the Boston metro area, we identify social interactions by comparing the propensity of individuals
on the same versus nearby blocks to work together. We find significant evidence of social
interactions: residing on the same block increases the probability of working together by over 33
percent. This finding is robust to the introduction of detailed controls for socio-demographic
characteristics as well as across various specifications intended to address biases caused by
sorting below the block group level and housing market referrals exchanged between people who
work together. Furthermore, the relationship between socio-demographic characteristics and the
strength of social interactions make sense. Social interactions tend to be stronger when the match
involves individuals who are likely to interact because they are similar in terms of education, age,
and presence of children, which is consistent with the notion of assortative matching in social
networks. Interactions also appear to be stronger when they involve at least one type of individual
who is strongly attached to the labor market leading to weaker interactions when both members
of the pair are high school drop-outs or married females.
In the second half of our analysis we use our heterogeneous referral effect estimates to
construct an individual-specific measure of the availability of referral opportunities on her block
of residence. Even after controlling for individual attributes, observable block attributes, and
block group fixed effects, this measure is a statistically significant determinant of all of the labor
market outcomes considered across all of our specifications. In terms of economic magnitude, a
one standard deviation increase in referral opportunities raises expected labor force participation
by 1.0-1.6 percentage points and earnings by 2.7-3.8 percentage points.
More generally, this paper provides a new approach for examining the effect of social
interactions based on variation in geographic scale. In presenting the results related to
neighborhood referrals and labor market outcomes, we also provide direct evidence on the
reasonableness of this new design by testing whether its key assumptions hold on observable
characteristics. In particular, we demonstrate that based on their observable characteristics, pairs
of individuals residing on the same block would actually be slightly less likely to work together
than pairs in the same block group but not the same block. This provides strong evidence that our
research design is likely to be robust to within-block group sorting.
This evidence also suggests that the research design proposed in the paper may be useful
in a variety of contexts. For example, in the case of welfare participation, the block of residence is
unlikely to greatly influence access to public service providers after controlling for the block
group. More generally, this design might be extended to the study of neighborhood effects in
specific contexts (e.g., specific types of neighborhoods), on specific populations (e.g., youths),
and for alternative outcomes (e.g., education, teenage fertility, health, welfare participation),
31