Provincial public expenditure in China: a tale of pro-cyclicality

Difficulties in accessing capital markets during recessions often force governments in developing countries to refinance or repay debt, obliging them to embark on contractionary policies (Gavin and Perotti 1997; Caballero and Krishnamurthy 2004).

Weak institutions encourage ineffective and often inappropriate spending of revenue windfalls, leaving countries with inadequate buffers to cope with adverse shocks (Tornell and Lane 1999; Talvi and Vegh 2005).

Theoretical work by Christiano et al. (2011) clearly suggests that this shift in the cyclical properties of fiscal policy is welfare improving since the optimal fiscal policy in a stochastic model with sticky prices is countercycical.

Rodden and Wibbels (2010) showed that procyclical fiscal policy among provincial governments offset the modest insurance against regional income shocks provided by central governments.

With the 1994 reforms, the autonomy of provincial governments increased significantly. In 1994, China introduced a major tax and fiscal reform program known as the tax assignment system to replace the discretionary fiscal contract system. This new scheme recentralized much of the tax revenues. The upper-level governments have the decisive power over the fiscal schemes of their directly subordinate governments with little discretion of the central government (Jia et al. 2014). There are five hierarchical levels: Central, provincial, prefectural, county, and township. This fiscal reform was a milestone in China’s intergovernmental relations, and largely reshaped the incentives of local government in economic development. Changing the resource-sharing system was to the advantage of the provincial governments; and the Central government focused more on the coordination and implementation of national economic policies (Naughton and Yang 2010).

Oversight from the central government certainly plays an important role, as emphasized by Xu (2011). Local officials are appointed by higher levels of government officials, and are evaluated by the central government based on a range of criteria. On the one hand, this system of appointment and promotion of provincial leaders serves as a powerful instrument for the national government to induce regional officials to follow the central government policies. On the other hand, the hierarchical administrative organization induces a shared authority, which is beneficial to the local governments.

For instance, from 1992 to 2008, every president, every premier, every newly elected Politic Bureau Standing Committee member of the four CCP congresses during this period was promoted from a provincial position (Xu 2011).

The ideal way to analyze the cyclicality of tax revenue is to focus on the cyclicality of tax rates because they are the true measures of the fiscal actions without the effect of automatic stabilizers (Kaminsky et al. 2005). A procyclical fiscal policy with respect to taxes is defined as increases in tax rates during recessions and reductions in tax rates during expansions (i.e., a negative correlation of tax rates with the business cycle). Unfortunately, there is no readily available data on conventional tax rates for Chinese provinces. Hence, we focus on the expenditure data which are less prone to the effect of automatic stabilizers.

Results with alternative filters (Baxter-King, Christiano-Fitzgerald, and Hodrick-Prescott filters) are provided in the robustness section.

However, controlling the time invariant factors can at most solve part of the omitted variables problem, as some omitted variables may be changing over time.

The ideal control for the initial fiscal position is local public debt which has increased markedly in the recent period (Zhang and Barnett 2014; Ambrose et al. 2015; Ang et al. 2016).

In 2003, provinces with the highest scores are Beijing, Shanghai, Tianjin, Jiangsu, Jilin, Liaoning, Zhejiang and Heilongjiang, provinces that tend to be in eastern coastal regions or on China’s borders. In contrast, inland provinces, such as Shanxi, Jiangxi, Henan, Hunan, Guangxi, Chongqing, Sichuan, Guizhou, Yunan and Gansu, tend to have relatively lower efficiency scores.

Simply by being numerous, instruments over fit the instrumented variables, failing to correct for their endogenous components and biasing the coefficient estimates toward those from non-incremental estimators. Therefore, the over-identification test is biased toward the null hypothesis of the validity of instruments. We found that the probability of the Hansen J-statistic test equals 1 in all cases.

Results with classic filters are broadly unchanged. Estimates are available upon request.

To ensure that endogeneity is not concern in the estimates, we also re-estimate Eq. (4) with one year lagged covariates. Our results remained unchanged. Results are available upon request.