Livelihood security among rural poor: Evaluating the impact of Rural Livelihood Mission in Odisha, India

Livelihood insecurity remains a prime concern for low household income countries. To provide secured livelihood to rural poor, the government of India has introduced a self-employment type poverty alleviation programme namely National Rural Livelihood Mission (NRLM). This paper empirically examines the effect of participation in NRLM on the livelihood security of rural poor. Data were collected from 220 respondents (including both beneficiaries and non-beneficiaries) through a structured questionnaire from Sonepur district of Odisha (India). A livelihood security index (LSI) was constructed to capture the livelihood security of the respondents taking habitat security, health security, food security, and economic security into account. Further, the impact of the programme has been estimated using propensity score matching (PSM) method. The study finds a positive and significant effect of participation in the programme on livelihood security. Therefore, poor should be encouraged to participate in the programme to strengthen their livelihood security. Subjects: Economics and Development; Economics; Development Economics


PUBLIC INTEREST STATEMENT
Poverty alleviation has been a prime concern for the Government of India (GoI) since independence. Therefore, various wage employment, self-employment, and food security type poverty alleviation programmes have been introduced by GoI across the time. In the year 2013, the GoI introduced a self-employment type poverty alleviation programme namely National Rural Livelihood Mission (NRLM) replacing previous Swarnajayanti Gram Swarozgar Yojana (SGSY) programme with the objective of providing secured and sustainable livelihood to the rural poor. To achieve the objective, various financial and livelihood services are provided to the beneficiaries under the programme. This study empirically examines the effectiveness of the programme in providing secured livelihood to the rural poor through a field survey conducted in one of the backward districts of Odisha, India. The outcome of the study revealed that participation in the programme helps the beneficiaries in achieving secured livelihood. The findings of the study have implications for the researchers and policy makers.

