1 Introduction

Food security is a central theme in development agendas, as many people suffer from hunger. Given this, the second Sustainable Development Goal (SDG) of zero hunger sets targets for food security and improved nutrition. Furthermore, the first two targets are to ensure access by all people to safe, nutritious and sufficient food all year round and to end all forms of malnutrition by 2030. According to the definition of the Food and Agriculture Organization (FAO), food security exists when all people, at all times, have physical and economic access to sufficient, safe and nutritious food to meet their dietary needs and food preferences for an active and healthy life.

Providing opportunities for improving their livelihoods is key to reducing undernourishment [1]. Given that a considerable number of people leave their birth countries for employment to support families back home, the paper investigates whether migration and subsequent remittances can be part of the solution to achieve these SDGs in developing countries. In line with the large number of people migrating, the developing world receives massive remittances. In 2019, international remittances to developing countries through official channels were estimated at US $ 551 billion [2].

Migrants and remittances are two distinct channels through which international migration can influence food security in the home country (Fig. 1). The next section discusses these two distinct ways of how migrants and remittances influence food security. Therefore, the objective of the current study is to examine the effect of migrants and remittances on food consumption in the country of origin, with special attention paid to the difference between migrant and remittance-receiving households.

Fig. 1
figure 1

Source: Authors’ illustration based on various literatures

Impact of Migrants and Remittances on the Home Country Food Security.

Full size image

1.1 The conceptual framework and pathways of migration and remittances impact on household food security

As mentioned above, migrants and remittances have separate effects on household food security. Firstly, remittances increase the purchasing power of recipient households, creating higher demand for food through market purchases and relaxing the financial constraints for agriculture [3,4,5,6,7,8,9,10,11]. However, increased purchasing power may adversely affect food security if they spend more on less nutritious food. Moreover, Jimenez [5]and Craven and Gartaula [12] found that migrant households consume more industrialized and imported foods, affecting the consumers’ access to locally produced food. This will eventually threaten the local food production in the migrant-sending and remittance-recieving country (Fig. 1).

Secondly, labour loss due to migration will also affect food consumption and production [3]. On the consumption side, the physical disappearance of people from a nation is expected to decrease the overall demand for food. Scholars have argued that per capita food expenditure or per capita food demand decreases with household size [13,14,15,16].

On the other hand, authors believe that migrant workers’ exposure to the new environment(s) and information will influence food consumption in their home country by increasing their awareness of food nutrition. Further, the authors assume that if the major income earner of a household migrates, the sudden reduction of income may adversely affect the demand for food through the household’s decreased purchasing power.

On the production side, a reduced labour endowment will negatively affect agriculture due to lack of labour [17,18,19]. Excess demand for labour may increase wages, which leads to increases in the cost of production and reduces agricultural profits. However, this should be analysed with special care as Fig. 1 explains that remittance income may release the credit constraints in agriculture, enhancing productivity. Further, new exposure of migrant workers also influences food production by increasing the awareness of investment opportunities and changing the perceptions of agriculture [20]. Understanding how the two different channels of migrants and remittances affect household food security is important, as there is a significant number of households which have migrants only, but do not receive remittances and vice versa [7, 21]. Therefore, the correctly identified unit of analysis is important. If the analysis is to identify the impact of migrants, the unit of analysis should be migrant households. If the analysis is about the impact of remittances on agriculture, unit of analysis should be remittances-receiving households. If the analysis is about migration and remittances, the unit of analysis should be the households with migrant workers and remittances-receiving households.

1.2 Contribution of the study

The contribution of this study is twofold: empirical and a policy-level contribution. In terms of the empirical contribution, first, the current study addresses identification issues of the unit of analysis. Second, according to the authors’ best knowledge, the current study is the first to use a multi-valued treatment effect model to analyse the impact of migrants and remittances on household food security in the remittances receiving/migrant-sending countries.

