CPI Bias and Experiment Bias
By David Henderson
As Bryan admits, his experiment to figure out whether the economy is stagnating suffered from a small-sample problem. Various commenters when Bryan first asked for volunteers pointed out that there was huge selection bias. The people who are even aware of the experiment are going to be among those who benefit most from the IT revolution. And, as you could see from the comments, a huge percent of the examples were from the IT revolution.
I think I agree with Bryan that things are getting better quickly. Here’s one of the pieces I wrote on this–in 1997. But I want to point out two hedges on that.
First, Robert Murphy on his blog pointed out a bias in Bryan’s experiment that not one of the commenters noted. Bryan asked:
1. Was my experience during the last hour noticeably better as a result of an innovation introduced from 1990-present? [Yes/No]
2. Was my experience during the last hour noticeably better as a result of an innovation introduced from 1950-1989? [Yes/No]
What do you notice missing? A chance to say that the experience was worse.
Now it’s unlikely, though not impossible, that the experience will be worse as a result of innovation as the term “innovation” is commonly understood. But things could be worse the old-fashioned way: products get worse or merchants sell slightly smaller versions of the product and the small reduction in size doesn’t get picked up by the BLS.
Which brings me to my second point. If you care about the issue of BLS bias, read a comment on Bryan’s first blog that came along well after other people had weighed in. The author, “randy,” works for the BLS and gives us a pithy review saying why he thinks that the CPI is not biased much in either direction. If randy is right, by the way, “Consumer Price Indexes,” the article I commissioned by Michael Boskin for the Concise Encyclopedia of Economics is off.