The Capability Approach to Adolescent Poverty in China: Application of a Latent Class Model

Abstract

After 2020, poverty alleviation in China will shift from eliminating absolute poverty to alleviating unbalanced and insufficient relative poverty. Based on Amartya Sen’s capability approach, this article aimed to realize the freedom of “capability” of children and ensure the well-being and development of Chinese adolescents, who are often ignored in child poverty research. The study sought to estimate the 10–15-year-olds in a multidimensional capabilities poverty group. Our analysis was based on the adolescent capability methodology framework, using data from the 2018 China Family Panel Studies (CFPS) to investigate the types and influencing factors of adolescent capability poverty in China. The result of latent class analysis showed that there are four types of capability poverty among the Chinese adolescents, namely, Extreme Capability Poverty Class, Goal Capability Poverty Class, Opportunity Capability Poverty Class, and Capability Non-Poverty Class. Multinomial logistic regression showed that the personal factors of gender, ethnic minority, hukou, and pocket money; the family factors of parents’ marital status, parents’ education level, and region of residence; and the family economic factor of per capita family income had a significant impact on the types of China’s adolescent capability poverty. The article provides empirical and quantitative evidence for the adolescent (ages 10–15) class of capability poverty. The findings provide a reference for policy makers to target poverty-alleviation efforts according to different types of poverty and to interrupt the intergenerational transmission of poverty.

1. Introduction

Children are the future and hope of a nation. Yet, in reality, child poverty is widespread. According to the United Nations International Children’s Emergency Fund (UNICEF), 1 billion children worldwide live in multidimensional poverty—without access to education, health, housing, nutrition, sanitation, or water; 356 million children live in extreme poverty and are forced to live on less than $1.90; children are twice as likely as adults to suffer from extreme poverty [1].

In addition, child poverty is a complex and unique problem. It is not only a continuation and a derivative of family poverty [2]. At the same time, the children are not attachments of their parents, also not the be recipient by any decides [3]. The study of child poverty must not deviate from the independence and special needs of children, as this might set the stage for the intergenerational transmission of poverty and fall into the poverty trap. The experience of poverty in childhood has a fundamental impact on the future, weakening future development opportunities, and affecting future health development. Child poverty is more than just material poverty, it also means that children are unable to lead healthy lives, develop intellectual skills, and have adequate physical and emotional conditions [4]. Child poverty is, therefore, a problem facing every country and its eradication is a goal of every country.

Great achievements have been made in the development of child welfare and protection systems since the founding ceremony of the the People’s Republic of China.According to China’s Child Welfare and Protection Policy Report of 2019, the incidence of extreme poverty was 1.7%, and the incidence of poverty among rural children was only 3.9% in 2018 [5]. In 2020, China completed the arduous task of eradicating absolute poverty and achieved the poverty reduction target of “The 2030 Agenda for Sustainable Development” 10 years ahead of schedule [6]. Therefore, after 2020, China’s poverty alleviation work shifted from eliminating absolute poverty to alleviating unbalanced and inadequate relative poverty [7]. There has been a gradual increase in consensus that “Spiritual Poverty Alleviation” is the next stage of China’s poverty control focus [8]. While Amartya Sen proposes a capability approach to understanding relative poverty, proponents of the capability approach argue that a more comprehensive understanding of human well-being should not be limited to material projects. They argue that it should extend to how much freedom people have in their lives [9]. Therefore, to carry out a more comprehensive analysis of poverty, the study of child poverty should be shifted from the perspective of material deprivation to the perspective of capability as well as to the perspective of freedom, opportunity, and development.

International research on child poverty started early through the perspective of capability. For instance, Biggeri et al. [10] studied multidimensional poverty among children aged from 5 to 15 in Afghanistan. They considered child poverty as the deprivation of basic capabilities and related achieved functionings. The situation of Afghan children was analyzed in accordance with the A-F method. This method takes into account the concept of child well-being—including love and care, which is often missing in standard surveys—through a multidimensional approach. The results of the study showed that children between the ages of 5 and 7 are the poorest when non-material aspects such as care and respect are taken into account. Wüst and Volkert [11] studied the multidimensional poverty analysis of 5-to-6-year-old children from the perspective of capability using the A-F method to study the situation of German children in the four dimensions of education/leisure, health, social participation, and income. Potsi et al. [12] presented an approach based on a fuzzy methodology applied to data from the EU-SILC 2009 ad hoc module on children; the study revealed the Italian children’s living conditions and their capabilities deprivation.

