1. Introduction
The hill and mountain areas in China occupy close to 69% of China’s total land area and are home to 45% of its population. The mountain areas are an essential component of regional development in the country [
1,
2], but are lagging behind. The incidence of poverty is higher in the mountains than in the plains as a result of the poor accessibility, fragile ecological environment, and tense relationship between people and the land, as well as low social economic development. Poverty in China is mainly found in rural areas, and in the mountains is particularly prevalent, especially in the southwest mountain areas [
3,
4]. Following the economic system reforms of the 1980s, China has gradually entered a stage of rapid urbanization and industrialization, and now a large number of rural migrant workers leave their villages every year for industries in the cities [
5]. Simultaneously, farmers’ livelihood strategies are diversifying. For example, the share of agriculture in household net income is decreasing, while the share of off-farm income is increasing. According to recent statistics, the total number of rural laborers working off-farm nationwide reached 262.6 million in 2012, 163.4 million (62.2%) of them as migrant workers [
6]. Farmers’ wage income represented 43% of net income, an increase of 23% since 1990 [
7]. For development planning and poverty reduction, it is important to know how these broad statistics relate to the actual situation in specific areas. What is the proportion of migrant farmers in mountain areas, especially in the poor settlements in the southwest? What proportion of household income is provided by agriculture and labor? Is the common assumption that the poorer an area is the greater is the dependence on agriculture true or not? And what are the major factors affecting a household’s choice of livelihood strategy? This study has been designed to answer some of these questions.
A number of authors have shown that poverty can be assessed on the basis of the capital owned by a household [
8,
9]. At the same time, the portfolio of a household livelihood capital will affect the household livelihood strategy [
2,
10,
11,
12,
13]. Researchers have used various approaches to classify livelihood strategies. For example, Fang classified livelihood strategies into two types, farm and off-farm [
2]; Diniz classified into three types, livestock-oriented, diversified-oriented, and off-farm-oriented [
13]; and Alemu into four types—only farm, farm and non-farm, only non-farm, and non-labor [
11]. Others also consider the share of forestry and animal husbandry in household income, for example household livelihood strategies can also be classified into less dependence, moderate dependence, high dependence, and very high dependence on forest/livestock income [
10,
14]. Few studies have considered household livelihood strategies in terms of the share of agricultural income to household net income. China has an ancient history as an agricultural country. In the poorly accessible mountain areas, agricultural income has always been a major part of household income. However, this situation started to change recently when large numbers of people started migrating to work in other areas. In this context, to formulate a sound poverty alleviation policy, it is important to explore the extent to which the income of households in mountainous areas still depends on agriculture (
i.e., the share of agricultural net income to household net income) and to clarify the key factors that influence the selection of household livelihood strategies.
Research regarding the relationship between farmers’ livelihood capital and livelihood strategies is quite abundant [
2,
10,
11,
12,
13], whereas studies of farmers’ livelihood capital and livelihood strategies in mountainous areas are relatively limited. This situation provides a valuable reference for implementing this research. However, compared with the relevant research areas [
12], China’s mountainous areas (especially the mountainous areas in the west) have their own peculiarities (such as land fragmentation and the obvious contradiction between people and land). Therefore, the relevant research conclusions are not necessarily consistent with the research conclusions of China’s mountainous areas. The specific relationship between farmers’ livelihood capital and selections of livelihood strategies in China’s mountainous areas requires further discussion. Due to differences in nature, economy, culture, and other aspects of studies in various countries, there will be some differences when various types of farmers’ livelihood capital are specifically quantified. For instance, Babulo’s [
10] and Bhandari’s [
12] studies did not consider the influence of the social relation network on farmers’ livelihood strategies, whereas the relevant research in China showed that farmers’ social relation network could have a significant influence on the selection of household livelihood strategies [
1,
8]. From this perspective, establishing an indicator system of farmers’ livelihood capital that conforms to the actual situation in China’s mountainous areas, clarifying the relationship between this capital and the selection of livelihood strategies, and comparing with the relevant research results could provide some implications for future research. From two dimensions, namely the research areas (
i.e., China’s typical mountainous areas) and the quantification of the indicator system of livelihood capital, the research has enriched the field and could inspire future research.