Third, most studies in the relevant literature use one food security indicator such as food expenditure, or daily calorie intake, coping strategy index (CSI) or dietary diversity [22,23,24,25,26]. Not only is one indicator insufficient to capture the multidimensional aspects of food security,Footnote 1 but as Carletto, Zezza and Banerjee [27] argue, some of these indicators themselves have limitations. Therefore, the authors addressed the issues regarding food security measurements. Fourth, even though some literature employs several food security indicators to capture the multidimensional aspects of food security, they have not considered the identification issues of the unit of analysis explained earlier. For example, Regmi and Paudel [28] utilize both the Food Consumption Score (SCI) and the Household Hunger scale, but it does not consider the identification issues of the point of analysis. Further, Crush [29]) employs a household food insecurity scale (HFIAS), household food insecurity access prevalence indicator (HFIAP) and the household dietary diversity scale (HDDS). Fifth, in many studies, those single indicators have been used for a blanket analysis of overall food expenditure or overall calorie consumption [8, 11]. These blanket analyses are too broad to allow us to understand how different households spend their income on and obtain calories from different types of food groups.

This analysis will shed light on areas where the government and policymakers should focus on improving the impact of migrants and remittances on food consumption in Sri Lanka. Understanding the separate impact will clear misunderstandings at some policy implementation levels.

1.3 Why is Sri Lanka important to conduct this study?

The importance of choosing Sri Lanka for this study is that the separate estimates for migrants and remittances are more significant in Sri Lanka. Both migration and household food security are significant phenomena. Annually, many Sri Lankans leave the country for employment [30]. In line with this large sum of emigrants, remittance inflows to Sri Lanka have increased steadily from US$ 60 million in 1979 to US$ 7 billion in 2018, which comprises around 9 per cent of GDP [31] (CBSL 2018). On the other hand, there is an issue of food insecurity at the household level in Sri Lanka. Around 4.5 million Sri Lankans, which amounts to around 20 per cent of its population, are undernourished. Malnutrition is a challenge, and Sri Lanka ranks at a lowly 84th place among 118 countries in the Global Hunger Index (GHI) and 65th place in 113 countries in the Global Food Security Index (GFSI). Acute malnutrition is also prevalent in children—at around twenty per cent in the age range of 6 to less than 60 months. Moreover, dietary energy consumption has remained appallingly poor, as almost half of the population consumes below the minimum level of dietary energy requirement [32]. More importantly, the prevailing economic crisis worsens the food security status of the country after 2021, mainly due to food inflation, lack of foreign reserves and loss of production [33]. Considering the facts related to remittances and food security, Sri Lanka is an ideal case to achieve the objectives of this study.

2 Materials and methods

Household Income and Expenditure Survey (HIES) (2016/17 data is employed in this study, covering a sample of 25,000 households on the island of Sri Lanka. This data is collected by the government agency of data collection: Department of Census and statics of Sri Lanka. Sample selection was mainly two-stage and stratified; district was the main domain for stratification, while sectors–urban, rural and estate–in each district were the selection domain. The sampling frame was the list of housing units prepared for the Census of Population and Housing (CPH) 2011. Primary sampling units (PSUs) –census blocks–were selected systematically from the CPH by allocating selection probability which is proportionate to the number of housing units available in the census block. Secondary Sampling Units (SSUs)–housing units–were randomly selected from the selected PSUs. HIES data has questions on migrants and remittances. Therefore, HIES data can be used to achieve the objectives of this study and as these data collected by the government agency, data can be considered as reliable.

Before analysing the model employed in this study, this section introduces the unit of analysis, followed by an explanation of how food security was measured. The model and analysis methods are discussed at the end of this section. As this study does an impact evaluation method, authors searched for all the impact evaluation methods and selected the propensity score matching technique which is widely used.

2.1 Measuring migration (unit of analysis)

To address the measurement issue in migration, this study used four different types of households; only receive remittances (OnlyRRHHs), only have migrants working in another country (OnlyMiHHs), households which have both migrants and remittances (BothMRs) and households which have neither migrants nor remittances (normal households).

In the HIES (2016/17) data, 17 per cent of households were identified as MiHHs,Footnote 2 of which 72 per cent did not receive remittances (Table 1). While eight per cent of the total Sri Lankan households receive remittances, 38 per cent do not have migrant members. Lastly, around five per cent of the total households have both migrants and remittances. If remittance-receiving household (RRHHs) is the unit of analysis, a significant percentage (12%) of OnlyMiHHs which record only migrantsFootnote 3 is classified as non-RRHHs. This classification underestimates the impact of migrants on food security, such as information flow, labour availability and consumption requirements. Similarly, if MiHHs are the unit of analysis, a significant percentage (3%) of OnlyRRHHs which does not record migrants are categorized as non-MiHHs.

