Does contract farming participation promote household’s food security for smallholders? Empirical evidence from Indonesia

: Contract farming has been extensively adopted as a strategy to overcome limitations in the market and enhance the well-being of farmers. Nevertheless, the extent to which it affects food security has not been sufficiently examined and is uncertain. Moreover, there is a lack of empirical evidence regarding the impact of contract farming on food security in Indonesia. To fill the existing research gap, this study employs propensity score matching (PSM) to mitigate selection bias in examining the impact of contract farming on the food security of smallholder farm households in Indonesia. It utilizes the 2021 Indonesia Agricultural Integrated Survey (SITASI) data, designed to monitor the indicators of sustainable development goals (SDGs) in the agriculture sector. The food insecurity experience scale (FIES) is used to measure food security. Our research shows that contract farming can potentially improve the food security of smallholder farm households in Indonesia. However, the overall impact can be considered minor. Contract farming has the potential to enhance food security, but it may not be adequate as a standalone solution. A comprehensive strategy, complemented by related policies such as innovative farming practices, technology adoption, and income-generation measures, is essential. Furthermore, our investigation revealed that this beneficial effect is particularly prominent among farmers residing in rural areas, female farmers, and farmers who do not own land or livestock. It indicates that contract farming can be a feasible tool for poverty alleviation, rural development, and woman empowerment. This study also found that factors such as location, market access, credit availability, gender, education, and exposure to agricultural training influenced contract farming participation.


