COVID‐19 Induced Income Loss among Migrant Workers: Evidence from Eight Villages of Bihar

Abstract

The COVID‐19 crisis re‐shaped many livelihood options and placed significant burdens on those with precarious incomes exacerbating persisting vulnerabilities, especially among a large section of the migrant population. This group faced a dual threat – both to their livelihood and health. To understand the consequences of this pandemic on the income of the migrant population, a household level survey was conducted in the state of Bihar, India, which is one of the highest migrant‐sending states. We examine the role of differences in the socio‐economic status of migrants and their households in determining the extent of vulnerability caused by the COVID‐19 crisis. Vulnerability is proxied by the income lost by migrants during the lockdown. The results suggest that households with diversified income portfolio, larger landholdings, and those receiving government benefits suffered significantly lower income loss whereas, larger household size and greater distance from town tended to escalate income loss. Additionally, private salaried workers faced higher income loss and an increment in years of education lowers the losses significantly. It is observed that individual‐level characteristics also played a significant role in determining economic loss due to the lockdown. Our findings suggest a binding necessity to actively shape policies considering the financial insecurity of vulnerable migrants at their destination and the household members at the origin.

1. Introduction

The recent COVID‐19 induced economic crisis has had a profound and pervasive impact across all sectors of the economy. The challenges posed by the pandemic in terms of the sudden outbreak, inadequacy of medical infrastructure, high‐fatality rates, and time lag in vaccine production had prompted governments across the world to impose stringent restrictions on social and economic processes. The Government of India too declared a complete lockdown1 from 25 March to 3 May 2020, and a partial lockdown which continued till 31 May 2020. This prolonged period of lockdown has paralysed all the economic activities and brought every sectors of the economy to a near standstill. Consequently, massive disruptions in production processes took place which impacted workers employed in both the formal and informal sector. Exposing the deep fault lines in the Indian labour market, this pandemic brought to the fore the plight of the migrant workers who were particularly hit hard since they were majorly employed in the informal sector industries (Misra & Gupta, 2021) like construction, manufacturing, transports and hospitality services. Besides this, migrant workers also tend to work as domestic help and petty street vendors. With the closure of working sites and residential colonies due to the lockdown, most of such workers were out of job. According to the Centre for Monitoring Indian Economy, about 122 million workers went jobless during April 2020 itself and about 90% of migrant workers did not receive their salary (Hindu, 2020), were left stranded, starved, homeless and uncertain about the future (Slater & Masih, 2020).The majorityy of these workers lacked social security and basic amenities (Mukherjee et al., 2009; Tiwary et al., 2012; Bhagat et al., 2020), and are often denied proper healthcare, sanitation and nutrition (Nair & Verma, 2020). The loss of jobs and limited saving for survival in urban centers triggered mass exodus of migrant workers who started walking hundreds of kilometres to reach back their native places as lockdown was imposed (Dandekar & Ghai, 2020; Pandey, 2020). Many died due to accidents, starvation, exhaustion and lack of medical aid (Guha et al., 2021).

To the best of our knowledge, no study has been conducted to assess the actual income loss of migrant individuals and households due to the imposition of the lockdown in India. Therefore, to understand the plight of such workers, we study eight villages from the state of Bihar. It is the second highest migrant sending state and one of the poorest states2 of the country, with 51.91% of the total population living below the poverty line. It was identified as a state with the highest number of multidimensional poor in India. In 2021–22, the state reported lowest per capita income in the country (Economic Survey of Bihar, 2021–22). The failure of the State to provide livelihood opportunities forces its local workforce to migrate in search of alternative livelihood opportunities. This trend of outmigration for work still persists, and Bihar's economy is often termed as “remittance economy”. Studies such as Datta (2016) found that 52% of rural households in Bihar reported remittance to have contributed more than half in their total income. The COVID‐19 crisis disrupted all income sources, including remittance inflow to these households. This unprecedented crisis accentuated the precarity of such migrant households making Bihar the appropriate place to provide unique insights about the impact of the COVID‐19 crisis on such households at the bottom of the income pyramid.

1.1. The Mass Exodus and Government's Strategy during the Pandemic

Job loss during the pandemic, exclusionary nature of welfare provisioning, invisibility as a citizen and lack of labour rights triggered a mass exodus of migrant workers from urban and industrial centers after the imposition of lockdown ‐ a phenomenon observed across most of the low and middle income countries such as India, Malaysia, Taiwan, China, Pakistan, Bangladesh, Latin American countries, Gulf regions and Sub‐Saharan Africa (KNOMAD, 2020). India with more than 100 million internal migrants became vulnerable. Although experts claim underreporting of figures, the Government of India confirmed the number of return migrants to be about 11 million3. In response to this huge exodus, the government organised ‘Shramik’ trains exclusively for the migrants but to avail this facility the migrants needed identity proof, which many of them were lacking. For instance, Jan (2020) – an NGO reported that 94% of migrant construction workers in India do not have any identity cards. Gram (2019) found that a significant portion of workers working in factory units did not have any documents to prove their employment status. The lack of legal identification, permanent residency and formal work documents forced the migrants to choose longer and tedious routes to reach their home, which were comparatively costly, long and painful (Jesline et al., 2021). The Government of India also announced a relief package of $22.6 billion to provide food, cash transfer and job creation for return migrants in rural areas through the Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA). But the existing invisibility of migrants in official statistics of labour, underfunding in rural infrastructure, healthcare and employment generating opportunities over the decades made the initiative difficult to execute. This lack of identification acted as a hindrance in accessing any welfare assistance provided by the state or central government during the lockdown (Bhagat et al., 2020). Migrants were mainly denied welfare assistance because of the lack of portability of domicile and identity cards which belong to their native place. For example, through the public distribution system, the Government of India provided ration to all individuals, but this scheme requires location‐specific document in the form of ration card, thus excluding the migrant workers. This limitation was clearly visible during the lockdown when despite the announcements of food relief programs migrants suffered from hunger.

