How to get good instrument variables?

[Theresia Verena]

Hello,

I have to find an instrument variable that has an effect on the level of work participation, but not on wages. How to test these variables, so that they can be used as instrument variables?

Thank you

[Carlo Lazzaro]

Theresia:
too broad questions are usually left unreplied, due to the lack of useful details (please red and act on the FAQ on how to post effectively. Thanks).
That said:
- what does the literature in your research field suggests when -wage- is an endogenous regressor?
- you can test the correlation of potential instruments with the endogenous regressor.
- you must test whether the selected instruments are not weak ones, otherwise the bias produced by the instrumental-variable regression is more severe than the one of the regression affected by endogeneity.

[Felix Bittmann]

I would argue finding a really good IV is an art and requires a lot of knowledge about the theory but also luck and genius... so I think waiting for general advice here will probably not yield much. I would start with your dataset and check out what is available.