INTRODUCTION
The dairy sector constitutes one of the most important components of Indian agriculture, contributing significantly to agricultural gross value added, nutritional security, employment generation, and rural livelihoods. Its contribution extends beyond value addition in allied sectors to generating employment and income opportunities for millions of people in both urban and rural areas (Sellappan et al., 2025). India remains the world's leading producer of milk, accounting for nearly 22% of global milk output (FAO, 2021). However, sustaining this leadership will require substantial improvements in livestock productivity, input use efficiency, and investment in dairy production systems to meet the growing domestic demand for milk and milk products. In recent years, dairy farming has become increasingly capital-intensive, owing to rising expenditure on improved dairy animals, feed and fodder, veterinary healthcare, mechanization, housing infrastructure, and breeding services. Simultaneously, dairy farmers continue to face production risks arising from fluctuating feed prices, animal diseases, climate variability, and market uncertainties. These challenges often constrain farmers' capacity to invest using their own resources, making timely access to affordable institutional finance increasingly important for sustaining and expanding dairy enterprises. Consequently, institutional credit has emerged as a key policy instrument for facilitating productive investments, technology adoption, asset creation, and improved farm productivity (Chen et al., 2021).
Institutional credit, provided through commercial banks, cooperative banks, regional rural banks, Primary Agricultural Cooperative Credit Societies (PACS), and the Kisan Credit Card (KCC) scheme, plays a crucial role in supporting dairy development by meeting the working capital and investment requirements of farm households. Access to affordable credit enables farmers to purchase quality dairy animals, improve housing facilities, invest in farm machinery, adopt improved feeding and breeding practices, and manage production risks more effectively (Raj et al., 2025). Previous studies have demonstrated that institutional credit contributes positively to farm productivity, asset creation, income generation, and food security (Misra et al., 2016; Kumar et al., 2020; Shivaswamy et al., 2020). Nevertheless, access to institutional credit remains uneven across regions and farmer categories, which is due to constraints such as collateral requirements, procedural complexities, limited financial literacy, and differential access to formal financial institutions (Deepika et al., 2026).
The livestock sector in Tamil Nadu demonstrates significant potential for expanding dairy production. The sector plays a vital role in the socio-economic landscape of this state, contributing significantly to agricultural growth, rural employment and rural livelihoods. The state has witnessed a positive trend in the milk production, with an average annual growth rate of 2.83% over the last decade (2010–2020). According to the 20th Livestock Census, Tamil Nadu recorded a 4.75% increase in milk production and a 5.51% rise in livestock population. Despite this growth, Tamil Nadu ranks 11th in milk production, with a per capita availability of 316 grams of milk per day which is below the national average. Notably, the state boasts a robust banking network with 12,659 bank branches (NABARD, 2023). This extensive banking infrastructure indicates an environment conducive to credit disbursal and financial services accessibility throughout Tamil Nadu. Despite these favourable institutional conditions, milk productivity and per capita milk availability in the state remain below the national average, indicating the need for greater investment in dairy production systems. Strengthening access to institutional credit is therefore expected to play a vital role in facilitating investment-led productivity improvements and enhancing the competitiveness of the dairy sector. Although a substantial body of literature has examined agricultural credit and its influence on farm performance, relatively few studies have comprehensively evaluated the impact of institutional credit on multiple dimensions of dairy farm performance in Tamil Nadu. Existing studies have largely focused on credit accessibility, repayment behaviour, or sectoral credit flow, with limited attention to the causal impact of institutional credit on farm income, employment generation, and farm mechanization after accounting for farmers' self-selection into borrowing. Addressing this gap is important for designing evidence-based credit policies that promote sustainable dairy development.
Against this backdrop, the present study evaluates the impact of institutional credit on farm income, employment generation, and farm mechanization among dairy farmers in Tamil Nadu using an Endogenous Switching Regression (ESR) framework that accounts for both observable and unobservable selection bias.
METHODOLOGY
Study area, sampling design and data collection
Tamil Nadu is one of India's leading dairy-producing states and has witnessed steady growth in milk production over the past decade. The state accounted for approximately 5.39% of the country's livestock population and recorded an average annual milk production growth rate of 2.83% during 2010–2020, highlighting its potential for dairy sector expansion. Owing to its well-developed dairy cooperative network and variation in institutional credit disbursement across districts, Tamil Nadu was selected as the study area. A multistage sampling technique was employed for selecting the respondents. In the first stage, districts were classified into high and low institutional credit disbursement categories based on official credit disbursement statistics. One district was randomly selected from each category, namely Coimbatore from the high credit disbursement category and Perambalur from the low credit disbursement category. In the second stage, two blocks were randomly selected from each district. Accordingly, Pollachi and Thondamuthur blocks were selected from Coimbatore district, while Perambalur and Veppanthattai blocks were selected from Perambalur district.
The study was based on primary data collected during 2023 through a structured interview schedule administered personally by the researcher. Information was collected on household socio-economic characteristics, institutional credit utilization, dairy production, income, employment generation, farm mechanization, and other farm-specific attributes.
