1 Introduction

In most developing countries, credit rationing represents a significant obstacle to growth and development (Bond et al. 2015; Luan et al. 2015 and Massenot and Straub 2015). The lack of credit available from formal institutions (i.e. banks, international organisations, microcredit programmes, etc.) induces the economic agents rationed in the formal market to resort to informal credit. This set of informal sources includes relatives, friends and other moneylenders. However, the diffusion of such a practice is generally a concern: informal moneylenders may follow illegal procedures, such as, for example, charging usury interest rates. Informal credit is mostly diffused in the agricultural sector (Linh et al. 2019), mainly because of the small household producers that characterise it as well as the fact that commercial banks have fewer branches and international organisations have less penetration there.

Access to credit generally requires the provision of some collateral, which in rural areas is generally represented by land in the case of formal credit, while informal moneylenders are much less likely to ask for such a guarantee (see Linh et al. 2019 for a review of the literature). However, the literature has considered essentially only the extent of the land cultivated by the borrowers, and only a few articles include land ownership in their analyses. Yet, farmers often do not own all the land they cultivate, but rent some, and this may have different consequences on their access to credit and the interest rate applied to it. Renters may have less collateral to provide and therefore may be more rationed, so may more often resort to informal credit (Khoi et al. 2013) and pay higher costs than owners.

Using data provided by the World Bank for a sample of farmers from the Mekong Delta, this paper examines the relationship between land ownership, access to informal credit and its cost. Vietnam is an interesting country, as the redistribution of land from collectives to farmers was one of the first and most relevant policies enacted as part of the wave of liberalisations started in 1988 (Stampini and Davis 2009). In particular, concerning the agricultural sector, a reform enacted in 1993 transferred land-use rights from the state to private farmers through land-use certificates, which assign exclusive and renewable rights to private individuals for a period between 20 and 50 years, depending on the crops grown (Le 2020 and Ho 2021).Footnote 1 These certificates are freely tradable as if they are property rights and no payment is due by the holders to the state. Holders are entitled to rent land to tenants. Land-use rights and rent contracts produce two levels of tenure insecurity. While the state keeps ownership of the land, the renewability of use rights renders the holders of such certificates more secure than those renting land from them. In addition, Thuy et al. (2008) showed that poor households cannot afford land-use certificates, meaning that they are forced to choose to rent the land that they cultivate.

In this context, the role of credit rationing in rural areas be crucial in shaping and maintaining this dual situation. The area of the Mekong Delta is particularly interesting because this is the region where the land reform was the fastest: Hare (2008) highlights that, in 1997, more than 71% of the farmers living there had acquired land-use rights, in comparison to 7.6% of those in the other large river delta region of Vietnam (that of the Red River). Moreover, the Mekong Delta is the most important agricultural region in the country (Van Hon and Ninh 2020).

Linh et al. (2019) demonstrate that the credit market in rural Vietnam is very fragmented, with insufficient governmental interventions and a large diffusion of informal lenders, and highlight the importance of land ownership in the credit market. In contrast to most of the other works, the authors focus on the share of the land cultivated by a household that the household itself owns through land-use rights. Their results show that such ownership significantly affects whether farmers access informal credit and the interest rates applied on loans.

The next section presents the relevant literature. Section 3 presents the data and the empirical methodology, followed by Sect. 4, which provides the results and discussion. Finally, Sect. 5 concludes this paper.

2 Related Literature

Dual credit markets exist, as banks, international institutions and official non-governmental organisations are scarcely present in rural areas of developing countries and often request collateral that farmers cannot provide. Such a situation leads to formal credit rationing; consequently, informal moneylenders represent a large share of the total credit supply, especially in the countryside (Stampini and Davis 2009).

