A Study of Multidimensional and Persistent Poverty among Migrant Workers: Evidence from China’s CFPS 2014–2020

Abstract

Poverty is a critical social problem in numerous countries. It is a result of many aspects and has been addressed worldwide for a long time. In this study, we construct the multidimensional poverty index (MPI) for migrant workers in China based on Amartya Sen’s capabilities approach. Using the Chinese Family Panel Studies (CFPS) data during 2014–2020 and the Alkire–Foster methodology, our study examines the multidimensional poverty of migrant workers using static to dynamic analyses. The results show the following: (1) The multidimensional poverty of migrant workers shows a general decreasing trend over time. (2) Over one third of migrant workers are in severe multidimensional poverty. (3) The in-work poverty of migrant workers is serious, which is reflected in the indicators of employment security, housing provident fund, labor contract, and labor union. (4) Approximately 30% of migrant workers’ multidimensional poverty is persistent. It is difficult for them to rid themselves of persistent poverty by their own effort. This study suggests that the government should pay more attention to poverty reduction and capability improvement for migrant workers.

1. Introduction

As a significant economic index, poverty is a persistent social problem, and has been a source of concern across the world. Despite the political systems and living standards in different countries, poverty remains a serious social problem [1]. Multidimensional poverty is a composite indicator. It reflects inequality, sustainability, and economic development [2]. The theoretical basis of multidimensional poverty can be traced to Amartya Sen’s capabilities approach. The concept and measurement of multidimensional poverty have been regarded as an effective method. This method can complement or replace the traditional income poverty line [3]. Poverty is multidimensional. Additionally, it should combine material hardship with broken institutions and human frailty [4]. In the state’s welfare assessment system, the incorporation of non-monetary dimensions would have positive significance for national poverty reduction efforts and policy interventions [5,6]. In Sen’s capabilities approach, the income–capability approach was proposed. This indicated that poverty can be judged in a multidimensional way by combining an income indicator at the monetary level and capability indicators at the non-monetary level [7].

In many countries, migrants are often regarded as poor [8,9,10]. China has a large group of migrant workers. This large-scale rural-to-urban migration constitutes a hallmark of modern society. Migrant workers have become an important labor force in the Chinese urbanization and modernization process. In 2021, the number of migrant workers reached 170 million [11]. The monthly income levels of migrant workers are gradually increasing, but the poverty rate is also increasing. Income can alleviate the temporary hardship of migrant workers, but it cannot help them escape from poverty. Due to non-citizenship, migrant workers suffer from the deprivation of employment, welfare, social security, public services, and so on. Additionally, it leads them into multidimensional poverty. Illness, COVID-19, and unemployment increase the poverty risk and poverty vulnerability of migrant workers [12,13], which will affect social stability and the progress towards poverty alleviation in China.

The poverty of Chinese migrant workers is often hidden [14]. According to the Chinese poverty line standard in rural areas, they have long been excluded from the rural poverty alleviation program. They are also excluded from the urban low-income social assistance system based on the income level of migrant workers [15]. Meanwhile, migrant workers still have a large gap compared with the average income level of urban residents [16]. This hidden poverty puts migrant workers in a vacuum of state regulation between the rural and the urban [17]. For several decades, Chinese society has inappropriately depicted the negative effects of migrant workers. For example, migrant workers are considered as a trigger of the rise in crime, the decline in the level of public services, the rise in unemployment of local residents, and the reduction of agriculture [18].

A unique feature of migrant work in China is the hukou system [19]. In China, migrant workers’ rights, welfare, and public services are seriously deprived by the hukou system [20]. Most of the migrant workers are marginalized groups in society [21], representing the disadvantaged and disenfranchised [22]. However, there are few studies that have focused on the issue of migrant workers’ survival. Additionally, few studies have measured their multidimensional poverty. Poverty is not only a static state but also a dynamic process. The poverty situation of migrant workers changes over time. Therefore, a dynamic analysis is important for the evaluation of multidimensional poverty, which can help us examine the changing trends in the poverty of migrant workers [23]. However, few studies have discussed the dynamic evolution of multidimensional poverty among Chinese migrant workers. Our study analyzes the poverty persistence of migrants through the duration of poverty, which is rarely seen in existing studies. Based on Sen’s capabilities approach and the A–F method, we measure the multidimensional poverty index (MPI) of migrant workers, in order to provide a comprehensive framework for assessing the multidimensional poverty of Chinese migrant workers using static to dynamic analyses.

