Joint decision-making, technology adoption and food security: Evidence from rice varieties in eastern India

Highlights • Spouses are increasingly making farming-related decisions.• Join-decision making (husband and wife) on farming is increasingly important.• Assess married couples’ joint decision-making on rice variety selection and rice productivity.• Farms with joint decision-makers and wife) would have higher yields when selecting older rice varieties.• Farm families with joint-decision making have higher food security.


Introduction
Increased crop productivity through the adoption of modern technologies (improved seeds, synthetic fertilizer, irrigation, and mechanization) is considered one of the greatest legacies of the Green Revolution in South and Southeast Asia (Tsusaka and Otsuka, 2013) and a source of food security.The introduction of modern technologies has increased food security and reduced poverty in developing and emerging economies worldwide (de Graft-Johnson et al., 2014).Rice is one of the crops that benefited from genetic improvement and resulted in estimated economic gains between US$ 296 million-US$ 9.9 billion (Raitzer et al., 2015).In India alone, there are more than 900 modern rice varieties  and around 47 hybrid rice varieties 1 (1994-2010) released by the government of India (DRD, 2020).
Rice plays a vital role in Indian agriculture as a major supplier of calories in the Indian diet, covering approximately 35% of the total area under food grains (GOI, 2023).Despite the rice varietal developments, there has been noticeable slow productivity growth in food grains in recent decades compared to the early decade of the Green Revolution (Khush, 1999).Eastern India 2 is one region that experienced slow rice productivity and food insecurity.Two possible reasons may contribute to the low growth in rice productivity.These include a lack of desirable traits of high-yielding rice varieties (Hossain et al., 2003) and adverse effects of climatic conditions (Tsusaka and Otsuka, 2013).For example, Pandey et al. (2007) estimated that drought events between 1970 and 2000 resulted in an average loss in rice production of 5.4 million tonnes or $162 million.This low yield growth will substantially impact the region since rice farming is dominated by fragmented and smaller holdings, lack of irrigation facilities, and frequent adverse climatic conditions.Given the conditions in rice farming states in eastern India, this may trigger the continuing vicious circle of low input-low output agriculture and, consequently, food insecurity and loss of livelihood.
To reduce the variability in farm income and uncertainty in livelihood, male heads of households have sought off-farm or dual employment to increase family income.The labor movement from agriculture has increased the daily nominal wage rate for various farm activities, including plowing, sowing, and rice transplanting. 3 A report by the International Crops Research Institute for the Semi-arid Tropics (ICRISAT) shows that the nominal wage rate increased 3.6 to 4.2 times during the 2004-2014 period (Bhattarai et al., 2014).The movement to the non-rural sector was further enhanced by government programs like the Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA), 4 which has led to labor shortages in agricultural production.
In most developing economies, farming decisions (such as selecting crops, technology, and labor) traditionally have been made by the male household heads (Orr et al., 2016).However, in the absence of male decision-makers, spouses are increasingly responsible for making farming-related decisions.Studies by Datta and Mishra (2011) and Maharjan et al. (2012) reveal that the Indian rice farmers' income sources have become more diversified in recent years, which has led to significant changes in gender roles within the households.Further, existing studies tend to rely on household heads' information in analyzing the adoption of technologies (Mehar et al., 2017).
Today, it is more likely than in the past that farming-related decisions are made jointly.Joint decision-making is gaining significant traction in literature (Ibrahim et al., 2012;Aregu et al., 2011).To this end, studies investigating the adoption of technologies focusing on male decision-makers may lead to biased estimates (Agarwal et al., 2013).However, there needs to be more studies on the married couple's participation in decision-making regarding farming-related activities, especially rice seed varieties in India.
Thus, the objective of this study is twofold.First, to investigate the factors affecting married couples' participation in the joint decisionmaking of adopting rice varieties.Second, to assess the impact of the joint decision-making strategy on food security-measured by rice productivity.This paper used the 2016 Rice Monitoring Survey, a large nationally representative household-level survey data than was previously reported from India.Our study contributes to the literature in several ways.First, the study focuses on why married couples choose joint decision-making regarding rice varieties.Since this study is based on observational data, the decision-making strategy choices are not distributed randomly, making the two groups systematically different.The study employs endogenous switching regression (ESR) to account for selection bias and endogeneity (Alene and Manyong, 2007;Mishra et al., 2017;Bairagi, Mishra, and Durand-Morat, 2020).Second, the impact of a joint decision-making strategy on rice productivity is generated using counterfactual estimation.Findings from this study will guide policymakers regarding the implementation of agricultural development strategies, technology uptake, and extension programs.In particular, developing outreach materials appropriate for spouses and their involvement in farming to increase food security (or rice productivity) in the eastern region of India.

