Price Responsiveness of Farmers and the Effectiveness of Grain Support Policies in Yan Jin, Cornelis Gardebroek, and Nico Heerink Abstract

Declining arable land and yield stagnation challenges China to feed its growing population. To increase national food security and farmers’ incomes, China has introduced various grain support policies since 2004. This paper evaluates the effectiveness of price supporting policies on the Chinese rice acreage. To investigate whether these policies were effective given that global rice prices increased after 2004, we perform our empirical analysis in two phases. First, using data on rice prices and acreages, we test whether supply responses of rice farmers have changed since 2004 because price responsiveness is required for the policy to be effective. Second, we perform Monte Carlo simulations to construct counterfactual acreages to assess the impact of the policy. The results show that acreages respond positively to farmers’ expected price for early, middle, and late indica rice. For early and late indica rice, there is a statistically significant positive change in acreage response to expected prices after the policy intervention in 2004 than before. Monte Carlo simulations show that the probability of effective policy for early, middle and late indica is 75.60%, 72.68% and 53.32%, respectively. Therefore, although there is heterogeneity in different rice varieties, the policy is in general effective in increasing rice acreages given that global rice prices increased after 2004.

Keywords

Acreage response, price intervention, expected price, price volatility, Monte Carlo, China

1. Introduction

The agricultural challenges for China include maintaining growth of farmers’ income, achieving sustainable agricultural development and ensuring national food security (Huang and Yang, 2017). With only 6% of fresh water and 7% of arable land in the world, China has to feed nearly 20% of the world’s population (Wong and Chan, 2016). The population in China reached 1.39 billion in 2019 (World Bank, 2020). Therefore, ensuring food security, especially grain security, is important for economic development and social stability (Li et al., 2013).

To increase national food security and increase farmers’ income, China’s policymakers have implemented several policies. The first set of policy measures consists of the abolition of taxes and fees and introduction of subsidy programs since 2004. To promote grain and other major crop production, China introduced the temporary storage program for maize, rapeseeds, and soybean since 2008. A price intervention program was introduced to ensure the minimum procurement price for rice since 2004 and wheat since 2006 (Huang and Yang, 2017).

1 This paper focuses on rice support policies since rice is the main staple food for more than half of the population. China consumes more rice than any other country, with 143.8 million metric tons in 2018 (USDA, 2018). Due to increased grain yield and improved crop management practices such as fertilization and irrigation, rice production in China has tripled in the past three decades (Peng et al., 2009). However, there was yield stagnation in the past ten years. Together with increasing population, China will need to produce around 20% more rice by 2030 to meet the domestic consumption (Peng et al., 2009).

Apart from the yield stagnation and increasing population, from 1976 to 2004, there was a declining trend in rice acreages in China (Ricepedia, 2020). The continuously decreasing rice acreages in combination with China’s desire for national food security triggered the introduction of policy interventions since 2004. After 2004, rice acreages stabilized and even increased by 8.3% in the period 2004-2017 (NDRC, 2017). However, at that time international rice prices were also on the rise, culminating in the 2007/08 price peaks. It is not clear whether the increasing rice acreage is due to the policy intervention or due to the increasing domestic rice price in line with the international rice price. In other words, the question is whether this policy was effective given that rice prices increased after 2004.

The objective of this paper is to investigate whether the rice support policy had a positive effect on rice acreages in China given the rising global rice prices. In order for the minimum procurement price policy to be effective, there are two questions to be answered. First, did rice acreages before and after 2004 relate to rice prices? If rice acreages did not respond to prices, we would not expect the policy to be effective at all. Moreover, we also examine whether the relation between acreages and prices was different before and after 2004, since the minimum procurement price policy may have reduced price risks and make farmers’ planting decisions more responsive to prices. This will be analysed with a fixed effects panel data model using acreage data and domestic prices for 3 main rice varieties (early, middle, and late indica rice) for 27 Chinese provinces for the period 1988-2017. Second, was there a positive impact of the minimum procurement price policy on rice acreages? To analyse this, we need to construct counterfactual acreages after 2004. These counterfactuals are based on international rice prices. A few econometric studies show the dependence of Chinese rice prices on international prices. Using monthly time series data from 2005 to 2013, Lee and Valera (2016) found a small positive and statistically significant effect of lagged world rice prices on Chinese domestic rice prices. Moreover, their analysis shows absence of Chinese rice price volatility spillovers to global prices, mainly because the Chinese prices are rather stable. Using cointegration analysis and an error-correction approach, Huang et al. (2014) also found that Chinese domestic rice prices were affected by international rice prices in the period 2003-2006. The long-run transmission of international rice price to China’s domestic prices faded during the global financial crunch (2007-2008), but after that, the domestic Chinese rice price in relation to the international trend returned to levels that had existed prior to the global food crisis (Huang and Yang, 2014). Based on the findings in the literature, we assume that Chinese rice prices would have closely followed international rice prices in the absence of domestic Chinese policies as they did up to 2004 (Huang and Yang, 2017). Therefore, we construct counterfactual acreages based on the international prices. Recognizing the inherent

2 uncertainty in constructing counterfactuals, we perform Monte Carlo simulations to include the market volatility and derive simulated counterfactual acreages that are compared with observed acreages to analyse the impact of the minimum procurement price policy.

