Does rural financial development facilitate food production? Evidence from major wheat–producing provinces of China

Abstract

The allocation of agricultural credit is an exceptional strategy for ensuring sustainable agricultural production. It increases farmers’ purchasing power and technical efficiency. In this study, we examine the long-run impact of agricultural credit on wheat production in nine major wheat—producing provinces of China from 1992–2020. The research also takes into account agrochemical inputs (such as fertilizer and pesticide), planting area, agricultural machinery, and rural labor force. Several advanced econometric techniques, such as cross—sectional dependence (CSD), Westerlund co—integration (second—generation approach), Feasible Generalized Least Squares (FGLS), Dynamic Ordinary Least Squares (DOLS), One—step Generalized Method of Moments (GMM), and panel causality tests, are utilized to conduct empirical estimation. The findings show that agricultural credit significantly increases wheat production. It has been proved that agricultural credit is an important factor in increasing grain production and ensuring food sustainability. Regarding agrochemical inputs, fertilizer use positively affects wheat production, whereas pesticide usage negatively affects wheat production. Other determinants influencing wheat production included planting area, agricultural machinery, and rural labor force. China’s agricultural credit and related policies have played a significant role in grain production since the country’s reform and opening up. China must consolidate poverty alleviation achievements and further develop the countryside under the current rural revitalization and food security strategies. Agricultural credit is essential in this process and should be expanded.

Keywords:

• Rural credit
• grain production
• food security
• FGLS approach
• China

1. Introduction

In many developing countries, escaping poverty is dependent on agricultural expansion and development (Ahsan et al., 2020; Dai et al., 2022). Agricultural expansion and development are unfeasible without yield—enhancing technological options because increasing cropland is no longer sufficient to meet the rising food demands of growing populations (Rehman et al., 2017). As a result, research and technological advancements are critical for increasing agricultural output and reducing poverty while preserving the agroecosystems that sustain livelihoods (Chandio et al., 2017; Enimu et al., 2017). In this context, financial services allow farmers to invest in and adopt new agricultural innovations, thereby enhancing agricultural output. This program allows farmers in need to purchase inputs such as seeds, fertilizers, and agrochemicals. For the agricultural sector to become more productive, it is necessary to make financial services accessible and affordable (Saqib et al., 2018). The experience of developed agricultural nations (Aryal et al., 2018) demonstrates the significance of adopting and employing advanced technology in agricultural operations. Agriculture research, agricultural loans, irrigation systems, highways, and storage facilities all contribute significantly to the improvement of technologically advanced agricultural inputs (Ali Chandio et al., 2022; Brizmohun, 2019).

The agriculture industry is the most dependent on credit among all economic sectors. Since farmers’ incomes fluctuate seasonally, shifting from subsistence to commercial farming, agricultural credit facilities are crucial for agricultural and rural development growth. Credit can be used to modernize agriculture and accelerate economic growth, but only if it is easy accessible, and affordable (Chandio et al., 2017). This would suggest that the demand for formal agricultural financing would increase as a result of the increased agricultural output. All farmers, but especially small and large farms, must have access to official agricultural credits (Li et al., 2011). It is crucial for economic growth and living standards in less developed rural regions (Bashir & Mehmood, 2010; Kassouri & Kacou, 2022). Access to credit is crucial for smallholders in developing nations. All of this is linked to farmers’ overall output and the increase in agricultural output per unit of input. Rural credit is essential in every aspect of agricultural operations. It is in high demand in many parts of the world, primarily for the capital required to improve and support farms with agricultural supplies and labor payment (Bahşi & Çetin, 2020; Chandio et al., 2017; Omoregie et al., 2018). In addition, the availability of credit eliminates the financial constraints associated with cash inputs, improves the technical efficiency of farmers, and increases resource allocation and profitability (Bahşi & Çetin, 2020).

Due to the lack of savings by farmers, there is a significant reliance on official and informal credit sectors in emerging nations. According to the production and financial structure hypothesis, it is possible to increase the overall output of low-income farm families if they have access to credit (Elahi et al., 2018). Since smallholder farmers lack adequate financial understanding, collateral, and significant institutional barriers, the agricultural sector has seen the emergence of informal loan channels (Hartarska et al., 2015; Kassouri & Kacou, 2022). Credit constraints may have both direct and indirect effects, directly reducing farmers’ purchasing power and indirectly influencing farmers’ risk-averse behavior, preventing them from investing in modern technology or taking other risks (Duong & Thanh, 2019; Rabbany et al., 2022; Saqib et al., 2016; Shew et al., 2019). Agriculture funding output, primarily through loans to small-scale farmers, remains the key to agriculture-induced macroeconomic growth (Chandio et al., 2018; Saqib et al., 2018).

Relevant economic research emphasizes three aspects of the theoretical significance of the relationship between credit and agricultural production. First, access to financial markets mitigates the risks and uncertainties associated with agricultural production and significantly sustains productivity growth (Eswaran & Kotwal, 1986). Second, informal credit is associated with high interest rates, making loans prohibitively expensive for the majority of borrowers (Narayanan, 2016; Saqib et al., 2016). Consequently, farmers may be able to increase agricultural output if they have access to affordable and readily available formal financing. By supplanting informal credit, access to formal finance may enhance agricultural output positively. Credit may also be utilized to support employment and management decisions, as well as to increase capital stock by investing in fixed capital. Despite the continued importance of physical capital and labor to crop production, the availability of agricultural credit influences crop output via investment in agrarian inputs (Linh et al., 2019; Saqib et al., 2016, 2017). Figure 1 depicts the distribution of rural credit in a selection of China’s wheat-producing provinces, revealing that Henan and Shandong benefit the most from rural credit distribution than other provinces.

Figure 1. Distribution of rural credit in selected wheat producing provinces of China.

