I am estimating female labour participation rates using panel data for 7 countries. I have data period from 1991 to 2021. I have reviewed the literature, and it suggests using GMM only when you have larger Ns and small Ts. Can you please help in this regard that which advanced or dynamic panel model be used?
Dynamic panel models like Arellano bond (xtabond,xtdpdsys) are used for large N and small T. For small N and large T, you can use FE, or Zellner’s seemingly unrelated regressions (SUR)-with Stata command sureg. or FGLS (xtgls).
Chuck A Arize Thank you very much for your prompt response. Further, I would like to know that Would it be a good procedure;
1. Estimate FE and RE
2. Hausman test
3. Zellner’s seemingly unrelated regression
Would it be the sequence?
For small N and large T panel data, you can use the Fixed Effects (FE) model or the First Difference (FD) model. The FE model is preferred when there is no serial correlation in the idiosyncratic errors. However, if there is serial correlation in the idiosyncratic errors, then the FD model may be more efficient.
There are also other advanced panel data models that can be used for small N and large T data such as the Generalized Method of Moments (GMM) and the Maximum Likelihood (ML) estimator. However, these models require more assumptions and may not be appropriate for small N and large T data.
Can I recommend Chapter 8 (Panel Time Series) In Croissant and Millo (2019), Panel Data Econometric with R, Wiley. This deals with large t and small n and has some good examples and explanations. in particular it deals with heterogenous coefficients, fixed coefficients, random coefficients, Swamy estimator, testing for poolability, cross-section dependence and common factors, non-stationarity and cointegration. If you do not use R it should be possible to implement many of the examples in your software.
This book also contains examples of routines like SUR and routines mentioned by Chuck A Arize
@John C Frain Thank you for the insights and introducing this book. I would definitely see this chapter.
Regards,
In addition to the fixed and random effects methods, another possibility for estimating coefficients in a panel model with a small number of individuals and a large number of periods is to use the instrumental variables method ( IV, instrumental variables)
Chuck A Arize Thank you for your response. After the First Difference model, would you recommend Anderson and Hsiao (1981) method to estimate the dynamic panel model?
Kuessi Prince Houssou Anderson and Hsiao (1981) method use IV for the same.
And kindly let me know if Anderson and Hsiao (1981) would be enough to present our results or I need to explore more.
to my opinion you should use the second generation models like Mean group, pooled mean group or augmented group models. first these models are efficient in presence of cross sectional dependence and second they can give robust results no matter if you have T or N smaller or greater.
- Badges
- Science topic
- Similar topics
- Mathematics