2.1. Sources of poverty

There are four typical sources causing poverty, including regional characteristics, community characteristics, household characteristics, and individual characteristics [20]. The regional sources of poverty are related to country or regional characteristics, which are involved with macroeconomic factors and geographical and natural conditions. Ethnic minorities and regions which are isolated and are inhospitable climatic conditions tend to be easily vulnerable to poverty. Besides, poor infrastructure and lack of public service availability are also correlated with a high probability of poverty. Inversely, urban regions with high population density, the Mekong Delta and coastal regions with developed infrastructure have advantages to overcome poverty.

Regarding household characteristics, household size, employment status of household members, dependency ratio, and the head’s gender are considered significant determinants of household poverty [20]. A larger household size implies a higher efficiency in household expenditure toward the economy of scale. On the other hand, household income relatively depends on the quantity of economically active members in the household. As such, adding these economically active members to the household tends to increase household net income as per capita income is higher while per capita expenditure reduces. As a result, a higher proportion of households can avoid poverty. In contrast, dependent members who cannot work do share household expenditures. However, they do not earn income for the households. Ethically, dependent members are family members and shall be treated fairly like others. As such, the household size and the dependency ratio contribute to household wealth differently [21].

Besides, a recent study from [5] confirms the gender gap in household wealth in Vietnam. These authors also confirm the heterogeneous wealth across different marital statuses. Particularly, widowhood is a special marital status which is not controlled by the couple, while other family statuses, including single, married, separated or divorced, are. WHHs are in an exceptional situation where the head’s partner, who played a significant role in the household, passed away. Losing a member reduces income sources and the efficiency of numerous families resulting in many difficulties in financial balance in WHHs. However [5], did not particularly focus on the difficulties of WHHs.

2.2. The overviews of the economic circumstances of widows

Early studies have shown that widows face a high risk of poverty because of the loss of family wealth after their husband’s death. The death of a spouse can be a devastating and life-altering experience, and widows often face numerous challenges after their loss. These challenges include financial difficulties [7,22–26]. Besides, widowhood leads to higher poverty rates for all women, regardless of pension eligibility or annuity choice for retirement. This is because when a husband dies, much family wealth is lost, including private pension income. Also, the loss of wealth is not always compensated by other resources such as life insurance, and poor widows often had little housing wealth when they were married [7]. In their study regarding the economic situations of middle-aged and older widows [26], found that widows in these age ranges face fewer economic resources after their spouse’s death than their married counterparts. Recent research by [24,25] also confirms the negative impact of widowhood on household income. They conclude that the magnitude of the effect depends on individual and household characteristics as well as social and contextual factors.

A female’s loss of a spouse can also mean a loss of social support and the need to assume responsibilities previously handled by their spouse [27]. Thus, the labour participation of women is significantly affected by widowhood. [28] studies the impact of widowhood on Indian women’s labour force participation using panel data that tracks women before and after the event of widowhood. The results confirm a strong age pattern, where women widowed before the age of 52 saw an increase in the number of days worked, while women widowed after the age of 52 saw a reduction. In addition, widowed women are more likely to be employed in the non-agricultural sector. They have higher incomes than married women if they become the household head after their spouse’s death.

The economic gap between widows and married individuals may be due to the lower levels of financial literacy and experience in financial planning among women in general [25,29]. Therefore, couples need to prepare psychologically and have a clear retirement plan to avoid the risk of poverty after widowhood. Many studies found that the most significant risk of falling into poverty occurs in the initial stage of widowhood [23,25], highlighting the importance of having sufficient resources before retirement to decrease the chance of facing poverty in retired years and after the loss of a spouse.

