这篇外文说 他参照wooldridge（2002）的做法
先预测x1的概率 以这个概率是连续的，可以作为作为工具变量，进行2sls。这个我似乎看懂了。但他只解决了文章的一个模型，这个模型是不带交叉变量的。紧接着他又说如何解决第二个模型，带交叉变量的，我没看明白：This approach has three steps. In the first step, we predict auditor
quality as in Table 3. In steps 2 and 3, we use a standard instrumental approach, with the
predicted value obtained from step 1 as an instrument. 到这里我看懂了，第一个模型完全可以按他说的解决，接下来就不懂了。In our subsequent analysis we interact
BigSix （六大事务所）with past stock returns（上期股票收益）, we follow Wooldridge (2002) and used the demeaned value of
past returns.13 Wooldridge (2002) indicates that, under fairly general conditions, this procedure
provides efficient estimations. In addition, we are using the fitted probability of auditor quality
as an instrument, the effect of misspecification in auditor quality model is mitigated. We also
correct the standard errors for heteroskedasticity and serial correlation.
扫码加好友，拉您进群
全部版块
我的主页
收藏