Studies till now have focused extensively on the exogenous factors that led to vulnerabilities of migrant workers, while ignoring endogenous factors such as household and individual level characteristics which could have an important role in amplifying the vulnerability of migrants. For example, Deshingkar (2017) highlighted the importance of caste in migrant's employment; the elite caste often holds powerful positions while backward castes work as a manual worker in low‐remunerative jobs. Similarly, households with diversified sources of income rather than singular dependency on remittances are expected to be in a better position with respect to handling any external shock. A few studies have tried to highlight the impact of pre‐existing characteristics on vulnerability (see Couch et al., 2020; Deshpande & Ramachandran, 2020; Abraham et al., 2022), but as Graeff (2002) have pointed out, they rely on online/telephonic survey response which could not capture the true picture due to uninformed response bias. This is primarily because telephonic surveys tend to increase efforts in generating higher item response rates which may have detrimental effects on the quality of the results (ibid.). Furthermore, telephonic surveys suffer from coverage bias and non‐response bias (Ambel, 2021) in the sense that individuals, especially females who may not own a separate mobile phone gets excluded from the surveys (Agarwal, 2021; Leo et al., 2015). In‐person surveys are also better as in such surveys the researcher and respondent often engage in long discussions thereby providing the later with important anecdotal evidence and information regarding the respondent's condition and leaves a much wider scope for repeat questions to clear doubts. Furthermore, much of the evidence on the disastrous impact of COVID‐19 has come from aggregate data, which has failed to reflect the significant heterogeneity that exists in its effect. Studying the dynamics of COVID‐19 induced economic shock by household‐specific factors as against aggregative studies can complement the later by providing much finer estimates of the distributional impacts. It can also present a better account of households' coping capacity, the role of people's saving, and the higher resilience of multi‐job households.

The rest of this article takes the following structure. Section 2 discusses a framework of vulnerability, and section 3 lays out the study design, the choice of the study area and sampling techniques. Section 4 discusses the pattern of migration from the study area. The empirical strategy for our analysis is presented in section 5. Section 6 discusses the results. Section 7 highlights the key findings and summarises them.

2. The Vulnerability Framework

Vulnerability is the likelihood that an individual or household is being affected by a given risk and is determined by the location, size, composition and asset status of the household (Farrington et al., 2006). The ability of households or individuals to prevent, mitigate or cope with the shocks is essentially an important question that needs understanding in the backdrop of the COVID‐19 pandemic. According to Maimbo (2021), jobs can be transformational as a source of “income and sustenance, of identity and fulfillment and of routine and learning”. It provides support to cushion the blow of major multidimensional shocks. Therefore, the loss of jobs during COVID‐19 made the world's most vulnerable people even more susceptible. The direct effect of the pandemic induced containment measures resulted in severe contraction of economic activities leading to income loss (ILO, 2020). As pointed out by Midões and Seré (2022), such income shock results in a substantial decline in individual well‐being, and reductions in disposable income of the household leads to high vulnerability (He & Zhou, 2022). A key factor shaping vulnerabilities and coping strategies related to COVID‐19 is individuals' and households' integration into markets (Eriksen & Silva, 2009), both factor market and product markets. The persistence of large low‐skill migration and the predominance of such migrant workers in the informal sector jobs not only highlight the inherently exclusionary nature of post‐colonial capitalist growth and development process in India (Sanyal, 2007), but also points to the structural vulnerabilities that migrant workers face. For example, while casual wage work is known to be precarious, even among regular salaried workers, 71% did not have a written job contract (NSSO, 2019). Again, informalisation within the ambit of the formal sector has resulted in an increase in the proportion of regular salaried workers with low job security in the past two decades (Amit et al., 2018; Bhattacharya et al., 2013). This ‘structural paradigm’4 emphasises the constraints that migration and informality doubly placed on the individuals by the more powerful institutional forces rendering them vulnerable and exposed to the risks of the pandemic.

Consequently, the vulnerability arising out of being a migrant worker with a limited degree of adaptive capacity because they suffer from a lack of entitlements in the social, economic and institutional spheres has predisposed and exposed a major section of this workforce to stress (Turner et al., 2003). Hence this large section of the working population is subject to the risk of poverty, deprivation and, in some cases, destitution. In normal circumstances, one of the highly acclaimed coping strategies of vulnerability is diversification of employment and income sources that would reduce the risk by increasing options. However, the COVID‐19 pandemic has provided a unique circumstance in which all kinds of economic activities were shut down for a considerable period of time. The absence of any economic activity rendered diversification of employment and income, both at the individual and household level, an ineffective coping strategy. In such circumstances, the well‐functioning institutional structure and state‐run welfare measures play a key role in developing coping mechanisms.