Women's participation in decision-making
Women's participation in rice farming is often associated with their share in production labor.In eastern India, women provide labor in nurseries, transplanting, weeding, and harvesting, comprising at least 60% of total rice labor requirements (Pandey et al., 2012).However, it is not always guaranteed that women significantly influence decisionmaking regarding critical issues about farming and household matters.Behura et al. (2012) and Bagchi and Bool-Emerick (2012) found that while women in eastern India contribute substantial labor, the male household head still makes decisions on which technology or technology practices to adopt.Women are responsible for decisions about the selling of rice production.However, in the absence of male decision-makers, spouses are increasingly responsible for making farming-related decisions.Most of the literature tends to rely on household heads' information to analyze the adoption behavior and exclude women who are not household heads.Due to its simplicity, focusing on the household head is a standard method used in most existing literature, particularly in the South Asian setting.However, with the changes in the sources of income to nonfarm rural sources among rice farming households, there is evidence of the changing household roles, which result in the high participation of women in decision-making regarding household and farming decisions (Paris et al., 2010).
Most of the studies that examine intrahousehold decision-making are from sub-Saharan Africa and Latin America.There is little evidence of the joint decision-making process between married couples in determining rice varietal choice in India.Though the Indian government has implemented several programs to improve women's status, it is still necessary to know women's involvement, particularly in farming decisions.Thus, investigating decision-maker constraints is an effective way to understand household technology adoption (Deerie, 2005), particularly in India's rice farming as a staple food.The characteristics of women who are not household heads but are involved in making decisions about technology adoption, food consumption, and food security5 is missing.

