Comment by azernik | Hacker News Reader

Comment by azernik | Hacker News Reader

Parent Full thread

This illustrates a big problem in econometrics, specifically inflation measurement ("price indexes"): when comparing the prices of equivalent (baskets of) goods and services, "equivalent" gets harder and harder to define over longer time and space intervals.

The classic example is light bulbs [1]. If you use naive comparisons of like goods across years, with weighting by percentage of expenditures, to measure consumer/producer price changes from e.g. 1890-1930, you'll see:

1. decreasing weight in the basket and increasing price of whale-oil lamps over time 2. increasing weight in the basket and decreasing cost of light bulbs over time

But what you won't capture is that these two goods are both providing lighting, and that the switch from one to another provides a sharp drop in the price of light that is not captured by the change in prices of either in isolation.

Here it's the opposite effect, in that looking at Li-Ion battery prices in isolation overestimates the deflationary effect on energy storage prices.

This problem has also popped up in the context of measuring US manufacturing output over time. Specifically, the correction factors used to deal with this are called "price deflators", and their inability to deal with the pricing and business model of computer hardware may have produced an incorrect impression that US manufacturing output is holding steady or increasing, when it's actually decreasing [2]

[1] "Classic" as in used to draw the attention of the field in this important 1998 paper: https://lucept.files.wordpress.com/2014/11/william-nordhaus-...

[2] https://research.upjohn.org/cgi/viewcontent.cgi?referer=&htt...

'Overestimates' because at the start of the time series people were using cheaper alternatives to the very-expensive li-ion batteries.

> But what you won't capture is that these two goods are both providing lighting, and that the switch from one to another provides a sharp drop in the price of light that is not captured by the change in prices of either in isolation.

I wonder if there was any way to compute light emitted per source. Then you could calculate a price per unit of light and whale-oil lamps would be obliterated by incandescent light builds which would in turn be obliterated by LED light bulbs.

There is! That paper I linked does exactly that. It uses light to illustrate the point because it's particularly easy to calculate the comparative amount of utility per unit of cost regardless of technology, and show how it diverges from what naive price indices would show.

It's harder for something like batteries, where there are multiple relevant performance parameters to reduce to one utility heuristic. And good luck when the utility is subjective (clothes), still under study (food and nutrition), or ill-defined (media).

Planet Money had an episode where they used the metric "minutes of light per day's wages". Here's the transcript: https://www.npr.org/transcripts/503769949

But there are qualitative changes that aren't captured by this -- street lights are brighter and whiter now than they were in the 1990s for example, and those were brighter than oil lamps before them.

Note: that's a more accessible podcast/write-up of the paper I linked in my comment as [1]. If people want the denser in (in both good ways and bad) version, look up there.

You might like this as well:

https://www.humanprogress.org/how-the-cost-of-light-has-fall...

Yup! That's a condensed version of a BBC write-up of that first paper I linked. Super influential in the field, but only percolated into popular media in the last decade.

Reply on news.ycombinator.com