1.1. Why wheat crop?

Wheat is the world’s third-largest crop and an essential source of energy for human diets. Its global expansion is 757,7 MT and 220,9 MH (Le Gouis et al., 2020). Therefore, maintaining and increasing global wheat production has a significant impact on food security (Rehman et al., 2015). Wheat yields vary widely among regions with comparable climatic conditions, indicating significant yield gaps in a number of countries and the potential for increased regional flexibility in wheat production. Future climate changes may have a significant impact on wheat production, as wheat is susceptible to global warming (Balkovič et al., 2014; Le Gouis et al., 2020). Wheat will continue to make a significant contribution to human nutrition; increasing its production is necessary to ensure food security. Currently under discussion are two crucial aspects of safeguarding and sustaining future global wheat production. To meet the rising demand for food, wheat systems must be sustainably intensified. The second factor is the impact of future climate change on wheat yields (Balkovič et al., 2014). Compared to the current output of 642 MT, the world will need 840 MT of wheat by 2050 (Sharma et al., 2015). Therefore, there is no longer a need for animal feed, and climate change will not negatively impact wheat production. To meet more than 80% of the wheat demand, emerging countries must increase their wheat production by 77% (Ghafoor et al., 2021). Asia and Europe produce the largest percentage of the world’s wheat (44 and 34 percent, respectively). Approximately 12 percent of the world’s output comes from North America, while only 3 percent comes from the rest of the world (Le Gouis et al., 2020).

The world’s largest wheat-growing region is China, and wheat is one of the country’s two most significant staple food crops (Zhao et al., 2019). Wheat cultivation occupied 24,26 MH in 2010, accounting for 11.18 percent of the world’s 216.97 MH wheat cultivation area. China produces 115.18 MT of wheat, which accounts for 17.70% (one-third) of the global total of 650.88 MT. It is the world’s largest consumer and producer of wheat, followed by India and the United States. Wheat provides around one-fifth of China’s total calorie intake (Xu et al., 2013; Yuan et al., 2018). In 2017, 134.3 MT of wheat were grown, accounting for about 15% of global wheat production, with India producing the most (98.51 MT), followed by the Russian Federation (85.86 million tons), the United States (47.37 MT), and France (36.92 MT) (Zheng et al., 2021).

Despite being the world’s leading wheat producer, China’s wheat output is being hampered by excessive chemical fertilizer use and low yields (Zheng et al., 2021). Despite the fact that China’s wheat production has been steadily increasing, the country is experiencing significant problems in a number of areas. These include a decrease in soil fertility, a lack of water, pollution of the environment, and more high temperatures and droughts due to changes in the global climate, as well as a growing labour shortage and a shift from growing grain to growing cash crops. Furthermore, average yields and grain quality must improve as China’s arable land area shrinks and its quality of life improves (Ghafoor et al., 2021; Wang et al., 2009). In China, there are two types of wheat. Winter wheat and spring wheat are named after the seasons in which they grow. Wheat is primarily grown in Hebei, Shanxi, Henan, Shandong, Anhui, Hubei, Jiangsu, Sichuan, and Shaanxi. Figure 2 shows that Henan, Shandong, and Hebei produce more wheat than other provinces in 2005, 2010, 2015, and 2020.

Figure 2. Province-wise wheat production in China.

China’s food security is a global concern due to the country’s large population. Over the last several decades, China has made enormous strides in increasing food production, feeding 22% of the world’s population on 7% of the world’s cultivated land. Food shortages in China, on the other hand, have significantly decreased in recent years. In comparison to 1997, when food imports totaled 4.17 MT, imports now total 19.51 MT, with a cost increase from $916.14 million to $6,217.31 million (Yu et al., 2019). There will be a continuous decline in food self-sufficiency in China as a result of urbanization (He et al., 2022). The domestic grain self-sufficiency rate must maintain at or above 95%. To fulfil the increasing food demand, annual grain output should be 580 MT by 2030 at a pace of at least 2 percent yearly (Fan et al., 2012; XU et al., 2013). Also, there is a growing demand for wheat in new markets outside the climate zone where it grows best. China is a major agricultural nation with a population that will reach a high of 1.5 billion around 2030. Due to its teeming population and unparalleled urbanization, global attention has been focused on its food supply capabilities (Wang et al., 2018).

1.2. Research gap/contribution

Climate, precipitation, drought type, soil, organic matter content, and nitrogen fertilizer input contribute to wheat yield and quality. Several studies regarding China have examined the effect of the climatic change on cereal production (Pickson et al., 2020), rice production (Pickson et al., 2021), wheat production (Zhai et al., 2017), soybean yield (Zhang et al., 2022), and agricultural output (Chandio et al., 2020; Rehman et al., 2020) using various approaches, modeling procedures, and obtained mix results/outcomes. Nevertheless, none of them has investigated the impacts of rural credit on crop production, particularly for China’s major nine wheat-producing provinces. The current paper contributes to the existing literature in the following ways: First, the current study scrutinized the relationship between rural credit and wheat production in nine selected provinces (i.e., Hebei, Shanxi, Henan, Shandong, Anhui, Hubei, Jiangsu, Sichuan, and Shanxi). Rural economic development and the modernization of agriculture rely heavily on agricultural credit. Modernizing agriculture refers to introducing new technologies, improving human resources, managing natural resources, and managing the environment; however, agriculture credit remains the most crucial factor. To enhance food security and eradicate poverty, agricultural credit allows farmers to use inputs efficiently, adopt modern technologies, and allocate resources more efficiently. Globally, wheat is one of the most widely consumed staple foods. Fertilization and intensive wheat cultivation promoted during and after the green revolution played a significant role in improving yield levels and ensuring food security. To meet the food requirements of the growing population within the bounds of resource availability, the focus of farming must shift toward sustainable production systems. In the existing literature, fertilizer is largely ignored as a factor in wheat production. Fertilizer’s impact on wheat production is examined in this study. Second, this study incorporated the other main determinants of wheat products, including fertilizer consumption, pesticides, planting area, tractors and rural labour. These determinants are also used in previous studies (Gul et al., 2022, 2022; Jan et al., 2021; Ozdemir, 2021; Warsame et al., 2021; Zhang et al., 2022). Third, for empirical investigation, we applied various econometric methodologies, including panel second-generation unit root test (CADF test), panel second-generation long-term co-integration approach (Westerlund ECM), and the FGLS long-run estimator. In last, the DOLS, One-step GMM, and panel causality tests verify the outcomes of the FGLS method.