Introduction
Households' livelihood security remains a pertinent issue in the economic development of low household income communities (Bhandari & Grant, 2007). It has been defined as adequate and sustainable access to the income and resources required to meet basic needs.
The basic need includes adequate access to food, health facilities, educational opportunities, housing, community participation and social integration (Frankenberger & McCaston, 1998). India has emerged as one of the fast growing economies (Dreze & Sen, 2014). Despite this growth, insufficiencies in terms of income, education, health care are also visible (Kasthuri, 2018;Patidar, 2019). For instance, high level of malnutrition and hunger is prominent in the country; as around 60 million of undernourished from the total of 178 million (in world) reside in India (Yadav et al., 2016). Moreover, in terms of per capita income and poverty head count ratio India's performance is poorer than other countries. Therefore, to meet the above insufficiencies various wage employment and self-employment programmes like Integrated Rural Development Programme (IRDP), Swarnajayanti Gram Swarozgar Yojana (SGSY), Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA) have been introduced by the Government of India across time (Mishra, 2017a). The success rates of these programmes were low as a large group of vulnerable did not get benefits from the programmes (Breitkreuz et al., 2017;Mishra, 2017a). The effectiveness of poverty alleviation programmes will be high if the cause of poverty is rightly identified and the programmes are introduced in accordance with that (Bradshaw, 2007). Thus, it is imperative to identify the cause before the implementation of the programme.
In India, poverty is caused by interdependencies of multiple factors (Mishra, 2017b) and thus the poverty alleviation programmes should be implemented with the objective of providing secured and sustainable livelihood (Chamber, 1995;Krantz, 2001) through i) providing varieties of services, ii) collaborating with other organisations and iii) organising the communities adopting the cycle of poverty perspective (Bradshaw, 2007;Mishra, 2017b). Further, to enhance households' livelihood security, three approaches can be taken: a) livelihood promotion, b) livelihood protection, and c) livelihood provisioning (Das, 2016). These strategies, whether applied alone or in combination, help to enhance households' livelihood security (Frankenberger & McCaston, 1998).
Recently, with the objective of providing secured livelihood to rural poor, the government has introduced a self-employment type poverty alleviation programme (for the promotion of livelihood) namely National Rural Livelihood Mission (NRLM). The programme is introduced by restructuring the previous Swarnajayanti Gram Swarozgar Yojana (SGSY) programme. NRLM is a centrally sponsored scheme with a three tier structure-national, state, and district level. At the national level, Ministry of Rural Development takes the responsibility. In the state level, Ministry of Panchayati raj undertakes the responsibility and at district level, District Rural Development Agency handles the functionality of the programme. The financing of the programme is shared between the centre and state. The ratio of sharing between the centre and the state is 75:25 in all the Indian states except North Eastern states, where it is shared in the ratio of 90:10.
NRLM ensures that at least one member from the identified rural poor household is brought under the self-help group (SHG) network in a time bound manner. This clarifies that NRLM tries to address poverty through SHGs building strong institutions of poor. The programme further provides financial and livelihood services to the beneficiaries. The financial services are provided to the beneficiaries in the form of community investment fund, bank linkage and capital received under interest subvention. In the form of livelihood services, the programme imparts short duration learning (self-employment) programmes to the beneficiaries. Moreover, to make youths confident and self-employed, the programme collaborates (in partnership) with other organisation for trainings and post-training follow-up (Mishra, 2017b). The above description clearly depicts that NRLM tries to alleviate poverty following cyclical perspective in terms of providing a variety of services, collaborating with other organisations and organising communities.
The extent of benefits derived from a programme can be known after impact evaluation as it addresses questions related to the causal effect of an intervention on an outcome (Mishra & Das, 2017). In the context of impact evaluation, Mohapatra and Sahoo (2016) report that the impact of a programme is better assessed if the study is conducted in a backward region [1] . On the contrary, a majority of the studies have been conducted in examining the impact of poverty alleviation programmes in developed states and the literature on backward region is scant. With this background, this study chooses Sonepur district of Odisha (India) as an ideal selection for the study (to bridge the gap) due to the following reasons. First, Sonepur is an indispensable core area of Kalahandi Balangir Koraput (KBK) region. The KBK region has been found to be the most underdeveloped and poverty-driven belt in India (Dash, 2007), which signifies the backwardness of this region. Second, the share of Gross District Domestic Product (GDDP) to Gross State Domestic Product (1.01% based on the Odisha Economic Survey of 2014-2015) of Sonepur is minimal in comparison to other districts of KBK region. The above scenario speaks about the backwardness of Sonepur that justifies the selection to be an ideal one.
The present study is carried out with the objective to examine the impact of NRLM on livelihood security of rural poor. The study constructs an index of livelihood security considering health security, habitat security, food security, and economic security. The study using propensity score matching method finds a positive impact of NRLM on livelihood security. This study empirically examining the impact of NRLM on livelihood security in a backward region overcomes the limitations of past literature (Kurian, 1987;Paul, 1998;Ray, 2008;Panda et al., 2012), which were limited to the impact assessments of self-employment programmes (in developed states) on poverty alleviation in terms of income generation and agricultural production which are few dimensions of livelihood security. Thus, it is believed to be an endeavour to enrich the existing literature as to our knowledge this study is the first ever empirical study on examining the impact of NRLM on livelihood security specifically in a backward state region. The remainder of the paper is as follows. Section 2 provides the theoretical framework and hypothesis formulation. Section 3 portrays the data sources and methodology used in the study. Section 4 explains about the impact of NRLM on livelihood security and section 5 concludes.

Theoretical framework and hypothesis formulation
Poverty alleviation programmes with SHG-linkage model is gaining popularity as participation in SHG empowers the beneficiaries (Desai & Joshi, 2012;Deininger & Liu, 2013;Mohapatra & Sahoo, 2016;Kumar et al., 2019;Nayak & Panigrahi, 2020). Further, the long-term exposure of the beneficiaries to the programme through SHG positively impacts their consumption, nutritional intake, and asset accumulation (or wealth accumulation) (Deininger & Liu, 2009), which ensures their economic and food security. Moreover, involvement in SHG networks increases the social capital (a resource derived from people's social participation) of members, which enhances the awareness regarding basic necessities including education (a component of educational security), sanitation, and drinking water (which are subcomponents of health security). Thus, it is clear that participation in SHGs can ensure educational security and health security to the rural poor through the creation of social capital. Therefore, broadly, five dimensions of livelihood security, i.e., empowerment, health security, economic security, educational security, and food security can be achieved through the participation in SHG. Earlier it was mentioned that NRLM tries to alleviate poverty by bringing beneficiaries into SHG network. Thus, it is articulated from the above literature that through the participation in SHG, NRLM can meet the five dimensions of livelihood security.
Availability of capital in the right quantity at an affordable cost has the potential to reduce the high incidence of poverty and contribute to household well-being (Mohapatra & Sahoo, 2016). NRLM provides financial assistance to the beneficiaries in the form of capital and bank linkages. Along with the financial assistance, the programme also provides self-employment trainings to the beneficiaries. The duo effect of training and capital can help the beneficiaries in either starting their own business or making investment in agriculture, which may increase their income level and can provide economic security. The increased level of income will permit families to consume quality food leading to a state of food security. Economic security coupled with social capital will enhance their social awareness and due to which they will look for better accommodation (house), which will ensure their habitat security. In rural India, the educational attainment of children is significantly affected by their parents' financial constraints. The increased level of income after participation in the programme can help parents educating their children and thus can ensure educational security. Taking a cue from the above literature, we hypothesise that livelihood security (considering the dimensions of educational security, economic security, health security, empowerment, educational security & food security) of people can be enhanced through the participation in NRLM. This theoritical framework is depicted in Figure 1.