2. Data and Methods

2.1. Data Source

In this study, we use data from the Chinese Family Panel Studies (CFPS). The CFPS is a nationwide comprehensive social survey conducted by the China Social Science Research Center of Peking University. The CFPS is followed by a formal nationwide interdisciplinary survey. It has launched a survey every two years since 2010. The CFPS is a representative national survey, covering family, work, economy, and other aspects. It is carried out as sampling surveys on residents of 25 provinces (municipalities directly under the central government and autonomous regions) in China, except Hong Kong, Macau, Taiwan, Xinjiang, Tibet, Qinghai, Inner Mongolia, Ningxia, and Hainan, covering about 95% of China’s population (except Hong Kong, Macau, and Taiwan). The data in the CFPS are sampled with implicit stratification, over multiple stages, multiple levels, and with probability proportionate to size sampling (PPS). Thus, it is suitable for multidimensional poverty measurement. Our study selected a sample of migrant workers over four years: 2014, 2016, 2018, and 2020. We processed the data samples as follows: (1) First, we considered migrant workers with agricultural household registration who engaged in non-agricultural work or business in cities. Additionally, their ages ranged from 16 to 60 years old. The unemployed samples were excluded accordingly. (2) Secondly, the samples with missing values were deleted. Our final valid sample consisted of 18,724 migrant workers (

Table 1. Sample composition of migrant workers.

Year	2014	2016	2018	2020	Total
Valid sample	4853	5214	4482	4175	18,724

).

2.2. Approach

We use Sen’s capabilities approach as the basis of multidimensional poverty [24]. It provides a broader view of poverty, considering poverty as the deprivation of wellbeing. Functions and capabilities are the core contents of Sen’s approach. Functions means the concepts required for a person to succeed, such as health and a social life. Capabilities refers to the set of all functions a person can choose from [25]. This approach evaluates the capability space for a person without considering whether the person would exercise these capabilities or not [26]. Sen states that a person is regarded as deprived when the person has no capability [24,27]. This deprivation means the lack of freedom and opportunities. Additionally, poverty is a result of many aspects, not only income.

Sen’s capabilities approach provides the theoretical basis for studies on multidimensional poverty, which investigate poverty throughout the life spectrum. Therefore, this approach can be used to study poverty not only in developing but also in developed countries [7,28]. For its scientific and precise results regarding multidimensional poverty, scholars in China and abroad have been paying more attention to this approach [29,30].

2.3. Measurement

Many studies use the MPI to measure multidimensional poverty at the household level. However, some studies have shown that household-based measures often lead to an underestimation of the overall poverty level in society, because they cannot capture the inequal arrangement and resource allocation within households [31]. The 2030 Agenda for Sustainable Development advocates that poverty should be examined at the individual level, that is, “leaving no one behind” [32]. Therefore, our study chose individual indicators to measure multidimensional poverty among migrant workers.

According to Sen’s theory, poverty can be seen as the deprivation of capabilities [27]. Additionally, Sen proposed the income–capability supplementary approach. This is a nonradical and practical approach, which uses income as a traditional measure, and uses capability considerations as supplement. The explanatory power and applicability are necessary for practical evaluation and policy analysis. The approach examined five categories of instrumental freedoms—economic level, political freedoms, social opportunities, protective security, and transparency guarantees—which can expand human capability. These five instrumental freedoms present a general framework to determine inequalities [33]. This framework also provides a broader and more detailed perspective on poverty analysis, which allows development and poverty reduction policies to be evaluated more thoroughly [34].

In addition, many economists evaluate welfare or poverty on subjective grounds (subjective well-being or life/economic satisfaction) [35], which is also a feasible complementary method to conventional measurement. In the process of poverty alleviation, Sen also noted that it is important to judge well-being and poverty using a mental-metric way [36], to use the results of the subjective evaluation as a basis of judgment in public policy, and to use the subjective evaluation as a tool for identifying the target of poverty. Therefore, establishing the relationship between welfare, poverty, and the individual’s subjective evaluation has been shown to be crucial [37]. In other words, the method of subjective poverty determination is a synthesis of both absolute and relative poverty concepts, which makes the explanation of poverty more comprehensive [38].

