Random effect vs GMM

[Paul Leeds]

I am doing a research which involve a panel data analysis with dependant variable that take the value of 1-100. This value is a result of a readability analysis called "Flesch reading ease" each score represent a reading description " ease/difficult/very difficult, etc"

This table helps to understand the scoring method (the Raw score are actually the dependent variable values and could take any value from 1 to 100):



The main independent variables is service fees (continuous and in logs)
Control variables are: Firm age (logs) + return (ratio %) + industry (sic codes) + std dev of profit + market capitalisation (logs)

Now the issue is that I am in two minds about choosing the right model. The tutor recommends me to apply Generalised method of moments (GMM) panel data since he suspects the presence of individual heterogeneity but I argue that basic random effect would do the trick!
My main contribution is the introduction of another methodological procedures regarding the computation of readability scores and this drove me to neglect applying GMM since I am coming from qualitative research background and learning such a model would be time consuming.
So, will it make any difference?
In other words, can I survive with random effect model and ignoring GMM?

Any posts would be highly appreciated

Paul

[Stephen Jenkins]

You have a combination of a fractional response outcome and panel data (a better title for your post, by the way), which suggests that the following paper is relevant: L Papke and J Wooldridge "Panel data methods for fractional response variables with an application to test pass rates", Journal of Econometrics, 145 (2008), 121-133. Also, have a look at Leslie Papke's home page at http://www.econ.msu.edu/faculty/papke/ : there is a downloadable zip file via the link labelled "Papke and Wooldridge (2008) Stata Files". (I have not used this code.)