Understanding the Health Capacity to Work among Older Persons in Rural and Urban Areas in the People’s Republic of China

I. Introduction

The People’s Republic of China (PRC) has been aging at one of the most rapid paces in the world. The share of older persons (60 years and above) increased from 8.6% in 1990 to 18.7% in 2020 (Figure 1). Their growing population resulted from the combination of a decline in fertility and an increase in life expectancy, which reached 77.9 years in 2020 (Figure 2). Greater life expectancy reflects an overall improvement in the population’s health.

While people have become healthier, they work less than in the past (Figure 3). The employment rate at every single age decreased between 2000 and 2015. This raises questions about whether people leave the labor market even when their health allows them to continue working. The answer to this question has important policy implications. The aging population has brought challenges to the financial sustainability of the social security system, and policymakers have signaled an intention to raise the retirement age. A deeper understanding of the work capacity of older persons can inform the design of reforms that can release their untapped work potential.

This paper applies two widely used methods—the Milligan–Wise (MW) method (Milligan and Wise 2015) and the Cutler–Meara–Richards-Shubik (CMR) method (Cutler, Meara, and Richards-Shubik 2013)—to study the potential work capacity of older persons. We define older persons as those who are above 60 years old. This paper aims to address the following questions: (i) How does health status affect employment decisions? (ii) What is the “tapped” and “untapped” work capacity among older persons? (iii) Is there heterogeneity in untapped work capacity across different groups—in particular, are there urban–rural and sex differences?

Health has long been understood to be a main determinant of retirement in the PRC, particularly in rural areas (Benjamin, Brandt, and Fan 2003). Yet the relationship between health and work among older persons is understudied. Some studies focus on the effect of health on work and find a positive effect (see, for example, Zhang et al. 2013 and Mitra et al. 2020). Mitra et al. (2020) reveal that negative changes in self-reported health, depression, and hypertension all reduce the probability of working. However, different studies find that health has little effect on the labor supply of older persons in rural areas (see, for example, Tan et al. 2022). Other scholars emphasize the effect of work on health. Retirement can improve health by releasing stress from work, but it may hamper health by reducing social interactions. Che and Li (2018) provide evidence that retirement improves the self-reported health of older persons.

Among all health measurements, self-reported health is the most studied. Other health indicators are examined to a lesser extent, mainly due to data limitations. Recently, depression has attracted more attention since the prevalence of depression has been found to be quite high in the PRC (Lu et al. 2014). The China Health and Retirement Longitudinal Study (CHARLS) conducted in 2011 reported that 30% of men and 43% of women have depression symptoms. Depression may result from work pressure (Sun et al. 2019), but how employment itself affects depression remains unclear. Not working has been found to increase the probability of depression for women aged 30–60 years (Shi et al. 2014), while retirement with a pension reduces depressive symptoms for older men (Fernández-Niño et al. 2018). Thus, existing evidence on health and work is mixed and requires further study.

The retirement decision is also affected by the social security system (Mitra et al. 2020). In the PRC, retirement patterns vary between rural and urban areas, which could lead to differences in additional work capacity. While half of urban residents retire before age 60, 80% of rural residents remain working at age 65 (Appendix 1).¹ One possible reason is that they are covered by different pension programs with different benefit levels and work incentives. There are two main pension schemes in the PRC: the Employee Basic Pension (EBP) and the Resident Basic Pension (RBP).² The former is designed for those who are formally employed, and the latter covers everyone else. The EBP program requires individuals to retire when they reach mandatory retirement age—which is 60 for all men, 55 for white-collar women, and 50 for blue-collar women—and provides significant pension benefits to retirees (typically at least 60% of their final wage) (Organisation for Economic Co-operation and Development 2021). The RBP program, developed from the New Rural Social Pension, was introduced between 2009 and 2012, and the Urban Resident Pension was introduced in 2011. The RBP allows people to claim pension benefits beginning at age 60 regardless of work status. However, the benefits are relatively low and generally insufficient for older persons to live on (typically 12% of average annual income per capita).³ Table 1 shows the pension coverage of older persons by program and for rural versus urban locations. Half of urban residents qualify for the EBP, while around 80% of rural residents are covered by the RBP. It is thus not surprising that the two groups have different retirement behaviors, with older persons in urban areas retiring much earlier than those residing in rural areas. The paper explores further the differences between urban and rural residents, and the extent to which such variation results from differences in the two pension programs.