Data and methodology
This study is based on both primary and secondary data. Due to dearth of secondary data on the functioning of NRLM to assess the impact, the study relies on primary data. To collect the primary data, a field survey was conducted at the household level using a structured questionnaire. A three-stage probability (and non-probability) sampling was followed to collect data, with blocks as the primary units, Gram-panchayats and villages as the secondary units, and beneficiaries and non-beneficiaries [2] as the ultimate sampling units.
Under the NRLM, two types of blocks were found in Sonepur district, intensive and non-intensive [4] . In the case of intensive blocks, all components of the programme work intensively, whereas in the non-intensive blocks all components do not operate to that extent. To obtain an unbiased outcome, data were collected from two blocks, one intensive and one non-intensive. In the year 2015 (when the study was undertaken), Tarbha was identified as the only intensive block in Sonepur district. Therefore, based on the judgemental sampling this intensive block (Tarbha) was chosen. Sonepur block (a non-intensive block for NRLM) was chosen from the rest five blocks at random. In six different Gram Panchayats of Tarbha block, intensive works were underway. Out of those, five panchayats were chosen purposively for this study. Further, one village from each intensive panchayat was chosen at random. In parallel, five villages from five different panchayats of Sonepur block (non-intensive) were also chosen at random. Lastly, a total of 220 respondents (120 beneficiaries and 100 non-beneficiaries) were interviewed.  Based on Soper's (2020) sample size calculator for multiple regressions with anticipated effect size of 0.15, desired probability level of 0.005 and desired statistical power of 0.8 for seven (in eq n 12) and two (in eq n 11) independent variables were found to be 103 and 67, respectively. Hence, sample size of 120 beneficiaries is adequate to achieve the satisfactory effect size. In the present study, it was imperative to include non-beneficiaries as they act as a counterfactual [3] . Moreover, in the use of propensity score matching method, it is evident from the past literature that the percentage of non-beneficiaries in the total sample size in comparison to the beneficiaries varies from 30% to 90% (Gebrehiwot & Van der veen, 2015; Mohapatra & Sahoo, 2016;Satapathy et al., 2020). This justifies the inclusion of 100 non-beneficiaries (which constitutes 83% of beneficiaries) in the study. Non-beneficiaries were selected from the same village or nearby villages at random from the below-poverty-line (BPL) lists that were collected from the panchayat office.

Socio-economic profile of sample households
In the context of social stratification, caste has a significant importance (Mohapatra & Sahoo, 2016); it has been found to be one of the key markers of socio-economic status in rural India (Rao & Ban, 2007). Scheduled Castes (SCs) and Scheduled Tribes (STs) have largely remained outside the mainstream of modern Indian society. In the present study, most beneficiaries belonged to the SC category (85%), followed by Other Backward Classes (OBC, 12%), and general (3%). Approximately the same distribution pattern is found among non-participants.
In addition to caste, occupation, household income, and land holdings are important indicators of economic status. Occupation denotes the economic condition of a household, which in turn describes their social status. Casual laborer was the primary occupation for a majority of respondents (72%), followed by agricultural laborer (20%), and self-employed (8%). The average annual income of beneficiaries was Rs.38,300 ($538.02), while the average annual income of non-beneficiaries was Rs. 35,200 ($494.48). Land is a primary asset in rural areas, as it is used not only for agricultural production but also as a security against shocks by selling those assets in the market (Mohapatra & Sahoo, 2016). The study revealed that 45% of beneficiaries and 69% of non-beneficiaries do not have cultivated land. The above discussion is summarised in Table 1.