The following research questions are addressed in this study: Does agricultural credit boost wheat production in China’s major wheat-producing provinces? How do input factors such as planting area, fertilizer, pesticide, machinery, and labor affect wheat production in China’s major wheat producing provinces?

Notwithstanding, the organization of the study is as follows: the second section discusses the literature review, which is based on the formal credit and agricultural output nexus; the third section discusses the data and econometric strategy; the fourth section reports results and discussions; and the final section discusses conclusion and policy implications of the study.

2. Literature review

2.1. Rural formal credit and agricultural output nexus

In both developed and developing nations, agricultural credit is crucial to the growth of the production and investment structure of the farming sector. It is a critical tool for agricultural advancement. Given the importance of agricultural output to the national economy and the importance of rural formal credit to agricultural development in developing countries, numerous studies have examined the relationship between rural formal credit and agrarian performance at the macro level. Chandio et al. (2020) argue that rural formal credit has a substantial impact on Pakistani agricultural performance over the short and long run using an auto-regressive distributed lag procedure. Anh et al. (2020) utilized the same methodology and concluded that agricultural financing affects agricultural gross domestic output in both the short and long run. From 1998 to 2016, Bahşi and Çetin (2020) assessed the impact of agricultural financing on agricultural output value in Turkey. According to the findings, agricultural output value and agricultural credits are positively correlated. Bashir and Mehmood (2010) investigated the impact of institutional lending on rice crop yields in Pakistan. They claimed that credit is critical for improving and increasing agricultural production, particularly rice productivity, and that rural formal credit has a positive impact on rice production.

Chandio et al. (2018) investigated the effect of short-term and long-term institutional loans on the amount of wheat that small farms in Pakistan are able to cultivate. Using the Cobb—Douglas production function method, it was determined that agricultural credit had a positive and significant effect on wheat yield. Short-term loans have a greater impact on wheat production than long-term loans. According to the authors, farmers used multiple inputs such as improved seeds and fertilizers to produce wheat in the same year.

Recent research by Ali Chandio et al. (2022) examined the short- and long-term formal credit, its effect, and the technical development of rice in Nepal between 1991 and 2019. Using the ARDL methodology, it was determined that formal credit significantly improved both long- and short-term rice cultivation. Similarly, Chandio et al. (2016) used the Ordinary Least Square (OLS) method to examine the direction of causality between rural formal credit distribution and agricultural GDP in Pakistan between 1992 and 2016. The empirical findings indicate a significant positive correlation between formal farm loans and agricultural gross domestic product. According to Oyakhilomen et al. (2012), there is no co-integration between cocoa output in the context of Nigeria and the agricultural rural credit guarantee program fund. It was often diverted by cocoa farmers whose loans were guaranteed by the fund, which guaranteed a limited number of credits, limited the amount of credit, and limited the number of loans. Similarly, Hartarska et al. (2015) investigate how rural commercial banks and Farm Credit System institutions affect economic development. According to the projected results, the agricultural loan has a beneficial influence on agricultural GDP growth. Kassouri and Kacou (2022) found that agricultural credit is important in West African agriculture’s growth.

Based on a constant credit elasticity of agricultural GDP, Narayanan (2016) found that rural formal credit plays only a minimal role in determining agricultural GDP in India.

Nascimento et al. (2022) found that agricultural credit had a positive long-term impact on agricultural gross value-added production in Brazil. A study of rice production in Nigeria between 1981 and 2016 used the vector error correction model by Omoregie et al. (2018). As credit increased, rice production increased. Bangladeshi rice growers’ efficiency was evaluated by Rabbany et al. (2022). The empirical findings reveal that credit-constrained rice cultivators are less technically efficient than credit-free rice cultivators. Using cross-country data, Seven and Tumen (2020) scrutinized the link between farming finance and crop production. The findings indicate that doubling agricultural credits increases crop production by around 4–5 percent. Shahbaz et al. (2013) examined the relationship between financial development and agricultural growth from 1971 to 2011 using the Cobb-Douglas function. According to the data, financial development facilitates agricultural expansion. In addition, Yazdi and Khanalizadeh (2014) analyzed the causal relationship between Iran’s dynamic financial expansion, economic growth, and instability by analyzing annual time series data from 1970 to 2011. The authors conclude that the relationship between agricultural economic growth and financial expansion is two-way. From 1973 to 2015, Zakaria et al. (2019) analyzed the impact of financial growth on South Asia’s agricultural output. The projected results indicated that agricultural production initially increases as financial development advances, but then decreases as financial development advances further. Table 1 reports a summary of earlier studies. This study also suggests hypothesis 1 in light of the analysis above.

Table 1. Summary of empirical studies on the impact of agricultural credit on agricultural production

Country	Time	Model	Crop/Agri. pro	Agri- credit	References
China	1978–2018	ARDL	Cereal production	+/*	He et al. (2022)
China	1990–2017	ARDL	Grain production	+/*	Chandio et al. (2022)
Pakistan	1983–2016	ARDL—VECM	Agri—production	+/*	Chandio et al. (2020)
Vietnam	2004–2016	ARDL	Agricultural GDP	+/*	Anh et al. (2020)
Turkey	1998–2016	OLS	Agri—production	+/*	Bahşi and Çetin (2020)
Pakistan	2017	CDPF	Wheat	+/*	Chandio et al. (2018)
Nepal	1991–2019	VECM	Rice	+/*	Ali Chandio et al. (2022)
China	2000–2020	DOLS-FMOLS	Soybean yield	-/*	Zhang et al. (2022)
US	1991–2010	DID	Agricultural GDP	+/*	Hartarska et al. (2015)
West Africa	1995–2014	CSD	Agricultural GDP	+/*	Kassouri and Kacou (2022)
India	1996–2012	OLS	Agricultural GDP	NS	Narayanan (2016)
Brazil	2022	VECM	Agricultural GDP	+/*	Nascimento et al. (2022)
Nigeria	1981–2016	VECM	Rice	+/*	Omoregie et al. (2018)
Bangladesh	2021	Stochastic frontier	Technical efficiency	+/*	Rabbany et al. (2022)
Cross-country	1991–2014	OLS	Agri—production	+/*	Seven and Tumen (2020)
Pakistan	1971–2011	CDPF	Agricultural GDP	+/*	Shahbaz et al. (2013)
Iran	1970–2011	JCI	Economic growth	+/*	Yazdi and Khanalizadeh (2014)
South Asia	1972–2015	JCI	Agri—production	+/*	Zakaria et al. (2019)