There are two related papers that also explore the health capacity to work in the PRC. Hou et al. (2021) apply the CMR model and use rural residents as the reference group to estimate the additional health capacity to work for urban residents. Zhan et al. (2022) use both the MW and CMR methods. Our study extends their study in a few aspects. First, we examine urban–rural differences and investigate potential explanations, particularly how pensions and types of work influence older persons’ retirement decisions. Second, we examine different reference groups and methods to discuss the sensitivity of the results to these choices. In particular, we look closely at urban women who normally leave the labor market much earlier than other groups. For this group, choosing a proper reference group leads to more sensible estimates. Third, we use more recent data from the 2018 wave of CHARLS and examine the changes between 2015 and 2018 in untapped work capacity.

The rest of this paper is organized as follows. Sections II and III present the results from applying the MW model and the CMR model, respectively. Section IV offers the results of additional analysis, including a discussion on the urban–rural distinction and changes in additional work capacity between 2015 and 2018. Finally, section V concludes.

II. Health Capacity to Work among Older Persons Using the Milligan–Wise Method

In this section, we apply the MW method to predict health capacity. The MW method assumes that people today can work as much as in the past if they have the same health status as measured by the mortality rate. We can then simulate today’s expected work capacity using the employment–mortality relationship in the past and compare this to actual work decisions today. The mortality and employment rate by age and sex are generated from census data. In the PRC, a census is conducted every 10 years, and a 1% population mini census is conducted 5 years after the last full census. To predict work capacity in 2015, 2000 is used as the reference year.⁴ Census data from other years (e.g., 1995, 2010) are also examined to explore how estimated work capacity varies when different reference years are used. The census data do not provide the employment rate by age directly. We compute it by dividing the number of employed persons by the total population. Figure 4 plots mortality rates and the proportion of those with poor self-assessed health since the 1990s for those aged 45–75 years. Self-assessed health is generated from two nationally representative surveys—the China Health and Nutrition Survey (CHNS) and the CHARLS.⁵ Summary statistics for CHNS and CHARLS survey data are provided in Appendix 2.

As seen from Figure 4, the fraction of the population with poor health in the 1990s (using CHNS data) follows a similar trend with respect to age as mortality rates in 1995 and 2000. The trend for the fraction with poor health with respect to age in the 2010s (using CHARLS data) is also comparable with mortality rates in 2010 and 2015. This provides some support for the mortality rate being a good measure of health.

Figure 5 shows employment rates for different age-specific mortality rates in 2000 and 2015. The mortality rate at every age decreased between 2000 and 2015, indicating the improved health of the population. The gap between the two lines shows that people in 2015 worked less than in 2000 for a given mortality rate. The employment–mortality relationship in 2000 was used to predict employment in 2015. For example, the employment and mortality rates of 55-year-olds are 0.6240 and 0.0042, respectively, in 2015. The mortality rate (0.0042) is the same as for 49-year-olds in 2000, who had an employment rate of 0.8404. Assuming that people with the same mortality rate (0.0042) have the same employment rate, 55-year-olds in 2015 would have the health capacity to increase their employment rate by 21.6 percentage points (i.e., 0.8404–$0.6240=0.2160$). We follow the same procedure for every single age in 2015 (Table 2). At age 50, the additional employment capacity is 14.3%, and it increases with age, peaking at 30.7% at age 61 before falling to 25.7% at age 64. Overall, people aged 50–64 years in 2015 could work 3.3 years more, given the 2000 employment–mortality relationship. Considering that the average years of employment remaining for those between the ages of 50 and 64 is 8.6 years, this indicates that employment for this age group can increase by 37.4%.

