Why Market Orientation Matters for Agriculture and Fishery Workers? Unravelling the Association between Occupational Background and Caloric Deprivation in India

Background: Given the presumptions of low-productivity, developmental policies in low- and middle-income countries usually display a pro-manufacturing bias and largely undermine the potentials within the agriculture sectors in promoting sustainable growth and development. As a consequence, skill development is usually aligned with preferences of the manufacturing-led economic growth and is accorded high priorities in policymaking. However, the success of such skill development policies essentially hinges upon the assumption of a quick and successful structural transformation towards non-agricultural sector (manufacturing and services). But contrary to expectations, the pace of economic transition has been rather slow in India and almost one-half of the workforces continue to be engaged in agriculture and allied activities. Moreover, it is unclear whether such shift away from agriculture can necessarily lead to reductions in widespread deprivations in the form of food insecurity and under nutrition. In fact, it is also feasible that market orientation of skilled agricultural and shery sector may display favorable impact on poverty and nutritional well-being of households. Methods: Therefore, drawing upon data from 68 th round (2011-12) of nationally representative cross-sectional Household Consumer Expenditure Survey (HCES) of National Sample Survey (NSS), Government of India, this study aims to examine the association between occupational backgrounds and nutritional deprivation (average caloric consumption as well as low calorie intake) among Indian households. Results: Evidences show that agricultural and shery labor households have lowest calorie intake (2086 kcal) across all the occupational groups. However, market oriented skilled agricultural and shery workers’ (2261 kcal – rural, 2165 kcal - urban) have higher calorie intakes than those belonging to subsistence agricultural (2165 kcal – rural, 2149 kcal - urban) and shery workers or agricultural and shery laborer (2086 kcal – rural, 2071 kcal - urban). Further, the multilevel logistic regression estimates suggest that in rural areas, households engaged in skilled agricultural and shery works have signicantly (at 5% level) lower odds ratio (OR: 0.72, with 95% CI: 0.63; 0.82) of having insucient calorie intake compared to the unskilled agricultural and shery laborer households. Conclusion: These insights when combined with the occupation-specic random-effects suggests that policy focus to promote market oriented skilled agricultural and shery workers can be an equally good option as direct investment in professional and managerial skills for manufacturing and services.


Introduction
Given the presumptions of low-productivity, developmental policies in low-and middle-income countries usually display a pro-manufacturing bias and largely undermine the potentials within the agriculture and shery sectors in promoting sustainable growth and development [1][2][3][4][5]. As a consequence, skill development is usually aligned with preferences of the manufacturing-led economic growth and is accorded high priorities in policymaking [6,7]. For instance, in India, the thrust on such approach is evident from the creation of a separate ministry for skill development and entrepreneurship (MSDE) that is concerned with policies to impart employable skills to the working-age (15-59 years) population that accounts for about two-third of the total population. It is expected that skill development can enhance employability quotient of the labor force and thus facilitate rapid reductions in poverty and deprivation.
However, the success of such skill development policies essentially hinges upon the assumption of a quick and successful structural transformation towards non-agricultural sector (manufacturing and services). But contrary to expectations, the pace of economic transition has been rather slow in India and almost one-half of the workforces continue to be engaged in agriculture and allied activities [8]. Moreover, it is unclear whether such shift away from agriculture can necessarily lead to reductions in widespread deprivations in the form of food insecurity and under nutrition. For instance, despite rapid economic growth in recent years, India is poorly ranked 97th of 118 countries in the Global Hunger Index 2016 [9]. In fact, sustained declines in nutritional intake (calories as well as other nutrients) are identi ed as a major developmental and food security concern [10][11][12].