Compared with other countries, there have been many empirical studies on child poverty from the perspective of capability. However, Chinese researchers have not significantly engaged in the measurement of child poverty from the perspective of capability, and instead most of them focused on the functional level of poverty [2,13]. Feng et al. [14] used the A-F method to study the current situation of multidimensional poverty among children in poverty-stricken areas focusing on five dimensions: survival, development, health, education, and participation. Li and Yang [2] used the A-F method to explore the degree, characteristics, and mechanisms of multidimensional poverty among Chinese children based on six dimensions: nutrition, health, education, security, culture, and social interaction. However, some researchers believe that addressing poverty in the new era must focus on capability building [15,16]. Lin, Chen, and Zhao [9] analyzed the Chinese adolescents in the 2014 CFPS study from the perspective of capability. Using the A-F method viewed through the capability lens, they mapped various profiles of functionings and capabilities poverty according to the features of these adolescents.

In terms of poverty measurement, the Multidimensional Poverty Index (MPI) of Alkire and Foster (A–F method), first published by the United Nations in 2010, is an important reference for studying multidimensional poverty of children. However, the MPI focuses on solving transnational absolute poverty assessment and is not necessarily suitable to China’s new era of relative poverty assessment. The A-F method can reflect the extent of poverty and can also analyze the depth of poverty. The decomposability of the index provides the possibility for comparative analysis of the mechanism of poverty and of different regions [17]. The contribution of each dimension can be examined, but since it is transformed from a different dimension to a single dimension, this can result in a loss of information about which dimensions are relatively more serious [18]. As a result, to provide effective measures to eliminate poverty, we can classify different types of poverty through different levels, and then provide targeted poverty alleviation to different types of poverty groups.

It is against this background that this study primarily focuses on the following two research questions: (1) what types of capability poverty can be classified for Chinese adolescents and (2) what factors will affect the type of ability poverty among Chinese adolescents.

The capability of children is different from that of adults, and to a large extent it depends on the age of the children [11]. The age of the adolescent child (10–15) is the time when the world view, the outlook on life, and the values begin to take shape. It is the peak stage of physical, emotional, and spiritual development. Therefore, it is significantly important to study the types of child poverty in this particular age group for targeted poverty alleviation with the view to eliminating intergenerational poverty. This article analyzes the poverty of Chinese adolescents and children from the perspective of capability. The latent class model was used to perform endogenous aggregation of the sample to explore the types of poverty among the Chinese adolescents. In addition, multidimensional logistic regression analysis was used to study the factors affecting the types of capability poverty among Chinese adolescents. As indicated earlier, the purpose of this article is to explore the types of Chinese adolescents’ capability poverty and to identify Chinese adolescents’ poverty according to the capability approach. The findings provide a reference for policy makers to target poverty-alleviation efforts according to the different types of poverty and to interrupt the intergenerational transmission of poverty through the capability approach.

The remainder of the article is organized as follows: Section 2 covers the theoretical background of the study; Section 3 reviews the data and methods used; Section 4 introduces the findings of the study; and lastly, Section 5 presents the discussion and conclusions.

2. Theoretical Background

The capability approach emerged from the criticism of material resources which cannot reflect the key factors such as human rights and capability, among others. Sen argues that “The role of the capability approach in poverty analysis is to enhance understanding of the nature and causes of poverty and deprivation by shifting the focus from material concerns to well-founded ends and, accordingly, to freedoms that satisfy those ends [19]”. The proponents of the capability approach argue that the capabilities approach begins with a very simple, yet, at the same time, highly complex question: ‘what are people really able to do and to be?’ The answer to that question is the set of capabilities or real opportunities the individual person has [20]. From the perspective of capability, child poverty is defined as the deprivation of basic capabilities and related achieved functionings; in this case, functionings refer to a person’s achieved doing and being [9]. Essentially, this focuses on the achieved life status, such as being healthy, being educated, and living a long life, among other aspects [21]. Capability is defined as “the various combinations of functionings that the person can achieve”; that is, “The substantive freedom he or she enjoys to live the kind of life he or she has reason to value [22]”. This includes the ability to lead a healthy life and the ability to achieve education, among others. The concept of capabilities embraces the duality of the “Freedom of opportunity” provided by the environment and the “Freedom of choice” imposed by the adolescent subject [23]. In this sense, capability is both an opportunity and an individual’s ability. In addition, the capability approach researchers are interested in concerns which factors influence adolescents’ capabilities. Robeyns [24] refers to these factors as conversion factors, which enable a person to successfully convert his or her resources into capabilities; these include gender, social class, and geographical location.