The additional work capacity varies somewhat when using different reference years. When the reference year is 1995, the additional work capacity increases to 4 years (from 3.3 years when using 2000 as the reference year). As expected, additional work capacity decreases as the reference year moves closer to 2015. Using 2010 as the reference year, the additional work capacity is only 2.1 years. The findings also differ by sex. Using 2000 as the reference year, men and women can work 2.7 years and 4.2 years more, respectively (Figure 6). This shows that women have more additional health capacity to work.

Next, we examine heterogeneity between urban and rural residents. Before estimating the additional health capacity to work, we first describe the employment patterns. Appendix 3 shows the employment rate by rural versus urban location and by sex, based on CHARLS 2018 data. While the sex difference is evident, the urban and rural difference is remarkable. In general, older persons in rural areas have a much higher employment rate than their urban counterparts. The urban–rural distinction is relatively small before age 50 when both rural males and urban males have employment rates above 90%. The urban–rural employment gap widens by age 55 and becomes much larger after age 60. At age 64, 79.2% of rural men are working compared to only 31.7% of urban men. The difference in employment rates between urban and rural women starts at an even earlier age. Though the employment rate of rural women declines gradually with age, the majority still work at age 65. In contrast, urban women start to leave the labor force before age 50. The employment rate drops quickly from ages 50 to 60, and less than 30% remain in the labor market after age 60.

The employment pattern by age based on the census shows that the urban–rural difference persists over time (Appendix 4). In both 2000 and 2015, older persons living in rural areas were working more than older persons in urban areas. Although overall employment rates are lower in 2015 than in 2000, the changes differ between urban and rural areas. The employment rates for rural women and men decreased at every age over this 15-year period. In contrast, the employment rate for older persons in urban areas actually increased slightly, especially for urban women, who started to leave the labor market at age 45 in 2000, while their departure from the labor force was delayed until around age 50 in 2015. One possible explanation for the increase in the urban employment rate between 2000 and 2015 is greater rural-to-urban migration since most migrants will only stay in urban areas if they are working.

We further apply the MW method to explore the health capacity to work by rural versus urban location and by sex (Figure 7). The gap in the employment rate for a given mortality level is found for all four groups. The gap for urban residents widens earlier than for rural residents. For urban men, the employment–mortality gap increases before age 60, while for urban women, the gap starts to increase at around age 50. For older persons in rural areas, the gap becomes larger after age 60. Figure 8 plots the tapped and untapped work capacity by subgroup. While the current employment rate limits the possible additional work capacity, the employment pattern in the past determines how much people will work given their current health status. Of the four groups, rural men have the least untapped work capacity (1.7 years). On average they work 12.3 years over a 15-year period (ages 50–64), thus they have less potential to increase their employment rate. The next smallest untapped work capacity is for urban men at 2.5 years. Women have relatively more untapped health capacity to work: 3.0 years for urban women and 3.1 years for rural women. Although urban women worked less than their rural counterparts in 2015, their additional health capacity to work is still slightly less than for rural women, mainly because urban women with similar health did not work as much in the past either. Rural women have the most untapped work capacity. Currently, they work 9.5 years on average between the ages of 50 and 64. If they worked as much as in 2000, and given the same mortality level, they would work 32.7% more.

III. Health Capacity to Work among Older Persons Using the Cutler–Meara–Richards-Shubik Method

In this section, we apply the CMR method to quantify additional work capacity. This method (i) first explores the relationship between employment status and health status for a younger reference group using regression analyses, and (ii) then uses the estimated regression coefficients and actual characteristics of older persons to predict the employment capacity of older persons. By subtracting the actual employment rate (tapped work capacity) from the predicted employment (potential work capacity), we obtain the untapped work capacity.

