18
0.456). This represents an over 100% increase in the likelihood the service will be
brought in- house.
On the other hand, there is no evidence to support the hypothesis that for-profit
contracting is more likely to result in vertical integration than other forms of contracting.
In the Ferris and Graddy and Stein models, the coefficient is not distinguishable from
zero and in the Brown and Potoski analysis, we find statistical as well as substantive
significance in the wrong direction (i.e., services with for-profit contracts are less likely
to be integrated).
Turning to the measures of service type, each typology receives mixed support.
Ferris and Graddy’s “public works” category is, as expected, negatively related to the
likelihood of movement in- house (influence = -0.068). But, their “health and human
services” grouping defies expectations and also exhibints a negative relationship of
similar magnitude. While Stein’s “collective” class was expected to be positively
associated with vertical integration, and his “private” category negatively correlated,
both, in fact, are found to have negative and significant relationships. Finally, high asset
specificity is associated with lower levels of integration (influence = -0.065), but service
meterability is not significantly related with the decision to move to in- house production.
Council- manager cities were argued to be less likely to pull services in-house due
to the market perspective held by many professional managers. Under the Brown and
Potoski specification, this assertion is supported. Holding all else constant, cities with
this form of government are nearly 5% less likely to integrate service delivery than their
counterparts. However, under the other two specifications, this variable is not related to
the propensity for vertical integration. The reverse is true regarding contracting capacity.
Convention