2.1 Credit Markets in Rural Vietnam

Informal moneylenders are widespread in the rural areas of Vietnam (Linh et al. 2019; Nguyen et al. 2020), mainly because of rationing in the formal market; indeed, potential borrowers have insufficient assets to provide as collateral and the absence of bank branches in many villages hampers the access to formal moneylenders (Pham and Lensink 2007; Linh et al. 2019). Nguyen-Viet and van den Berg (2011) show that, in 2006, 21.5% of poor farmers and 10.0% of those classified as non-poor resorted to informal credit. Barlsund and Tarp (2008) report that, according to some studies, credit rationing in the formal market pushes Vietnamese farmers to borrow from both formal and informal lenders to obtain enough money to finance their projects and needs; in other words, the two types of credit may appear complementary. The same authors demonstrate that formal loans mainly finance production development and asset accumulation. Instead, informal borrowing is used to smooth consumption over the year. Moreover, they also note that formal lenders provide larger amounts for longer time spans than informal lenders. This may also explain why the former ask for real rights as collateral much more often than the latter. In addition, the authors find that the extent of land owned by the borrower does not affect the probability of resorting to informal credit and has a negative and significant effect on the amount borrowed. However, for some households the two types of lenders are not complementary; in particular, poor farmers cannot provide collateral to access formal credit and instead borrow from informal lenders, often exploiting their social capital (i.e. personal and family members’ relationships).Footnote 2 In addition, the same authors highlight that individuals with a past bad credit history are unlikely to obtain loans from formal lenders, while the informal type may be more willing to provide them with some credit.

Truong et al. (2022) show that informal loans in rural areas of Vietnam are more commonly used to respond to unpredictable and urgent needs, as the procedures to obtain them are faster and people are less able to provide collateral in such circumstances. This result is more evident for households with higher levels of social capital (Dang et al. 2023b). In this sense, informal credit complements loans from formal lenders. However, other reasons can also explain this phenomenon: Nguyen (2022) demonstrates that education and labour diversification (i.e. working both in and outside the household’s farm) is positively correlated with access to formal credit in rural Vietnam. Labour diversification reduces the debtor’s risk, while higher education renders farmers more productive and more aware of the opportunities provided by formal lenders. Finally, Dang et al. (2023a) highlight that social trust, which households build over time, is valuable collateral for informal lenders, while it is not for the formal ones; therefore, when farmers lack other forms of collateral or simply do not want to pledge them (Trinh et al. 2022), they can borrow from informal moneylenders, relying on their social trust as collateral.

The extant literature shows that the aforementioned obstacles to accessing formal credit have been persistent over time. Recently, Anh et al. (2022) have demonstrated that distance from bank branches, poor endowments of land and capital and low levels of human capital are the main reasons why Vietnamese farmers struggle to access formal credit, although, during the first two decades of the twenty-first century, the government has tried to increase formal credit supply in the countryside (Linh and Hang 2022).

While formal credit rationing is a persistent problem in rural Vietnam, the government has attempted to mitigate it. Nguyen (2008) and Tran and Dinh (2021) highlight that government programmes aimed at enhancing microcredit positively impact poverty reduction. However, the procedures intended to reduce the use of informal credit are still too complex to be effective, helping this source of loans to survive, especially among poor people (Nguyen and van der Berg 2014). For this category in particular, Phan et al. (2019) show that the burden of interests may be heavy enough to reduce consumption in the short term—although they may have a positive impact in the long run—thus reducing the demand for microcredit. In addition, some of the same issues preventing farmers from accessing formal credit affect the microcredit market: Khoi et al. (2013) confirm that distance from main roads reduces the ability to access microcredit programmes. Moreover, Mookherjee and Motta (2016) reveal that the standardised procedures used by formal credit institutions (including microcredit) are not always optimal for small farms.

2.2 The Importance of Land as Collateral

Land constitutes a major form of collateral in the formal market, but informal moneylenders rarely request it (Feder et al. 1988; Barslund and Tarp 2008; Duy et al. 2012). In addition, agricultural machinery is also an important type of collateral in the countryside (Pham and Lensink 2007). However, this evidence is not conclusive: Guirkinger (2008) finds that as the size of the farm increases, so does the probability of accessing informal loans; on the one hand, this may be simply because larger farms need more credit and therefore they borrow in both the formal and the informal markets. Conversely, this does not exclude the possibility that relatively large plots are a kind of collateral that is also appreciated in the informal market. However, the author focusses on the total size of the farm, without distinguishing between owned and rented land. For this reason, the first interpretation of her results—i.e. larger farms need larger loans—is the most tenable. The extent of cultivated land is also relevant for other reasons: Dalla Pellegrina (2011) shows that the larger the farm size, the greater the value of the working capital employed, which could also serve as collateral. Giné (2011) states that land titling programmes in Thailand have reduced the share of economic agents resorting to informal credit. The reason for this seems to be that land ownership allows farmers to circumvent the obstacle of credit rationing. Shoji et al. (2012) reach the same conclusions by studying the credit market in Sri Lanka. Khoi et al. (2013) demonstrate that owning land (assessed by the means of a dummy equal to 1 if the household owns some land) has a positive and statistically significant influence on accessing informal credit in rural areas of the Mekong Delta region. This article thus provides further evidence that land ownership frees farmers from the need to resort to the informal market. However, as already noted, the authors confirm that formal and informal credit are complementary, as they serve different purposes.

