Place of Work and Place of Residence: Informal Hiring Networks and Labor Market Outcomes

average correlations for the full sample: the first column reports unconditional correlations, while
the second conditions on block group fixed effects, and the third includes, in addition,
specifically, whether the house is rented or owned and its corresponding rent or self-reported
value, respectively.²² In each case, both the individual and block measures are first regressed on
the corresponding variables (e.g., block group fixed effects) and the correlation between the
residuals is reported.

The results indicate a significant amount of sorting on the basis of education, race, age,
and the presence of children across the neighborhoods of the metropolitan area as a whole. The
correlation between whether an individual is a college graduate and the fraction of neighbors that
are college graduates is 0.21, while that between whether an individual is black and the fraction
of black neighbors is 0.56. The second and third columns provide an explicit test of our
identification strategy, providing a measure of sorting on observables within block groups. As
these successive columns clearly demonstrate, the correlation between observable individual and
neighbor characteristics falls to near zero as only within-block group variation is isolated. The
inclusion of block group fixed effects reduces the estimated correlations by 70-75 percent on
average, with a remaining maximum correlation of 0.07 across all characteristics and 0.04 across
all characteristics except race. The inclusion of housing characteristics, which is intended to
control for the fact that some within-block group sorting would be expected if the housing stock
differed significantly across blocks within a block group either in terms of prices or tenure of
occupancy, drives these estimated correlations slightly closer to zero.

The second set of three columns in Table 3 reports average correlations for a sample of
blocks with at least five sampled workers. We drop blocks with a small number of workers at
various points throughout our analysis for two reasons. First, blocks with a small number of
residents are largely non-residential and, consequently, interactions among neighbors may be
limited on such blocks. Second, as we discuss in greater detail below, a measurement error arises
related to the use of the 1-in-7 sample of individuals observed in the Census to estimate
neighborhood effects. In this case, blocks with only a small number of workers may be

magnitude because an individual’s own characteristics contribute a significant amount to the average
neighborhood characteristics of others within the same block group. By sampling only one individual per
block, we report an unbiased estimate of the correlation between individual and neighborhood
characteristics at the block level.

²² The housing controls include whether both individuals reside in owner-occupied housing, whether both
individuals reside in rental housing, the average rent or house price for two households if both are owners
or both renters, and the absolute value of the difference in rent or house price if both are owners or both
renters.

12

<<   11   12   13   14   15   16   17   >>

More intriguing information