Vietnamese studies that specifically address the economic circumstances of widows are rare, and typically, when widows are mentioned, it is only as part of a broader analysis of the situation of women. For example [30], includes widow households in his study of Vietnamese FHHs, and [31] mentions them in her study of the impact of lone motherhood on children’s education outcomes. [32] also include widows in their broad study of older persons in Myanmar, Vietnam, and Thailand, finding evidence that widows are more likely to live solo, which makes them more vulnerable to stress, disabilities, and financial pressure. [31]’s study adds evidence to the disadvantaged economic position of widows by showing that widowed lone mothers have a relatively high prevalence of disability, a relatively low level of educational attainment, and a relatively high level of economic activity. However, they mostly work in low-skilled sectors. [31]’s study also shows that children of widowed lone mothers have the lowest level of school performance, as measured by school enrolment and completion rates, pointing to concerning impacts of widowhood on the outcomes of both the widows themselves and their children.

Widows were also an important group included in [33]’s qualitative investigation of gender and land entitlements in northern Vietnam. This study provided more pointers to the possible sources of poverty among widows. It highlighted the barriers faced by widows in exercising their inheritance rights over land and other assets. Scott also found that widows faced barriers in accessing the information on land allocation. Thus, they commonly experienced difficulties acquiring land for production and as collateral to secure bank loans [33–35].

Recently [5,21], approached widowed households as a moderating factor while investigating a gender wealth gap and a wealth return on education in the Vietnamese context, respectively. In these recent studies, widowhood significantly and negatively impacts household wealth accumulation compared to other households with different marital statuses. Particularly, at the upper asset threshold (from 440.25 million VND in 2018) [21], found that wealth accumulation favours single households over widowed households. However, the authors found relatively weak differences in wealth accumulation between the two cohorts. On the other hand [5], found that WHHs consistently stay in the vulnerable group across all net worth quantiles.

To our knowledge, the only prior Vietnamese study with a primary focus on widows is [36], in which the authors use the 2009 census data. This study identified that widows who were relatively old, had a disability, or attained a relatively low education level were more likely to live independently. In addition, ethnicity was also identified as an important correlate of living on one’s own in widowhood, with women in minority groups more likely to be in this situation than other Vietnamese widows.

Our study adds to [36] analysis by focusing squarely on the factors linked to widows’ vulnerability to poverty and expanding the analysis to widows living in FHHs and other household situations. First, we compare the likelihood of poverty for widows and other Vietnamese women. We also assess whether widows who head up their households face different risks of poverty from those who live in other households. Finally, we examine the effect on the risk of poverty of a range of social, demographic and locational characteristics of widow households in Vietnam.

Our literature review indicates that selected social and economic aspects, such as the wealth and welfare of widows, have been investigated in developed countries [25,27,28]. In addition, efforts have been made to propose several solutions to mitigate poverty after widowhood [23,25,29]. However, widowed households suffering from poverty have largely been neglected in the existing literature, especially in Confucian countries such as Vietnam. This research gap warrants the analysis below to be conducted.

Barriers due to traditional Confucian norms prevent women, especially widows, from asset ownership leading to vulnerability. We employ the probit regression for various samples, including the entire sample, a sub-sample of widowed households where at least one widow is living, and a sub-sample of female-headed households to examine the probability of suffering from poverty of the widowed households. Besides, other proxies such as land value and income are also used as the alternative dependent variable to investigate the effects of widowhood on a household’s wealth [5]. We also extend our analysis by estimating the marginal effects of widowhood on several forms of insufficiency, including suffering from food, foodstuff, electricity, water, housing, and clothing. Our analytical approach is discussed in detail in the next section.

3.1 Data

This study utilizes data from the 2018 Vietnam Household Living Standard Survey (VHLSS2018) to examine the economic outcomes of widows in Vietnam and to assess the factors that influence their risk of poverty. The VHLSS has been conducted by the General Statistics Office of Vietnam since 2002 with technical assistance from the United Nations Development Programme and the World Bank [37]. It plays a key role in Vietnam’s attempts to monitor the living standards of Vietnamese households, and it is widely used to inform policy decision-making. Data from the VHLSS has been used to evaluate developmental programs associated with the Comprehensive Poverty Reduction and Growth Strategy, the Millennium Development Goals, and Vietnam’s socioeconomic development goals [38].

