political groups. Finally, we present (Section 6) our data and results from empir-
ical tests of the theoretical predictions. Conclusions and suggestions for further
work are collected at the end (Section 7).
2. The model
2.1. Economic policy
A population consists of 4 economic groups of equal size, normalized to unity and
indexed by J. Individuals in group J have preferences represented by:
VJ(q)=1-τ+H(gJ) . (2.1)
Here, τ is a lump sum tax constrained to be non-negative, gJ is a local public
good that only benefits economic group J, H is a well-behaved concave utility
function, and q denotes the full vector of policy instruments. Individual income
is normalized to 1 for all individuals. The government budget constraint equates
total tax revenue to total spending:
4τ=XJ gJ+XP rP , (2.2)
where rP denotes political rents (here, literally taken out of tax revenue) appro-
priated by political party P.
This is a simple and standard economic policy setting in the literature: see for
instance Persson, Roland and Tabellini (2000), Persson and Tabellini (2000). The
vector of policy instruments, q = τ, gJ , rP , induces a three-way conflict of
interest: (1) among economic groups over the allocation of the (targeted) spending
on local public goods, gJ ; (2) between politicians and citizens at large over the
total size of political rents, ΣrP , versus (non-targeted) taxes, τ ; and, (3) among
politicians over the allocation of these rents, rP .
A benevolent and utilitarian social planner, who assigns no value to the rents
captured by politicians, would implement the following policy (subscripts denote
partial derivatives and -1 an inverse function):
gJ = H-1(1), rp = 0 (2.3)
with taxes residually determined from the government budget constraint (we as-
sume an interior optimum). All groups are treated equally and the marginal