st: FW: clustering of errors with unbalanced clusters

["Alexeev, Michael V." <malexeev@indiana.edu>]

Hello,
 
I am working on the issue of the effect of product market competition on corruption. The basic question is whether firms that face stronger competition in their product markets are more likely to pay bribes to officials. My cross-sectional firm-level data come from the World Bank enterprise surveys and cover about 50 countries and 13 manufacturing industries. I have relatively strong results when I do not cluster errors by industry and/or country and weaker results when I do. Depending on specification, I have from about 250 to about 350 country-industry clusters. I noticed that Nichols and Schaffer pointed out that when cluster sizes are very unbalanced, "the cure can be worse than the disease." (Nichols, Austin, and Mark Schaffer, 2007, "Clustered Errors in Stata," mimeo). My clusters are indeed very unbalanced. Some have fewer than 10 observations and some are over 100. What are the arguments for not clustering errors when clusters are unbalanced and is there any test for!
  when the unbalanced nature of the clusters makes clustering not appropriate?
 
Thanks,
 
Michael
 

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/