In summary, based on Sen’s capabilities approach, the CFPS data, and the situation of Chinese migrant workers, we measure multidimensional poverty using the following six dimensions: (1) economic level, (2) social opportunities, (3) transparency guarantees, (4) protective security, (5) political rights, and (6) spiritual poverty. Specifically, we include 11 indicators; the explanations of each dimension and indicator are presented in

Table 2. Dimensions, indicators, and deprivation thresholds of multidimensional poverty among migrant workers.

Dimension	Indicator	Poverty Line (Deprived if…)
Economic level	Income (Inc)	Personal annual income is less than 50% of the per capita disposable income of urban residents
Social opportunities	Education (Edu)	Failure to complete Nine-Year Compulsory Education in China
	Health (Hea)	Self-rated health as “unhealthy”
	Chronic disease (CD)	Have doctor-diagnosed chronic disease
Transparency guarantees	Labor contract (LC)	No employment contract
Protective security	Medical insurance (MI)	Not participating in any medical insurance
	Employment security (ES)	No employment security whatsoever
	Provident fund (PF)	No housing provident fund
Political rights	Labor union (LU)	Non-unionized
Spiritual poverty	Life satisfaction (LS)	Self-life satisfaction is very low
	Subjective poverty (SP)	Self-income rating is very low

.

The choice of indicator weight is key. Our study uses principal component analysis (PCA) to determine the weights of each indicator. First, we conducted factor analysis on 11 indicators. The results of the Kaiser–Meyer–Olkin (KMO) values and cumulative variance contribution rates are shown in

Table 3. Weighting of multidimensional poverty indicators for migrant workers.

Dimension	Indicator	wj*
		2014	2016	2018	2020
Economic level	Income (Inc)	0.042	0.073	0.095	0.080
Social opportunities	Education (Edu)	0.104	0.087	0.095	0.088
	Health (Hea)	0.107	0.095	0.102	0.119
	Chronic disease (CD)	0.096	0.079	0.056	0.085
Transparency guarantees	Labor contract (LC)	0.103	0.108	0.117	0.117
Protective security	Medical insurance (MI)	0.098	0.112	0.121	0.112
	Employment security (ES)	0.100	0.117	0.091	0.074
	Provident fund (PF)	0.101	0.074	0.083	0.093
Political rights	Labor union (LU)	0.050	0.028	0.044	0.021
Spiritual poverty	Life satisfaction (LS)	0.100	0.114	0.098	0.105
	Subjective poverty (SP)	0.100	0.114	0.098	0.105
KMO value		0.781	0.798	0.774	0.783
Cumulative variance explained		59.28	63.29	70.25	71.18

. This shows that (1) the KMO results reflecting our sample data were suitable for factor analysis; and (2) the cumulative variance contribution rates had strong explanatory power for information variation. Second, the coefficients of the 11 indicators in each linear combination of principal components were calculated using the loadings and the characteristic root values. We take the variance contribution rate as the weight. The weighted average of the coefficients was derived from the variance contribution rate. Finally, the weight coefficients w_j (j = year) of the 11 indicators were calculated. w_j* was obtained after normalization, which made the weights of each dimension between 0 and 1. The weights of each indicator changed over time (Figure 1). Compared with 2014, the weights of Inc, Hea, LC MI, LS, and SP increased in 2020. The weights of Edu, CD, ES, PF, and LU decreased in 2020. The weight represents the explanatory power of an indicator. The greater the weight, the higher the importance of the index [39].

2.4. The A–F Methodology

The A–F method is the most mature and widely used method, especially in policies and field surveys [40]. The A–F method was proposed by Alkire and Foster and adopted by the United Nations Development Program. Alkire and Foster proposed the A–F method based on Sen’s capability approach [41]. As poverty is explained according to the vectors of various dimensions of the standard of living in Sen’s capability approach, the A–F method measures poverty axiomatically. The A–F method has the following steps: dimension setting, poverty identification, and poverty summation.