Some literature studying the effectiveness of the price supporting policies focused on grain subsidy programs. Huang et al. (2011) concluded that grain subsidy programs did not distort farmers’ decisions in terms of grain area or input use decisions. Yi et al. (2015) examined the effects of the grain subsidy programs on grain-sown areas. By introducing households’ liquidity constraints, they concluded that liquidity-constrained households were likely to distribute more land to grain production while households that are not liquidity-constrained were not significantly affected by the grain subsidy. There are other measures within the price supporting domain, such as minimum procurement price, which also play an important role. A few recent papers studied the effectiveness of price supporting policies in a broader sense. Li and Chavas (2018) used quantile auto-regression to investigate the effects of China’s price support policy on price enhancement and price stabilization in the Chinese rice and corn markets. The results show that the price support policy increased the price of corn and contributed to stabilizing the domestic rice market. Wang and Wei (2019) developed an aggregate structural econometric model of China’s soybean market to analyse the worldwide impacts of China’s soybean price support policies from 2008 to 2016. The results show that the soybean price support policies played an effective role in stabilizing the domestic price and the welfare distribution. Lyu and Li (2019) used a regime-switching model to evaluate the effectiveness and sustainability of the grain price support policies in China. They found that there is a structural change since 2004 when the price support policies were established, and since then, Chinese grain prices have followed a regime with significantly lower volatility.

So far, the literature focuses on the effectiveness of price supporting policies in stabilizing domestic grain prices. As far as we are aware, there is not much research on evaluating the effectiveness of price supporting policies on the acreage through price responsiveness of farmers. Supply response cannot be simply tested by acreage change before and after the policy intervention since structural change does not necessarily mean the policy takes effect. Even though the acreage does not increase significantly after the policy intervention, the policy may still be effective since acreage could have even decreased more without the policy. Considering the increasing global price after 2007/08 due to the global food crisis, it is even more unclear whether the acreage change is due to the policy intervention or due to the increasing international rice price. Therefore, our study investigates whether the Chinese rice supporting policies were effective given that global rice prices increased after 2004

2. Background information on Chinese rice policies and acreages

Different regions have different climate leading to different rice planting practices. The paddy rice planting in China can be distinguished into four main kinds: early, middle, and late indica rice, and . Early indica grows primarily in provinces along the River and in provinces in the south. The early indica is planted from February to April, and harvested in

3

July. The taste of early indica rice is inferior to other rice, and therefore, it is mainly processed. Middle indica grows primarily in the southwest, and is planted from March to June, and harvested in October. Late indica is planted after the early indica is reaped inthe same regions and is harvested in November. Late indica needs relatively more time to ripen due to the cold weather. Because of the difference in temperature between day and night in late autumn, late indica is considered as good-quality rice since it contains higher nutrition value (ChinaBaoGao, 2019). The other popular rice is japonica, most of which is planted in the northern part of China. Figure 1 shows the regional distribution of different kinds of rice in 2017.

Figure 1. Distribution and planting density of four rice varieties in 2017 Data source: Authors, based on Ning and Hu (2017) Note: The acreage data for japonica is not available. Therefore, the coloured area for japonica only means where it is planted based on the reported market price of japonica in 2017.

The annual total acreage for rice1 was 30 million hectares in China in 2019, which is 18% of the world’s rice acreage (FAOSTAT, 2020). Three-quarters of the acreage in China is planted with indica varieties, and the rest with japonica varieties. More than 40% of acreage is planted with high-quality rice due to the improvement of living standard and increasing demand for high-quality rice (Peng et al., 2009). Up to 2004, the acreages of main rice varieties in China kept decreasing by 17.4% from 34.4 million hectares to 28.4 million hectares (NDRC, 2017). To stabilize the total rice acreage, the government established various policy interventions.