3. Results and Discussion
3.1. Household Types and Description of Livelihood Strategies
Table 2 shows the number of sample households in each of the four livelihood strategy classes, the share of total household net income from each of the different sources, and the net value of total household income. The share of miscellaneous income in total household income was extremely small and not further considered. The great majority of households (56%) were less dependent on agriculture (income share < 20%); less than 18% were extremely dependent on agriculture (income share >60%).
Table 2.
Household income share in total income (by income source) and net total income (n = 349).
Table 2.
Household income share in total income (by income source) and net total income (n = 349).
LS Class | Agriculture Income Share | Dependence on Agricultural Income | No. of hh | Av. Income Share a | Net Household Income from All Sources |
---|
Agriculture | Off-Farm | Transfer | Mean b | Range |
---|
LS1 | ≤20% | Less dep. | 197 | 0.094 (0.045) | 0.895 (0.170) | 0.011 (0.168) | 69,713 (34,278) | 12,000–233,800 |
LS2 | (>20%–40%] | Moderately dep. | 65 | 0.291 (0.062) | 0.748 (0.301) | −0.039 (0.289) | 42,747 (24,672) | 6450–104,578 |
LS3 | (>40%–60%] | Highly dep. | 24 | 0.482 (0.062) | 0.328 (0.255) | 0.191 (0.260) | 26,610 (22,456) | 3160–87,300 |
LS4 | >60% | Extremely dep. | 63 | 1.011(0.732) | 0.130 (0.296) | −0.142 (0.679) | 9985 (9062) | 685–35,837 |
| Total | | 349 | 0.337(0.485) | 0.679 (0.386) | −0.016 (−0.363) | 52,435 (38,172) | 685–233,800 |
On average, off-farm net income comprised 68% of total income, more than twice as the average share of agricultural income (34%). This suggests that farmers in the region are on average only moderately dependent on agriculture and are more oriented towards labor. This is consistent with the finding that 95% of households (331) that had at least one family member who worked outside reported that they had worked outside in the survey year. The share of the transfer net income to total income was negative (−1.6%); in other words, farmers’ monetary gift expenditure is more than the total income from policy subsidies and monetary gifts from friends and relatives. Monetary gifts have always been an important part of farm household expenditure in Southwest China. According to the survey, a household with extensive social networks can spend more than 80,000 Yuan ($12,924) in monetary gifts, whereas the average for all households was 3311 Yuan ($535).
In LS1, by far the largest group, approximately 91% of the share of an average household’s total income came from off-farm work, and dependence on agriculture was very low. In LS2, the share of off-farm work in net income was still high at 75%, but agriculture was a much more important source of income, providing more than a quarter of the net total (29%). There were only 24 households in LS3, 7% of the sample. These households were more dependent on agricultural income (48% share of net income) than off-farm income (33%). Farmers in LS4 were almost wholly dependent on agriculture, with off-farm income accounting for only 13% of the net total, and this was absorbed by the very high losses in farmers’ transfer net income, with a negative balance of −14%. Farmers in LS2 also spent more than they received in transfer income (net share: −4%), farmers in LS1 received approximately the same amount that they spent, and farmers in LS3 received a substantial amount that provided 19% of their net income. The average total household net income was strongly associated with the level of dependence on agriculture, with low dependence households having an average net income more than seven times that of the extremely dependent households (69,713 Yuan ($11,262) per year compared to 9985 Yuan ($1613) per year).