3. Data and Methods

3.1. Data

In this section, a series of methodological issues are covered including databases, measurement dimensions, indicators, and statistical methods used.

The study involved using secondary data and quantitative methods. The data used in this article were from a large-scale data survey—China Family Panel Studies (CFPS) implemented by the institute of Social Science Survey (ISSS) of Peking University in 2010 to estimate Chinese adolescents’ capability poverty.

The CFPS survey covered 25 provinces, cities, and autonomous regions which represent 95% of China’s population, and utilized a multi-stage, implicit hierarchical and population-scale sampling method to ensure the representativeness of the sample. The questionnaire not only covered a wide range of topics, but also included comprehensive interview modules in both rural and urban areas. The information was collected on family structure and family members, migrant population, event history (e.g., history of marriage, education, and employment), cognitive ability, and child development [25].

In this study, we obtained answers from a total of 2607 adolescents (aged between 10 and 15) in our 2018 survey analysis. However, for some unanswered questions we needed to discard them and eventually used 1732 eligible adolescents (aged between 10 and 15) for further analysis in our study.

3.2. Indicators and Dimensions

Although Nussbaum [20] lists 10 capabilities based on the capabilities perspective, research experience shows that capability may be age-related, and the degree of correlation can vary with age [19]. The setting of capability indicators depends to a large extent on the age of the child [11].

First, we draw on Lin’s selection of measurement dimensions for functionings and capabilities. The functionings are divided into four dimensions, including physical health, mental health, education, and parental care. Lin believes that capabilities can be divided into skill-based capabilities which refer to the opportunities for children to acquire hard skills and soft skills that are useful for future development. Opportunity-based capabilities refer to the opportunities that young people have to participate in meaningful activities; goal-based capabilities refer to the importance of people’s goals and aspirations for the future; and lastly, potential-based capabilities refer to people’s talents in life [9].

Based on the availability of data, knowledge of the types of indicators used in other national indicators, shared capability approach supporters, and indicator data available for multidimensional indicators in China [19,26,27], we selected and measured 9 appropriate indicators.

Deprivation was calculated as a dichotomous variable and for each indicator, deprived adolescents were given a value of 1 (have fallen into poverty) and non-deprived adolescents were given a value of 0 (have not fallen into poverty) (see

Table 1. Measuring dimensions, indicators, and cutoffs for functionings and capabilities.

Domain	Indicators	Cutoff
Physical health	Exercise for teens	The adolescent is judged deprived if he/she reported he/she exercised 0 exercises in the past week
	Unhealthy items	The adolescent is judged deprived if he/she reported that he/she has consumed unhealthy items (e.g., cigarettes, alcohol) in the past month
	Adolescents self-rated health	The adolescent is judged deprived if he/she considered his/her health status was generally low
Mental health	Depression	Adolescents were judged deprived if they scored too high on the depression scale
	Adolescent’s interpersonal relationship	Adolescents were judged deprived if they perceived bad or poor relationships
Education	Adolescent’s satisfaction with their studies	The adolescent was judged deprived if he/she was dissatisfied or very dissatisfied with his/her study
	Adolescent’s satisfaction with their school	Adolescents were judged deprived if they were dissatisfied or very dissatisfied with the school
Parental care	Parents know the whereabouts of adolescent	The adolescent was judged deprived if he/she answered that his/her parents did not know about his/her whereabouts
	Parents are concerned about adolescent learning	If the adolescent’s parent never or rarely cared about the adolescent’s homework, he/she would be identified as deprived in this indicator
Skill-based capabilities	The performance of adolescent in Chinese and mathematics	Adolescents were judged deprived if they performed poorly or very poorly in any two courses
	Adolescent’s rule-abiding behaviors	The adolescent was judged deprived if he/she had little or no adherence to the rules
Opportunity-based capabilities	Opportunities for adolescent to read after-school books	Adolescents were judged deprived if they had never had the chance to read entertainment books in the past 12 months
	Internet access for adolescent	Adolescents were judged deprived if they never accessed the Internet in their daily life
Goal-based capabilities	Adolescent’s goal-setting	Adolescents were judged deprived if their education target was high school
	Adolescent’s schedule	Adolescents were judged deprived if they did not follow their own schedule
	Adolescent’s self-control in his/her life	Adolescents were judged deprived if they had low self-control of their life
Potential-based capabilities	Adolescent’s self-report ofhis/her own good quality	The adolescents were judged deprived if they reported that they had no or little good qualities
	Adolescent’s self-affirmation	The adolescent was judged deprived if he/she had low or no self-affirmation

).