Policy interventions include different kinds of subsidies, starting with the “direct grain subsidy”, “quality seed subsidy”, and “machinery subsidy” in 2004, extended to the “aggregate input subsidy” in 2006. As their names indicate, these direct subsidy programs subsidize planting grain and the costs of buying quality seed, machinery and inputs, such as chemical and fuel (Huang and Yang, 2017). However, according to Huang and Yang

1 For the area planting double-seasoned rice (early and late indica), the acreage is counted twice.

4

(2017), these subsidies had negligible impact on grain production since the subsidies are mostly given to the land contractor instead of tiller due to the identification difficulty.

Important policy interventions include the temporary storage program for maize, rapeseeds, and soybean since 2008, and the price intervention program to ensure a minimum procurement price, which has been implemented for rice since 2004 (Gale, 2013; Huang and Yang, 2017) and for wheat in 2006 (Lyu and Li, 2019). The Chinese government set the minimum procurement price for rice to ensure that it is high enough to cover the production cost and earn a profit. The minimum procurement price is valid in the main production provinces (Lyu and Li, 2019), and is set annually for different kinds of rice based on the production cost, market demand and supply, as well as prices at home and abroad. The minimum procurement price is announced in January before rice is planted. Purchases are made by state-owned grain enterprises only when the market price is below the minimum value. Purchased rice is kept in storage until it can be auctioned at a grain exchange at a higher price. The Chinese government subsidizes storage and operation costs (Gale, 2013). Figure 2 shows the trend of the minimum procurement price for early, middle, late indica and japonica in China between 2004 and 2020. The minimum procurement price has been gradually increased till 2014, maintained in 2015, and decreased since 2016 (NDRC, 2020).

Figure 2. The minimum procurement price for early, middle, and late indica and japonica Data source: Authors, based on NDRC (2020)

3. Data and methodology

3.1 Description of the data

The data used for the analysis are yearly provincial acreages (1988-2017) from Ning and Hu (2017), provincial market prices of early, middle and late indica2 (1975-2017) in 27 provinces in China (see Appendix 1) from the National Development and Reform Commission (NDRC)

2 The provincial acreage for japonica is not available, and therefore, we only focus on the early, middle and late indica in the study.

5

(2017) and the yearly Thai international rice price (1978-2017) from the World Bank. The unit of acreage is thousand hectares and the unit of domestic and international rice price is RMB (Chinese currency) per 50 kilograms adjusted by the Consumer Price Index with base year 1978. Figure 3 shows the development of provincial acreages for early, middle and late indica rice. As observed, acreages in general were decreasing before 2004, especially for early and late indica. After 2004, the acreages stabilized and in some provinces even increased. However, since market prices also increased after 2004 in line with international rice prices, it is not clear whether the acreage increase was due to the policy intervention or due to the increasing market prices.

2000 3000

1500 2000

1000 1000

500

acreageearly of indica acreagemiddle of indica

0 0

1990 2000 2010 2020 1990 2000 2010 2020 year year

Zhejiang Anhui Fujian Hunan Chongqin

2000

1500

1000

500

acreagelate of indica

0

1990 2000 2010 2020 year

Zhejiang Anhui Fujian Jiangxi Hubei Hunan Guangdong Guangxi Hainan

Figure 3. Provincial acreages of early, middle and late indica in China from 1988 to 2017 Data source: Ning and Hu (2017)

25 25

20 20

15 15

10 10

real market priceearly of indica(1978) 1990 2000real market pricemiddle of indica (1978) 2010 2020 1990 2000 2010 2020 year year

Zhejiang Anhui Anhui Fujian Fujian Jiangxi Hunan Jiangsu Hubei Hunan Henan Yunnan Guangdong Guangxi Shaanxi Guizhou Hainan Sichuan Chongqin

6

30

35 30

25

25 20

20

15 10

15 1990 2000 2010 2020 real market pricejaponica of (1978) year

10

real market pricelateindica of (1978) Zhejiang Anhui 1990 2000 2010 2020 Hubei year Neimenggu Zhejiang Anhui Jilin Fujian Jiangxi Jiangsu Hubei Hunan Henan Yunnan Guangdong Guangxi Tianjing Hainan

Figure 4. Provincial market prices of early, middle, late indica and japonica in China, 1988-2017 Data source: NDRC (2017)

Table 1 provides descriptive statistics of data in our study. The mean is followed by the standard deviations in parentheses. The row below provides a breakdown of the standard deviation in the between standard deviation (given in italic) and the within standard deviation (underlined). Within variation captures the variation over time while between variation captures the variation across provinces. Average real price and average expected price3 for all kinds of rice were higher after 2004. Although in general, volatility after 2004 stabilized (Li and Chavas, 2018; Wang and Wei, 2019; Lyu and Li, 2019), Table 1 shows average expected volatility of early and middle indica was higher after policy intervention than before while average expected volatility of late indica was lower. This is due to the extreme values during global food crisis. Average rice acreage of middle indica was higher while average rice acreage of early and late indica was lower after 2004. For average real price, average expected price and average expected volatility, the between variation is smaller than within variation in the dataset, indicating there are less variation between different provinces than variation of individual provinces throughout the years. However, between variation of average rice acreage is larger than within variation.