3.2. Factors Influencing Farmers’ Livelihood Strategies
Table 3 shows the Mean and Standard Deviation of the livelihood capital indicators under various types of livelihood strategies. The average household size was 4.1, including 2.7 laborers. Household size was related to dependence on agriculture and ranged from 2.9 for the most dependent households (LS4) to 4.6 for the least dependent (LS1). Thirty percent of the households have 6–13-year-old children, which ranged from 0.4 for the least dependent (LS1) to 0.2 for LS3. The mean age of the household heads was 48.8 years, with little variation among the groups, except for the least dependent group, which had somewhat younger heads of household. The average years of education of the household heads (7.0) was markedly lower than that of other household members (9.0); there was little variation among the groups except for the group most dependent on agriculture, in which both the household heads and other family members had lower levels of education (5.9 and 6.8 years, respectively).
Physical capital was assessed in terms of the market value of the house and fixed assets and distance from the nearest county center. As shown in
Table 3, farmers with a greater dependence on agriculture tended to have less valuable fixed assets, whereas those with less dependence on agriculture tended to have more valuable houses. Households with the highest dependence on agriculture also tended to be located furthest from the county center (49.5 kilometers cf. 41.2 kilometers).
The natural capital comprised the area of cultivated land. The average land per capita cultivated per household was 0.6 mu (0.4 ha), or 2.5 mu (1.67 ha) for an average household, which is relatively low. The households most dependent on agriculture had the most land per capita (1.25 mu (0.83 ha)), and the households least dependent on agriculture had the least (0.38 mu (0.25 ha)).
Financial capital was assessed in terms of loans. As shown in
Table 3, the households that were less dependent on agriculture had greater lending ratios (LS1 > LS4), whereas households that were more dependent on agriculture had greater borrowing ratios (LS4 > LS1), although the greatest amount was for LS2.
Social capital was determined in terms of the size of the social network of family and friends available to assist financially or in finding a job, as well as participation in a farm association. There was considerable variation between the groups, but no obvious relationship with the dependence on agriculture (LS1 had the most developed social network to assist with work, and LS2 had the most developed social network to help financially). Few household heads participated in a farm association, only 4% overall, with 11% of the LS3 households and 2% of the LS1 households.
Table 3.
Summary statistics for the explanatory variables (by LS class).
Table 3.
Summary statistics for the explanatory variables (by LS class).
Variable | Total Sample | LS1 (Less Dependent) (n = 197) | LS2 (Moderately Dependent) (n = 65) | LS3 (Highly Dependent) (n = 24) | LS4 (Extremely Dependent) (n = 63) |
---|
Mean | S.D. | Mean | S.D. | Mean | S.D. | Mean | S.D. | Mean | S.D. |
---|
pop | 4.10 | 1.62 | 4.55 | 1.46 | 4.28 | 1.62 | 3.17 | 1.55 | 2.87 | 1.40 |
lab | 2.74 | 1.26 | 3.01 | 1.15 | 3.17 | 1.22 | 2.04 | 1.23 | 1.75 | 1.02 |
children a | 0.34 | 0.48 | 0.40 | 0.49 | 0.34 | 0.48 | 0.17 | 0.38 | 0.24 | 0.43 |
hhage | 48.75 | 10.51 | 45.53 | 9.01 | 50.86 | 9.19 | 55.75 | 13.34 | 53.95 | 11.26 |
hhedu | 6.98 | 3.29 | 7.28 | 2.83 | 7.37 | 3.55 | 6.29 | 4.25 | 5.92 | 3.74 |
maxhedu | 9.01 | 3.74 | 9.63 | 3.23 | 9.57 | 3.58 | 8.17 | 4.57 | 6.79 | 4.23 |
ln(fiasvalue) | 8.68 | 1.05 | 8.85 | 0.90 | 8.72 | 0.82 | 8.40 | 1.20 | 8.17 | 1.43 |
