[Mikeb85]

IMO Card/Krueger's results are flawed for a different reason. They rely too much on real world results meaning that if all the other factors don't stay the same, their conclusions will eventually fail too.

For example, they say immigration doesn't affect wages. In Canada our government and top banks literally all say it does. There's also a wide body of evidence that increased labour supply does put downward pressure on wages (now whether they actually fall or stagnate depends on a combination of other factors).

Card/Krueger's work is like perfectly fitting a regression line to historical data. It has predictive power until it doesn't...

[tomrod]

I'm not sure I've heard people complain that economics got too much data. Fun. To be clear, their model is to explain, not to predict.[0]

Causal inference has progressed a lot since their paper.

[0] https://projecteuclid.org/journals/statistical-science/volum...

[Mikeb85]

Just go to the thread on HN discussing their Nobel Prize. They're most known for their study on Cuban migrants to Miami, where many migrants came and wages didn't fall.

Many, many economists cite this study to say that immigration doesn't reduce wages. And it is true, often. However it's not because an influx of workers doesn't create downward pressure on wages. It's because there's other factors which keep wages rising.

Now in Canada we have a situation where our immigration levels are so high that we don't even pretend it doesn't affect wages anymore.

https://www.baystreet.ca/economiccommentary/3472/Canada-Wage... https://financialpost.com/news/economy/immigrant-influx-is-s... https://www.bloomberg.com/news/articles/2021-11-24/immigrant...

[SantalBlush]

>However it's not because an influx of workers doesn't create downward pressure on wages. It's because there's other factors which keep wages rising.

So one of the main theories about that paper is that the immigrants were in a largely separate labor market in terms of skills/experience from local workers. When they raised the labor supply in their particular market, it increased entreprenurial activity and more labor was hired across many labor markets as a result.

[tomrod]

I got you. Yeah, reduced form models will have the issue of disentangling factors. This is why diff-in-diff and synthetic control models are valuable, but identification assumptions have to be clear.