Conceptual and empirical framework
The rice variety that the farmer adopts is an important factor in increasing rice productivity.This study focuses on the decision strategies of married couples of rice farming households in eastern India.The decision strategy will depend on whether the couple has joint participation in selecting rice varieties or the household head 6 (husband) is the decision-maker.Since the survey queried married couples on seven farm production-related decisions, 7 we only considered the joint decisionmaking regarding selecting rice seed varieties.Each of the couples was asked about their involvement in decision-making.Only couples who answered the question were included in the analysis.We categorized decision-making into the following categories: (1) husband only decides in the presence of the wife; (2) wife only decides in the presence of the husband; (3) both husband and wife participated in determining the choice of a rice variety to be used in the coming season.In our data category, 2 is non-existent.Thus, the joint decision-making regarding rice variety takes a value of 1 when both husbands and spouses choose the rice variety and 0 if the husband solely decides on the rice variety.
Several factors are included in estimating the Probit model, in which the data description is presented in Appendix Table 1.The choice of decision-making strategy follows a random utility maximization framework where the latent variable J * i , describes the i th household decision strategy on whether the couple jointly participates or only the husband solely decides the rice variety.The latent variable, J * i , captures the expected benefits that household i receives by jointly participating and can be determined by the observed attributes X, and unobserved characteristics, ε and expressed as: The household i will choose a joint participation strategy in deciding on rice variety if J * i > 0, or 0 otherwise.The error term is μ i with mean zero and variance σ 2 .
The endogenous switching regression (ESR) is used to analyze the 4 Enacted as the National Rural Employment Guarantee Act of India ( 2005) is a public policy in India that pays people to seek employment.The wage rate is higher than the daily wages of agricultural workers (Sharma, 2011).availability, food access, food utilization and food stability.Food availability comprises of food production (total output or yield) and crop diversity.See Barrett (2010), Gregory et al. (2005), and Jenkins and Scanlan (2001) for different measures and dimensions of food security. 6Household head is household member who is responsible in rice farming.In the married couple sample, all the respondents identified household head as the husband. 7We use the operator interchangeably with husband and spouse as a wife.
impact of the couple's joint decision-making strategy on rice yield.This method was developed by Lee (1978).Since then, this method has been used in several empirical studies (Alene and Manyong, 2007;Mishra et al., 2017;Bairagi et al., 2020.Separate outcome equations are specified if the couple has joint decision-making and the husband is the sole decision-maker: (2) where Y i is the outcome variable (yield) of the i th household when using couple's participation strategy (1 = joint decision-making; 0 = husband sole decision-making), Z is a vector of explanatory variables (farmer and plot characteristics), and α are parameters to be estimated.The outcome variable Y 1i when the couple jointly decides on the rice variety while Y 2i is observed when the husband solely decides on the rice variety.In using OLS, the estimates α 1 and α 2 in Equations ( 1) and (2) will suffer selection bias since the choice of strategy is endogenous.This implies that error terms in Equations ( 2) and (3) will have non-zero expected values (Lee, 1978;Madala, 1983).The error terms in the Probit model (first stage), 1, and 2 are assumed to have a tri-variate normal distribution with mean zero and non-singular covariance matrix, which given as where σ 2 μ is the variance of the error term of the selection equation; σ 2 ε1 and σ 2 ε2 are variances of the error terms of the outcome functions in 1 and 2; σ ε1μ and σ ε2μ are the covariance of μ i , ε 1i , and ε 2i .According to Maddala (1983), since Y 1i and Y 2i are not simultaneously observed, the covariance between μ 1i and μ 2i are not defined.Based on the given assumptions, the expected values of ε 1i and ε 2i conditional on sample selection are non-zero: (5) where ϕ is a standard normal probability density function and Φ standard normal cumulative functions.The ratio between ϕ and Φ evaluated at αX i is the inverse Mills ratio (λ 1i and λ 2i in Equations ( 5) and (6).
Substituting λ 1i = ϕ(αXi) Φ(αXi) and λ 2i = ϕ(αXi) 1− Φ(αXi) in Equations ( 2) and (3), then the outcome equations can be expressed as where ε 1i and ε 2i have zero conditional means.If the estimated σ ε1μ and σ ε2μ are statistically significant we can reject the null hypothesis (absence of sample selectivity bias).The result suggests that there is evidence of endogenous switching.Since the generated regressors arising from two-stage estimation often result in heteroscedastic error terms ε 1i and ε 2i , OLS estimates for Equations ( 7) and (8) will be inefficient (Antle, 1983;Khonje et al., 2018).An efficient method in estimating endogenous switching models uses the full information maximum likelihood ([FIML] Loskin and Sajaia, 2004).The FIML simultaneously estimates the selection and outcome equations to have consistent standard errors.The FMIL estimates are obtained using movestay command in STATA (Loskin and Sajaia, 2004).On the other hand, for the model to be identified, exclusion restrictions need to be included.Thus, at least one variable in X, which is not included Z.
The choice of instruments is considered valid if they can influence the selection (joint decision-making strategy) equation but not the outcome (yield).The instruments include access to credit and the differences in the couple's age (husband-wife).Access to credit among women has been proven to benefit women by increasing household assets and savings (Amin and Becker, 1998), resulting in self-confidence and recognition of her role in the household (Sharma and Varma, 2008).Khush (1999) found that access to credit among households in Bangladesh impacted women's participation in the household decisionmaking processes.However, not all credit loans were allocated for the improvement in production.Chavas et al. (2005) found that Gambian farmers who availed of loans from Osusu 8 often used the funds for nonfarming-related activities, and only a few were used to purchase inputs and equipment.Only a few have existing agricultural loans, most of which were used for medical and school expenses.Thus, this can be used as an instrument since it may influence participation in decision-making but may not directly determine rice productivity.
The difference in the couples' age is the second instrument that affects how a decision is made-joint or solo (husband).This represents the power relation between the couple.For instance, Kantor (2003) examines Indian female participants in the home-based garment sector and found that a couple with a large age difference would place more power on the husband.In addition, Schneebaum and Mader (2013) found that smaller age differences initiate a joint decision-making process among married couples rather than one person making critical decisions.The age difference may affect women's participation in deciding on rice variety but not necessarily the outcome variables (yield).
We use Di Falco et al.'s (2011) falsification test for admissibility by to test for the exclusion restrictions, and the result is presented in Appendix Table 2. Results confirm that the instrumental variables jointly affect participation in joint-decision making or the husband solely decides (χ 2 = 19.860;p = 0.000) but do not affect the outcome variable (yield).Additional tests for the inclusion instrument are presented in Appendix Table 3.The Durbin and Wu-Hausmann (DWH) test for exogeneity is found to be insignificant.To test the identification of the model, the Anderson canonical correlation statistic by Baum et al. (2007) was used, which rejects the null hypothesis that the model is underidentified (LMstatistic = 21.25,χ 2 p = 0.000).The Anderson-Rubin's test by Baum et al. (2007) was used to test for weak instrument robust inference and found to be insignificant (χ 2 = 0.58; p = 0.746).
The above results are used to estimate the counterfactuals and average adoption effects using cross-sectional.The conditionally expected outcome (yield) was computed to generate the average treatment effect on treated (ATT) and average treatment effect on untreated (ATU) with joint decision-making (treated group) and husband decisionmaking (base group) in actual and counterfactual scenarios (Shiferaw et al., 2014).The conditional expectations for each outcome expectations are the following: Household with joint decision-making (adopters) (actual): Households with husband (or male operator) decision-making (nonadopters) (actual): Households with husband decision-maker that decided to make decisions jointly (counterfactual): A household with joint decision-makers that decided for a husband to 8 Osusu is local rotating saving and credit association in Gambia (Chavas et al., 2005).

