RELATIONS BETWEEN YOUTH EMPLOYMENT AND ELDERLY EMPLOYMENT IN CHINA

Transcription

1 RELATIONS BETWEEN YOUTH EMPLOYMENT AND ELDERLY EMPLOYMENT IN CHINA A Thesis submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Master of Public Policy in Public Policy By Meiyi Jiang, B.A. Washington, DC April 11, 2017

2 Copyright 2017 by Meiyi Jiang All Rights Reserved ii

3 RELATIONS BETWEEN YOUTH EMPLOYMENT AND ELDERLY EMPLOYMENT IN CHINA Meiyi Jiang, B.A. Thesis Advisor: Terry Kenndey, Ph.D. ABSTRACT China s rapidly aging population is burdening the nation s social security system. As a result, State Council is considering raising the retirement age. However, economists worry that raising the retirement age keep more elderly people in the labor force crowding out the youth employment. Using the data from China s 2000, 2005 and 2010 censuses of 83 provincial observations in basic OLS model, a fixed effect model and 2SLS model, this paper concludes that there is no crowd-out effect between elderly employment and youth employment. Rather, there may even be a positive relationship for female between these two factors. Thus, the State Council would be less concerned crowd-out effects if they want to raise the retirement age. iii

4 TABLE OF CONTENTS Introduction and Background...1 Literature Review...5 Benefits of Raising the Retirement Age...5 Impact of Elderly Labor Force Participation...6 Data and Conceptual Model...9 Data Description...9 Conceptual Model...12 Empirical Results...17 The Basic Model...17 The Fixed Effect Model SLS Model...18 Difference in Developed and Developing Provinces...20 Conclusion...23 Appendix...25 Bibliography...31 iv

5 List of Tables Table 1. Descriptive Statistics, by Province...25 Table 2. Developed Provinces vs. Developing Provinces...26 Table 3. Relations between Youth Employment and Elderly Employment, Male...28 Table 4. Relations between Youth Employment and Elderly Employment, Female...29 Table 5. First Stage Result of 2SLS Model...30 v

6 Introduction and Background In recent years, scholars and policy makers are paying increasing attention to rapid population aging in China. According to the Chinese National Bureau of Statistics, 222 million of the country s citizens were above 60 years old in 2015, 16.1 percent of the total population. And it is estimated that the over-60 population cohort will exceed 300 million in 2025, one fifth of the whole population. China s aging population raises several concerns. The first is about the sustainability of the nation s social security system. The current retirement age, determined by the National People s Congress in 1978, is 60 for males, and 55 for females. This implies that people usually work for 30 to 40 years before they are eligible for a pension. But with life expectancy increasing, people on average receive a larger pension without working longer or paying more tax, which would place huge burden on the pension system. Meanwhile, partly due to the past four decades of the one-child policy, the elderly dependence ratio actually increased from 9.2% in 1995 to 13.7% in In other words, in 2014, every hundred working age person supported 13.7 elderly people. The financial gap of the pension system the difference between the relevant tax revenue, and outlays used to pay pensions and government expenditures on programs serving the elderly is increasing. Second, the aging population in becoming a burden for ordinary families. The elderly get a basic pension of ($470.6) a month. But this is not enough to cover living expenses, especially in big cities and when considering elderly people s health status. Historically, the gap has been filled by their children. But the number of families 1

7 is increasing in China (families with 8 grandparents, 4 parents, a working married couple, and 1 child). In this case, each working couple may need to support up to 13 other people. The third concern is the slowdown of the Chinese economy. China s demographic dividend cheap and productive workers is diminishing. In the past three years, the labor force has lost 10 million people. At the same time, labor costs are increasing rapidly, which makes the manufacturing sector less profitable. According to Deloitte report (2009) average hourly wages in the manufacturing sector increased from $4.35 in 2004 to $12.47 in Policy makers are concerned about how to keep enough workers in the labor force to maintain future economic development. To buffer the negative impacts of population aging, Chinese economists and demographers have proposed raising the country s retirement age. This will not only increase the labor supply but also make the social security system more sustainable. People would work longer and, therefore, pay more tax, and postpone the years in which they would draw a pension. However, this possible solution has been widely criticized and strongly opposed. One criticism is that increasing the supply of elderly cohorts will lead to a reduction in demand for young or prime age workers. This is known as the lump of labor view, according to which, the number of jobs in the economy is fixed; if more jobs are taken by the elderly, fewer will be left for young and prime age workers. Also, the expansion of Chinese higher education from 1998 to 2005 has increased supply of university graduates in the labor market and many have had difficulty finding job. This problem may become even worse if new competitors (the elderly) remain in or return to the labor market. 2