Table 1 Remittances receiving and migrant sending households in Sri Lanka
Full size table

This ignorance of a portion of RRHHs underestimates the impact of remittances on food security. With this classification of households in the model, none of the proportion of households is ignored. Therefore, this study corrects the underestimation/overestimation issues regarding the impacts of migrants and remittances on food security.

2.2 Measuring food consumption

As there is an empirical puzzle between per capita food expenditure and household size, the household size is controlled in analysing the multivalued treatment effect. The current study measures food consumption by measuring per capita food expenditure, per capita calorie consumption, and food consumption diversity.

Per capita food expenditure sheds light on overall food expenditure and spending on specific food categories [34].Per capita food expenditure can be used as an indicator to know whether increased income releases the financial constraint for purchasing food or whether sudden income loss due to migration of the income earner reduces food expenditure. Total food expenditure was calculated by summing consumption expenditure on 280 consumption items and was aggregated into 11 groups (Table 2). Monthly per capita food expenditure was calculated by dividing total food expenditure by the adult equivalent household size employed by Fiegehen and Lansley [35].

Table 2 Monthly Food Expenditure and Daily Food Calorie Share by Households Type
Full size table

The per capita daily calorie consumption indicator translates these expenditure values into calorie consumption and enables us to measure the impact of migrants and remittances on the nutritional requirements of households. Hoddinott and Yohannes [36] emphasise that per capita calorie intake is a reasonable food security indicator. The quantity of food consumed is converted into calories, by using the National Nutrient Database for Standard Reference provided by the United States Department of Agriculture and the dietary guidelines provided by the Ministry of Health in Sri Lanka [34, 37].

As can be seen in Table 2, households’ food expenditure is dispersed among different types of food groups, but calorie consumption is concentrated around a few types of foods. Households mainly spend on cereal and other starches, animal protein, food away from home, milk and other dairy, condiments, fat and oil, and vegetables.Footnote 4 However, the main calorie sources of these households are cereal and other starches, fat and oil and food away from home. Although households spend around 18 er cent of their food budget, they consume around 4 per cent of calories from animal protein. These findings are more or less similar among the three types of households. With slight differences, all types of households spend only around 20 per cent of their food expenditure on ‘cereal and other starches’ but consume more than 50 per cent of their daily calories from that food group. Further, households spend less money on ‘fats and oil’ but their calorie contribution is more than 16 per cent.

Furthermore, Evang et al. [38], point out that low diet quality is a driver of general and micronutrient malnutrition. Jayawardena et al. [39] indicate that a substantial proportion of Sri Lankans do not consume a varied or balanced diet. Sri Lankan diets are mostly composed of starchy staples such as rice, wheat flour. The paper adds to the literature by taking into account the diversity of diets as well. In addition to calorie consumption, we take into account diversity indices to measure food security based on both consumption and expenditure by using the Simpson index and the Shannon index. The Simpson index evaluates diversity in terms of the number, as well as the distribution of different food items [34, 40]. It is defined as \(1-\sum {w}_{i}^{2}\) where \({w}_{i}\) is the expenditure share or calorie share of food group i. The Shannon index is defined as \(-\sum {w}_{i}log\left({w}_{i}\right)\) and evaluates the concentration of food groups [34].

As a result, expenditure-based dietary diversity is higher compared to calorie-based dietary diversity indices in all types of households (Table 3). As can be seen in Table 3, asset indices vary between households in terms of ownership of farming assets and livestock. OnlyRRHHs and BothMRs have fewer farming assets compared to other types of households. However, BothMRs have more livestock assets than the other three types of households.

Table 3 Dietary Diversity Index and Assets Index by Household Type
Full size table

2.3 Choice of the model

The analysis of this paper is based on descriptive statistics (provides insight into current food security status), multinomial logit regression and the multivalued treatment effect technique. Multinomial logit regression, an extension of binary logit regression, analyses which households are more likely to be OnlyRRHs, OnlyMiHHs, BothMRs, compared to be a normal household. The Multinomial logit model (Cameron & Trivedi, [41]) is specified as;

$${P}_{ij}=\frac{{e}^{{x}_{i}{\prime}{\beta }_{j}}}{{\sum }_{1}^{3}{e}^{{x}_{i}{\prime}{\beta }_{j}}} j=1, 2, 3$$
(1)

where \({P}_{ij}\) is the probability of being OnlyRRHs, OnlyMiHHs, or BothMRs, \({x}_{i}\) is a set of socio-economic characteristics of migrant-sending and remittance-receiving households.

