Compare coefficient across sub-samples in Dynamic Panel Data estimations

[Jan Ohlenbusch]

Hi all,

I have a question regarding the comparison of coefficients across different sub-samples in a DPD model, using xtabond2's Arellano-Bover estimator.

Background
I am studying the impact of network centrality on company performance (using three dependent variables) - and separating the sample into different groups (industries, since very distinct characteristics). However, it would be very interesting to discover if the effects are stronger in one industry than in another (comparing these groups). All models have standardized dependent variables and the same independent variables & instrumented variables.

Question
Is it possible to compare the coefficient of different samples (industries), when using the exact same model specification?
Is there something specific about the model (Arellano Bover estimation) that needs to be considered in terms of the coefficients?

Thanks!

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

I don't use xtabond so I don't help you much. While suest and sureg do this for other estimators, I don't know if they work for xtabond or xtabond2. Note, they're more likely to work with xtabond (or other supplied estimators) than user written xtabond2. One alternative is to use factor variable notation to allow different estimates by industry. In regression, this might look like regress y i.industry##(x) Whether this is works and is legit in xtabond, I don't know.

[Jan Ohlenbusch]

Hi Phil,

Thanks you for your answer. I am new to Statalist, so the advice is appreciated.

As you stated correctly, suest and sureg do not work for xtabond2 - initially, we wanted to prevent the use of industry dummies as factor variables (since many industries), but we are heading in this direction now.

Thanks again!

Jan