8 However, this concern is disputed by some economists who argue that the number of jobs in the labor market is not fixed and the market will adapt to the changes in labor supply (Walker, 2007). Additionally, the elderly labor force may be a complement rather than a substitute for young and prime age workers. Currently, no empirical evidence shows that elderly people have a crowd-out effect on young and prime age workers. In fact, population aging is a pressing problem not only in China but also in many developed countries. Japan, the United States and most of the European countries either have already implemented or are considering social security system reform to mitigate the adverse effects of an aging population on pension programs and the labor market. At the same time, these countries also have concerns about the relationship between raising the retirement age and the employment of young and prime age persons. In order to examine this relationship, Gruber and Wise (2010) conducted a series of cross country analyses. These authors examined the relationship between the labor force participation of elderly persons and the labor force participation of young and prime age persons in twelve OECD countries. However, they found no evidence that increasing the labor force participation of older persons reduces the job opportunities of young and prime age persons. Moreover, they argued, greater labor force participation by older persons is usually associated with greater youth employment and with reduced youth unemployment. These results shed light on the ongoing discussion about retirement reforms in China. However, unlike OECD countries, China is still a developing country and its labor market is still evolving. The nature of the jobs and the skills it needs change quickly. For example, labor intensive industries that originally located in China are gradually moving to South East Asian countries or Africa (Deloitte, 2009). Low skilled workers in China 3

9 are finding it more difficult to get a job. In the meantime, companies complain that it is difficult to hire high skilled workers, even at considerable wages. If China wants to reform its social security system through raising the retirement age, empirical examination is required to determine whether its labor markets will, like the OECD countries, make the necessary adaptive changes. Using a basic OLS model, a fixed effect model and a 2SLS model, my empirical results show that there is no crowd-out effect for either genders, in both developing and developed provinces. Moreover, the increase of female elderly employment may have a positive effect on female youth employment. Thus, policy makers should be less concerned about raising the retirement age. 4

10 Literature Review Benefits of Raising the Retirement Age In China, as determined by National People s Congress in 1978, the current retirement age for men is 60, and for women is 50 (if they do physical work) or 55 (in other cases). If these retirement ages continue, in 2050, China will have 79 retired people for every 100 working age people, which would place immense pressure on China s retirement financing system (Jackson et al., 2009; Sin, 2005). Additionally, resulting from decades of one-child policy, the country faces a labor shortage. According to the World Bank, China s labor force declined to 929 million in 2013 and will decline to 847 million by Raising the retirement age can keep older persons in the labor force and re-utilize their skills. For example, raising retirement age for women from 50 to 60 would increase the proportion of the working-age population (persons above aged 15) by 10.9 percent by If the retirement age were to be further increased to 65 for both genders, the working-age population would increase by 18.4 percent by 2039 (Dorfman et al. 2012). However, according to the current projected retirement reform plan, public opposition makes it impossible to raise the retirement age to 65 all at once. Instead, the Chinese State Council plans first to raise the retirement age incrementally by a specific number of months based on the number of working years and then gradually increase it to 65. My paper will focus on the relationship between increases in the retirement age and changes in the labor market. 5

11 Research by Chen and Groenewold (2017) shows that raising the retirement age generates extra revenue for local governments which are assumed to be ultimately responsible for the payment of pensions. Thus, raising Chinese retirement ages would relieve local government s burden and fill part of the pension gap. However, Chen and Groenewold is the only study of the economic benefits of raising the retirement age, and their research area only includes two of China s provinces. Impact of Elderly Labor Force Participation The claim that raising the retirement age will crowd out youth employment is often based on two assumptions: first, that the labor forces in different age groups are substitutes; and second, that the total number of jobs in the economy is fixed. Therefore, if more jobs are taken by the elderly, fewer jobs will be left for the young (Hunt & Katz, 1998; Gruber & Milligan, 2008). Regarding the first assumption, while some empirical studies have found that elderly workers and younger workers are substitutes (Card & Lemieux, 2000), others have found that elderly workers and younger workers are complements (Hebbink, 1994). There is no conclusive empirical evidence to show the estimated degree of substitution between elderly workers and youth. One possibility is that the work that they do is similar. In that case, elderly workers and younger workers are highly substitutable. While as suggested by Freeman (1998), people with different skills are likely to be complements. Elderly workers probably have more specific skills than younger workers, because specific skills are usually acquired over an extended period of employment and closely attached to the firm in which workers are employed. 6

