We have three specifications of IVs. First, we only use two IVs:
(1)living in the same building and (2)living on the same floor as the
2011-diagnosed smoker. Second, we include three IVs: (1) living
in the same building as the 2011-diagnosed smoker; (2) living on
the same floor as this smoker; and (3) whether the person ever
knew a lung cancer neighbor. Third, we use five IVs. In addition to
the above three IVs, we add interaction terms for: same building
and knew the diagnosed smoker; and same floor and knew the
neighbor.
If there are systematic differences in prior information between
residents of the treatment and control buildings, these differences
might invalidate our identification strategy. This is unlikely to be a
problem for several reasons. First, we presented empirical evidence
above indicating that control and treatment group respondents had
similar acquaintance with lung cancer patients one month before
the 2011-cancer diagnosis. Second, although spatially separated,
the median distance between treatment and control buildings is
115 m (