by From Gerald Holtham Econometrics like more casual empiricism can be done well or badly, intelligently or stupidly, dogmatically or with an open mind. But are these gentlemen saying that statistical analysis can never reveal anything in economics that is not obvious to simple observation? Evidently that is untrue. What is revealed is never a “law” and will obviously be contingent in space and time. That follows from the nature of society and economic data. Statistical analysis nevertheless has an indispensable place in social studies whether economics, psychology, medicine or sociology. Econometrica may be boring and contain articles of no evident application but there is a sound reason for the growth of statistical theory. Physicists do not have to contend with lousy data. They can set
Topics:
Editor considers the following as important: Uncategorized
This could be interesting, too:
Editor writes Water Flowing Upwards: Net financial flows from developing countries
John Quiggin writes Machines and tools
Eric Kramer writes An economic analysis of presidential immunity
Angry Bear writes Protesting Now and in the Sixties and Seventies
from Gerald Holtham
Econometrics like more casual empiricism can be done well or badly, intelligently or stupidly, dogmatically or with an open mind. But are these gentlemen saying that statistical analysis can never reveal anything in economics that is not obvious to simple observation? Evidently that is untrue. What is revealed is never a “law” and will obviously be contingent in space and time. That follows from the nature of society and economic data. Statistical analysis nevertheless has an indispensable place in social studies whether economics, psychology, medicine or sociology.
Econometrica may be boring and contain articles of no evident application but there is a sound reason for the growth of statistical theory. Physicists do not have to contend with lousy data. They can set up a controlled experiment and repeat it to generate as much data as they need. If the data are well conditioned simple statistics is good enough. Economists do not have that luxury. Economic data contains measurement errors, multicollinearity, erratic distributions and a host of other issues. Econometric theory is concerned with how to extract information as efficiently as possible from a sparse and noisy data set. Economists use more fancy statistical methods than physicists not out of perversity but because their data is lousier. Statistical sophistication cannot make up for lousy data and you cannot extract information that isn’t there but that is no reason for using simple statistics that you know are inappropriate, ignoring problem of which you are aware.