3.3. Analytical Statistical Method

Currently, poverty is regarded as a multidimensional concept. This study used latent class analysis (LCA) as a measurement method to explore capability poverty. LCA is a statistical method that can explain the relationship between indexes through the category latent variable and then realize the local independence of indexes. The LCA method is similar to factor analysis (FA), but there are essential differences between them. The focus of FA is to classify the variables, while the focus of LCA is to classify the observations [28]. The LCA establishes and identifies the relationships among the observed variables through the presence of latent variables; however, it is difficult to observe directly because poverty is an abstract concept [29]. In this particular analysis, poverty was considered as a latent variable and was measured through clear and observable socio-economic indicators [30].

LCA is not a division of groups by subjective judgment but rather an ex post segmentation in which the groups are classified together by subdividing them into the sum of distinct sub-distributions with individual means and variances, and then determining their nature; statistical validation is also possible [31]. At the same time, LCA analysis does not rely on traditional modeling assumptions, so it is not necessary to meet a series of assumptions (linearity, multivariate normality, or homoscedasticity) which cannot be met by the types of indicators normally available [32]. Furthermore, the LCA is well suited for the goal of multidimensional analysis, separating the population into different groups that reflect varying degrees of poverty, and into different groups based on similar socioeconomic characteristics; this allowed us to study the size and distribution of poor populations [30]. The statistical principles of LCA analysis are mainly conditional probability and Bayesian formulas; unlike other classification techniques, such as cluster analysis or k-means clustering, LCA is based on the model, and LCA is a cross-sectional model with definite results in which binary indicators are most commonly used in practice [33]. Suppose that the model contains a total of $r$ observed items, all with 0–1 values, and that the association between the indicators is explained by the latent class variable $c$. With $k$ classes, as shown in Equation (1) [34]:

$$P\left(y_j=b_j\right)={\displaystyle \sum }_{k=1}^{K}P\left(c=k\right)P\left(u_j=1|c=k\right)$$

(1)

In LCA, there are two main parameters: $P\left(u_j=1|c=k\right)$ is the conditional probability specific to a given class, which describes the probability of the observation of the $k$ class taking $b_j$ on the $j$ index. The $P\left(c=k\right)$ class probability specifies the relative size of the k class and the probability that any observation belongs to the $k$ class [35].

Assuming that the conditions are independent, the joint probability of all $r$ observed items is:

$$P\left(u_1,u_2,\dots u_r\right)={\displaystyle \sum }_{k=1}^{K}P\left(c=k\right)P\left(u_1=1|c=k\right)P\left(u_2=1|c=k\right)\dots P\left(u_r=1|c=k\right)$$

(2)

Based on the joint probability, conditional probability, and potential class probability, we used Bayes’ formula to estimate the posterior probability of a class, which represents the probability that a given observation is classified into a given class [35], as follows [34]:

$$P\left(c=k|u_1,u_2,\dots u_r\right)=\frac{{{\displaystyle \sum }}_{k=1}^{K}P\left(c=k\right)P\left(u_1=1|c=k\right)P\left(u_2=1|c=k\right)\dots P\left(u_r=1|c=k\right)}{P\left(u_1,u_2,\dots u_r\right)}$$

(3)

Because we can get two or more classes by LCA, we used multinomial logistic regression analysis to explore the conversion factors that affect the adolescent capacity poverty. The multinomial logistic regression model is a simple extension of the binary logit model, which allows multiple dependent variables [36], and it can be regarded simultaneously as an estimation of multiple binary logistic regression models composed of two pairs of each class in the explained variables.

$$logit\left(P\left(Y=j\right)\right)=\ln \left[\frac{P\left(Y=j\right)}{P\left(Y=i\right)}\right]={\alpha }_i+{\beta }_{1j}{X}_1+\dots +{\beta }_{pj}{X}_j+\mu \left(j\ne i\right)$$

(4)

In Equation (4), $Logit(P\left(Y=j\right)$ is the dependent variable, indicating the class of capability poverty to which the Chinese adolescents and children belong; $X_j$ is the explanatory variable indicating each influencing factor affecting the class of capability poverty; ${\alpha }_i$ indicates the model intercept;${\beta }_{pj}$ indicates the corresponding regression coefficient; $j$ is the number of explanatory variables; and $\mu$ is the random error term that follows an extreme value distribution.

