A Study on Mandarin Proficiency and Multidimensional Poverty in Ethnic Areas

To study the role of language in anti-poverty, based on the theory of multidimensional poverty and the economic value of language, this paper analyzes the correlation between Mandarin proficiency and multidimensional poverty in ethnic minority areas by using CFPS data from 2010 to 2018. The regression results show that the improvement of Mandarin proficiency is beneficial to solve multidimensional poverty problems. The endogenetic analysis and robustness test are carried out, and the conclusion is consistent with the baseline regression. Further, the influence of language ability on different samples is analyzed. The study shows that Mandarin proficiency has a greater influence on male, 18-28 and 29-50 years old, high poverty groups and non-agricultural workers.


Literature Review
With the development of education economics and human capital theory, Jacob Marchak (1965), a pioneer of the West and professor of economics at UCLA, put forward the concept of linguistic economics, which explains the essence of language economics: value and utility, cost and income, which opens the prelude of the study of world linguistic economics. As a human capital, language acquisition and use have an impact on income and employment. Therefore, most scholars focus on the relationship between language and income to study language of poverty alleviation. Carliner (1981) takes Quebec province as an example, and finds that people who speak French or English have a greater economic return than those who can't speak both languages, while English speakers have a greater financial return than French speakers. Outside Quebec, the wage level of English speaking people alone is significantly higher than that of those who can neither speak English nor speak French. English are 34% and 13% higher respectively; the wage increase for the English speaking workers is equivalent to the hourly wage of the high school educated workers, and the hourly wage of the college degree workers. Di the impact of Mandarin on their income. The results show that the impact of language ability on the income of workers in China's labor market is between 11.62% and 15.6%, and the impact of Mandarin listening ability and Mandarin expression ability on the occupational income of workers. Jinjiang, Yin Feifei and Lian Jie (2017) based on the 2010 China (Guangzhou) family dynamic tracking survey data, using probit model to investigate the relationship between Mandarin proficiency and employment of Guangdong residents, the research shows that Mandarin proficiency can significantly improve men's employment situation, which is conducive to workers engaged in mining, manufacturing and construction industry , trade services and communications. From the perspective of Mandarin proficiency and income and employment of migrant workers, Qin Guangqiang (2014) analyzed the impact of Mandarin proficiency of migrant workers in Beijing on their income. The study found that migrant workers who are proficient in Mandarin can get 21% -40% of their monthly income higher than those who are not proficient in Mandarin, and those who are proficient in mandarin have better performance in job training, vocational skills, self-worth, job adaptation, etc. Jiang Shan (2017) and Xia Li (2009) found that Mandarin has a significant positive impact on non-agricultural income, and the hourly wage of migrant workers with Mandarin standard is 15.73% higher. More than 70% of migrant workers think that "if they speak Mandarin well, they can find a good job". Wang Hailan (2019)  the impact of language ability on poverty alleviation, the study found that the average annual income of the people who can't speak Mandarin is 17034 yuan less than those who master Mandarin, which is nearly 1500 yuan lower than the average monthly income. Xie Zhiju and Li Qiang (2020) analyzed the ability, demand and income of Mandarin in villages of 12 counties in G Province, and found that Mandarin mainly affects the income of poor households. The income of those who can speak Mandarin is 6965 yuan higher than those who can't speak Mandarin, which is 32.8%. At the same time, the improvement of Mandarin proficiency increases the poverty alleviation rate by 20%.
To sum up, as a human capital attribute, the improvement of lingua franca is conducive to the improvement of individual income and employment level of workers. However, most of the above literatures only studies language and anti-poverty from the single perspective of language, income and employment, while the multidimensional research of language and health, education, living standard and employment is less. Therefore, this paper selects 15 indicators from the five dimensions of economy, health, education, living standards and employment to build a multidimensional poverty index, and analyzes the relationship between Mandarin proficiency and multidimensional poverty.

Data Resource
The data used in this paper are mainly from the national sample of the China Family Panel Studies (CFPS) from 2010 to 2018, which is a biennial follow-up survey conducted by the Chinese Center for Social Science Survey at Peking University. The survey aims to comprehensively reflect the social changes and economic development in China by collecting nationally representative information on villages, families and family members. According to the research of this paper, rural residents in eight ethnic provinces of China are selected as the research object to analyze their multidimensional poverty in the dimensions of economy, education, health, living standard, work.

Indicators Selection
In order to comprehensively present the situation of multidimensional poverty in ethnic areas, we analyze the situation of multidimensional poverty in ethnic areas from five dimensions: economy, education, health condition, work condition and living standard. The economic dimension is replaced by per capita disposable income, while the education dimension is represented by years of education and dropout rate. The health condition is measured from three aspects: self-rated physical health, chronic disease and mental health. Work condition is measured from five aspects: employment, work formality, satisfaction with work environment, satisfaction with job promotion and information resources. Living standard from medical insurance, cooking water, cooking fuel, culture, education and entertainment expenditure ratio of 4 aspects to measure.  The proportion of cultural, educational and entertainment expenditure in total expenditure. If it is less than 40% of the proportion, it is 1, otherwise it is 0.

Multidimensional Poverty Index
Based on the calculation method of multidimensional poverty index, using the CFPS data from 2010 to 2018, we get the multidimensional poverty index at different critical values. The results are shown in Table 2. When k is 30%, the multidimensional poverty index is 0.2836, the incidence of poverty is 0.6385, and the poverty degree index is 0.4442, indicating that 63.85% of individuals are in multidimensional poverty, and the multidimensional poverty degree is 0.4442.  usually sets 30% as the poverty threshold, when the poverty score of the sample is more than 30%, it is considered as poor, otherwise it is non multidimensional poverty. Most of China's multidimensional poverty measurement also takes 30% as the critical value of poverty (Zhang Quanhong, 2015; Gao Ming, 2018; Xiao Rongrong, 2018). Therefore, this paper also takes 30% as the critical value of multidimensional poverty, calculates the multidimensional poverty index, reflects the multidimensional poverty situation in ethnic areas, and discusses the correlation between Mandarin and multidimensional poverty.