Note: ARDL: Auto Regressive Distributed Lag, VECM: Vertical Error Correction Model, ADF: Augmented Dickey Fuller, P—P: Phillips-Perron, Dynamic, OLS: Ordinary Least Square, JCI: Johansen Co—integration, DID: Difference in Difference, CSD: Cross-sectional dependence, +/*: Positive and significant, NS: Not Significant, CDPF: Cobb—Douglas production function.

Hypothesis 1:

In the major wheat-producing provinces of China, agricultural credit has a positive effect on wheat production.

3. Methodology

3.1. Data

We examined the long-run impacts of rural credit on wheat production in selected nine provinces of China from 1992 to 2020, considering fertilizer consumption, pesticide consumption, wheat planting area, large and medium tractors, and rural labour force. Hebei, Shanxi, Henan, Shandong, Anhui, Hubei, Jiangsu, Sichuan, and Shanxi are the nine provinces selected. We measure wheat production, fertilizer, and pesticide consumption in 10,000 tons. Agricultural credit is expressed in 100 million Yuan, and the wheat planting area is measured in thousands of hectares. In rural areas, the labour force is measured by 10,000 people, and large and medium-sized tractors are measured in numbers. A rise in wheat imports could result in a reduction in the price of wheat in the domestic market and, therefore, a reduction in resources directed toward its production. Globally, these factors have resulted in improved GDP, trade balance, and investments, and as a result, private consumption has decreased. In recent years, wheat yields have been able to increase in a positive manner as a result of agricultural credits because there has been an increase in the usage of farming inputs such as seeds of improved varieties and fertilizers which can, in turn, be translated into higher wheat yields (Chandio et al., 2018). It is imperative to have access to credit on time if modern technologies are to be adopted, fertilizers purchased, and improved seeds bought. A timely introduction of agricultural credit will increase the production of farms, thus increasing the economy’s growth rate; thus, agricultural credit is a crucial component of modernizing the agricultural sector. For humans to be able to eat enough food, fertilizer is added to crops in order to increase production. Potassium, phosphorus, and nitrogen are essential nutrients for crops to grow, as they enable them to grow bigger, faster, and produce more food (Sedlacek et al., 2020). The yield losses associated with a reduction in pesticide use in the wheat production industry were quantified on the basis of several experimental crop systems. Since 1960, US crop yields have increased by more than 98%, and in France, they have increased by more than 187% between 1960 and 1990 as a result of pesticide use (Hossard et al., 2014). The data have been collected from the Chinese Statistical Yearbooks and transformed into logarithmic form before econometric investigation is performed.

3.2. Econometric strategy

In addition to fertilizer and pesticide consumption, wheat planting area, and large and medium tractors, we looked at the long-term effects of rural credit on wheat production in nine Chinese provinces from 1992 to 2020.

(1) $W{P}_{it}={\beta }_0+{\beta }_1(CRE{D}_{it})+{\beta }_2(FER{C}_{it})+{\beta }_3(PES{C}_{it})+{\beta }_4\left(WP{A}_{it}\right)+{\beta }_5\left(TR{A}_{it}\right)+{\beta }_6(RL{F}_{it)}+{\epsilon }_{it}$(1)

Where ${\beta }^{\prime }s$ show the slop of explanatory variables, and selected nine provinces are cross sections denoted by $i$, whereas, $t$ shows time period from 1992 to 2020. Wheat production is represented by $WP$, rural credit by $CRED$, fertilizer consumption by $FERC$, pesticides by $PESC$, planting area by $WPA$, tractors by $TRA$, and rural labour by $RLF$. The dependent variable is wheat production, whereas the independent variables are rural credit, fertilizer consumption, pesticides, planting area, tractors, and rural labour. However, $\epsilon$ is the residual. The logarithmic form of equation$\left(1\right)$is expressed by equation $\left(2\right)$.

(2) $lnW{P}_{it}={\beta }_0+{\beta }_1(lnCRE{D}_{it})+{\beta }_2(lnFER{C}_{it})+{\beta }_3(lnPES{C}_{it})+{\beta }_4\left(lnWP{A}_{it}\right)+{\beta }_5\left(lnTR{A}_{it}\right)+{\beta }_6(lnRL{F}_{it})+{\epsilon }_{it}$(2)

3.2.1. Cross-section dependence tests

All units within the same data cross-section can be correlated, indicating that all panel data can be subject to pervasive cross-sectional dependence. In the majority of cases, this phenomenon can be attributed to unobserved common factors that affect all units differently, despite the fact that these common factors affect all units equally (Patton, 2013). Tests of the first- or second-generation panel unit root tests have been decided using cross-sectional dependency tests (Arslan et al., 2022; Khan et al., 2022). In order to accomplish this, four tests were undertaken: The Breusch and Pagan LaGrange Multiplier (LM), the Pesaran Scaled LM test, and Bias-corrected scaled LM and the Pesaran CD test. Breusch and Pagan’s (1980) LM test formed as:

(3) $LM=\sum _{i=1}^{N-1}\sum _{j=i+1}^N{T}_{ij}{\hat{\rho }}_{ij}^2\to x^2\frac{N\left(N-1\right)}{2}$(3)

where, $x^2$ is an asymptotical transmission for constant $N$ as $T_{ij}\to \mathrm{\infty }$ for all $\left(i,j\right)$ and $N\to \mathrm{\infty }$ the ${\hat{\rho }}_{ij}^2$ is the correlation.