Given the context, it is critical to unravel the association between occupational background and nutritional deprivations and thereby develop insights regarding the scope and focus of skill development policies. For instance, it is noted that poverty and nutritional deprivations are generally concentrated among households belonging to unskilled occupational categories such as agricultural labor, casual labor or shery workers [13][14][15][16][17][18]. But it is unclear whether only selected sections or the entire agricultural and shery sector is vulnerable to such risks. Also, there is no evidence to understand the relative advantages of skilled agricultural and shery sector vis-à-vis other occupations. In fact, it is also feasible that market orientation of skilled agricultural and shery sector may display favorable impact on poverty and nutritional well-being of households [19][20][21][22][23][24]. Given such possibilities, this paper aims to examine the association between occupational backgrounds with nutritional deprivation (average caloric consumption as well as low calorie intake) among Indian households. As such, a focus on caloric intake is critical because it is a fundamental indicator of food security and has a major in uence on economic well-being [25,11]. Moreover, adequate calorie intake has high relevance for nutritional well-being and is regarded as a fundamental marker of public health [10,26]. Given such relevance, an analysis of average per capita calorie intake can highlight the patterns of nutritional deprivation across various occupational groups and can effectively outline the differences and similarities therein. Thus, these results can also provide vital insights to decide upon policy approach towards skill development and diversi cation. In particular, this can help to comprehend whether promotion of market oriented skills among the unskilled agricultural and shery households can have signi cant in uence on nutritional and food security. The relevance of these ndings increases manifold because of its wider implications for other developing countries. In fact, similar patterns of income and nutritional deprivation are observed across other lowand middle-income countries and reiterate the need and relevance of a comparative analysis of nutritional well-being across various occupational groups [27,15,3,[28][29].

Data And Methods
This study is based on data from nationally representative cross-sectional Household Consumer Expenditure Survey (HCES) conducted in 2011-12 (68th round) by National Sample Survey O ce (NSSO), Ministry of Statistics and Programme Implementation (MOSPI), Government of India (NSSO 2014). The HCES is widely used by developmental practitioners and policymakers to assess the levels and patterns of food consumption across various population subgroups in India. The HCES uses a strati ed, multistage cluster design at state-level to provide reliable estimates at state and for rural and urban areas.
Households for survey are selected on the basis of circular systematic sampling. The results are estimated using data from HCES schedule 1.0 (Type 1/ Mixed Reference Period) which altogether has a sample of 1, 01,662 households (59695 rural + 41967 urban).
The HCES provides occupational categories of households based on the code structure provided in the National Classi cation of Occupations, 2004 [30]. Based on this coding structure, the households are categorized into 27 mutually exclusive occupational categories (Table 1). Based on a mixed recall period of 30 days and 365 days, the HCES provides direct information on household food and non-food consumption. In this paper, the information on food consumption is used to estimate the average per capita per day per caloric intake (in kilocalories, kcal) across households. The estimation is based on intake conversion parameter prescribed by the NSSO and which is derived from a nutrition chart that provides details regarding energy content of different foods in the Indian diet [31]. For certain food items the intake quantity is unavailable but has been replaced by information on average energy contents per Indian rupee. It is worth noting that the consumption details are available at the household level and thus cannot be speci cally associated with caloric intake of each household member. Notwithstanding this limitation, the household per capita per day calorie intake has been an important indicator to examine levels, trends and patterns in nutritional intake in India. Further, we de ne prevalence of under nutrition or low-calorie intake as the situation when households are estimated to be consuming less than 80% of the prescribed calorie norms (2400 kcal in rural areas and 2100 kcal in urban areas). In other words, households with per capita per day consumption of less than 1920 kcal in rural areas and less than 1680 kcal in urban areas are regarded as undernourished household. Use of this 80 percent threshold is motivated by the approach adopted NSSO in its analysis of levels and patterns in nutritional intake in India diet [31]. In fact, the estimates of under nutrition obtained using this benchmark is more or less similar to the proportion of nutritional deprivation estimated using other anthropometric measures such as prevalence of low body mass index among men and women in India [32]. Based on this transformation of caloric information we arrive at two outcome indicators for the analysis: average per capita per day caloric intake across households (continuous outcome) and prevalence of low-caloric intake or under nutrition across households (binary outcome). The analysis is conducted separately for rural and urban areas as they have different average calorie intake norms. It may be noted that the analysis is based on a sample of 94,157 households (56536rural, 37621 -urban) after ltering out observations where household NCO codes or other correlates are missing or not speci ed. Following a bivariate analysis, we use multilevel linear and logistic regression models to understand the association of occupational groups with continuous and binary outcomes, respectively. The analysis is adjusted for sampling weights as prescribed by the NSSO. For brevity, the beta coe cients and standard errors (SE) for the fully adjusted linear regression models and odds ratios and con dence interval (at 95%) for logistic regression model are reported. We also estimate the variance partition coe cient for both set of regressions to highlight the between-occupational group differences in calorie deprivations (Browne et al., 2005, Goldstein et al 2002. Using the estimated variance of random effects, the variance partition coe cients (VPCs) at each level for the respective models (variation in calorie intake or variation in the log odds of receiving insu cient caloric intake) is computed. The VPC for the concerned level is calculated by dividing the estimated variance at the concerned level by the total variance. While calculating the total variance in the logistic regression, a latent variable methods approach is used whereby the between-household variance is assumed to follow a standard logistic distribution with a value of 3.29 [33,34]. The regression analysis is adjusted for the following indicators of household socioeconomic status: age and sex of household head, household size, and education of household head, religion, social group, wealth quintile, sampling weights and standard errors clustered at the district and state level. The wealth quintiles are constructed using principle component analysis (PCA) on the 20 household durable items from NSS 68 Schedule 1.0. The analysis is performed in Stata 15.0 using runmlwin module [35][36][37].

Results
The average per capita calorie intake is very similar across rural (2172 kcal) and urban (2163 kcal) India (Table 2). In rural areas, households under high level services and managerial professions (particularly science, life science and health professionals) report the highest average per capita calorie intake (in excess of 2300 kcal). The lowest calorie intake (2086 kcal) is estimated for agricultural and shery labor households. In urban areas, the similar group of professionals and managers has highest levels of caloric intake (in excess of 2400 kcal) whereas households belonging to low-end workers and laborers report low intake (below 2100 kcal). It is worth noting that across both rural and urban areas, market oriented skilled agricultural and shery workers' (2261 kcal -rural, 2165 kcal -urban) have higher calorie intakes than those belonging to subsistence agricultural (2165 kcal -rural, 2149 kcal -urban) and shery workers or agricultural and shery laborer (2086 kcal -rural, 2071 kcal -urban). Nevertheless, across both rural and urban areas, the calorie intake has a wider distribution around the mean and can be con rmed by glancing through the box-plots presented in Fig. 1 or at the standard deviations reported in Table 2.
Further, Table 2 also reports the percentage of households with insu cient per capita calorie intake is 33.8% in rural India and 18.5% in urban India. This percentage varied signi cantly across occupational groups. For instance, in rural India the highest level of insu ciencies is noted among workers in precision, handicraft, printing and related trades (45.2% households). Agricultural and shery laborer as well as extraction workers also display higher levels of caloric deprivation (39% households). In urban areas, caloric deprivations are highly concentrated among households engaged in elementary occupation related to sales and services (27% households) whereas legislators, professionals and managers have very low estimated prevalence of caloric deprivation.  (Table 3) indicate that compared to agricultural and shery laborers, households of legislators and senior o cials, life science and health professionals, and market oriented skilled agricultural and shery workers have signi cantly higher average per capita calorie intake. While a number of other service sector professionals depict higher household calorie intake but the differences are not statistically signi cant. Among urban households, a large number of households from service sector background as well as those engaged in market oriented skilled agricultural and shery work report signi cantly higher levels of calorie intake. There is no signi cant difference in calorie intake of low-end occupations and laborers in urban or rural areas. Further in Table 3, we also present the multilevel logistic regression estimates for the association of caloric intake and occupational background while adjusting for demographic and socioeconomic factors such as age and sex of household health, household size, and education of household head, religion, social group, and household wealth quintile. The estimates suggest that in rural areas, only households engaged in market oriented skilled agricultural and shery works have signi cantly (at 5% level) lower odds ratio (OR: 0.72, with 95% CI: 0.63; 0.82) of having insu cient calorie intake compared to the agricultural and shery laborer households. While households with professionals and managers also depict lower odds but the effects are signi cant only at 10% level. In urban areas, a similar relative advantage is discernible for market orientation and skills among agricultural and shery worker households (OR: 0.72, with 95% CI: 0.59; 0.86). The odds of receiving insu cient calorie intake are also much lower for the service sector professionals, particularly life science and health professionals (OR: 0.33, with 95% CI: 0.13; 0.81).