Another dimension of credit which affects a household’s financial condition and—more generally—rural development is the interest rate paid by the borrower. Fewer works exist on the relationship between the cost of credit and land ownership than those linking land ownership and access to credit. Kochar (1997) finds a negative correlation between the extent of the land owned and the cost of formal credit. Mallick (2012) shows that, as the percentage of households with less than 10 decimals of land in a village increases, the interest rates charged by informal moneylenders tend to decrease. However, this result is not robust to the different specifications of the estimated model.Footnote 3

Given its legislation on land, in Vietnam land-use certificates—not land property—can be provided as collateral for loans. Indeed, they are tradable; therefore, if a debtor cannot repay the loan, the creditor may take a land-use certificate as repayment. As previously highlighted, the extant literature considers the certificates equivalent to ownership, given their duration and renewability. In line with this, current studies on Vietnam show that households with more assets (particularly land) are less credit-rationed in the formal credit market (Duong and Izumida 2002) and those cultivating more land borrow more in the formal market and less from informal lenders (Barslund and Trap 2008). Indeed, people who are less rationed in the formal market may decrease the demand for informal credit. The same results also emerge in the Mekong River Delta: Khoi et al. (2013) find that access to both formal and informal credit is easier for landowners than for tenants. More recently, Thu and Goto (2020) have demonstrated that land ownership correlates positively with higher investment in irrigation and soil and water conservation; the evidence that they provide suggests that land ownership increases the need to borrow money to finance such investments. However, Ho (2021) argues that, as land-use certificates involve more insecure rights than full ownership, the Vietnamese legislation may hinder both renters and owners of use rights investments.

3 Data and Methodology

3.1 Data

This paper uses data from the World Bank covering a rural area in the Mekong Delta region.Footnote 4 The main purpose of the survey was to inquire about the use of pesticides and the socio-economic conditions that lead to their misuse. The questions about the economic situation of the households included access to credit, wealth, income, employment, ownership and rental of cultivated land. The World Bank, the Centre of Occupational and Environmental Health, the Vietnam Association of Occupational Health and the University of Ho Chi Minh City designed the survey. The Vietnamese partners of this study (i.e. the last three mentioned organisations) were responsible for administering the questionnaire through face-to-face interviews. All the collected answers were anonymised, and all the interviewees were aware of this. This survey covered the following districts: An Phu, Cai Lay, Chau Thanh, Cho Gao, Tan Thanh, Thot Not, Thu Thua, Tieu Can, Tra Cu and Vi Than.

In total, 603 households responded to the questionnaire; while this sample size is limited, on average it is not very different from other studies on formal and informal lending in developing countries (e.g. Feder et al. 1988; Duong and Izumida 2002; Barlsund and Tarp 2008; Dalla Pellegrina 2011; Duy et al. 2012 and Khoi et al. 2013).

Amounts borrowed from formal and informal lenders are reported in dongs; the interest rate is monthly. The percentage of land owned refers to the land cultivated by the household and the surface used for residential purposes. The last is included as a control for two reasons: first, it works as a proxy for the household’s wealth (other indicators of it are included and are discussed further in this section); second, it is part of the total land used by the household, which could also borrow money for residential improvements. However, the main variable of interest is the share of cultivated land for which the household has a land-use certificate. The use of the share of ownership of an asset as collateral instead of the value of the collateral itself is not uncommon in the economics literature (e.g. Liu et al. 2016; Bellucci et al. 2021).

Friends, relatives or credit associations may supply informal credit (e.g. Kumar et al. 2017; Ojong 2019). Some of these potential lenders (such as relatives, for instance) may not apply interest rates; therefore, money borrowed from relatives is excluded from the analyses. In addition, information on this type of source is fragmentary in the dataset, as the interviewees preferred not to answer the question about money borrowed from relatives. The survey includes all the other informal lenders under the same “informal lenders” option.