ln(hovalue) | 1.81 | 1.19 | 2.04 | 1.04 | 1.99 | 1.00 | 1.21 | 1.33 | 1.11 | 1.42 |
disc | 43.83 | 18.27 | 41.24 | 17.32 | 48.20 | 19.30 | 38.45 | 18.63 | 49.46 | 18.06 |
area | 0.60 | 0.84 | 0.38 | 0.46 | 0.75 | 0.67 | 0.58 | 0.56 | 1.17 | 1.49 |
loan a | 0.35 | 0.48 | 0.35 | 0.48 | 0.43 | 0.50 | 0.29 | 0.46 | 0.30 | 0.46 |
borrow a | 0.46 | 0.50 | 0.45 | 0.50 | 0.49 | 0.50 | 0.46 | 0.51 | 0.48 | 0.50 |
sonwohelp a | 1.87 | 0.83 | 1.95 | 0.84 | 1.92 | 0.82 | 1.54 | 0.72 | 1.71 | 0.83 |
sonmohelp a | 2.26 | 0.75 | 2.31 | 0.75 | 2.34 | 0.69 | 2.17 | 0.70 | 2.06 | 0.78 |
whpafaass a | 0.04 | 0.19 | 0.02 | 0.14 | 0.06 | 0.24 | 0.08 | 0.28 | 0.05 | 0.21 |
3.3. Determinants of Livelihood Strategies
The results of the regression analysis are presented in
Table 4. Collinearity diagnostics and tolerance statistics were used to diagnose potential multicollinearity problems [
30]. No evidence of multicollinearity was found in the data. Meanwhile, to avoid the heteroscedasticity of the independent variable influencing the results, when building the z statistic, we used robust standard errors. The Wald chi
2 is 119.94 (Prob > chi
2 = 0.0000), which indicates that at least one of the independent variables has a significant influence on the dependent variable. Additionally, the Pseudo
R2 of the model is 0.2032.
Table 4.
Ordinal logistic regression estimates and test statistics for the agriculture dependence model a.
Table 4.
Ordinal logistic regression estimates and test statistics for the agriculture dependence model a.
Explanatory Variables | Regression Coefficient | Robust Std. Err. | z | P > |z| | OR b |
---|
pop | −0.072 | 0.141 | −0.51 | 0.611 | 0.931 |
lab | −0.544 *** | 0.176 | −3.09 | 0.002 | 0.580 |
children | 0.225 | 0.339 | 0.66 | 0.507 | 1.252 |
hhage | 0.328 *** | 0.125 | 2.63 | 0.009 | 1.388 |
hhage2 | −0.002 ** | 0.001 | −2.04 | 0.042 | 0.998 |
hhedu | 0.093 * | 0.053 | 1.76 | 0.078 | 1.097 |
maxhedu | −0.155 *** | 0.050 | −3.10 | 0.002 | 0.856 |
ln(fiasvalue) | 0.034 | 0.197 | 0.17 | 0.864 | 1.034 |
ln(hovalue) | −0.170 | 0.141 | −1.21 | 0.228 | 0.843 |
disc | 0.023 *** | 0.008 | 2.98 | 0.003 | 1.023 |
area | 0.653 *** | 0.196 | 3.33 | 0.001 | 1.922 |
loan c | −0.030 | 0.269 | −0.11 | 0.911 | 0.970 |
borrow c | 0.681 *** | 0.263 | 2.59 | 0.010 | 1.975 |
dum_sonwohelp_2 | −0.543 | 0.422 | −1.29 | 0.197 | 0.581 |
dum_sonwohelp_3 | −1.110 ** | 0.557 | −1.99 | 0.046 | 0.329 |
dum_sonmohelp_2 | 0.636 | 0.438 | 1.45 | 0.147 | 1.889 |
dum_sonmohelp_3 | 1.033 * | 0.620 | 1.67 | 0.096 | 2.812 |
whpafaass c | 2.315 *** | 0.625 | 3.71 | 0.000 | 10.121 |
3.3.1. Impact of Human Capital on Household Livelihood Strategy Selection
Inconsistent with hypothesis H1, the size of a household had no significant influence on household livelihood strategy, whereas the number of laborers did. The results suggest that dependence on agriculture is negatively associated with the number of laborers; the partial correlation coefficient is −0.544. Specifically, a one-person increase in the family labor force corresponds to a reduction of household income dependence on agriculture by 0.420 levels. One possible reason for this relationship is that the scattered land and lack of significant economic crops mean that agricultural income is generally low and does not feature as strongly as expected as a proportion of net income, even when agriculture is a major strategy. Furthermore, even when older people and children can help in farm work, the increase in agricultural income is relatively small, and when additional household members are aging, sick, or very young, they make demands on the labor force and actually reduce the output. Thus, the number of members available for labor is much more significant than household size alone.
