1. We began by posing the question: how can we evaluate the impact of the odd-
even road rationing scheme on Delhis air quality?
2. We agreed that a natural idea is to compare air quality before and during the
road rationing scheme.
4. The pitfall in the pre and post comparison was it does not hold all the other
variables constant i.e., it violates the ceteris paribus condition. Then it is hard to
know whether the observed change in air quality is due to the policy or due to other
factors.
5. The ideal experiment is to compare the observed change with the change in air
quality in the counter-factual, i.e., the world in which the policy never happened.
7. If all other variables are held constant, then the simple pre and post comparison
is valid.
8. In the hours of study example, we end up comparing students who studied more
with students who studied less. But here too other things may not be the same
between the more and less hard working students. In particular, the more hard
working students may also be more motivated and more able. Then the grade gap
between the more and less hard working students cannot be fully attributed to the
difference to the hours in study.
10. Sometimes it is impossible to hold all other things constant. The fertilizer effect
on yield example showed that if all other variables cannot be held constant, another
strategy of estimating the counter-factual would be to choose fertilizer applications
randomly in each plot (independent of the other factors that determine yield). This
is possible when one can design an experiment. However, in many situations, it is
not possible to design an experiment. In that case, we have to work with
observational data.
11. In the air quality example, a strategy better than the pre-and post comparision
is to employ difference-in-difference (DID); i.e., compare the change in Delhi with
the change in the surrounding regions. DID is one way of estimating the counter-
factual and it is valid under certain conditions.