However, households engaged in sales and services based elementary occupations (OR: 1.35, with 95% CI: 1.04; 1.75) are 35% more likely to have insu cient caloric intake compared to the agricultural and shery laborer households.

AOR -Adjusted Odds Ratio
For rural and urban India, Table 4 presents the variance partition coe cients (VPC) for the multilevel linear regression model for average per capita calorie intake and multilevel logistic regression model for households having insu cient caloric intake. The models use ve levels wherein the nesting runs in a hierarchical manner starting from households, occupational groups, districts, region and state of residence. The VPC can reveal the between-group variations attributable at the various levels. In this regard, the null model for average per capita calorie consumption in rural India shows that 10.2% of the total variance in this indicator is attributable to differences in occupational groups whereas state-related differences account for 7.4% of the variation in calorie intake. After adjusting of demographic and socioeconomic factors, the VPC of occupational groups declines to 7.8%. However, in urban areas, a greater proportion of variability in calorie intake is attributable to occupational group related differences (VPC 18.1% null model; VPC 14.8% fully-adjusted model). The geographic boundaries of states and districts have low relevance in explaining variability across urban areas. In fact, region of residence has very low relevance in explaining variations in either outcome across rural or urban India. Further, the VPCs from logistic regression for households having insu cient caloric intake also present similar insights.
Finally, the occupational group-speci c random intercepts from the four respective null models are plotted in Fig. 2. It is inferred that in rural India caloric intake of about two-thirds of the occupational groups cannot be distinguished from the overall average (Fig. 2a). However, more signi cant differences are apparent in urban areas (Fig. 2b). In particular, most of the legislators, professionals and managers have a higher average intake. Figure 2c and 2d reveal that households engaged in mining, construction, manufacturing and transport labor activities are at an elevated risk of insu cient caloric intake.

Discussion And Conclusion
This analysis of nationally representative survey across India (2011-12) nds that the patterns of caloric intake and deprivations are signi cantly associated with occupational background of the households among both rural and urban households. Households dependent on occupations such as casual labour in agricultural and non-agricultural activities as well as those involved in low-end sales and services consumed fewer calories than others and also were at an elevated risk of caloric deprivation. In contrast, households engaged in market oriented skilled agricultural and shery as well as the high-level professionals and managers had more than adequate calorie consumption and also were at lowest risk of such deprivations. These patterns mirror the evidence on disproportionate burden of poverty and deprivation among unskilled agricultural and non-agricultural workers in rural and urban India [13, 38, 17, 18, 39 -40].
The regressions, particularly the occupation-speci c random effects (Figure 2), reveal that market oriented skilled agricultural and shery workers are among the select few occupations which display robust association with food and nutritional security. The households belonging to this occupational category display signi cantly higher levels of calorie consumption and also a lower risk of caloric deprivation. The nutritional well-being of this group is matched only by households belonging to legislators, professionals and managers. The occupation-speci c random-intercepts, con rm the stark inter-occupational disparities in nutritional intake with highest disadvantage for unskilled mining and construction laborers as well as those engaged in elementary sales and services workers. The betweenoccupation differences are also higher in urban areas. This is an important concern because higher degree of occupational diversi cation has not led to more equitable nutritional intake although there is evidence to suggest its favorable in uence on poverty reduction [41].