Interestingly, the age of the household head had a significant influence on household livelihood strategy, and the relationship between the two is represented by an inverted U curve. This result is inconsistent with Babulo’s: his study shows that the age of the household head had no significant influence on household livelihood strategy [
10]. In our study, when the age of the household head is below 82 years (0.328/(2 × 0.002)), the older the household head, the greater the dependence on agriculture (the partial correlation coefficient of the household head is 0.328). Specifically, with a one-year increase in the age of the household head, household income dependence on agriculture increases by 0.388 levels. The low educational level meant that the only outside options for most of the household heads in the survey were industries that demand high-intensity labor, such as construction. When age and poor health limited their ability to continue in such jobs, one option was to return to farming, with household income again more dependent on agriculture. When the household head is under 82 years of age, the older the household head, the lower the dependence on agriculture (the partial correlation coefficient of the household head square is −0.002). This is consistent with the actual situation in the study area. When the household head is too old to farm, there is naturally a reduced dependence on agriculture.
Interestingly, inconsistent with hypothesis H3, whether the household has children had no significant influence on the household livelihood strategy, although the partial regression coefficient is positive. This finding differs from the findings in other developing countries [
12,
22,
23]. One possible reason for this difference is a different division standard of household livelihood strategy. For example, Bhandari’s study [
12] divided the household livelihood strategy into two categories, left farming and continued farming, and drew a conclusion that working-age children are statistically important human capital measures in the decision to exit farming. This study separates household livelihood strategy into four categories according to the share of agriculture net income to total net income. The span between the four categories is large. For example, the agricultural income share of LS1 is less than 20%, whereas that of LS2 is more than 20% and less than 40%. So, even if children had a significant influence on whether the adult labor force engaged in agriculture, the degree should not significantly affect the proportion of household income. Whether children had a significant influence on labor transfer in this study area can be discussed in future studies.
Interestingly, consistent with the results of Babulo’s study [
10], the number of years of education of the household head had a positive significant influence on household livelihood strategy, which is inconsistent with hypothesis H4 (more years of education, more likely to work outside). In contrast, the maximum years of education of any household member had a negative influence on household livelihood strategy, consistent with hypothesis H5. Specifically, with a one-year increase in education of the household head, the household income’s dependence on agriculture increases by 0.097 levels. With a one-year increase in the maximum years of education of any household member, the household income’s dependence on agriculture decreases 0.144 levels. One possible reason for this relationship is that the low level of rural economic development and relative scarcity of advanced agricultural equipment meant that agricultural production in the region was labor-intensive, which generally limited the years of education of all household heads. Generally, when new agricultural production tools and technologies emerge, households with higher education will adopt them earlier [
8,
20], and this will make household income more dependent on agriculture. Meanwhile, the household members who had received the most education were mostly young or middle-aged (79 of the 97 people with a high school background were under the age of 45) and probably more inclined to migrate, partially to be able to support children and elderly parents financially. Thus, having a family member with a higher educational level was reflected in income from labor being more important than income from agriculture.