Moreover, Pesaran (2004) advocated the use of alternative indicators to manage the proportional bias that is associated with the pairwise correlations, as presented below.

(4) $C{D}_{\rho }=\sqrt{\frac{2}{N\left(N-1\right)}}\sum _{i=1}^{N-1}\sum _{j=i+1}^N{T}_{ij}{\hat{\rho }}_{ij}\to N\left(0,1\right)$(4)

in any desired sequence $T_{ij}\to \mathrm{\infty }$ and $N\to \mathrm{\infty }$with asymptotic and standardized normal distributed. For an extensive panel data modeling $CD=0$ for all $T_{ij}>k+i$ and all $N$, Pesaran described it as a broad range of data. In order to capture the significant thresholds of approximation, we use all these cross section dependence techniques in our study.

3.2.2. Panel unit root

In the context of probability theory and statistics, it can be difficult to infer statistical insight from time series models. Unit roots are associated with linear stochastic processes if 1 is included in the characteristic equation of the panel unit root-testing framework (Menegaki et al., 2021). In the panel unit-testing framework, there are two generations of root unit tests. In contrast to the first generation of panel unit root tests, which assumed cross-sectional independence of units, the second generation relaxes this assumption (Yang & Khan, 2022; Zakari & Khan, 2021, 2022). Using additional information in unit root testing improves the precision of test regressions and the accuracy of inferences. This study employs a second generation augmented cross-sectional ADF (CADF) test presuming heterogeneity and a typical autoregressive structure of all cross-section units. Hansen (1995) suggests that the CADF test assumes that real economic phenomena are typically not univariate in nature. Hansen (1995) formed CADF unit root test as:

(5) $y_t=d_t+s_t$(5)

(6) $a\left(L\right)\mathrm{\Delta }{s}_t=\delta s_{t-1}+v_t$(6)

(7) $v_t=b{\left(L\right)}^{\prime }\left(\mathrm{\Delta }{x}_t-{\mu }_x\right)+e_t$(7)

Where $d_t$ is constant and linear trend deterministic term, $a\left(L\right)(1-a_{1}L-a_2{L}^2-\dots a_p{L}^p$ legal operator polynomial $L$.

3.2.3. Panel co-integration tests

A co-integration test can be used to determine the degree of sensitivity of two variables to the same average price over a given time period and to establish conditions under which two or more non-stationary time series have been integrated in such a way that a deviation from equilibrium cannot occur over an extended period of time (Khan et al., 2022; Li & Taghizadeh-Hesary, 2022). It is, therefore, imperative to determine whether the time series has a stable, long-term relationship if they are non-stationary. Pedroni (1999) and Kao (1999) co-integration techniques are popular in the empirical literature for testing co-integrating relationships among variables also used in this study. Through the use of these basic tests, we are able to gain a better understanding of the long-term relationships that exist between variables, as well as an understanding of what the fundamental framework upon which the entire theory rests is all about. Based on Engle and Granger (1987), Kao (1999) tests are formed as follows:

(8) $D{F}^{\ast }=\frac{t_{\rho }+\frac{\sqrt{6N}{\hat{\sigma }}_v}{2{\hat{\sigma }}_{0v}}}{\sqrt{\frac{{\hat{\sigma }}_{0v}^2}{2{\hat{\sigma }}_v^2}+\frac{3{\hat{\sigma }}_v^2}{10{\hat{\sigma }}_{ov}^2}}}$(8)

where, ${\hat{\sigma }}_{ov}^2={\hat{\sigma }}_{ov}^2-{\hat{\sigma }}_{ov\epsilon }^2{\sigma }_{0\epsilon }^{-2}$ reflects the long run relationship, $t_{\rho }$ is identical deterministic $t$ value,, ${\hat{\sigma }}_v^2$, ${\hat{\sigma }}_{0v}^2$ are ${\sigma }_v^2$, while ${\sigma }_{0v}^2$, ${\hat{\sigma }}_v^2={\hat{\sigma }}_v^2-{\hat{\sigma }}_{v\epsilon }^2{\sigma }_{\epsilon }^{-2}$ are short term equilibrium adjustments. One of the main reasons for estimating Kao (1999) co-integration analyses is that many studies fail to reject the no-co-integration null despite the apparent co-integration implied by the theory. Additionally, we applied Westerlund’s (2007) panel co-integration approach. By using the panel co-integration technique derived by Westerlund (2007), we can find the residuals from static as well as dynamic panel regressions that have cross-sectional and unit root dependences. Moreover, it takes into account the cross-section dependency issue based on structural dynamics rather than residual dynamics, and it does not impose any common-factor restrictions (Khan et al., 2022).

(9) $K_{it}=\prod _i^{/}{M}_t+{\epsilon }_{it}$(9)

(10) $M_{jt}={\lambda }_j{M}_{jt-1}+v_{jt}$(10)

Where, ${\epsilon }_{it}={\vartheta }_i{\epsilon }_{it-1}+{\mu }_{it},$ $j=1,\dots ,k$, $j=1,\dots ,k$ and ${\prod }_i$ is factor amenable, and ${\lambda }_j<1$ to all $j$. The process for estimating data is as follows:

(11) $y_{it}={\mathrm{\varnothing }}_{1i}+{\mathrm{\varnothing }}_{2i}t+z_{it}$(11)

(12) $x_{it}=x_{it-1}+v_{it}$(12)

where $t=1,\dots ,T$ and $i=1,..N$ index the time series and cross sectional units respectively.

3.2.4. Long-run modeling

To calculate the long-run estimates, we used a cross-sectional time-series Feasible Generalized Least Squares (FGLS) and robust dynamic ordinary least squares (DOLS) as well as a panel generalized method of moments (GMM). The error term in FGLS regressions with heteroskedastic cross-sectional correlations and panel-specific $AR\left(1\right)$speculations reflects a large number of factors, heteroskedasticity, serial correlations and cross-sectional correlations (Bai et al., 2021). Moreover, FGLS method eliminates cross-sectional as well as serial correlation biases by using a high-dimensional error covariance matrix estimator.