Table 1. Descriptive statistics of data, 1975-2017 Early indica Middle indica Late indica Before 2004 Real price 12.33 (2.74) 12.76 (2.55) 13.84 (3.40) (1.42-2.46) (0.69-2.49) (1.02-3.27) Expected price 12.73 (3.05) 13.10 (2.69) 14.26 (3.87) (1.55-2.75) (0.87-2.62) (1.18-3.70) Expected volatility 1.58 (1.07) 1.58 (1.03) 1.93 (1.34) (0.42-1.00) (0.36-1.02) (0.40-1.28) Acreage 869.68 (488.91) 1019.84 (801.89) 941.56 (526.21) (489.58-156.46) (812.21-172.30) (531.28-155.69) After 2004 Real price 19.54 (2.72) 19.80 (2.90) 20.87 (3.14)

3 Please refer to section 3.2 about the definition of expected price and expected volatility.

7

(1.13-2.50) (1.31-2.66) (1.55-2.78) Expected price 19.18 (2.66) 19.49 (2.95) 19.65 (2.80) (0.06-2.66) (0.36-2.93) (0.04-2.80) Expected volatility 1.73 (0.78) 1.77 (0.82) 1.74 (0.85) (0.10-0.77) (0.24-0.79) (0.13-0.84) Acreage 634.28 (499.09) 1063.33 (682.72) 692.20 (520.65) (524.38-52.14) (713.56-67.57) (547.50-49.94) Source: Authors, based on Ning and Hu (2017) and NDRC (2017).

3.2 Model specification

Acreage response has been widely studied in literature (Weersink et al., 2010; Chavas et al., 1983) that prices, risks, and policies play an important role. Nerlove (1958) argues that farmers react not to last year’s price but rather to the expected price. Similar to the models developed by Haile et al. (2016), we specify a fixed effects panel data model to analyse the provincial acreage responses to changes in farmers’ expected price, expected volatility, and time trend capturing the technological changes. We do not consider inputs in the model since the input subsidy has negligible impact on production (Huang and Yang, 2017). We focus on rice acreages instead of production since production is directly affected by other factors, such as weather, which can influence the production but actually not related to farmers’ decision. Therefore, acreage is a better indicator capturing farmers’ responses to the policy.

e acreagePtntntntnnt0123  (1)

where n denotes the provinces, t denotes year,  n denotes the provincial fixed effects, and

 nt is the residual. We include time t in the model to control factors that change through time, such as industrialization and urbanization. Due to difference in soil type between rice and other crops, as well as farmers’ inherent planting practice inherited from the past (personal communication), it is difficult for rice farmers to switch to other crops. Therefore, we do not e include prices of competitive crops in the model but only the expected price of rice Pnt . Shonkwiler and Maddala (1984) argue that the presence of a price intervention program should directly affect farmers’ expectations. Therefore, expectations should be conditioned by both market conditions and the intervention. In this study, we include the information of the past five years. Nerlove (1958) argues that although in theory all past prices are supposed to be included, we can ignore prices in the very distant past. The expected price can be defined as a weighted function of past prices with declining weights. In this study, we define the j weights using an exponential format, j ,j 1,2,3,4,5 (Richardson et al., 1998, Bollen, 2015). Since the weights should sum up to 1 (+ 2345 +  +  +  =1), we calculate   0.51. Besides the weighted lagged prices, the expected price is also based on the announced minimum procurement price. Combining both elements, the expected price can be expressed in equation (2) which allows for the possibility that the expected market price can be either higher or lower than the announced minimum procurement price.

8

5 ejmin ppjj 1 pppttt max,11   (2)  p j1 j1

 nt denotes expected volatility, which is the rolling volatility of the real rice price with the same weight of volatility in the past five years. We include price volatility to capture its impact on acreages since Haile et al. (2016) argue that although higher prices serve as an incentive to improve supply, price volatility acts as a disincentive.

For the model in equation (1), we are interested in the parameter 1 since it indicates whether the expected price has a significant positive impact on the acreage response. Price responsiveness is a precondition for the policy to be effective, i.e. in maintaining (or even increasing) the rice acreage. If farmers do not respond to expected prices in their acreage decisions, a minimum procurement price policy is not expected to work. However, if farmers 4 do respond to expected prices, we cannot conclude that the policy is necessarily effective . 3 indicates whether the expected volatility has a significant negative impact on the acreage response.