Measure
A livelihood security index (LSI) was constructed to capture the livelihood security (understood as outcomes of the NRLM) of the respondents, as LSI helps in providing a clearer profile of livelihood security at the community level (Lindenberg, 2002). Furthermore, it provides information regarding all the dimensions in a single number (Sarma, 2008).

Determining variables and indicators
The measurement of livelihood security requires considering the best variables (or indicators) to capture it. CARE, the internationally recognised non-governmental organisation, groups the subcomponents of livelihood security into economic security, food security, health security, educational security, and empowerment. Due to variations in people's bio-physical and socio-economic conditions, the indicators for studying livelihood security should be location specific and be constructed based on the socio-economic particulars of the location under examination (Bhandari & Grant, 2007).
In the present study, a majority of both beneficiaries and non-beneficiaries are poor with low household income affecting not only their consumption but also the education of their children. Consumption of quality food is essential in maintaining a balanced life failing which will affect the health (Mishra et al., 2020). This economic scenario entails to include food security, economic security, health security, and educational security. On the contrary, the literacy rate in study area was found to be better than the state's literacy rate. Therefore, the study drops educational security as one dimension of livelihood security. Thus, the study proposes to include economic security, health security, empowerment, and food security as the dimensions of livelihood security. Along with the above, based on the results of the pilot study [5] habitat security was taken as an additional variable and dropped empowerment as one dimension. The study follows Mishra (2017b) and considers the sub-indicators to measure the habitat security. Details on indicators and sub-indicators undertaken to construct the LSI are presented in Figure 2.
After the identification of indicators and sub-indicators, indices of individual dimensions were computed using the equations presented below.
Where, d i is the index of i th dimension and A i , M i , and m i are the actual value, maximum value, and minimum value in i th dimension. The overall index of livelihood security is computed as Where, LSI i represents the overall livelihood security index of i th individual and W i stands for the weight allotted to each dimension. Here, equal weights are given to each dimension, so The present study gives equal weights to all the dimensions following Samanta and Nayak (2015).

Livelihood security amongst sample households
After calculating the index of livelihood security, the difference of livelihood security among beneficiaries and non-beneficiaries was assessed. The livelihood security index has been categorised into three types -low, medium, and high. When we look at medium level livelihood security index, the percentage of beneficiaries is more than non-beneficiaries. But, in the lowlevel index, the percentage of non-beneficiaries is more than beneficiaries. High level of livelihood security index is found only among beneficiaries. The result is presented in Table 2.

Selection of method
The evaluation of a policy is based on two theories, the structural approach and the treatment effect approach. The former is applicable where there is universal participation, while the treatment effect approach is applicable where there are two groups: (a) a treatment group that takes part in the programme and (b) a comparison group that does not participate in the programme (Heckman & Vytlacil, 2005;Mishra & Das, 2017). Evaluation methods in empirical economics fall into five broad categories, each of which provides an alternative approach for constructing the counterfactual and helps to minimise selection bias Ravallion (2008) advocates that studies based on a randomised selection method mitigate the selection bias problem to the greatest possible extent. Randomised selection is applicable when the programme is a large one and needs to be implemented in a phasewise manner (Gertler. et al., 2007). This method also ensures that each individual has an equal opportunity to take part in the programme. The NRLM is a flagship programme that has been implemented in a phase-wise fashion. Thus, randomisation method can be suitable in examining the impact of NRLM. On the contrary, participation in the NRLM is not random, and when the participation is not random, there is a possibility of self-selection bias, which can be addressed by creating a suitable counterfactual. This can be done by matching each programme beneficiary to a non-beneficiary with similar covariates (Mohapatra & Sahoo, 2016). Further, matching method provides similar result like randomisation method (Dehejia & Wahba, 1999). Thus, this study uses propensity score matching (PSM) as an alternative to randomisation method in examining the impact of the NRLM on livelihood security.