Introduction
Literature has widely endorsed contract farming as a viable approach to help smallholder farmers overcome their market constraints (Mishra et al., 2018).Participating in contract farming presents small-scale farmers with the opportunity to enhance their earnings and expand their production capacities.The potential for contract farming is being credited for facilitating market access (Meemken & Bellemare, 2020;Soullier & Moustier, 2018), enhancing inputs and credit accessibility (Debela et al., 2022;Gatto et al., 2017), and facilitating the adoption of technological advancements (Ganewo et al., 2022;Ragasa et al., 2018).For this reason, policymakers in numerous developing countries have utilized contract farming to stimulate agricultural development and alleviate rural poverty (Bellemare & Novak, 2017).
Studies on the impacts of contract farming have predominantly centred on its impact on income (Hoang, 2021;Mwambi et al., 2016;Khan et al., 2019).The prevailing consensus among these studies is that contract farming positively impacts farmers' income.Conversely, the scholarly discourse surrounding its impact on food security has been comparatively limited.It is an important issue since its impact on food security can be ambiguous.According to Soullier & Moustier (2018), contract farming improves the food security of rice farmers in Senegal through increased income.Suppose contract farming is capable of increasing farmers' income and considering that those with higher income can acquire a greater amount of better quality food.In that case, it is reasonable to expect that contract farming will improve food security.
On the contrary, it is possible that contract farming can adversely impact food security.Andriamparany et al. (2021) and Olounlade et al. (2020) discovered that contract farming adversely affects the food security of vanilla farmers in Madagascar and rice farmers in Benin.According to Soullier & Moustier (2018), contract farming may lead to competition between own consumption and sales.Participating households may be forced to allocate a portion of their production, formerly allocated for their consumption, to meet their agreed-upon quota.In some circumstances, contracted farmers must grow a crop or raise livestock not commonly consumed within the household (Olounlade et al., 2020).Contract farming may also redirect the farmer's attention towards the contracted commodities, decreasing the time allocated to subsistence farming, off-farm jobs, and domestic duties such as food preparation (Andriamparany et al., 2021).As a result, these households need to procure food from the market for their consumption.If the market is not easily accessible or the market price is high, it will result in lower food security status.All prior empirical studies on the impact of contract farming on food security focus on a particular contract scheme (Andriamparany et al., 2021;Soullier & Moustier, 2018), specific commodity (Debela et al., 2022;Ganewo et al., 2022;Olounlade et al., 2020), or for limited regions such as several districts (Bellemare & Novak, 2017).Hence, the conclusion is not generalizable to other cases because it is specific to the common scheme and commodities in that region.As each nation may have its contract farming schemes and commodities vary, conducting country-specific research is necessary.
The Minister of Agriculture's Decision Number 484/KPTS/RC.020/M/8/2021has included contract farming as a component of the National Economy Recovery Programs or Pemulihan Ekonomi Nasional (PEN) to strengthen food security and improve the welfare of farmers.Our data indicates, however, that the participation rate in Indonesia remains extremely low at 1.64%.In comparison, the participation rates in contract farming in the United States and China were 9% (Whitt, 2022) in 2020 and 24% in 2017 (Lixia et al., 2021).
On the other hand, Indonesia is far from attaining the SDGs' objective of eliminating hunger or food insecurity.According to Statistics Indonesia (Badan Pusat Statistik), in 2022, approximately 4.85% of the Indonesian population, equivalent to over 13.3 million individuals, still experience food insecurity.Food security-wise, Indonesia lags significantly behind the United States and China.According to the Global Food Security Index (GFSI), Indonesia is ranked 63rd, while China and the United States are ranked 26th and 13th, respectively.A higher rate of contract farming participation, while not the only determinant, could contribute to better food security in both countries, given its beneficial impact on food security.This disparity highlights the necessity for Indonesia to reassess and potentially promote the adoption of contract farming to strengthen food security.Regrettably, there is a lack of studies examining the relationship between contract farming and food security in Indonesia.To the best of our knowledge, there is only one study conducted by Milinia et al. (2023) that specifically addresses this topic.However, this study narrowly focuses on coffee production and is limited to two districts in Java.The broader context remains unexplored.
In order to address this research gap, this study attempts to evaluate the impact of contract farming on the food security of smallholder farm households in Indonesia.In contrast to prior research focusing on a limited number of commodities, this study covers all agricultural commodities (including crops and livestock) cultivated or reared in Indonesia.The analysis utilizes data from the SITASI 2021 dataset, which covers all 34 provinces in Indonesia.Given the non-random nature of contract farming participation in our non-experimental/observational data, it is acknowledged that numerous factors could confound the relationships between contract farming and food security, resulting in selection bias.Propensity score matching (PSM) is utilized to address this concern.
This study makes several significant contributions to the literature.First, it provides empirical evidence on the impact of contract farming on the food security of farmers in Indonesia, addressing a topic that has been insufficiently investigated.Second, as mentioned, all previous studies were conducted for particular contract schemes, certain commodities, or limited regions such as several districts.In contrast, this study contributes novel evidence employing nationwide data with broad coverage of commodities.This approach allows for generalization beyond specific contracts or commodities.Third, the study investigates an unexplored area: how the impact varies by village type (urban or rural) and asset ownership (land or livestock).The findings offer crucial insights for refining government policy targeting.
Our findings reveal that participating in contract farming can potentially improve food security.Contract farming reduces the probability of participating households experiencing food insecurity by 0.92% to 1.17%.However, this impact can be considered minor.Our results also show that the positive impact is more pronounced among rural farmers, female farmers, and farmers who do not own land or livestock.Regarding factors influencing contract farming participation, the study highlights that location (urban/rural), market and credit access, gender, education, membership in farmer associations, household size, and exposure to agricultural training play significant roles.

Data
This study utilizes data from the Indonesia Agricultural Integrated Survey (SITASI) which conducted in 2021 by Statistics Indonesia.The data collection covered the period of one year, from September 2020 to September 2021.SITASI is the only survey that collected information related to contract farming participation and the food security conditions of farm households in Indonesia.It was first conducted in 2021, and no subsequent surveys have been conducted since then.Unlike previous studies limited to limited regions such as districts, this study uses data from all 34 provinces in Indonesia to obtain a broader national-level analysis.Therefore, the findings from this study can be considered representative of farm households across Indonesia.Although SITASI 2021 also collected data for the forestry and fishery sectors, this study only uses data for the agriculture sector.The agriculture sector consists of crops and livestock production.The analysis only accounts for farms that conducted crop or livestock production activities during the survey period.
There are some missing items in the food security data.In order to fill in the missing items on the FIES questions, an imputation procedure was implemented using the IMPUTERASCH command in STATA.However, due to the large number of covariates and the considerable effort required to perform imputation for each of them (while also raising doubts about the validity of the imputation results), observations missing items on their covariates are excluded from this study.A total of 3.76% of the observations are excluded from the analysis because of missing data.The final dataset comprised 230,189 farm households.