Table 1 presents the descriptive statistics of the variables used in the analyses. The figures show that, on average, more money is borrowed from formal than informal lenders; however, the latter charge higher interest rates than the former. The largest share of income comes from crops, although off-farm jobs and remittances represent more than one third of the total household income. These figures are consistent with the distribution of working time: about three quarters of it is spent on the farm. Plots are generally small (less than 2 ha) and the main crop is rice, followed by various types of fruits, which are very relevant for the analysis. Indeed, all the farms included in the survey maintain orchards, extending the land-use rights to 50 years (i.e. the maximum allowed by law). Households generally own at least one refrigerator (indispensable in the climate of the Mekong Delta), while other durable possessions—TV sets in particular—are less common. Less than 60% of the households declared that they owned a motorbike. Given the difficulty of accessing formal credit because of the distance between villages and bank branches, this figure is relevant, as it helps explain why such issues exist and are common to several households.

Table 1 Descriptive statistics of the variables used in the regressions (s.d. in brackets)

3.2 Empirical Methodology

The first research question is whether land ownership increases the probability of accessing formal or informal credit. In this case, the dependent variable is a dummy, taking the value of 1 if the household borrowed from a formal or informal lender in the 12 months before the interview. The dummy equals 0 otherwise. The second research question is relative to the relationship existing between land ownership and the amounts borrowed from the two types of lenders. In this case, the dependent variable is the amount of dongs (in millions) borrowed. The equation to be estimated using the instrumental variable (IV) probit is thus the following:

$$P\left({A}_{j}^{f,i}=1 |{L}_{j}, {X}_{j}\right)=\Phi \left({\beta }_{0}+ {\beta }_{1}{L}_{j}+ \gamma {X}_{j}+ {u}_{j}\right)$$
(1)

where \(P\left({A}_{j}^{f,i}=1 |{L}_{j}, {X}_{j}\right)\) is the probability that household \(j\) accesses (\({A}_{j})\) either formal (\(f\)) or informal (\(i\)) credit conditioned on the share of land owned (\({L}_{j}\)) and a set of other controls (\({X}_{j})\) at the household level.

The effect of land ownership is estimated using tobit regressions, censored at 0, as many farmers declared that they had not borrowed any money in the last 12 months, but they may have simply preferred not to disclose their loans to the interviewer—a World Bank employee—especially when their source was informal. There is no upper censoring, as loans may potentially reach any sum and the literature does suggest over-reporting as a possibility in this case. The following equation is thus estimated:

$${BM}_{j}^{f,i}= {\beta }_{0}+ {\beta }_{1}{L}_{j}+ \gamma {X}_{j}+ {u}_{j}$$
(2)

where \({BM}_{j}^{f,1}\) is the money borrowed by household \(j\) from either formal (\(f\)) or informal (\(i\)) lenders. The other elements are the same as in (1).

The third research question relates to the relationship between land ownership and the interest rate paid to informal lenders. In this case, the dependent variable is the monthly interest rate. While the relationship between land ownership and the amount borrowed is likely to be linear, as the value of the collateral (land) increases linearly with ownership and, therefore, the same is expected to happen to the lenders’ willingness and availability to lend, interest rates likely do not depend linearly on collateral size, but present diminishing marginal value (Karapetyan and Stacescu 2014; Liu et al. 2016). Hence, the logarithm of land ownership enters the estimated equation:

$${r}_{j}^{i}= {\beta }_{0}+ {\beta }_{1}log\left({L}_{j}\right)+ \gamma {X}_{j}+ {u}_{j}$$
(3)

where \({r}_{j}^{i}\) is the monthly interest rate paid by borrowers to informal lenders, while the other elements are the same as in (1) and (2).

Several controls allow accounting for observable factors, which may affect the quantity and price of the borrowed money. The first set of variables captures the extent of cultivated (either owned or rented) land assigned to different crops (rice, other vegetables and orchards) or fish farming (an important activity in the Mekong Delta). Indeed, the value of land may depend on its use: in a region where the main crop is rice, paddies may be valued differently than land destined for other crops.

Other controls include the amount of money borrowed from both formal and informal lenders. The first enables capturing the total indebtedness of the household; this may indeed increase the interest rate requested, as the debtor may have already pawned a part of its land as collateral for the formal creditor, so decreasing the solvability of the debtor. Moreover, the inclusion of the amount of money borrowed from formal actors in the regressions, where the quantity of the money borrowed from informal lenders is the dependent variable, allows for testing the hypothesis of complementarity or substitutability between different credit sources. Finally, if a formal institution has lent large amounts to a household, this may entail that this household has a high degree of solvability and reliability; hence, this may have some negative effect on the interest rate required by the informal lender. The interest rate paid by the household on the amounts borrowed in the formal or informal market is also included, depending on the outcome variable. This serves two purposes: first, it measures the financial resources absorbed by the other lines of credit already open. Second, it is a synthetic measure of how risky that household is perceived as being by formal lenders and helps capture some possible variables—not included in the dataset—that affect creditworthiness.