Equation (3) allows for testing whether the acreage responds differently to expected prices before and after the policy intervention:

ee acreagent0   1 P nt   2 t   3 P nt D   4 D   5  nt   n   nt (3) where D is a dummy variable which equals 1 after 2004 and equals 0 otherwise, which we e include for a better fit of the model. PDmt is the interaction between the dummy variable and the expected price. We are interested in the coefficient 3 , which indicates whether the acreage responds differently to expected prices before and after the policy intervention .

3.3 Monte Carlo simulations for analysing the policy impact

To assess the impact of the policy on the acreage response, we perform Monte Carlo simulations to construct counterfactual acreages as if there were no minimum procurement price policy since 2004. The advantage of Monte Carlo simulations is to include the inherent uncertainty in constructing counterfactuals. Monte Carlo simulations consider thousands of random routes that form a distribution. By doing the simulations, although we do not know the exact counterfactual acreages when there were no policy intervention, we are able to estimate the probability of the effectiveness of the policy considering uncertainty of domestic price and the evolution of international price. Equation (4)5 is the simple model based on which we simulate acreages with expected prices.

4 The extreme case would be farmers do respond to expected prices, but the minimum procurement price is way below the market price each year. 5 When constructing the counterfactual acreage, we use the simplest model only with the explanatory variable of expected price.

9

**e acreagePntnnt,20031 (4)

where n,2 0 0 3 is the intercept for each province in 2003 which is the year before the policy intervention. 1 captures the general relationship between acreages and expected prices, and is based on random draws from the confidence interval of the parameter estimate of 1 in equation (1) regressed with the data before 2004. The random draws are generated by a PERT distribution using the maximum, minimum, and mode from the real occurrence happened in the past to generate a distribution that more closely resembles a realistic probability distribution (Davis, 2008) (see Appendix 2). Better than other similar distributions (e.g. triangular distribution), the PERT distribution constructs a smooth curve with the expectation e* that the resulting value is around the most likely value. Pnt is the simulated expected price based on equation (2) calculated with data before 2004. The detailed steps for generating the set of simulated farmers’ expected price and acreages between 2004 and 2017 are as follows.

First, with the data before 2004, we generate 10,000 random draws for each province with a PERT distribution, based on the difference between the expected price of one year and the previous year. This provides the dispersion of the volatility for simulating the expected prices in the next step.

Second, for each province between 2004 and 2017, expected prices are generated using the ee** formula PPtntn tnt,1 , that is by adding the simulated expected price from the previous year, a drift from the international price, and the volatility of the expected prices before 2004 in the first step.  is the drift of the international price after 2004 and is regressed with the ww w w formula PPtttt 1 , where Pt and Pt1 denote the international rice price from the current year and previous year, respectively, and  t is the residual. The real provincial expected prices in 2003 are used as the starting points of the simulation. This generation mechanism guarantees the resemblances of a drift from the post-policy period based on the international price and the volatility from the pre-policy period6.

Third, simulated acreages for each province between 2004 and 2017 are generated by e* plugging in n,2003 , 1 , and Pnt in equation (4), where is the provincial intercept e calculated based on equation (4) (nnn,2003,20031,2003acreageP ).

Finally, we test whether there is a significant difference on average between the observed acreages and simulated acreages between 2004 and 2017. If there is, we can conclude that the minimum procurement price policy was effective. The effectiveness is tested by assessing the estimated parameters of equation (5):

6 We realize this might be a strong assumption since the policy intervention actually dampened the volatility. On the other hand, however, international prices were more volatile after 2004 than before, so that this pre-policy volatility could be an underestimate.

10

diff * acreagetntnt 01 (5)

diff * where a c r e a g ent is the difference between the observed acreages and the simulated acreages for each province in each year between 2004 and 2017. We are interested in the overall intercept for the within model  0 indicating the effectiveness of the policy. If  0 is positive and significantly different from zero, we can conclude that the observed acreages are on average significantly larger than the counterfactuals, and therefore, the policy intervention has been effective. The parameter t indicates whether the difference between observed acreages and simulated acreages has changed over time.  nt is the residual.

The second to fourth step described above are included in a loop. The same process is e* repeated 10,000 times but with 10,000 different random draws of n,2 0 0 3 , 1 and Pnt , resulting in 10,000 simulated acreages. We compare the observed acreages with simulated acreages 10,000 times and are interested to see how many times out of the total number of loop, observed acreages are significantly larger than simulated acreages. The calculated percentage can be interpreted as the probability of effective policy intervention.