12 Moreover, in China, due to the expansion of higher education and the establishment of compulsory nine-year education, young people usually receive more education but limited skills training. Meanwhile, due to rapid progress in technology, the nature of jobs has also changed largely making skills acquired in firms even more specific. Thus, I argue that because of the large differences in specific skills and the gap in education level between young persons and old persons, the two groups are less likely to be homogeneous, and thus, less likely to be substitutes compared to similar groups in developed countries. The second assumption the lump-of-labor theory (Walker, 2007) is usually not supported by empirical analyses (Kalwij et al., 2009; Kapteyn, et al., 2004). The labor market is a dynamic market rather than a static one, which means it can adapt to large changes in labor supply, especially in the long term. For example, Hunt and Katz (1998) found that higher marginal labor costs will cause substitution away from labor and toward other inputs. Under this theory, the declining employment of the elderly will decrease with employment of younger persons together, since owners will transfer the money of hiring people into other uses. In addition, the National Bureau of Economic Research has sponsored several studies about OECD countries showing that greater labor force participation of older persons is associated with greater youth employment and with reduced youth unemployment (Gruber et al., 2010), which is contrary to the lump-oflabor theory. However, only two similar studies have been conducted on China (Munnell & Wu, 2012; Zhang & Zhao 2012). Both concluded that greater elderly labor force participation will not result in less youth employment. But they neglect the fact that when prospects in 7

13 the markets are dim, young persons may choose to continue to receive more education. Thus, actual youth unemployment may be higher than the measured one. The ability and willingness of elders to work is another concern, which would greatly affect elderly persons employment rates. Moreover, because a women s retirement age is determined by the type of work she does, the two studies set different age groups for elderly females. In Munnell and Wu s research, elderly women are defined as 45 to 59-year-olds, while in Zhang and Zhao s study, elderly women are 50 to 59 years old. But, in fact, many women do not retire at 55, but keep working into their 60s or 70s. Thus, the divisions of age may suffer from an endogeneity problem. That is, if people do not reach retirement age and keep working, but research include them, the estimation on raising retirement age may be biased. Finally, both studies use data before Things have changed a lot since then, and we can get 2010 census data now to reflect the recent relationship more accurately. 8

14 Data and Conceptual Model Data Description My analysis of the relationship between youth employment and elderly employment uses data from China s population census in 2000, 2005 and 2010 in 31 provinces (excluding Hong Kong, Macao, and Taiwan). The 2005 census is a 1% population sample survey. And the 2000 and 2010 census are a 10% population sample surveys. Other controls, including health status self-reported data, mortality rates, share of higher education, share of young population, share of urban population, and ethnic minority share of each province are from the same source. All economic data in my model i.e., GDP per capita of provinces are from the China Statistical Yearbook of the corresponding years. As stated in the previous sections, the current retirement age in China for men is 60, and for women it is 50 or 55 (depending on the nature of work). Based on retirement age, national division of age groups, and relevant research in OECD countries (Gruber & Milligan et al. 2008), I define youth as individuals aged 20 to 24, for both men and women, and old people as individuals aged 55 to 65 for men, and 50 to 65 for women. The following analysis is conducted for men and women separately because the age groups are different for each gender and the estimated effects for each gender may be different. The dependent variable in the model is youth employment to population ratios. They are calculated as the ratio of young people employed to the total population in each age group. Students are excluded and do not count as being employed. The independent 9

15 variable is elderly employment to population ratios which calculated as the ratio of old people employed to the total population in each age group. The control variables in the model are GDP per capita, health status data, mortality rate, share of higher education, share of urban population, share of minorites and share of young population. Health status data is self-reported data, which are divided into three categories, healthy, fair, and no. It is percentage data that is calculated as the number of people belonging to one health status category divided by the whole population in each age group. Mortality rate is calculated as the ratio of people who die to the population in each age group and is the average rate for the elderly of each gender. Share of higher education is the ratio of people who receive a bachelor degree or above to the whole population in each age group. Share of urban population is people who live in the urban area to the whole population in each province. Share of young population is the ratio of people who are aged 20 to 24 to the whole population in each province. Ethnic minorities are the ratio of people who are not Hans in each province to the whole population in provinces. Generally, Hans consists of 92% of the whole population and there are total 56 ethnic groups in China spreading all over the country. Table 1 shows descriptive statistics for 83 province level observations in the 2000, 2005 and 2010 censuses. On average, the youth employment rate is 0.77 for men, and 0.69 for women. Elder employment rates are 0.65 for men and 0.49 for women. The employment rate is higher for men than women in both age groups. GDP varies greatly among different provinces and there is huge gap between the maximum ( ) and minimum ( ). The gap is around 71,000 (which equals about $10000), with a standard deviation of Also, according to their self-report, generally women s 10