Income at the household level is another important control; it is included and categorised by source: crops, remittances and in kind (transformed into dongs by the interviewer). Additional information about earnings comes from the share of working time that each of the household members spends inside or outside the farm. These people represent both an additional source of income and a way of diversifying its sources.Footnote 5 Moreover, following the extant research (Luan et al. 2015), the household’s wealth is approximated by a series of dummies, which capture the ownership of different goods: TV, radio, motorbike and refrigerator. To avoid the inclusion of 18 dummies, a principal component analysis (PCA) is run on these variables. The components with an eigenvalue larger than 1 are included in at least one of the model's specifications.Footnote 6 Table 2 shows that the first component correlates positively and statistically with almost all the variables included in the PCA; therefore, it is a summary of the household’s wealth, as approximated by the ownership of the items listed in the table. The second and the third components chiefly capture the ownership of electrical and electronic goods; the fourth represents mainly transportation means (bicycles and motorbikes) and ownership of PCs. Finally, the fifth component represents furniture, including related electronic goods such as radios and TVs. The questionnaire also indicates (through a dummy variable) whether the household is poor in absolute terms or not, according to the World Bank standards, i.e. a per capita income lower than 1 USD a day. Other controls include the yield per hectare and the price of the commodity grown in the third- and in the second-to-last crop seasons (i.e. the last two seasons before the interview). Bad crop seasons may indeed increase the interest rates required by the lenders because of both the worse solvability conditions of the farmers and an increase in the demand for loans (Barham et al. 1996).

Table 2 Correlations between the principal components retained for the analyses and the originating variables (s.e. in parentheses)

3.3 Endogeneity

Unobservable factors, however, may affect both the share of owned land and the quantity and price of credit borrowed from formal and informal lenders. An example of such possible factors is labour productivity, which is not present in the dataset used, but might be observed by local informal lenders, so affecting their attitude towards borrowers. For this reason, estimates resort to an IV approach to try to clean the results from the effects of unobserved, though relevant, factors. Two variables are identified as instruments: the leave-one-out mean of the owned land and the average quantity of pesticides per hectare sprayed during the crop season. The first considers the average share of land owned by the other farmers in the same district, calculated for each i-th household excluding the share of the i-th household itself. Such a technique is widely used in empirical economics (e.g. Dobbie and Song 2015; Rysman 2019; Cette et al. 2022). The choice of the second IV is based on Migheli (2017) and Thu and Goto (2020), who show that farmers tend to spray fewer agrochemicals on owned than rented land. If farmers spray fewer chemical pesticides, one might claim that they need less money; therefore, this instrument is also correlated with the dependent variable. However, the correlation coefficient between these two variables equals -0.05 and is not statistically different from zero.

A second issue worth addressing is the relationship between formal and informal credit. Indeed, the summarised literature highlights that the two sources of money may at times be substitutes for one another and at others be complementary. To check for this hypothesis, the distance between the farm and the nearest bank branch would be useful. Unfortunately, data on the distribution of bank branches in the Mekong Delta for past years are unavailable. Nevertheless, regressions may still try to capture the relationship between the two credit types. Indeed, if they were complementary, regressing informal on formal credit amount (or access) would result in a positive and statistically significant coefficient for the independent variable. Instead, if they were substitutes, the coefficient would be negative and statistically significant. To address this issue, the following section presents regressions using access to informal credit as the dependent variable and the quantity of money borrowed from formal lenders.Footnote 7 In this case as well, the regressions instrument the share of owned land using the leave-one-out mean described afore.

The empirical results shed some light on (1) the role of land ownership in the choice of accessing informal credit and (2) its price. Third, they allow for understanding whether formal and informal sources are substitutes, complements or independent from each other. First, IV probit regressions analyse the probability of demanding informal loans, with the main variable of interest being the share of cultivated land owned by the household. The second set of regressions presents IV tobit estimates examining the relationship that exists between formal and informal sources of credit. A third set of regressions provides the results of IV tobit regressions, where the dependent variable is the cost of the informal credit (i.e. the monthly interest rate paid by the farmer). The reason for using a tobit estimation stems from the fact that several farmers declared that they had not borrowed from informal lenders, but this information may be voluntarily underreported: not all the borrowers may want to disclose that they resorted to informal credit to the interviewer (an officer of the World Bank). Therefore, running tobit regressions—censored at 0—on the whole dataset is a better strategy than deleting the households declaring no informal borrowing (García and Labeaga 1996; Humphreys 2013).

