The Immediate Impacts of COVID-19 on Low-Income Households: Evidence from Malaysia

Abstract

This study unravelled the economic impacts of the coronavirus disease 2019 (COVID-19) on low-income households. The asymmetric economic impacts of the pandemic that are biased towards the poor, young, and women have been well established. However, micro evidence on the poor is limited, thus demanding detailed understanding to design an effective targeted assistance. In this study, data were gathered from face-to-face interviews using a sampling frame provided by the Department of Statistics Malaysia (DOSM). Online data collection was dismissed to ensure all low-income households had the same chance to participate, as some might have no online access. Logistic regressions were estimated to identify the characteristics of households that suffered job loss and income reduction. The findings revealed that one in ten households experienced job loss during the pandemic, while one third survived with lower income. The extent of income reduction was rather severe, as the pandemic had reduced income generation by more than half among the affected households. The regression outcomes showed that the higher-income households among the low-income households had higher chances of experiencing income reduction. A similar scenario was noted for less-educated households. Notably, the adverse impacts were not biased toward female-headed households, as is widely perceived. There was no evidence that economic sectors explained job losses, but households involved in the agriculture, domestic, and transportation sectors had higher chances of suffering from income reduction. These results suggest that monetary government assistance should not rely on general indicators, such as female-headed households and below-poverty-line income (PLI). Instead, a more effective measure is to look at other characteristics, such as employment type, education level, and job sectors.

1. Introduction

The COVID-19 pandemic left a massive economic impact across the globe. However, as suggested by most existing studies using aggregate-level data, the impacts are generally asymmetric, with the socially and economically disadvantaged segments of society—with the low-income, the relatively uneducated, the youth, and female-headed households bearing the brunt. For instance, a survey of 37 countries showed that the pandemic resulted in three quarters of households losing income, with 82 percent of poorer households adversely affected [1]. Single-country studies have reported similar findings in Italy [2], Germany [3], the US [4], South Korea [5], and Japan [6].

In the case of low-income households, they are mostly involved in sectors that require physical proximity, intensive routine, and manual tasks, hence the low possibility of remote working. Economic shutdown has led to either job loss or income reduction in some circumstances. For low-income households, which in most cases are already living on a hand-to-mouth basis, facing either job loss or income reduction affected their ability to afford basic necessities, with some having to change their food consumption patterns. The government launched various assistance programs for those adversely affected to mitigate their sufferings. However, guidance in designing the assistance programs or interventions is generally drawn from studies using aggregate data. The availability of micro-level studies is rather limited. In addition to that, prior studies have often failed to capture well those at the very top and the very bottom of the income distribution, making it difficult to establish the intensity of the impact on each group, particularly the most vulnerable [7]. Hence, assistance programs are broad-brush rather than targeted, although the former is relatively inefficient [8].

Following that, this research aims to focus only on low-income families and examine how the pandemic has affected them. Affected households experienced job loss or have to work with reduced income. Hence, this study examines who among the low-income households suffered job loss and who among them are forced to work with reduced income. The ultimate aim is to ascertain whether the impacts on low-income households are discriminated by certain characteristics or are similar across the board. Understanding the characteristics that differentiate those affected and those who are spared, if there are any, could shed some light on how to design an effective government intervention mechanism.

The contributions of this study are as follows: This study adds knowledge to the field of research by concentrating on the very bottom of the income distribution. While there are various studies on the economic impacts of the pandemic, most are on the population as a whole. Placing focus on a specific segment of society—the economically disadvantaged—enables the extraction of minute details, such as the characteristics of those most severely affected within the studied group. Finer details allow for a more accurate identification of the affected households, hence better targeted assistance measures.

While most surveys on the impacts of the pandemic were carried out online due to the imposed restrictions, the survey deployed in this study was conducted physically. The survey took place immediately after the initial lockdown restrictions were relaxed. A physical survey is more inclusive as it enables researchers to reach households without internet access. This is crucial for studies that focus on low-income households, as most of them might not have internet access. Online surveys might lead to their omission, when they could be the ones most affected.

The remainder of this article is organised as follows: The next section narrates the findings reported in prior studies. Section 3 elaborates the methodology employed in this study. Section 4 presents the empirical findings, and Section 5 discusses the outcomes. Study implications, limitations, and recommendations for future research are elaborated in Section 6, and Section 7 concludes the study.

2. COVID-19 and Economic Impacts: Narrative from the Literature

Studies on the impacts of COVID-19 were thriving well amidst the lockdown being imposed in various parts of the world at different times, as deemed necessary by respective countries. The studies offered conclusive support for the incidence of job and income losses following the COVID-19 pandemic, to varying degrees from one economy to another, as well as from one group of households to another. Most researchers reported that this stemmed from the economic lockdown measures imposed on businesses, such as the restriction of operating hours or mandatory closure for some time. Some businesses had to shut down permanently, thus resulting in employees being laid off or a reduction in working hours that led to loss of income. While some studies, such as [9,10,11], found that job or income loss is more severe with longer and stricter shutdown orders, ref. [5] revealed a similar economic slowdown and higher unemployment rates in South Korea without any lockdown orders. Similarly, ref. [12] asserted that even if no containment measures were implemented, a recession would occur anyway, fuelled by the precautionary and panic behaviour of households, whereas firms needed to face the uncertainty of dealing with a pandemic as well as with inadequate public health responses. In general, findings at the micro level corroborate aggregate observations. This crisis hit the most vulnerable the hardest [1]. However, some exceptions were noted, suggesting that micro-level studies are needed in each society to understand the idiosyncratic impacts.

Disparities of the Effect

Household-level studies on the effects of lockdowns due to the outbreak of COVID-19 revealed the heterogenous nature of the impact based on the characteristics of the households. Factors often cited to explain the disparities of the impacts of COVID-19 are income level, education, type of job, employment sector, gender of the household head, and age.

