Syndicate content

Regression

Questioning the External Validity of Regression Estimates: Why they can be less representative than you think.

David McKenzie's picture
A common critique of many impact evaluations, including those using both experimental and quasi-experimental methods, is that of external validity – how well do findings from one setting export to another? This is especially the case for studies done on relatively small samples, although as I have ranted before, there appears to be a double standard in this critique when compared to both other disciplines in economics and to other development literature.

Phil Howard on Information Technology and Political Islam

Anne-Katrin Arnold's picture

Last week I went to listen to a talk by Philip Howard of the University of Washington. He spoke about the "Digital Origins of Democracy: Information Technology and Political Islam." The story was mainly the one we keep hearing about ICT and the Arab Spring, although Howard cautioned that ICT don't actually topple dictators, they rather catch dictators off-guard. And while ICT don't cause political change per se, they provide "capabilities and impose new constraints."

Howard went on to show a table of Arab countries with a few characteristics that may or may not be helpful in predicting future civic unrest. The variables in the table were: country, years of ruler in power, approximate proportion of people connected through ICT, average age of the population, and next elections. This kind of collection of variables is seductive because it seems so easy to use them to predict civic uprisings in the Arab World.