4 Results

Table 3 presents the results (marginal effects) of IV probit regressions, where the dependent variable is a dummy taking the value of 1 if the household resorted to informal (upper panel) or formal (lower panel) credit during the last 12 months. The share of owned land is instrumented using its leave-one-out mean, as described in the previous section. The figures demonstrate that the probability of resorting to informal or formal credit correlates negatively with the share of cultivated land owned by the household. This evidence seems to suggest that farmers renting the land that they cultivate are more likely to borrow money from either formal or informal lenders. For the sake of conciseness, the table does not show the marginal effects of all the controls; however, those relative to the amount and cost of money borrowed from the other type of lenders do not reveal any statistically significant association. This result seems to suggest that formal and informal credit are perceived as different, as well as that households renting land are financially more fragile than the others because they need to access credit more, and are thus more indebted. Income from crops is positively and significantly (though not robustly) associated with accessing formal, but not informal, credit. This seems to suggest that households whose productions yield more in monetary terms have more ability to access formal credit. Though not robust through different specifications, this result may suggest that credit rationing is present and only farmers who can provide collateral in the form of income as well can access formal lenders.

Table 3 Relationship between land-use rights and resorting to informal credit

Table 4 presents the results of IV tobit estimates, where the dependent variable is the quantity of money borrowed from informal (left panel) or formal (right panel) lenders. The figures suggest that there is no association between the amounts borrowed from two types of lenders and therefore the two forms of credit are apparently not substitutes or complements. Instead, the link between informal credit and the share of owned land is negative and statistically significant for informal credit, while the opposite holds for money borrowed from formal sources. These results suggest that land is valuable collateral in the formal credit market, as the quantities borrowed increase with ownership. In contrast, the quantity of money demanded in the informal market decreases as land ownership increases. Taken together, these results suggest that, in reality, formal and informal credit are substitutes and land ownership mediates such substitutability: tenants, who cannot provide enough collateral to formal lenders, resort to the informal agents; owners instead borrow from formal lenders, to which they can provide land as collateral.

Table 4 Relationship between the amount borrowed from informal and formal lenders (IV tobit estimates, standard errors in brackets)

The evidence highlighted above supports the previous conclusion that tenants are more vulnerable than owners, as the informal credit market may protect borrowers less than the formal market, which is subject to regulation and monitoring. Instead, land-renting farmers are exposed to the risks of an unregulated and unmonitored market.

A major risk for borrowers in informal markets relates to the interest rate applied by lenders. Regulation and the existence of contracts ensure that formal lenders cannot charge usury interests, while in the informal market this is a likely possibility. However, the debtor’s riskiness may also play a role in this market in determining the interest rate applied by lenders. In other words, while it does not seem to be the main collateral in the informal market, land ownership may nevertheless play a role as a sort of insurance against usury.

Table 5 presents the results of regressions where the dependent variable is the monthly interest rate applied by informal lenders and the regressor of interest is the share of land owned by the borrower. In line with economic theory, the first outcome is that the interest rate level depends positively on the amount borrowed. As expected, the share of land owned by the household has a negative and statistically significant impact on the interest rate applied to informal loans. This effect is robust to different specifications and increases after adding additional controls (although the addition of variables controlling for the household’s wealth and land yields decreases it slightly). The coefficient has the expected sign: the cost of the informal debt drops as land ownership rises, suggesting that land constitutes a form of collateral in the informal credit market. The data do not allow for disentangling whether land is direct collateral (i.e. if the farmer does not pay back the loan, then the land-use right is transferred to the creditor), or indirect (i.e. it is a proxy for the household’s ability of producing income). However, the magnitude of the coefficient taken in absolute terms surges after the addition of controls for other assets owned by the household. This suggests that land ownership represents a direct rather than an indirect collateral for the informal lender.Footnote 8 This interpretation of the results is particularly relevant: for a given amount borrowed in the informal credit market, households owning small shares of land pay higher interests and therefore have a greater risk of defaulting. In the last case they would lose portions of the land owned, risking a sort of “social downgrade” from being independent farmers (who cultivate owned fields) to salaried farmers. An alternative to this “downgrade”, which is common in developing countries, is migrating from the countryside to the industrialising towns, usually engendering additional social problems in these areas (Lipton 1980; Marx et al. 2013).