Borrower households were identified using records maintained by dairy cooperative societies and local bank branches, including commercial banks, cooperative banks, regional rural banks, Primary Agricultural Cooperative Credit Societies (PACS), and beneficiaries of the Kisan Credit Card (KCC) scheme. Non-borrower households were selected from villages within the same study blocks that operated under similar agro-climatic and dairy production conditions but had not used institutional credit for dairy farming. These farmers were identified through dairy cooperative societies and village-level milk collection centres to ensure that borrowers and non-borrowers belonged to comparable production environments. From each selected block, 25 borrower and 25 non-borrower dairy farmers were selected, resulting in a total sample of 200 respondents comprising 100 borrowers and 100 non-borrowers.
Endogenous switching regression
To identify the determinants of institutional credit access and estimate its impact on farm income, employment generation, and farm mechanization among dairy farmers, the study employed the Endogenous Switching Regression (ESR) model. The ESR approach was considered appropriate because access to institutional credit is a voluntary decision rather than a randomly assigned treatment. Farmers self-select into borrowing based on both observable characteristics (such as age, education, landholding, household income, farming experience, and cooperative membership) and unobservable characteristics, including managerial ability, entrepreneurial orientation, risk preferences, and motivation. These unobserved factors may simultaneously influence both the decision to access institutional credit and farm performance, resulting in endogeneity and sample selection bias (Khan and Kim, 2025).
Conventional estimation techniques, such as Ordinary Least Squares (OLS), assume that participation in institutional credit is exogenous and may therefore produce biased and inconsistent estimates when self-selection exists. Likewise, matching approaches such as Propensity Score Matching (PSM) account only for selection arising from observable characteristics and are unable to control for unobserved heterogeneity. The ESR model overcomes these limitations by jointly estimating the participation decision and separate outcome equations for borrowers and non-borrowers using the Full Information Maximum Likelihood (FIML) estimation procedure. Simultaneous estimation allows the error terms of the selection equation and outcome equations to be correlated, thereby correcting for both observable and unobservable selection bias and yielding consistent estimates of the treatment effects (Kumar et al., 2017; Dhakal, 2022; Malaiarasan et al., 2025).
An additional advantage of the ESR framework is that it permits the estimation of counterfactual outcomes. Specifically, it estimates not only the observed outcomes of borrowers and non-borrowers but also the hypothetical outcomes that borrowers would have experienced had they not accessed institutional credit, and vice versa. These counterfactual estimates enable the computation of the Average Treatment Effect on the Treated (ATT), the Average Treatment Effect on the Untreated (ATU), and heterogeneity effects, thereby providing a more comprehensive assessment of the impact of institutional credit than conventional regression models.
Selection equation
The first component of the ESR model estimates the probability of a dairy farmer accessing institutional credit. The participation decision is represented by an unobserved latent variable, which reflects the expected utility associated with borrowing. A farmer is observed as a borrower when the latent utility exceeds a threshold value. The selection equation is specified as follows:
That is, a dairy farmer is a borrower of institutional credit (Ai = 1), if Y* > 0, where Y* represents the expected benefits of borrower farmers compared to non-borrower farmers.
– is a latent variable which is not observed but can be expressed as a function of the observed characteristics and attributes.
Ai – binary variable that equals 1 for farmers who have taken institutional credit and 0 otherwise
α – vectors of parameters to be estimated (age of the respondent, gender, education level of the household head, farming experience, size of landholding (ha), annual income, household size, number of machineries, membership in dairy cooperatives and distance to bank)
Zi – vector of variables that affect the decision to access credit
ηi – error term with mean 0 and σ2η variance.
Outcome equations
Depending on the farmer's borrowing decision, separate outcome equations were estimated for borrowers and non-borrowers. The ESR framework therefore allows the production process to differ across the two groups rather than imposing a common production function. Separate models were estimated for each of the three outcome variables considered in the study, namely farm income from dairying, farm mechanization, and employment generation.
The general outcome equation is specified as follows:
where:Y – farm income/farm mechanization/employment generation
A – binary variable that equals 1 for farmers who has taken institutional credit and 0 otherwise
X – vector of explanatory variables
β – vector parameter to be estimated
ɛ – error term.
The ESR model estimates two outcome regimes corresponding to borrower and non-borrower households.
Here, Y1i and Y2i are the mentioned parameters of borrowers and non-borrowers for which the impact of credit has to be estimated (i.e. Farm income from dairying / Farm Mechanization/Employment generation). β1 and β2 are parameters to be estimated & ɛ1i and ɛ2i are the error terms. If the estimated covariances between δ and ɛ's (ρ1 and ρ2, respectively) are statistically significant, then the borrowed farmers' characteristics and the outcome variables are correlated. The selection of instrumental variables should directly affect the selection variable but not the outcome variable (Ullah, 2021). The model assumes that the error terms of the selection equation and the two outcome equations follow a trivariate normal distribution. If the estimated correlation coefficients between the selection equation and the outcome equations are statistically significant, the presence of endogenous self-selection is confirmed, thereby justifying the use of the ESR framework. The log likelihood function for the system (1–2) was estimated using full-information maximum likelihood method (FIML) to fit both binary and continuous equation simultaneously and to account for endogeneity. The model can also be fitted using two-stage least squares or maximum likelihood estimation. However, employing these estimation methods to yield consistent standard errors necessitates cumbersome adjustments and these methods are inefficient (Cariappa and Sendhil, 2021).