Contract Farming Definition
Contract farming is an agreement between farmers and purchasers, typically processing or marketing firms.This agreement outlines the terms and conditions for producing, purchasing, and selling agricultural goods.The agreement is made in advance and typically includes predetermined prices (Bellemare & Lim, 2018;Eaton & Shepherd, 2001).The foundation of such a contract is the farmer's commitment to produce agreed-upon quantities and quality and the buyer's commitment to purchase the products.Frequently, the contract specifies the delivery date, the amount and standard of the product that purchasers demand, and the amount of money to be paid to the farmer.Occasionally, additional information, such as the production method or whether the buyer would provide inputs like seeds, fertilizer, and technical assistance, may be included in the contract.
Two distinct types of contract farming exist, namely marketing and production contracts (Bellemare & Lim, 2018).In a production contract, buyers are the ones who make decisions on production and supply essential inputs like seeds and fertilizer, as well as technology, technical assistance, and loans.On the other hand, marketing contracts give farmers control over production, while buyers have authority over pricing and quantity requirements.Marketing contracts do not provide provisions for input and other forms of support.
The definition of contract farming participation in this study is operationalized as follows.The SITASI questionnaire asked whether the farm or agricultural holdings had a production or marketing contract.A household might own multiple farms.A household is coded as participating in contract farming if it possesses a minimum of one farm or holding with a production, marketing, or both contract arrangement.

Food Security Measurement
Food security is attained when individuals have consistent access to an adequate, nourishing, and safe food supply that fulfils their needs and enables them to maintain good health (World Food Summit, 1996).This study uses the Food Insecurity Experience Scale (FIES) to measure food security.It is a metric that assesses food security using experiential data.The FIES consider not only the dietary quality and quantity of food but also the psychological elements associated with worry or uncertainty regarding the ability to access enough food.It is a feature that is absent in other measures.There are three levels of food insecurity resulting from the FIES: food security, moderate food insecurity, and severe food insecurity.In this study, "food insecurity" refers to moderate and severe food insecurity, whereas "food security" has the exact definition of FIES.All the information about FIES in the remainder of this section is from the Voice of Hungry (VoH) 2016 Technical Report (FAO, 2016).
The methodology of FIES depends upon the individual's perceptions of their encounters with limited food accessibility, as expressed through their responses to the eight questions of FIES.Each question relates to a different experience and corresponds to the different severity of food insecurity.The questions inquire whether, in the last 12 months, the respondent experienced worries about not having enough food, inability to access healthy and nutritious food, limited food options, skipping meals, eating less than desired, depletion of food supplies in households, experiencing hunger without eating, and even going without food for an entire day.
The Rasch model is used to construct FIES based on the responses to the questions.This model uses a logistic function to estimate the likelihood of a responder reporting a specific episode of food insecurity based on the distance between the responder's condition ( ℎ ) and an item's position (  ) on the severity scale: where  ℎ, is the respondent ℎ response to the item , with the value of 1 if "yes" and 0 if "no".The model estimation is conducted by employing conditional maximum likelihood (CML) using FIES estimation software built by VoH.The software can be accessed through the VoH website (https://www.fao.org/in-action/voices-of-the-hungry).