3.2 Measuring poverty

There are two typical kinds of poverty in the literature: (i) absolute poverty and (ii) relative or multi-dimensional poverty. Absolute poverty is usually measured by income. In this criteria, the poverty line is a benchmark to identify "a poor household". Besides the poverty line, the bottom 10 or 20 per cent of income is usually perceived as the ’a poor group’. However, assigning a proportion of households whose income is lower than the average or median income within the region is sensitive to personal preference. On the other hand, income-based measurement is typically underestimated.

Regarding the second kind of poverty measurement [39], argue that poverty is essentially the lack of the means to live. The fundamental of poverty relies on a serious problem of what is needed to live "a decent life" and, more critically, what it is to be human. From this perspective, multi-dimensional poverty is more appropriate because it captures many social-economic aspects (including incomes, social needs, and many others) to identify "a poor household" [40–43]. Therefore, measuring poverty by a constant poverty line or a proportion of income is inadequate.

In 2015, the Vietnamese government formally released official documents to identify a poor household by multi-dimensional measurements which are officially announced throughout the Decision number 59/2015/QD-TTg dated 19/11/2015 by the Prime Minister about the multi-dimensional poverty standards. There are two major pillars to assign whether a household is poor: (i) the income-based measurement and (ii) norms on deprivation of access to basic social services. First, regarding the income criteria, the Vietnamese government assigns a poor household based on monthly income. Second, regarding the norms on deprivation of access to basic social services, there are five categories (including (i) health; (ii) education; (iii) housing; (iv) clean water and sanitation; and (v) information), which are measured by ten indices (including (i) access to medical services; (ii) health insurance; (iii) education level of adults; (iv) school attendance of children; (v) housing quality; (vi) average housing area per capita; (vii) residential water sources; (viii) hygienic latrines and toilets; (ix) telecom services; and (x) assets to serve information access). The government also assigns different benchmarks of poverty in rural and urban areas. In the rural area, a poor household is identified if its monthly income is lower than 700 thousand VND or if its monthly income is from 700 thousand VND to 1 million VND, and the household fails in at least three indices of the second criterion as discussed above. In urban areas, the two milestones of monthly income are 900 thousand VND and 900 thousand VND to 1.3 million VND, respectively. The conditions of norms on deprivation of access to essential social services in rural and urban regions are identical. The VHLSS 2018 has a question to identify a poor household: “Have the local authorities classified your household as ‘poor’ in the commune/ward in 2018?”. As such, poor households have been classified by the local authorities. The local authorities have to strictly follow the official guidance from the prime minister’s decision to classify a poor household.

Based on the above ground, the multi-dimensional poverty measurement is appropriate to identify poverty. Besides, the process of identifying a poor household is strictly managed by the government authorities in the Vietnamese context. As such, we consider it appropriate to use the current identification of poor households in the VHLSS for our analysis.

3.3 The analytical framework

In the first part of our analysis, we examine whether Vietnamese households with a widow are more vulnerable to poverty than households that do not include a widow. Our key outcome measure is based on the responses to a VHLSS question: “Have the local authorities classified your household as ‘poor’ in the commune/ward in 2018?” In this sense, the outcome variable is binary, at which the value of “0” is proxied for non-poor households and the value of “1” denotes the poor household.

In further stages of our analysis, we extend the set of outcome variables to include household income and land ownership. We assess the household income and land holdings differences between the widow and non-widow households. These additions allow us to account for the possibility that there may be widow households that are not subject to poverty per se but still experience a relatively high level of economic hardship due to low income/wealth. It is important to identify these risks if they exist, especially when the poverty threshold is so low. Household income is measured as the total labour income of all household members, plus other income from non-labour sources, over the previous 12 months. Labour income comprises salaries, wages, cash, benefits, allowances and in-kind payments from all jobs. Other sources of household income include agricultural activities, non-agricultural activities, gifts and presents. Regarding land ownership, the VHLSS has a range of measures, including the size and value of each land plot owned by the household. In this study, we use the total value of all land plots owned within the household.