We study vulnerability ex‐post, using the information on the actual income loss of migrant workers, both at the individual and household level, to understand how income loss may have had an effect on the vulnerability of such households. Studying vulnerability using the extent of income loss not only helps to account for the risks faced by individuals and households but also the options and ability they are left with to handle such risks. Income loss due to multi‐dimensional shocks as COVID‐19 also tends to highlight the deeply rooted structural weaknesses of the economy that interact and amplify the distress of the already susceptible group of people. In our analysis, we did take into account the multifarious, interrelated, socio‐economic and institutional factors affecting individuals and households. This is primarily because analyses of COVID‐19‐related impacts have emphasised the importance of institutional factors such as state support (Dutta & Fischer, 2021; Krauss et al., 2022), credit and food relief (Kansiime et al., 2021) or existing structures of government neglect in shaping vulnerabilities, especially in rural contexts (Puerta Silva et al., 2020).

3. Study Area and Village Selection

The data for this study come from a primary survey conducted at the household level between December 2020 and February 2021 in eight villages situated across four districts of Bihar – an eastern Indian state. The choice of the state is guided by the fact that 51.9% of its total population is multidimensionally poor (NITI Aayog, 2021). The state has neither witnessed any commendable expansion in the farm sector nor in nonfarm sectors over the last decades, thereby creating an extremely worrying unemployment rate of 7.2%; higher than the national average of 6.1% (Kumar & Kumar, 2020; PLFS, 2020). This poverty and lack of employment opportunities compels a sizable portion of its population to migrate to other states for their livelihood (Choithani, 2017). According to the 2011 Census, there were a total of 7.5 million migrants who reported Bihar as a place of origin, making Bihar the second largest migrant‐sending state. Moreover, based on NSS 64th round (2007–08), Keshri and Bhagat (2012) highlighted that the temporary and seasonal migration of the working age population was highest in Bihar among the major migrant‐sending states. During the lockdowns, it was evident that circular, temporary and seasonal migrants suffered the most and returned back to their hometown (Srivastava, 2020)5. Prevailing poverty, lack of employment opportunities, high return migration during the lockdown, and no immediate government relief made the working population highly vulnerable to the pandemic.

Within the state, we study four districts, namely Muzaffarpur, West Champaran, Rohtas and Sheohar, situated across the two‐agro climatic zones of Bihar. This selection of the districts was made to represent the diversity in the cropping pattern, and therefore, it also accounts for the diversification of productive forces in the state. All the villages from the four districts were listed according to the Census (2011), and a stratified sampling technique was employed to select the sample villages. In the first stage, we listed all the villages within the district according to the number of households in each village. To create the sample frame, we retained the villages with a median range of households from the distribution of villages in a district. The final selection of the villages within a district was done from this list of the villages that have a median range of households. This has also given us the opportunity to maintain a consistent representation of the population through the chosen sample size across the study villages. In the second level of stratification, caste compositions were used, and villages with the presence of at least 20–30% historically deprived castes (SCs and STs) were sampled. The final level of stratification is based on the extent of irrigation (low, moderate and high); 15 villages were listed out, five from each category. Then from any of the three categories, two villages were selected randomly, considering that both the selected villages are not from the same category and not from the same blocks of the study districts. This gives us a total of eight villages, namely, Adalpur, Harpur (Muzaffarpur district); Sirji Barhampur, Bishnupur Raghunath (West Champaran district); Bisi Khurd, Nanhu (Rohtas district); Bishunpur Bindi, Azrakbe Pota (Seohar district) from the four selected districts. The sampled households from all the study villages exhibit diversified employment and income profiles and were selected by simple random sampling without replacement from within each different employment and income groups of the villages. The survey framework and household listing are detailed in Tables A1 and A2.

Our survey covered 378 households. Of these, 168 households had sent out 282 migrants. Of them, 231 (or 82%) were workers, and the others were mostly accompanying spouse or children. Of the 231 migrant workers, we could obtain information on income loss due to COVID‐19 from 172 individuals who comprise our final sample6. The study was conducted almost 7–8 months after the stringent lockdown was lifted and the economic and social processes were returning to normalcy, giving us an opportunity to examine the immediate impact of the lockdown and the re‐shaping of livelihood options adopted by the migrants as a consequence of the lockdown.

4. Pattern of Migration from the Study Areas

Across villages, the mean comparison of characteristics of migrant and nonmigrant households confirm that that migrant households have significantly larger family size, as reflected from household size, and an average number of dependents and workers. They are comparatively richer than nonmigrant households (Table 1).

Table 1.

