I have eleven homogenous firms to investigate
1. The effect of 3 rgressors with 1moderator on their (aggregate) performance over 8-year period (1995 – 2002) and compared to their performance in another 8 year period (2005 – 2012).
My issues:
a. Would I have to estimate a panel data model/equation for each period and compare the results? Or is there a way I can determine the diference in performance using one panel data model?
b. Again I am also worried if 11 firms with 8 year period of observation are sufficient for PhD research. The study is focusing on listed insurance firms in a developing economy. Total listed firms are 26 but data is available for only 11 firms.
Please I need help!!
Thank you
Dear Sunday
a) Two methods. you can use only one model, and use a dummy variable that takes the value of '1' for the 2005-2012 period, and takes the value of '0' for the 1995-2002 period. by using this dummy variable, the coefficient of the dummy variable should tell whether there is a significant effect on performance in the 2005-2012 period compared to the 1995-2002 period.
Alternatively, you can use the same model, and then divide the sample into two sub-samples. the first sub-sample will consist of firms during 1995-2002 while the second sub-sample will consist of firms during 2005-2012
b) with regard to sample size, 11 firms for a phd thesis is just too small. In the UK, a phd student cannot pass a phd with 11 companies sample. High quality journals will not accept a study with 11 companies sample. Normally, for a good Phd, the minimum number of sample is around 200 firms. You can increase your sample size by downloading data for non-insurance firms(e.g. microfinance firms or financial firms) and then compare the performance of insurance companies and the performance of non-insurance companies, in order to increase the sample size
Hope this helps!
Regards
P K Ozili
Dear Peterson,
Thank you.
I agree with your suggestion on methods. Looks Ok!!
But, the sample size is very troubling and i know this pretty well. But when i look at articles from quality journal and past PhD studies in some schools, some use 14 firms (for PhD work) and some within 5-10 firm (in journals) in Panel data studies. These further exacerbated my worries. I am dealing with sector specific study. Even if i would have to extend to non-insurance companies, say all financial companies i cannot still get the minimum of 200 companies, How do i handle this sample size issue please?
First of all, as a journal reviewer, there is a big difference between a quality journal and a "high-quality journal", and I maintain that the latter will not accept studies that have less than 20 firm sample or survey.
Going forward, a sample of 80 firms may be reasonably okay.
However, if you can provide a good justification for why you cannot obtain data for more than 11 insurance companies, then you could pass your viva if your supervisors and external examiners accept your explanation - if its good enough.
One thing that is clear is that larger sample size is always better because it improves the chance of making good generalisation about the population where the sample is obtained from. Try to see if you can obtain data for 80 firms say 11 insurance companies and 69 non-insurance companies.
Regards
Ozili
Ok PKO...he smiles!!!
Thanks for building my confident.
I kindly solicit from others on this platform to help with more suggestions.
- Badges
- Science topic
- Similar topics
- Filing