Migration and remittances are not random processes but are self-selected. Migration and remittances may be influenced by a range of factors, including food consumption. That means there is a possibility for reverse causality in that lack of food security situation leads to migration and remittances rather than vice versa. This issue can be summarized as establishing causality or self-selection. The propensity score matching (PSM) method, an alternative method of addressing the issue of self-selection, was employed in the current study. PSM method controls for the selection of observable characteristics by comparing treatment groups with similar control groups. Moreover, this method controls for unobservable characteristics based on the conditional independence assumption (CIA) [42]; the potential outcomes are independent of treatment status and therefore assume that there is no unobservable characteristics bias. To ensure that CIA holds, a large set of pre-treated independent variables (covariates) which are independent of treatment are required. The reliability of the CIA or the selection of covariates can be tested with a post-match balance test which indicates how well matching has replicated. The present study included a wider range of covariates to avoid the possible risks of selection on unobservable characteristics, and the results of the balance test are presented in Table 4. They include individual and household characteristics such as age, gender, and employment status of the heads of the households, average years of schooling separately for male and female household members, household size, and dependency ratioFootnote 5 and assets index. The differences between the treated group average and the control group average of each covariate are reduced significantly with the Kernal matching score. This reduction in the bias of each covariate is significant in all three types of households: OnlyRRHHs, OnlyMiHHs and BothMRs compared to normal households. One major advantage of Kernel matching, among other matching systems in PSM, is the lower variance, which is achieved because more information is used.

Table 4 Results of the balance Test (OnlyRRHHs and Normal households)
Full size table

PSM allows correcting possible selection by comparing each treatment household with similar non-treated households based on their propensity scores, which is the probability of being in the treatment group based on observed baseline characteristics. According to some scholars, PSM gives more accurate non-experimental estimates when there is a self-selection problem [43,44,45].

Although Rosenbaum and Rubin [46] considered using a binary treatment, such treatment was found to be unsuitable for the current study, as treatment (T) takes more than two values, as explained earlier in Sect. 2.1 (t = 0, 1, 2, 3). Therefore, the multi-valued treatment technique, an extension of the PSM, neatly fits with the objectives of the current study [41, 47, 48]. The model chosen is derived from a study by Imbens [48].

Associated with each household i and each value of treatment t, there is a potential outcome\({Y(t)}_{i}\). The outcome is associated with the treatment level and a vector of pre-treatment variable\({X}_{i}\). \({Y(t)}_{i}\) in the current study is one of the eight outcome variables of household i; log of per capita monthly food expenditure, log of per capita daily calorie intake, expenditure-based Simpson index, expenditure-based Shannon index, calorie based Simpson index, calorie based Shannon index, farming assets index and livestock assets index. Pre-treatment variables (covariates), \({X}_{i}\) includes individual and household characteristics such as age, gender, and employment status of the heads of the households, average years of schooling separately for male and female household members, household size, and dependency ratioFootnote 6 and assets index. These variables cover most of the important determinants of migration as identified in the literature [26].

Average outcomes can then be estimated by averaging these conditional means [48]:

$$E\left\{Y(t)\right\}=E\left[E\left\{Y(t)|X\right\}\right]$$
(2)

Imbens [48]has modified Rosenbaum and Rubin’s definition of propensity score and renamed it as the generalized propensity score. The generalized propensity score is the conditional probability of receiving a particular level of the treatment given the pre-treatment variables [48]:

$$r\left(t, x\right)\equiv pr\left(T=t|X=x\right)=E\left\{D(t)|X=x\right\}$$
(3)

As there is no logical order in the treatment in the current study, multinomial logit is used to estimate the \(r\left(t, x\right)\). The problem is to estimate the average impact of migration and remittances.

$$E\left\{Y|T=t, r\left(T, X\right)=r\right\}-E\left\{Y|T=s, r\left(T, X\right)=r\right\}$$
(4)

Equation 4 measures the average treatment effect (ATE) of exposing all units to treatment t rather than treatment s.