16 health status is worse than men s. But women s mortality rate is lower than men. It seems women live longer than men but report slightly worse self-reported health. The health condition and mortality rate may affect the elderlys ability to work. Moreover, as Table 1 shows, men receive more higher education than women. Education levels vary greatly among different provinces and over time. The minimum rate for male who receive higher education is only 0.6% while in some provinces there are over 20% males who receive higher education. Similarly, only 0.3% female receive higher education in Tibet in 2000 while there are nearly 20% female going to college in Beijing. The share of urban population is around one third for both genders and is highly related to the economic development level. Also, the minority share is quite different in each province. In some provinces, over 90% people are not Hans, while in other provinces, there may be only 0.3% people are minorities. Provinces having more minorities are often provinces in the western part of China, like Tibet, Gansu, Yunnan, and Xinjiang, whose GDP is not high. Also, there may be some conflicts between different ethnic groups in these areas. This huge difference in the share of minority groups might cause different policies in different provinces to encourage employment and mitigate any adverse relationship between different ethnic groups. 11

17 Conceptual Model In order to examine the relationship between young persons employment and elderly persons employment, I estimate the following basic OLS model: Youngemp = β + + β - Olderemp + β 3 GDPper + X 8 + ε 8 (1) where youngemp is the employment rate of younger persons, olderemp is the employment rate of older persons. Both the employment rate of young persons and older persons are separated by gender. GDP per capita is used to control the economic development of provinces, which is believed to have a large effect on employment rates, X j represents all other control variables in the model, like the share of young people, the share of urban population, the share of minority and the share of those with higher education in each province. It is acknowledged that it is difficult to control other variables affecting employment of persons in different age groups because some factors are not observable, or hard to measure, or the data have not been collected. Some provinces, for example, those that have a large minority population, may have their own employment encouragement policies aimed at different age groups. With these policies working at the same time, other changes are difficult to differentiate and measure their effects separately. Also, even if provinces have the same policy, some provincial governments may be more effective than others, thus leading to different impacts. Similarly, as time goes by, there might be an influence of aggregate (time series) trends in different province which also will create bias. 12

18 In order to eliminate potential bias, I construct data in panel form at the province level using 2000, 2005 and 2010 data, and then employ a fixed effects estimation strategy. A fixed effect model can help to control for unobserved factors and remove time invariant factors. Also, I need to add year dummy variables to control for the influence of time.thus, through a fixed effect model and year dummy variables, I can remove the unobserved provincial differences over time, like effectiveness of government and policies, and get an unbiased result. The estimated equation is: Youngemp = β + + β - Olderemp + β 3 GDP + X 8 + p 8 + γt + ϵj (2) where pj represents the province fixed effects, and γt are time dummy variables. Year 2000 will be the base line of other year dummy variables. Estimation using these two kind of fixed effects variables allow us to control time-related and province-related factors. Even if I can control for obvious variables which affect youth employment, it is reasonable to suspect that there are some factors omitted that may influence both youth and elderly employment. At the same time, there are other factors, like health status or income, that influence older persons ability or willingness to work, and therefore, influence the employment rate of elderly persons. Ignoring these factors may raise omitted variable bias problems and lead to bias. In order to solve this bias, I take an instrumental variable approach. A successful instrumental variable should be correlated to older people s employment, but have no direct relationship to youth employment except through elderly s employment. I will use this instrumental variable to predict the employment rate of the elderly persons. The prediction of elderly persons employment 13

19 rate will then be used in a new regression to estimate the relation between youth employment and elderly employment. These activities can be achieved by a 2SLS (Two Stage Least Squares) model with good instrumental variables. Health status is a good candidate for an instrumental variable. Studies show that health status is strongly correlated with labor supply. (Mcgarry, 2004; Smith 2006; Haider & Stephens 2007). It is acknowledged that poor health will result in lower productivity. On the one hand, employers do not want to hire low productivity people. On the other hand, health status limits the elderly s ability to go to work. People with bad health may choose to leave jobs early while people in good health may continue to work until their body does not allow them to continue. This may be expected to particularly influence the elderly and have negligosble impact on the youth. Since 2005, the census had begun to record the elderly self-reported health status, and divided them into three categories healthy, fair, and no. I will use these data to make a prediction of elderly employment rates. Moreover, research (Stevens, Miller, & Page, 2011) suggests that own-age group mortality rates of the elderly are negatively associated with employment rates because it shows the most extreme situation of health deterioration. It can give us a general idea that how many people in this age group that still have the ability to work. Therefore, it will be a good indicator to infer the health status of elder persons. Meanwhile, research also reports that the mortality rate of other age groups are independent of employment rate. In other words, mortality rate of other age groups should not be related to the youth employment rate. 14