The explanatory variables included the individual factors, family factors, and family economic factors. In order to further study the factors that affect the types of Chinese adolescents’ capability poverty, a multinomial logistic regression was conducted using the Capability Non-Poverty Class as the reference class.

Table 2. Descriptive statistics.

	Variable	Variable Definition	Mean/Percentage (%)	Standard Deviation
Individual factors	Gender	Female = 0	47.11
		Male = 1	52.89
	Age	10–12 = 0	49.13
		13–15 = 1	50.87
	Ethnic minority	No = 0	88.86
		Yes = 1	11.14
	Hukou	Rural = 0	81.93
		Non-Rural = 1	18.07
	Pocket money	No = 0	22.23
		Yes = 1	77.77
Family factors	Parents’ education level	Below primary school level = 0	17.38
		Primary-junior high school level = 1	54.04
		Above junior high school level = 2	28.58
	Parents’ marital status	Married = 0	94.23
		Widowed and Divorced = 1	5.77
	Family size	1–3 = 0	17.21
		4–6 = 1	69.00
		>6 = 2	13.80
	Region of residence	West = 0	33.60
		Central = 1	32.56
		East = 2	33.83
Family economic factors	Per capita family income	Ln (family monthly income)	9.46	0.86

provides descriptive statistics for these conversion factors.

3.4. Analysis

In this study, we used Mplus8.3 statistical software to carry out the latent class analysis (LCA). From the initial model, the number of classes was gradually increased in the model until the model with the best fit to the data was identified. Stata15.0 statistical software was used to analyze the data, and descriptive statistical analysis for measurement and count data was performed to explore the conversion factors affecting adolescent capability poverty. We used a multinomial logistic regression model.

4. Results

4.1. The Number of Poverty Groups

The selection of the model was based on fitting statistics according to the Akaike information criterion (AIC), Sample-Size Adjusted Bayesian Information Criterion (SABIC), entropy, and Lo–Mendell–Rubin adjusted likelihood ratio test (LMR); the number of models with the best goodness of fit and their potential classes were determined [33]. The smaller the AIC, BIC, and SABIC, the better the classification results when the entropy is close to 1. LMR is mainly used for fitting the difference between K-1 and K classes; a significant LMR value indicates that the k-class model is superior to the k-1 class model. The homogeneity of classes should include at least 5% of the total observed values [4]. Furthermore, since the fit index does not necessarily point to the only solution, the interpretability of the statistical fit index and the model should be considered jointly when selecting the appropriate class [37].

Latent class analysis was used to fit the model of 1–5 classes (children were divided into 1–5 classes). As shown in

Table 3. Model fit. AIC: Akaike information criteria; BIC: Bayesian information criteria.

Number of Classes	AIC	BIC	SABIC	Entropy	LMR	Class Proportions
1	16,407.26	16,450.91	16,425.50			1.00
2	16,156.93	16,249.70	16,195.69	0.460	0.0000 ***	0.74/0.26
3	16,133.06	16,274.94	16,192.34	0.581	0.1077	0.06/0.76/0.18
4	16,127.78	16,318.78	16,207.58	0.682	0.0392 **	0.05/0.73/0.09/0.13
5	16,127.1	16,367.21	16,227.43	0.659	0.1344	0.06/0.05/0.12/0.64/0.13

** p < 0.05, *** p < 0.01.

, the AIC value decreased with the increase of the number of classes, and the 4-class model fit better; BIC showed that 2-class models fit better; SABIC showed that 3-class models fit better; entropy showed that 4-class models fit better; and LMR showed that 4-class models fit better than 3-class models. Considering the simplicity and practical significance of the model, we chose 4-class models as the final model.

4.2. Types of Latent Classes

For the latent classes of adolescent capability poverty based on the conditional probability of each dimension, Figure 1 and

Table 4. Results in probability scale.