(13) ${\hat{\beta }}_{FGLS}={\left[X{ }^{\mathrm{\prime }}{\hat{\mathrm{\Omega }}}^{-1}X\right]}^{-1}X{ }^{\mathrm{\prime }}{\hat{\mathrm{\Omega }}}^{-1}Y$(13)

Robustness checks are often performed when regression model assumptions are violated, and residuals are independently distributed (Hoechle, 2007). DOLS is an alternative (parametric) method of solving problems in which lags and leads are introduced to deal with problems regardless of how they are integrated and whether or not they are co-integrated (Mark & Sul, 2003). Using the system GMM estimator in dynamic panel data modeling, moment conditions for the differenced equation and moment conditions for the model at each level are combined in the estimation process. For this estimation procedure, it is not known what the initial optimal weight matrix would be non-serial correlation and under homoscedasticity (Windmeijer, 2000). In view of the fact that the estimations of all these advanced economic models in one study represents an important contribution to avoid any sporadic estimations. To explore the causal connections between the studied variables, we also performed Dumitrescu and Hurlin (2012) panel causality tests. An overview of the entire process of econometric estimation is shown in Figure 3.

Figure 3. Estimation steps.

4. Results and discussions

4.1. Summary of descriptive statistics and correlation estimates

Table 2 displays the descriptive statistics and correlation matrix for the variables chosen. The average wheat production, agricultural credit, fertilizer consumption, pesticide usage, wheat planting area, agricultural machinery (tractors), and rural labor force are 6.64, 6.94, 5.55, 1.83, 7.52, 11.19, and 7.87, respectively, according to the results. The average of agricultural machinery (tractors) is higher, while the average pesticide usage is lower than the other variables. Table 2 also shows that agricultural credit, fertilizer consumption, pesticide use, wheat planting area, agricultural machinery (tractors), and rural labor force are all significantly and positively related to rice production. Figure 4 depicts the data trend for stack cross sections for all of these variables.

Figure 4. Stack cross sections data trend.

Table 2. Summary of descriptive statistics and correlation estimates

	lnWP	lnCRED	lnFERC	lnPESC	lnWPA	lnTRA	lnRLF
Mean	6.644	6.941	5.554	1.834	7.529	11.195	7.870
Maximum	8.230	11.713	6.573	9.313	8.655	13.311	8.593
Minimum	5.018	2.679	4.158	−0.150	6.283	8.455	6.834
Std. Dev.	0.834	2.131	0.512	0.936	0.640	1.092	0.463
lnWP	1.000
lnCRED	0.230*	1.000
lnFERC	0.783*	0.418*	1.000
lnPESC	0.496*	0.175*	0.646*	1.000
lnWPA	0.974*	0.057	0.734*	0.481*	1.000
lnTRA	0.428*	0.749*	0.555*	0.327*	0.279*	1.000
lnRLF	0.735*	0.338*	0.840*	0.587*	0.717*	0.310*	1.000

Note: *p < 0.01.

4.2. Cross-sectional dependence (CSD) tests results

In this study, we chose six variables that have a relationship with wheat production. If the CSD issue is ignored when estimating data, the results will be unreliable and unstable. We used four CSD tests to report more reliable results (i.e., Breusch-Pagan LM, Pesaran scaled LM, Bias-corrected scaled LM, and Pesaran CD). Table 3 shows that all of the CSD tests are significant at the 1% level, indicating that the presence of CSD in residuals is confirmed.

Table 3. CSD tests results

Tests	lnWP	lnCRED	lnFERC	lnPESC	lnWPA	lnTRA	lnRLF
BPLM	412.534*	985.138*	646.532*	443.010*	415.945*	852.079*	583.058*
PSLM	44.3750*	111.857*	71.9519*	47.966*	44.777*	96.175*	64.471*
BCSLM	44.214*	111.696*	71.7912*	47.805*	44.616*	96.0151*	64.310*
PCD	5.831*	31.379*	24.473*	12.742*	5.245*	28.992*	13.695*

Note: *p < 0.01.

4.3. Panel unit root test results

Table 4 reports the stationarity of each variable based on the outcomes of the Cross—Sectional ADF (CADF) second generation panel unit root test. The findings of the CADF test reveal that the series under investigation is not stationary at I(2). Further, results show that some variables, e.g., wheat production (lnWP), agricultural credit (lnCRED), fertilizer consumption (lnFERC), usage of pesticide (lnPESC), wheat planting area (lnWPA), and rural labour force (lnRLF) are not stationarity at the level, while tractors (lnTRA) is stationarity at the level. However, after using the first difference, all the studied variables become stationary and highly significant at the first difference.

Table 4. CADF test results

Tests	lnWP	lnCRED	lnFERC	lnPESC	lnWPA	lnTRA	lnRLF
Level	−1.910	−2.088	−2.050	−1.894	−1.314	−2.382**	−1.973
P-value	(0.321)	(0.155)	(0.185)	(0.339)	(0.916)	(0.027)	(0.149)
First difference	−3.638*	−2.902*	−2.882*	−2.991*	−2.503*	−2.695*	−3.455*
P-value	(0.000)	(0.000)	(0.000)	(0.000)	(0.011)	(0.002)	(0.000)

Note: *p < 0.01 and **p < 0.05.

4.4. Co-integration testing outcomes

The Pedroni and Kao panel co-integration tests are used to investigate in depth the long-term relationships between the variables under study. The results of panel co-integration tests are presented in Table 5. At the 1% and 5% levels of significance, the results of both tests indicate a long-term association between the sample variables.

Table 5. Panel co-integration tests outcomes

Panel (A) Pedroni co-integration test
Statistics	Panel v-Stat	Panel rho-Stat	Panel PP-Stat	Panel ADF-Stat
Value	−0.706324	−0.390726	−11.64023*	−2.601809*
P-value	0.7600	0.3480	0.0000	0.0046
		Group rho-Stat	Group PP-Stat	Group ADF-Stat
Value		1.255665	−9.940556*	−1.718817**
P-value		0.8954	0.0000	0.0428
Panel (B) Kao co-integration test
	ADF
Value	−4.969448*
P-value	0.0000

Note: *p < 0.01 and **p < 0.05.