2.4.1. Dimensional Setting

M^n,d represents the n × d dimensional matrix. Let the elements of the matrix y ∈ M^n,d represent the values obtained by n individuals in d different dimensions. In the formula, for any element y_ij in y, this represents the value taken by individual i in dimension j, i = 1, 2, …, n; j = 1, 2, …, d.

The row vector y_i = (y_i1, y_i2, …, y_id) includes the values of individual i in all dimensions. Similarly, the column vector y_j = (y_1j, y_2j, …, y_nj) represents the distribution of the values obtained by different individuals in j dimensions.

2.4.2. Poverty Identification

(1)

Unidimensional poverty identification

Let z_j (z_j > 0) represent the threshold or poverty line at which the jth dimension is deprived.

For any matrix y, we can define a deprivation matrix: g⁰ = [g⁰_ij]. g⁰_ij is defined as follows: when y_ij < z_j, g⁰_ij = 1; when y_ij ≥ z_j, g⁰_ij = 0. For example, for the ijth element, when an individual i is deprived in the jth dimension, the value is assigned a value of 1; when the individual is not deprived in that dimension, the value is 0.

For this deprivation matrix g⁰, a column vector can be defined to represent the total number of dimensions of poverty suffered by individual i; that is, the value of the ith element is c_i = |g⁰_i|.

(2)

Multidimensional poverty identification

Each element of the above deprivation matrix, g⁰ = [g⁰_ij], represents the presence or absence of deprivation for each individual in each dimension, in a single-dimensional approach. When introducing a multidimensional identification method, it is important to consider whether an individual is deprived when k dimensions are considered simultaneously.

Let k = 1, …, d; p_k be a function used to identify the poor when k dimensions are considered; p_k (y_i; z) = 1 when c_i < k, and p_k (y_i; z) =0 when c_i < k. In other words, p_k defines individual i as being in poverty when the total number of dimensions in which individual i is deprived, c_i, is greater than or equal to k; otherwise, p_k defines individual i as being in poverty when the total number of dimensions in which individual i is deprived, c_i, is less than k and p_k defines individual i as not being in poverty in dimension k. In other words, p_k is influenced by both z_j (deprivation within dimensions) and cross-dimensional c_i deprivation. This is known as the dual cut-off approach. Usually, the multidimensional identification of k lies between 1 and d.

2.4.3. Poverty Summation

After each dimension of deprivation was identified, they were summed to obtain a multidimensional composite index. The simplest summation method is the incidence of poverty (H) on a per capita basis: H = H(y;z); H = q/n, where q is the number of poor individuals under z_k (i.e., the number of individuals with k simultaneous dimensions of poverty). This method is known as the Foster–Greer–Thorbecke (FGT) method. The advantage of the FGT method is its simplicity and clarity, and its disadvantage is that it is insensitive to the distribution of poverty and the depth of deprivation.

To overcome these disadvantages of the FGT method, Alkire and Foster proposed a new measure of multidimensional poverty using a modified FGT method. The formula used is as follows:

M₀(y;z) = μ[g⁰(k)] = HA

(1)

where M₀ is the adjusted multidimensional poverty index. It consists of two parts: H (multidimensional poverty incidence) and A (average deprivation score), also known as the poverty intensity index, which is equal to the ratio of the average number of deprived dimensions to the total number of dimensions of all poverty individuals. The formula used is as follows:

$$H_{\left(k\right)}=\frac{\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{q}}_{\mathrm{i}\mathrm{j}}\left(\mathrm{k}\right)}{\mathrm{n}}$$

(2)

$$A_{\left(k\right)}=\frac{\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{c}}_{\mathrm{i}}\left(\mathrm{k}\right)}{\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{q}}_{\mathrm{i}\mathrm{j}}\left(\mathrm{k}\right)·\mathrm{m}}$$

(3)

Combining Formulas (1)–(3), the multidimensional poverty index, denoted as M(k), can finally be derived as follows:

$${\mathrm{M}}_{\left(\mathrm{k}\right)}=\frac{\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{c}}_{\mathrm{i}\mathrm{j}}\left(\mathrm{k}\right)}{\mathrm{n}\mathrm{m}}=\frac{\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{q}}_{\mathrm{i}\mathrm{j}}\left(\mathrm{k}\right)}{\mathrm{n}}\times \frac{\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{c}}_{\mathrm{i}}\left(\mathrm{k}\right)}{\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{q}}_{\mathrm{i}\mathrm{j}}\left(\mathrm{k}\right)·\mathrm{m}}={\mathrm{H}}_{\left(\mathrm{k}\right)}\times {\mathrm{A}}_{\left(\mathrm{k}\right)}$$