M.L.L. Malabayabas et al.
make decisions regarding the choice of rice varieties (counterfactual): Recall that the ATT estimates the effect of participation strategy on the actual yield of households that make decisions jointly.Specifically, the difference between Eq. ( 10) and ( 12): The impact on yield for husband decision-makers had they jointly made decisions is estimated using the average treatment effect on the untreated (ATU) is the difference between Eq. ( 12) and Eq. ( 10) specifically: The treatment effects can be further identified through heterogeneity effects (Carter and Milon, 2005).A household with joint decisionmakers (actual) may have a higher outcome (yield) than those households with husband decision-makers regardless of their strategic decision but due to other unobservable characteristics.This effect is termed "effect base heterogeneity" (BH) and is defined as: Therefore, the BH for a household with a joint decision-maker is the difference between Eq. ( 9) and Eq. ( 11), while BH for a husband (male) decision-maker is the difference between Eq. ( 10) and Eq. ( 12).
The robustness of ESR estimates was compared to Propensity Score Matching, commonly used to address selectivity bias due to observable characteristics.This technique matches households with joint and male decision-making based on observable characteristics.Table 2 shows the descriptive statistics of the variables used in the PSM.Appendix Table 5 presents the results of the probit regression, which shows that the spouses'(wife) education and share of assets owned, and the difference in age of the husband and wife has more likely to affect the joint decision-making strategy choice of the household.Appendix Table 6 shows a comparison of two algorithms, shows that the nearest matching (NNM) algorithm with a maximum of three matches with caliper 0.25 σ p best suits the analysis.The matching procedure yielded a total of 2,454 matched.A comparison of means was used to assess if there is no significant difference between two groups in the matched sample, thus fulfilling covariates' balancing condition using pstest command in STATA 17 (Appendix Table 7) (Caliendo and Kopeinig, 2005).A visusal inspection of the overlap in the distribution of the propensity scores shows that the common support is satisfied (Fig. A1).The PSM estimation regarding the effects of yield is presented in Appendix Table 8, which shows that the joint decision-making households have 103.45kg/ ha lower rice yield than male decision-making households with p ≥ 0.548.The results suggest no yield impact regarding who decides the rice variety to plant.In terms of magnitude, the results of PSM estimates are lower compared to ESR estimates.The same results were found by Tesfaye (2016) when comparing the PSM and ESR estimates due to understating the effects leading to a downward bias of results as compared to ESR that accounts for unobserved heterogeneity.