20 In sum, self-reported health status and mortality rates meet two criteria when selecting instrument variables that the instrumental variable should be related to the selected independent variable, while having no connection with the dependent variables except through its impact on the selected independent variable. Thus, these two are good candidates as instrument variables to develop a 2SLS model. I specify first stage and second stage as: Olderemp = α + + α - mortality + α 3 healthsts + α H GDP + X 8 + p 8 + γt + ϵj (3) Youngemp = β + + β - Older emp + β 3 GDP + X 8 + p 8 + γt + ϵj (4) In addition to controlling for GDP, and time and province differences, we want to know the influence of different development levels on youth employment. Because there is a large amount of worker migration in China, as people are more willing to go to big cities to work, especially when they are young, the influence on youth employment will be different in developed cities and other cities. Thus, according to the division in 2010 China Statistical Yearbook, I split the provinces into developed provinces and developing provinces to see the different influence. I define developed provinces as 10 provinces whose GDP per capita was the largest among China s 31 provinces in 2010 (average GDP per capita is ). Most of the developed provinces are coastal provinces, cities directly-controlled by the central government and provinces with abundant mineral resources. Other provinces will be treated as developing provinces, whose average GDP per capita is The average youth employment of developing and developed provinces is similar for both 15

21 genders while the elderly employment rate in developing provinces are higher than in developed provinces for both genders. When comparing other control variables, there are much more people in developed provinces receiving higher education. The average share of higher education in developed provinces is 76 men per thousand and 69 women per thousand. While in developing areas, it is only 36 men per thousand and 29 women per thousand. Also, developed provinces have higher share of young population, higher share of urban population and lower share of minority population. There is not much differences between females and males in the control variables. 16

22 Empirical Results The Basic Model Table 3 shows estimation results for males, and Table 4 shows the results for females. As I discussed above, the retirement age is different for each gender, and I defined old people as individuals aged 55 to 65 for men, and 50 to 65 for women. For this reason, the different gender s effect of elderly employment on youth employment is likely to be different. Thus, I need different models for each gender. The basic model is predicted in column (2) in Table 3 and Table 4. We cannot reject that there is no crowd-out effect of elderly employment on youth employment of both genders as the coefficient on the elderly employment rate is insignificant. Except GDP per capita and the share of urban population, all other variables are statistically insignificant. One thousand yuan increase in the GDP per capita is associated with decrease in male youth employment and decrease in female youth employment. And one percentage point increase in share of urban population is associated with decrease in male youth employment. The Fixed Effect Model When I use a fixed-effect model with province and year dummy variables to control the influence of years and time invariant factors in provinces (the third column in Table 3 and Table 4), the results indicate there is still no crowd-out effect for both genders. The coefficients on year dummy variables are highly significant, but surprisingly, years have a negative effect on youth employment, and as time goes by, this effect becomes more 17

23 negative. One plausible explanation of this finding is that, as times goes by and economy has developed, more young people choose to continue to study or become freelance, while in earlier years, more young people needed to work to support their families even though they did not finish junior school or high school. Also, when calculating employment status, freelance sometimes is considered as unemployed. Thus, the current employment rate is lower than previous ones and has had a tendency to become even lower in recent years. Moreover, in the fixed effect model, higher education has a significant positive effect on youth employment. It seems reasonable that the more education you receive, the easier it will be to find a job and also to find a better one. Also, when one province has more people who receive higher education, it is more competitive in attracting business. More businesses allow more people, especially the young, to find jobs there, which increase the employment rate. But other variables are not statistically significant. It seems that, when we control for provinces and years, the GDP per capita does not play an important role any more. That may be because we control for differences across provinces and the biggest differences across provinces is their economic conditions. For similar reasons, the share of urban also becomes insignificant. 2SLS Model To address endogeneity problem resulting from potential factors that influence elderly and youth employment, I develop a 2SLS model. The results are presented in 18

24 column (3) in Table 3 and Table 4. I find that when using mortality rate as aninstrumental variable, the results indicate no crowd out effect on youth employment for both genders. For my 2SLS model, I use three categories of self-reported health status and the mortality rate of each age group as instrumental variables. The mortality rate alone is highly significant and has a negative effect. However, though the literature review shows that health status and mortality rate could be useful factors that influence the ability of the elderly to work, but the first stage of my 2SLS model indicates that my overall instruments relevance is poor. None of the self-reported health status variables are statistically significant enough and the overall F-statistic is 5.31 and 4.3 for each gender. (Generally, only when F-statistic is over 10 do we think the instrumental variables are good.) Thus, using all variables as instrument variables may not be a good idea. That may be because the weakness of the self-reported data. Different people have different ideas about the health status and there is no standard definition. Even if their health indicators show they are in good health, people may not think they are in good condition. Thus, I just use mortality rate as my instrument variable in the second stage to predict the elderly employment. As Table 3 and Table 4 show, the results are similar to those of my fixed effect model: no relationship between youth employment and elderly employment for men, and a positive relationship for women. And the degree of female elderly employment effect is much larger than previous models. A percentage point increase in female elderly employment is associated with a 0.8 percentage point increase in female youth employment, which is more than twice that of previous models. This may indicate that our previous estimations are biased. 19