We utilize data from the CHARLS beginning in 2011, with additional waves conducted in 2013, 2015, and 2018.⁶ We use data from the most recent survey in 2018. The group of interest is older persons aged 60–79 years ($N=8,080$). To implement the CMR method, we choose the reference group to be those aged 50–59 years ($N=3,832$), which is before people can claim pension payouts in most pension schemes. We also test how the results change using other reference groups with different age restrictions.

The main analysis and simulation exercises are based on the linear regression model estimated using ordinary least squares (OLS). In addition, we employ the logistic regression model as a robustness check. We postulate the linear regression model as follows:

where the dependent variable, Employed, is an indicator of employment status that takes the value of 1 if employed, and 0 otherwise. To define each respondent’s employment status, we use responses to three questions: (i) whether a person was engaged in agriculture work for at least 10 days in the past year, (ii) whether a person worked for other farmers and got paid for at least 10 days, and (iii) whether a person worked for at least 1 hour in the past week doing nonagricultural work. If a respondent answers yes to any of the three questions, the respondent is categorized as being employed.

The main independent variables are health indicators, including self-reported health status, functional limitations, and prevalence of specific health conditions and risk factors. Self-reported health is grouped into four categories: very good, good, fair, and poor or very poor. Functional limitation is measured by whether a person has difficulty in activities of daily living (ADL), instrumental activities of daily living (IADL), or has other types of limitations. We use two measures of ADL difficulties. The first (ADL 1) takes a value of 1 if any of four activities (walking, dressing, taking a bath, and eating) cannot be done independently, and 0 otherwise. The second measure (ADL 2) also takes into account two additional activities (going to the toilet and getting out of bed). Three dummy variables are constructed to indicate visual impairment, hearing impairment, and physical limitations. Physical limitations take a value of 1 if a person has difficulty with either walking 100 meters or kneeling or squatting, and 0 otherwise. The prevalence of health conditions is measured by the number of diagnosed chronic illnesses among three specific diseases (diabetes, hypertension, and heart disease). A series of dummy variables are also created to measure whether a person suffers from diabetes, hypertension, heart disease, cancer, asthma, arthritis, back pain, or depression. Risk factors consider whether a respondent currently smokes or drinks alcoholic beverages. Individual characteristics also are controlled for, including educational attainment, whether married, whether living in an urban area, and the respondent’s region and ethnicity.

Table 3 presents summary statistics by sex and age group. Among the four categories, men in the reference group (ages 50–59) have the highest employment rates. For both males and females, older persons have worse health status than the younger group. There are a few differences in health status between the sexes. Older women have a higher proportion of self-reported poor health (33.5%) compared with older men (26.7%). Older women also are more likely to have functional limitations (ADL, IADL, kneeling, and physical limitations). They also have more diagnosed illnesses and a higher probability of having elevated depressive symptoms. Men are more likely to engage in risky behaviors. More than half of men smoke or drink alcohol, though older men are slightly better than the younger group in this regard. Turning to individual characteristics, women have less educational attainment than men for both age groups. In the bottom of the table, we present the coverage rates of health insurance and pensions. Almost everyone is covered by health insurance, and around 90% have a pension.

Appendixes 5 and 6 present the OLS regression results for men and women aged 50–59 years. Columns 2–4 show the results from different model specifications (0, 1, and 2) using different health indicators and control variables. Model specification (0) includes self-reported health; ADL (adl_1); number of illnesses among three specified diseases (diabetes, hypertension, and heart disease); smoking; marital status; whether living in an urban area; and province dummies. Instead of using the number of illnesses, model specification (1) includes a series of dummy variables indicating diabetes, hypertension, and heart disease. Model specification (1) also adds depression and educational attainment into the regression. Model specification (2) includes the second version of ADL (adl_2). Additionally, IADL, visual impairment, hearing impairment, physical limitation, cancer, asthma, arthritis, back pain, and drinking are included. The ethnicity dummies are also controlled for.

