Research of Decision Variables of Tax Revenue Based on Nonparametric Selection Approach - Innovative Management in Information and Production - page 310

Tabl e 2 Comparison of parameter models with nonparametric methods

Nonparametric Stepwise Linear

MSE method regression regression

Before eliminating variables 0.06535574 0.1181059 0.1181059

After eliminating variables 0.05892305 0.1181143 0.2854998

5 Conclusion

The paper has explored the decision variables of tax revenues. The main contribution
of nonlinear method of the nonparameters compared to the more classical linear
method of the parameters is presented as follows:

1. Robust estimation. The nonparametric method not only can apply to the variable
selection of nonlinear model but also suit for that of the linear model the mean
square error of the model after the variable selection without any increase in
error, but a slight decrease. It shows that the nonparametric kernel estimation is
more robust than linear parameters.
2. The precision of fitting model improved. The result shows after the variable
selection, the MSE of nonparametric method has reduced.
3. Automatic dimensionality reduction. It is the nonparametric method that
selected out six decision variables of tax revenue from seventeen variables based
on data driven.
4. Identify linear and nonlinear relation. The nonparametric kernel method
not only obtained a relatively robust estimation but also made sure of the less
decision variables of tax revenue. The decision variables selected respectively
by three approaches were not completely consistent. The industrial gross output
index, per capita GDP, the proportion of the second industry to GDP, and area
are the common decision variables in three approaches. The results may indicate
that tax revenue mainly depends on the development of the second industry
and different regional economic structures have different decision variables. The
variable of Real estate investment is only one not the decision variables in three
approaches, it provides negative empirical evidence whether to control real-estate
investment decisive influence to the tax revenue.

Acknowledgements The study was carried out with the financial support of the Most Important
Soft Science Subjects of Zhejiang (2011C25075).

References

S.J. Cao, X. Liu, W.J. Zhang, Fiscal decentralization, local government competition And actual tax
burden of listed companies. J. World Econ. 4 , 69-83 (2009)
H.R. Davoodi, D.A. Grigorian, Tax potential vs. tax effort: a cross-country analysis of Armenia's
stubbornly low tax collection. International Monetary Fund 38-41 (2007)