How to deal with Omitted Variables

[Vilson Kodra]

Hi everyone

I am trying to run a Linear Regression Model with Panel and Aggregate Data, but i have some problems.

My promlem is as follow: 3 of 4 independent variablen (Xi) omitted because of collinearity, as well as 5 of 6 Instrumental Variables (Zi) omitted because same reason (because of collinearity)
I had two datasets, and first I merged (many to one) Individual data (wages) with aggregate trade data. and when I run a regression almoust all variables are omitted.
in the attachment I uploaded the regression table and data

Can you help me how to deal with this problem?

Thankyou in advance!

[Joao Santos Silva]

Dear Vilson,

I am not the best person to help you with this, but I suppose that your variables are being dropped simply because they do not vary over time and therefore are collinear with the fixed effects. This is a well-known feature of fixed effect estimators; please check any decent textbook on the subject, OK?

Best of luck,

Joao

[Carlo Lazzaro]

Vilson:
as an aside to Joao's helpful insight, I'm also noticing that you have a very low number of observations per group.
Moreover, the F-test at the foot of the output table states that you can reject the null hypothesis that all your fe are indeed zero.
Are you sure that -fe- is the best specification available for your model?
I would recommend you to reconsider your model according to the existing literature in yuou research field, too.

[Vilson Kodra]

Thankyou you Joao and Carlo for your sugestions! Except fixed effects, i have trayed Random Effects and Between Effects too, but I received the same ommitted variables. Even I received omitted variable by runing a normal linear regressen without instrumental variables.

My variables varying over time, but the low number of observations per group can be the main problem. Maybe I should think to reconsider my model.

Thnx. once again!

[Joao Santos Silva]

Vilson,

I have just noticed that all your variables are dummies with a 1,116/915 split. That is, it may well be that all your xs and zs are the same; is this the case?

Joao

[Vilson Kodra]

Hi Joao

Thankyou for your interests to go deeply in my problem. Since months i am at a loss to understand the problem, and to finish the Regression Analysis.

In general I have 3 years aggregate data (3 observation for each variable) as Independent and Instrumental Variables, and 3 years Panel Data (19 903 Observation) as Dependent Variable.

My independent variables (Xi), and Instrumental Variables (Zi) are aggregate trade data, such domestic demand, exports, imports, migration (the logarithmized Values after the first diference). The only dummy on the regression table above is the variable "alter" (age groups as a Individual characteristic).

My dependent variable (Y) is Wages as Panel Data. After a merge data sets, I have calculated first difference and log. values, as are attached above. The aim of this analysis is to find Effects of Trade and Migration on Wages.

The model has worked for one researcher that he analyse 3 years data for one another time period. I am trying to do the same thing but for an another time period!

Vilson

[Jorrit Gosens]

I think Joao points you to the most important issue. It seems all your x and z variables take on only one of two possible values. That would be very strange if they are continuous economic data. If these are transformations of original data you may have made an error in the process?