Mathematical interpretation of methodology
The mathematical methodology discussed below is borrowed from Rosenbaum and Rubin (1983), Cameron and Trivedi (2005) and Mohapatra and Sahoo (2016). In the present study, let T represent participation in the programme. T = 1 indicates a particular respondent participates in the programme (treatment/beneficiary), otherwise T = 0 (control/non-beneficiary). 'LS' denotes the livelihood security. LS 1 is the livelihood security of the beneficiaries and LS 0 is the livelihood security of non-beneficiaries. Total sample size N is 220, of which beneficiaries (N b ) are 120 and non-beneficiaries (N nb ) are 100. X is the vector of covariates and P(x) is the propensity score.
The average impact of treatment (AIT) on the livelihood security is expressed as: and the estimated sample AIT is: The average impact of the treatment on the population may provide a biased result; therefore, there is a need of finding out the average impact of the treatment on the treated or beneficiaries (AITB). This is expressed as: and the estimated AITB on the sample in this study is: However, the second term of equation 5, i.e., E LS 0 =T ¼ 1 ð Þ is not observable. On the other hand, which is due to selection bias. Therefore, the observed outcome given the participation is: In the present programme, participation is not random. This leads to the presence of self-selection bias, for which the estimation of the average impact of treatment on beneficiaries (AITB) and AIT are not feasible. This is feasible under conditional independence assumption of PSM. The conditional independence assumption ensures that the selection is solely based on a set of observable covariates (X) (Wooldridge, 2002). Symbolically, LS 0 ?T=X, where ? denotes independence. The common support or overlap condition ensures that the person with the same X values have a positive probability of being beneficiary and non-beneficiary (Heckman, LaLonde, &Smith, 1999).
It may be noted that under conditional independence assumption and set of observable covariates (X), both AITB and AIT are identical (Wooldridge, 2002;Mohapatra & Sahoo, 2016). Thus, estimation of AITB(X) will be sufficient for examining the effect of participation on livelihood security. The equation is presented below.
Furthermore, Rosenbaum and Rubin (1983) have proved that if conditional independence assumption holds, then replacing conditional on propensity score over conditional on observable covariates will also hold. Symbolically, (Overlap) 0 < P (T = 1=X) < 1.
Hence, based on the conditional independence assumption and overlap condition, the average impact of treatment of participation is: and the estimated AITB for the sample is: In equation 10, LS 1;N À LS 0;N PSM � � is the difference between observed livelihood security of the beneficiaries and matched (PS). Following Mohapatra and Sahoo (2016), different degrees of bandwidths are also used to check the sensitivity or the presence of hidden bias in the estimate.
Further, when the outcome variable (in this case livelihood security) is continuous and treatment is a discrete variable (NRLM participation), the impact of the treatment can be evaluated by adding the estimated propensity score P (X) in the ordinary least square (OLS) estimation (Wooldridge, 2002;Mohapatra & Sahoo, 2016). Symbolically, the regression function can be represented as The estimated impact is said to be positive when the value of P in eq n 11 is positive and statistically significant.

Results
The results of the present study are obtained following the steps mentioned in Caliendo and Kopeinig (2008) for using PSM. Based on those methods, first variables are identified based on hit and trial and statistical significance method to run a binary probit model for finding out the propensity score.
The estimable regression equation is In equation 12, A represents the age of the respondents, DR represents the dependency ratio (calculated by taking the total number of earning members and total number of members into account), and E, AH, SOC, and PO represent electricity facility, asset holdings, source of cooking, and primary occupation, respectively. Dummy variables were used for electricity (yes = 1, 0 = otherwise), livestock holding (yes = 1, 0 = otherwise), and landholding (yes = 1, 0 = otherwise). In the present study, most of the respondents reported their primary occupation as casual laborer. Thus, a dummy variable was also created for primary occupation, with 1 assigned for casual laborers and 0 otherwise. Further, the present study follows Mohapatra and Sahoo (2016) and adds source of cooking as an instrumental variable. The result is presented in Table 3. Table 3 shows that source of cooking and primary occupation are the significant factors in influencing the decision to participate in the programme.
After the estimation of propensity score, the common support region was identified and was found to be [0.28, 0.92]. This implies that propensity scores below 0.28 and greater than 0.92 were eliminated based on the maxima and minima approach. This elimination results in the loss of seven respondents (both beneficiaries and non-beneficiaries). The balancing property test is carried out to assess the matching quality using pseudo R 2 and the likelihood ratio test (LR chi 2 ). The matching quality is said to be good when the pseudo R 2 value decreases (Sianesi, 2004) and likelihood ratio test becomes insignificant (Smith & Todd, 2005). In the present study, the pseudo R 2 value decreases from the raw to the matched sample and that the likelihood ratio test becomes insignificant. This signifies that the matching performance appears to be good (Table 4).
In the next step, after determining the propensity score, a matching algorithm was used to estimate the impact. The present study used kernel matching as a matching estimator because it provides a better estimate in comparison to the nearest neighbor and interval matching (Mohapatra & Sahoo, 2016). The robustness result mentioned in Table 4 shows the coefficient of participation in NRLM is positive (0.16) and also statistically significant. Therefore, based on eq n (11), it is inferred that participation in NRLM is found to have positive and significant effect on livelihood security. The results obtained from the study are also found to be robust (Table 5).