Socio‐Economic Characteristics of Surveyed Households According to Migration Status

Socio‐economic characteristics	Nonmigrant household	Migrant household	Mean difference
Household size	5.9	7.6	−1.75***
Number of dependents	4.00	4.87	−0.87***
Number of workers	2.01	2.94	−0.93***
Log of land size (acres)	0.23	0.04	0.19*
Years of education of head	6.12	4.06	2.06***
Household income (per annum)	Rs. 148390.4	Rs. 234712.4	−86321.96***
Caste groups †
Others	20%	15%
BC‐1/EBC	30%	32%
BC‐2	30%	33%
Scheduled Caste/Scheduled Tribe	20%	21%
Landholding category
Landless households	28%	35%
Households with less than 1‐acres land	38%	36%
Households with 1 to 4 acres of land	26%	24%
Households with more than 4 acres of land	9%	5%
Total number of households	210	168

Note: ***Refer to 1% significance level, *Refer to 10% significance level.

Source: Field survey, 2020–21.

^† Forward Castes (Others), Extremely Backward Classes (EBC/BC‐1), Backward Classes (BC‐2), Scheduled Caste (SC) and Scheduled Tribe (ST) are the terms used by the Government of India to classify groups that are historically disadvantaged and educationally, economically or socially deprived.

Ownership of agricultural land has been a crucial productive asset in rural India and is considered to be an important security against any crisis. The average landholding size is significantly higher for nonmigrant households. Absolute landlessness has been very high among the surveyed households and particularly prominent for the migrant‐sending households. The extent of land poverty among the migrant sending households was high, with 35% of such households being landless and 36% of households with less than one acre of land. Such small landholding cannot provide enough to maintain a minimum standard consumption basket. This landholding pattern indicates that the majority of the households are dependent on nonfarm (including migration) sources of income or diversify their activities to supplement the limited income‐generating capacity of agricultural lands (Vanwey, 2003; Keshri & Bhagat, 2012; Dutta & Dhar, 2022).

The migrant‐sending households can be classified into three groups, households with temporary or seasonal migrants, households with permanent migrants and households with both temporary and permanent migrants. Across the study villages, 84% of households have sent out temporary migrants, 13% of households have sent permanent migrants, and only 3% of households have sent both types of migrants. Of the different types of migrants, average per capita income loss was the highest for temporary migrant‐sending households.

Across the study villages, 44% of households had sent out migrants. Among the migrant‐sending households, about 21% belonged to the SC and ST, almost 33% belonged to Extremely Backward Classes (EBC/BC‐1), another 32% belonged to Backward Classes (BC‐2), and 15% belonged to the social group Other7 (see Table 2 and Figure 1).

Table 2.

Proportion of Households Sending Different Types of Migrants and their Average Income Loss during the Lockdown

Type of migration	Proportion of households (%)	Average income loss (Rs)
Temporary	84	54,603
Permanent	13	48,438
Both (temporary and permanent)	3	51,333

Source: Field survey, 2020–21.

Figure 1.

Proportion of Migrant Sending Households by Social Groups.

Source: Field survey, 2020–21. [Color figure can be viewed at wileyonlinelibrary.com]

The reasons for migration across different social groups are presented in Figure 2. The reasons have been classified into four categories (i) lack of employment opportunities in and around the villages, (ii) lack of matching skills, (iii) irregular payment and/or part‐payment of wages and (iv) to access better education. The data suggest that lack of employment opportunities in the villages was the major reason for migration across all castes followed by migration for obtaining a better education. This indicates insufficient development of industries which could absorb this pool of workers in and around the study areas. An ever increasing workforce in Bihar, lack of gainful employment in agriculture, and absence of any manufacturing or service sector, left a vast majority of the workforce with no alternative but to seek work outside the state. The survey data is also indicative of the poor educational infrastructural development in terms of both quality and quantity in the study villages.

Figure 2.

Reason of Migration among Different Social Groups.

Source: Field survey, 2020–21. [Color figure can be viewed at wileyonlinelibrary.com]

Our survey data highlights the precarity inflicted by the pandemic, especially for migrant workers who were left stranded during the lockdown. Due to the lockdown, 92% of migrant workers from these villages lost their jobs. Loss of livelihood compelled most of the migrant workers to move back to their hometown, where they relied on meagre agricultural income or on saving and/or borrowings to maintain sustenance. The occupational profile of the migrant workers suggests that 57% were engaged in nonagricultural wage employment and 18% in agricultural wage employment. Of the rest 25% who reported having salaried jobs, the majority worked as drivers, security guards or waiters and did not have any formal job contract or social security benefits. Only one person reported a regular salaried job in the Indian Army (Figure 3). This shows that almost 99% of the migrant workers were engaged in the informal sector. The unskilled or semi‐skilled jobs that the migrant workers are engaged in are low paid, insecure, entail exploitative conditions and are without social security and welfare schemes8 (Deshingkar et al., 2022). Such kind of recruitment do not have provisions for advance payments and make them dependent on the recruiter for daily needs just as bonded labour (Mosse et al., 2005). As the lockdown was imposed, most of such migrant workers were abandoned by their contractors and employers leaving them stranded without job and money.

Figure 3.

Proportion of Migrants Engaged in Different Type of Employment Done at the Destination.

