Hacker News Reader

Comment by ein0p | Hacker News Reader

CPI is designed to make the government look good, hence the article.

> CPI is designed to make the government look good, hence the article.

The CPI is run by the Bureau of Labor Statistics:

> The "food price index" evolved into what we now call the Consumer Price Index (CPI). During World War I, rapid increases in prices, particularly in shipbuilding centers, made a more comprehensive index essential for calculating cost-of-living adjustments in wages. Studies of family expenditures were conducted in 92 industrial centers in 1917–19 in order to provide appropriate weighting patterns for the index. Periodic collection of prices was started and, in 1919, the Bureau of Labor Statistics (BLS) began publication of separate consumer price indexes for 32 cities. Regular publication of a national index, the U.S. city average, began in 1921, and indexes were estimated back to 1913.

* https://www.bls.gov/opub/hom/cpi/history.htm

It is one of most examined numbers in statistics, and while the "best" way† to do things is debated, are there any non-tin-foil hat people that think that the numbers are actually wrong given the published methodologies? Entire research projects (which have released source code) have examined the official numbers and found that they track things fairly well:

* https://en.wikipedia.org/wiki/MIT_Billion_Prices_project

† Often no such thing exists, but rather trade-offs in various metrics.

>> CPI is designed to make the government look good, hence the article. >The CPI is run by the Bureau of Labor Statistics: ... https://www.bls.gov/opub/hom/cpi/history.htm

You do notice that your link to back up disagreeing with the assertion of potential government bias is to a .gov domain, surely?

I really like how they put Labor in italics. As if that disputed the fact it's part of the government.

> are there any non-tin-foil hat people that think that the numbers are actually wrong given the published methodologies?

No, but the whole point of the poster above, and of the article we're commenting on, is that the methodologies themselves are wrong (i.e. that they don't measure what people really care about, they measure something that the government thinks will look better than reality).

This is a very common problem in economics: numbers agree very well with the models, but the models themselves are barely applicable to reality.

If the CPI was designed to make the government look good it wouldn’t use a modified Laspeyres index which introduces overestimation bias by assuming substitution does not happen between basket adjustments. It would be trivial to use a different index type which introduces the opposite bias, but they don’t.

> It would be trivial to use a different index type which introduces the opposite bias, but they don’t.

What would be an example(s) of this different type of index? Do you know of any (online?) resources that explain these differences about different types?

The Paasche index would be the alternative to the indexes used today that would tend to understate inflation. It is not commonly used however. If you wanted to be move clever you could use the Fisher index or Marshall–Edgeworth index, which would give lower inflation values than the methods commonly used today while also being easier to justify since you can argue that you believe they’re more accurate than other methods.

I did make a mistake in my comment above - I said that we use a Laspereys index, but we actually use a Lowe index, which is modification to the Laspereys formula.

The “Formal Calculation” section of the Wikipedia article on price indexes describes all of these: https://en.m.wikipedia.org/wiki/Price_index

Reply on news.ycombinator.com