:26 UTC](https://forum.dynare.org/t/forecasting-variance-decomposition-vs-historical-decomposition/13051/9 "2020-06-07T17:26:27Z")
Thanks! Could you please advise some literature that addresses this issue. ---
**Author:** ![jpfeifer](https://forum.dynare.org/user_avatar/forum.dynare.org/jpfeifer/32/5044_2.png) [@jpfeifer](https://forum.dynare.org/u/jpfeifer)\ **Post date:** [8 June 2020 07:08 UTC](https://forum.dynare.org/t/forecasting-variance-decomposition-vs-historical-decomposition/13051/10 "2020-06-08T07:08:14Z")
No, there isn’t because researchers usually don’t point to these conflicts. And you as the researcher on your paper need to decide what you are interested in.