In econometrics, an endogeneity problem occurs when an explanatory variable is correlated with the error term.[1] Endogeneity can arise as a result of measurement error, autoregression with autocorrelated errors, simultaneous causality (see Instrumental variable) and omitted variables. Two common causes of endogeneity are: 1) an uncontrolled confounder causing both independent and dependent variables of a model; and 2) a loop of causality between the independent and dependent variables of a model.

For example, in a simple supply and demand model, when predicting the quantity demanded in equilibrium, the price is endogenous because producers change their price in response to demand and consumers change their demand in response to price. In this case, the price variable is said to have total endogeneity once the demand and supply curves are known. In contrast, a change in consumer tastes or preferences would be an exogenous change on the demand curve.

Endogeneity is a fancy word for a simple problem. So fancy, in fact, that the Microsoft
Word spell-checker does not recognize it.
Technically, in a statistical model you have an endogeneity problem when there is a
correlation between your X variable and the error term in your model.

What does this mean? Well remember that the error term in your model is due to all of
the stuff in your dependent variable that is not due to the variables you have in your
model.
So in the broadest sense an endogeneity problem arises when there is something that is
related to your Y variable that is also related to your X variable, and you do not have that
something in your model. Call that something Z, although notice that I have not claimed
that Z is a variable – it is just a “something.”

For example, endogeneity in this broad sense can be caused by omitted variables, or
unobserved heterogeneity. In this case, the endogeneity complaint is a complaint that
you left a variable (or two) out of your model.
This is obviously very familiar to everyone in this room. One reason why it is familiar is
that we all know how to deal with it: measure the variable and put it in. And we all know
how to fight with the reviewer about such things.

Unfortunately, there are other sources of endogeneity that are not so easily dealt with.
And in fact, I think that in most cases where the charge of endogeneity is filed, people are
not so much worried about omitted variables. Rather, what they are worried about is
things like simultaneity – i.e., X causes Y but Y also causes X, -- and self-selection. The
problem with such endogeneity problems is that no amount of control variables will
address them.

For an example of simultaneity, consider a very nice paper by Simcoe and Waguespack
on status signals. Sociologists are somewhat obsessed with the idea that rewards accrue
to actors because of their status, and claim that status affects the performance of those
actors – i.e., quality. This is the Matthew Effect – because people defer to high-status
actors and wish to affiliate them, these actors reap higher rewards and get more
recognition. Yet the problem is that quality also affects status – people get recognized
because they do good work.

But if all there is to the status effect is the effect of accumulated quality, then there really
is not a whole lot for sociologists to talk about, at least with respect to status. So trying to
see whether the status signal has an independent effect is very important – but controlling
for variables (even measures of quality) is not good enough.

I think we are starting to see in organizations and management research an increased
concern with these kinds of endogeneity issues. If you think about the different kinds of
endogeneity concerns, what you see is an increasing concern with more complex forms of
endogeneity – everyone takes it for granted that we should be worried about omitted
variables, a fair number of people think seriously about self-selection problems, and we
are seeing the beginnings of an emphasis in management research on worrying about
simultaneity problems.