The Westerlund (2007) co-integration approach verifies the long-term association among the sample variables. The Westerlund’s (2007) co-integration test in Table 6 indicates highly significant test statistics that establish the existence of the long-term co-integration in the model. As a result, wheat production has long-run interrelationship with agricultural credit, fertilizer consumption, pesticide usage, wheat planting area, agricultural machinery, and rural labour force in the context of selected nine provinces of China. The occurrence of the long-term associations makes assessing the long-run elasticity/coefficient estimates straightforward.

Table 6. Westerlund ECM panel cointegration test results

Statistic	Gt	Ga	Pt	Pa
Value	−3.679	−8.287	−10.696	−8.853
Z-value	−3.685	2.003	−3.619	0.395
P-value	0.000*	0.977	0.000*	0.654
Robust P-value	0.005*	0.223	0.003*	0.123

Note: *p < 0.01.

4.5. Estimating the long-run relationship

The current study applied the FGLS method to examine the long-term impact of agricultural credit, fertilizer consumption, pesticide use, wheat planting area, agricultural machinery, and rural labor force on wheat production in selected provinces of China. Table 7 displays the results of the FGLS methodology. The results indicated that agricultural credit had a substantial and positive effect on wheat production. The estimated long-run elasticity/coefficient is 0.035, indicating that a 1% increase in rural agricultural credit results in a 0.035 percentage point increase in wheat production. This empirical analysis supports the effective policy of rural agricultural credit as a crucial institutional implementation for accelerating agrarian production and enhancing the welfare of rural households. China is currently one of the largest producers of wheat in the world. Around the globe, China ranked third in terms of wheat production in 2020, after India (31.36 MH) and Russia (28.86 MH) (FAO, 2020). In 2020, the total wheat production was 134.25 MT, ranking first worldwide. Despite this, China’s wheat yield was only 5.74 tons per hectare, which is lower in comparison to other major wheat-producing nations for example New Zealand (9.93 tons/ha), Belgium (8.95 tons/ha), and the Netherlands (8.56 tons/ha) (FAO, 2020). The decreased yield hinders the economic performance of China’s wheat industry, which should be enhanced. Finding of the current study is reliable with Rehman et al. (2019). They argued that institutional credit improves food production through purchasing of farming modern inputs (i.e., fertilizers, insecticides, improved seeds, and agricultural machinery). Various empirical studies, including Das et al. (2009) for India, Chisasa and Makina (2015) for South Africa, Sarker (2016) for Bangladesh, Narayanan (2016) for India, Anetor et al. (2016) for Nigeria, Ahmad et al. (2018) for Pakistan, and Nascimento et al. (2022) for Brazil revealed that rural credit plays a pivotal role for rural economic development and improves agricultural production. He et al. (2022) highlighted that rural credit substantially increased cereals yield in the long-run in the case of Sichuan-China. Similarly, Chandio et al. (2022) concluded that farm credit significantly increased China’s long-term and short-term grain yields.

Table 7. FGLS regression results

Variables	Coef.	St.Err.	t-value	p-value	[95% Conf	Interval]
lnCRED	0.035*	0.006	5.93	0.000	0.023	0.046
lnFERC	0.032	0.042	0.76	0.445	−0.050	0.114
lnPESC	−0.002	0.005	−0.37	0.711	−0.012	0.008
lnWPA	1.135*	0.029	39.62	0.000	1.079	1.191
lnTRA	0.037*	0.007	5.30	0.000	0.023	0.050
lnRLF	0.083**	0.041	2.01	0.044	0.002	0.163
Constant	−3.363*	0.164	−20.46	0.000	−3.686	−3.041
Wald chi2(6)	10934.76*
Prob > chi2	0.000

Note: *p < 0.01 and **p < 0.05.

Wheat production is positively impacted by agrochemical inputs like fertilizer. The estimated long-run elasticity/coefficient is 0.032, which indicates that a 1% increase in chemical fertilizer results in a 0.032% increase in wheat yield. In contrast, it has been observed that pesticide use is detrimental to wheat production. Fertilizer is an essential input factor that accelerates the plant’s growth and development. To increase production, Chinese farmers rely heavily on agrochemicals such as pesticides and fertilizers. However, excessive use has negative effects on the environment and human health. More recent studies, for example Gul et al. (2022) for Pakistan, Ozdemir (2021) for Asian countries, and Rayamajhee et al. (2021) for Nepal reported that the usage of fertilizer significantly enhanced cereals production in the long-term. Zhai et al. (2017) examined the influence of technical progress on wheat production in China and highlighted that fertilizer consumption significantly boosts wheat production. Similar outcomes are also revealed by Gul et al. (2022) for Pakistan.

In addition, planting area, agricultural equipment (tractors), and rural labor force have a substantial impact on wheat production. The estimated long-run coefficients are 1.135%, 0.037%, and 0.083%, indicating that a 1% increase in the planting area, agricultural machinery (tractors), and rural labor force resulted in a 1.135%, 0.037%, and 0.083% increase in wheat production in certain Chinese provinces. These outcomes are consistent with Gul et al. (2022), Kumar et al. (2021), and Warsame et al. (2021). They argued that planting area significantly contributed to cereals production. Furthermore, Zhai et al. (2017), Pickson et al. (2020), and Ozdemir (2021) reported that power consumption for agricultural machinery positively and significantly augmented food crop production.

4.6. Robustness check using DOLS and one-step GMM estimators

We also estimate the DOLS and One-step GMM methodologies to verify the results of the FGLS method. The outcomes of both models are reported in Table 8. The findings of the DOLS and One-step GMM methods indicate that agricultural credit is statistically significant with a positive coefficient, suggesting that agricultural credit boosts wheat production. Further findings show that other determinants also have the same positive and negative influence on wheat production in China. Hence, the outcomes of the DOLS and One-step GMM methods are consistent with the outcomes of FGLS method.

