Landlessness, Economic Activities and Household Income in the Red River Delta, Vietnam

The main aim of the current study is to investigate the inuence of landlessness and landholding on the choice of livelihoods among rural households in the Red River Delta. Among ve livelihoods adopted by local households, we nd that the highest income derives from formal wage earning, the lowest from agricultural and informal wage-paying livelihoods. The middle income group comprises livelihoods based on formal wage-paying jobs and other sources, and nonfarm self-employment and other income. Notably, the study provides evidence that landlessness or land shortage is not a potential barrier preventing rural households from pursuing gainful livelihoods in the Red River Delta. Specically, households affected by landlessness or a shortage of land tend to adopt non-farm livelihoods that are more protable than agricultural livelihoods. The nding suggests that landlessness or shortage of land should not be viewed as an absolutely negative phenomenon in the region.


Introduction
Over the past three decades, Vietnam has transformed itself from an agricultural into an industrializing economy. This transformation has had the result that Vietnam, at one time one of the world's poorest countries, became a lower-middle income country by 2011. This structural transformation has emerged not only as a key driver of change in the economy but also as a major factor contributing to poverty alleviation in Vietnam (López Jerez, 2019). The initial reduction in the incidence of poverty was the result of the improved earnings of farming households. Given that 70% of the population was engaged in agriculture in 1990, agricultural growth is re ected in the rural poverty rates, from 70.9% in 1993 (López Jerez, 2019) to 18% in 2006 and 9.6% in 2018 (GSO, 2018).
While rice is the main staple and is planted everywhere Vietnam, it is mainly produced in the two "rice bowls," the Red River Delta (RRD) in the north and the Mekong River Delta (MRD) in the south. The rapid conversion of agricultural land to urban use has taken place unevenly across regions, and is found mostly in the peri-urban areas of the RRD and MRD, leading to a large decrease in the paddy rice area (Van Dijk et al., 2013). Consequently, this process has had a major effect on the livelihood of local people. On the one hand, the loss or decrease of arable land (due to urbanization) might have a negative effect on agricultural production and income. On the other hand, rapid urbanization in the peri-urban areas provides farmers with a wide range of non-agricultural opportunities, thereby enabling them to change their livelihoods and improve their economic well-being (Tran, 2014).
In Vietnam, numerous studies have examined the consequences of landlessness for rural households (Ravallion & Van de Walle, 2008)  Also, a study by Hoang et al. (2019) revealed that a lack of annual cropland is a key factor for adopting a pro table livelihood in the MRD, while annual cropland is found to have a negative link with household income in the Highland Central region .
The literature suggests that most empirical studies focus on the direct impact of cropland on household well-being, while only a few investigate its effect on the choice of livelihood in rural Vietnam. Also, to the best of my knowledge, no study exists for the Red River Delta, where cropland size per capita is the lowest, so that most rural households have diversi ed their livelihoods towards non-farm activities over the past decades (GSO, 2018(GSO, , 2019. This gap in the literature inspired us to conduct the current research to answer two main research questions. First, what kinds of livelihood are adopted by local households? Second, are landlessness and landholdings the main factors affecting the choice of high-return livelihoods in this region? Using the sub-sample of the Red River Delta from the Vietnam Household Living Standard Survey 2016, we nd that rural households adopted ve livelihood strategies, namely formal wage employment (FEW), FEW and other income sources (OIS), nonfarm self-employment (NSE), agriculture and informal wage employment (IWE). Also, both descriptive and inferential statistical analyses con rm that households that pursued agricultural or informal wage employment livelihoods, earned lower income per capita on average than those adopting any of the other remaining livelihoods. Notably, we nd that households characterized by landlessness or land shortage are more likely to adopt high income livelihoods. This suggests that the lack of access to land can act as a push factor while higher returns from some non-farm activities can operate as a pull factor, combining to lead to a household's choice of livelihood.

Land reform
Agricultural de-collectivization was initiated in 1981, when farming households were allowed to sell their agricultural products after contributing a required amount of produce to the state. Then, in 1988, the adoption of Resolution 10 removed most features of collective production, transferring land use rights (LUR) from collectives to individual households, enabling farmers to lease land from the state for up to 20 years and thus securing their land tenure (World Bank, 2016). Also, land allocation from collectives to households and individuals was made subject to two main criteria: (1) the number of family members, and (2) the quality of land in terms of irrigation, distance among plots and other farming conditions (H. Nguyen, 2014

