Hacker News Reader

Comment by jcranmer | Hacker News Reader

When you have a model that says something should happen under certain conditions, and every practical instantiation of those conditions have failed to cause the predicted effects, perhaps your model is wrong.

I'm reminded of the Phillips curve which relates inflation and unemployment. If you graph these data points for the 1950s and the 1960s, you do get a great correlation. But extend the graph to the 1970s and the 1980s and the data instead looks a lot more like a random scatterplot. Yet the concept is important enough to be covered in Economics 101 textbooks. To me, it seems a case of economists clinging to models that predict the opposite of reality, and it contributes to my general impression that economists prefer mathematical models even if their correlation to reality is poor.

View on HN

What in tiny experiments like the article's? Inflation and even hyperinflation have occurred many times in many places in the last century. In Venezuela the inflation rate is something like 2000 percent. It's not just theory. Even the US had double digit inflation when I was in school. We were taught things about the time value of money (using numbers like 8 percent interest rates) that seem nuts now.

As for interpreting trends unwisely, isn't that what you're doing by noting a lack of inflation and arguing therefore that fundamentals don't matter. Common sense says if you give out more cash, people can bid up prices higher. There are undoubtedly variables not accounted for in all economic models. As the statisticians say, "all models are wrong". But those unknown variables can turn on you too. Every bubble consists of people ignoring fundamentals because they don't fit the recent curve, then getting burned when it crashes.

Or, the model is right, but the system is being gamed in a way that isn't clear to us in the public. Then when we find out what the game is it'll be too late.

Aren't you worried that there's really no model that you can't support with that logic, no matter how dumb it might be?

I'm more worried about the implicit economics textbooks cover how an old model was falsified in the 1970s, therefore we should assume every other economic model is also false argument he was responding to...

Have the metrics used by models changed? E.g. can you hypothetically define away inflation by excluding property prices from the basket?

Yes. It comes with the territory of realizing that there are actors gaming the biggest systems around, and part of the game is building support for those gaming the system, without knowing which segments of our society are participating.

You have to actually examine the specifics of the issue and see whether the measures actually reflect the reality. But in this specific case there's a pretty strong case that what's broken down is not the relationship between how many people are able to find work and the rate of inflation, but the relationship between how many people are able to find work and the "unemployment rate".

Reply on news.ycombinator.com