Dimensions	State of Poverty	Latent Classes
		C1 (n = 90)	C2 (n = 1266)	C3 (n = 160)	C4 (n = 216)
Physical health	Non-deprivation	0.353	0.65	0.744	0.909
	Deprivation	0.647	0.35	0.256	0.091
Mental health	Non-deprivation	0.427	0.687	0.776	0.954
	Deprivation	0.573	0.313	0.224	0.046
Education	Non-deprivation	0.274	0.832	0.977	0.99
	Deprivation	0.726	0.168	0.023	0.01
Parental care	Non-deprivation	0.441	0.604	0.778	0.75
	Deprivation	0.559	0.396	0.222	0.25
Skill-based capabilities	Non-deprivation	0.335	0.889	0.978	0.934
	Deprivation	0.665	0.111	0.022	0.066
Opportunity-based capabilities	Non-deprivation	0.35	0.529	0	1
	Deprivation	0.65	0.471	1	0
Goal-based capabilities	Non-deprivation	0.046	0.38	0.953	0.788
	Deprivation	0.954	0.62	0.047	0.212
Potential-based capabilities	Non-deprivation	0.530	0.651	0.960	0.947
	Deprivation	0.470	0.349	0.040	0.053

show the conditional probability of each dimension. The first class is Extreme Capability Poverty Class, which had a higher deprivation probability in other dimensions except the potential capability dimension. Although not deprived in the dimension of potential capability, the probability of being deprived was higher than in other classes. The LCA model estimated that 5% of adolescents fell into this class.

Class 2, the Goal Capability Poverty Class had an incidence of 73% and the majority of adolescents fell into this class. This class showed deprivation only in the goal-based capability dimension, but had a lower probability than the Extreme Capability Poverty Class. In other dimensions, although it did not show deprivation, the deprivation probability is much higher than the third and fourth classes.

Class 3 is the Opportunity Capability Poverty Class, which showed a higher probability of deprivation only in opportunity-based capabilities and a lower probability in other dimensions. A total of 9% of adolescents were in this class.

Class 4 is the Capability Non-Poverty Class, which had a low probability of being deprived in all dimensions, and had a total of 13% of adolescents belonging to this category.

4.3. Factors Influencing the Children’s Capability Poverty

Because the capability poverty of adolescents was classified into four classes—Extreme Capability Poverty Class, Goal Capability Poverty Class, Opportunity Capability Poverty Class, and Capability Non-Poverty Class—the dependent variable was a four-category variable. We used a multinomial logistic regression model to test the impact of various factors on the type of adolescent capability poverty, and the reference class of the regression model was the Capability Non-Poverty Class. The model estimates are shown in

Table 5. Multinomial logistic model results.

Variable	Extreme Capability Poverty Class		Goal Capability Poverty Class		Opportunity Capability Poverty Class
	RR-Radios	S.E.	RR-Radios	S.E.	RR-Radios	S.E.
Individual factors
Gender: Male	1.922 **	0.519	1.114	0.170	0.668 *	0.145
Age: 13–15	1.444	0.393	0.855	0.133	0.394 ***	0.088
Ethnic minority: Yes	3.418 ***	1.417	1.637	0.533	1.064	0.465
Hukou: Non-Rural	0.292 ***	0.133	0.548 ***	0.104	0.824	0.235
Pocket money: Yes	0.526 **	0.168	0.681 **	0.136	0.915	0.253
Family factors
Parents’ marital status: Widowed and Divorced	1.626	1.129	2.943 **	1.311	2.816 *	1.574
Parents’ education level: Primary-junior high school level	0.500 *	0.181	0.619 *	0.162	0.939	0.330
Above junior high school level	0.385 **	0.163	0.527 **	0.148	0.656	0.256
Family size: 3–6	0.627	0.208	1.156	0.225	1.674	0.534
>6	0.805	0.385	1.449	0.439	1.824	0.799
Region of residence: Central	0.756	0.269	0.952	0.201	0.876	0.245
East	0.808	0.268	0.754	0.154	0.542 **	0.154
Family economic factors
Per capita family income	0.551 ***	0.093	0.667 ***	0.069	0.562 ***	0.081
Constant	372.349 ***	622.201	675.502 ***	717.381	363.084 ***	530.045
Log likelihood	−1401.006 ***
Pseudo R2	0.062

*** p < 0.01, ** p < 0.05, * p < 0.1.

.

Among the individual factors, gender, ethnic minority, hukou, and pocket money had significant effects on the types of adolescent capability poverty. Boys were more likely to be in the Extreme Capability Poverty Class than girls. However, girls were more likely to be in the Opportunity Capability Poverty Class. At ages 10–12, adolescents were more likely to belong to the Opportunity Capability Poverty, and the likelihood of exposure to internet access and reading increased with age. Ethnic minorities were more likely to belong to the Extreme Capability Poverty Class compared with the Capability Non-Poverty Class. In terms of hukou, the agricultural household registration was more likely to be the Extreme Capability Poverty Class and the Opportunity Capability Poverty Class. Regarding having pocket money, people without pocket money were more likely to be in the Extreme Capability Poverty Class and the Opportunity Capability Poverty Class.