Most studies showed that low-income households were disproportionately affected by the crisis. Most low-income households were involved in jobs or traditional economic sectors that were badly hit [2,3,6,13,14,15]. In addition, the lower-income households are most likely those that are also financially vulnerable and thus faced a greater risk of losing income, as shown by [16]. However, certain cases, such as in Ghana [17] and Uganda [18], reported that high- and middle-income classes suffered more in terms of the severity of income reduction, which was well reflected in their consumption patterns. In addition, ref. [19] reported that middle-income households in Rwanda were most affected in terms of the magnitude of lost income. Studying the impact using a larger data set from 13 developing countries, the authors of [11] suggest that the impact on income losses and job losses are different if one considers low versus middle income. In terms of job losses, lower-income countries recorded higher disparities, while middle-income households suffer more compared to low-income households, because most low-income households in low-income countries were self-employed in non-farm sectors. On the other hand, in middle-income countries, the impact had a widespread impact on low- and middle-income households, because low-income households in middle-income countries mainly worked in the informal sectors. However, when it comes to income losses, the lower-income countries suffered more for the same reason. Countries with a large proportion of lower-income households depending on remittances could also explain the likelihood of income losses, as found by [20,21]. The outbreak of COVID-19 had an abrupt negative effect on the transfer of remittances, which led to immediate income loss for these households.

The findings on the ability of education to provide job and income protection appear to be mixed as well. Although generally, less-educated workers have a higher possibility of job and income loss [2,3,6,11,22], findings have suggested that economic structure is the deciding factor as to whether education offers job protection. For instance, ref. [23] reported that education is the strongest predictor of job safety during the crisis, with the exception of India. India displayed a different pattern since most jobs available in the country were nonflexible in terms of remote performance. Another study in India by [24] also reported a severe impact in terms of job and income losses among illiterate workers, and the impact lasted longer as compared to more educated workers.

In the context of Uganda, ref. [18] found that households with more educated heads had a higher probability of suffering job and income loss. They were more likely to rely on wage labour, and this group was the most affected in the country relative to farm workers, who still kept working on their own farm despite the lockdown. In Vietnam, ref. [25] stressed that the highly educated households (tertiary) suffered a lot more than those with lower education levels. This corresponds to the emergence of the “new poor” during the pandemic, whereby the highly educated were not spared from the adverse economic shock. In addition, ref. [26] showed that even though lower-educated households suffer more, the assistance that they received from the government provided a cushion to their household economy, a privilege that was not experienced by middle-income households, which resulted in a higher negative impact on them.

Given the different nature of this crisis, which led to atypical sectoral recession, many studies have considered job nature and economic sector as predictors for how households were affected. In this crisis, sectors that operated with a high degree of physical proximity or that had less ability to be performed remotely were hit hard during the economic lockdown. It was concluded by [23] that “the ability to work from home is a key determinant of employment outcomes given widespread shutdowns, mobility restrictions, and social distancing policies”. Jobs with the highest potential for working from home were noted to be in finance, insurance, management, and professional services [8,27]. In addition to that, workers performing high-level non-routine analytical and interpersonal tasks can work remotely, as opposed to those with jobs involving intensive routine and manual tasks. The latter were more likely to face income and job loss [23]. It was cautioned by [23] that the findings might have limited applicability in developing countries given that all studies had been based on data from developed countries. For instance, ref. [3] used data from the US and the UK, while [28] applied data from Italy.

Some studies have identified the economic sectors most affected by the pandemic. A large fraction of jobs in these sectors either cannot be performed from home or are contact-intensive. The hardest-hit sectors were construction, retail trade, and transportation, as well as arts and entertainment [3,29,30]. Generally, the sectors that were relatively protected from shock were primary sectors that provide necessities, such as food, which reflects agriculture. As for the information and telecommunications industry, it possesses a higher ability to be performed remotely and without close physical proximity, so it was spared from the adverse shock. On the contrary, sectors that cannot be performed remotely and require physical proximity (e.g., construction, tourism, hotels and restaurants) were badly affected in all countries. Studies by [11,21,30] revealed that business owners or self-employed persons were the most vulnerable group. This is especially important in countries with a large proportion of micro, small, and medium enterprises.

The impacts on female-headed households have also been mixed. Most studies have found that female-headed households suffered greater income and employment loss due to two reasons. First, female workers are more likely to work in “customer-facing” services, such as the hospitality, leisure, and food industries, which were closed down during the economic lockdown. In the UK, women were one third more likely to be involved in these sectors [31]. Second is the traditional role of caregiver ascribed to women. Women were expected to be in charge of children when childcare centres and schools were closed during the lockdown [1,3]. This discrimination effect was traced by [17] in the case of low-income households in Ghana. On the contrary, ref. [18] found no gender impact on income and job losses in rural Uganda. Instead, the empirical evidence revealed that women increased their labour supply following the pandemic. A similar scenario was noted in Quebec, Canada [32]. In contrast, ref. [33] found that male-headed households in Ethiopia were more likely to be affected in terms of their livelihood due to COVID-19.

In terms of age, studies found that youth-headed households faced higher probabilities of suffering income and job loss [34]. The younger households were more economically unstable during the pandemic when compared to their older peers because they were more likely to be in jobs affected by the crisis, such as leisure and hospitality, and were less likely to have jobs suitable for the work-from-home concept [23]. Although it is generally agreed that the youth suffered the most during this pandemic, micro-level evidence is still rather limited. It was reported by [25] that households headed by those older than 25 years faced a lower probability of income loss, while [32] showed that older households (aged 55–64 years) were more likely to be laid off in Canada. The impact of COVID-19 on the older Thai population was examined by [35], given the fact that the labour market of the country is characterised by a large participation from the older cohort. Since the largest proportion of old-age labour worked in the informal sector, the outbreak of COVID-19 exposed them to the risk of losing their jobs and income. At the same time, two out of five of the older population in Thailand are considered poor, and social assistance may be out of reach due to digital illiteracy. Meanwhile, in the US, ref. [36] reported an increased rate of retirement among workers aged 55 and older after the pandemic. On the other hand, ref. [2] found that age was not a significant predictor for household job and income loss in Italy.