(4)

2.5. The Measurement of Multidimensional Poverty Persistence

For most people who have experienced poverty, the degree and condition of their poverty changes over time, repeatedly and continuously [42]. National poverty alleviation efforts need to track the dynamic changes in migrant workers’ poverty situation over time [14]. In our study, we used the panel data to examine the persistence of migrant workers’ poverty. These data can not only be used to estimate the poverty level, but also to dynamically identify the poverty persistence of migrant workers.

In this study, we evaluated the multidimensional poverty persistence of migrant workers via poverty duration, which refers to the number of consecutive periods in which migrant workers fall into multidimensional poverty during the observation period [43]. The data used in our study are CFPS samples from 2014, 2016, 2018, and 2020. Therefore, there are several classes of poverty duration: (1) two periods of poverty persistence are found for 2014 and 2016, 2016 and 2018, and 2018 and 2020, which represents being in poverty for two consecutive survey years; (2) three periods of poverty persistence are found for 2014, 2016, and 2018, or 2016, 2018, and 2020, which represents being in poverty for three consecutive survey years; and (3) four periods of poverty persistence are found for 2014, 2016, 2018, and 2020, which represents being in poverty for four consecutive survey years.

3. Results

3.1. Socio-Demographic Characteristics

Table 4. Characteristics of migrant workers.

Year	2014	2016	2018	2020
Variables/samples	4853	5214	4482	4175
Age	33.80 ± 11.44	33.69 ± 10.44	34.76 ± 10.60	35.75 ± 10.47
Gender
Male	62.60	61.78	59.95	61.51
Female	37.40	38.22	40.05	38.49
Sex ratio	167.38	161.64	149.69	159.81
Education
Primary school or below	35.26	32.92	25.19	22.81
Junior high school	38.94	39.05	38.29	37.17
Senior high school	16.06	16.32	19.25	19.26
University or above	9.74	11.71	17.27	20.76

shows the demographic characteristics of the migrant worker samples in this study. The average age of the samples ranged from 33.80 to 35.75. The sex ratio ranged from 167.38 to 159.81. The ratio of male to female migrant workers was approximately 6:4. The proportion of migrant workers with junior high school education was the largest. The proportion of migrant workers with elementary school education or below decreased with time. On the contrary, the proportion of migrant workers with university education or above increased over the years. The descriptive statistics of demographic characteristics suggest that the samples of our study are consistent with the basic situation of migrant workers in China, thus making them suitable for further multidimensional poverty measurement.

3.2. Poverty Rate

Table 5. Poverty rate of Chinese migrant workers.

Dimension	Indicator	2014	2016	2018	2020
Economic level	Income (Inc)	40.72	39.81	26.48	26.18
Social opportunities	Education (Edu)	35.22	32.91	25.19	22.80
	Healthy (Hea)	6.14	5.72	7.01	6.92
	Chronic disease (CD)	7.77	6.87	7.18	7.40
Transparency guarantees	Labor contract (LC)	49.95	49.14	49.78	45.68
Protective security	Medical insurance (MI)	9.07	8.90	9.66	10.44
	Employment security (ES)	65.57	51.02	53.41	54.61
	Provident fund (PF)	74.22	55.10	63.16	67.52
Political rights	Labor union (LU)	82.88	86.65	94.69	93.34
Spiritual poverty	Life satisfaction (LS)	10.30	13.25	7.07	6.35
	Subjective poverty (SP)	38.64	42.14	31.30	29.44

and Figure 2 show the poverty rate of the Chinese migrant workers for each indicator in 2014, 2016, 2018, and 2020. From the trend of the unidimensional poverty rate, the poverty rate of migrant workers in terms of Inc, Edu and LS decreased over the years. The difference between the poverty rate for Inc and Edu in 2014 and 2020 was over 10.0%, which indicates that the income treatment, overall educational level, and life satisfaction of migrant workers are improving. Although the poverty rate for ES, PF, and SP exhibits several peaks, there was still a large decline from 2014 to 2020. The poverty rate for ES, HP, and SP in 2020 was 10.96%, 6.70%, and 9.20%, respectively. These values are lower than those in 2014. The changes in Hea, CD, and MI remained stable, indicating that migrant workers were in a healthy condition or engaged in medical insurance.