Discussion
The present study was carried out with the objective of examining the impact of NRLM on livelihood security. In the process of examining the impact, the study examined the determinants which affect the participation decision using probit model and found that respondents' whose primary occupation was casual labourer and who were using firewood and chips as their source of cooking had a greater possibility of participation in the programme. Moreover, though source of cooking was taken as an instrumental variable, the study finds that households with the occupation of causal labourer were using firewood and chips as their source of cooking due to their short-term engagement resulting in lower income. This situation indicates their level of vulnerability. That means, poor people are more interested to participate in the programme. From this outcome, it can be inferred that correct identification of poor for the programme was done in Sonepur district of Odisha, India. Furthermore, the present study reveals a positive impact of NRLM on livelihood security. That means, participation helps the beneficiaries in securing their economic security, food security, health security and habitat security. The outcome of the present study is in line with the past literature (Kurian, 1987;Derze, 1990;Paul, 1998;Ray, 2008;Panda et al., 2012) in terms of providing positive impacts on the outcome variables [6] . Thus, to meet the sustainable development goals, government should bring more vulnerable people into the NRLM network to alleviate poverty, make the nation hunger free, and provide education, and clean sanitation; as NRLM provides livelihood security through which all these goals can be achieved.

Conclusion
In post-independence India, several policies have been implemented to alleviate poverty.
NRLM, a self-employment programme was implemented with the objective of providing secured livelihood to rural poor. Using primary data, this study attempted to examine the impact of NRLM on livelihood security. The present study captures the livelihood security through an index, livelihood security index taking habitat security, health security, food security and economic security into account.
Participation in NRLM is voluntary or self-selecting. In that case, the socio-economic factors that determine the participation decision were studied. A binary probit model was used to examine the participation decision. The result showed that occupation and source of cooking had significant effect on participation. Additionally, the impact of NRLM was examined using propensity score matching method and found a significant positive impact of NRLM on livelihood security.
Irrespective of this positive impact, the findings of the study are not free from limitations. First, it relies on a purposive sample with a relatively small sample size. Second, for the computation of food security, data on consumption of foods were collected from respondents on a recall basis, which may entail inaccuracies. Notwithstanding any limitations, this study provides solid evidence that the NRLM positively affects the livelihood security of the rural poor in the study area. Irrespective of the positive impact of the programme, Mishra (2017b) identifies that the restriction clause imposed in the programme in terms of age, hindrance the programme in achieving its outcome. Therefore, in line with Mishra (2017b), this study too provides suggestion to policy makers in relaxing the rigidity clause imposed in terms of age for providing secured livelihood to rural poor to a great extent. This is one of the early studies on examining the impact of NRLM on livelihood security in a backward region. The study documents a positive and significant effect of people participation in the programme on their livelihood security. This could be helpful to policy makers to design suitable policy through which the participation in the programme will increase, which in turn, will enhance livelihood security. [4] The government requires huge amount of funds and manpower to implement theprogramme intensely across all districts of the state. For that reason, the blocks are segregated as intensive and non-intensive. Further, in due course of time government is bringing more blocks into intensive category.

Author details
[5] A pilot study was conducted taking the interviews of 50 respondents. Thereliability and the validity tests were conducted taking the pilot data into consideration. From the pilot study lack of housing was found among the respondents. Therefore, habitat security was considered as an additional dimension to measure livelihood security. Further, in consistent with Mohapatra and Sahoo (2016), the study had prepared the questionnaire on women empowerment considering autonomy, economic empowerment and gender empowerment and had collected data in the pilot study. It was found from the pilot study that all the respondents were empowered in the dimension of autonomy, 83% were empowered in economic dimension, 65% were empowered in gender dimension and 80% were empowered in both autonomy and economic dimensions. Overall 82% respondents were empowered considering all the three dimensions. Therefore, based on the above result, the present study dropped the empowerment dimension for measuring livelihood security.
[6] The outcome variables considered in the past literature and the outcome variable of the present study are different. The past literatures were confined to the impact assessments of self-employment programmes in terms of income generation and agricultural production, which are few dimensions of livelihood security. The present study considering a broader concept overcomes the limitations of past literature.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
The authors received no direct funding for this research.