The Causal Inference Framework
Participation in contract farming is determined by an array of factors.As a result, contract farming is not random and happens upon selection.The problem is that the factors determining a farm household's choice to participate in contract farming may also be related to its security status.If this claim is valid, the model will suffer from selection bias because the error term is associated with the treatment assignment.Utilizing a simple binary regression would result in biased estimations.
Adding the source of selection to the regression as control variables might solve the issue of observed selection bias.However, this method is susceptible to model specification.Therefore, this study will use PSM, which is widely used to address the issue of selection on observable.Unlike the former method, PSM is robust to model specification.PSM eliminates the selection bias by matching the treatment and control group units based on their similarity in observed characteristics (Khandker et al., 2009).The matched data should establish a balance in which observed characteristics of the treatment group units are similar to those of the control group.These characteristics will be reflected in the probability of receiving treatment.Units with similar characteristics should have equal probabilities of receiving treatment.The estimated value of this probability is called the propensity score.
The analysis begins with estimating the propensity score, which represents the probability that a farm household participates in contract farming using the probit model.The probit model is given by: where   is the treatment variable, which in this study represents the household's status of contract farming participation.After obtaining the propensity score, participants will be matched with non-participants with similar propensity scores.The treatment effect will then be estimated by comparing matched and non-participant outcomes.As a measure of treatment effect, this study focuses on the average treatment effect on the treated (ATT), which is defined as: Before using the matching result, one needs to check the validity of the PSM.Common support is the primary assumption that needs to be satisfied to ensure the validity of PSM results.The common support condition requires that treatment group units are similar to the control group units in terms of observed covariates.This assumption is considered satisfied when a significant overlap exists between the propensity scores distribution of the treatment and control groups.One can also conduct balancing tests to assess whether the mean of propensity scores within each distribution quantile is the same.For PSM to be effective, it is necessary to ensure that the treatment and control groups exhibit balance.
As PSM only considers selection based on observable factors, it is necessary to examine the possibility of unobserved selection that might bias the estimated treatment effect.Within the framework of PSM, this is the conditional un-confoundedness assumption.In order to evaluate this assumption, this study conducts sensitivity analysis using Rosenbaum bounds (Rosenbaum, 2002) for binary outcome variables.The bounds are calculated using the MHBOUNDS function in STATA.The test identifies "hidden bias" caused by unobserved confounders that might affect the estimation of treatment effects.This study conducted PSM using the STATA PSMATCH2 command by Leuven & Sianesi (2003).The common support assumption and balancing property are assessed using the STATA PSTEST command.The common support result is considered satisfied if the standardized difference of the mean propensity score between the treatment and control group is less than 25% and the variance ratio is between 0.5 and 2.0 (Rubin, 2001).
Covariates are considered to be balanced if the mean difference of the variable between the treatment and control group is not significant according to the t-test result, and the variance ratio is between 0.94 and 1.07.
This study used a double-adjustment strategy to eliminate residual selection bias from imbalanced covariates (Nguyen et al., 2017).Double-adjustment refers to applying regression of the outcome variable on the treatment variable using matched data obtained from PSM.This step added covariates that have not yet achieved balance after the matching procedure as control variables.The probit regression will be employed since the outcome variable is binary.In order to enhance its robustness, the matching weight acquired via PSMATCH2 is also utilized as a probability weight to correct for bias arising from dropping unmatched observations.The estimated ATT is equivalent to the marginal effect obtained from the probit regression analysis using the weight resulting from PSM.

Summary Statistics
Appendix 1 shows summary statistics of our samples, and Appendix 2 shows the mean of each variable based on contract participation.The last column of Appendix 2 shows the test results of the equality of proportion or mean between contract participants and nonparticipants.Only 1.64% of farm households in our sample participate in contract farming.This participation rate is likely underestimated, as those who engaged in verbal or informal contracts (without written legal contracts) may not be aware they are participating in such arrangements.Marketing contracts are more common than production contracts.Approximately 1.37% and 0.92% of farm households participate in marketing and production contracts, respectively.
The majority of farm households are food secure, with only 4.29% experiencing food insecurity.Based on the test of the equality of two proportions, A significant difference exists in the percentage of households experiencing food insecurity between those who practice contract farming and those not involved, with the former exhibiting a comparatively lower rate.It is possible that this finding suggests a positive correlation between contract farming participation and food security.
As anticipated, more than half of agricultural households in our sample reside in rural areas.Regarding market and credit accessibility, most of these households reside in villages that lack direct access to conventional markets but possess favourable access to credit.Most farmers in our sample are male, indicating the continued male dominance in the agricultural sector.Most of them also have not completed compulsory education.Their mean age is 52 years, providing evidence of the phenomenon of ageing farmers.Contrary to popular belief, most of them have their land or livestock.Regarding institutional connection, most farm households are not members of farmer associations or cooperatives.The level of agricultural training exposure is minimal, with less than one member per household on average having received training.The typical household size is approximately three individuals, indicating that most farm households are relatively small.
Based on the test of equality of proportions or means, it can be concluded that contract participants and non-participants differ in various characteristics, including the type of domicile area (urban or rural), access to markets and credit, gender, age, education level of the farmers, membership in associations and cooperatives, ownership of assets, household size, and exposure to agricultural training.There appears to be a potential selfselection issue with participation in contract farming, as both groups possess distinctive characteristics.