Treatment effects and counterfactual analysis
The estimated ESR model enables the computation of both observed and counterfactual expected outcomes for borrowers and non-borrowers. Specifically, it estimates the expected outcomes for borrowers under observed conditions, the expected outcomes for non-borrowers under observed conditions, the hypothetical outcomes that borrowers would have experienced had they not accessed institutional credit, and the hypothetical outcomes that non-borrowers would have experienced had they accessed institutional credit.
The endogenous switching regression model enables us to compute four comparable expected outcomes (equation (a) to (d)). The actual expected farm income, level of farm mechanization and employment generation of the dairy farmers that are borrowers (a) and non-borrowers (b); and these same outcome variables in counterfactual scenarios, i.e., outcomes of borrowers if they had not taken institutional credit (d) and that of non-borrowers if they had taken institutional credit (c). The simple difference of (a) and (b) could be misleading and result in wrongful conclusions (Di Falco and Veronesi, 2013).
δ1η and δ2η represent the covariance between ηi, ɛ1i, and ɛ2i. We also calculate the average effect of the treatment on the treated (borrowers) (ATT) as the difference between (a) and (c), and the average effect of treatment on untreated (non-borrowers) (ATU) as difference between (d) and (b).
The heterogeneity effects for the group of farmers to be borrowers and those not to be borrowers were calculated as the differences between (a) and (d), and (c) and (b), respectively, which are presented in Table 1. The result of this computation explains the unobserved differences, such as skills, between borrowers and non-borrowers that result in outcome differences regardless of credit taken out or not. Transitional heterogeneity i.e., whether the effect of institutional credit is larger or smaller for the borrowers or for non-borrowers in the counterfactual case are estimated by finding the difference between the equations (e) and (f).
Table 1.
Treatment and heterogeneity effect: decision stage
| borrowing | not borrowing | ||
|---|---|---|---|
| Borrower | (a) E(Y1i | Ai = 1) | (c) E(Y2i | Ai = 1) | TT |
| Non-borrower | (d) E(Y1i | Ai = 0) | (b) E(Y2i | Ai = 0) | TU |
| Base heterogeneity | (e) BH1 | (f) BH2 | TH |
[i] (a) and (b) – represents observed expected outcome indicators; (c) and (d) – represents counterfactual expected outcome indicators; Ai = 1 if dairy farmers have taken out institutional credit; Ai = 0 if farmers have not taken out institutional credit; Y1i – outcome indicators if dairy farmers have taken out institutional credit; Y2i – outcome indicators if farmers have not taken out institutional credit; TT – the effect of the treatment on the treated (borrowers); TU – the effect of the treatment on the untreated (non-borrowers); BHi – the effect of base heterogeneity for dairy farmers who have taken out institutional credit (i = 1), and farmers who have not taken it out (i = 2); TH = TT – TU (that is, transitional heterogeneity).
Table 2 presents the descriptive characteristics of borrower and non-borrower dairy households included in the study. Borrower households were relatively older and had greater farming experience than non-borrowers. They also exhibited higher levels of farm mechanization, greater participation in dairy cooperatives, wider livestock insurance coverage, higher milk productivity, and substantially higher farm income per animal. In contrast, non-borrower households reported relatively higher annual household income, marginally larger landholdings, and were located farther from the nearest bank branch, suggesting that physical access to formal financial institutions may influence credit participation. Overall, the two groups operated under similar agro-climatic conditions but differed in several socio-economic and institutional characteristics, highlighting the potential for self-selection in credit participation decisions and justifying the use of the Endogenous Switching Regression (ESR) framework for impact estimation.
Table 2.
