# Estimation of variables

I am trying to estimate some variables, but I would like to know if what I am doing is right. I am using some data about labour productivity (in log) and, by using Stata, I filtered (Hodrick-Prescott) the data. Then, I fit a AR(1) process for labour productivity. Therefore, I created a lag variable for labour productivity and then run the following regression:

reg labprod lag(labprod)

Is it okay to write the results of this regression as: $labprod_{t}=coeff*labprod_{t-1}+\varepsilon_{t}$ ?

You can find here (pag.17) the paper with the results by Gali & Monacelli. Did I correctly replicate the results? Do I need to undertake further steps?

The reg command always includes a constant.

$$labprod_t =\beta_0 +\beta_1 labprod_{t-1}+\varepsilon_t$$.

I know nothing about the paper you are citing, but be aware that OLS is biased when estimating AR(1) processes. You can see some discussion of that here.

• Thanks for your help! Do you have any idea how to solve this problem?
– AVR
Jul 14 at 13:27
• It'd help if you provided more details. People don't want to sift through a paper to try to figure it out and interpret it in your context. If all you are trying to do is estimate the AR(1) process with OLS, yes, this is it. If you are trying to do something more complicated, can you describe it in your question rather than referencing us to a paper? Jul 14 at 17:27
• I just want to know if there is an estimator which is not unbiased for AR(1) processes. For instance, I heard that Arellano-Bond could be an option to get this kind of result.
– AVR
Jul 15 at 7:23
• Arellano-Bond is an IV estimator for fixed effects with a lagged dependent variable. It is an IV estimator. IV estimators are in general biased but consistent. Jul 15 at 12:23