4. Results and discussion

4.1 The effect of expected prices and volatility on acreages

Equation (1) and equation (3) were estimated for acreages of 27 rice-producing provinces in China between 1988 and 2017. Table 2 shows parameter estimates and test statistics.

Table 2. Factors related to the acreage response (dependent variable) in China, 1988-2017. Variables Fixed effects model Fixed effects model Fixed effects model (early indica) (middle indica) (late indica) Equation 1 Equation 3 Equation 1 Equation 3 Equation 1 Equation 3 Expected price 14.24*** 10.14*** 11.49*** 9.88* 10.95*** 5.07* (2.67) (3.78) (4.78) (7.03) (3.07) (3.90) Expected volatility -23.72*** -16.12* -34.23*** -32.97*** 3.89 12.98 (8.92) (9.94) (13.71) (14.85) (9.60) (10.11) Trend -21.48*** -22.43*** -5.59*** -7.63*** -18.64*** -18.90*** (1.27) (2.05) (2.34) (3.43) (1.56) (2.48) Interaction term 11.52** 1.33 17.06*** (expected price and period dummy) (6.72) (10.16) (6.56) Period dummy -186.03** 24.57 -312.85*** (105.57) (157.91) (106.26) Constant 1195.31*** 1258.93*** 1119.66*** 1178.52*** 1174.72*** 1255.66*** (35.58) (68.04) (57.09) (118.60) (43.07) (76.94) R2 (within) 0.60 0.60 0.04 0.04 0.52 0.54

11

F-test joint 122.48*** 74.44*** 3.24*** 2.09* 86.8*** 55.44*** significance F-test unit-specific 447.72*** 448.24*** 414.08*** 411.21*** 426.06*** 437.77*** effects N 259 259 253 253 251 251 Note: Standard errors are reported in parentheses. *, **, and *** indicates statistical significance in a one-tailed test at 10%, 5%, and 1% level, respectively.

For early, middle and late indica, expected prices have a statistically significant positive effect on acreages. In other words, the acreages respond to farmers’ price expectations for all kinds of rice. Price responsiveness is a precondition for the policy to be effective in maintaining, or even increasing the rice acreages. Therefore, it is reasonable that we later test the policy effectiveness for all three kinds of rice.

For early and middle indica, expected volatilities have a statistically significant negative effect on acreages, which confirms that price volatility acts as a disincentive to improve supply. Interestingly, the expected volatility of late indica does not have an effect on acreages. For early and late indica, there is a statistically significant positive change in acreage response to expected prices after 2004 than before. Although middle indica also has a small positive change in acreage response to expected prices after 2004, it is not significantly different from zero at 10% significance level. The time trend has a statistically significant negative impact on the acreages for all three kinds of rice because there is less and less arable land due to industrialization and urbanization. The within R2 values are 0.60 and 0.54 for early and late indica, respectively; however, it is only 0.04 for middle indica indicating the explanatory power of the model for middle indica is low, which might be caused by omitted explanatory variables from competitive rice varieties of middle indica (e.g. japonica rice) ignoring the substitution effect.. Further research is needed to study the heterogeneity of different rice varieties. The null hypothesis that all slope parameters are jointly equal to zero is firmly rejected for all kinds of rice. The use of a fixed effects estimation approach is justified by the outcome of the F-tests on the province-specific effects in all cases. The null hypothesis that these effects are all similar (absence of constant regional differences) is firmly rejected in all cases.

4.2 Monte Carlo results of effective policy intervention

Table 3 shows the distribution of the results of 10,000 Monte Carlo simulations for different quintiles.

Table 3. Distribution of the results of Monte Carlo simulations Min. 1st Qu. Median Mean 3rd Qu. Max.

Early Intercept ( 0 ) -97.52 0.61 25.54 25.43 50.15 146.78 indica

Slope (1 ) -17.96 -7.08 -5.10 -5.11 -3.14 5.21 Number of significant 7,560 out of 10,000 positive intercepts

12

Middle Intercept ( 0 ) -216.98 -5.42 47.26 46.54 100.16 328.09 indica

Slope (1 ) -23.42 -1.98 1.95 2.00 6.03 23.03 Number of significant 7,268 out of 10,000 positive intercepts

Late Intercept ( 0 ) -194.24 -21.73 2.99 2.08 27.20 125.9 indica

Slope (1 ) -17.28 -6.11 -4.19 -4.26 -2.37 5.76 Number of significant 5,332 out of 10,000 positive intercepts

Table 3 shows that on average, both early and late indica have positive intercepts and negative slopes, indicating that on average, observed acreages are larger than simulated acreages with a decreasing acceleration through the time. Therefore, the policy intervention was in general effective while its effect faded through time. Interestingly, middle indica has both positive intercept and slope, indicating on average, observed acreages are larger than simulated acreages with an increasing acceleration through the time. Calculating with the number of significant positive intercepts out of 10,000 loops, Monte Carlo simulations recognize the inherent uncertainty in constructing counterfactuals and provide us with a distribution of p values resulting in a probability of effective policy intervention instead of a deterministic yes- no answer. We conclude that there is a probability of 75.60%, 72.68%, and 53.32% that the minimum procurement price policy was effective for increasing the acreages of early, middle, and late indica, respectively, given that global rice prices increased after 2004.