Descriptive statistics of borrower and non-borrower dairy households
| Variable | Borrowers (Mean) | SD | Non-borrowers (Mean) | SD |
|---|---|---|---|---|
| 1 | 2 | 3 | 4 | 5 |
| Age (years) | 59.89 | 5.05 | 55.38 | 4.99 |
| Education level | 3.67 | 1.30 | 3.63 | 1.47 |
| Farming experience (years) | 13.21 | 4.01 | 10.48 | 3.69 |
| Annual income (₹ lakh) | 3.23 | 0.79 | 4.48 | 1.49 |
| Household size | 4.82 | 0.87 | 4.85 | 1.02 |
| Total landholding (ha) | 1.13 | 0.65 | 1.27 | 0.58 |
| Distance to nearest bank (km) | 5.89 | 1.83 | 7.64 | 1.90 |
| Membership in dairy cooperatives | 0.78 | 0.33 | 0.69 | 0.37 |
| Number of machineries used | 4.13 | 0.78 | 2.89 | 1.14 |
| Area under fodder cultivation (ha) | 0.38 | 0.22 | 0.42 | 0.19 |
| Area under dairy farm (ha) | 0.57 | 0.32 | 0.64 | 0.29 |
| Livestock insurance coverage | 0.72 | 0.45 | 0.44 | 0.50 |
| Farm income (₹/animal/day) | 103.02 | 10.17 | 78.56 | 10.17 |
| Milk productivity (L/animal/day) | 8.68 | 1.30 | 7.83 | 1.30 |
RESULTS AND DISCUSSION
Impact of credit on the farm income of farmers
The determinants of access to institutional credit and its effects on farm income from dairying per animal of dairy farmers were shown in Table 3. The estimated results in columns 1 and 2 of the table reflect the impact of access to institutional credit on farm income from dairying for both borrowers and non-borrowers. Column 3 reports the results of selection equation that represents the determinants of access to institutional credit. The net farm income per animal per day was used as the outcome variable for the analysis. It was found that the explanatory variables regarding the farmer's education level, annual income, number of machines, farming experience, and size of landholding (ha) had a significant effect on the farm income of the dairy farmers. The coefficient of the variable education level of the household head was found to be significant and positive for both borrowers and non-borrowers. The positive influence of education suggests that better-educated farmers are more capable of adopting improved dairy management practices, maintaining financial records, accessing extension services, and efficiently utilizing institutional credit. These capabilities improve resource allocation and enhance the productivity of borrowed capital, resulting in higher farm incomes among borrower households (Dalei et al., 2025). Together with this, the coefficient of variables regarding annual income, number of machineries used on the farm, and farming experience were found to be both significant and to affect the farm income of borrower dairy farmers positively. Annual income, number of machines used for farming, household size and size of landholding (ha) were found to significantly affect the farm income of non-borrower dairy farmers. The coefficient of the household head's age was positive and statistically significant, indicating that older farmers were more likely to access institutional credit. In contrast, farming experience exhibited a negative and significant coefficient, suggesting that more experienced dairy farmers were less likely to rely on institutional credit, possibly because of accumulated savings, stronger internal resource mobilization, or a greater preference for self-financing. Similarly, the positive and significant coefficient for number of machines used in farming and net returns from agriculture implies that mechanization and increased farm profitability from agriculture contribute to enhanced creditworthiness, thereby facilitating greater access to institutional credit for dairy farming activities. The coefficient of the variable ‘annual income of the farm household’ was negative and significant, indicating that households with a greater annual income are less likely to access institutional credit for dairy farming. This finding was in accordance with the studies by Kimkong et al. (2023), and by Vimefall and Levin (2023), which showed that higher income households may have more diversified income sources or assets that provide financial stability, making them less dependent on loans for dairy farming. Although institutional credit significantly improves farm performance, not all dairy farmers choose to borrow. This may reflect behavioural and institutional constraints that are not fully captured by the econometric model. Farmers may avoid institutional borrowing because of risk aversion, concerns regarding repayment obligations under uncertain milk prices and production conditions, procedural complexities associated with loan applications, collateral requirements, limited financial literacy, or previous experiences with formal financial institutions. Similar barriers have been reported in studies on agricultural credit adoption in developing countries. These findings suggest that improving credit access requires not only expanding the supply of institutional finance but also reducing transaction costs and strengthening financial inclusion among smallholder dairy farmers (Birthal et al., 2017; Mwonge et al., 2025).
Table 3.