We start with model specification (0) for males. Compared with those who report very good health, those with fair or poor health are less likely to work. In particular, those having poor or very poor health have employment rates that are 12.8% lower. Having ADL limitations also reduces the probability of working as does having a higher number of illnesses. Surprisingly, smoking and drinking are positively associated with employment, perhaps because they are part of social activities in the workplace. Urban residents work less than rural residents. Turning to model specification (1), adding more health indicators into the regression does not change the effects of health indicators included in model specification (0) very much. Diabetes reduces the probability of working by 4.7%. Compared to those with a high school education, those with both less education (primary school) and those with more education (college) work more. When more covariates are included in model specification (2), we find that IADL difficulties and physical limitations also reduce employment.

The regression results for women show different associations. Across the board, self-reported health status does not affect employment. Unlike men, hypertension hampers working for women. This could reflect the fact that compared to men, women are more likely to be aware of hypertension and receive treatment (Lu et al. 2017). Surprisingly, depressive symptoms positively affect working. One possible explanation is reverse causality: Working might bring about psychological pressures leading to depression (Sun et al. 2019). Smoking has little influence, while drinking is associated with an increased likelihood of working.

Using the regression coefficient from Appendixes 5 and 6, we predict the employment rates for different groups of older persons. Figure 9 shows the tapped and untapped employment by sex and age group based on model specification (1). The estimated additional work capacity ranges from 13% to 45% for men, with greater untapped work capacity among those who are older. Across all age groups, women have smaller tapped and untapped work capacities than men. The estimated additional work capacity ranges from 6% to 31%.

To test the robustness of the results, we repeat the calculations using different regression models. Table 4 presents the expected untapped work capacity by sex and age group. It shows that the untapped work capacity using model (0) and model (1) are very similar for all groups. We further predict employment rates using the logistic regression rather than the OLS. The predicted values from both model (0) and model (1) are similar to the ones from the OLS-based simulations. Thus, our simulation results are largely robust to using different models. Without loss of generality, we use OLS model specification (1) in the following analysis.

We next explore heterogeneity in work capacities among different education groups and types of work. Figures 10 and 11 present the additional work capacity by level of educational attainment. Since the sample for high school and college graduates is limited, we split the sample into two groups: (i) elementary and below, and (ii) middle school and above. The latter group includes middle school, high school, and college graduates. For men, the group with less education has more tapped and less untapped work capacity compared to those with more education. For those aged 60–64 years, tapped and untapped employment does not vary much between the two different education levels, but the differences become noticeable for those aged 65 years and older. The results suggest that people with better educational attainment leave the labor market earlier.

The pattern for women is a bit different. First, differences in work capacity already can be seen in the youngest group (ages 60–64). The actual employment rates for the less educated and more educated groups are 66% and 40%, respectively, which is consistent with more educated workers leaving the workforce much earlier than the less educated. Only 12% of middle school and above graduates still work when they reach the ages of 75–79. Adding tapped and untapped work capacity together, the more educated group has less estimated health capacity to work. However, this does not mean they have worse health than those with only a primary education or less. One possible reason is that middle school graduates are more likely to be covered by the EBP program and have the opportunity to retire between the ages of 50 and 55. Estimated work capacity is generated based on the coefficients for the younger group, many of whom may have already stopped working. If this is the case, the predicted employment rate based on the reference group may not be very different from the actual employment rate.

Figures 12 and 13 show the additional work capacity by type of work for men and women, respectively. For those who report working in 2018, we use their current type of work. For those who do not report working in 2018 but were recorded as working in earlier waves of the CHARLS, their type of work is based on their previous responses. Since some individuals do not have information on their previous jobs, the analysis sample ($N=11,703$) is slightly smaller than the full sample. We classify work into three types: (i) farmwork (58.5%), (ii) nonfarmwork (25.7%), and (iii) both farmwork and nonfarmwork (15.8%). Men employed in a nonfarm job work less than the other two types of work and have the most additional work capacity. Those doing both farm and nonfarmwork have the most tapped work capacity. The pattern for women is a bit mixed. Just as for men, those not involved in farmwork are less likely to work. It is evident that retirement behavior is related to the type of work, which may help explain the urban–rural differences that are analyzed in the following section.