It should be noted that the overall value of the labor contract poverty rate is high. From 2014 to 2020, the labor contract poverty rates were close to 50%, which indicates that at least nearly half of migrant workers do not have a labor contract. In addition, compared with other indicators, the poverty rate of the union indicator has the highest value with an increasing trend. The union poverty rate in 2020 was 10.46% higher than that in 2014, which indicates that approximately 90% of migrant workers do not participate in unions. This low union participation demonstrates a high poverty rate in terms of the political rights dimension for migrant workers.

Figure 2 shows the mean values of the poverty rate from 2014 to 2020. The indicators of ES, PF, and LU are over 50%. The indicators of Hea, CD, MI, and LS are lower than 10%. The indicators of Inc, Edu, LC, and SP are between 10% and 50%.

3.3. The Trend towards Multidimensional Poverty

Figure 3 presents a line graph for the multidimensional poverty measurement results of Chinese migrant workers based on the A–F methodology, including the multidimensional poverty index (M), multidimensional poverty incidence (H), and average deprivation score (A). In general, as the value of deprivation k increases, the M-values and H-values of migrant workers gradually decrease, while A-values gradually increase. As mentioned in Section 2.3, the A-values are the result of the number of deprived dimensions to the total number of dimensions. This indicates that the larger the value of deprivation k, the lower the H-values and the M-values. Furthermore, with high values of the deprivation dimension numbers, the A-values increase.

From the comparison results from 2014 to 2020, when k = 0.4, the M-values of migrant workers show a decreasing trend (0.1924 > 0.1800 > 0.1730 > 0.1576). When k = 0.1, 0.2, 0.3, and 0.5, the M-values decrease. When k = 0.6, 0.7, 0.8, and 0.9, the M-values are stable. This indicates that the multidimensional poverty of migrant workers stays at the same level within the high range of the deprivation value domain. These results demonstrate that the M-values of the overall migrant worker sample have a decreasing trend in the lower dimension (k ≤ 0.5) and a stable state in the higher dimension (k > 0.5). When k = 1, the A-value equals 1, representing the full deprivation. However, based on the H-values of migrant workers in each year (H = 0 when k = 1), none of migrant workers faced extreme deprivation in the full dimension.

The mean results of the H-values, A-values, and M-values are shown in

Table 6. The mean of multidimensional poverty outcomes.

k	Hmean	Amean	Mmean
0.1	0.9069	0.3532	0.3207
0.2	0.7480	0.3978	0.2979
0.3	0.5677	0.4469	0.2539
0.4	0.3422	0.5134	0.1758
0.5	0.1584	0.5924	0.0938
0.6	0.0561	0.6793	0.0381
0.7	0.0168	0.7607	0.0128
0.8	0.0044	0.8357	0.0037
0.9	0.0006	0.9286	0.0005
1	0.0000	1.0000	0.0000

and Figure 4. When k = 0.1, the H-value, M-value, and A-value are 90.69%, 0.3207, and 0.3532, respectively. This reveals that 90% of migrant workers are deprived in at least four indicators. When k = 0.4, the H-value, M-value, and A-value are 34.22%, 0.1758, and 0.5134, respectively. This indicates that at least one third of migrant workers are deprived in three dimensions. When k = 0.6, the H-value, M-value, and A-value are 5.61%, 0.0381, and 0.6793, respectively. This shows that only 5.61% of migrant workers face deprivation in more than seven indicators. When k > 0.6 and the H-value < 2%, migrant workers suffered from higher multidimensional deprivation from 2014 to 2020.

3.4. The Persistence of Multidimensional Poverty

Figure 5 presents pie charts representing the multidimensional poverty persistence rate among migrant workers. In previous studies, k = 0.3 or k = 1/3 was used as the cutoff to define whether individuals or households were in multidimensional poverty [44,45]. Since our study measures multidimensional poverty in 6 dimensions and with 11 indicators, we chose k = 0.3 and k = 0.4 as the poverty cutoffs to examine the multidimensional poverty persistence among migrant workers.