By employing this model, the authors tested the main three hypotheses.

Hypothesis 01: There is a strong relationship between out-bound migrants and household food consumption measured by expenditure on food, calorie consumption and dietary diversity, as explained in Fig. 1.

Hypothesis 02: There is a strong relationship between remittances inflow and household food consumption, measured by expenditure on food, calorie consumption and dietary diversity, as explained in Fig. 1.

Hypothesis 03: There is a strong relationship between out-bound migrants and remittances inflows and household food consumption (expenditure on food, calorie consumption and dietary diversity) as explained in Fig. 1.

3 Results and discussion

This section is divided into two main sections: multinomial logit regression results and results of the multi-valued treatment effect.

3.1 Migration and remittances in Sri Lanka

Table 5 presents the results of the multinomial logit regression analysis (Eq. 1), that is, the key characteristics of migrants and remittances. are consistent with the insurance smoothing hypothesisFootnote 7 of the new economics of labour migration (NELM), which postulates that the purpose of remittances is to work against risks.

Table 5 Migration Analysis (Multinomial Logit Model)
Full size table

The most vulnerable groups, such as female headed households and elders, have a high likelihood of sending migrants and receiving remittances (Table 5). For example, female-headed households have a high likelihood of sending migrants and receiving remittances. In other words, female-headed households have 1.14 per cent probability of being in BothMRs than being in a normal household.Footnote 8, and this result is significant at 1 per cent significant level. Moreover, the age of the head of the household has a positive relationship with sending migrants. When the age of the head of the household is increasing by 1 unit, the probability of having an outbound migrant in the household is 0.01 per cent higher than not having a migrant member in the family and it is significant at 1 per cent significance level. This finding is consistent with Nguyen and Winters [34]. Nevertheless, employed-headed households and households with more educated memberss are less likely to send migrants and receive remittances. For example, an extra years of schooling of the male household members decrease the probability of being in BothMRs by 0.16 per cent. The result is the same for educated female household members, with a higher magnitude. These findings align with Fransen and Mazzucato [26] and Nguyen and Winters [34], respectively. Overall, these findings are consistent with the fact that the majority of Sri Lankan emigrants are lower-skilled or unskilled. In 2016, the unskilled and housemaid categories accounted for more than 56% of total departures for foreign employment.

All three types of households own more consumer durables, which aligns with the findings of Fransen and Mazzucato [26]. For example, households with higher assets index have higher probability to receive remittances and sending migrants (0.45%) compared to a normal household. In one hand, this may indicate that wealthier households have more opportunities to migrate and, therefore, to receive remittances. On the other hand, these households may have a history of migration. If they have returned migrants, or if migrant members are long-term migrants, those households are likely to have accumulated more consumer durables. The International Labour Organization (ILO) [49] reports that returned migrants tend to purchase more consumer-durable assets such as electrical appliances.

Given the characteristics of migrant and remittance-receiving households, the next section provides the findings regarding the impact of migrants and remittances on food consumption. status of those households separately.

3.2 Impact of migrants and remittances on food consumption

The impact of migrants and remittances on immediate food consumption is presented in three groupings: impact on food expenditure, impact on calorie consumption, and impact on dietary diversity, measured by ATE compared to normal households.

3.2.1 Impact on food expenditure

Table 6 shows the results of Eq. 4, which compares the food expenditure patterns of OnlyRRHHs, OnlyMiHHs and BothMRs to normal households. As can be seen in Table 5, remittances increase household food expenditure. Compared to normal households, OnlyRRHHs spend 4 per cent more on food, and this disparity is accounted for by ‘fruits’, ‘animal protein’, ‘milk and dairy’, ‘sugar and beverages’, ‘food away from home’, ‘condiments’ and Liquor (Table 6). OnlyRRHHschange food expenditure without concern for the nutritional contents and healthiness of food. For example, remittances increase demand for all types of healthier and nutritious foods (‘fruits’, ‘animal protein’, ‘milk and dairy’) as well as unhealthy foods (‘sugar and beverages’ and ‘food away from home’). OnlyRRHHs increase expenditure on both nutritious and less nutritious foods and reduce their expenditure significantly on healthy foods such as ‘vegetables’. Three main conclusions can be drawn from this result of increasing food expenditure due to remittances. First, because of remittances, OnlyRRHHs can spend on food categories which can be categorised as nutritious, expensive and luxury foods such as animal protein and dairy products. Second, OnlyRRHHs increase food expenditure in almost all food types without concern for the nutritional contents and healthiness of food. Third, not only do OnlyRRHHs increase expenditure on less nutritious and less healthy foods, but they also reduce their expenditure significantly on healthy foods.