IV. Additional Analysis

This section includes two parts. The first discusses one of the most remarkable internal differences in work capacity in the PRC, the urban–rural distinction, and explores possible explanations. The second reviews changes in additional work capacity between 2015 and 2018.

A. Urban–Rural Differences

As discussed above, the urban–rural difference in untapped work potential is as significant as the sex difference. To explore these differences in more depth, we divide the sample into four groups: rural men, rural women, urban men, and urban women. Since urban and rural residents are covered by different pension schemes, the retirement ages may differ, especially for women. As seen in Appendix 3, the four groups leave the labor force at different ages. Specifically, labor force departures start earlier for urban women than for the other groups. This suggests that the most appropriate age range for the reference group may differ across groups given that the CMR method works best if the reference group is not yet experiencing many retirements. As we will see, the choice of reference group can significantly influence the estimated additional work capacity of some groups.

Table 5 presents the estimated work capacity for older persons aged 60–79 years using different reference groups. Besides the age range of 50–59 years, we also use ages 50–55, 55–59, and 45–49 for urban females since some urban women retire starting at age 50. Not surprisingly, compared to the baseline of using ages 50–59 as the reference group, when a younger reference group (ages 50–54) is used, the predicted untapped work capacity increases, and when an older reference group (ages 55–59) is used, the predicted untapped work capacity decreases. If we look closely at the magnitudes of the differences, Table 5 reveals that estimated additional work capacity is relatively stable for rural males and rural females, regardless of the choice of reference group. This likely reflects the fact that rural residents normally do not stop working before age 60, so ages 50–59 serve as a good reference group for older persons in rural areas.

We find that the estimated work capacity of older persons in urban areas is much more sensitive to the choice of reference group. For urban men, the untapped work capacity ranges from 40.5% to 51.9%. The variation is even larger for urban women. When those with ages 45–49 are used as the reference group, the estimated untapped work capacity is 51.1%, which is twice as much as when ages 50–59 are used as the reference group. We use ages 45–49 as the reference group for urban women and ages 50–59 for other groups to estimate the additional work capacity for men and women. Comparing Figure 14 with Figure 9, changing the reference group for urban women yields a much larger additional health capacity to work for women. The total additional health capacity to work for women is even slightly larger than for men (22.6% versus 22.5%). Using this reference group, we conclude that women have slightly more untapped work potential than men, consistent with the results using the MW method.

Since employment rates for urban men and women already decrease to a certain extent for ages 55–59, using a younger reference group (ages 50–55) may be preferred. Although urban women may retire starting at age 50, using those aged 45–49 years as the reference group for those aged 60 years and older may raise issues of comparability. We next present results when the reference group for both urban men and women is changed to ages 50–54. Figure 15 shows the tapped and untapped employment rates by urban versus rural location and sex. Older persons in urban areas have greater untapped work capacity than their rural counterparts. Moreover, the older groups have significant untapped potential, with urban males having untapped work potential of 39%–74% and urban females having relatively less (but still significant) untapped work potential of 21%–36%. Rural men have the highest tapped employment capacity and the least untapped employment capacity. Rural women work a bit less than rural men and so have slightly higher estimated additional work capacity of between 4% and 30%.