All prior investigations have concluded that this pandemic registered an unequal impact on households and economic sectors, where job loss and earnings are concerned. Most studies concur that the impact is biased towards younger, low-income, and low-educated households. However, household-level studies did paint a different picture; the high-income group in Ghana and Uganda suffered greater impact to the extent that those households altered their consumption, and in Vietnam, the more educated households suffered more. Individual country studies suggest that the disparities of impact from the pandemic depend largely on the pre-existing socioeconomic structure and characteristics of the labour market of the country. This highlights two main points: (1) the need for household-level studies to identify the exact impact, as the general impact might not hold for all societies, and (2) the need to determine whether the unequal impact holds within low-income households.

3. Data and Methodology

This study employed a quantitative research design to determine the research objectives and examine the correlation between the selected variables. Quantitative research is a good way to finalise results and to prove or disprove a hypothesis. This study employed the Statistical Package for the Social Sciences (SPSS) version 28 to perform preliminary and descriptive analyses. The research design also assisted in testing the hypothesised relationships between the independent variables, the intervening variable (mediator), and the dependent variables proposed in this study’s model.

The national census data from the year 2020 were used for this study because it was designed around the pandemic that occurred during that year. Malaysia has a population of 32.66 million citizens and 8.2 million households as of 2020. According to the Malaysian Department of Statistics, the number of low-income B40 households climbed from 405.4 thousand in 2019 to 639.8 thousand in 2020 [37]. After conferring with the Malaysian Department of Statistics (DOSM), low-income B40 households were selected using the National Household Sampling Frame (NHSF) list.

As part of a collaborative research project under the Malaysian Research University Network (MRUN), data were collected among low-income households in Malaysia. Based on statistics gathered from the Department of Statistics Malaysia, the population of Malaysia in 2021 was projected to be 32.7 million, with an annual growth rate of 0.2% [38]. On the other hand, the number of low-income B40 households increased from 405.4 thousand in 2019 to 639.8 thousand in 2020, as reported by [39].

This study adopted the Malaysian definition of low-income households, which refers to the B40 category. This denotes households living with a monthly income of not more than MYR 4850 (USD 1169). Referring to [38], the monthly income range of the M40 category is MYR 4851–10,959 (USD 1169–2642), while those earning more than MYR 10,959 monthly (USD 2642) are classified as T20. Those earning a low income after the first wave of the pandemic were also included in this study, despite being categorised in the middle- or high-income group prior to the pandemic.

Following a consultation with DOSM, low-income or B40 households were identified based on the National Household Sampling Frame (NHSF) list. The current study then adopted a two-stage sampling technique to perform a more in-depth analysis of this demographic framework. Using a multi-stage random sampling technique, 1928 households from five zones in Peninsular Malaysia and East Malaysia were sampled in the first stage (Central, Northern, Southern, East Coast, and East Malaysia). Consequently, six states (Selangor, Johor, Penang, Pahang, Sabah, and Sarawak) were selected from these five zones using a random multi-stage sampling technique. Figure 1 presents a map of the studied areas in order to provide a better understanding of their demographics.

In the second stage, the selected states in each zone were then targeted to obtain households from urban areas based on Enumeration Block (E.B.) and Residential Places (R.P.) through random sampling. In the survey, respondents were heads of households living in urban cities and earning less than MYR 4850 per month. These respondents were categorised as low-income individuals prior to data collection, and the researchers have obtained approval from the university’s research ethics committee.

The physical survey was performed between September 2020 and March 2021 across Malaysia, including East Malaysia (i.e., Sabah and Sarawak). The survey was conducted immediately after the first lockdown (six months) was lifted. This enabled the researchers to capture the immediate impacts of the pandemic on low-income households. Face-to-face interviews were held using a questionnaire approved by the ethics committee of University Putra Malaysia. Six states in Malaysia (out of twelve) were selected to represent the country. The four states chosen to represent the four zones in Peninsular Malaysia are: central region (Selangor), Southern (Johor), East Coast (Pahang), and Northern (Penang), whereas Sabah and Sarawak represented East Malaysia. Sampling was conducted in consultation with the DOSM. Enumeration Block and Residential Places in each state were selected through random sampling. The B40 households were identified using the National Household Sampling Frame (NHSF). The unit of analysis refers to the head of household. In cases where the head of household was not available, the interview involved either the spouse or any adult that could represent the household. Initially, 2125 households were identified and interviewed. After data cleaning, only 1928 households had fulfilled the income criterion.

Demographic Profile of Respondents

Table 1. Tabulation of the profile of respondents from the B40 families. The data display significant heterogeneity within the B40 households.

Profile of Respondents from B40 Families
Variable	Category	Number (Percentage)
Gender	Male-headed households Female-headed households	1309 (67.9) 619 (32.1)
Income group before COVID-19	Less than MYR 2500 MYR 2501–3170 MYR 3171–3970 MYR 3971–4850 More than MYR 4850	1296 (67.2) 225 (11.7) 179 (9.3) 182 (9.4) 47 (2.4)
Employment Status	Employed Self-employed Pensioner Unemployed	1165 (63.4) 340 (18.5) 79 (4.3) 254 (13.8)
Type of Employment	Public Sector Private Sector Self-employed	221 (14.7) 944 (62.7) 340 (22.5)
Employment Sector	Agriculture Manufacturing Construction Services	25 (1.8) 271 (20.1) 103 (7.6) 949 (70.4)
Education level of Household Head	High school or less Matriculation or diploma Tertiary education Others (religious education, etc.)	1398 (74.6) 304 (16.2) 152 (8.1) 21 (1.1)
Age of household Head	Less than 25 26–40 41–60 More than 60	134 (7.3) 599 (32.6) 826 (45.0) 277 (15.1)
Marital Status of Household head	Single Married Divorced	291 (15.1) 1345 (69.9) 289 (15.0)
Household Size	1–2 2–4 5–7 More than 7	503 (27.7) 773 (42.5) 406 (22.3) 136 (7.5)

presents the baseline household information. Most households were headed by males, while about 30 percent were headed by females. Most of the households were composed of three or four members. Most of the heads of households were married. Only 7.3 percent were young households, while the majority were mature households, with the head aged between 41 and 60 years. The sample was consistent with the education characteristics of low-income households, with 75 percent possessing only high school or less education. Overall, 63 percent were salaried workers and 18.5 percent were self-employed. The percentage of self-employed among the respondents was smaller than the expectation. A large portion of the households were involved in the services sector. This also contradicted earlier expectations, as most of them were expected to be involved in the agriculture sector. Most of the low-income households lived with income below the poverty line (67 percent). Concurrently, 47 households were identified as the “new poor”. They were earning more than the B40 income prior to the COVID-19 pandemic but were living with monthly incomes of less than MYR 4850 as a result of the pandemic.

