# Variables with negative values - linearization vs loglinear

**URL:** https://forum.dynare.org/t/variables-with-negative-values-linearization-vs-loglinear/2565
**Category:** Dynare help (legacy posts)
**Created:** [12 June 2012 22:49 UTC](https://forum.dynare.org/t/variables-with-negative-values-linearization-vs-loglinear/2565 "2012-06-12T22:49:51Z")
**Posts on this page:** 2
**Page:** 1

<div class="post-metadata">

### Author: ![AGM](https://forum.dynare.org/letter_avatar_proxy/v4/letter/a/bcef8e/32.png) [@AGM](https://forum.dynare.org/u/AGM)
#### Post date: [12 June 2012 22:49 UTC](https://forum.dynare.org/t/variables-with-negative-values-linearization-vs-loglinear/2565/1 "2012-06-12T22:49:51Z")

</div>

Hi,

I have a question concerning my model and the way we enter it into Dynare.

I loglinearized my model by hand and entered it into Dynare. However, now I read that if a variable can take also negative values, you cannot loglinearize it. And I have such a variable in my model!

Can I simply linearize then the 1 equation defining this variable and leave the rest in the loglinearized form? How will then Dynare interpret it with respect to other equations containing this variable and how should I then interpret the IRF? Because for the rest, these are percentage deviations from ss. And for linear models, simply level deviations from ss? Can I then have both in one code?

Or should I linearize all equations containing this variable? Or maybe the whole model?

I would be thankful for help.

---

<div class="post-metadata">

### Author: ![YHS](https://forum.dynare.org/letter_avatar_proxy/v4/letter/y/59ef9b/32.png) [@YHS](https://forum.dynare.org/u/YHS)
#### Post date: [3 July 2012 17:00 UTC](https://forum.dynare.org/t/variables-with-negative-values-linearization-vs-loglinear/2565/2 "2012-07-03T17:00:08Z")

</div>

Please refer to the following post:

> [@Non-linear vs. exp() vs. log-linearized models](https://forum.dynare.org/t/non-linear-vs-exp-vs-log-linearized-models/2366/2):
>
> Regarding 2: you cannot take logs of variables whose steady state is 0. Hence, putting them in exp() is not possible. The same by the way applies to variables with negative steady states like e.g. the net foreign bond position of an indebted country. Usually you just leave those variables without the exp(), i.e. only linearized. Regarding 3: Log-linearizing the model by hand only shifts the steady state computation problem to setting parameters like Cbar in your example. For those you still hav…