This PR updates the JJZ (2017) Fig 4 plot and the Saez (2001) Fig 2 plots.
In addition, I've found that a quartic spline does a much better job smoothing marginal tax rates than a cubic spline, which misses key variations in the marginal tax rate schedule. A kernel regression is also added as a method to smooth marginal tax rates over the income distribution. A bandwidth around 10000-15000 works well here.
The candidate comparison notebook is updated to include these new plots and mtr smoothing.
This PR updates the JJZ (2017) Fig 4 plot and the Saez (2001) Fig 2 plots.
In addition, I've found that a quartic spline does a much better job smoothing marginal tax rates than a cubic spline, which misses key variations in the marginal tax rate schedule. A kernel regression is also added as a method to smooth marginal tax rates over the income distribution. A bandwidth around 10000-15000 works well here.
The candidate comparison notebook is updated to include these new plots and mtr smoothing.