Guidelines and principles of did

DID First, the parallel trend test must be satisfied , If not satisfied , Use triple DID test .

Why should we test it ？
Upper figure ：

If you don't check , It means that the early development characteristics are different , Two individuals can't play together , How can we compare differences ？
D1 After the policy ,D2 Before the policy .

The video screenshot comes from ：

https://www.bilibili.com/video/BV1Ut4y1z7YE?share_source=copy_web

Ha ha ha ha ha ha , How to do DID Parallel trend test ？
That's true
Put the link ：

https://www.jianshu.com/p/fb88cdfdf9f7

A lot of links , In fact, it's quite simple , Is to take the dummy variable of the year for regression , If the dummy variable of the year before the policy occurs is not significant , The dummy variable of the year after the policy occurred is significant , That means , In accordance with the parallel trend test .
Understand the basic principle .python You can run

Screenshot Links ：

https://www.jianshu.com/p/fb88cdfdf9f7

The original said ：

And, of course, there's graphical representation . Not for the time being .

Next, let's have a look DID How to do , Very simple! .

DID How to do

https://www.bilibili.com/video/BV1Ho4y1d7za?share_source=copy_web

The above link is clear , But don't look at the trouble , Actually , No matter how long or did Variable , Just set the local policy dummy variable to 1 That's all right. .

picture source ：

https://www.bilibili.com/video/BV1Ho4y1d7za?share_source=copy_web

Then analysis DID It's all right