Table 6 Average Treatment Effect on Per capita Monthly Food Expenditure (Percentage difference)
Full size table

Food expenditure patterns of OnlyMiHHs are affected in two ways. First, sudden income loss due to the emigration of an income earner affects OnlyMiHH’s capacity to survive only with basic staples, which are relatively cheaper and reduces their consumption of expensive foods. Second, as per Fig. 1, migrants are already aware of information concerning healthy eating, therefore OnlyMiHHs tend to increase the consumption of healthier food. OnlyMiHH’s overall food expenditure is 7 per cent higher than a normal household, this change of the food expenditure is mainly coming from the consumption of ‘vegetables’, ‘fruits’, and ‘confectionaries’. OnlyMiHHs spend 16 per cent higher than normal households on ‘confectionaries’. Even though, OnlyMiHHs do not reduce the consumption of unhealthy foods, they have increased the spending on some healthy food groups such as ‘vegetables’ and ‘fruits’. However, it is ambiguous that this increase in the expenditure on healthy food is exactly due to the information received from the migrants as it is not measurable.

The food expenditure patterns of BothMRs should be affected by three channels (Fig. 1). First, increased income through remittances should increase food expenditure. Although there is no evidence to support this channel in terms of overall food expenditure, expenditure on some food categories has increased. For example, expenditure on animal protein among BothMRs is 20 per cent higher than in normal households.

3.2.2 Impact on calorie consumption

Table 7 compares OnlyRRHHs, OnlyMiHHs and BothMRs with normal households in terms of the impact of migrants and remittances on per capita daily calorie consumption. Remittances positively affect calorie consumption in Sri Lanka. OnlyRRHHs consume 17 per cent higher calories than normal households on overall food. As OnlyRRhhs spend more on expensive, nutritious and luxury food types, their calorie consumption from these food types is also high. For example, calorie consumption from ‘animal protein’ and ‘milk and dairy’ products among OnlyRRHHs is higher than in normal households by 7 per cent and 8 per cent, respectively, in Sri Lanka. Calorie consumption from fruits is also significantly higher in OnlyRRHHs than in normal households. Apart from these healthy food categories, OnlyRRHHs also consume some more ‘bad’ calories. For example, OnlyRRHHs consume significantly higher calories from ‘sugar and beverages’, and ‘food away from home’ as well. Not only does this reveal an increase in less healthy calorie consumption, but also OnlyRRHHs reduce their consumption of some healthy calories, such as ‘vegetables’. Per capita, daily calorie consumption of OnlyMiHHs is not much affected by migration, except for the higher calorie consumption from ‘fruits’ and ‘sugar and beverages’.

Table 7 Average Treatment Effect on Per Capita Daily Calorie Consumption (Percentage Difference)
Full size table

As for the expenditure pattern, BothMRs’ per capita daily calorie consumption is also affected by the ‘three channels’. First, it is predicted that increased income through remittances should increase calorie consumption. Although migrants and remittances together do not increase overall calorie consumption, calorie consumption of BothMRs from some food groups is higher than in normal households. For example, calorie consumption among BothMRs from animal protein is 23 per cent higher than in normal households.. They consume 36 per cent higher calorieks from fruits in the rural sector than in normal households..

3.2.3 Impact on food diversity

The impact of migrants and remittances on dietary diversity, based on both expenditure and calorie consumption, is shown in Table 8. Both OnlyRRHHs and OnlyMiHHs spend on diverse foods in Sri Lanka. However, the diversity of calorie consumption among OnlyRRHHs is significantly lower (0.02 per cent according to the Simpson index and 0.04 per cent according to the Shannon index) than in normal households, while there is no significant difference between OnlyMiHHs and normal households. This pattern remains in the rural sector as well. However, in the urban sector and the estate sector, remittances and migrants do not have a significant impact on food diversity based on both expenditure and calorie consumption, except in OnlyRRHHs in the estate sector.

