Fiscal Behavior Volatility, Economic Growth, and Urban-Rural Income Disparity

2.1 Data description

One of the hypothesis (hypothesis I) is that policy volatility exerts a negative impact on urban-rural income disparity through direct and indirect effect. This paper chooses panel data of 30 provinces from 1994 to 2011 according to the data availability. Base data of urbanization (UN), open economy (OP), financial development (FD) are from the regional statistical yearbook from 1995 to 2012, while GDP, urban-rural income disparity (y) and dual economic structure (S) come from the China statistical yearbook in 2012 and CEINET database. This paper changes these variables into real variables based CPI of 1994 to eliminate the influence of inflation. Table 1 reports summary statistics of the sample.

2.2 Measurement of fiscal behavior volatility

Fiscal behavior mainly includes the fiscal revenue policy and fiscal expenditure policy. In general, there are three measures of fiscal behavior volatility: first, standard deviation of fiscal policy variables. Second, conditional variance through GARCH model (Henisz, et al.[18]), but this method is usually used for policy changes caused by volatility. While the above two methods of measurement may cause inaccurate estimation results because they don’t exclude the policy fluctuations caused by economic fluctuations. Third, remaining residual volatility by regression analysis. The fiscal behavior is decomposed into three parts including discretion, persistence and volatility. Excluding discretion and persistence, the remaining residual volatility is the fiscal behavior volatility. In this section we construct a measure according to the third method based on Fatas and Mihov([4,19]), Afonso, et al.[5].

In order to distinguish the components of responsiveness, persistence and discretion of government spending and revenue, we estimate with the following regressions:

where G is real government spending, R is real government revenue and X is a set of controls including a time trend, the level value and square of inflation.

GAP is the regional output GAP, the values can be estimated by HP filter method:

The natural logarithm of real output ln Y_i is decomposed into trend components (the natural logarithm of potential output lnYi∗ ) and periodic components (the output gap lnYi−lnYi∗ ). The smoothing parameter λ = 25 by using the annual data.

When it comes to the analysis of fiscal behavior volatility, we focus on fiscal expenditure policy. Because fiscal revenue policy in China is relatively stable, while fiscal expenditure policy has greater volatility, which leads to the structural imbalance of the central finance and local finance. And some studies indicate that the fiscal revenue policy has small impact on the income disparity, while transfer payments in government spending policy may expand the competitiveness and efficiency of the differences between regions (Liu[20]).

Equation (1) is a variable coefficient and dynamic model, and will be estimated by GMM (Generalized the Methods of Moments) method using instrumental variables because of endogenous problem. We use two lags of output growth and lagged inflation as instruments. We get each regression residuals by grouping regression, the standard deviation of residual stands for fiscal behavior volatility, and will be carried out further the exponential regression. They all pass the correlation test of endogenous variable and relative to exogenous test of the residual.

3.1 Regression model

In recent years, many scholars have studied the impact factor of urban-rural income disparity from the perspective of dual economic structure, urbanization level, the level of economic growth and opening, the most significant impact factor is the urban-rural dual economic structure, including binary tax system, dual fiscal investment system, dual social security system, which lead to widen the urban-rural income disparity. At the same time, in order to study the nonlinear effects between economic growth, financial development and the urban-rural income disparity, their square are considered in the model. The regression model as follows:

where y_it is the urban-rural income disparity of the ith provinces in time t, S_it and FBV_i are dual economic structure and fiscal behavior volatility of local government, respectively. A series of control variables include economic growth (GDP), financial development (FD), level of urbanization (UN), open economy (OP); λ_t is specific time fixed effect, which has nothing to do with the region; u_i is fixed effect for particular region, which has nothing to do with the time.