There is significant difference in poverty duration between k = 0.3 and k = 0.4. When k = 0.3, the rates of two periods, three periods, and four periods of poverty persistence are 35.48%, 21.51%, and 43.01%, respectively. In other words, the rate of four periods of poverty persistence is the highest, referring to a long-term poverty situation from 2014 to 2020. When k = 0.4, the rates of two periods, three periods, and four periods of poverty persistence are 50.65%, 29.22% and 20.13%, respectively.

4. Discussion

The rapid modernization and transformation of China has brought about the large-scale movement of migrant workers. China’s marketization process has provided opportunities for the rural surplus labor force to enter into cities. However, due to the discriminatory rural–urban dualization policy, migrant workers are blocked from employment, social security, welfare, and public services. Therefore, many of them are forced into the informal labor market. Some studies have shown that educational level is an important factor influencing migrant workers’ poverty, but the multidimensional poverty of migrant workers is rooted in their exclusion from the household registration system [46]. In this study, we combined the characteristics of the Chinese migrant workers and the social environment. We constructed a multidimensional poverty index for migrant workers in China in six dimensions. Our study found that the multidimensional poverty of Chinese migrant workers is more serious than was expected. According to the results of our full-sample measurement, approximately 90% of migrant workers are deprived with respect to at least four indicators. This high poverty rate is basically consistent with the measurement results of Zhou et al. Our findings suggest that poverty among migrants requires urgent national and academic attention [47].

With the improvement of the urban economy, although the income of migrant workers has increased, it is difficult for them to catch up with the growth of median urban per capita income. Even if the income growth brings the decrease in the multidimensional deprivation rate, migrant workers are still cannot afford the high costs of urban living [48]. This study measures the multidimensional poverty index of Chinese migrant workers. It shows a decreasing trend over the years, with a decrease of 18% in 2020 compared to that in 2014. However, it needs to be considered that the COVID-19 pandemic after 2020 has posed a serious challenge to poverty reduction worldwide. COVID-19 has caused mass unemployment, a return to poverty, and the vulnerability of individual and family livelihoods, which has pushed some back into poverty [16]. Therefore, post-pandemic migrant poverty and poverty reduction efforts should be considered in future poverty studies.

Compared to income poverty, multidimensional poverty is less volatile [44]. The statistics of the multidimensional poverty incidence of migrant workers in our study also show that income is not sensitive to poverty evaluation. This shows that income is not the main reason behind the multidimensional poverty measurement for migrant workers, but it is also an indispensable indicator as a classical poverty measurement index. Some studies have shown that income poverty does not have strong explanatory power for the deprivation of living standards and employment rights [49]. This conclusion is further confirmed in our study. Table 6 shows the results of Spearman’s correlation test among different deprivation indicators of multidimensional poverty. As shown in Table 6, the Spearman correlations between income and other indicators are not always strong, although income is generally claimed to be a measure of poverty that reflects deprivation in other dimensions. Our analysis shows that income deprivation is not correlated with some aspects of deprivation, including chronic disease and life satisfaction. Additionally, the correlation between income and health, medical insurance, and labor union poverty was not significant during 2014–2020 (

Table 7. Spearman correlation coefficient between deprivations.

Index	Income
	2014	2016	2018	2020
Education	0.0376 ***	0.0468 ***	0.0443 ***	0.1192 ***
Health	0.0515 *	0.0247	0.0690 *	0.0759 *
Chronic disease	0.0235	0.0127	0.0092	0.0210
Labor contract	0.1179 **	0.0751 **	0.0396 *	0.1485 **
Medical insurance	0.0657 **	0.0193	0.0211	0.0372 **
Employment security	0.1287 **	0.0579 **	0.1297 **	0.1621 **
Provident fund	0.0944 **	0.2212 ***	0.0385 ***	0.0721 **
Labor union	−0.0035	0.0085	0.0745 *	0.0891 **
Life satisfaction	0.0138	0.0073	0.0257	0.0036
Subjective poverty	0.0921 **	0.0359 **	0.0670 *	0.0900 **

Note: *** p < 0.001, ** p < 0.01, * p < 0.05.