Background of the region of study
Our study focuses on the Red River Delta (RRD), which is a at plain in the northern part of Vietnam. It is surrounded by mountains in the east and west and by hills in the north. The Delta is created by recent alluvium deposited by the two main rivers, namely the Red River and its distributaries, and the Thai Binh River. The Delta is characterized by a slight slope from the northwest to the southeast, from 15 m down to sea level, and by a quite unpronounced topography, including the alternation of natural levees and depressions (Devienne, 2006). The RRD has long been a densely populated wet-rice cultivation area (Labbé, 2019), and is often referred to as the "cradle of Vietnamese civilization" (Seto, 2005).
The region includes ten provinces, namely Bac Ninh, Ha Nam, Ha Noi, Hai Duong, Hai Phong, Hung Yen, Nam Dinh, Ninh Binh, Thai Binh, and Vinh Phuc. It covers 21000 km 2 and is home to 21 million people, with a very high population density of 1090 people/km 2 on average (Labbé, 2019). Annual cropland covers about 41.7% of the RRD's total area, slightly less than that of the Mekong River Delta (MRD) (43.5%) but much higher than that of other regions. Also, the region had the largest area of irrigated annual cropland (93.7%), followed by the MRD (92.3%), the North Central and Central Coastal regions (73.1%), the Southeast (66.5%), Central Highlands (54.4%) and Northern Midland and Mountain regions (44.2%). The region is endowed with good human capital, with 32.4% making up its trained labour force (15 years of age and above) in 2019, higher than the average level for the whole country (22.8%) and that of all other regions (GSO, 2019).
The o cial data show that over the 2008-2018 period, the RRD always attained the second highest income and second lowest poverty levels, just behind the Southeast region (GSO, 2018). Speci cally, in 2018 the monthly household per capita income for the RRD was 4.775 million Vietnamese dong (VND), lower only than that of the MRD (5.526 million VND) but much higher than the average level for the whole country (3.874 million VND) and that of all other regions.
Also, the poverty rate in 2018 was much lower in the RRD (1.9%) than in all other regions, except for the Southeast (0.6%) (GSO, 2018).

Dataset
Our study utilizes a sample of 6227 households in the RRD, drawing on secondary data from the 2016 VHLSS. The dataset contains rich information on various characteristics of households and their family members, and includes details about demography, education, employment, income activities, housing and expenditure, and ownership of various types of land. Data from multiple les were collapsed, re-organized, merged and nally combined in one le that could be used for both descriptive and inferential statistics.