In each stage of our analysis, we incorporate controls to account for the likely influence of a range of individual and household characteristics on widows’ chances of poverty. We use VHLSS data on household size, the number of absent members, the household dependency ratio, and information on the household head’s characteristics, such as their level of education, age, injury status, and labour force participation status. These factors have been discussed as determinants of a household’s consumption and welfare by [44].

Following [44], we measure the household size by its equivalent scale adjustment, considering the number of individuals in the household, their age and likely household economies of scale. We measure the equivalent scale adjustment recommended by Deaton and Zaidi (2002): where EA denotes the equivalent-adjusted household size; A represents the number of adults, and C is the number of children under the age of 15 years; α, which ranges from 0 to 1, represents the weighted expenditure requirement of a child relative to an adult, and θ, ranging from 0 to 1, measures the household economies of scale.

[44] argue that raising a child can be less costly in developing countries. The economies of scale can be larger because they spend a larger proportion of income on food (while maintaining a fixed expenditure). Thus, these authors suggest that in developing countries, α shall be low, in the range of 0.3–0.5, and θ shall be high, from 0.9 to 1. In addition, since 2002, Vietnam’s average income has increased substantially, and raising a child has also been more costly. For this reason, we use a value of θ = 0.9, which is less than 1, and α takes a value of 0.5 as the highest weight suggested by [44].

We measure absent household members as members who have not stayed in the household for at least six months. The dependency ratio measures the proportion of all household members that do not participate in the labour market, do not do housework and do not receive pensions, unemployment allowances, one-off severance pays, or an allowance for loss of working capacity the past twelve months. Finally, the injury variable identifies the total times household members have had such a severe injury that they have needed care or had to stop working, studying, or participating in other normal activities in the previous 12 months.

We also include locational variables to help control for spatial heterogeneity [45]. Of particular relevance in the current study is the relatively high rate of poverty in the Midlands and the Northern Mountainous Area, as well as in the Northern and central coastal and Highlands regions. The high prevalence of poverty in the Northern Mountainous Areas is partly due to its geography, which prevents the region from diversifying its economic activities, accessing major markets and improving its infrastructure. In addition, these northern regions are also the home of several ethnic minorities, among the poorest of the poor in Vietnam [46–48].

The regional breakdown of our data is eight-fold, comprising the Red River Delta, the Northern and Southern regions, the Highlands, and the Mekong Delta. An urban/rural categorization of locations identifies cities in Vietnam according to their official ranking, which reflects their function and role, the structure and level of their social-economic development, and their size. The highest-ranked cities (type I) have central roles in the social-economic development of their region and the nation. These cities also have the largest populations (more than 1,000,000 inhabitants), the highest population density (more than 2,000 inhabitants per square kilometre) and the highest proportion of non-agricultural labour (more than 65%). The cities of lower ranking perform less substantial political and economic roles and have lower populations and population densities.

3.4 Research methodology

In the above section, poverty has been described as a dummy variable at which “0” stands for a poor household, and “1” represents a non-poor household. Then, to analyze the vulnerability of widow households to poverty, probit regression and logit regression are the two appropriate methods for dealing with the binary dependent variable. On the one hand, logit regression can perform well regardless of whether the dependent variable is multinomial or binary. On the other hand, the probit regression mainly focuses on the binary dependent variable. In our study, both models are appropriate to analyze the probability of suffering poverty in widow households. As such, the probit regression is used to conduct our empirical analysis as the basis case, whereas logit regression is used as the robustness analysis. The empirical results of this robustness analysis are presented in the Appendix.

Derived from a latent variable model [49], the regression model takes the form:

Where y* is a latent variable, the household is classified as poor when y*>0. Widowhh is a categorical variable identifying whether the household includes a widow, and ε is the error term (assumed to have a standard normal distribution where its cumulative density function is f(x)).

