Hmm.... This is probably a good platform to exchange some ideas...
Unfortunately, the identification strategy in this paper cannot convince me.
Above all, if you read carefully about how the instrument is constructed, you will find that the timing if CSL (Compulsory schooling law) across the provinces are almost consistent. The main variance the author use is the before&after the 1986. However, during this period, there were a bunch of issues happening....
Maybe RD is a good way to address it. The authors may realize it so they take two-year as band to estimate the causal effect again. I have to say that it's not as convincing as it looks. I would like to see a very obvious discontinuity at the cut-off. Unfortunately, no evidence~
Though 20% is a bit large, yet understandable. What worries me is CWL itself is not good as it looks... like all the other policies in China.