3.3.2. Impact of Physical Capital on Household Livelihood Strategy Selection
Surprisingly, inconsistent with hypotheses H6 and H7, the value of houses and fixed assets had no significant influence on the household livelihood strategy. The reason the former was not significant is that houses have become a symbol of wealth for farmers: valuable and attractive houses are both a cause for pride and are extremely beneficial for the children’s marriage prospects, so farmers usually spend years of savings or go into debt to finance building a house [
20]. So, except for a few rich households, there is no significant difference among the house value of most households. Fixed assets had no significant influence on household livelihood strategy. This is consistent with the result of Babulo’s study [
10]. This finding may be in part because the value of agricultural production tools (such as pumps, dozen rice machine,
etc.) is generally low and there is no obvious difference between farmers. At the same time, there are very few families that can afford expensive fixed assets (such as cars, computers,
etc.) due to low education levels and low income. Furthermore, most of the families can afford the commonly used fixed assets (such as mobile phones, televisions, refrigerators,
etc.); however, the purchasing power of the wage-based income households is relatively stronger. Thus, the value of these commonly used fixed assets in these households is generally higher than in households that are highly dependent on agriculture, but there is still no obvious difference between farmers. This relationship can be confirmed from the LS1 to LS4 households’ descending value of fixed assets. Additionally, another reason that the value of houses and fixed assets had no significant influence on the household livelihood strategy is that the livelihood vulnerability of households in southwest mountainous areas is strong, whereas the strike capabilities of households are weak. So, households often save their cash income to resist external shocks rather than invest in houses or fixed assets. This has been proved by Chen’s [
8] and Tai’s [
20] studies.
The distance from a household to the nearest county center had a positive influence on the household livelihood strategy, consistent with hypothesis H7 (the greater the distance, the greater the dependence on agriculture). Specifically, with a 1-km increase in the distance to the nearest town center, the household income’s dependence on agriculture increases by 0.023 levels. Greater distance means reduced access to labor and other markets, an increased need for self-sufficiency, and greater dependence on agriculture.
3.3.3. Impact of Natural Capital on Household Livelihood Strategy Selection
Consistent with the results of Babulo’s [
10] and Jansen’s [
31] studies, the per capita cultivated land area had a positive influence on livelihood strategy, consistent with hypothesis H8 (more land is related to increased dependence on agriculture). Specifically, with a 1-mu increase in per capita cultivated land area, the household income’s dependence on agriculture increases by 0.922 levels. One possible reason for this relationship is that having more land means that more labor is required and that family members are less able to migrate; thus, the dependence on agriculture is greater.
3.3.4. Impact of Financial Capital on Livelihood Strategy Selection
Interestingly, on the one hand, consistent with the findings of Babulo’s study [
10], whether a household had loans had no significant influence on the household livelihood strategy (inconsistent with hypothesis H10). On the other hand, consistent with hypothesis H10, whether a household had borrowed had a positive influence on the livelihood strategy. Specifically, the households with loans are 1.975 times more dependent on agriculture than the households without loans. One reasonable explanation for this phenomenon is that farmers who are dependent on agriculture are more likely to be exposed to risks from factors such as extreme weather, pests, and natural disasters, especially given the lack of support facilities, such as irrigation. They are also less likely to have savings available to cope with demands for events such as weddings, funerals, children’s schooling, and abnormal medical charges. Thus, households dependent on agriculture are more likely to have to borrow. The link did not exist for formal loans. This may be because taking a loan from a bank is the least preferred option, partially because interest payments are high. Equally, those dependent on agriculture may find it more difficult to obtain a bank loan as they have few assets to offer as security and may have to leave assets that they need for farm work as a pledge or must find a creditworthy guarantor, which is also problematic in a poor area.