Survey data
The study uses the 2016 Rice Monitoring Survey, which focuses on rice farms in eastern India.A rice-producing household is defined as a household that produced rice during the past 12 months.The survey targeted the rural population of eastern India by randomly selecting rural areas based on the 2011 Census of India.Four states in the eastern part of India are considered in the study: eastern Uttar Pradesh, Odisha, Bihar, and West Bengal (Fig. 1).A multi-stage sampling technique was adopted in selecting the respondents.In the first stage, the number of districts was randomly selected in each state using the Census of 2011.9On the other hand, the second stage involves determining the number of villages based on the proportion of each state's total rice area, keeping the total number of villages at 720.Among the selected villages, household samples are randomly selected using the household census village data.A total of 101 districts and 2,471 rice-producing households are included in the survey (Table 1).Only families that reported with married couple are included in the study.All households in the sample considered the husband as their household male head and wife as their female head.A structured questionnaire was used to interview two household members.Information regarding rice production and farmrelated decision-making were collected from husbands, and information regarding livestock, household assets, and farming decisions were also collected from the wife (spouse).The survey employed male and female enumerators in the interview process to elicit unbiased responses.The male enumerator interviewed the household head, while the female enumerator interviewed the spouse.The study focused on information regarding the 2015 wet season, the primary rice-growing season in eastern India.
The summary statistics and definitions of the variables are presented in Table 2 and Appendix Table 1, respectively.The identified household head of all the samples is the husband (male) with an average of 48 years old.Regarding choosing the rice varieties, results show that half of the sample follows a husband decision-making strategy in selecting the varieties.A husband under husband decision-making strategy has higher educational attainment than joint decision-makers.Regardless of decision-makers, it shows that there is a higher number of scheduled tribe/caste (36%) and other backward caste (OBC) (43%) in joint decision-making households than in husband decision-making households.There is a considerably higher number of households under joint decision-making in Odisha and Uttar Pradesh than in male decisionmaking households.Aside from rice production, respondents have other sources of income or off-work like employment in the private sector and small businesses.Results show that more households with joint decision-making have off-farm work than husband decisionmaking households.Regarding asset ownership, it shows that spouses (wife) under the husband decision-making households have a higher share of farm asset ownership than in joint decision-making households.
The households in our sample are considered marginal rice farmers, with an average rice area of 0.43 ha.Families under joint decisionmaking have larger rice areas than households under male decisionmakers.Rice farmers allocate at least one plot for rice production during Kharif (wet) season to mostly medium the land part of the landholdings.More than half of the respondents (58%) use supplemental irrigation such as shallow or deep tube wells, implying that some farmers still rely on rainfall as their primary source of irrigation.Since the rice ecosystem in the sample location is considered rainfed, the farmers encounter several water-related problems.In 2015, around 57% of the farmers experienced floods and droughts, particularly in households with husbands' decision-makers.Recall that Pandey et al. (2007)   note that one reason for the low adoption of technology in eastern India is the frequent flooding and droughts, thus hindering productivity.The rice yield in the sample is lower than the reported national average yield of 3,700 kg/ha (IRRI, 2019).Male decision-making households have significantly higher yield by 134 kg/ha than joint decision-making households.Rice farmers use several significant inputs, including seeds, labor, fertilizer (NPK, DAP, and Urea), machinery, and pesticides.The households under joint decision-making apply significantly larger amounts of seeds than their counterparts.Regarding fertilizer, the households with the husband decision-maker use more fertilizer than joint decision-making households by 46 kg/ha.Regarding machinery and pesticide application, the table shows that a significantly higher number of husband decision-making households use these inputs than their counterparts.Table 2 also presents small-scale rice producers' labor (family, hired, and contract) requirements.Results show that most households employ the same amount of labor except for contract labor.Families with husband decision-makers use significantly higher contract labor than joint decision-maker households by 7 (persons per day/ha).
Finally, the adoption of rice variety varies depending on the major decision-maker.It shows a significantly higher adopter of MRV5 and MRV6 (mixed) among husband decision-makers and joint decisionmaking households, respectively.Clear distinctions regarding the two household types are very evident in rice variety adoption.For instance, Gauchan et al. (2012) and Behura et al. (2012) found that farmers use multiple rice varieties depending on the land types that vary according to topographical sequence and moisture level to ensure production that will supply their family consumption needs.Since the sample area is flood/drought-prone, MRV6 (mixed generations) serves as a safety net for household production since most households produce rice for home consumption.