Impact of credit on farm income of farmers
| Variables | Borrower = 1 | Non-borrower = 0 | Borrower =1 Non-borrower = 0 |
|---|---|---|---|
| Farm income per day (per animal) | Selection | ||
| Dependent variable | |||
| Age of respondent | −0.098 (0.933) | 0.175 (0.328) | 0.107*** (0.033) |
| Education level | 43.807*** (2.484) | 20.853*** (2.195) | −0.089 (0.153) |
| Annual income (₹) | 2.870*** (0.985) | 0.498* (0.286) | −0.281** (0.112) |
| Number of machineries | 7.418** (1.594) | 2.546** (3.347) | 0.291*** (0.100) |
| Farming experience | 7.557*** (0.947) | −0.221 (0.411) | −0.067** (0.043) |
| Household size | 3.067 (2.844) | 2.833* (1.543) | 0.017 (0.142) |
| Land holding (ha) | 0.001 (0.001) | 0.001*** (0.001) | 0.001** (0.001) |
| Instrumental variables | |||
| Distance to nearest bank branch | −4.539*** (1.193) | ||
| Membership in dairy cooperatives | 1.637*** (0.446) | ||
| Sigma 1 | 20.408*** (1.694) | ||
| Sigma 2 | 12.405*** (1.089) | ||
| Rho 1 | 0.254 (0.415) | ||
| Rho 2 | 0.898*** (0.098) | ||
| Constant | 259.396*** (49.965) | 369.405*** (23.294) | −1.727 (1.681) |
| Observations | 100 | 100 | 200 |
Distance to nearest bank branch and membership in dairy cooperatives were used as the instrumental variables in the selection equation. These variables were based on the assumptions that distance to a bank may influence a farmer's access to credit due to transportation costs and convenience. However, it might not directly affect farm income for non-borrowers who do not rely on institutional credit. Moreover, dairy farmers who are members of cooperatives may have easier access to credit through group lending schemes or cooperative initiatives. However, membership of cooperatives may not result in significant direct effects on farm incomes for dairy farmers. The coefficients of instrumental variables were found to be significant at different levels of significance. The coefficient ‘distance to the nearest bank branch’ was negative and significant, indicating that farmers residing further away from the nearest bank branch exhibit a reduced likelihood of accessing institutional credit for dairy farming. The coefficient of the variable ‘membership in dairy cooperatives’ was positive and significant, indicating that households with membership of dairy cooperatives are more likely to access institutional credit for dairy farming. This is in line with the research finding that cooperative participation can empower farmers, strengthen their financial capabilities, and facilitate sustainable growth in the dairy farming sector (Wossen et al., 2017; Yu et al., 2023). The significance of variables such as distance to the nearest bank branch and membership of dairy cooperatives indicates that institutional accessibility continues to influence farmers' borrowing decisions. Farmers located further from banking facilities incur higher transaction costs and face greater procedural difficulties in obtaining formal credit. Conversely, cooperative membership facilitates access to information, documentation, and institutional networks that improve credit availability. These findings suggest that strengthening last-mile financial services and cooperative-based credit delivery mechanisms may further enhance institutional credit outreach (Gaku et al., 2025; Bhat et al., 2026).
The estimates for the average treatment effects on the treated and untreated (ATT & ATU), which present the overall effects access to institutional credit on farm income, are visible in Table 4. This presents the estimates of the expected farm income given the actual and counterfactual scenario. Both borrowers and non-borrowers of institutional credit face two possible decisions: to access or not to access. The diagonal cells (a and b) present the expected farm income from dairy in the actual scenarios; the off-diagonal cells (d and c) show the expected farm income from dairy in the counterfactual scenarios. These estimates based on the ESR model account for selection bias resulting from both observable and unobservable characteristics (Courtois and Subervie, 2015). The expected average farm income of farmers from dairy, who had accessed institutional credit (a), was ₹ 103.02 per day per animal and that of non-borrowers (b) was ₹ 78.56. In the counterfactual scenario, the expected average farm income from borrower dairy farmers, assuming if they had not accessed the credit (d), is ₹ 85.42, and that of non-borrowers assuming they accessed it (c) is ₹ 94.30. The average treatment effect on the treated (ATT), which is the difference between cells (a) and (c), is 8.72. This shows that by adopting institutional credit, dairy farmers significantly increase the farm income by 9.24 per cent compared with their counterparts. The treatment effect on the untreated (ATU) obtained by the difference between cells (d) and (b) is 6.86. This shows that when institutional credit is taken out by dairy farmers who had not taken out credit previously significantly increases their farm income. The estimated ATT indicates that institutional credit enabled borrower households to undertake productive investments that translated into higher dairy income. Such credit likely facilitated the purchase of better dairy animals, quality feed, veterinary services, and labour-saving equipment, thereby increasing milk productivity and farm profitability. A positive ATU further suggests that non-borrower households could also experience marked income gains if barriers to institutional credit were reduced. The difference between the ATT and the ATU is the transitional heterogeneity effect (TH), which is 1.86. However, the transitional heterogeneity effect of the farm income of a dairy farmer is positive, meaning that the effect would be greater for borrower farmers than for non-borrower farmers. The statistically significant correlation coefficients confirm the presence of endogenous self-selection, thereby justifying the use of the Endogenous Switching Regression model over conventional regression approaches. These findings indicate that access to institutional credit for dairy farming enhances the farm income of dairy farmers.
