Strengthen test_sub_effect_independence

[martinholmer]

@MattHJensen, Maybe I don't understand the goal of pull request #37, but it seems to me the code changes in that pull request don't fully achieve the goal. The problem is when a filing unit has nontrivial long-term capital losses. Look at the changes I made in this PR to the test_sub_effect_independence test, which cause the test to fail because of a lack of independence. Am I confused about this? Or do the changes in PR #37 need to be fine-tuned?

[MattHJensen]

@martinholmer, thanks for identifying this. I'll think about how to fine tune the changes from #37.

[martinholmer]

@MattHJensen said in B-R PR #41:

I'll think about how to fine tune the changes from #37.

Isn't it just a matter of the $3000 limit on the amount of long-term capital losses that can be included in AGI?
In other words, maybe we should be subtracting max(-3000., ltcg) from the taxable-income base instead of subtracting ltcg. If you did that wouldn't the strengthened test pass? Another way to put this is that if the record in the strengthened test had ltcg of -3000 (instead of -50000), everything works as expected, right?

[MattHJensen]

That’s the first thing I’m going to try, but I’m worried the convolution of short and long term capital gains when the loss limitation is implemented might complicate things.

[MattHJensen]

@martinholmer, thanks for identifying this problem with #37. I have resolved the test failure on my branch: https://github.com/MattHJensen/Behavioral-Responses/tree/martinholmer-strengthen-test_sub_effect_independence.

Would you prefer that I open a PR to martinholmer:strengthen-test_sub_effect_independence for inclusion in this PR, or open a new PR to PSLmodels:master?

[martinholmer]

@MattHJensen said:

thanks for identifying [a] problem with #37. I have resolved the test failure on my branch: https://github.com/MattHJensen/Behavioral-Responses/tree/martinholmer-strengthen-test_sub_effect_independence.

Would you prefer that I open a PR to martinholmer:strengthen-test_sub_effect_independence for inclusion in this PR, or open a new PR to PSLmodels:master?

A new PSLmodels/Behavioral-Responses pull request makes more sense to me.

[martinholmer]

This strengthening of the independence test is now incorporated in pull request #44.