Introduction
Rapid urbanization, climate change and disaster risks place pressure on agricultural production and baseline vulnerability. While there has been a steady increase of population growth during the past decade, it is very important for future generations to be able meet sufficient quantity of appropriate food available in an inclusive and sustainable manner. Obviously, food security and sustainability can be achieved by increasing productivity in agricultural production and finding alternative sustainable ways to produce more food on less land. In addition to parameters such as fertilizer, irrigation, medicine, seed, labor, soil, tool-machine and technology usage, agricultural financing with low-interest for elimination of farmers‘ financial constraints to carry out their activities more comfortably and efficiently is vital to sustainable rural development, particularly for the least developed countries. Besides, advances in technology within agriculture have forced businesses to use more modern inputs and increased their capital requirements to access new knowledge, investments and innovative farming methods. Furthermore, price risks due to the low elasticity of demand in agricultural products can result in uncertainties in the incomes of producers and thereby, necessitate the use of policy tools such as measures to support producer prices, subsidies for agricultural inputs, and short-run financing loans to producers (Thompson, 1916; Leatham and Hopkin, 1988; Binswanger, 1989; Tweeten and Zulauf, 2008).
First and foremost, agricultural sector plays a strategic role in reaching social welfare development through spreading employment and income gains. Coinciding with steady growth in real wages, increasing productivity growth in all sectors is critical to increase domestic demand. To increase agricultural productivity, as well as inefficient production methods, labor market distortions should also be prevented. Besides, in agricultural development process, agricultural support policies such as cash subsidies, credits, tax benefits make agriculture competitive, improve investments, increase resource allocation and profitability, and thereby, continue farming in a sustainable manner (Kahan, 2013). Specifically, stimulating agricultural credit policies ensure the quantity and continuity of agricultural production by providing input to higher productivity gains. Especially, the existence of mostly small sized farms in low-income countries, most of which are family- owned and operated, the timing, techniques and conditions of harvesting, adopting new technology and diversifying their production are responsible for agricultural credit demand. Consequently, the uncertainties inherent in agricultural production raises the possibility that the producers do not always get the expected income and lead to reduced productivity by increasing vulnerability (Kahan, 2013; Rajan and Ramcharan, 2015).
The main challenge for developing countries is to improve their agricultural credit systems. For those countries, first and foremost, access to credit, especially for small-sized family farms, needs to be improved. Generally, lending to agricultural sector can be classified into mainly two types according to their intended use: Short-run or farm operating loans to help with the day-to-day expenses and investment loans to support agricultural enterprises to finance their investment expenses (Salami and Arawomo, 2013). Another aspect that deserves attention is the relationship between the agricultural credit and agricultural growth, even its impact on economic growth. As a matter of fact, agricultural sector has continued to show its important role as a ”buffer sector ” in the economy by providing employment during economic downturns in other sectors. However, above and beyond this, the ability of credit to induce agricultural productivity is also an issue for many developed countries worldwide. Besides this, access to adequate credit affects farm output by contributing to sustainability of farming systems and technology. Since managing risk to stabilize farm income is an important aspect of farming business, access to credit is a key component in protecting farmers against uncertainties. That is, easy availability and access to credit increases farmers and entrepreneurs lead to more diversified options to undertake new investments. One of the arguments for the contribution of agricultural credits on agricultural productivity by Braverman and Guasch (1986) is that agricultural loan interest rate should not be market rate, otherwise these types of programmes may result in some kind of subsidy and income transfer; that is, since income transfer and subsidy is proportional to the size of loan, larger landholders and larger farms receive larger income transfers and subsidies. However, it is crucial to determine to what extent financing of agricultural sector, especially through increasing the availability of bank credits to agricultural businesses and farmers have an impact on agricultural productivity. But more importantly, to what extent should we consider that agricultural financing may be more effective in contributing to higher agricultural productivity as well as stimulating growth? Going one step further, do making efforts to subsidize agricultural credits in developing countries and providing credit cheaper and easier to finance new technology adoption give an opportunity to those countries to promote rural development rapidly increasing agricultural output and productivity?
