# Instrument Variables and Exclusion Restriction from a Mediation perspective

I'm having trouble making sense of the exclusion restriction in instrumental variables.

I understand that the unbiased treatment effect is $B = \frac{Cov(Y, Z)}{Cov(S, Z)}$, where $Y$ is the outcome, $S$ is the treatment, and $Z$ is the instrument. In other words, $B = \frac{ITT} {\text{Compliance Rate}}$.

However, if I think about this in a mediation framework, and apply the exclusion restriction, this makes less and less sense.

In a mediation framework, ITT = the total effect, or the $Cov(S,Z)\cdot Cov(Y,S) + Cov(Y,Z|S)$. So, the unbiased treatment effect is:

$\frac{(Cov(S,Z)\cdot Cov(Y,S) + Cov(Y,Z|S))}{Cov(S,Z)}$, which reduces to:

$Cov(Y,S) + \frac{Cov(Y, Z|S)}{Cov(S, Z)}$,

so the unbiased causal estimate is the effect of the biased treatment + the effect of the instrument ($\frac{controlling for the treatment} { compliance rate}$.

However, with the exclusion restriction, there is not supposed to be an effect of the instrument once we control for the treatment.

An example, from Gelman's Sesame Street example. First, obtaining the unbiased treatment effect via 2SLS:

fit.2s <- lm(regular ~ encour, data = df)
watched.hat <- fit.2s$fitted fit.2b <- lm(postlet ~ watched.hat, data = df) summary(fit.2b) which gives the answer, 7.934. And now, within an SEM framework: library(foreign) library(lavaan) mod <- ' regular ~ a*encour postlet ~ b*regular + c*encour ind := a*b total := a*b + c ' fit <- sem(mod, data = df) summary(fit) Regressions: Estimate Std.Err Z-value P(>|z|) regular ~ encour (a) 0.362 0.051 7.134 0.000 postlet ~ regular (b) 13.698 2.079 6.589 0.000 encour (c) -2.089 1.802 -1.160 0.246 Defined Parameters: Estimate Std.Err Z-value P(>|z|) ind 4.965 1.026 4.840 0.000 total 2.876 1.778 1.617 0.106 13.698 - 2.089/.362 = 7.92 So, the only reason that the unbiased treatment effect is not just the biased treatment effect is the there is still an effect of the instrument when controlling for the treatment, which, according to the exclusion restriction should not happen. Am I missing something here? ## 1 Answer You correctly state that under the LATE-style IV assumptions with a causal effect of the IV Z on the treatment S, exogenous instrument, and no direct effect on the outcome Y, your treatment effect B of S on Y is identified as$Cov(Y,Z)/Cov(S,Z) = ITT/Compliance Rate$So clearly,$ITT = Cov(Y, Z)$, and not$Cov(S,Z)⋅Cov(Y,S)+Cov(Y,Z)\$, as you mistakenly state. This is also intuitively clear because if the instrument is exogenous with respect to the treatment and the outcome, its causal effects are identified and (under linearity) are simply the correlation between the instrument and the outcome.

You further seem to imply that the IV and Y should be unrelated in a regression of Y on the treatment and the IV. This is not the case if treatment S is actually endogenous. Then, it is a collider, as it is caused by the IV and the unobserved error term of Y. Conditioning on the treatment makes the IV and the error term dependent, and therefore also the IV and Y. So you get a non-zero regression coefficient even if the exclusion restriction is valid. This should be very clear if you draw the causal graph.

If it was not the case, we could actually test the exclusion restriction, but of course we know we usually can't test it! (At least not that easily).

• I added an edit to clarify, but from a mediation perspective, the ITT is (Cov(S,Z)⋅Cov(Y,S)+Cov(Y,Z|S))/Cov(S,Z). This is verifiable with Gelman's example: coef(summary(lm(postlet ~ encour, data = df)))['encour',]['Estimate'] = 2.88. From the SEM output, (Cov(S,Z)⋅Cov(Y,S)+Cov(Y,Z|S))/Cov(S,Z) = 2.88 (the estimate next to total, which is a*b + c)
– sam
Sep 26, 2016 at 13:09
• Then how can it be also true that Cov(Y,Z)/Cov(S,Z) = ITT/ComplianceRate? Your equation for the ITT and the IV formula cannot be true at the same time. With the exclusion restriction, your formula is not true. Sep 26, 2016 at 13:11
• Also note that "from a mediation perspective", the ITT is a causal quantity, so needs to be defined in potential outcomes or with the do-operator, not as a function of the PDF of observable. So the ITT is defined to be E(Y|do(Z)), for example. See en.wikipedia.org/wiki/…. Sep 26, 2016 at 13:13
• Because Cov(Y,Z) can be decomposed into (Cov(S,Z)⋅Cov(Y,S)+Cov(Y,Z|S)). Maybe I'm missing something here, but that's kind of my point: identification of the causal impact relies on the exclusion restriction to be false, or so it seems.
– sam
Sep 26, 2016 at 13:17
• I've edited my answer. I still do not fully understand your confusion. You might find it very illuminating to read Pearl, Judea. "Linear models: A useful “microscope” for causal analysis." The tools in there are extremely easy and make it clear how simple this problem is. Sep 26, 2016 at 13:29