Hacker News new | ask | show | jobs
by giantg2 1316 days ago
No historical data is going to tell us the future. But it can give us an idea of the trend, which is useful since in a large system like the economy the trend tends to change relatively slowly as compared to individual compentents.
1 comments
lottin 1316 days ago
A statistic is only useful if you know how to interpret it. Case in point: you can have double digit year-over-year inflation, while simultaneously falling prices for the last 11 months.
link
giantg2 1316 days ago
This doesn't make sense. Falling prices for what? CPI is a measure of cost of a basket of goods. You can't have net aggregate falling prices and a rising CPI unless you are measuring different prices.
link
danielmarkbruce 1316 days ago
CPI up 20% month on month in month 1. CPI down 1% month on month for the remaining 11 months. That would result in what they are saying.
link
giantg2 1316 days ago
Ah, so not mismatched prices/goods, but mismatched temporally.
link
danielmarkbruce 1316 days ago
It's just about how are you measuring the slope? Like, when you first learn derivatives and the math teacher draws a line between two points on a curve, then draws another from points closer together, then draws the tangent, then goes through how you get to a derivative function etc.

The yoy is just using two points very far apart. The mom is using two points closer together. It would be nice but practically impossible to measure it instantaneously, mom is about as good as we get.

link
fragmede 1315 days ago
I can get the prices for goods online, and then get them delivered to my house. Getting the CPI more often seems entirely feasible to me if we really wanted to.
link
dcow 1316 days ago
Which is still stupid to even be arguing about. It’s like people are trying to find a problem with data that is objectively reported by arguing that if you change the x axis you get an entirely different conclusion. No shit… that’s how data works.
link
danielmarkbruce 1316 days ago
No one is suggesting there is a problem with the data. They are discussing how best to interpret it.

And nobody is suggesting changing the x-axis. It's still time. The distinction people are making is which points does one use to calculate the slope? The most recent two? The most recent one and the one from a year ago? What are the implications of each?

link