A substantial number of recent studies were devoted to investigating the effects of agricultural credits on agricultural productivity. As regards agricultural productivity, a great deal of research (Siriram, 2007; Das et al., 2009; Ali et al., 2014; Misra et al., 2016; Rehman et al., 2017; Seven &Tumen, 2020; Tuan Anh et al., 2020; Chandio et al., 2021, Manoharan & Varkey, 2021) identified the role of credits on agricultural productivity. While some of these studies used farm-level data, some of them has identified and addressed linkages focusing specifically on country-level data. For instance, Chandio et al. (2017) provided evidence on the positive impact of formal credit on sugarcane productivity with a farm-level data for Pakistan. Rehman et al. (2017) did a discrimination among loan types using time series data and concluded that total food production, loan disbursed by Modern agriculture technology machinery and Agriculture loans in Pakistan (ZTBL) and the total loan disbursed by various institutions had a positive and significant influence on the agricultural gross domestic product whereas cropped area and cooperatives loan had a negative but insignificant influence on the agricultural gross domestic product using time series data in Pakistan. Misra et al. (2016) observed a positive impact of the intensity of agricultural credit on total factor productivity in agriculture under state-level panel model for the Indian economy. A more comprehensive study by Seven and Tumen (2020) used country-level data covering 104 countries for the 1991-201 period. Accordingly, agricultural credit expansion contributes to high agricultural growth rates in almost all countries; however, this positive effect may vary according to their level of development. To the best of our knowledge, no prior work has yet been carried out to investigate the nexus between agricultural credits and agricultural value added as a proxy for agricultural productivity under a heterogenous panel cointegration and panel autoregressive distributed lag (ARDL) approach.
In the discussion of the role of agricultural credits in agricultural productivity both in short-run and long-run, the ARDL of pooled mean group (PMG), mean group (MG) and dynamic fixed effects (DFE) approaches are employed using a dataset comprising 53 countries in the period from 2000 to 2018. FMOLS (Fully Modified OLS) and Dynamic OLS (DOLS) and panel pairwise causality tests are used as robustness tests. According to this narrative, we are interested in studying the effect of agricultural credits on agricultural productivity with inflation, net foreign direct investments and general government total expenditures as our control variables to avoid omitted variable bias.
The rest of this study is structured as follows: section 2 presents the methodology; section 3 presents the data; empirical results are summarized in section 4 followed by conclusion in section 5.
Empirical Methodology
The aim of this empirical analysis is to investigate the short-run and long- run effects of agricultural credits on agricultural productivity by considering a global sample of 53 countries over the period 2000-2018. The countries in the panel are listed in Appendix A. We initially present the conceptual framework that we follow and then outline the statistical approach that we implement to estimate the long-run equilibrium parameters.
Before the estimation, it is necessary to investigate the characteristics of the cross sections and time series, as well as to control for the existence of specificities which may lead to inconsistent and incorrect results. In this context, a set of preliminary tests should be performed before estimating the model of interest as Variance Inflation Factor (VIF) to check for the existence of multicollinearity, Cross-section Dependence (CSD) test (Pesaran, 2004) to account for serial correlation of an unknown form in the error term, second generation unit root test (CIPS-test) (Pesaran, 2007) to test the stationarity of the data, and finally, second generation cointegration test (Westerlund, 2007) to check the order of integration of the variables under consideration for establishing long-run relationship among them.
Based on the results of these tests, we consider a heterogeneous dynamic panel model to estimate the relationship between agricultural credit and agricultural productivity. A combined autoregressive distributed lag (ARDL) panel approach, namely, the Mean group (MG) developed by Pesaran and Smith (1995), the Pooled mean group (PMG) developed by Pesaran et al. (1999) and the Dynamic Fixed Effect (DFE) estimator to estimate the short-run and long-run linkages between agricultural value added and agricultural productivity with other control variables.