Selection Into Contract Farming Participation
Before looking into the impact of contract farming on food security, the factors influencing the choice of farm households to participate in contract farming are analyzed using probit estimation of equation ( 1).The findings are presented in Table 1.Farmers' direct access to traditional markets reduces their willingness to engage in contract farming.This result indicates that producers would rather sell their products directly on the spot market than enter a contract with an agribusiness.Given that producers are contractually bound to conform to the specified quality standards, and they cannot seek alternative buyers, they may be discouraged from participating if the contract does not offer more favourable pricing compared to the market price, risk sharing, or input provision (Widadie et al., 2021).
Contractor farming is more prevalent among farmers who have access to credit.Contract farming may necessitate investment in novel technologies or particular inputs.Farmers with access to credit have an edge as they can secure a loan to finance this investment.They will, therefore, be more likely to accept a contract opportunity (Ganewo et al., 2022).Contract farming is less prevalent among female farmers.Since contract farming typically favours farmers who own land ownership and female farmers lack land ownership (Quisumbing et al., 2015), they are more often marginalized in terms of contract farming opportunities.Contract farming also often involves the production of traditional cash crops or products intended for export, typically cultivated by male farmers.Compared to farmers who do not complete compulsory education, those who accomplish it are more likely to participate in contract farming.According to Kutawa (2016), farmers with a better education level would comprehend the benefits of contract farming.Higher education also helps farmers understand the contract terms better and makes them better at negotiating (Ganewo et al., 2022).Consequently, they will feel more confident and enthusiastic about contract farming.
Membership in farmer associations and cooperatives positively affects contract farming participation.Both are a viable source of various information related to agricultural activities, including contract farming opportunities.Cooperatives also frequently provide contractual arrangements to their members.The members will produce products that will be sold to cooperatives, which will, in turn, sell them to end consumers.Farmers may also collectively participate in contract schemes via associations or cooperatives to strengthen their ability to negotiate favourable terms or secure a better price (Ganewo et al., 2022).A positive correlation has been observed between the size of a household and the probability of participating in contract farming.This favourable outcome can be attributed to the ability of households to employ their family members as a source of labour (Ganewo et al., 2022).Contract farming frequently comes with an opportunity to expand production scale.Scaling up requires additional labour.If more members are employed as family workers, farm owners do not need to take on additional paid labour.Hence, they may find it easier to accept contract farming offers.
Exposure to agricultural training positively impacts contract farming participation.The favourable outcome can be attributed to the knowledge and expertise individuals gain through training (Ba et al., 2019).By acquiring these competencies, farmers can meet the buyers' requirements and effectively utilize the new technologies they provide.They would appeal to more potential buyers, boosting their chances of being offered a contract.

The Impact of Contract Farming on Food Security
The analysis begins with estimating the baseline estimation using probit regression without (Baseline 1) and with control variables (Baseline 2).All factors that influence participation in contract farming are included as control variables and covariates in all following analyses.The results are displayed in the first and second rows of Table 2.Both estimates produce the same results regarding the direction, with slight differences in the magnitude of the impact.Both estimates show that those participating in contract farming may have a significantly lower probability of experiencing food insecurity.approach is employed after obtaining the matching data to rectify any remaining imbalance that cannot be resolved through matching.The estimated ATT, shown in Table 2, is the marginal effect from the double adjustment regression for each matching method.

Tabel 2 Esrimation of ATT of Contract Farming on Food Security
The estimated ATT from all PSM estimates demonstrates similar outcomes, indicating that the estimation is robust.All estimates consistently suggest that those who practice contract farming experience a significantly lower probability of experiencing food insecurity, thereby highlighting its beneficial impact on food security.In addition, the outcomes do not deviate significantly from the baseline estimates.It raises the question of the effectiveness of the PSM in mitigating potential biases.The effectiveness of PSM is greatly dependent on the choice of covariates.The similarity between PSM and baseline estimates might suggest that there may be limitations in the selected covariates to account for selection bias.It suggests that there is a possibility that the PSM omits variables that play a role in contributing to the remaining source of selection bias.
Among the estimates, Radius (0.001) yielded the best-matched data with the lowest mean difference of propensity score between the treatment and control groups while retaining most observations.This method also portrays good balancing properties, as shown by the result of the balancing test in Appendix 3 and the propensity distribution graph in Figure 1.Based on this estimate, participating in contract farming lowers the probability of a household experiencing food insecurity by 1.17%.The size is about 27% relative to the control group's mean (see column 4 of the row in Appendix 2).It appears that this impact is relatively small.

