[biren34]

I work with data a lot, and my main issue with hedonic adjustment is that anytime I've ever made any such adjustment to "improve" a metric that wasn't purely mechanical (like a moving average or a ratio to another number), it's just one step away from making the number say anything I want it to.

The entire concept of "fixing" a metric relies on an inhuman level of self-truthfulness to not just lead you down the path of hearing only what you want to hear.

The point of metrics is that they should slap you in the face with objective reality, and improvement is just too subjective a concept.

[SilasX]

Yes, and this confirms one suspicion of them I had, that they only work in one direction. They ever adjust inflation upward when products get worse:

>Interestingly, hedonic adjustments only act as deflators. Say the airline crams another seat in your row, eliminates carry-on bags and otherwise makes your flight less happy and hedonic. Does the hedonically-adjusted price of your airfare increase? Nope.

[dragonwriter]

> Yes, and this confirms one suspicion of them I had, that they only work in one direction. They ever adjust inflation upward when products get worse:

Well, it asserts that but doesn’t provide a basis for the claim, which directly contradicts BLS: “The hedonic quality adjustment method removes any price differential attributed to a change in quality by adding or subtracting the estimated value of that change from the price of the old item.” [0]

Given that the author focus almost the entirety of his piece on hedonic adjustments discussing new car price data as illustration without ever mentioning that this is a CPI component to which the hedonic quality adjustment method is not applied, I would trust its unsourced claims not at all, and be extra sure to double check any sourced claims.

[0] https://www.bls.gov/cpi/quality-adjustment/

[ericd]

Hm, I wonder if anyone’s tried to popularize an alternative inflation measure to CPI that either discounts hedonic adjustments, or tries to be more “honest” about it somehow, perhaps by trying to measure relative user happiness or something (though that’d be super shaky).

[jerf]

Yes: http://www.shadowstats.com/alternate_data/inflation-charts

However, I'm just answering your question, not endorsing that per se. Here's some skepticism, which I am also not endorsing, just balancing and emphasizing my lack of endorsement: https://munknee.com/debunking-validity-shadowstats-inflation...

Personally I suspect the answer is somewhere in between. Arguably it's not even a well-defined question. Clearly something like inflation and deflation can exist, even across large time periods, but there may be no single number that can ever capture it accurately. I am somewhat reminded of the difficulties of trying to fix an absolute reference frame in physics, especially when trying to fix it to locations on Earth over large periods of time.

[ericd]

Thanks for the links! Agreed, it’s extremely hard to do well - the basket of goods that one would consider reasonable doesn’t even stay the same over time, computers get added, gasoline (hopefully) drops out.

[hnzix]

Not CPI, but Western China-watchers (eg Morningstar) use various proxies for China's GDP since they assert that the official numbers are often politicized.

[spaniard89277]

It's a lot of work honestly. I once played with that idea (not for the US) but I have a life and so on...

[em500]

Yup, productionalizing a comprehensive CPI is a boring, thankless and expensive (household surveys, not just webscraping!) undertaking. Which is why we pay a big government department to do it for us.

The sceptical but honest truth-seeker would discover, after months of labour that the official quality-adjusted BLS numbers are reasonably sane, maybe understated by a few tens of pct-points, and the hedonic adjustments will add about 0.5 to 1%, which is probably a reasonable estimate of overall quality improvements. That truth is so mundane that your months of labour will barely net you a Medium blogpost. Or you take a shortcut, just add 1.5 to 2%-points to the BLS numbers and sell it in a newsletter, like shadowstats...

[SilasX]

You think "TVs are 10% of their quality-adjusted price from 1998" is sane?

[labcomputer]

Erm... yes? Am I supposed to think it isn't?

Here's a Best Buy ad from 1998: https://www.chron.com/news/article/Best-Buy-ad-from-1998-sho...

Here's a Best Buy listing from today: https://www.bestbuy.com/site/insignia-32-class-n10-series-le...

In 1998, a 27" CRT TV was $300 in 1998 dollars. Today's 32" LCD TV is $120.

Just in nominal dollars, not adjusted for inflation, the new TV costs 40% as much as the one from 1998. Then the offical CPI says $300 1998 dollars are worth around $500 2021 dollars... so that brings us down to about 24% before making any hedonistic adjustment. [Edit: And, really, we should be comparing a 32" TV from 1998. That's not in the ad, but it is plausible to imagine it costs more than any of the 27" models. Let's call it $450 just to be conservative. That gets you to around 15% w/ CPI and no hedonistic adjustment]

But the new TV has twice as many lines of resolution, a bigger screen, better color, uses a fraction as much power, is lighter and is physically smaller. So the operating cost is lower, you can put it anywhere you want, you have a bigger and nicer picture to look at. I'm sort of baffled how someone could not agree that the newer one is very clearly better.

I suspect that if you offered most people a CRT version of today's TV, nobody would take it at even half the price. That gets you to within spitting distance of 10%.

You can perform the same exercise with other size tubes, but the result is basically the same.

[OscarCunningham]

Look at these TVs that cost $3000 in 1997: https://www.cnet.com/news/are-tvs-really-cheaper-than-ever-w.... You can definitely get a better TV now for $300.

I think the point of this article is that the TV you could buy for $300 in 1997 you can't now buy for $30. So the 10 times increase in quality hasn't allowed a 10 times reduction in cost of living.

[ajuc]

Ask a colleague that has different political beliefs to fix it for you :)

[KirillPanov]

Er, isn't OP's whole point that these numbers and statistics should be politically invariant?

We don't have liberal integrals or conservative arctangents.

[InitialLastName]

I have seen `conservative_arctan` in a codebase before, but it had to do with the speed/precision tradeoff.

[iagovar]

Yes, but hedonic adjustment is basically an attack of the bottom line of inflation as an useful metric IMO.

Inflation is typically used for measuring purchasing power.

IDK much about CPI because I'm not from the US, but in my country the IPC index (which is similar) has a lot of massage really hurts its usefulness.

[gruez]

> Yes, but hedonic adjustment is basically an attack of the bottom line of inflation as an useful metric IMO.

The underlying premise makes sense though. If your country got richer and everyone started buying high end variants of stuff, and the low-end variants got discontinued, is everyone suddenly worse off?

[makomk]

The trouble is that the reality is probably closer to it simply not being financially worthwhile for manufacturers to offer the lower-end variants anymore - in the case of cars, partly because they're legally required to include these features, but also because the incremental cost of adding those new features shrinks. That is, rather than the cost of a 1990s base-model remaining the same and people just getting higher-end models, the cost of a 1990s base model increases closer to that of more feature-filled models to the point it doesn't make sense to make the features optional. If you think about the cost of making, say, a fancy entertainment or sat-nav interface, that's gone down in a way that the cost of huge chunks of carefully machined, stamped and welded metal hasn't.

Also, the definition of a high-end variant changes over time too. For example, today a 52-inch TV fills the same price niche that a 32-inch one did, say, a decade ago but it also fills the same quality niche - you need to buy the 52-incher to get the same kind of audio and picture quality as a 32-inch set from years gone by, and the 32-inch sets have deteriorated in quality in ways that aren't necessarily visible from the obvious spec-sheet numbers.

[ur-whale]

> really hurts its usefulness.

Well, since I don't know what country you hail from, I am going to speculate, but my bet is it hurts its usefulness for you, but very probably not for the government that publishes it and uses it to demonstrate how "effective" their policies have been.