Post Your Answer

Can anyone suggest some solutions for removing cross sectional dependence in panel data?

3 years ago in Panel Data , Sectioning By Farheen Ahmed

Panel Data Reliability

Hi everyone, I am conducting research in which I came across a few issues in the panel data due to the cross sectional dependence. Moreover, after studying more about it, I came to the point that cross sectional dependence can raise correlated errors (endogeneity).

Using lag term or series can help resolving the issue of time-invariant effect with the help of the first difference. However, I am not so sure about using the same method to solve this problem of cross sectional dependence. Moreover, by using this, it is easier to lose data on a country. In fact choosing that country becomes an even tedious task.

So, can anyone please tell me the possible solutions, methods or literature to choose the right technique.

Thank you in advance.

All Answers (4 Answers In All)

By Kevin Answered 3 years ago

Hi Dear, I would suggest you find the right dummy independent variables first as it can help in solving the issue. Also, do not see it as a problem, the researchers usually apply the IV approach whenever there is an endogeneity issue. You can go through any textbook which includes the econometric topic.

Replied 3 years ago

By Farheen Ahmed

Hi Kevin, I understand what you are trying to imply but this is not the same issue as the fixed effect issue. There can be different sources for endogeneity and how one treats it depends upon those sources. In this case, the source of endogeneity is the regional factors and I am dealing with the potential endogeneity.

By Aaftab Answered 3 years ago

Hi Farheen, I will suggest you to go for the correction of group dependence in case it is the units present in the same cross section which are causing the cross sectional dependence. You can choose a software which provides the standard error correction for the same.

Replied 3 years ago

By Farheen Ahmed

Hi Aaftab, thank you for your suggestion. However, can you please suggest some studies if possible to research more on this area?

Replied 3 years ago

By Aaftab

Although there are many studies available on sources, you can still try going through these papers if you’d like. https://scholar.google.it/scholar?q=correcting+cross+sectional+dependence+in+panel+data&hl=en&as_sdt=0&as_vis=1&oi=scholart#d=gs_qabs&u=%23p%3Dd_4_OCy6B7MJ

Replied 3 years ago

By Farheen Ahmed

Thank you so much for the help.

By Chris Answered 3 years ago

Hello Farheen, there are many approaches which you can use when it comes to cross sectional dependence and endogeneity. For example, you can either choose instrumental variables and the Hausman Taylor approach. However, when it comes to dynamic panel estimators, the impact of cross sectional dependence can be really high. This becomes even more difficult when the cross sectional dependence is noted in the disturbances of the small dynamic panel-data. The generalized method as well as the IV dependent estimation procedures become inconsistent due to the growing N and T remaining fixed. Therefore, either choose the instrumental variable carefully or use another estimator to perform the analysis. Best regards.

By Sandeep Answered 3 years ago

Hi Farheen, in order to choose the right method you can try identifying whether it is spatial patterns or unobserved common factors that are resulting in cross sectional dependence. Based on this, the distinction between the strong and weak cross sectional dependence measures can be made to draw the models effectively. Shedding light on this can also help in changing nature as shown by the cross sectional dependence. After this, the exogeneity of the regressors is estimated under both the weak and strong measures for the fixed T factor and large T factors. Then you may go for the panel unit root test to introduce a simple and new panel unit for the cross sectionally augmented Sargana- Bhargava statistics. I hope it helped. Good luck for your project.

View Related Questions