5. Conclusions

In the past decades, the acreages of main rice varieties in China kept decreasing from 34.4 to 28.4 million hectares in 2004. This induced the Chinese government to implement various price interventions. The price support policy is assumed to have increased farmers’ profits by setting a minimum procurement price. This has led to concerns on whether the policy had a positive effect on rice acreages in China, or the observed rise in rice acreages is simply due to the increase of market prices after 2004.

Our study shows that there exists price responsiveness from rice farmers planting early, middle, and late indica. For early and late indica, there is a statistically significant positive change in acreage response to expected priced after the policy intervention in 2004. Based on generating counterfactuals by performing Monte Carlo simulations, we conclude that in general rice supporting policy was effective in increasing rice acreages given that global rice prices increased after 2004. The probability of effective policy for early, middle, and late indica is 75.60%, 72.68% and 53.32%, respectively.

Policy interventions are widely used by countries worldwide to reach political and economic targets. Assessing the effectiveness of policy intervention is important for stakeholders. While

13 counterfactuals are not observable, assessing the policy intervention without known counterfactuals can be broadly considered as a missing data problem. In this paper, we demonstrate a methodological innovation for generating counterfactuals based on computational simulations and estimating the probability of effective policy interventions, which can be applied to a wider context where counterfactuals are unknown.

However, further research is needed in explaining the heterogeneity of different rice varieties, such as middle indica, which might be due to substitution effects of japonica that we lack data in this study. When constructing the counterfactual acreage, we use the simplest model only with the explanatory variable of expected price, which is constructed based on a constant international drift. Further studies may expand the model by including other explanatory variables and by constructing the counterfactual expected price based on the international drift incorporating time-dependency.

References

Bollen, B. (2015). What should the value of lambda be in the exponentially weighted moving average volatility model? Applied Economics 47 (8): 853-860.

Chavas, J. P., Pope, R. D. and Kao, R. S. (1983). An analysis of the role of futures prices, cash prices and government programs in acreage response. Western Journal of Agricultural Economics, 27-33.

Davis, R. (2008). Teaching Project Simulation in Excel using PERT-Beta distributions. INFORMS Transactions on Education 8 (3): 139-148.

FAOSTAT. (2007). FAO Statistical databases. www.fao.org. Food and Organization (FAO) of the United Nations, Rome.

Gale, F. (2013). Growth and Evolution in China’s Agricultural Support Policies. Economic Research Report Number 153. Washington, D.C.: USDA - Economic Research Service.

Haile, M. G., Kalkuhl, M. and von Braun, J. (2015). Worldwide acreage and yield response to international price change and volatility: a dynamic panel data analysis for wheat, rice, corn, and soybeans. American Journal of Agricultural Economics 98 (1): 172-190. Huang, J., Wang, X., Zhi, H., Huang, Z. and Rozelle, S. (2011). Subsidies and distortions in China’s agriculture: evidence from producer-level data. Agricultural and Recourse Economics 55: 53-71. Huang, J. and Yang, J. (2014). The political economy of food price policy in China. Food Price Policy in an Era of Market Instability: A Political Economy Analysis. UNU-WIDER Studies in Development Economics, Oxford University Press, Oxford, http://dx.doi.org/10.1093/acprof:oso/9780198718574.001.0001

Huang, J. and Yang, G. (2017). Understanding recent challenges and new food policy in China. Global Food Security 12 (2017): 119-126.

14

Lee, J. and Valera, H. G. A. (2016). Price transmission and volatility spillovers in Asian rice markets: Evidence from MGARCH and panel GARCH models. The International Trade Journal 30 (1): 14-32.

Li, J. and Chavas, J. P. (2018). How Have China’s Agricultural Price Support Policies Affected Market Prices? A Quantile Regression Evaluation. Paper presented at the 30th International Conference of Agricultural Economists, Vancouver, Canada, July 28 – August 2, 2018.

Li, Y., Zhang, W., Ma, L., Wu, L., Shen, J., Davies, W. J., Oenema, O., Zhang, F. and Dou, Z. (2013). An analysis of China’s grain production: looking back and looking forward. Food and Energy Security 3 (1): 19-32.