Pesaran et al. (1999) developed two estimators to estimate the panel ARDL model: MG (Mean Group Estimation) and PMG (Pooled Mean Group Estimation). The MG estimator places no restrictions on the coefficients in the long-run. PMG approach, associated with pooling and averaging of the coefficients over the cross-sectional units, allows a greater degree of parameter heterogeneity then the usual estimator procedures by imposing common long-run relationship across countries while allowing heterogeneity for the short-run. In other words, PMG restricts the long-run coefficients but allows the constants, error term variances, and short-run coefficients to vary. Therefore, PMG allows short-run coefficients and error variances to vary across different groups while assuming a homogeneous long-run relationship between dependent and explanatory variables (Tong et al., 2016). By contrast, DFE model further limits the speed of adjustment coefficient and the short-run coefficient to be the same or equal and subject to the bias between the error term and the lagged dependent variable. However, the model features country-specific intercepts allowing different intercepts and groups.
As shown in Pesaran and Shin (1996), the aim of panel ARDL approach is to estimate the relationship between agricultural productivity and agricultural credit and can be specified by the following equation:
\(Y_{\text{it}}=\sum_{j=1}^{p}{\beta_{i,j}Y_{i,t-j}+\sum_{j=0}^{q}{\gamma_{i,j}X_{i,t-j}+\mu_{i}+\varepsilon_{\text{it}}}}\)(1)
By rearranging terms such as:
\(Y_{\text{it}}=\varnothing_{i}\left(Y_{i,t-1}-\theta_{i}X_{i,t}\right)+\sum_{j=1}^{p-1}{\beta_{\text{ij}}^{{}^{\prime}}Y_{i,t-j}+\sum_{j=0}^{q-1}{\gamma_{\text{ij}}^{{}^{\prime}}X_{i,t-j}}}+\mu_{i}+\varepsilon_{\text{it}}\)(2)
with i and t representing country and time respectively, Y is the agriculture, forestry, and fishing value added as a percentage of GDP, \(X_{i,t}\) is a \(kx1\) vector of explanatory variables containing the credit to agriculture, net foreign direct investments as a percentage of GDP, annual rate of changes in CPI (%) and general government total expenditure as a percentage of GDP,\(\varnothing_{i}\) is the group-specific speed of adjustment coefficient, \(\theta_{i}\) is the long-run coefficients of explanatory variables, ECT=[\(Y_{i,t-1}-\theta_{i}X_{i,t}]\) is the error correction term and finally, \(\beta^{{}^{\prime}}\) and \(\gamma^{{}^{\prime}}\) are represent the short run
coefficients linking agriculture, forestry, and fishing value added with its past values and the variables of interest\(X_{i,t}\)
Further, one more estimation technique is employed as a part of robustness check. The Dynamic Ordinary Least Square (DOLS) proposed by Stock and Watson (1993) and later extended by Kao and Chiang (2001). DOLS method can be applied with mixed and higher orders of integration and frequently used in estimating long-run nexus for heterogeneous panel by correcting simultaneity, endogeneity, serial correlation and small sample bias among the regressor. Fully Modified OLS (FMOLS) is also commonly used to check the robustness of DOLS results in the literature; however, this approach essentially requires that all variables must have the same integration order (Yahyaoui and Bouchoucha, 2021). As Kao and Chiang (2001) proposed, DOLS outperforms FMOLS estimators in terms of mean biases. Ali et al. (2017) notes that the most significant benefit of DOLS is that the test considers the mixed order of integration of variables in the cointegration framework. (Stock and Watson, 1993, Masih and Masih, 1996; Kumar et al., 2021).
Data
An annual balanced panel data set of 53 countries over the 2000-2018 period was used. The data for a selection of countries are drawn from the World Development Indicators (WDI) database provided by the World Bank (2019) and listed in Appendix A. The descriptions of the variables used in this study are presented in Table 1. The selection of the countries and time period is limited by data availability. Agriculture, forestry, and fishing value added series (AGV), the output of the agricultural sector less the value of intermediate inputs as identified by FAO statistical annex, were used as a proxy for agricultural productivity. (For more information, please see https://www.fao.org/3/a0050e/a0050e10.htm.) AGC is the Credit to agriculture, forestry and fishing, in constant LCU, INF is Inflation as the annual percentage change in consumer prices, FDI is the net foreign direct investments as a percentage of GDP, INF is the annual rate of changes in CPI (%) and GOV is the general government total expenditure as a percentage of GDP. AGV and AGC are converted into natural logarithm for consistent and reliable empirical results. As it can be seen on Figure 1., there is a strong correlation between agriculture productivity and agricultural credits.