Dealing with bad/unobservable data

Sorry, I don’t know if my question goes here or in “Dynare Help”, so I posted in both. Here’s my question, please help, thank you! :

How does one deal with bad/unobservable data? For example, in Smets-Wouters (2003), the authors said “as we do not have good measures of the area-wide capital stock, the value of capital or the rental rate on capital, we will assume these variables are not observed”. Similarly, my problems is that I have a bad real wage data, which I want to treat as unobservable. So how does one goes about that? How does one estimate/do it in dynare? any reference/readings you can recommend on the actual estimation implementation (I know in principle it is possible in Bayesian estimation, but my question goes more on actual (Dynare) estimation, step by step.)

Thank you in advance for any help/recommendation!

Dear Pert,

what if you specify the list of observable variables in the Dynare command “varobs” and exclude the area-wide capital stock and real wage there. Does it work in your case?

Pavel

You can also add a measurement error (possibly autoregressive) on real wage if you believe that the theoretical variable does not match exactly its empirical counterpart. Even if the theoretical real wage is not fully consistent with the observed real wage you can benefit from the use of this information. To do this you just have to add an equation in the model block:

ObservedRealWage = TheoreticalRealWage + MeasurementError ;

where MeasurementError is an exogenous variable, and to declare ObservedRealWage as an observed variable (with varobs).

Best Stéphane.

Great discussion. Even if the theoretical real wage is not fully consistent with the observed real wage you can benefit from the use of this information.

Thanks for sharing useful information.

I’ve been researching this topic for a few hours and found this post pretty helpful. Thanks for compiling the info in one spot.