Thus in order to identify causal effects we need to condition on variables that

# Thus in order to identify causal effects we need to

• 30

This preview shows page 11 - 13 out of 30 pages.

Thus, in order to identify causal effects, we need to condition on variables that are effectively causes and not on consequences. This criterion is called the Back- door criterion . 29 It asserts that a causal effect is identified if all back-door paths are blocked. A path is any sequence of connections in any direction between two variables. For example, in Figure 3, the sequence INC t INC t - 1 PG t CR t - 1 S t S t +1 is the path connecting the levels of income at the t-th period and segregation at the next period. Moreover, a back-door path is a path where a directed edge points towards the initial variable. It is important to control for these back-door paths because they generate bias in the estimation of the true causal effect, in our example INC t S t +1 . The previous example is a back-door 27 Pearl[45] discusses this point at length. 28 Cholecystitis is the inflammation of the gallbladder. It was believed to cause or aggravate diabetes (see Berkson[7]). 29 See Pearl[45] and, Morgan and Winship[37] for more details on this point. 10
path for instance. We will further say that a path is blocked if we can find a set of control variables Z to condition on which are either the middle variable of a chain of mediation or the common cause of two consequences. If a control variable Z is a consequence of two independent causes, it will spoil the analysis due to the spurious correlation mentioned earlier. However, if we can additionally control for another variable Z’ which is not a consequence of two causes, we will thus still block this back-door path. Then the back-door criterion identifies causal effects by controlling for all the other potential causes. Our causal graph in Figure 3 has a lot of back-door paths potentially threat- ening the identification of causal effects. In fact, every path is a back-door path because a determinant is itself caused by another determinant at the previous period. Then, a directed link is pointing towards our starting variable, thus cre- ating a back-door path by definition. But we can block all these back-door paths by conditioning by all the determinants at the previous period. Due to the def- inition of a cause, we have this nice autoregressive structure that allows only descendent links from the previous period to the actual one. Thus, we have only two types of back-door paths, those with common consequences and those with- out. For instance, between CR t S t +1 we can have the following back-door paths, without common consequences CR t CR t - 1 S t S t +1 and with CR t CR t - 1 PG t INC t - 1 SD t PG t +1 S t S t +1 . For the first back-door path, we can block it by conditioning either on S t or CR t - 1 as S t is in the chain of mediation CR t - 1 S t S t +1 , and CR t - 1 is the common cause of S t and CR t . For the second back-door path, there are two common consequences on which we should not condition, PG t in CR t - 1 PG t INC t - 1 and PG t +1 in SD t PG t +1 S t . However, it can be blocked by conditioning on all the other variables, for instance S t .

#### You've reached the end of your free preview.

Want to read all 30 pages?

### What students are saying

• As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

Kiran Temple University Fox School of Business ‘17, Course Hero Intern

• I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

Dana University of Pennsylvania ‘17, Course Hero Intern

• The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

Jill Tulane University ‘16, Course Hero Intern