Among the family factors—apart from the influence of family size—parents’ marital status, parents’ education level, and region of residence had significant influence on the type of adolescent capability poverty. Adolescents and children with divorced parents and widowed parents were more likely than those with a happy family to belong to the Goal Capability Poverty Class and Opportunity Capability Poverty Class. Adolescents whose parents were below primary school were more likely to belong to the Goal Capability Poverty Class and Opportunity Capability Poverty Class. In the model of the Opportunity Capability Poverty Class/Capability Non-Poverty Class, adolescents who lived in the East were more likely to belong to the Capability Non-Poverty Class than those who lived in the West.

Among family economic factors, adolescents living in families with higher per capita household income were more likely to belong to the Capability Non-Poverty Class.

5. Discussion of Findings

First, among the four classes, 5% of adolescents belonged to the Extreme Capability Poverty Class. The Extreme Capability Poverty Class had a higher probability of deprivation in all dimensions, and adolescents belonging to this class experienced multidimensional capability poverty in both functionings and capability. The reason for the low proportion of such class is inseparable from China’s poverty alleviation work [5].

At the same time, in addition to the Extreme Capability Poverty Class, the deprivation probability of the Goal Capability Poverty Class and the Opportunity Capability Poverty Class in the four dimensions of function (physical health, mental health, education, and parental care) was very low. This finding is consistent with previous research which indicates that China’s child poverty has gradually improved, and children’s multidimensional poverty has been effectively improved in functional dimensions such as housing, care, education, health, and nutrition [38].

The second type is the Goal Capability Poverty Class, which accounted for the highest proportion among adolescents, reaching 73%. It is the type of poverty where we need to address the intergenerational phenomenon of poverty. Although the implementation of the nine-year compulsory education policy has enabled almost everyone to study, there is a cultural system where the poor are often isolated from the rest of the population and have their own codes of conduct, worldviews, and values, such as the uselessness of reading [2]. Therefore, it is necessary to correctly guide the development of adolescents’ goal capability and improve their educational goals.

The third type is the Opportunity Capability Poverty Class, and this accounts for 9% of all adolescents. This type of poverty can be regarded as a special case of the Capability Non-Poverty Class. Apart from the opportunity-based capabilities dimension, the probability of deprivation in this class is not too different from the Capability Non-Poverty Class. There may be two reasons for this: On one hand, internet access remains unequal according to the geographic area where children live [39]. On the other hand, in the multinomial logistics regression, we found that younger adolescents were more likely to be in this class, which suggests that they may have been limited by their parents’ opportunities because of their age, and as they got older, this class became the Capability Non-Poverty Class [40].

The regression results found that Chinese adolescents from ethnic minorities and agricultural hukou were more likely to belong to the Extreme Capability Poverty Class; this finding resonates with existing studies [41,42]. The reason for this may be that in recent years, with the rapid development of urbanization in China, there has been a massive movement of laborers from less-developed areas (such as ethnic areas and rural areas) to urban areas, resulting in a large number of children being left behind in rural areas. Firstly, some left-behind children lack parental enlightenment during infancy, resulting in delayed cognitive development, and lack of parental care can lead to various psychological disorders. Secondly, the lack of good food and medical services makes rural left-behind children more vulnerable to illness and poor health. Finally, the absence of parents also has a negative impact on the education of these children, as they may be discriminated against and bullied at school due to lack of supervision and poor academic performance, resulting in a high dropout rate [41,43]. Pocket money has made a significant contribution to reducing Chinese adolescent capability poverty, and this is consistent with the findings of some capability approach supporters [9,44,45]. Chen and Lin [45] considered that if parents give their children pocket money, it helps to increase the likelihood that they will exit capability deprivation. Lin, Chen, and Zhao [9] reported that there is no significant gender difference in terms of the indexes for functionings and capability poverty. However, this article is inconsistent with their findings; in the classification, boys were more likely to belong to the Extreme Capability Poverty Class and girls were more likely belong to the Opportunity Capability Poverty Class. The younger age group is more likely to be in the Opportunity Capability Poverty Class. The reasons for this may be related to the young age of the children, the technical limitations imposed by the parents, and the increased active mediation of the parents as the children grow older [40].