). Therefore, it is necessary to identify the poverty of migrant workers from multiple dimensions and with long-term study.

An unstable labor relation is a key factor for poverty, which leads to a low social status and social security [50]. This is corroborated by our study. Migrant workers have a high poverty rate in terms of employment security, labor contracts, and housing fund indicators in our study. In addition, education is considered the main determinant of poverty [51], but our study shows that in terms of the poverty rate of migrant workers, the education indicator is not an essential cause. The education poverty rate is between 20% and 40%, which is not the highest among all indicators. In the unequal labor market, some studies suggest that a high education level cannot ensure freedom from poverty and unstable employment. Additionally, this instability in work is gradually becoming a “new normal” [52]. The growing rate of employment without labor contracts results in not only fewer working opportunities but also poverty. Such unequal labor relations have been internalized as a market “norm” and passively accepted by most workers, which covers up the seriousness of the poverty problem [53]. By comparing the results of different years, our study found that the poverty rates related to the labor contract indicator are high. This confirms the findings of Jo McBride and Andrew Smith, suggesting that there is a “routinization” of in-work poverty. The precariousness of employment reduces an individual’s ability to resist risks, increases their vulnerability to poverty, and weakens the adaptation and social integration of migrants [54].

Poverty is a dynamic process with continuity, and should be analyzed from both a static and dynamic perspective [55]. While the angle of multidimensional poverty has been widely used in poverty research, the Chronic Poverty Research Center (CPRC) has proposed the concept of “chronic poverty” based on the poverty trap theory, which aims to examine the dynamics of poverty from the perspective of longitudinal research. Chronic poverty is defined as the situation in which people live below the poverty line for five years or more. Causing difficulty in earning a livelihood, chronic poverty represents a severe problem in a country [56]. Chronic poverty has a close relationship with multidimensional poverty. Multidimensional poverty can be estimated in both the long and short term [57]. Chronic poverty can be measured from a multidimensional perspective [58]. The results of our study confirm the strong connection between chronic poverty and multidimensional poverty. In terms of poverty duration, about 30% of migrant workers were in continuous poverty from 2014 to 2020, which reveals a possibility that they might have been in multidimensional poverty for seven years. This indicates that the multidimensional poverty of migrant workers has a strong persistence. That is to say, the problem of chronic poverty is serious. This study also found that an increase in the poverty cutoff leads to a decrease in the value differences among the poverty indicators. This result shows that migrant workers in severe multidimensional poverty suffered from comprehensive deprivation, which made it more difficult for them to leave poverty. Therefore, ignorance of multidimensional and persistent poverty may result in extreme poverty and irreversible social effects. This poverty problem related to migration also exists in other countries [59,60,61]. However, many studies have researched migration poverty based on household or income poverty. Our study measures poverty and its persistence for Chinese immigrant workers individually. This could be used as a methodological sample for other countries to study individual and chronic poverty.

5. Conclusions

Based on Sen’s capabilities approach, the A–F methodology and the Chinese Family Panel Studies data during 2014–2020, this study examines the multidimensional and persistent poverty of migrant workers in China. The multidimensional poverty of migrant workers was analyzed from a static to a dynamic perspective. This provided a comprehensive framework, which has been seldom covered in current research. From a static perspective, one third of migrant workers are in multidimensional poverty. The in-work poverty of migrant workers is serious, which is reflected by continuous and unstable labor relations. In addition, income, education, and health are not the main determinants of migrant workers’ multidimensional poverty. Income poverty is not a key factor in the multidimensional deprivation of migrant workers. From a dynamic perspective, the multidimensional poverty of migrant workers was obviously alleviated. The multidimensional poverty index decreased by 18% from 0.1924 in 2014 to 0.1576 in 2020. However, approximately 30% of migrant workers were in multidimensional poverty in 2014, 2016, 2018, and 2020, which may indicate that these migrant workers have suffered from persistent poverty. Specifically, because the data are sampled every two years, we cannot analyze the poverty of these immigrants annually, which is also the main limitation of this paper. The persistent poverty problem of these migrant workers needs attention and help from the government, as they are unable to leave poverty through their own efforts.