Lyu, J. and Li, X. (2019). Effectiveness and sustainability of grain price support policies in China. Sustainability 2019 11, 2478; doi:10.3390/su11092478

National Development and Reform Commission (NDRC). (2017). Compilation of Costs and Benefits of National Agricultural Products. China Statistics Press.

National Development and Reform Commission (NDRC). (2020). The minimum procurement price of rice in 2020. Available at: https://www.ndrc.gov.cn/fzggw/jgsj/jgs/sjdt/202002/t20200228_1221748.html (accessed on February 29th, 2020)

Ning, Q. and Hu, L. (2017). China Agricultural Statistical Yearbook, China Agriculture Press. Nerlove M. 1958. Adaptive Expectations and Cobweb Phenomena. The Quarterly Journal of Economics 72 (2): 227–40.

Peng, S., Tang, Q. and Zou, Y. (2009). Current status and challenges of rice production in China. Plant Production Science 12 (1): 3-8.

Ricepedia (2020). Available at: http://ricepedia.org/china (accessed on February 27th, 2020)

Richardson, M.P. and Boudoukh, J. and Whitelaw, R.F. (1998). The Best of Both Worlds: A Hybrid Approach to Calculating Value at Risk. Available at: https://ssrn.com/abstract=51420 (accessed on February 1st, 2020 )

Schonkwiler J.S. and Maddala G.S. (1985). Modelling expectations of bounded prices: an application to the market for corn. The Review of Economics and Statistics 67 (4): 697-702.

United States Department of Agriculture (USDA) (2018). Available at: https://www.fas.usda.gov/commodities/rice (accessed on February 27th, 2020)

Wang, W. and Wei, L. (2019). Impacts of agricultural price support policies on price variability and welfare: evidence from China’s soybean market. Center for Agricultural and Rural Development Publications at Iowa State Univeristy, Working paper 19-WP 592, available at: https://lib.dr.iastate.edu/cgi/viewcontent.cgi?article=1614&context=card_workingpapers

15

Weersink, A., Cabas, J. H. and Olale, E. (2010). Acreage response to weather, yield, and price. Canadian Journal of Agricultural Economics/Revue canadienne d'agroeconomie, 58(1), 57-72.

Wong, A.Y. and Chan A.W. (2016). Genetically modified foods in China and the United States: a primer of regulation and intellectual property protection. Food Science and Human Wellness 5: 124-140.

World Bank. 2020. Available at: https://data.worldbank.org/indicator/SP.POP.TOTL?locations=CN (accessed on Januray 11, 2020)

16

Appendix 1. Provincial distribution of rice acreages

The provinces planting early and late indica include Zhejiang, Anhui, Fujian, Jiangxi, Hubei, Hunan, Guangdong, Guangxi, Hainan, , Sichuan, Guizhou, and Yunnan Province. However, our study only focuses on the first nine provinces and does not include the last four provinces since their market prices are not available.

The provinces planting middle indica include Jiangsu, Anhui, Fujian, Henan, Hubei, Hunan, Chongqing, Sichuan, Guizhou, Yunnan, Shaanxi, , , Hebei, , Neimenggu, Liaoning, Jilin, Heilongjiang, Shanghai, Zhejiang, Jiangxi, Shandong, Guangdong, Guangxi, Hainan, Xizang, , Ningxia and Province. However, our study only focuses on the first eleven provinces and does not include the rest since their market prices are not available.

We use number 1-27 to denote 27 provinces in the study. n=1, Zhejiang; n=2, Anhui; n=3, Fujian; n=4, Jiangxi; n=5, Hubei; n=6, Hunan; n=7, Guangdong; n=8, Guangxi; n=9, Hainan; n=10, Hebei; n=11, Neimenggu; n=12, Liaoning; n=13, Jilin; n=14, Heilongjiang; n=15, Jiangsu; n=16, Shandong; n=17, Henan; n=18, Yunnan; n=19, Ningxia; n=20, Tianjin; n=21, Shanghai; n=22, Shanxi; n=23, Beijing; n=24, Shaanxi; n=25, Guizhou; n=26, Sichuan; n=27, Chongqing.

17

Appendix 2. Distributions of deviation of expected prices (as example: early indica)

Zhejiang Province Anhui Province

Fujian Province Jiangxi Province

18

Hubei Province Hunan Province

Guangdong Province Guangxi Province

19

Hainan Province

We use the deviation of expected prices before 2004 to construct the counterfactuals after 2004 based on a PERT distribution. The PERT distribution needs three parameters: maximum, minimum, and mode, which we obtain from the distributions based on the real data before 2004.

20