Results and discussion
The results of the ESR model are presented in Appendix Table 4. Due to brevity and space constraints, we only discuss the effects of the impact model.Table 3 shows the expected quantity of rice produced (kg/ha) under actual and counterfactual conditions.For instance, the cells (A) and (B) represent the expected rice yields (kg/ha) observed in the sample.Cells (C) and (D) represent the expected rice yields (kg/ha) in the counterfactual case.Since there is no selection effect 10 for the husband as decision-maker, we only focus on the expected rice yield of the joint decision-maker and its counterfactual.In the case of "all rice varieties," results show that the expected rice yield of households under joint decision-making was about 830 kg/ha and 1,029 kg/ha for husband sole decision-making households.However, this simple comparison could be misleading in attributing the different values of expected yields for both groups to joint decision-making.Columns 5 and 6 of Table 3 report the treatment effects of participation in joint decisionmaking about variety selection.In the counterfactual case (C), joint decision-making households would have produced less (about 160 kg/ ha, or 24%) if they had not adopted a joint decision-making strategy.Similarly, in the counterfactual case (D), rice farmers who did not adopt joint decision-making adopted (or husband decision-maker households) would have produced about 81% less if they had adopted joint decisionmaking.
In terms of adopted rice varieties, Table 3 shows that yield advantage differs depending on rice varieties.Most rice smallholder households show a yield advantage when adopting joint decision-making in all rice Standard deviations in parentheses: *** p < 0.01, ** p < 0.05, * p < 0.1.
1 Husband and spouse are making farming-related decisions jointly. 2 The respondents are the household head who are knowledgeable in rice farming.In this study, all the household heads are husbands.
3 Includes designated groups of historically disadvantaged indigenous people in India.The terms are recognized in the Constitution of India (GoI), and the various groups are designated in one of the categories.Since independence, the scheduled castes and tribes were given Reservation status, guaranteeing political representation. 4Includes castes that are socially and educationally discriminated.
5 At least one of the household members has off-farm labor like a salaried job (government or service industry) and small business.
12 Extended family members include parents, siblings, and relatives. 10The estimated correlation coefficient terms (ρ i ) are presented in Appendix Table 4. Results show that there is positive and significant self-selection in both decision-making strategies, which suggests that observed and unobserved factors affect the joint decision-making strategy choice and rice yield.On the other hand, the correlation coefficient for husband decision-maker is not statistically significant, which means that households under joint decision-makers and husband decision-makers will have the same rice yield using the old decisionmaking strategy.
varieties except for MRVGen2 (1986 and older) and MRV5 (Hybrid).In other words, joint decision-making on rice variety selection positively impacts rice yields for most of the rice variety types particularly MRVGen1 (before 1986).Table 3 (Panel 3) show that the expected rice yield of households producing MRVGen1 under joint decision-making was about 791 kg/ha and 767 kg/ha for husband sole decision-making households.In the counterfactual case, joint decision-making households would have produced less (about 224 kg/ha, or 39%) if they had not adopted a joint decision-making strategy.Similarly, in the counterfactual case (D), rice farmers who did not adopt joint decision-making adopted (or husband decision-maker households) would have produced about 78% more if they had adopted joint decision-making.Indeed, the study shows exciting findings.A study by Paris et al. (2008) found that male and female farmers in eastern Uttar Pradesh have preferred traits in choosing a particular variety based on varying factors (e.g., environmental, socio-economic, and cultural gender roles).Results showed that farmers prefer rice varieties based on agronomical traits (e.g., yield, tolerance to submergence, pest resistance, and fertilizer responsiveness).In contrast, spouses like the intrinsic qualities of varieties (e.g., taste, cooking qualities, and grain shape).Indeed, results underscore the importance of rice quality when farmers choose rice variety, especially during the joint-decision-making process.Our findings extend the literature findings (Mottaleb, Rahut, and Mishra, 2017) that argue for innovating rice varieties with consumer preferences in mind.However, both men and women farmers prefer high yielding, good taste and aroma, and postharvest quality.One popular mega-variety is Swarna, which covered almost 30% of the total rice area in eastern India in 2015 (Tsusaka et al., 2015) and belongs to the MRVGen1 category.High productivity and consumer preference may be driving factors.MRVGen1 contains attributes that are attractive to both farmers and spouses.For instance, studies (Tsusaka et al., 2015;Mehar et al., 2017) have shown that farmers prefer most mega-varieties due to their higher yield and good eating quality.A sensory evaluation analysis by Champagne et al. (2010) indicates that Swarna has a rough aftercooking surface suited for the thick sauce prominent in Indian cuisine.Thus, it is no surprise that the study found positive effects of joint decision-making on rice yields in MRVGen1-rice varieties bred for grain yields, grain quality, and consumer preference attributes.
Finally, Table 3 reveal transitional heterogeneity effects in adopting a joint decision-making strategy in most of rice variety types.Results indicate that smallholder rice producers with joint decision-makers produced significantly higher rice yields (or more food secure) than households in the counterfactual case (C).Thus, findings show some essential heterogeneity sources that make the rice farming smallholders under joint decision-makers better producers than their counterparts.
The yield advantage also differs depending on the caste where the family belongs (Table 4).For example, households under Other Backward caste (Table 4, panel 2) shows that in counterfactual case (C), joint decision-making households would have produced less (about 113 kg/ ha, or 27%) if they had not adopted a joint decision-making strategy.Similarly, in the counterfactual case (D), rice farmers who did not adopt joint decision-making adopted (or male decision-maker households)   would have produced about 83% less if they had adopted joint decisionmaking.The participation of women decision-making among the marginalized are becoming more evident.A study by Paris et al. (2008) found that women from marginalized groups in Eastern Uttar Pradesh are now more empowered, enabling them to choose the rice variety depending on their needs.Results show a transitional heterogeneity effect in adopting a joint decision-making strategy in all caste groups.Results indicate that smallholder rice producers with joint decisionmakers produced significantly higher rice yields (or more food secure) than households in the counterfactual case (C).