Among family factors, parents’ marital status and parents’ education level had significant influence on Chinese adolescents’ capability poverty. This finding is consistent with previous research that groups of children have a strong dependence on family adult individuals and are highly vulnerable to influences from their families in terms of growth and developmental dimensions [4,46,47]. Adolescents with divorced parents and widowed parents were more likely than those with a happy family to belong to the Goal Capability Poverty Class and Opportunity Capability Poverty Class. Previous studies showed that children in single-parent families are more likely to grow up poor [48,49]. Parental divorce can cause psychological damage to adolescents and produce negative psychology, leading to deprivation of adolescents’ capability to achieve goals and opportunities, lack of higher educational goals, and poorer autonomy [50,51,52,53]. Higher levels of parental education were negatively associated with adolescents experiencing opportunity capability poverty and goal capability poverty, and the results of this study were fully consistent with previous research findings [53]. Parents with higher educational levels are more likely to choose to educate their children and to provide support and guidance in their children’s learning, so that children have higher educational goals [54].

Data from region of residence showed that adolescents living in the western region were more likely to be in the Opportunity Capability Poverty Class. This is consistent with existing research [9]. Compared with the developed areas in the East, western China may face adverse factors, such as traffic congestion and underdeveloped technology, making the flow of information much more difficult than in the East [55,56].

Although the capability approach places emphasis on income, people with the same income can have different experiences. For instance, differences can arise due to their personal potential and eventually the different social circumstances in which they will live. However, income is still a very important indicator; depending on the economic situation a person can satisfy many wishes and have more opportunities [11,57]. Therefore, raising people’s economic income is still an important means to improving people’s well-being.

6. Conclusions

In this study, we used the concept of multidimensional poverty to study the capability poverty of Chinese adolescents (aged 10–15) through the capability approach. According to the eight dimensions of physical health, mental health, education, parental care, skill-based capabilities, opportunity-based capabilities, goal-based capabilities, and potential-based capabilities, the latent class analysis showed that the poverty types of Chinese adolescents can be divided into the Extreme Capability Poverty Class, Goal Capability Poverty Class, Opportunity Capability Poverty Class, and Capability Non-Poverty Class.

When studying the conversion factors affecting adolescent capability poverty in China, we found that the individual factors of gender, ethnic minority, hukou, pocket money; the family factors of parents’ marital status, parents’ education level, region of residence; and the family economic factor of per capita family income had a significant impact on adolescent capability poverty. This provided instructive empirical evidence for our approach to alleviating child capability poverty. In view of the above findings, we put forward the following recommendations:

First, the capability poverty children should be accurately identified, classified, and registered. This helps to solve the problem of “whom to support”. At the same time, it is important to understand the children’s situation, that is, the causes of poverty and the types of poverty, to facilitate the deployment of appropriate poverty-targeted interventions.

Second, parents should increase their confidence in their children, pay attention to the development of their children’s autonomy, and give their children the opportunity to pursue their own interests and to relax by giving them pocket money. It is important to avoid limiting children’s opportunities for other meaningful activities.

Third, in order to reduce child capability poverty and eliminate intergenerational transmission of poverty, we should also increase adult education and cultural publicity, enhance the spiritual civilization of parents, develop problem-oriented child discipline, cultivate parents’ ability of early education, and provide care services such as recreation, psychological counseling, life care, and home instruction for single parents and children in difficult situations.

Finally, it is important to strengthen the construction of infrastructure in the western region of the country, ethnic minority areas, and rural areas. This will help ensure full coverage of public transportation, water and electricity, communication networks, and universal education. At the same time, the state should strengthen macro-economic regulation and control, raise the people’s economic income, and narrow the income gap.

Compared with other studies, the contributions of this study are: first, using the capability approach to construct a capability and functioning framework for adolescent poverty. Second, the latent class model was used to classify the capability poverty of Chinese adolescents. Unlike the studies of cluster analysis and the A-F method, LCA has advantages in dealing with measurement errors and effectively identifying clusters, and LCA can classify the types of poverty through different dimensions. This, in turn, can help different types of poverty classes by providing precise poverty alleviation interventions. Finally, the further discussion on the influencing factors on different classes of Chinese adolescents provide guidance for policy makers to target poverty-alleviation efforts according to different types of poverty and to interrupt the intergenerational transmission of poverty through the capability approach. Despite these important findings, this study had some limitations that can be addressed in future studies. The research design of this study was cross-sectional, which neither determined the causation nor offered a dynamic perspective. Additionally, most of the capability dimensions were measured by respondents’ self-reports and subjective evaluations; future research should develop indicators that use more objective norms to measure children’s capabilities.