Impact of credit on the employment generation in dairy farms
The determinants of access to institutional credit and its effects on employment generation among dairy farmers are shown in Table 5. Columns 1 and 2 reflect the parameters of employment generation in terms of man-days per animal for borrower and non-borrower dairy farmers, respectively. Column 3 shows the selection equation that represents the determinants of access to institutional credit. The number of man-days per animal serves as the outcome variable in the analysis. It was found that the explanatory variables age of respondent, household size, farming experience, number of machineries used, annual income, fodder cultivation area (ha), and dairy farm area (ha) had a significant effect on employment generation in dairy farm households. Household size and farming experience were likewise found to have a positive and significant effect on the number of man-days per animal for both borrower and non-borrower farmers. This can be attributed to farm households with greater farming experience and a larger number of members being able to allocate more family labour to farm operations. The coefficient of the variable number of machineries used for farming was found to be negative and significant at the 10- and 1-percent level for borrower and non-borrower dairy farmers, respectively. This indicates that for a unit increase in the number of machines used in farm, labour in terms of man-days per animal was found to be decreasing. Additionally, the coefficients ‘fodder cultivation area’ and ‘dairy farm area’ were positive and significant for borrower and non-borrower farmers, respectively. This indicates that as dairy farm size increases, demand for more labour per animal also increases. The significant and negative coefficient of farming experience in the selection equation suggests that experienced farmers are less likely to rely on institutional credit. Conversely, the number of machines used on a farm, annual income, fodder cultivation area (ha), and dairy farm area (ha) were found to be positive and significant. The positive coefficient indicated that dairy farmers who utilize more machinery in their farming operations are more likely to access institutional credit. Farmers with a larger area under fodder cultivation and dairy farms were also more likely to access institutional credit. Explanatory variables such as adoption of livestock insurance and membership of dairy cooperatives were used as instrumental variables in the selection equation. The coefficients of instrumental variables were also significant. The coefficient of membership of dairy cooperatives was positive and significant, indicating that households which are members of dairy cooperatives are more likely to access institutional credit for dairy farming.
Table 5.
Impact of credit on employment generation in dairy farms
| Variables | Borrower = 1 | Non-borrower = 0 | Borrower = 1 Non-borrower = 0 |
|---|---|---|---|
| Mandays per animal | Selection | ||
| Dependent variable = Mandays per animal | |||
| Age of respondent | 0.002*** (0.001) | 0.001*** (0.001) | −0.001 (0.001) |
| Household size | 0.002*** (0.002) | 0.007*** (0.005) | −0.156 (0.208) |
| Farming experience | 0.021*** (0.001) | 0.005*** (0.002) | −0.223* (0.082) |
| No of machineries | −0.001* (0.001) | −0.014*** (0.004) | 0.631*** (0.166) |
| Annual income (₹) | 0.001 (0.001) | 0.008*** (0.003) | 0.002*** (0.001) |
| Area under fodder cultivation (ha) | 0.002** (0.005) | 0.001 (0.000) | 0.549*** (0.154) |
| Area under dairy farm (ha) | −0.007 (0.048) | 0.058** (0.048) | 15.673*** (2.803) |
| Instrumental variables | |||
| Livestock insurance coverage (0/1) | 0.322*** (0.047) | ||
| Membership in dairy cooperatives (0/1) | 0.185** (0.152) | ||
| Sigma 1 | 0.015*** (0.001) | ||
| Sigma 2 | 0.035*** (0.002) | ||
| Rho 1 | −0.631** (0.206) | ||
| Rho 2 | −0.836 (0.234) | ||
| Constant | −0.132*** (0.022) | −0.051 (0.035) | −0.737 (1.517) |
| Observations | 100 | 100 | 200 |
Table 6 presents estimates of expected employment generation given the actual and counterfactual scenarios. The diagonal cells (a and b) present expected employment generation in dairy farms in actual scenarios in Tamil Nadu. The off-diagonal cells (d and c) show the expected employment generation in the counterfactual scenarios. The expected value of employment generation on dairy farms was greater for borrower dairy farmers than non-borrowers. This simple comparison, however, could be misleading in attributing the effect of access to credit on outcome variable of borrowing and non-borrowing households (Kumar et al., 2020). The expected average employment generated for borrower and non-borrower farmers were 0.18 and 0.12 man-days per animal, respectively. In the counterfactual scenario, employment generation in dairy farms of borrower farmers, assuming they had not accessed the credit (d), was 0.17, and that of non-borrowers, assuming they accessed credit (c), was 0.12 man-days per animal. The average treatment effect on the treated (ATT), which is the difference between cells (a) and (c), was 0.05. This reveals that adoption of institutional credit by dairy farmers increases the additional labour generation on dairy farms by 37.87 percent compared with non-borrower households. It also states that in the counterfactual case, farm households that borrowed money would have an employment generation per farm of approximately 0.050 man-days per animal lower than if they had not borrowed money. The treatment effect on the untreated (ATU) obtained by the difference between cells (d) and (b) is 0.04. The positive mean difference of (d) and (b) shows that on average, the control group farmers would have generated an additional 0.04 man-days per animal if they were in the treatment group. The positive contribution of access to credit to employment generation in dairy farms is reinforced by the transition heterogeneity (TH). This is calculated as the difference between the ATT and the ATU, which was estimated at 0.007. However, the transitional heterogeneity effect for employment generation was positive, meaning that the effect would be greater for the treatment group than for the control group. The positive value of transition heterogeneity also suggests that borrowers and non-borrowers are significantly different. Moreover, it confirms the presence of some unobserved heterogeneities that leave borrowers better off than non-borrowers. The findings suggest that access to institutional credit contributes to generating employment, as evidenced by a slight increase in man-days per animal on dairy farms in Tamil Nadu. This aligns with the results obtained by Ghimire et al. (2017), who found that securing bank financing for agricultural endeavours boosted employment opportunities among the beneficiaries in the post-loan period. However, the observed increase in employment generation does not necessarily indicate labour inefficiency. Rather, institutional credit appears to encourage expansion of dairy enterprises through increased herd size, improved feeding practices, and greater investment in dairy infrastructure, all of which increase labour demand for routine management activities. Thus, credit contributes not only to farm productivity but also to rural employment generation (Kumari et al., 2025).