Source: Field survey, 2020–21. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 4 shows the distribution of return migrants during the lockdown by their social groups, suggesting that migrants belonging to the lower strata of the caste hierarchy were more likely to lose their job. It is also indicative of the fact that these groups of people are more likely to be employed in casual and daily wage work which renders them more vulnerable to job loss.

Figure 4.

Return Migrants during the Lockdown by Caste.

Source: Field survey, 2020–21. [Color figure can be viewed at wileyonlinelibrary.com]

About 30% of migrants reported retaining their employment with the same employer, however, of these only 3% had stayed back during lockdown at their place of work. The average income loss of households during the lockdown varied by their income quintiles, landholding status and educational attainment of the migrants. Figure 5 show that households falling in the lowest income quintile (Q1) reported the highest income loss during the lockdown. This implies that the poorest households were more vulnerable to income loss as they were engaged in casual or daily wage employment and lost earnings immediately with the imposition of the lockdown. With no earnings and no social securities, they opted for longer routes to reach back their home, thereby incurring huge travelling and other costs (Hindu, 2020), which further added to their burden of income loss. Surprisingly, households in the top income quintiles faced higher income loss compared to the middle income quintile. This is because members of the top income quintile households were likely to be earning more and once the lockdown was imposed and jobs were lost, they suffered a higher loss in terms of the higher salary.

Figure 5.

Average Income Loss by Household by Income Quintiles (in Rs.)

Source: Field survey, 2020–21. [Color figure can be viewed at wileyonlinelibrary.com]

Of the total income loss, the proportion of income loss by landless households was 49%, income loss of households with marginal landholding was 37%, and that of small households were 11%. The share of income loss by large household was merely 3% (Figure 6).

Figure 6.

Share of Income Loss of Household by their Landholding Status.

Source: Field survey, 2020–21. [Color figure can be viewed at wileyonlinelibrary.com]

The average income loss suffered by individuals according to their years of education show that income losses were higher among respondents with less than 10 years of education (Table 3). Low‐educated workers are less likely to work in jobs that allow remote working arrangements, and are less likely to possess sufficient digital skills to be able to take advantage of such arrangements when these are available therefore resulting in disproportionately more income loss for such groups.9

Table 3.

Average Income Loss of Individuals by their Years of Education

Years of education	Average loss of income (Rs.)
0–4	37657.89
5–9	38225.40
10–15	29791.67
>15	19000.00

Source: Field survey, 2020–21.

5. Empirical Estimation

The New Economics of Labour Migration (NELM) by Stark (1991) emphasised the importance of the household as a separate unit in migration decision. Following the NELM theory, we expect vulnerability to differ according to the heterogeneity in characteristics of both individuals and households. The corresponding estimating equation for the household level is given by:

$$\begin{array}{c}\ln Y_i=\alpha +{\beta }_1{\mathit{SID}}_i+{\beta }_2\mathit{\text{Employment}}{\mathit{\text{status}}}_i+{\beta }_3{\mathit{\text{Landholding}}}_i+{\beta }_4{\mathit{\text{Benefits}}}_i+{\beta }_5\mathit{\text{free}}\_{\mathit{\text{foodgrains}}}_i\\ +{\beta }_6\mathit{\text{Agricultural}}{\mathit{\text{Income}}}_i+{\beta }_7{\mathit{\text{Asset}}}_i+{\beta }_8\mathit{\text{Household}}{\mathit{\text{Size}}}_i+{\beta }_9\mathit{Dis}\tan {\mathit{ce}}_i+{\beta }_{10}{\mathit{\text{Caste}}}_i+{\epsilon }_i\end{array}$$ (1)

For the individual level:

$$\begin{array}{c}\ln Y_i=\alpha +{\beta }_1{\mathit{\text{Occupation}}}_i+{\beta }_2{\mathit{Age}}_i+{\beta }_3\mathit{\text{Years of}}{\mathit{\text{education}}}_i+{\beta }_4\mathit{\text{transport}}\_\cos t_i+{\beta }_5\mathit{\text{Food}}\_\cos t_i\\ +{\beta }_6\mathit{\text{Medical}}\_\cos t_i+{\beta }_7\mathit{\text{other}}\_\cos t_i+{\beta }_8\mathit{\text{stay}}\mathit{at}{\mathit{\text{destination}}}_i+{\epsilon }_i\end{array}$$ (2)

where, Y _i is the income loss of an individual or a household during the lockdown, and ε _i is the error term.

5.1. Variable Descriptions

1. Income loss = It is the outcome variable representing the total income loss of an individual or a household during the lockdown. This primarily includes the reported loss of earnings from all sources which were affected due to the lockdown. We have included it in the logged form.

5.2. Household Level Variables

1. Income diversification index (SID) = We have calculated and used the Simpsons Index of Diversity (SID) to estimate the degree of income diversification among the households10. The SID ranges between 0 and 1, with 0 implying specialisation and 1 the extremity of diversification. The more the SID value is closer to one, the more diversified is the household.

2. Employment status = It takes the value 1 if migrant member(s) of the household retained their job during the lockdown and 0 if job was lost.

3. Household Landholdings (in acres) = Landholding is the key economic force driving migration in India (Connell et al., 1976), particularly for those who are landless or have marginal landholdings (Vanwey, 2003). The extent of cultivable land owned by surveyed households is measured in acres.