4. Estimation Model

The following equation was applied to examine the effects of COVID-19 on low-income households:

$$EcoEffect={\alpha }_0+{\beta }_{1 }{X}_{1 }+{\beta }_2{X}_{2 }+\delta$$

(1)

where $EcoEffect$ = job loss and income changes; ${\alpha }_0$ = constant; $X_{1 }$ = vector of socio-economic characteristics of households including age, gender, and highest education level of the heads of households, and household baseline income; $X_2$ = head of household’s economic sector; and $\delta$ = residuals.

The variable selection is based on findings from previous studies. The focus variable is the head of household’s economic sector. Following [23], employment outcome is determined by the ability to work from home. Those with work that requires physical proximity, intensive routine, and manual tasks—such as agriculture, manufacturing, and service industries such as restaurants and hotels, as well as fitness and tourism—are expected to be affected. On the other hand, those involved in jobs with the largest potential of working from home, such as finance, insurance, management, and professional services [8,27], are expected to be better off.

The inclusion of socio-economic characteristics of low-income households is based on the obvious findings from previous studies that, in general, disadvantaged households suffers the most [1,8]. The task is to examine whether this applies to low-income households; for instance, do female-headed households have a higher possibility of suffering job or income loss as opposed to male-headed households? The initial hypothesis is that female-headed households, as well as more elderly and less educated heads of households are more likely to suffer from job and income loss. The more elderly and less educated heads of households are usually involved in work with repeated manual tasks; hence, they have less ability to work from home. They are also less resilient in facing economic shocks as it is harder for them to venture into new jobs that employ recent technologies [8].

Two sets of separate regressions were conducted using two different estimation strategies. First, binary logistic regression was deployed in this study to identify the characteristics of households that suffered job loss during the pandemic. Job loss is defined as a binary variable, which is equal to one if the worker lost his/her job and is zero otherwise. For this regression, the sample was restricted to those working before the COVID-19 pandemic only, following [32].

In the case of income change, multinomial logistic regression was executed to assess households that suffered income loss or increment, as opposed to those with unaffected income. This stemmed from a preliminary analysis that examined households with an increase in income during the pandemic. Multinomial logistic regression allows for the inclusion of three categories of dependent variables, in this case, households that are forced to accept income reduction, households whose incomes are not affected, and those with an increase in income. A similar technique was used by [22,40]. The reference category is household whose income was not affected.

5. Findings

Economic Impact of COVID-19

Table 2. Economic impact of COVID-19.

Impact of COVID-19	No	Percent
Job loss	196	10.2
Changed Job	48	2.5
Reduced income	707	36.6
Percentage reduction
80–100	176	24.9
60–79	57	8.1
40–59	139	19.7
20–39	176	24.9
0.1–19	159	22.5
Changed from T20/M40 income bracket to B40 income bracket	47	2.4

summarises the immediate economic impacts of the pandemic on low-income households. Notably, 10.0 percent of households experienced job loss due to the pandemic, while 36.6 percent suffered from income loss due to fewer working hours or pay cuts. Reduction of income among the low-income households in Malaysia is comparable to low-income households in Kenya but is better than in Uganda as the low-income households suffered a 60 percent decline, whereas in India weekly earnings declined by 86.2 percent [6].

In terms of percentage decline, the average decline was 11.5 percent. The distribution of percentage income decline is presented in Table 2. Although the average was relatively low, the overall distribution appeared to be alarming. About fifty percent of the households that reported income reduction had lost more than 50 percent of their income during the pandemic. Besides that, 154 households reported total income reduction. The median for percentage of income decline was 50 percent (38 households). On the other hand, 28.3 percent of low-income households reported an increase in income.

Table 3. Regression output: job losses due to COVID-19.

	I Job Loss (Basic Model)				II Job Loss (Controlling for Employment Sector)				III Job Loss (Controlling for Economic Sectors)
Variables	B	p-Value	SE	OR	B	p-Value	SE	OR	B	p-Value	SE	OR
Female-headed	0.190	0.293	1.104	1.210	0.121	0.508	0.183	1.129	0.200	0.300	0.193	1.222
Age Group Age (more than 60 years old)		0.001	16.981			0.003				0.000
Age (less than 25 years old)	−1.395	0.003	9.036	0.248	−1.270	0.007	0.473	0.281	−1.428	0.003	0.474	0.240
Age (26–40 years old)	−0.717	0.008	6.967	0.488	−0.542	0.054	0.281	0.582	−0.734	0.009	0.279	0.480
Age (41–60 years old)	−1.003	0.000	14.261	0.367	−0.882	0.001	0.270	0.414	−1.051	0.000	0.270	0.349
Education Level Education (other)		0.000	20.833			0.000				0.001
Education (high school and below)	−1.320	0.055	3.683	0.267	−1.185	0.088	0.696	0.306	−1.192	0.089	0.701	0.304
Education (diploma and matriculation)	−2.093	0.004	8.347	0.123	−1.925	0.008	0.731	0.146	−1.951	0.008	0.737	0.142
Education (tertiary)	−2.766	0.001	11.259	0.063	−2.510	0.003	0.832	0.081	−2.504	0.003	0.845	0.082
Income above PLI	0.043	0.801	0.063	1.044	0.136	0.001	0.175	1.146	0.064	0.718	0.177	1.066
Employment Sector Public-sector employee					−2.193	0.126	0.640	0.112		0.302
Private-sector employee					−0.751	0.285	0.491	0.472		0.500
Self-employed					−0.534	0.436	0.499	0.586		0.296
Economic Sector Manufacturing									0.299	0.121	0.289	1.348
Agriculture									0.407	0.250	0.605	1.503
Construction									0.390	0.571	0.373	1.478
Restaurants and hotels									0.655	0.156	0.422	1.924
Transportation and warehousing									0.386	0.077	0.335	1.471
Communications and financial services									−0.603	0.875	1.063	0.547
Domestic services									0.637	0.152	0.448	1.890
Health security defence									−1.004	0.649	0.567	0.366
Administrative support									0.071	0.647	0.451	1.073
Professional technical									0.752	0.672	0.524	2.120
Education									−0.358	0.954	0.786	0.699
Trade and repair									0.145	0.300	0.318	1.157
Business services									0.138	0.000	0.325	1.148
Constant	0.359		0.256	1.432	0.859		0.827	2.360	0.044	0.003	0.760	1.045
Nagelker R-squared	0.059				0.087				0.079
Hosme Lesme statistics	0.413				0.846				0.317