Conventionally, poverty and nutritional deprivation in India is largely discussed as a state-level phenomenon [42][43][44][45][46][47]10]. However, there is limited evidence to understand whether it is more associated with occupational differences or other forms of disparities across states and regions. In this regard, the variance partition coe cients (VPCs) provide an overwhelming evidence to emphasize on occupations approach towards poverty and nutritional well-being. The VPCs highlight that occupational groups have the greatest effects on caloric intake across households and these effects were greater than the statelevel in uence. In particular, it is noted that the amount of variation in caloric intake attributable to the occupations (10.2% and 18.1% for calorie consumption in rural and urban India, respectively) is substantial even when adjusted for standard household-level socioeconomic correlates. These insights when combined with the occupation-speci c random-effects suggest that policy focus to promote market oriented skilled agricultural and shery workers can be an equally good option as direct investments in professional and managerial skills for manufacturing and services. Moreover, these ndings reiterate the need for a balanced approach towards skill development in India whereby a focus on agricultural sector is not undermined because of narratives favoring manufacturing and services.
Yet, the success of skill development and training programmes depends on a range of factors. In particular, occupational mobility declines with age, therefore it is important to impart skill and training at the right age to increase the output of skill development programmes. As such, in 2010, for the rst time the share of working age population in India increased to 60% of the total population [48]. Such favorable population age-structure is expected to have signi cant impact on economic growth and development [50][51][52][53]. However, despite such favorable outlook, the developmental narrative for India is less optimistic largely because of a number of growth constraints. In particular, inadequate labor skills and low employment opportunities are regarded as a major challenge for higher growth and sustainable development in India [54][55][56]. Given the challenges, recent policies have emphasized on skill development to realize the full demographic dividend. In particular, the skill development campaign with major initiatives like Pradhan Mantri Kaushal Vikas Yojana (2015) and Kaushal and Rozgar Mela (2016) are apt examples of speci c policy interests. But a skill-based approach is more likely to favor the younger generations whereas these policies also have to accommodate the needs and concerns associated with cohorts of older adults. In the absence on an all-encompassing approaching, it is likely that poverty and deprivation can turn out to be an inter-generational affair whereby only the young generation within poor households is presented with any potential chances to improve upon household well-being.
Furthermore, it is equally important to entail a gendered-perspective for skill development. In fact, there are success stories (such as Kudumbshree, the Kerala State Poverty Eradication Mission) to demonstrate that a gendered-approach towards market oriented and skilled agrarian workers can be an effective approach to enhance the income and nutritional status of households [57,58]. This also implies that the Skill India agenda should focus on mainstreaming agrarian occupations by promoting professional and technical education among those engaged in unskilled and subsistence agriculture and shery. In particular, despite a vast coastline, the shery sector in India is relatively unexplored for its potential impact on nutritional and income security. It is no surprise that the policy paradigm is rather in sync with the developing world whereby poverty among small-scale sheries has remained a neglected aspect of development [15,59,60,38,61]. Whereas, there is increasing evidence to support that modernization of the sheries sector offers tremendous potential for development and growth [62 -65].
Nevertheless, in concluding, it is worth noting the three limitations of the analysis that can be largely associated with the nature of survey and the data. First, given the cross-sectional design, the results do not necessarily reveal the casual direction of association between occupation and caloric intake even though this does not impact the results regarding occupation-speci c disparities and advantages in caloric intake. Second, although the NCO 2004 classi cation is su ciently disaggregated to arrive at some meaningful inferences but further disaggregation is advisable to understand the intricacies associated with skilled occupations within agriculture and shery sectors. In fact, in the survey the NCO