Table 5 Cost of credit and land-use rights

A relevant issue related to accessing informal credit and its cost is social capital. Indeed, households who have lived in the same place for long periods have higher levels (i.e. denser and stronger local networks) than recently settled ones. In the rural areas of developing countries, access to informal credit also depends on social networks (Guirkinger 2008) and social capital constitutes a major collateral for loans (Giné, 2011),Footnote 9 which reduces the risk premium (Feder et al. 1988). In addition, Pham and Lensink (2007) show that social proximity with informal lenders is negatively associated with the risk of default. However, the fourth specification of the estimated model includes the number of years that have elapsed since the household’s settlement in the village, minimising the risk that the unobserved level of social capital (which is proxied by this control) substantially biases the results.

A final noteworthy result is the role of the different crops grown by farmers; while they were expected to affect the value of land as collateral for the reasons mentioned in Sect. 3.2, no statistically significant effect of them on any of the dependent variables emerges. The reason may be that lenders who foreclose land-use certificates sell them within a short period of time. In addition, all the certificates were issued around 1993 (the year of the reform) and all the farmers included in the survey maintain orchards—thus the life of their certificates is 50 years. Consequently, all the land-use rights had more or less the same residual duration at the time of the interview.Footnote 10

The empirical evidence presented in this section suggests that the contribution of land ownership to the cost of informal credit and the probability of resorting to it is larger when the share of owned land is small. Because of this, to offer a clearer interpretation of the results and to show the relevance of land ownership in determining the interest rate paid by borrowers in the informal credit market of Mekong Delta, Fig. 1 provides a graphical summary of the impact of the ownership share on the interest rate. The solid curve in the graph represents the difference between the average interest rate paid by households whose share of owned land is equal to x and the average interest rate paid by the farmers who own all the land that they cultivate (for which, indeed, the difference is equal to 0). This representation highlights that the premium does not decrease linearly and that it declines fast when the share of owned land is relatively very small or very large. Figure 2 represents the marginal variation in the premium for any given ownership share. This marginal rate is always negative, confirming that, as the share of cultivated land owned increases, the interest rate paid moves in the opposite direction. The form of the marginal rate curve is concave, graphically depicting a previously observation: the marginal rate is large in absolute terms for ownership shares below 10% and becomes considerable again for ownership shares greater than 80%. The dashed lines in the two graphs represent the linear interpolations of the data. They are useful for comparing the punctual values of the curves to the average trends indicated by these dashed lines.

Fig. 1
figure 1

Land ownership and change in the interest rate paid on informal credit

Fig. 2
figure 2

Marginal variation of the interest rate paid by share of land ownership

The cost of informal credit represents a non-negligible burden for the farmers. The average amount borrowed by households who access the informal credit market in the Mekong Delta is 3,508,125 dongs, slightly less than the average yearly per capita income of the farmers in the sample. Given an average household’s size of 5.02, the figures show that the average informal loan amounts to approximately 23% of the total yearly household income. This implies that a 1 percentage point differential in the monthly interest rate paid represents—on a yearly base—an absorption of 2.30% of the household yearly income. In other words, a household owning 50% of the cultivated land would face an additional burden of about 4.50% of its yearly income with respect to a household that owns all the cultivated land. In addition, the estimates suggest that the monthly interest rate in the informal market grows by approximately 2 percentage points for each additional million dong borrowed, which represents a non-negligible burden.

Given the previous results, a caveat is needed. The possibility of accessing credit may influence farmers’ choices between renting and buying cultivable land. In other words, there may be reverse causality between the variable used as the dependent in the previous regressions and the regressor of interest. On the one hand, this possibility does not change the descriptive results presented in the previous tables, which show that, as the share of owned land rises, the interest rate paid for money informally borrowed decreases. On the other hand, because of this, farmers may deem it more convenient to buy land-use certificates than to rent plots, as the first choice would reduce the cost of credit. Therefore, while the absence of formal credit may push farmers to approach informal lenders, the bias induced by this situation should go in the direction of increasing land ownership. Nevertheless, the IV approach should mitigate the possible existence of reverse causality.