3.2 Results and discussion

One problem of estimating model (4) using OLS is that the lagged dependent variable is endogenous to the fixed effects (u_i), which will increase the estimated error. Thus, OLS estimates of this model will be inconsistent, even in the fixed or random effects settings, because would be correlated with the error term, ∊, even if the latter is not serially correlated. First-difference removes the individual effect (u_i) and thus eliminates a potential source of bias. While, variables become endogenous when they are first differenced if they are not strictly exogenous, since the first-difference will be correlated with the error term. Following Holtz-Eakin, et al.[21], Arellano and Bond[22] developed a Generalized Method of Moments (GMM) estimator for linear dynamic panel data models that solves this problem by instrument variables: levels of the dependent and endogenous variables, lagged two or more periods; levels of the predetermined variables, lagged one or more periods. The exogenous variables can be used as their own instruments.

The results of system-GMM estimation of the model are shown in Table 2. Column (1)~(4) give the basic estimation results of system-GMM, column (5) and (6) are the results by mixed regression and random effects, which are for the comparison and analysis. AR (1) and AR (2) results show that there is no second order autocorrelation between the residual error, Hansen test results show that the instrumental variable is effective, and the proposed system-GMM method is also effective.

The symbols of main explanation variable are consistent with rational expectations. There is a strong positive relationship between fiscal behavior volatility (FBV) and urban-rural income disparity. In addition to statistical significance, impact of fiscal behavior volatility on the urban-rural income disparity is quite considerable. From column (2), one point increase in the fiscal behavior volatility increases urban-rural income disparity by 0.193%. Comparing with column (1) and (3), there is a weaker affect through the dual economic structure, but it is still a positive correlation. Coefficient of interactive items FBV*S is positive, which instructs that the larger urban-rural income disparity is, the greater the volatility of financial behavior is, especially for the provinces with more obvious dual economic structure.

Hypothesis II is verified by the estimation results in column (1) and (2), which shows that the coefficient of the square of economic growth (GDP2) and financial development (FD2) are significantly negative with the control another variable. There is a inverted U-shaped relationship between economic growth, financial development and urban-rural income disparity, which further verifies the Kuznets inverted U-shaped curve.

Dual economic structure (S) and opening (OP) are related positively to the urban-rural income disparity, these variables expand income disparity, which is consistent with the research conclusions of Lu and Chen[13]. Urbanization (UN) is negatively related to the income disparity, which means that the higher the level of urbanization, the smaller the urban-rural income disparity, which is also the cause of improvement of the urbanization in China. The income disparity is sustainable because the lag of income disparity is highly significant in the results.

Column (4) is a reduced version which removes all of the square and interaction of variables, however, there is also positive correlation between the fiscal behavior volatility and urban-rural income disparity, and the influence of the dual economic structure is more significant in urban and rural income gap. GDP is negatively related with the income disparity, the faster the economic growth, the smaller the income disparity, maybe it is inconsistent with status quo, but the conclusion is established in the long run (a combination of column (1) and column (2)). Financial development and urban-rural income disparity has significant positive correlation, financial development contributes 7.3% in expanding the income gap between urban and rural areas.

Column (5) and (6) are estimated results by static panel data regression method respectively. Due to model (4) the effect of individual FBV is needed to be estimated, in order to compare with the results of the first four columns, column (5) uses the method of mixed regression (POOL), and column (6) uses random effect. Results show that the fiscal behavior volatility and urban-rural income disparity are still positive correlation, and square and interaction of variables are significant negative.

3.3 Robustness test

In order to further test the robustness of the conclusion of fiscal behavior volatility effect on the urban-rural income disparity, this paper analyzes the robustness from the following two aspects. 1) Using the standard deviation of fiscal expenditure to depict fiscal behavior volatility. The results show that the symbols of explanatory variables and control variables and significance remain unchanged. At the same time, the Hansen test and the value of AR(1) and AR(2) also meet the requirements, which shows that there is a positive correlation between fiscal behavior volatility and urban-rural income disparity, namely the greater the financial behavior volatility, the greater the income gap between urban and rural areas. And the square of the economic growth and financial development are significantly negative, which shows the inverted U-shaped curve relationship between the financial behavior volatility and urban-rural income disparity. 2) By increasing the annual virtual variable control for the year effects, and the empirical result is consistent with the prior.