The probability of a household being classified as poor is defined as:

Because the heterogeneity in household and individual characteristics can potentially explain differences in poverty status, we extend the probit model to include measures of these characteristics as control variables.

As noted, the control variables in X′ consist of: measures of household size, the number of absent household members, the household dependency ratio, the head’s age (including a squared term to account for possible non-linearity), the value of the land owned within the household, and locational and educational variables. Educational attainment is measured by years of education, based on the highest grade of education the person completed. Respondents who left school at Grade 12 are assigned 12 years of education, whilst those with a College, Bachelor, Master, or PhD degree are assigned 14, 15, 17, or 18 years of education, respectively. Urban location is measured with dummies for large and medium cities, with the rural location being the reference category. Also included are eight regional variables, with the Red River Delta being the reference category.

Because we are also interested in whether WHHs are more vulnerable to poverty than other widow households, we run a similar probit regression model on the VHLSS data, including variables to identify households headed by a widow (WHH) and, secondly, other widow households. The other elements of the model remain unchanged. To address whether widow-headed households are more susceptible to poverty than other female-headed households (FHHs), we apply the regression model to the sub-set of FHHs.

Our analysis of the correlation between widowhood and household income and land ownership deploys Ordinary Least Square (OLS) regression models, with both household income (hhincome) and the value of land owned by the household (land) logged to enable us to focus on relative differences where zero values are replaced with a number infinitesimally close to zero (e.i. 0.000000001). The models use the same key explanatory variable (Widowhh) as the probit models. The set of control variables is also similar. However, in the model of household income, land value is one of the control variables, whilst, in the land value model, household income is retained as a control variable.

In the final step, we examine the factors that affect widows’ relative chances of being exposed to different forms of poverty. Different probit regression models target food, housing, energy and other forms of insufficiency. The control variables identify household and individual characteristics, including the head’s education, age, injury status, whether he/she is participating in the labour force, and household size, location, and dependency ratio.

Summary statistics on the variables included in the regression models are provided in Table 1. These clearly show a relatively high incidence of poverty amongst WHHs. In all samples, Columns 2 and 3, 10.6 per cent of the households are considered poor. Within the sub-sample of the WHHs, 18.4 per cent of them is considered poor household, as presented in the last column. On the other end, for the widow households where the head is not a widow, results presented in columns 6 and 7 indicate that 16.1 per cent of the households are considered poor. In contrast, for the sub-sample of non-widow households, only 9.4 per cent are considered poor, as presented in columns 4 and 5.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics of Vietnamese households.

https://doi.org/10.1371/journal.pone.0285595.t001

As could be expected, the heads of WHHs are relatively old. The average age of household heads in this group is 64.7 years, while in the broader group of widow households, the average age of heads is 58.5 years. In non-widow households, the average head age is 49.1 years. The head of widow-headed households also has a relatively low level of educational attainment. The average years of education for widows who are household heads are only 4.7, while heads in the broader group of widow households have, on average, 5.9 years of education. On average, the heads of non-widow households have had 7.8 years of education.

Widows who are household heads also have a relatively low average level of labour income. As a result, these households tend to be more reliant on the labour income of other household members. The average annual labour income of the heads of widowed-headed households is 6.5 million VND, and the average earnings of other household members in these households is 29.3 million. In comparison, in non-widow households, the average annual earnings of household heads are 21.4 VND million, and other members’ earnings only reach 29.5 VND million.

Pointing to the possible role of assets in determining the risk of poverty, non-labour income in WHHs is also relatively low (83.8 VND million per annum, compared to 120 million VND in non-widow households). The data in Table 1 show that WHHs appear to be most susceptible to food and foodstuff poverty, with this rate reaching 2.6% and 7.1% compared to the average rate in other households of 2% and 5.6%, respectively. The incidence of other forms of poverty (clothing, electricity, housing) in WHHs is generally similar to that experienced in the other groups.