1。共线性。也就是解释变量里有比较高的相关关系。（这个是不是所谓的伪回归？）
解决方法是找工具变量，进行2SLS等类似回归。
I think it should not be a reason, because colinearity suggests that additional variables have no explaining power, which may not lead to high R^(2).
A possibility is that the regressand is highly related to specific country or year, i.e., country--spefic effect may affect the regressand.

非常感谢！很有道理·！我尝试了一下我自己的回归，发现加入country dummy后，R^2增加了很多。加入year dummy就没太大改变。

另外就是，如果考虑autocorrelation，比如我用xtpcse y x, c(ar1)，R^2也增加很多。

请问很多变量同时都很显著，会是因为 multi-colinearity吗？

另外，您知道为什么考虑autocorrelation，R方也会增大么？比如从0.4增大到0.8。

想明白了，有关autocorrelation的原因，被解释变量是贫困率，随时间变化不大，所以和上一年的高低相关度比较高：

The high R^2 in our regression could also come from the control of series autocorrelation. Since poverty rate doesn’t change remarkably over time, it make sense that the poverty rate this year highly depends on the poverty rate last year. We use the command ‘xtpcse’ in STATA which controls for series autocorrelation,  heteroskedasticity and simultaneous spatial correlation. Therefore, r^2 could also become higher when take the autocorrelation into consideration.

路过看看学习学习