4. Benefits received = It is the amount of assistance received by migrant households under various central and state government schemes during the year 2019–2020.

5. Free foodgrains = It is a binary variable that takes the value 0 or 1, capturing whether or not households received the benefits of free foodgrain distribution as provided under government schemes during the lockdown period. We study this variable separately from other state sponsored welfare schemes to understand the impact of free food grain distribution by the government during COVID‐19.

6. Agricultural income = It captures the net household income obtained from crop cultivation during the year preceding the survey.

7. Asset index = Asset index is created using 27 different asset indicators for the households. Since the Kiser–Meyer–Olkin test for sample adequacy was found to be 0.80, we employed Principle Component Analysis to generate the asset index. Nine components were found to have eigen values greater than 1 and explained 61% of the variance (see Figure A1).

8. Household size = Number of persons in a household during the survey.

9. Distance from the nearest town (in km) = It measures the distance of the study villages from the nearest town. It is indicative of the available economic opportunities nearby.

10. Caste = This represents the social groups of the surveyed households. Social groups in the state broadly consist of five categories: EBC, BC, SC, ST and others. This categorization is based on the historical, socio‐economic superiority or disadvantaged position of these groups (Desai & Dubey, 2012). Such stratification of the population is likely to affect the economic position of migrant households (Mosse et al., 2005; Keshri & Bhagat, 2010). For regression analysis, we dropped ST as they were very few in numbers in the study villages. The composition of these groups in our study villages is detailed out in Table A1.

5.3. Individual Level Variables

1. Age = Age of the migrant member(s).

2. Years of education = The number of years of education completed by the migrant(s)

3. Occupation = Occupation category of the individual. We divided the workers into three groups, agricultural worker, nonagricultural workers, and salaried.

4. Transport, food, medical and other costs = The costs incurred by the migrant member(s) on transportation, food and medical facilities while returning back to home from their place of work to their home once the lockdown was imposed.

5. Stay or returned = This takes the value 0 if the migrant returned back home and 1 if they stayed back at their place of work.

We tested the possibility of the correlation between study variables, and found the correlations to be lower than the 80% as suggested by Kennedy (2008). The Variance Inflation Factor confirms the absence of multicollinearity among explanatory variables. Finally, the results from Bruesch–Pegan/Cook–Weisberg test necessitated the rejection of the null hypothesis of constant variance (p < 0.05), thereby confirming the presence of heteroscedasticity in the dataset (see Appendix Tables A3, A4, A5). Consequently, we assume strict exogeneity of explanatory variables and to obtain consistent results in presence of autocorrelation and heteroscedastic errors, we undertake an FGLS estimate. Furthermore, each household has its own individual characteristics that may impact the outcome variable (loss in household income). Since we cannot account for all the household specific factors, these omitted variables are likely to be correlated with the variables in the model resulting in omitted variable bias. Therefore, we also report the results from FE estimation, i.e., we assume that there is a correlation between the error term and predictor variables and we control for this. Additionally, the time‐invariant characteristics of the households are unique to themselves, and therefore, the household's error term and the constant (which captures individual characteristics) should not be correlated with the others (Wooldridge, 2002)11. This renders the FE estimates worthwhile to report.

6. Result and Discussions

6.1. Impact on Household Earnings

First, we study the correlates of income loss at the household level (see Appendix Table A3). The distribution of household income across the study villages and the distribution of remittance earning, in particular, is presented below (Figure 7a,b). Household income is quite skewed in the villages like Sirji Barhampur and Bisanpur Bindi. Nanhu has the lowest levels of household income.

Figure 7.

(a) Distribution of Household Income across the Study Villages. (b) Distribution of Income from Remittance across the Study Villages.

Source: Field survey, 2020–21. [Color figure can be viewed at wileyonlinelibrary.com]

Remittances account for about 11% of household income per annum across all the study villages, with the highest being in Adalpur (about 15%). In almost all the villages, the distribution of income from remittances was highly skewed across households.

To identify the household characteristics that determine the extent of income loss during the COVID‐19 induced lockdown, we use the log of total income lost by migrant households as our outcome variable (Table 4). This captures losses from two aspects. First, a household with more number of migrant workers will be suffering higher losses and hence will be more vulnerable compared to others. Second, a household which faces the loss of more number of income generating sources will tend to be more vulnerable than the households with fewer but surer sources of income.

Table 4.

Results of Existing Household Characteristics in Determining Income Loss during COVID‐19

Variables	FE	FGLS
Income diversification index	−0.56**	−0.64***
Employed during the lockdown (Ref: No)
Yes	−0.44**	−0.45***
Landholding (in acres)	−0.11*	−0.12*
Benefits_amount	4.75	−6.12*
Free food grains (Ref: No)
Yes	−0.38**	−0.63**
Income_agriculture	0.85	1.73
Asset_index	0.09	0.02
Number of persons in household	0.01	0.03***
Distance from the nearest town	0.05**	0.05**
Caste (Ref: SC)
Others	0.14	0.09
BC‐1/EBC	0.13	0.14
BC‐2	−0.10	−0.16
Constant	11.06***	11.08***
Wald chi2	26.39*	172.05***
Village fixed effects	Yes

Note: *p < 0.10; **p < 0.05; and ***p < 0.01.