Some additional reflections on the role of crop income may also be useful. Indeed, this paper has assumed so far that land is usable as a collateral to access credit, without considering income from crops within the set of possible collaterals. Both debts and interests are paid with money derived from selling crops; therefore, such flows might also be (at least partially) used as collateral. In other words, the diminishing interest rate paid by farmers who own larger plots may also depend on the fact that they earn more income from crops and, therefore, as they are more solvable, the risk premium applied to them is lower. This may certainly be true and suggests two additional considerations. First, crop income depends on the size of the owned land: indeed, the larger the plot, the higher the likely income from crops. However, this means that owned land is a proxy for crop income. In particular, one may wonder why the amount of owned instead of total (owned + rented) land is relevant. The reason is that farmers pay to rent land, thereby decreasing the income available for repaying debts. Therefore, owned land is a better proxy for the ability to repay debts than total land. The second consideration is related to the volatility of crop prices, which renders income from agricultural products uncertain and, therefore, a riskier collateral than land. One might argue that when crop prices fall so do land prices; however, if we assume that the price of land is equal to the discounted value of its future monetary yields, crop price volatility has less impact on the value of plots than on yearly farmers’ income. Consequently, land is a better collateral than crop income. A final technical comment may also help in better understanding the results presented in this section: crop prices are used as a control in some of the presented specifications. Therefore, the effect of owned land on the probability of accessing informal credit and its cost is (in the specifications that include crop income) net of the contribution of this income source.

5 Conclusions

The evidence provided in this paper suggests that land ownership decreases both the need to access informal credit and the cost of it. Moreover, it allows for borrowing greater amounts of money from formal lenders, although it also decreases the need for this type of credit.

The Vietnamese land reforms, which redistributed the ownership of the land from the state to small farmers, may have contributed to reducing the problem of informal credit and to alleviating its burden on farmers’ budgets. Therefore, the implementation of effective land reforms seems to have a positive impact on the cost and quality of credit. On the one hand, further analyses should aim to analyse the roles of different informal lenders. Such a strategy would increase the understanding of the phenomenon of informal credit in the Mekong Delta and elsewhere. On the other hand, although this study does not differentiate between the various informal lenders, it shows the link between the interest rates applied by them and the economic conditions of households in the rural Mekong Delta, suggesting the need for policies that may strengthen the presence of formal moneylenders.

Other possible policies may entail the provision of alternative jobs in the countryside of Vietnam and of developing countries in general. Indeed, the literature shows that diversifying the sources of income reduces both the demand for informal credit and its cost. Some developing countries have already planned and implemented programmes aimed at sustaining the income of the population living in rural areas (for example, the Mahatma Gandhi National Rural Employment Guarantee Act in India).Footnote 11 However, these programmes should also consider that people from relatively rich farms are more likely to access off-farm jobs, outside of government plans. Indeed, as Barrett et al. (2001a) point out, in the off-farm rural job markets, candidates from relatively rich farmer households are—on average—more skilled than candidates from poor households. This increases the probability of finding an off-farm job for the former in comparison to the latter.

An open question concerning land reforms is how much land should be allocated to each household (or, alternatively, how much land should be distributed per capita) to allow the farmers enough economic independence. The plots allocated in the past Vietnamese reform may have underestimated the farmers’ needs (for example, neglecting farmers’ fertility), and this may help to explain the persistence of informal credit markets in the Mekong Delta. In addition, land reforms require effective protection of property rights and land administration. Governments should also enact policies aimed at helping farmers to support investments and to use agrochemicals and natural products properly (Migheli 2017, 2021). In addition, the evidence provided by this paper shows the need for interventions aimed at increasing the presence of formal moneylenders and supporting farmers with the provision of collateral. Ho (2021) highlights that, while the land reform has increased the farmers’ well-being in Vietnam, the taxes levied on them are too high and land-use certificates are source of insecurity for the holders, which might be reduced by transferring property instead of use rights. In addition, Tran and Vu (2019) claim that farmers have been assigned plots that are too small, which do not allow households to significantly improve their conditions.

The analysis presented in this paper has some limitations. First, the results are based on cross-sectional data, while a panel would have allowed for a causal inquiry. Second, and strictly related to the previous point, the data do not provide information about the past credit history of the surveyed households; this information would have helped in understanding the reasons why farmers resort to informal lenders. Third, information about the nearest bank branches would have served the same purpose, enabling this study to control for distance from banking lenders. Unfortunately, neither the State Bank of Vietnam nor other sources provide information about past years that would allow for locating branches at the time of the survey.