Analytical methods
First, we use descriptive statistics for summarizing the main characteristics of households, including various parameters such as mean, median, standard deviation, and frequency. We then employ cluster analysis to categorize households into various clusters or groups of livelihoods and use ve income sources (measured by percentage) as input variables for this analysis, namely (1) agricultural income (cultivation, shery and livestock, forestry); (2) formal wage income (wage-paying work with a formal labor contract); (3) informal wage income (wage-paying work without a formal labor contract); (4) nonfarm selfemployment income (non-agricultural self-employment activities) and (5) other sources (remittances, public/private transfers, rentals and interest, etc.).
Our cluster analysis follows two steps. First, the optimal number of clusters was determined via the Calinski-Harabasz pseudo-F stopping-rule index (Halpin, 2016), which showed that the highest pseudo-F value was 4616.67 for the ve-cluster solution. We then used the K-mean cluster to identify households according to mutually exclusive livelihoods. Some main features of the ve livelihoods are given in Table 1. household characteristics, such as household size, dependency ratio, the age, gender and education of the heads of . The landlessness of is denoted by a dummy variable. includes three socio-economic commune variables, namely the rates of formal wage employment, informal wage employment and nonfarm-self-employment at the commune level, and nine provincial dummy variables controlling for provincial xed effects. A description and de nition of these variables are provided in Table 2.  Table 1 describes the income structure of ve livelihoods that were classi ed by cluster analysis. It shows that 29% of the total household sample pursued a livelihood from nonfarm self-employment and other sources, followed by those with an informal wage livelihood (20%). The proportion of households adopting a formal wage-paying livelihood is 18%, which is equal to the number of those pursuing a livelihood from formal wage-paying jobs and other sources, while those following an agricultural livelihood account for 15% of total households. On average, formal wage income makes up about 26% of total household income, followed by other income (20%), agricultural income (20%), informal wage income (17%) and nonfarm self-employment income (16%). On average, formal wage-earning employment contributes about 84% of total household income for those with a formal wage-paying livelihood. This employment provides about 49% of total household income among households with a livelihood from formal wages and other sources, and the other remaining sources contribute 51%. About 40% and 46%, respectively, of total household income, on average, derive from nonfarm self-employment and other income among those living from nonfarm self-employment and other income. Agricultural activities contribute about 70% while the remaining sources account for 30% of total income for those with an agricultural livelihood. Finally, for those with an informal wage-earning livelihood, an average of 68% of their total income was earned from informal wage-paying work and 15% from agriculture, whereas 17% came from other sources. Table 2 provides some of the major characteristics of households by livelihood. On average, 78% of total households are headed by men. The gure is higher for households whose livelihood comes from formal wages and other sources (84%) and lower for those living from nonfarm self-employment and other sources (69%). The average age of household heads is about 54.25 years for the whole sample. However, the average age is highest for those with a nonfarm self-employment livelihood (59 years) and lowest for those with an informal wage-earning livelihood (49 years). The average number of family members is 3.45 for the whole sample, 4.06 for households living from formal wages and other sources, but only about 3.00 for those living from agriculture or nonfarm self-employment and other income. The average dependency ratio is about 0.45 for all households, but the highest gure is 0.55 for those living from nonfarm self-employment and other income, whereas the lowest is 0.31 for those adopting an informal wage-paying livelihood.
Regarding the education level of household heads, the data in Table 2 show that for the whole sample, each household head had attained an average of about 9 years of formal schooling. The highest level of education among household heads is recorded for households adopting a formal wage-earning livelihood (10.81 years), followed by those with a livelihood from formal wage earning and other sources (9.64 years), while the lowest level of education is found among those whose livelihood depends on nonfarm self-employment and other income (8.03 years). The average size of annual cropland is about 1584 m 2 per household for all households. However, the largest and smallest average size, respectively, is found for those living from agriculture (2801 m 2 ) and livelihoods deriving from nonfarm self-employment and other income (1090 m 2 ). Households depending on an agricultural livelihood also owned more perennial cropland and water surfaces than did those in other livelihood groups. On average, about one fth of all households do not own any annual cropland. The percentage of households without annual cropland is found predominantly among those with formal wage-earning livelihoods (37%) and with livelihoods from nonfarm self-employment and other income (28%). Source: Author's estimations from the 2016 VHLSS. a, b, c the proportion of households with at least one member engaging in formal wage-earning employment, informal wage-earning employment and nonfarm self-employment at the commune level, respectively.    Table 4 compares the differences in per capita income across livelihood groups using the Bonferroni method for pairwise multiple comparisons. The second column shows that the income gap is negative and statistically signi cant (at 10% or lower levels), con rming that all other livelihoods secured lower income than did a formal wage-earning livelihood. Also, the results in columns 3 and 4 indicate that those living from agriculture and informal wages earn less, on average, than do those choosing the remaining livelihoods. Finally, we nd no statistical evidence for the income difference between those in agricultural and those in informal wage-earning livelihoods. The results in Tables 3 and 4 suggest that a formal wage-earning livelihood is the highest-earning strategy while the lowest return is found for agricultural or informal wage-earning employment strategies. The results also suggest that shifting from an agricultural livelihood to any other would signi cantly improve household income, except for an informal wage-earning livelihood.