Source: Field survey, 2020–21.

Households' engagement in multiple income generating portfolios is an attempt to create additional ways to increase income and is a survival technique against the susceptibility to external shocks (Ayana et al., 2021). The income diversification index of migrant households shows significant impact on income loss during the lockdown. A household with high degree of income diversification is less susceptible to such income loss. This implies that households with more diversified income sources are able to reduce income instability and livelihood vulnerability due to external shocks. Households with larger landholdings had about 12% less income loss during the lockdown compared to households with smaller land holdings. This implies that larger landholdings enabled households to accumulate wealth and be resilient against any external shock. Additionally, larger landholdings have the possibility of engaging family labour instead of hired labour during the lockdown for cultivation to maintain the same level of output.

If migrant household members remained employed during the lockdown, it decreased the income loss by 45%, implying that those with comparatively secured jobs or possibly those in essential services, who did not lose their jobs during the lockdown and hence were less susceptible. Receiving free food grains during the lockdown reduced the impact of income loss on households by 64%, providing relief to the rural households in general and the marginalised and migrant households in particular who were hit relatively hardest during the pandemic. As stated by Engel's law (1957), households with low‐income levels spend a large percentage of their income on food in order to sustain life, as income increases percentage share of food expenditure declines. Thus, the government's effort to minimise the adverse effects of the lockdown through free food grain distribution cushioned these households from food poverty and the effect of income loss.

An additional member in the household raises income loss by 3%. With larger household size, more working members might have lost employment during the lockdown, resulting in higher income loss for these households. A positive and significant coefficient of distance to town indicates that income loss during the lockdown intensifies as distance increases. Furthermore, the town from the villages, fewer are the economic opportunities available; therefore, in search of better economic opportunities, households send permanent or semi‐permanent migrants to distant towns. Short distance migration is typically exercised by temporary and seasonal migrants (Rajan & Bhagat, 2021), who works in a precarious condition and are paid less than permanent or semi‐permanent migrants. Migrants at distant centers earn a higher income, thereby incurred higher income loss during the lockdown. The distance also involves commuting or transportation costs, as the number of migrant members in the household increases, their transportation cost increases resulting in higher income loss (Hindu, 2020).

6.2. Impact on Individuals' Earnings

Table 5 reports the impact of individual‐level characteristics on income loss during the lockdown. Suspension of almost all economic activities due to the COVID‐19 induced lockdown is expected to impact the earnings of individuals severely. Among the agricultural labourers, non‐agricultural workers and salaried individuals, those employed in the nonagricultural sector suffered significantly more income loss compared to the others. Under normal circumstances, the non‐agriculture sector is known to pay higher wages than the agriculture sector, however, casual nature of employment in this sector makes hiring and firing comparatively easier, with the employer not having any obligation to pay for social security or medical benefits or paid leave to the employees. This makes the workers susceptible, and with the imposition of the lockdown, such workers immediately lost their jobs, thereby losing the ‘higher wage’.

Table 5.

Results of Existing Individual Characteristics in Determining Income Loss during COVID‐19

Variables	FE	FGLS
Job Type (Ref: Salaried)
Agri	−0.54**	0.57**
Non_agri	−1.19***	1.22***
Age	0.03	0.01
Years of education	−0.03*	−0.04**
Transport_cost	0.68**	0.32***
Food_cost	−0.07***	−0.07***
Medical_cost	0.19***	0.018***
Other_cost	0.0001	0.0002
Stay at destination (Ref: No)
Yes	−0.07**	−0.08***
Constant	10.63***	10.66***
Wald chi2	37.53***	111.58***
Village fixed effects	Yes

Note: * = p < 0.10; ** = p < 0.05; and *** = p < 0.01.

Source: Field survey, 2020–21.

As years of education increases, income loss declined by 4%, indicating that individuals with higher years of education are likely to have a more stable job with possibility of social security benefits, thus, resulting in lower income loss during the lockdown than those with less years of education. Migrant workers who stayed back at their destination during the lockdown had to face lower income loss by 8% than those who had to return back. This is because migrant workers who stayed at their destination were likely to be engaged in an occupation that remained functional or kept paying salary during the lockdown. Such workers also saved on spending large amount of money as transportation cost to return back. This is evident from the significant impact of financial hardships endured in terms of high‐transportation cost and food costs by the migrants while returning back to their place of origin. Transportation costs multiplied several folds during the time (Hindu, 2020; Deshingkar et al., 2022) adding to higher‐income losses. Our result resonates with such findings.

7. Conclusion

Using primary data of internal migrant workers from Bihar, this study made an attempt to analyse the consequences of pre‐existing characteristics and vulnerabilities of households and individuals on their income loss during the lockdown. The findings suggest that households with a diversified source of income were at lower risk of losing income during the crisis like COVID‐19. However, households with migrant workers who had secured employment suffered less income loss during the lockdown. Households with larger landholdings and availing of social welfare schemes made households less vulnerable to the loss of income. The association of distance to town with the status of income loss remains positive, indicating greater distance is associated with higher income loss. Similarly, individual level characteristics such as occupation type, education, etc., also played a greater role in economic loss due to the lockdown. From these findings, it may be concluded that the pre‐existing households or individual characteristics were pivotal in reducing or amplifying the vulnerability inflicted by any crisis such as the COVID‐19.

