China’s government finance and food security nexus: a regime switching analysis

ABSTRACT

This paper is an empirical assessment of the government finance and food security nexus in China from 1978 to 2016. Using autoregressive distributed lag for linear and threshold cointegration, the results suggest that China’s government finance and food security have reinforced each other throughout the reform era when a structural break was allowed in the model. However, the strength of cointegration and Granger causality in the government-led food security link is stronger, hence supporting the state interventionist view that a strong state has played a dominant role in fulfilling national food security in China. China’s government finance and food security changes indicate food sector reform probabilities under fiscal constraint, as regime-switching indicators have delineated the Chinese government’s difficult fiscal periods and the main reforms in the food sector. In practice, this result implies that further policy reforms can be effective for China’s future food security.

KEYWORDS:

• Food security
• government finance
• agricultural reform
• structural change

JEL CLASSIFICATION:

• Q18
• Q11
• P21
• O23

Acknowledgements

The authors thank Chien-Hsun Chen of the Chung-Hua Institution for Economic Research, Taiwan, for his advice and comments on this work. The authors also thank Tsang-Yao Chang of Feng Chia University, Taiwan, for his technical support in writing an early draft.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ The ‘basic’ food, on some definitions, refers to just rice and wheat; on others, the ‘basic’ food might be extended to include all cereals.

² According to the contemporary agricultural history study group (1998), during 1961–1965 (Great Leap Famine) and 1976–1978 (pre-reform food shortage) China’s net food import–export balance was negative at –13.44 and –11.29 million tonnes, respectively.

³ Calculations based on central government final expenditure divided by the same year’s fixed base cumulative consumer price index (1979=100).

⁴ ADL method is valid for either I(0), I(1) and a mix between them.

⁵ In accordance with the 1979 Decisions of the Fourth Plenary Session of the Eleventh Communist Party of China Central Committee, the government quota procurement price for grain increased by 20%, while above quota price for grain increased by 50%.

⁶ The 1984 wheat output increased by 63.1% from the 1978 level and rice increased by 30.2%.

⁷ This is indicated in Document No. 137 issued by the Chinese Communist Party Central Committee in 1982.

⁸ The central government lowered grain procurement price in 1985 using a ‘reversed 30:70 ratio’ (dao san qi) to relieve its fiscal burden. The term reversed 30:70 ratio was a new structure with 30% of government procurement price at a low baseline and 70% of procurement price fetching an award procurement price that was 50% higher than the baseline price. This reversed-30:70-ratio price was lower than the preceding year’s award procurement price which disincentivized Chinese grain farmers.

⁹ This is known as ‘purchasing and selling grain at the same price’ (gouxiao tongjia).

¹⁰ Compared to the preceding year, the 1993 and 1994 consumer price index increased by a respective 14.70% and 24.13%.

¹¹ The reason is, as marketization deepened in urban China, salaries in government-dominated service sectors increased.

¹² The Chinese government comprehensively increased rural subsidies from 2007 to 2010. See Document No. 1s in 2007 and 2010.

¹³ Fixed based cumulative consumer price index (1979=100) is used as the deflator.

¹⁴ The same lag lengths $p=q$ for the ADL model are used in this paper.

¹⁵ Here $\tau$ represents the percentile of the distribution of ${\epsilon }_{t-1}^k$ or $\mathrm{\Delta }{\epsilon }_{t-1}^k$, while $\tau$ is obtained by maximizing the statistic between 15% and 85% of the percentiles of the sorted error terms ${\epsilon }_{t-1}^k$ and $\mathrm{\Delta }{\epsilon }_{t-1}^k$.

¹⁶ Schwarz–Bayesian information criterion (SBC) was also considered for selecting the regime indicator; results are shown in Table 3(b).

¹⁷ Li and Lee (2010) gives two tests for threshold ADL cointegration: the BDM test (Banerjee, Dolado, and Mestre 1998) and the BO test (Boswijk 1994). As to the power and size, the BO test is argued better than the BDM test. The BO statistics are thus largely followed to conclude the presence of threshold cointegration. BDM test results are also included for reference. See Section IV for details.