Table 8. Robustness check results

	One-step GMM		DOLS
Variables	Coef.	St.Err.	Coef.	St.Err.
lnWP L1.	0.238*	0.054	–––	–––
lnCRED	0.030*	0.006	0.024**	0.014
lnFERC	0.097	0.074	0.073	0.123
lnPESC	−0.041	0.052	−0.072	0.104
lnWPA	0.843*	0.066	1.1372*	0.060
lnTRA	0.014	0.011	0.046**	0.023
lnRLF	0.249*	0.097	0.402**	0.210
Constant	−4.074*	0.743	–––	–––
Wald chi2(7)	1802.26*		R-squared	0.995
Prob > chi2	0.000		Adjusted R^2	0.991

Note: *p < 0.01 and **p < 0.05.

4.7. D-H causality test

Finally, we explore the direction of causality among the studied variables by using the D-H causality test. Findings are reported in Table 9. It can be viewed from Table 9, there is unidirectional causality from lnCRED and lnRLF to lnWP, suggesting that a change in lnCRED and lnRLF will affect lnWP, while any change in lnWP will not influence lnCRED and lnRLF. Further, we observed the bidirectional causality from lnFERC, lnPESC, lnWPA, and lnTRA to lnWP. This means that agrochemical inputs (i.e., fertilizer and pesticide), planting area, and agricultural machinery significantly lead to influence wheat production in the context of selected wheat producing provinces of China.

Table 9. Dumitrescu and Hurlin panel causality tests results

Null Hypothesis	W-Stat.	Zbar-Stat.	Prob.
lnCRED does not homogeneously cause lnWP	3.12481	3.72024	0.0002
lnWP does not homogeneously cause lnCRED	1.20283	0.21153	0.8325
lnFERC does not homogeneously cause lnWP	2.44542	2.47997	0.0131
lnWP does not homogeneously cause lnFERC	2.29204	2.19997	0.0278
lnPESC does not homogeneously cause lnWP	2.57138	2.70993	0.0067
lnWP does not homogeneously cause lnPESC	4.34042	5.93942	3.E–09
lnWPA does not homogeneously cause lnWP	3.56395	4.52193	6.E–06
lnWP does not homogeneously cause lnWPA	2.96467	3.42790	0.0006
lnTRA does not homogeneously cause lnWP	4.04839	5.40630	6.E–08
lnWP does not homogeneously cause lnTRA	2.25568	2.13359	0.0329
lnRLF does not homogeneously cause lnWP	2.24886	2.12113	0.0339
lnWP does not homogeneously cause lnRLF	1.42326	0.61394	0.5393

5. Conclusion and policy implications

This study examines the long-term effects on wheat production of rural agricultural credit, fertilizer consumption, pesticide use, planting area, agricultural machinery, and rural labor force. This paper utilizes panel data from 1992 to 2020 for nine of China’s most important wheat-producing provinces. The CADF test was utilized to confirm the stationarity level, and the results indicated that all variables under study are integrated at the first difference. After ensuring the stationarity of all variables, we employed the Pedroni, Kao, and Westerlund ECM panel co-integration techniques to investigate the long-term relationships. On the basis of the outcomes of three co-integration methods, it is confirmed that there is a long-term co-integration relationship between the variables under study. In addition, the FGLS, DOLS, and one-step GMM techniques are employed to estimate the long-term effects of regressors on wheat production. Consistent with the findings of the FGLS, DOLS, and one-step GMM estimation methods, it is important to note that agricultural credit significantly increases wheat production over time. Other factors, including fertilizer consumption, planting area, agricultural machinery, and rural labor force, have a positive and significant impact on wheat production in China’s nine major wheat-producing provinces, whereas pesticide use has a negative impact. Additionally, the D-H causality test findings reported a unidirectional causality relationship between rural credit and rural labour force to wheat production. On the other hand, a bidirectional causality is observed from fertilizer consumption, pesticide consumption, wheat planting area, and tractor to wheat production. The findings of this study have a number of significant policy implications. Given that access to credit increases farmers’ propensity to contribute to wheat production, the government should expand farmers’ access to institutional support facilities and provide financial literacy to ensure credit is used responsibly. This will ensure an increase in wheat cultivation and production, thereby satisfying wheat demand.

In addition, policymakers should encourage financial institutions to consider expanding their credit capabilities for rural farming families, so that a greater number of households can benefit. Efforts to increase the use of agricultural technology, such as agricultural machinery, require access to finance, which must expand beyond a single source, especially to traditional sources, because credit access may play a significant role in boosting agricultural production. Therefore, policies emphasizing easy access to bank financing will significantly improve farmers’ adaptability and utilization of modern agricultural technologies.

Furthermore, financial institutions should actively meet farmers’ capital needs, continue to make good use of agricultural loans and rediscounting, and provide differentiated credit support measures for all types of agriculture-related subjects around the links of grain production, storage, and processing. Additionally, they should look into ways to increase the credit availability of new agricultural business entities so that the prospects for agricultural growth can be improved. At the same time, China’s successful experience with agricultural credit will provide some guidance to other developing countries in increasing agriculture productivity and addressing poverty issues.

Despite its potential contribution, this study has several limitations. To begin, the current study focuses on nine Chinese provinces that are major wheat-producing regions (Hebei, Shanxi, Henan, Shandong, Anhui, Hubei, Jiangsu, Sichuan, and Shanxi); therefore, other provinces must be studied. Second, the current study examines the long-run impact of agricultural credit on wheat production, whereas corn, fruits, and vegetables can be studied in the same regions in the future. Third, the current study uses a panel dataset for nine wheat producing provinces in China, ignoring the province-specific estimation that is required. Finally, while the current study focuses on the impact of credit on wheat production, the Cobb-Douglass production function can be used to investigate the impact of physical inputs, fertilizer and pesticide expenditures, and climatic factors on corn production.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This study was supported by the National Social Science Fund of China (Project No. 21CGL026).