Impact of credit on farm mechanization on dairy farms
The determinants of access to institutional credit and its implications for mechanization on dairy farms are shown in Table 7. Columns 1 and 2 show the parameters of farm mechanization by borrower and non-borrower dairy farmers, respectively. Column 3 reports the results of the selection equation that represents the determinants of access to credit. The number of machines used for dairy farming was treated as the outcome variable for the regression model. Slurry pumps, water pumps, milking machines, chaff cutter machines, weighing balances etc. were the machinery commonly used by most respondents. It was found that the explanatory variables age of the farmer, annual income, total landholding (Ha), farming experience, and household size had significant effects on dairy farm mechanizations. The average annual income and farming experience of the respondent farmers was found to positively affect technology adoption on farms in terms of the number of machines used by both borrower and non-borrower dairy farmers at different significance levels. This indicates that the farmers with a higher annual income and greater experience tended to invest more in purchasing machinery for farm use. Institutional credit reduces liquidity constraints associated with purchasing capital-intensive farm equipment such as milking machines, chaff cutters, slurry pumps, and water pumps. These investments improve labour productivity, operational efficiency, and the timeliness of farm operations, thereby bolstering the productive capacity of dairy farms. The total landholding was positive and significant at the 5-percent level for both borrower and non-borrower famers. This was in line with the findings of Tiamiyu et al. (2009). Distance to nearest bank branch and membership of farmer cooperatives were used as the instrumental variable in the selection equation. The coefficients of instrumental variables were significant. The negative and significant coefficient of distance to the nearest bank branch suggested that there was an inverse relationship between the distance of farmers to the nearest bank branch and their likelihood of accessing institutional credit for dairy farming. The coefficient of membership in dairy cooperatives was positive and significant, indicating that households that are members of dairy cooperatives were more likely to access formal credit sources for their dairy farming activities compared to non-members.
Table 7.
Impact of credit on farm mechanization of dairy farmers in Tamil Nadu using endogenous switching regression
| Variables | Borrower = 1 | Non-borrower = 0 | Borrower = 1 Non-borrower =0 |
|---|---|---|---|
| No of machineries | Selection | ||
| Dependent variable = Number of machineries | |||
| Age | −0.059* (0.034) | 0.005 (0.006) | 0.222 (0.069) |
| Gender | −0.040 (0.123) | −0.123 (0.093) | −0.637 (0.580) |
| Education level | 0.021 (0.095) | 0.025 (0.027) | 0.820* (0.279) |
| Annual Income (₹) | 0.891*** (0.158) | 1.181** (0.049) | 2.591*** (0.535) |
| Total Landholding (ha) | 0.239** (0.117) | 0.095** (0.038) | −1.694** (0.398) |
| Household size | 0.004** (0.013) | −0.017 (0.008) | 0.269* (0.126) |
| Farming experience | −3.236 (1.430) | 0.916* (0.509) | 14.989*** (3.653) |
| Instrumental variables | |||
| Membership in farmer cooperatives | 0.195** (0.056) | ||
| Distance to nearest bank branch | −0.004*** (0.004) | ||
| Sigma 1 | 0.424*** (0.030) | ||
| Sigma 2 | 0.278*** (0.019) | ||
| Rho 1 | −0.031 (0.486) | ||
| Rho 2 | 0.724 (0.471) | ||
| Constant | 3.132 (2.515) | −1.669** (0.417) | −23.490** (6.503) |
| Observations | 100 | 100 | 200 |
Table 8 presents estimates of the expected technology adoption in terms of the number machines used by dairy farmers in the actual and counterfactual scenarios. Both borrowers and non-borrowers of institutional credit face two possible decisions: to access or not to access. The diagonal cells (a and b) present the expected technology adoption in dairy farms in terms of the number of machines used on the farm in the actual scenario. The off-diagonal cells (d and c) show the expected technology adoption in dairy farm in the counterfactual scenario. The expected average level of technology adoption in terms of the number of machines used for farming by borrowers (a) was 4 per farm, while that of non-borrowers (b) was 2–3 per farm. In the counterfactual scenario, the expected average level of technology adoption in terms of the number of machines used by borrower farmers, assuming they had not accessed the credit (d), was 3 per farm, and that of non-borrowers, assuming they accessed credit (c), was also 3 per farm. The average treatment effects on the treated (ATT) and on the untreated (ATU) were 0.83 and 0.44, respectively. This reveals that the adoption of institutional credit by dairy farmers increases farm mechanization by 25.15% compared to non-borrowers. Improved access to institutional credit enables farmers to overcome liquidity constraints associated with purchasing capital-intensive equipment such as milking machines, chaff cutters, slurry pumps, and water pumps. These investments improve operational efficiency, reduce drudgery, and support timely farm operations, thereby enhancing the overall productive capacity of dairy enterprises. The difference between the ATT and the ATU yields a transitional heterogeneity effect (TH) of 0.39. However, the transitional heterogeneity effect of farm mechanization of dairy farmers is positive, indicating that the effect would be greater for the treatment group than for the control group. The results suggest that access to institutional credit enhances the mechanization among dairy farmers. Similar research conducted by Rani et al. (2019) and Mukherjee (2020) also confirmed that improving access to banks increases the adoption of high-yielding variety (HYV) technology.