3.3.5. Impact of Social Capital on Livelihood Strategy Selection
Consistent with hypothesis H11, the social network of relatives and friends available for assistance when seeking non-farm work had a negative influence on household livelihood strategy. Specifically, there was no difference between the households with an undeveloped social network and those with no social network, whereas the households with developed social network were 0.329 times less dependent on agriculture than those with no social network. In China’s rural areas, social networks based on blood relationships and geopolitical and friendship links have a significant influence on farm household income. These networks not only provide information for workers but can also provide financial assistance, life care, and emotional support, especially for moderate external risks [
32,
33]. A greater network means more likelihood of accessing work and a higher proportion of income from migrant labor, with a concomitant reduction in dependence on agriculture.
Similarly, consistent with hypothesis H12, the social network of relatives and friends available for assistance when in urgent need of a lot of money had a positive influence on livelihood strategy. Specifically, there was no difference between the households with an undeveloped social network and those with no social network, whereas the households with a developed social network were 2.812 times more dependent on agriculture than those with no social network. Migrant laborers have fewer interactions in the village and weaker trust relationships than agriculture-dependent household workers who live in the village year round and are likely to have more relatives and friends available for assistance. Thus, the more developed the social networks of the households, the more the household income is dependent on agriculture.
Additionally, consistent with hypothesis H13 (participation of a household head in a farm association linked to dependence on agriculture), households whose household head participates in a farm association are 10.121 times more dependent on agriculture than the households whose household head does not participate. This result is consistent with Fang’s study [
2]; his study also shows that households that have participated in a farm association tend to choose farming as their main livelihood strategy.
4. Conclusions
This paper attempts to answer the following two key questions: “To what extent are the farmers in the Three Gorges Reservoir area still highly dependent on agriculture against a background of rapid urbanization and industrialization?” and “Which factors determine a household’s choice of livelihood strategy, with a particular focus on the production of and dependence on agricultural products?” Based on the DFID SLA framework, data were collected from a sample of 349 rural households in the Three Gorges Reservoir area. The analysis of the data and the ordinal logistic regression results showed the following:
- (1)
Households tended toward migrant work as their main livelihood strategy, with the majority of income from off-farm sources. Close to 56% of households showed low dependence on agriculture, and 19%, 7%, and 18% showed moderate dependence, high dependence, and extreme dependence on agriculture, respectively.
- (2)
The livelihood capital portfolio of farm households significantly affected their livelihood strategy.
The number of laborers in a household, the age of the household head, the education years of the household head, the maximum years of education of any household member, household location, per capita area of cultivated land, whether the household loaned or borrowed in the last five years, and formal and informal social networks had a significant effect on the choice of livelihood strategy.
Based on our findings, several recommendations can be made for research and policy.
The number of laborers in a household encouraged a strategy of migrant work and reduced the dependence on agriculture, whereas the number of household members had no significant influence. The maximum years of education of any household member and the years of education of the household head had an opposite and significant influence on household livelihood strategy. Many studies consider only the household size and the educational level of the household head; future studies should include the more differentiated indicators.
The distance of a household from the nearest county center increased the dependence on agriculture and reduced the options for a strategy of migrant work. Government departments should consider increasing investment in infrastructure, such as roads for remote and less developed villages, which could improve accessibility, increase the access to employment information, reduce transportation costs for village enterprises, and increase economic benefit.
Whether the household had borrowed in the last five years was linked to a strategy of increased dependence on agriculture, whereas whether the household had loaned was not. Government loan policies could be made more favorable for farmers, which could encourage them to lead other villagers toward achieving increased prosperity.
Farmers’ formal and informal social networks had a significant influence on household livelihood strategy. During the course of the study, it became clear that many migrant workers from the villages were working in the same city (for example, the large number of people from Baoping Village engaged in construction demolition in Wuhan). The government should consider not only strengthening support for farm associations but also encouraging rural able-person to use their social networks to actively establish businesses at migrant destinations (such as establishing a large demolition company) where they can work together. This would help farmers achieve a common prosperity and enhance their collective ability to resist external risks, as well as strengthening village resilience.