25 In this model, despite the coefficient being small, GDP per capita becomes significant again. A thousand yuan increase in GDP per capita is associated with a percentage point increase in female youth employment. The share of minority population also has positive effect on the youth employment of both genders. Year effects also play important roles and the effects are negative. Two more things are interesting in the 2SLS model. In male 2SLS model, a percentage increase in the share of urban population is associated with 0.44 percentage points decrease of male youth employment. That might be because labor market in big cities are more open to the idea that women go to work. Thus, young males face more competitors in the big cities. And women have more opportunies in the urban area. Thus, urban population does not affect, or even may increase, female youth employment. In female 2SLS model, one percentage point increase in the share of young population is associated with 2.74 percentage points increase in female youth employment while I do not observe similar trend in male model. One interpretation here is that if more young people migrate to one city to find job, and therefore the province has higher share of young population, the idea that women can go to work is more acceptable to the whole society. Differences in Developed and Developing Provinces In order to see the influences of different economic development levels on youth employment, I separate provinces into developing and developed to compare these two groups. As I stated before, developed provinces are the top 10 provinces by GDP per capita among China s 31 provinces in 2010 and other provinces are defined as 20

26 developing provinces. The results are presented in columns (4) and (5) in Table 3 and Table 4. In developed provinces, there is no crowd-out effect of elderly employment on youth employment when we examine each gender. Instead, there is a positive effect between female elderly and female youth employment. A percentage point increase in female elderly employment is associated with a 0.4 percentage point increase in female youth employment. In developing provinces, similarly, the relationship between male elderly employment and male youth employment is statistically insignificant, while the influence of female elderly employment is highly statistically significant on female youth employment. A percentage point increase in female elderly employment is associated with a percentage point increase in female youth employment, which is smaller than in developed provinces. It may be because that in big cities, living expenses are higher, so more young women need to go to work to support the family. Also, big cities have more opportunity for young women to find a job. Year still has a significant effect on the youth employment rate in developing provinces. And the effect is still negative, like we observed in the basic model. Furthermore, in developing provinces, the minority share of the population plays an important role in youth employment of both genders. It is possible that in developing provinces, which have higher shares of minorities, government may propose more generous policies to encourage employment to relieve the economic conflicts between Han and other ethnic minority groups. Also, provinces with a higher share of minority usually turn this uniqueness into a tourism resource because the minority concentrated areas in China are often remote places with beautiful scenery. People can experience 21

27 different cultures and customs there. Thus, more local people would be hired to support the development of the tourism industry which promotes the employment rate. For example, according to National Bureau of Statistics of China, there are more than 25 minority groups living in Yun nan province. In 2015, the tourism industry takes up 45.14% of its GDP and around 15% of its work force are related to trousim industry. However, in big cities and more developed regions, it is a different story. The share of minority has no relationship to male youth employment while there is negative effect of the share of minority on female youth employment. A percentage point increase in the share of minority for female is associated with a 0.6 percentage point decrease in female youth employment. One interpretation here is that in developed provinces, especially provinces with rich resource and with relatively higher share of minority, like the Inner Mongolia province, welfare and ethnic minority policies are generous. So, more women may tend to stay home to take care of the whole family. When comparing differences between males and females, I observe that the effect of female elderly employment rate on female youth employment rate seems to be more significant and larger than the effect on males. My interpretation here is that more working female elderly demonstrates that the society is more accptable to working women. Thus, if one place has more working female elderly, young ladies will find it eaiser to work, resulting in larger coefficients than males, who are always regarded as the pillar of their families and keep working regardless of time and economic changes. 22