Factors associated with livelihood choice
The multinomial regression estimates for the impact of landlessness and landholding on livelihood choice are given in Tables 5 and 6, respectively. We report and interpret the results in terms of relative risk ratios (RRRs). In our study, the RRR of a coe cient shows how the likelihood of a household choosing a given livelihood (the comparison group) can be compared to the likelihood of that household choosing an agricultural livelihood (the reference livelihood) with changes in explanatory variables. For instance, an RRR > 1 shows that the likelihood of a household choosing a formal wage-earning livelihood relative to the probability of choosing an agricultural livelihood increases as the explanatory variable increases. An RRR < 1 indicates that the probability of the household choosing a formal wage-earning livelihood relative to the likelihood of choosing an agricultural livelihood decreases as the explanatory variable increases (UCLA, 2020) Table 5 shows that the RRRs of landlessness in all livelihood choices are greater than one and statistically highly signi cant. For example, holding all other variables constant, the likelihood of choosing an informal wage-earning livelihood (relative to an agricultural livelihood) is 2.76 times higher for a household without annual cropland than for those with annual cropland. A similar but greater impact is found when other remaining livelihoods are chosen: 7.30 times for a formal wage-earning livelihood, 4.41 times for a livelihood based on nonfarm self-employment and other income, and 3.50 times for livelihoods dependent on formal wage-earning and other sources, respectively. A similar effect is also observed in the case of perennial cropland and aquaculture land.
Our research provides evidence that landlessness acts as a push factor inducing rural households to engage intensively in the rural nonfarm sector. Also, the higher returns from most nonfarm activities suggest that the rural nonfarm sector functions as a pull factor encouraging rural households to diversify their livelihoods towards non-farm activities. Our nding is consistent with that in peri-urban areas of Hanoi, that farmland loss (due to urbanization) increases the likelihood of rural households choosing non-farm livelihoods that are more pro table than farm activities (Tran, et al., 2014). Notes: Robust standard errors in parentheses*** p < 0.01, ** p < 0.05, * p < 0.1. Estimates are adjusted for sampling weights and clustered at the commune level. Hanoi is the reference group. Notes: Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. Estimates are adjusted for sampling weights and clustered at the commune level. Hanoi is the reference group.
The RRRs of land variables in Table 6 are smaller than one and statistically highly signi cant. They indicate that households with more agricultural land are less likely to pursue non-agricultural livelihoods or, in other words, more likely to concentrate their livelihood in agricultural activities. Speci cally, as can be seen in Fig. 1 We also nd that some household characteristics are closely linked with livelihood choice. Male-headed households are less likely to adopt a formal wageearning livelihood. Households with more members are more likely to pursue a non-agricultural livelihood or, in other words, are less likely to specialize in farming activities. This suggests that agricultural production may not require more family labour, possibly as a result of agricultural mechanization in the Red River Delta. As shown in Fig. 2, households whose heads attain better education also have more opportunities to choose a formal wage-earning livelihood or a livelihood from a formal wage-paying job and other sources, which offer high-return nonfarming activities. For instance, one additional year of formal schooling for a household head increases the relative probability of choosing a formal wage-earning livelihood by 29%, and that of choosing a livelihood from a formal wage-paying job and other sources by 12%. Similar ndings are also observed in some peri-urban areas of Hanoi (Tran et al., 2014) and the Mekong River Delta (Hoang et al., 2019), where better education enables rural households to diversify their livelihoods towards high-return activities that often require higher levels of human resources.
As expected, we nd that the likelihood of a household choosing a given livelihood is strongly associated with livelihood opportunities in the commune where the household lives. For instance, Table 6 shows that with a one percentage point increase in formal wage employment opportunities, the relative probability of a household choosing a formal wage-earning livelihood increases by about 4.7%, holding all other variables constant in the model 1 . We also nd that livelihood opportunities vary considerably across provinces. With the same household characteristics, rural households who live in Hai Duong, Bac Ninh, Hung Yen, Thai Binh and Ha Nam are more likely to pursue high return livelihoods than are those in rural Hanoi. 1 Exp (4.621634* 0.01) = 1.047301. Note that 4.621634 is the multinomial logit coe cient that can be obtained by taking the log of the RRR of 101.66, and RRR can be obtained by exponentiating the multinomial logit coe cient.

Conclusion And Policy Implications
The main aim of the current study is to investigate the impact of landlessness and landholding on the choice of livelihoods among rural households in the Red River Delta. The paper uses micro-data from the 2016 VHLSS and econometric analysis. Using cluster analysis, we rst discover what types of livelihoods are pursued by local households. We then compare per capita income across livelihoods in order to understand which brings a higher return. Finally, we measure the role of landlessness and landholding in choosing livelihoods, controlling for other household and commune characteristics.
Among ve livelihoods identi ed via cluster analysis, we nd that the highest income is earned by formal wage-paying livelihoods and the lowest derives from agricultural and informal wage-earning livelihoods. The middle-income group comprises livelihoods from formal wage-paying jobs and other sources, and from nonfarm self-employment and other income. We provide evidence that landlessness or land shortage is not a potential barrier preventing rural households from pursuing gainful livelihoods in the Red River Delta. Notably, households affected by landlessness or land shortage tend to adopt non-farm livelihoods that are more pro table than an agriculture. The nding suggests that landlessness or land shortage should not be viewed as an absolutely negative phenomenon in the region. The reason is that the situation may encourage land-limited households to engage intensively in pro table non-farm activities, which in turn reduce their dependence on farmland and improve their income.
We also nd a number of additional factors determining a pro table livelihood. Better education increases the probability of a household adopting a livelihood related to formal wage-earning employment, which brings much higher income than agricultural or informal wage-paying livelihoods. Also, livelihood choices are found to be in uenced by job possibilities at the commune level. For example, in a commune with greater opportunity for nonfarm self-employment, a household is more likely to pursue a livelihood from nonfarm self-employment and other income. In addition, we observe that livelihood opportunities vary considerably across provinces.

Declarations
Con ict of Interest Statement: The authors agree that this research was conducted in the absence of any self-bene ts, commercial or nancial con icts and declare absence of con icting interests with the funders.   Predicted probabilities of choosing various livelihoods by years of formal schooling