unconditional relationship between TAX and MNE. However, there may be
some endogeneity left, and this could be particularly important in the cross-
sectional models employing MNE rather than NEWMNE. For instance,
such a bias would arise from tax competition between municipalities in order
to attract foreign MNE headquarters. The source of the bias would be a
correlation of the explanatory variables with any possible omitted regressors,
time-variant or time-invariant.
The first problem can be overcome by means of instrumental variable
count data model regression and the second one by quasi-differenced count
data model estimation using panel data. The corresponding models are sum-
marized in Table 4. In one of the models, we use the stock of MNE head-
quarters per municipality (MNE) in 2005 with cross-sectional instrumental
variable GMM estimation (IV-GMM). With appropriate instruments, IV-
GMM may overcome the bias associated with the omission of time-variant
or time-invariant variables from the model. Alternatively, we employ panel
data and quasi-differenced GMM estimation.24
- Table 4 -
Theoretical work on spatial tax competition suggests that a jurisdiction’s
tax rate is a function of both local and other jurisdictions’ characteristics.
Strategic interaction among jurisdictions suggests that jurisdictions set their
tax rate in response to the one(s) applied in neighboring jurisdictions. In
equilibrium, tax rates in all jurisdictions are determined by economic fun-
damentals (such as region size, factor endowments, etc.). Accordingly, a
large fraction of empirical work addresses the problem of tax competition by
means of instrumental variables regressions where weighted tax rates of other
municipalities are modeled as a function of weighted economic fundamentals
there (see Brett and Pinkse, 2000; Egger, Pfaffermayr, and Winner, 2005a,b).
In a reduced form, local tax rates are a function of economic fundamentals of
the local jurisdiction and weighted tax rates of the “neighbors”. We use the
latter idea to instrument municipality-level tax rates with average charac-
teristics of neighboring municipalities. In the cross section, the instruments
used are the averages of the share of area covered with buildings and streets,
24Neither for IV-GMM nor for quasi-differenced GMM did the estimation procedure
converge when using NEWMNE instead of MNE as the dependent variable. The rea-
son for this is to be seen in the extremely large number of zeros in NEWMNE in any
considered year.
17
More intriguing information
1. Does Competition Increase Economic Efficiency in Swedish County Councils?2. A multistate demographic model for firms in the province of Gelderland
3. On Evolution of God-Seeking Mind
4. The name is absent
5. ISO 9000 -- A MARKETING TOOL FOR U.S. AGRIBUSINESS
6. Apprenticeships in the UK: from the industrial-relation via market-led and social inclusion models
7. The name is absent
8. SOME ISSUES CONCERNING SPECIFICATION AND INTERPRETATION OF OUTDOOR RECREATION DEMAND MODELS
9. The value-added of primary schools: what is it really measuring?
10. Applications of Evolutionary Economic Geography