* maru dubshinki ([EMAIL PROTECTED]) wrote: > Possibly. Was that bias taken into account? If then, I will shut the > heck up about this; if not, I'd suggest that the proper statistical > treatment would be to widen the margin of error. I think
No, if it is a significant effect which lowers the productivity growth, then you don't widen the margin of error. You would lower the productivity growth estimate across the board. Which puts SS in worse financial shape. Unless you are arguing that the countries that were left out actually had higher productivity growth than the ones that were included. Personally, I don't think it is a big effect. The usual way to guard against survivorship bias is to use all relevant data. In this case, you would choose your samples at the beginning of the period, rather than the end of the period. To see how one expects productivity to grow in a country with a well-developed economy and free market, one should start by going back to, say 1900, and identifying all of the countries in existence at that time which had a well developed free market economy. Then include all of them in the model. If you think there is significant survivorship bias in the study I referenced, then you should be able to point to a number of countries with well developed free markets in 1900 that were not included. -- Erik Reuter http://www.erikreuter.net/ _______________________________________________ http://www.mccmedia.com/mailman/listinfo/brin-l