tabulates the regression results of those who lost their jobs during the pandemic among the low-income households. The first set of the regression (1) relates to the results retrieved from basic regression, the second set includes types of employment as control factors, and the last set reports results after controlling for economic sectors. As opposed to the majority of findings [2,3,6], no evidence was obtained in this present study that female-headed households were hit harder than their male counterparts. Despite the consistently positive signs, they were insignificant.

All younger-headed households were more job resilient, as opposed to those headed by those more than 60 years of age. This is ascribed to the nature of jobs that the low-income households were engaged in, such as blue-collar jobs and those that required physical strength. In this case, the employers might be forced to choose the older ones to be laid off first. Apparently, older workers were the first to be terminated during difficult times. This scenario was noted in the UK, where only two in ten self-employed workers could sustain the employment they had prior to the COVID-19 pandemic [1].

The role of education emerged as a highly significant factor, as households with heads possessing more than a high school education faced a lower probability of job loss. The education impact remained even after controlling for the economic sectors. This is consistent with the conclusion derived from most findings that most job losses are concentrated among the relatively lower educated. Even among the low-income households, the higher-educated households had a lower probability of losing their jobs. However, no evidence was found that those with a lower or higher baseline income faced a greater probability of job loss. The economic sectors also did not hold any significant explanation for the characteristics of households that experienced job loss (see

Table 4. Regression output: economic (income) winners and losers from the pandemic.