Conclusions and implications
This study analyzed the factors affecting joint decision-making in selecting rice varieties and assessed the impact of joint decision-making on food security measures-rice yields.The study used a large and unique 2016 Rice Monitoring Survey and the ESR method.In general, findings showed that farming households with joint decision-making tend to have higher rice yields than their counterparts, but not all the time.Households with joint decision-making strategies positively impacted rice yield, particularly when adopting MRVGen1 (before 1986) varieties.Farmers who adopted joint decision-making tend to have higher rice yields of MRVGen1 (before 1986) than farmers who did not adopt a joint decision-making strategy in the counterfactual.Adoption of MRVGen1 (before 1986) among joint decision-making households increase rice yield due to familiarity and grain quality preference with the varieties.Similarly, a joint decision-making strategy on choosing rice variety has higher yield advantage than male decisionmaking households in all caste.
Since joint decision-making households perform well when choosing MRVGen1 rice varieties (composed of rice varieties like Swarna, 11 Lalat, MTU-1001, Moti, and Sambha Mahsori), increasing awareness about the  The years of education of the wife (years) Difference in the education (husband-wife) This is the difference between the husband and wife's education.Total rice area (ha) Total area allocated for rice (ha) Source: Laborte et al. (2015), Launio et al. (2008) and GOI (2023).
flood-tolerant, better grain quality and improved pest resistance rice varieties should be targeted in the above group.There are ways of expanding a spouse's participation in rice production activities.First is the involvement in varietal development.One way of verifying newly developed rice variety lines' acceptability is through Participatory Varietal Selection (PVS).In this method, husband and their spouses can participate in the initial screening before varieties are released (Paris et al., 2008;Paris et al., 2011).Usually, participants are selected based on the proportion of the male-headed and female-headed in the area responsible for making farming decisions in the household.Since our sample shows that eastern India is mainly composed of male-headed families, there is a possibility that women can also participate in choosing rice varieties.For example, Manzanilla et al. (2013) show that

