Statistical methodology can be a dry subject, but this article in today's Globe and Mail on changes to the way in which the Consumer Price Index (CPI) is calculated is a good illustration of how important it can be.
Estimates for the savings generated by correcting the well-known upward biases in CPI inflation are measured in hundreds of millions of dollars. Unfortunately, the Conservatives' decision to make the National Household Survey (NHS) a voluntary survey may take those savings away again. Or increase them – there's no way of knowing.
The "basket" whose price is measured by the CPI is constructed from the Survey of Household Spending (SHS), a voluntary survey. In the past, the problem of non-response was corrected by using the census's benchmarks in order to produce a sample that was representative of the population as a whole.
Economy Lab contributor Kevin Milligan and his UBC colleague David Green provide a very useful summary of the problem in last September's issue of Canadian Public Policy. SHS response rates have been steadily falling over time, so the need for proper benchmarks will become more important – and the NHS will not be able to provide them.
The move to annual CPI baskets means that more is going to be asked of the SHS, even as its ability to answer is steadily eroded. Worse, we won't even know in which direction any given sample may be biased.
In their CPP article, Mr. Green and Mr. Milligan remark on the "hidden ubiquity" of the census. The consequences of the decision to make the NHS voluntary have only begun to manifest themselves.