Decreased Income against Unchanged Income	I Basic Model				II Coping Strategies				III Economic Sectors
Variables	B	p-Value	SE	OR	B	p-Value	SE	OR	B	p-Value	SE	OR
Intercept	−0.914		0.556		1.275	0.074	0.714		1.451	0.043	0.719
Socio-Demographic Variables
Male-headed	−0.104		0.124	0.901	−0.101	0.422	0.126	0.904	−0.148	0.253	0.130	0.862
Age (more than 60 years old)
Age (less than 25 years old)	0.236		0.289	1.267	−0.040	0.895	0.303	0.961	−0.028	0.926	0.304	0.972
Age (26–40 years old)	0.509	0.013	0.205	1.664	0.283	0.194	0.218	1.327	0.207	0.337	0.216	1.230
Age (41–60 years old)	0.135		0.192	1.145	0.006	0.974	0.199	1.006	−0.054	0.786	0.198	0.948
Household size (7 and more)
Household size (1 to 2)	−0.636	0.005	0.228	0.530	−0.657	0.005	0.233	0.518	−0.657	0.005	0.233	0.518
Household size (3 to 4)	−0.248		0.214	0.780	−0.278	0.202	0.218	0.757	−0.315	0.150	0.219	0.730
Household size (5 to 6)	−0.187		0.226	0.829	−0.238	0.303	0.231	0.788	−0.259	0.264	0.231	0.772
Education (others)
Education (high school and below)	−0.526		0.505	0.591	−0.521	0.301	0.503	0.594	−0.423	0.400	0.502	0.655
Education (diploma and matriculation)	−0.771	0.014	0.524	0.463	−0.848	0.105	0.523	0.428	−0.758	0.146	0.522	0.469
Education (tertiary)	−1.367		0.554	0.255	−1.500	0.007	0.556	0.223	−1.438	0.011	0.563	0.237
Income above PLI	0.803	0.000	0.120	2.231	0.835	0.000	0.121	2.304	0.812	0.000	0.121	2.252
Salaried: public-sector employee	−0.193		0.257	0.825	−0.304	0.243	0.261	0.738
Salaried: private-sector employee	1.155	0.000	0.180	3.173	1.020	0.000	0.184	2.774
Self-employed	1.696	0.000	0.199	5.450	1.636	0.000	0.201	5.134
Coping Strategies
Transfer					−0.444	0.005	0.159	0.642	−0.460	0.000	0.158	0.631
Part-time job					1.289	0.000	0.333	3.629	1.356	0.000	0.336	3.881
Changed job					−1.902	0.000	0.440	0.149	−2.046	0.000	0.444	0.129
Economic Sectors
Manufacturing									0.820	0.000	0.184	2.270
Agriculture									1.838	0.000	0.467	6.282
Construction									1.122	0.000	0.256	3.070
Food, beverage, and accommodation									1.497	0.000	0.327	4.469
Transportation									1.662	0.000	0.230	5.268
Communication and finance									0.691	0.156	0.487	1.995
Domestic services									1.780	0.000	0.370	5.928
Health, security, defence and social									0.030	0.923	0.310	1.030
Administrative support									0.503	0.083	0.291	1.654
Professional/technical									1.569	0.000	0.448	4.801
Education									1.297	0.000	0.403	3.657
Trade and automobile repairs									1.445	0.000	0.199	4.240
Business services									0.869	0.000	0.201	2.384
Increased Income against Unchanged Income	I Basic Model				II Coping Strategies				III Economic Sectors
Variables	B	p-value	SE	OR	B	p-value	SE	OR	B	p-value	SE	OR
Intercept	−2.669	0.027	1.206		−0.706	0.613	1.395		−0.290	0.837	1.407
Socio-Demographic
Male-headed	0.313		0.270	1.368	0.357	0.198	0.278	1.429	0.144	0.612	0.284	1.155
Age (more than 60 years old)
Age (less than 25 years old)	2.250	0.000	0.591	9.492	2.262	0.000	0.627	9.601	2.209	0.000	0.627	9.111
Age (26–40 years old)	1.813	0.000	0.539	6.126	1.728	0.003	0.578	5.628	1.506	0.008	0.567	4.507
Age (41–60 years old)	0.944		0.522	2.571	0.893	0.095	0.535	2.442	0.801	0.131	0.531	2.229
Household size (2 and less)
Household size (3 to 4)	−0.612		0.502	0.542	−0.483	0.359	0.526	0.617	−0.363	0.493	0.528	0.696
Household size (5 to 6)	−0.257		0.476	0.774	−0.124	0.804	0.500	0.883	−0.093	0.852	0.501	0.911
Household size (7 and more)	−0.064		0.506	0.938	0.022	0.966	0.531	1.023	0.038	0.943	0.535	1.039
Education (other)
Education (high school and below)	−0.638		1.074	0.529	−0.725	0.500	1.075	0.484	−0.892	0.409	1.081	0.410
Education (diploma and matriculation)	0.007		1.107	1.007	−0.280	0.801	1.110	0.756	−0.484	0.664	1.117	0.616
Education (tertiary)	0.122		1.127	1.130	−0.217	0.848	1.134	0.805	−0.418	0.716	1.147	0.659
Income above PLI	−0.855	0.012	0.339	0.425	−0.765	0.027	0.345	0.465	−0.906	0.009	0.348	0.404
Salaried: public-sector employee	−1.239	0.014	0.506	0.290	−1.228	0.018	0.518	0.293
Salaried: private-sector employee	−0.752	0.018	0.317	0.472	−0.893	0.008	0.339	0.410
Self-employed	−0.202		0.393	0.817	−0.148	0.713	0.403	0.862
Coping Strategies
Transfer					−0.021	0.955	0.370	0.979	0.058	0.873	0.367	1.060
Part-time job					2.328	0.000	0.420	10.254	2.368	0.000	0.424	10.674
Changed job					−2.076	0.001	0.634	0.125	−2.429	0.000	0.659	0.088
Economic Sectors
Manufacturing									−1.185	0.035	0.563	0.306
Agriculture									1.329	0.107	0.826	3.779
Construction									−0.764	0.321	0.771	0.466
Food, beverage, and accommodation									−0.181	0.817	0.783	0.834
Transportation									−0.095	0.864	0.554	0.909
Communication and finance									0.051	0.952	0.857	1.052
Domestic services									−0.171	0.875	1.083	0.843
Health, security, defence and social									0.082	0.874	0.515	1.085
Administrative support									−2.260	0.037	1.083	0.104
Professional/technical									1.447	0.031	0.670	4.251
Education									0.101	0.890	0.726	1.106
Trade and automobile repairs									0.114	0.803	0.456	1.120
Business services									−0.988	0.041	0.483	0.372

).

Multinomial logistic regression was estimated to determine the characteristics of households that either suffered income loss or experienced an increase in income, relative to those with unchanged income. Ordered logistic was excluded because the model failed to meet the parallel regression assumption. The multinomial regression results are illustrated in Table 3. The first column reports the baseline regression, and the second reports regressions controlling for coping strategies. The third column presents results controlling for the economic sector, while the last column reports results combining coping and economic sectors. The odds ratio (OR) and p-values are reported as well.

A significant relationship was noted between head of household’s age and income change during the COVID-19 pandemic. Households with younger head (26–40 years) had higher chances of experiencing income reduction during the pandemic (OR₁ = 1.664, p₁ = 0.000). However, the significance diminished after controlling for both coping strategies and economic sectors. Simultaneously, younger households, those headed by persons aged less than 25 and between 25–40, had higher chances of experiencing income increment. The odds for households headed by persons aged less than 25 years remained 9 times higher than those aged 60 and above (OR₁ = 9.494, p₁ = 0.000; OR₂ = 9.601, p₂ = 0.000; OR₃ = 9.111, p₃ = 0.000). The odds for households headed by persons aged 26–40 years were slightly lower (OR₁ = 6.126, p₁ = 0.000; OR₂ = 5.628, p₂= 0.000; OR₃ = 4.507, p₃ = 0.000). Unlike other studies that reported younger households as being more economically vulnerable during the recent crisis, the present study revealed that younger low-income households had better chances of increasing their income, possibly through various part-time jobs (e.g., delivery jobs) that emerged during the crisis.

Smaller household size (single or married without children) was associated with lower chances of experiencing income reduction, but the odds were low (OR₁ = 0.530, p₁ = 0.000; OR₂ = 0.518, p₂ = 0.000, OR₃ = 0.518, p₃ = 0.000). Similarly, households headed by those with tertiary education consistently had lower chances of suffering from income reduction (OR₁ = 0.255, p₁ = 0.000; OR₂ = 0.223, p₂ = 0.000; OR₃ = 0.227, p₃ =0.000), with relatively low odds. Findings on income were rather interesting: those living above PLI had higher odds of suffering from reduction of income, as opposed to those earning lower than PLI by twice, relative to those whose income remained unchanged (OR₁ = 2.231, p₁ = 0.000; OR₂ = 2.304, p₂ = 0.000, OR₃ = 2.252, p₃ = 0.000). In a similar vein, ref. [18] found that for those already earning very low income, income would not fall by much. They also recorded lower chances of experiencing an increment in income (OR₁ = 0.425, p₁ = 0.000; OR₂ = 0.465, p₂ = 0.000; OR₃ = 0.404, p₃ = 0.000). The upper segment of B40 lost out during the crisis, whereby they faced a greater possibility of income reduction and lower chances of income increment. Both cases are in comparison to those whose income remained unchanged.