Robustness Test
For the robustness test, several alternative estimations were conducted: 1) direct NNM(1); 2) inverse probability weight regression adjustment (IPWRA); 3) probit regression using propensity score as control; and 4) placebo test using randomly generated treatment variable.The result is presented in Table 3.The first four estimations exhibit consistent and similar results regarding direction with previous estimates.Therefore, the findings in Table 2 are robust.If the placebo treatment significantly affects food security, the treatment effect estimated in prior estimates cannot be considered causal.The insignificant placebo effect implies that contract farming and food security are causally related.
Sensitivity analysis in STATA using the MHBOUNDS command was conducted to address potential unobserved selection bias, employing Rosenbaum bounds for binary outcomes.
The result is displayed in Appendix 4. It indicates that any unobserved bias in contract farming participation would have to modify the odds ratios of contract farming participation for the treated and control groups by a factor of 1.4 to 1.75 to undermine the interpretation of the impact of contract farming on food security.Such a level of bias is likely to manifest.Should such bias exist, it could result in the conclusion that contract farming does not significantly affect food security.Given that prior estimates indicate a minor impact of contract farming, it would not be so unexpected if the existence of any unobserved bias would lead to the conclusion that contract farming does not affect food security.Thus, instead of confidently stating that contract farming definitively improves food security, it is tentatively proposed that contract farming can potentially improve food security.

Mechanism Analysis
One mechanism in which contract farming positively affects food security is through the positive income effect.Competition between production for contract and own consumption or competition in time allocation might happen sometimes, but the positive income effect dominates the negative substitution effect.Literature suggests that contract farming increases productivity and farm income (Hoang, 2021;Khan et al., 2019;Selorm et al., 2023).With a higher income, farmers can purchase more and higher-quality food, thereby improving their food security.They can also save additional income from contract farming to buy food during the off-season (Bellemare & Novak, 2017).Thus, they can guarantee a continuous food supply and achieve food security throughout the year.
Due to the unavailability of data, testing the mechanism through efficiency improvements is not feasible.This study is also unable to assess whether contracted farms have a higher yield or income due to the poor quality of yield data and the absence of income data.However, the data contains a variable about whether the contribution of on-farm income to the total household income has increased since last year.The information provided is based exclusively on the respondent's recall, not the original time series data.
ATT of contract farming on the income comparison variable is estimated, and the resulting ATT is statistically significant.Our research reveals that participating in contract farming is associated with a 6.54% rise in the probability of experiencing a higher contribution of on-farm income.Upon further examination, the ATT of contract farming on food security is reassessed, considering the income comparison variable as a control.The ATT changed minimally to 0.0114 in absolute value, differing by only 0.0003 from the first estimate.It indicates that approximately 2% of the aggregate impact of contract farming on food security may be attributable to the positive effect on income.This contribution is relatively minor.This minor result could be attributed to the inadequacy of the income comparison variable in accurately assessing changes in income.The variable in question represents whether there is an increase in on-farm income contribution rather than capturing the actual change in income.Due to its reliance on the respondent's recall, the variable is highly subjective and susceptible to bias.Moreover, there may be a more substantial mechanism beyond the income effect, which provides a more precise justification for the favourable impact of contract farming on food security.

