Scott Alexander

Also, even if the CPI figures are right, the composition might still matter.

Broadly, essential goods are more expensive now than in 1970 and consumer goods are much cheaper.

Even if that looks like a wash in CPI figure, it could still mean a loss in well-being for several reasons.

1.

Maybe consumer goods are more likely to be affected by hedonic adaptation or have negative externalities.

By normal CPI measures, a smartphone might be worth a million dollars in 1970.

But what is the actual hedonic effect of smartphones?

Maybe negative.

Having a $1 million smartphone isn't going to make up for not having a $300,000 house.

Two, maybe consumer goods have steeper diminishing marginal utility.

The same way having three right shoes isn't better than one right and one left shoes, being able to afford 20 50-inch TVs isn't as good as being able to afford one TV and a house to put it in.

End quote.

Scott writes, I appreciate your work, but when I try to check ChatGPT's work, I find that its basket of goods assumes the median person buys one house per year.

This means the basket massively overweights the price of housing compared to everything else, and since the price of housing has increased the most, it massively overestimates inflation.

Woolly AI writes, For the first time, this is less than I wanted to know, very specifically regarding inflation measurements and how we adjust for it in wages.

Specifically, I can remember around 2013 when there was a proposal that social security COLA adjustments, annual cost of living increases, be tied to chained CPI rather than CPI.

And the AARP seemed very, very convinced that minor changes to how inflation was calculated would dramatically impact real people's social security checks.

I've seen calculations that chained CPI is 0.25 to 0.3% lower than CPI, which is small overall, but large relative to overall inflation, roughly 2%, and compounding.

I don't want to argue here that the gold bugs are right and the purchasing power of the US dollar has dropped by 80%, but it does seem plausible that different inflation measures, even if both equally valid, could dramatically alter the real median wage over this period.

For example, from Fred's Real Median Household Income series, from 2000 to 2024, income grew from 71,790 to 83,730.