Table A2
Parameter estimates-test on the validity of selection estimates.female farmers are as knowledgeable as male farmers in evaluating the lines/variety of visible characteristics.
To incorporate women's participation in PVS strategies for submergence tolerant varieties in Southeast Asia, researchers involved the participating households' wives by selecting only a sub-sample of the farmer participants (Paris et al., 2011).In India, where most are male-headed households, exploring the individual and the couple's preferences when selecting new variety lines is warranted.Thus, international research institutes like the International Rice Research Institute (IRRI) should include the preferred grain quality traits in the breeding programs, ultimately benefiting resource-poor smallholders.
Second, there is a high yield impact among joint decision-making strategies when adopting local varieties.Until now, there is a large proportion of farmers that still adopt local varieties despite an increased number of MRV releases in India.This paper defines local varieties as rice varieties that do not have seed certification from the government.It may be the case that the reported local varieties are unrecognizable among farmers due to their continuous use and poor seed systems that affect the rice variety's authenticity.Tsusaka et al. (2015) found that India's average varietal replacement age ranges from 18 to 19 years, meaning that farmers will continuously use the same rice variety that usually comes from their previous harvest.To assess the identity of the local varieties, DNA fingerprinting should be explored for more reliable and accurate identification of variety types.A study by Kretzschmar et al. (2018) found that using DNA fingerprinting on the farmer-reported rice varieties in Bangladesh shows that the farmer-reported local rice varieties come from two predominant matches: BR22 (Bangladesh rice variety) and Horidhan (Indian rice variety).Proper identification of these farmer-reported rice varieties will help rice breeders and policymakers accurately assess the impact of rice variety on rice productivity in India.
Third, targeting a woman's self-help group (WSHG) is one of the most natural pathways to reaching spouses.WSHGs serve as channels for disseminating information, particularly in areas hard for extension workers to enter and marginalized groups are majorly located.Since MRVs are composed of specific agronomical characteristics (e.g., potential yield, grain size, resistance to pests and diseases), information can be disseminated through farmer's field schools or demonstration plots.The demonstration trials would enable women particularly in the marginalized sector to be exposed to new rice varieties, labor-saving technologies, and proper farm management practices, leading to higher adoption rates.
The current study provided evidence from a large sample of Indian rice farmers.But, the study could be used by policymakers in other developing countries in South and Southeast Asia that are major rice producers.These countries have a similar profile regarding smallholder agriculture, rice production, and livelihood strategies.This study examined the impact of joint decision-making on rice yields, but it has a caveat.The study used cross-sectional data for one rice season, suggesting that the findings only apply on a short-run basis and should be interpreted accordingly.One needs to investigate this issue with panel data to capture long-term adoption impacts.The degree of the spouse's  4 Includes castes that are socially and educationally discriminated. 5At least one of the household members has off-farm labor like a salaried job (government or service industry) and small business.

Table A5
Results of Probit model.*** p < 0.01, ** p < 0.05, *0.10.a Standardized percentage balancemean of the treated (joint decision-making) minus the mean of the control (male decision-making) divided by the average of the square root of the variance of the treated and control samples.Matching is sufficient if standardized bias is below 3% or 5%.b Lower Pseudo R 2 after the matching, and there is no systematic difference.

Table A7
Means before and after matching., where 0 and 1 are standard deviations of estimated propensity scores of the control and the treatment groups, respectively (see Cochran and Rubin, 1973).

Table 1
Sample districts and smallholder households in eastern India, 2016.

Table 2
Summary statistics of the variables used in the rice variety selection model, Eastern India, 2016.
(continued on next page) M.L.L.Malabayabas et al.

Table 3
Average treatment effect on treated/untreated and heterogeneity effects for rice yield (kg/ha), by rice variety.

Table 4
Average treatment effect on treated/untreated and heterogeneity effects for rice yield (kg/ha), by caste.

Table A1
Variable definition used in the analysis.
This group is considered to have better grain qualities, particularly the one released in 1986-1995.In the later period, the varieties are designed for adverse environments.The popular varieties in this group are the following: MTU-1010

Table A2
Husband and spouse are making farming-related decisions jointly.2Therespondentsare the household head who are knowledgeable in rice farming.In this study, all the household heads are husbands.3Includesdesignatedgroups of historically disadvantaged indigenous people in India.The terms are recognized in the Constitution of India (GoI), and the various groups are designated in one of the categories.Since independence, the scheduled castes and tribes were given Reservation status, guaranteeing political representation.4Includescastesthataresociallyandeducationally discriminated.5Atleastone of the household members has off-farm labor like a salaried job (government or service industry) and small business.6Share of productive assets solely owned by women.7Atleast one member is a migrant. 1

Table A4
Parameter estimates of couples' decision-making strategies and yield (kg/ha) log, endogenous switching regression (ESR) approach.

Table A4 (
continued ) Standard deviations in parentheses: *** p < 0.01, ** p < 0.05, * p < 0.1.1Husbandandspouseare making farming-related decisions jointly.2Therespondentsare the household head who are knowledgeable in rice farming.In this study, all the household heads are husbands.3Includes designated groups of historically disadvantaged indigenous people in India.The terms are recognized in the Constitution of India (GoI), and the various groups are designated in one of the categories.Since independence, the scheduled castes and tribes were given Reservation status, guaranteeing political representation.

Table A6 Standardized
Percentage Balance, Pseudo R2, and Final Sample Size Using Different Matching.