L. LEE, G. VULETIN 813
Extrabudgetary funds, which is capital used for specific
projects assigned by the central government, are ex-
cluded.
3. Flypaper Effect Evidence (1980-2008)
Table 1 shows the regressions for Chinese provinces for
the period 1980-2008. We consider the following basic
specification:
,
f itit
f
ity it
gy
(1)
where g, y and f represent government spending, output,
and fiscal transfers all expressed in real and per capita
terms. Column 1 reports basic Ordinary Least Squares
regressions without controls and assuming that the re-
siduals are homoscedastic and have no autocorrelation.
The marginal propensity to spend out of fiscal transfers is
clearly larger than for local output; there is a flypaper
effect.
The regression reported in Column 2 allows for het-
eroscedasticity and autocorrelation. Column 3 allows for
provincial fixed effect and Column 4 also includes year
dummies to reduce the omitted variable bias that may
occur as a consequence of the processes of decentraliza-
tion. Column 5 controls for population density, a typical
variable used to proxy for cost of provision of public
goods. This variable is expected to have a negative sign
as such cost is expected to decrease with higher popula-
tion concentrations3. Columns 6 and 7 split the sample
into before and after the 1994 co-participation and tax
reforms. In every case the flypaper effect remains a
strong empirical regularity with a size close to 0.9. This
novel finding for Chinese provinces coincides with that
of federations.
4. Teflon Effect Evidence (1952-1979)
Table 2 shows the regressions for the central government
for the period 1952-1979. We find that, even after relax-
ing the assumption of homoscedasticity, the marginal
propensity to spend out of fiscal transfers is clearly
smaller than national output.
This finding is the opposite to the one of Section 3 at
the provincial level and novel in nature. No previous
study found that government spending responds less to
an increase in fiscal transfers than to an equal increase in
private income. We define this novel regularity teflon
effect. The term teflon effect aims to visualize the idea
that, contrary to the flypaper effect, money tends to slide
from where it hits: money from the private sector (i.e.,
from private income) is taxed and mostly allocated to
public spending, while money from the public sector (i.e.,
from provincial fiscal transfers) tends not to be spent and,
therefore, rebated to citizens4. In particular, while an in-
crease of $1 in national income raises central government
spending by $0.6, an equivalent increase in fiscal trans-
fers from provinces only triggers into an increase of
$0.015 in expenditures. Therefore, the size of the teflon
effect is 0.586.
5. Rationalizing the Teflon Effect
This section develops a simple optimal fiscal policy
model that rationalizes the teflon effect using the collec-
tion cost/distortionary taxation arguments developed by
Hamilton (1986) [6], Aragón (2009) [7], and Vegh and
Vuletin (2012) [8] to rationalize the flypaper effect.
These papers show that the flypaper effect occurs be-
cause subnational tax collection has higher collection
costs or are more distortionary than the central govern-
ment.
The endowment economy is inhabited by a benevolent
fiscal authority (FA) and a representative citizen (RC)
blessed with perfect foresight. Without the loss of gener-
ality and in order to obtain analytical solutions, we use
log preferences for RC’s utility function:
lnln ,Wg c
,yc
(2)
where c is private consumption and g is public spending.
The RC’s budget constraint is given by
(3)
where y is output and τ is lump sum tax collection.
Because the central government has two sources of fi-
nancing, own revenues and fiscal transfers, the central
FA’s budget constraint is given by
11,gf
(4)
where f is total provincial fiscal transfers to the central
government.
and
are the central government and
effective provincial collection cost/distortionary taxation
ii i
ff
. Hamilton (1986) [6] rationalizes them
as the real costs of distortionary taxation, while Aragón
(2009) [7] treats them as pure collection costs.
Solving the model we get
111.
2
gyf
(5)
From expression (5) we obtain
1dd1
,
2dd 2
gg
fy
(6)
which states that the relative optimal response of public
spending to each type of shck crucially depends upon o
3The estimated coefficient for population density would become nega-
tive if year dummies were not included.
4Naturally, funds are actually not rebated via checks or financial depos-
its but they rather decrease the amount being taxed in net terms.
Copyright © 2012 SciRes. ME