The findings on employment types were as expected. Those employed in the public sector were very unlikely to experience income increment (OR₂ = 0.290, p₂ = 0.001; OR₃ = 0.293, p₃ = 0.001). This is not surprising given the fixed remuneration for government servants. Those working in the private sector were also very unlikely to gain additional income during the crisis (OR₂ = 0.472, p₂ = 0.001, OR₃ = 0.410, p₃ = 0.000). However, those working in the private sector faced higher chances of suffering income reduction (OR₂ = 3.173, p₂ = 0.000, OR₃ = 2.174, p₃ = 0.000). In comparison, the odds of private sector workers experiencing income reduction were higher. The public-sector employees were better off relative to their private-sector counterparts. Those that were self-employed had higher odds of suffering from income loss by a factor of 5.450 (p₁ = 0.000), while the factor reduced marginally to 5.143 (p₂ = 0.000) when coping strategies were controlled.

In terms of coping strategies, households that received money transfers from children and changed jobs during the pandemic had a lower possibility of experiencing reduction in income. Nonetheless, those who worked part-time jobs still had a higher probability of experiencing a decrease in income (OR₂ = 3.629, p₂ = 0.000; OR₂ = 3.881, p₂ = 0.000) compared to those who did not, relative to households with unchanged income. This is attributed to the reduction in the main income source being large and income from moonlighting also failing to compensate for the lost income. On the contrary, a higher possibility was noted for those who moonlighted, as opposed to those who did not, to experience an increment in income (OR₂ = 10.254, p₂ = 0.000; OR₃ = 10. 674, p₃ = 0.000) in comparison to those whose income remained unchanged. Notably, those who changed jobs were associated with a lower possibility of experiencing increment in income (OR₂ = 0.149, p₂ = 0.000; OR₃ = 0.129, p₃ = 0.000). This suggests that job changing was undertaken as a coping strategy, not to earn extra income.

The findings revealed that the economic sectors significantly affected income change. Most economic sectors were linked with a high possibility of income reduction, except finance and communication, health, security, and social services, as well as administrative support services. The highest possibility of income reduction was related to the agriculture sector (OR₃ = 6.282, p₃ = 0.000), and this was followed by domestic services (OR₃ = 5.928, p₃ = 0.000). Similarly, ref. [41] reported that in Malaysia, most of the workers affected during the COVID-19 pandemic were those involved in the agriculture sector. This outcome is more pronounced in this study, as most of the workers in the agriculture sector earned a low income. On the other hand, workers in professional and technical services had a greater chance of experiencing increment in their incomes (OR₃ = 4.251, p₃ = 0.000). The findings on economic sectors and income change concurred with findings from [27].

6. Discussion

This study had assessed the immediate economic impacts of COVID-19 on low-income households or the B40 segment of society. This study investigated two pronounced economic impacts: job loss and change in income. About 10.2 percent of the households were laid-off, and 36.6 percent had to continue living with the lower income. In comparison, a study by the International Labor Organisation (ILO) on households, irrespective of income, revealed that the laid-off rate was 1.5 percent, and households that suffered from income reduction were 5.5 percent. This verified the findings of other studies that the low-income group is indeed the most affected social segment. However, the incidence paled in contrast to India, in which the low-income group suffered by as much as 72–86 percent [6], while in Uganda, a 60 percent reduction of income was reported [18]. The extent of the reduction suggested a rather unpleasant outlook. More than half of the households that were experiencing income reduction earned income less than half of what they used to receive prior to the pandemic. Given that they were already from the low-income bracket, a reduction of half or more than half of their initial income denotes a dire living state. Interestingly, a small fraction of the low-income households experienced higher income following the crisis. Some low-income households were resilient in facing the pandemic. Within the short term of the pandemic, the emergence of the “new poor” was noted. In fact, 2.4 percent of the households within the B40 category were initially earning either M40 or T20 income prior to the COVID-19 pandemic.

As for income group, no significant evidence revealed that baseline income had significantly predicted those who lost their jobs. Those earning above PLI (>MYR 2500) but who were still categorised under the B40 income group (maximum of MYR 4850) had a higher possibility of suffering from income reduction. Concurrently, there were fewer chances for this group to earn more during the crisis. This concurs with the experience of low-income households in Uganda [18] and Ghana [17]. This finding supports the policy options devised by the government of Malaysia, which indicated that all households within the B40 income bracket should be assisted. Government assistance should not be limited to households living under PLI only, as it is the upper segment of the B40 that were hit harder by the pandemic.

Among the low-income households, those headed by persons with tertiary education had a lower probability of losing their jobs. Notably, possessing tertiary education significantly reduced the possibility of households suffering from adverse income changes. Taken together, this signifies that education matters in protecting jobs during the pandemic. Despite being categorised in the low-income group, those who were highly educated did earn a stable income during the pandemic. This scenario is similar to findings in other studies, which have shown that this pandemic disproportionately affected the less educated (see [2,3,6]), given that the less educated were concentrated in certain jobs only, such as blue-collar work and the services sector, which were mostly affected due to the lockdown.

Gender did not matter in terms of job loss or income change during the crisis. This is comparable to findings from other household-level studies by [18] in Uganda and [32] in Canada. No evidence suggests that female-headed households suffered more than male-headed households did, as evidenced in most studies (see [2,3,42,43]). In terms of specific gender concentration in industries, there was no indication that females were more concentrated in sectors that had been badly hit compared to males. The statistics pointed out that women and men were both highly engaged in manufacturing, as well as the wholesale and retail trade domains. While the employment trend for women is to be hired more in the food and accommodation industries, men are hired mostly in automotive repair [44]. Both sectors were affected during the pandemic, hence the possibility for female-headed households to not suffer less in terms of job or income loss.