28 Conclusion China s latest census (2015) shows that the elderly dependence ratio is 14.3%, and people above 60 constitute 16.1% of the population. The Chinese government is having to put more resources into the social security system every year, while in coming years, due to decades of one child policy, the country will lack labor. Policy makers are worried about how to keep enough workers in the market. China s current retirement age is 60 for males, and 55 for female, which is lower than most developed countries. To ease financial pressure on the nation s social security system and maintain or even increase the labor supply, the option of raising the retirement age is an attractive approach for the Chinese government. Though there is a concern about elderly employment crowding out youth employment, my empirical results show that there is no crowd-out effect for either genders, in both developing and developed provinces. Moreover, for females, the increase of elderly employment may have a positive effect on the female youth employment. Thus, policy makers should be less concerned about raising retirement age, and do more empirical researches on the economic benefits such a policy may bring. However, there is one concern regarding the measurement of youth employment. When calculating employment, people who are in school are excluded. College enrollment has increased rapidly. Also, after graduating from college, more and more students choose to pursue masters and doctoral degrees, especially when there is an economic slowdown and few job opportunities. It is uncertain if this desire to receive more education will affect my regression results. That is, because people choose to 23

29 continue to receive education when they cannot find jobs, the youth unemployment rate might be under estimated. Another limitation of my study is the sample size. Because I only can get the provincial census data of 2000, 2005 and 2010, the number of provincial observations is only 83. Due to small smaple size, there may be Type II error and therefore, the statistical power, i.e. the ability to detect a particular effect may be low. Also, some small factors and influence may be neglected. Therefore, I need to be careful in treating my results. Also, I had difficulty finding effective wage variables for my regression. This may neglect the influence of elderly employment on the income of the youth. In many provinces, if a company hires a certain amount of young people, the government will give them tax credits. So companies try to hire more young workers to get the credits, but at the same time, they also hire skilled staff, including the elderly who have already acquired work experience and skills for many years, to keep the company working well. Thus, compainies may lower the wage for young workers to lower their overall cost. Therefore, even if there is no crowd-out effect for the employment rate, raising retirement age may affect the young finacially. 24

30 Appendix Table 1. Descriptive Statistics, by Province Mean Min Max Standard deviation Dependent Variable: Youth Employment Male Female Independent Variable Elder Employment Male Female GDP per capita ( ) Share of Higher Education Male Female Share of Young Population Male Female Share of Urban Population Male Female Share of Minority Population Male Female Male: Health Status Healthy Fair No Female: Health Status Healthy Fair No Mortality Rate Male Female Source: China s population census in 2000, 2005 and 2010 Notes: Dependent Variable:Youth as individuals aged 20 to 24, for both men and women; Independent Variable: Old people as individuals aged 55 to 65 for men, and 50 to 65 for women. Employment to population ratios are calculated as the ratio of people employed to the total population in each age group. Minority are population who are not Hans, taking 8% of Chinese population. 25

31 Table 2. Developed Provinces vs. Developing Provinces Youth Employment Rate Elderly Employment Rate Share of Higher Education GDP per capita( ) Male Female Male Female Male Female Developed Province Shanghai Beijing Tianjin Jiangsu Zhejiang Inner Mongolia Guangdong Liaoning Shandong Fujian Mean Developing Province Jilin Hebei Hubei Chongqin Shanxi( 陕西 )* Heilongjiang Ningxia Shanxi( 山西 )* Xinjiang Henan Hunan Qinghai Hainan Sichuang Jiangxi Guangxi Anhui Tibet Gansu Yun nan Guizhou Mean

32 Table 2. (cont.) Share of Yong Population Share of Urban Population Share of Minority Population Male Female Male Female Male Female Developed Province Shanghai Beijing Tianjin Jiangsu Zhejiang Inner Mongolia Guangdong Liaoning Shandong Fujian Mean Developing Province Jilin Hebei Hubei Chongqin Shanxi( 陕西 )* Heilongjiang Ningxia Shanxi( 山西 )* Xinjiang Henan Hunan Qinghai Hainan Sichuang Jiangxi Guangxi Anhui Tibet Gansu Yun nan Guizhou Mean * In Chinese, there are diffirent characters for shan and xi. Thus, despite their translation is the same, they are different provinces. In order to avoid confusion, I indicate their Chinese writing. Source: China s population census in 2000, 2005 and 2010 Notes: Dependent Variable:Youth as individuals aged 20 to 24, for both men and women; Independent Variable: Old people as individuals aged 55 to 65 for men, and 50 to 65 for women. Employment to population ratios are calculated as the ratio of people employed to the total population in each age group. Minority are population who are not Hans, taking 8% of Chinese population. 27