Table 8.
Treatment and heterogeneity effect
| Variables | Decision | Treatment effect | Percent change | |
|---|---|---|---|---|
| Borrowing | Not borrowing | |||
| Borrower | 4.13 (a) | 3.30 (c) | ATT = 0.83 | 25.15 |
| Non-Borrower | 3.33 (d) | 2.89 (b) | ATU = 0.44 | 15.22 |
| Base heterogeneity | BH = 0.80 | BH = 0.41 | TH = 0.39 | |
These findings also have important policy implications. Although India has introduced several institutional mechanisms to strengthen agricultural credit, including the Kisan Credit Card scheme, interest subvention programmes, cooperative banking networks, and NABARD refinance support, barriers related to awareness, documentation, collateral requirements, and transaction costs continue to limit credit access among dairy farmers. Therefore, policy efforts should focus not only on expanding the supply of institutional credit but also on simplifying loan procedures, strengthening cooperative-based financial intermediation, promoting digital credit delivery systems, improving financial literacy, and integrating institutional credit with livestock insurance and dairy extension services.
CONCLUSIONS
This study examined the impact of institutional credit on farm income, employment generation, and farm mechanization among dairy farmers in Tamil Nadu using the Endogenous Switching Regression (ESR) framework, which accounts for both observable and unobservable self-selection bias. The findings reveal that institutional credit has a significant and positive impact on all three outcome variables. Dairy farmers who accessed institutional credit experienced higher farm income, greater employment generation, and increased farm mechanization compared to non-borrowers. Furthermore, the estimated treatment effects indicate that non-borrower households could also achieve substantial gains in these outcomes if barriers to institutional credit were reduced. These findings demonstrate that institutional credit serves as an important catalyst for productive investment, technology adoption, and the overall economic performance of dairy enterprises. The study also identifies several socio-economic and institutional factors influencing farmers' access to institutional credit, including age, education, annual income, farming experience, household characteristics, landholding, and membership of dairy cooperatives. The significant roles of institutional factors such as proximity to bank branches and cooperative membership highlight the importance of improving the accessibility and outreach of formal financial institutions in rural areas.
This study's findings have important implications for agricultural credit policy. While expanding the availability of institutional credit remains essential, improving access alone may not be sufficient. Policy interventions should focus on simplifying loan application procedures, reducing documentation requirements, strengthening the outreach of commercial banks, cooperative banks, Regional Rural Banks and Primary Agricultural Cooperative Credit Societies, expanding the coverage of the Kisan Credit Card (KCC) scheme for dairy farmers, promoting digital credit delivery systems, and enhancing farmers' financial literacy. Greater integration of institutional credit with livestock insurance, dairy extension services, veterinary healthcare, and cooperative institutions would further improve the productive utilization of credit and enhance its effectiveness in strengthening dairy farm performance. At the same time, institutional credit should be viewed as one component of a broader dairy development strategy. Investments in quality breeding services, fodder development, veterinary infrastructure, milk marketing systems, and extension support are equally important for ensuring that farmers are able to generate adequate returns from productive investments and sustain timely loan repayments. A comprehensive approach combining financial inclusion with technological, institutional and market support would contribute substantially to improving productivity, competitiveness, and sustainability in the dairy sector.
This study has certain limitations that should be acknowledged. Firstly, the analysis is based on cross-sectional data collected from two districts of Tamil Nadu, which limits the generalizability of the findings to other regions with different institutional and production environments. Secondly, although the ESR framework accounts for both observable and unobservable selection bias, behavioural factors such as farmers' risk preferences, entrepreneurial orientation, financial literacy, and attitudes towards institutional borrowing were not explicitly captured. Thirdly, the study focused on farm income, employment generation, and farm mechanization as indicators of farm performance, while other dimensions such as milk productivity, profitability, and farm resilience were beyond its scope. Future research may employ longitudinal or panel data across diverse geographical regions and incorporate behavioural and institutional variables to better understand the long-term impacts of institutional credit on dairy farm development.
Overall, the study provides robust empirical evidence that institutional credit significantly enhances the productive capacity and economic performance of dairy farms. Strengthening inclusive, accessible and efficient rural credit delivery systems, alongside complementary investments in extension, infrastructure, and market development, will play a critical role in fostering sustainable growth and improving the competitiveness of the dairy sector in India.
ACKNOWLEDGEMENT
The lead author sincerely acknowledges the guidance of the Head of the Department and kind support of other faculties along with research scholars.