Heterogeneity Analysis
This study also estimates the impact heterogeneity based on domicile area type, farmer's gender, and asset ownership.4 shows that contract farming enhances the food security of farm households, irrespective of their location (urban or rural).However, the impact will be more significant for farm households residing in rural areas.Rural farm households experience nearly double the impact compared to urban areas.It might be attributed to the fact that rural farmers mostly rely on on-farm income as their primary source of income.Hence, income earned through contract farming might serve as the primary source of income for most rural farmers.In addition, rural farmers, especially those living in isolated areas, may face difficulties locating markets for their agricultural products.Less than a third of rural farmers lack direct access to traditional markets.Hence, contract farming may be the only way to market their agricultural products and earn income.Therefore, it would be unsurprising if contract farming were to yield more advantages for them.
Tabel 5 Impact Heterogeneity Based on Farmer's Gender While the differences may not be substantial, table 5 shows that contract farming has a more significant impact on food security for households with female farmers.Women often play a crucial role in managing household income and expenditures.Women also are often responsible for food production and distribution within households.If the beneficiary is a female, the income generated is more likely to be used for the benefit of the entire household, including food and nutrition.According to Debela et al. (2022), if women lose authority over income through contract farming, it might have a detrimental impact on food security, as women are known to allocate more of their money towards purchasing nutritious food.There is a potential for contract farming income to be used for personal expenses, such as alcohol or meat, rather than being allocated towards healthy food, mainly if men assume the responsibility of managing the income.Finally, Table 6 demonstrates that farmers who do not own land or livestock will experience more significant food security improvements due to contract farming.Farmers without assets experience roughly twice the impact.This can be attributed to the production risk encountered by farmers who lack land or livestock ownership.Farmers lacking land or livestock ownership must lease land or livestock for agricultural production.Consequently, the cost of production for these farmers will be higher than for those who possess their land or livestock, as they have to pay for rental expenses.They will suffer substantial financial losses if they cannot successfully market their products and generate income.Contract farming is potentially the optimal choice for guaranteeing their income.Therefore, contract farming will assume greater significance and yield more advantages.

Conclusion
Utilizing a causal inference approach with PSM and SITASI 2021 data covering all commodities and 34 provinces of Indonesia, this study empirically examines the impact of contract farming on the food security of smallholder farm households in Indonesia.Food security is measured using FIES.Our findings show that contract farming has the potential to improve food security.Participating in contract farming may slightly decrease the probability of experiencing food insecurity.However, the impact is relatively minor.The robustness of our estimations regarding the impact of contract farming on food security was demonstrated through various alternative methods and placebo tests.
However, it appears that our results are at some point influenced by unobserved selection bias.Contract farming may not significantly impact food security when this bias is present.
Regarding the mechanism, it is concluded that the suggested positive income effect, as indicated by multiple prior studies, only explains two percent of the total impact on food security.The remaining should be explained by mechanisms unrelated to the income effect.Regarding impact heterogeneity, our findings show that the positive impacts are more evident among farmers who live in rural areas, are female, and do not possess land or livestock.These findings demonstrate the effectiveness of contract farming in promoting rural development, empowering women, and alleviating poverty.
Regarding the factors influencing contract farming participation, this study shows that location (urban/rural), market access, credit access, farmers' gender and education, membership in farmers' associations and cooperatives, household size, and exposure to agricultural training affect contract farming participation.The fact that participation is more prevalent among male farmers, those who have completed compulsory education, and those who reside in urban areas provides further evidence that discrimination continues to exist with regard to contract farming opportunities.The likelihood of engaging in contract farming was positively correlated with credit accessibility, membership in agricultural associations or cooperatives, and agricultural training.
It is important to note that this study has several limitations.First, this study may not adequately capture the prevalence and dynamics of informal contract farming, especially regarding informal contract farming, such as verbal contracts.A dedicated and detailed study designed to examine informal contract farming is recommended, as it cannot be effectively captured through national-level surveys.Second, due to the absence of income data, this study cannot properly confirm and measure the presence and extent of the income effect of contract farming.In order to assess the degree to which contract farming affects food security via the income effect, it is essential to employ actual income data.Therefore, it is essential to incorporate such data to gain more comprehensive understanding of the correlation between contract farming, income, and food security.
From these results, several recommendations can be made.Given the findings indicating that contract farming has only a minor impact on food security, while contract farming can function as a tool to improve food security, it should not be relied upon as the sole strategy, as it may not be adequate on its own.A holistic approach to improving food security is needed.It may entail assisting with a combination of farming methods, adoption of technology, and other income-generating efforts to strengthen the total capacity of households to withstand food insecurity.Given our findings, it is evident that contract farming has a more significant impact on individuals who have been marginalized, such as those residing in rural areas, women, and those without assets.Therefore, it is recommended to prioritize targeted interventions that specifically address the implementation of contract farming for this marginalized group.In order to increase contract farming participation, the first thing that should be addressed is gender, education, and urban-rural discrimination in contract farming.Efforts should be made to create equal opportunities for all farmers.It can include providing support, access to land, inputs, and credit facilities specifically designed for female, uneducated, or rural farmers.

Table 6
Impact Heterogeneity Based on Asset Ownership