COVID‐19 is more than just a health crisis, in addition to minimising loss of life, it was essential to actively provide financial security to vulnerable migrants and their household members. Although several initiatives were taken by governments to reduce the vulnerability of migrant workers, it faced criticisms because of delay and inefficiency in executions. The Government of Bihar did provided financial assistance, offered a job under employment guarantee programs as MGNREGA, directed the concern departments to take necessary steps for providing trainings to migrant workers who returned back but apparently it was not sufficient to cushion the vast majority of returned migrants from the vulnerability induced by COVID‐19.

Table A1.

Sampling Frame, Census, 2011

District	Village	No. of household	Total population	Caste	Total irrigated land (ha)
Muzaffarpur	Adalpur urf Abdulpur	319	1527	SC: 19%, Others: 81%	58.2
Muzaffarpur	Ajitpur urf Harpur	329	1366	SC: 25%, Others: 75%	44
West Champaran	Sirji Barhampur	263	1745	SC: 10%, ST: 11%, Others: 79%	235.4
West Champaran	Bishnupur raghunath	274	1646	SC: 18%, ST: 0.1%, Others: 81.9%	101.2
Rohtas	Bisi Khurd	153	1189	SC: 17%, Others: 83%	265
Rohtas	Nanhu	157	945	SC: 19%, Others: 81%	148
Sheohar	Bishunpur Bindi	88	412	SC: 22%, Others: 78%	50.6
Sheohar	Azrakbe Pota	113	616	SC: 21%, Others: 79%	22.6

Source: Field survey, 2020–21.

Table A2.

Number of Households Surveyed

District	Village	No. of households
Muzaffarpur	Adalpur urf Abdulpur	54
Muzaffarpur	Ajitpur urf Harpur	54
West Champaran	Sirji Barhampur	54
West Champaran	Bishnupur raghunath	54
Rohtas	Bisi Khurd	51
Rohtas	Nanhu	51
Sheohar	Bishunpur Bindi	32
Sheohar	Azrakbe Pota	32

Source: Field survey, 2020–21.

Table A3.

Descriptive Statistics

Variable	Obs	Mean	SD	Min	Max
Incomelossln	110	10.73	0.58	9.61	12.12
Asset index	162	−0.41	1.93	−4.52	8.43
SID	168	0.43	0.23	0	0.75
Distance to town	168	10.42	3.04	6	16
Employed lockdown
No	126	0.86	0.35	0	1
Yes	126	0.14	0.35	0	1
Free food grains
No	168	0.17	0.38	0	1
Yes	168	0.82	0.38	0	1
Caste
BC‐1/EBC	168	0.33	0.47	0	1
BC‐2	168	0.32	0.47	0	1
Other	168	0.15	0.36	0	1
SC	168	0.19	0.39	0	1
ST	168	0.02	0.13	0	1
Household size	168	7.64	3.18	2	19
Benefits	116	15018.96	27614.33	500	134,800
Ownership landholding	168	1.09	3.37	0	37.5
Income Agri	168	967.69	43183.73	−113,850	404010.19

Source: Field survey, 2020–21.

Table A4.

Correlation Matrix

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
(1) Income_loss	1.000
(2) Employed_covid	−0.256***	1.000
(3) Occupation	−0.109	0.060	1.000
(4) Stayed_destination	−0.048	0.113	0.108	1.000
(5) Age	0.033	0.003	−0.123*	0.056	1.000
(6) Years_education	−0.101	0.081	0.035	0.040	−0.174**	1.000
(7) Transport_cost	0.119*	−0.062	−0.239***	−0.191**	−0.098	0.085	1.000
(8) Food_cost	0.251***	−0.093	−0.191**	−0.264***	−0.162**	−0.079	0.601***	1.000
(10) Other_cost	0.120*	−0.072	−0.199**	−0.112*	−0.131*	−0.046	0.501***	0.724***	0.697***
(9) Medical_cost	0.247***	−0.084	−0.247***	−0.171**	−0.161**	−0.065	0.547***	0.794***	1.000	1.000

Note: *p < 0.10; **p < 0.05; and ***p < 0.01.

Source: Field survey, 2020–21.

Table A5.

Breusch–Pagan/Cook–Weisberg Test for Heteroscedasticity

BP test Value	Degree of freedom	p‐Value	Remarks
13.40	1	0.0003	Null hypothesis of constant variance rejected

Source: Field survey, 2020–21.

Figure A1.

Scree Plot for Principal Component Analysis to Obtain the Asset Index.

Source: Field survey, 2020–21. [Color figure can be viewed at wileyonlinelibrary.com]

DATA AVAILABILITY STATEMENT

The data for ECPA‐2022‐060.R2 ‐ COVID‐19 induced income loss among migrant workers: Evidence from eight villages of Bihar was obtained through field survey funded by Indian Council of Social Science Research. Data may be made available upon permission from ICSSR.