significance of individual coefficients matter, or just joint significance and goodness of fit?
In 1st stage of 2sls, does

significance of individual coefficients matter, or just joint significance and goodness of fit?
All that matter is that the instrument is significant with f>10.
Not quite  your SEs in second stage depend on the Rsquare in first stage.
I have high r^2, but most variables are not individually significant.

If you have a single endogenous variable, then the F stats for IVs should be more than 10.
If you have more than a single endogenous variable, then this rule of thumb may not work and likely F should be even greater.Thank you, fstat is 50+ for both first stage equations.Is r^2 important or only fstat? R^2 is 0.6 in one and 0.8 in other.

If your excludable instrument is not highly significant, with a rule of thumb being a (separate) Ftest on its significance exceeding 10 or 12 (depending on the writer), then you have a weak instrument problem and should probably abandon the strategy (using that instrument anyway).significance of individual coefficients matter, or just joint significance and goodness of fit?
All that matter is that the instrument is significant with f>10.
Not quite  your SEs in second stage depend on the Rsquare in first stage.
I have high r^2, but most variables are not individually significant.
I think a helpful article for you will be Angrist and Krueger's 2001 JEP article on IV. One of the things one learns from reading it closely is that the best instruments come from *indepth familiarity with the institutional details of the thing you're studying*. Instruments only become apparent to someone who is already a master of the thing they're studying. Only then will you find something likely to be strong in the first stage *and* likely to satisfy the exclusion restriction. Most people are regression monkeys, though, and don't bother to master the thing they're studying so never get that far. But you should play the long game.

If your excludable instrument is not highly significant, with a rule of thumb being a (separate) Ftest on its significance exceeding 10 or 12 (depending on the writer), then you have a weak instrument problem and should probably abandon the strategy (using that instrument anyway).
I think a helpful article for you will be Angrist and Krueger's 2001 JEP article on IV. One of the things one learns from reading it closely is that the best instruments come from *indepth familiarity with the institutional details of the thing you're studying*. Instruments only become apparent to someone who is already a master of the thing they're studying. Only then will you find something likely to be strong in the first stage *and* likely to satisfy the exclusion restriction. Most people are regression monkeys, though, and don't bother to master the thing they're studying so never get that far. But you should play the long game.
https://www.aeaweb.org/articles?id=10.1257/jep.15.4.69Thank you very much. I am an expert in the area and am using such instruments that are clearly important exogenous shifters. I think the problem is that there is low variance in the individual shifters bit together they explain the endogenous variables very well.

I haven't seen people talking about R^2. Note that a very high R^2 does not mean much as the reason for high R^2 may be other exogenous variables. If you are looking for whether your IV is weak or not you are interested in measures based only on them such as F stats.
If you have a single endogenous variable, then the F stats for IVs should be more than 10.
If you have more than a single endogenous variable, then this rule of thumb may not work and likely F should be even greater.Thank you, fstat is 50+ for both first stage equations.Is r^2 important or only fstat? R^2 is 0.6 in one and 0.8 in other.

I haven't seen people talking about R^2. Note that a very high R^2 does not mean much as the reason for high R^2 may be other exogenous variables. If you are looking for whether your IV is weak or not you are interested in measures based only on them such as F stats.
If you have a single endogenous variable, then the F stats for IVs should be more than 10.
If you have more than a single endogenous variable, then this rule of thumb may not work and likely F should be even greater.Thank you, fstat is 50+ for both first stage equations.Is r^2 important or only fstat? R^2 is 0.6 in one and 0.8 in other.
If the R^2 is too high, your instrument is no good. One of the criticisms of the Family Ruptures paper was that their supposed innovation, using IV where the previous paper had used OLS, had an R^2 of like .98 in the first stage. At that point, even if you have a good story, you're effectively still doing OLS.

I haven't seen people talking about R^2. Note that a very high R^2 does not mean much as the reason for high R^2 may be other exogenous variables. If you are looking for whether your IV is weak or not you are interested in measures based only on them such as F stats.
If you have a single endogenous variable, then the F stats for IVs should be more than 10.
If you have more than a single endogenous variable, then this rule of thumb may not work and likely F should be even greater.Thank you, fstat is 50+ for both first stage equations.Is r^2 important or only fstat? R^2 is 0.6 in one and 0.8 in other.
If the R^2 is too high, your instrument is no good. One of the criticisms of the Family Ruptures paper was that their supposed innovation, using IV where the previous paper had used OLS, had an R^2 of like .98 in the first stage. At that point, even if you have a good story, you're effectively still doing OLS.
Ahh, yes. I remember reading about this. Thanks!

I haven't seen people talking about R^2. Note that a very high R^2 does not mean much as the reason for high R^2 may be other exogenous variables. If you are looking for whether your IV is weak or not you are interested in measures based only on them such as F stats.
If you have a single endogenous variable, then the F stats for IVs should be more than 10.
If you have more than a single endogenous variable, then this rule of thumb may not work and likely F should be even greater.Thank you, fstat is 50+ for both first stage equations.Is r^2 important or only fstat? R^2 is 0.6 in one and 0.8 in other.
If the R^2 is too high, your instrument is no good. One of the criticisms of the Family Ruptures paper was that their supposed innovation, using IV where the previous paper had used OLS, had an R^2 of like .98 in the first stage. At that point, even if you have a good story, you're effectively still doing OLS.
and they also failed to cite papers that did the exact thing they did. just to remind everyone