Control Variables
The selection of control variables mainly includes three levels, namely, individual level, family level and regional level. Specifically, the control variables at the individual level mainly include the individual's gender, age, marital status, social interaction, social status, non-farm work and other indicators. The main reasons for selecting the above control variables are as follows: gender and marriage will affect the individual's risk tolerance, opportunity and job choice; On the one hand, the increase of age will bring about an increase in working experience, which will have an impact on income. On the other hand, the increase of age will also be accompanied by a decline in physical health, which will affect the income and living standard of individuals. In addition, Zhang W G(2020) pointed out that non-agricultural employment, social status and social interaction were one of the important reasons for Mandarin's influence on economic poverty, health poverty and spiritual poverty. Therefore, the three indicators such as non-agricultural employment, social status and social interaction were taken as individual control variables. The family size was selected as the control variable at the family level, and the dummy variable of the province where the individual lived was added at the region level to control the impact of potential and unobserved regional differences on the regression results.  Table 4 shows the results of descriptive statistics for all samples. From the perspective of multidimensional poverty, the average of multidimensional poverty in ethnic areas is 0.44, which indicates that the degree of multidimensional poverty in ethnic areas is in the middle level; from the perspective of language ability, the average of Mandarin proficiency is 2.89, which is in the middle and lower level of 1 to 7, which indicates that the level of Mandarin in ethnic areas still has a large room for improvement. From the perspective of individual characteristics, the proportion of women in the sample is slightly higher than that of men, the average age is about 44 years old, 61% of the individual marriage is married, the average family population is about 5, and the degree of social interaction is 3.78, indicating that the individual lacks good social interaction, and the average social position is 2.66, which is in the middle level.

Model Specification
In this subsection, we try to analyze the relationship between Mandarin proficiency and multidimensional poverty using the following panel regression model : represents the Multidimensional Poverty Index , represents the individual's Mandarin proficiency, refers to the control variables of individual, including gender, age, marital status, social interaction, social status, non-agricultural employment, family size, represents a locale-level control variable, is a random disturbance term. At the same time, in order to reduce the risk of missing variables caused by the differences in cultural traditions and customs among provinces, this paper also controlled the provincial dummy variables, in addition to controlling the year fixed effect.     Table 5 of this section, indicating that the regression results are reliable and robust. Year fixed effect Y Y Note: Robust standard errors are in parentheses. The *, **, and *** indicate statistics significance at 10%, 5%, and 1% levels.

Quantile Regression
The least square method is used to analyze the impact of Mandarin proficiency on multidimensional poverty in ethnic minority areas, but the estimated result only reflects the impact of Mandarin proficiency on the mean value of conditional distribution of multidimensional poverty. To accurately describe the impact of Mandarin proficiency on the variation range and conditional distribution of multidimensional poverty index, quantile regression method is used to explain the influence of Mandarin proficiency on different multidimensional poverty levels. As shown in Table 7, the general level has a significant negative impact on the multidimensional poverty level of individuals. With the increase of the Multidimensional Poverty Index, the effect of Mandarin proficiency on poverty reduction increases. Specifically, in the 20% quantile, Mandarin increased by one unit, and the multidimensional poverty index decreased by 2.41%. In the 50% quantile, the poverty reduction effect of Mandarin was 4.29%, and in the 90% quantile, the poverty reduction effect of Mandarin was 8.53%, which was significant at the 1% level. But in the 10% quantile, Mandarin proficiency has a negative impact on the Multidimensional Poverty Index, but the impact is not significant. Note: Robust standard errors are in parentheses. The *, **, and *** indicate statistics significance at 10%, 5%, and 1% levels.

Analysis of Subsamples by Gender and Age
In order to further study the influence of Mandarin proficiency on the multidimensional poverty index of individuals of different ages and genders, the samples were classified according to gender and age. The genders were divided into two categories: male and female. In the age reference, the labor force was divided into four categories: 18-28 years old, 29-50 years old, 50-65 years old and above 65 years old. The regression results are shown in Table 8. Mandarin proficiency has a significant negative impact on the multidimensional poverty level of individuals, but the influence degree is different for different genders and different age groups. Specifically, on the one hand, from the perspective of the whole sample, the effect of poverty reduction brought by the improvement of male Mandarin proficiency is greater than that of female. When male Mandarin proficiency increases by 1 unit, the multidimensional poverty index decreases by 6.87%, while the multidimensional poverty index of female only decreases by 4.12%, and both of them are significant at the level of 1%. On the other hand, in order to study the poverty reduction effect of Mandarin on different age groups in different genders, on the basis of gender division, the samples are divided into four categories according to age for regression analysis.

Analysis of Subsamples by Employment
As one of the seven key poverty alleviation measures, a large number of literature proves that the transfer of workers' employment across departments is beneficial to the increase of income level and plays a positive role in solving poverty. Therefore, this part analyzes the influence of Mandarin proficiency on multidimensional poverty of agricultural and non-agricultural employment. The empirical results are shown in Table 9.  Note: Robust standard errors are in parentheses. The *, **, and *** indicate statistics significance at 10%, 5%, and 1% levels.

Conclusion
To study the correlation between Mandarin proficiency and multidimensional poverty, we use the CFPS data