Mature households headed by those aged from 41 to 60 years were also relatively safe from facing job loss. This group, however, was associated with a higher possibility to suffer income reduction. Hence, they had a better chance to hold on to their jobs but with reduced income. As for the younger households with heads of households aged up to 40 years old, they had greater chances of enjoying an increase in income during the pandemic. Younger households were the least likely to possess the ability to withstand income reduction, given that they might have a relatively lower financial buffer and student loans [45]. This could also be attributed to their motivation to grasp any chance to generate income during the pandemic.

It is undeniable that this pandemic did offer new job opportunities in areas related to online retails, e-commerce, and healthcare, as these sectors were considered crucial during the pandemic [1]. According to [44], those involved in food sector services, information and communication, as well as health-related sectors earned higher incomes, though only marginally, by 0.5 percent, 0.6 percent, and 1.0 percent, respectively. Being more technologically literate and relatively more energetic, younger adults are better able to take up these jobs. However, as this survey collected data at one only data point, it is not known if the increase is permanent and stable, or merely temporary. The point here is that only a fraction of young low-income households that reported an increase in income during the pandemic were observed within the short time frame of this study.

The economic sector appeared insignificant in explaining job loss, but it mattered for income reduction, at least after the first lockdown was lifted. This is ascribed to the helpful assistances offered by the Malaysian government, particularly to SMEs in terms of wage subsidies. This ensured that the firms could afford to keep their workers on the payroll. Besides, some firms might have had to resort to certain measures, such as wage reduction, indicating the significance of economic sectors with regard to income change. In terms of income change in relation to the economic sector, those involved in the agriculture, domestic services, and transportation segments faced the highest probability of suffering income reduction. This evidence dovetails with the findings reported by ILO [46] on sectoral impacts of the pandemic. The sectors categorised as highly affected in the Malaysian economy were accommodation and the food industry, retail trade and repairs, as well as the transportation domain. As these sectors were affected, the workers faced higher chances of being laid off or working with reduced income.

Implications and Limitations

The findings from this study highlighted a few implications that could be useful for the design of government assistance for low-income households in facing the consequences of the pandemic. Firstly, the economic impact of job loss and reduction in income is widespread, and the findings revealed that those who are in the upper bracket of the low-income group suffered the most in terms of income reduction. Hence, assistance should not be limited to those living under PLI.

Secondly, the prevalence of households who have to continue living with lower income is greater than those with job loss. The rate of income reduction is quite substantial. This indicates that assistance should also look at income reduction rather than job loss per se as reduction in income could also affect households’ standard of living, especially given that these households are in the low-income bracket.

Thirdly, the role of education stood high in discriminating between households who suffered job losses and those who did not. The higher-educated heads of household had a higher possibility of retaining their jobs during the pandemic. This highlighted the need to formulate assistance not only in monetary form, but also using other mechanisms to ensure that children from low-income households received much-needed education. In addition, the learning and education process had been disrupted during the pandemic. Online education did not reach the children from low-income households well due to the lack of gadgets and internet access. Findings from this study suggest that education is an important avenue to assist low-income households in facing economic shocks.

The limitation of this study is that it was carried out in-between the two massive lockdowns in the country. Hence, the findings only captured the immediate economic effects of the pandemic. Nonetheless, the findings can still serve as a benchmark of the short-run economic impact and identify the characteristics of low-income households that were affected.

Future studies should be conducted to capture the economic impact of the prolonged crisis on low-income households, especially with respect to its severity. The effect of the crisis is expected to linger since the economy takes time to recover. In certain sectors, the adverse impact might get more pronounced in the long run; hence, more job loss or income reduction might follow. There will also be changes in certain economic structures, such as the adoption of higher technology, that could pose restrictions on lower-income households. It is also interesting to examine the impact on low-income households from the consumption perspective, that is, whether they suffer to the point that they have to alter their consumption, such as reducing food quality and quantity, and forfeit their children’s education in order to accommodate basic needs.

One aspect that should also be considered is to look at the role of institutional quality as proposed by [47] since the role of government played a crucial role during the lockdown and recovery periods. This is especially true when taking into account the long-term effects of COVID-19 on poverty, which require not only proper intervention but also implementation and monitoring of the intervention programs.

7. Conclusions

This study unpacked the heterogeneous nature of low-income households. In contrast to the typical generalisations of the disproportionate impacts of the pandemic, this study revealed that within the low-income households, some were either unaffected or even winners during the pandemic in terms of income and job security. Notably, those living under the poverty line were the ones most affected, instead of those who were earning greater than PLI. In addition, there were also the “new poor”, who suffered from massive income shock due to the pandemic. Interestingly, no empirical evidence suggests that female-headed households suffered more than their male counterparts. Therefore, monetary government assistance should not be based on general indicators, such as female-headed or below PLI. As monetary assistance is supposed to help cushion the impact of the pandemic on income, a more effective measure is to look at the characteristics such as type of employment, education level, and involvement in job sectors.

The findings of household-level studies paint a different picture regarding the impacts of the COVID-19 pandemic on low-income households as opposed to the generalisations made from aggregate data. For instance, it is not necessarily the female-headed households, the younger households, or the lowest income quintile that had been laid-off or suffered income loss during the pandemic. This highlights the contribution of micro-level household studies to better comprehend the impacts of the pandemic on a specific segment of society.

Education, which remained a protection for job and income safety, emerged as a crucial policy prescription. In the long run, it is imperative to ensure that children from low-income households have access to proper education. This is because there is a very high probability that their education could have been disrupted by the pandemic. Poor access to equipment and a conducive environment might have hindered their learning during the pandemic, thus leaving a scarring effect on their future.

One caveat applies: the findings reflect the short-term impacts of the pandemic. The emergence of the second wave of COVID-19 led to various restrictions being re-imposed in early 2021, including a second lockdown in June 2021. This suggests a high possibility that this prolonged pandemic, together with the deepening of its severity, led to greater economic difficulties for the low-income households. Nonetheless, the study outcomes may serve as a benchmark in comprehending the impacts of the COVID-19 pandemic, specifically on low-income households in Malaysia.