33 Table 3. Relations between Youth Employment and Elderly Employment, Male Elderly Employment GDP per capita Share of Higher Education Share of Young Share of Minority Urban (1) Basic Model (0.19) * (0.001) (0.539) (0.598) (0.047) ** (0.155) Year Year Fixed Effects Control over Provinces (2) Fixed Effect Model (0.208) (0.0011) 1.934*** (0.706) (0.549) (0.599) (0.199) *** (0.018) *** (0.024) (3) 2SLS (0.307) 0.003** (0.0013) (0.38) (0.872) 0.121*** (0.042) -0.44** (0.211) ** (0.023) *** (0.023) (4) Youth Employment in Developed Provinces (0.296) (0.002) (0.5) (1.116) (0.211) (0.201) (0.037) * (0.075) (5) Youth Employment in Developing Provinces (0.15) (0.0028) * (1.562) (0.702) 0.141*** (0.038) -0.34** (0.15) (0.028) *** (0.047) No Yes No No No R-squared Source: China s population census in 2000, 2005 and Notes: significant level *** p<0.01, ** p<0.05, * p<0.1 Dependent Variable:Youth as individuals aged 20 to 24, for both men and women; Independent Variable: Old people as individuals aged 55 to 65 for men, and 50 to 65 for women. Employment to population ratios are calculated as the ratio of people employed to the total population in each age group. Minority are population who are not Hans, taking 8% of Chinese population. 28

34 Table 4. Relations between Youth Employment and Elderly Employment, Female Elderly Employment GDP per capita Share of Higher Education Share of Young Share of Minority Urban (1) Basic Model (0.13) ** (0.001) (0.701) (0.745) (0.049) (0.153) Year Year Fixed Effects Control over Provinces (2) Fixed Effect Model (0.147) (0.001) 1.543*** (0.568) * (0.536) (0.701) (0.24) *** (0.015) *** (0.024) (3) 2SLS 0.794*** (0.266) 0.003** (0.002) (0.443) 2.738** (1.042) 0.096** (0.038) (0.231) ** (0.025) *** (0.032) (4) Youth Employment in Developed Provinces 0.401** (0.172) (0.0026) (0.579) 1.997** (0.941) ** (0.237) (0.168) (0.044) (0.094) (5) Youth Employment in Developing Provinces 0.375*** (0.092) (0.0032) (2.009) (0.703) 0.143*** (0.032) (0.133) ** (0.025) ** (0.045) No Yes No No No R-squared Source: China s population census in 2000, 2005 and Notes: significant level *** p<0.01, ** p<0.05, * p<0.1 Dependent Variable:Youth as individuals aged 20 to 24, for both men and women; Independent Variable: Old people as individuals aged 55 to 65 for men, and 50 to 65 for women. Employment to population ratios are calculated as the ratio of people employed to the total population in each age group. Minority are population who are not Hans, taking 8% of Chinese population. 29

35 Table 5. First Stage Result of 2SLS Model Male Elderly Employment Female Elderly Employment *** *** Mortality (3.38) (7.67) R-squared Source: China s population census in 2000, 2005 and Notes: significant level *** p<0.01, ** p<0.05, * p<0.1 Employment to population ratios are calculated as the ratio of people employed to the total population in each age group. 30

36 BIBLIOGRAPY Card, David and Thomas Lemieux, (2000), Can Falling Supply Explain the Rising Returns to College for Younger Men? A Cohort-based Analysis, NBER working paper Chen, A., & Groenewold, N. (2017). An increase in the retirement age in China: the regional economic effects. Applied Economics, 49(7), Dorfman P. W., Javidan M., Hanges P., Dastmalchian A., House R. (2012). GLOBE: A twenty year journey into the intriguing world of culture and leadership. Journal of World Business, 47, Freeman, R., (1998), Working-sharing to Full Employment: Serious Option or Populist Fallacy? In: Freeman, R., Gottschalk, P. (Eds.), Generating Jobs: How to Increase Demand for Less-Skilled Workers, Chapter 6. Russell Sage Foundation Press, NY, pp Gruber, J., & Milligan, K. (2008). Do Elderly Workers Substitute for Younger Workers in the United States? Social Security Programs and Retirement around the World: The Relationship to Youth Employment, (February), Retrieved from Gruber, J., & Wise, D. A. (2010). Social Security Programs and Retirement Around the World: Micro Estimation. NBER Working Paper No Hebbink G.E., (1994), Production Factor Substitution and Employment by Age Group, Economic Modeling, Vol. 10, No. 3, pp Hunt, Jennifer and Lawrence F.Katz, (1998), Hours Reductions as Work-Sharing, Brookings Paper on Economic Activity, No.1, pp Jackson, Richard, Keisuke Nakashima, and Neil Howe. (2009). China s Long March to Retirement Reform: the Graying of the Middle Kingdom Revisited, Center for Strategic and International Studies, Washington, DC. Kalwij, Adriaan, Arie Kapteyn and De Vos, Klaas, (2009), Early Retirement and Employment of the Young, Discussion Paper 03. Kapteyn, Arie, Adriaan Kalwij and Asghar Zaidi, (2004), The Myth of Work sharing, Labor Economics, No.11, pp