Table 8 Food diversity estimates
Full size table

4 Summary, conclusion, and policy implication

Using the HIES 2016/17 data set for Sri Lanka, this study examines the role of migrants and remittances on food consumption measured by per capita monthly food expenditure, per capita daily calorie consumption, and food diversity.

The results of the multi-valued treatment effect analysis show a positive impact of remittances on food expenditure and per capita daily calorie intake. However, OnlyRRHHs increase their expenditure on all types of food and consume all types of calories without considering their nutritional status.

The findings of this study show that OnlyMiHHs spend and absorb calories from healthy food. This finding is different from the previous studies of Karamba et al. [6], Craven & Gartaula [12] and Nguyen & Winters [34]. Karamba et al. [6] and Craven and Gartaula [12] highlight that migration appears to increase overall food expenditure, resulting in a shift towards the consumption of potentially less nutritious categories of food, such as ‘sugar and beverages’ and ‘eating out from home’. Nguyen and Winters [34] find no significant impact of migration on calorie intake for ‘food away from home’. However, the results of this study show that remittances promote less healthy food consumption and that migrants help households to be away from them. These results show the significance of differentiating the impact of migrants from the impact of remittances on food expenditure and calorie consumption.

Migrants and remittances jointly affect BothMRs in Sri Lanka by causing them to purchase and absorb calories from healthy and expensive food such as ‘animal protein’. However, the impact of migrants on the food consumption pattern seems to dominate over the impact of remittances.

As a result of OnlyRRHHs purchasing all types of food items, their food diversity based on expenditure is significantly higher than that of normal households in Sri Lanka. However, the food diversity based on calorie consumption is significantly lower than others. OnlyMiHHs also show significantly higher food diversity based on expenditure, while there is no impact on food diversity based on calorie consumption. The joint impact of migrants and remittances does not impact food diversity. These findings are slightly different from those of Nguyen and Winters [34], who found that neither short-term nor long-term migration affects food diversity in Vietnam.

The results of this study validate the disaggregation of the influence of migrants and remittances on food security and its analytical framework presented in Fig. 1. This built-up and validated analytical framework can be used in any other country to analyse the impact of migrants and remittances on food consumption.

Given the findings of the current study, it is clear that awareness of food nutrition and food investment among remittance receivers and left-behind family members needs to be improved. Most of the awareness programs implemented in Sri Lanka target migrant workers and educate them to earn higher wages abroad and reduce their risk. However, promoting education on and awareness of households’ financial management, micro-level food security, and food utilization among remittance recipients and left-behind members is rare.

As remittances alone are more effective in influencing food expenditure and calorie consumption, further policy intervention is needed to raise remittance flows from migrants who do not send remittances back home. Policies that facilitate remittance flows and lower the transaction cost are possible solutions.

The targeting of food subsidy programs requires revisiting. According to Ilangasinghe and Bandara [50], migrant households are not eligible for the government’s poverty alleviation program.Footnote 9 Nevertheless, the current study’s findings show that a significant percentage of migrant households are not receiving remittances and that their food security is lower than that of normal households. Hence, the targeting criterion of subsidy programs should be revisited. The content of food subsidy programmes also should be improved. The inclusion of an awareness component is also recommended in cases where the food subsidy is given as a voucher instead of as a food basket.

This study recommends that the government make the subsistence agricultural sector attractive and profitable so that remittance recipients and left-behind family members of migrant households are encouraged to engage in this sector. The government could implement policies to provide incentives for home gardeners, such as the ‘Divineguma’ program, which should certainly be continued.

The major limitation of this study is that it analyses the impact of migrants and remittances on food consumption. However, the authors do not analyse the impact of migrants and remittances on food production, which is indicated in Fig. 1. A more comprehensive analysis can be conducted by addressing this limitation in the current study. Further, the result might be similar or different in the case of internal migration as the migrant is no longer in the household. Therefore, future research can be conducted on evaluating the impact of internal migration on food security.