One plausible explanation for the urban–rural difference in employment is the difference in pension systems. Figure 16 shows the additional work capacity by sex and pension. Comparing Figure 16 with Table 5 (reference: ages 50–59), the explanation seems plausible. For example, when males receive the RBP, their tapped and untapped work capacity is the same as rural males. To test whether the urban–rural difference exists after the type of pension is controlled for, we examine the health capacity to work by urban versus rural location and type of pension. Although rural residents are mainly covered by the RBP, there are some who are covered by the EBP (Table 1). Appendixes 7 and 8 illustrate the additional work capacity by urban versus rural location and pension type for men and women separately. For men, although there is a difference by pension type, the urban–rural difference also persists. Rural men work more than their urban counterparts even when they have the same type of pension scheme. For example, among rural men receiving the EBP, 59% work and there is an additional work capacity of 31%. Meanwhile, only 19% of urban men who receive the EBP work. Women in general work less than men, but the urban–rural difference also persists even after controlling for pension type.

To further identify the persons in rural areas that work more, we tabulate the types of work by pension (Appendixes 9 and 10). We find that those involved in farmwork tend to work more. About 54% of rural men who receive the EBP do farmwork, while only 5.4% of their urban counterparts do farmwork. The pattern is also evident for those with either no pension or the RBP, and for women. Farmers retire later than people who do not do farmwork, which pushes up the employment rate in rural areas. Taking into account all of these results, although the pension system indeed shapes people’s work behavior, large urban–rural differences remain even after controlling for differences in pension type, with farmwork being associated with higher employment rates at all ages.

B. Changes in Additional Work Capacity over Time

This section examines changes in additional work capacity from 2015 to 2018 using the CMR method. Such analysis can reveal recent trends in how potential work capacity has been changing and clarify the extent to which such changes can explain differences in our findings with those of Zhan et al. (2022), who used data from the 2015 wave of CHARLS. We replicate our results using the same methods applied to data from the CHARLS 2015 wave—see Appendixes 11 and 12 for sample statistics and the regression results—and then compare the estimated work capacity with that using the CHARLS 2018 wave. Although there was no major reform in the social security system after 2015, the state did expand existing schemes to cover more of the population. The coverage of the health insurance and pensions increased by more than 10% over the 3 years between the survey waves (compare Appendix 11 with Table 3).

Table 6 shows the estimated additional work capacity in 2015 and 2018. The general pattern is similar for both years. Urban people, especially urban males, have the highest additional work capacity, while rural males have the lowest additional work capacity. Table 7 presents the changes in additional work capacity between 2015 and 2018. It shows that additional work capacity decreased for most groups over the 3 years, except for urban men. We further investigate the changes in potential work capacity and actual employment rates to identify the source of the variations. For females, the potential work capacity decreased over the 3 years, while actual employment either increased or decreased by a smaller magnitude. Therefore, the changes in additional work capacity are negative. Urban men are the only group that experienced a positive change in additional work capacity. In particular, additional work capacity increased by over 8% for those aged 70 years and above.⁷ For most age groups, changes in both potential work capacity and actual employment are negative, but the actual employment rate decreases faster than potential work capacity, resulting in a negative change in additional work capacity. Our findings are consistent with earlier trends analyzed by Zhan et al. (2022), who find similar negative changes between 2011 and 2015 for most groups (except urban men). Combining this information, it appears that additional work capacity decreased steadily from 2011 to 2018.

V. Conclusion

This paper studies the health capacity to work among older persons in the PRC. Applying two widely used methods, the MW and CMR methods, we confirm that there is significant untapped work capacity among the PRC’s older population. The MW method results suggest that in 2015, people aged 50–64 years could work 3.3 years more and that women have higher additional work capacity (4.2 years) than men (2.7 years). When further exploring the heterogeneity between urban and rural residents, we find that older persons in rural areas work more and have less untapped work capacity than older persons in urban areas.

The CMR results confirm that health indicators affect people’s work decisions. For example, men reporting poor health are less likely to work. The simulation results show that in general, men have higher untapped work capacity (about 13%–45%) than women (about 6%–31%). However, the estimated additional work capacity can be sensitive to the age of the reference group, particularly for urban women. If we recalculate the additional work capacity using the younger reference group (ages 45–49) for urban women, the additional work capacity for women is slightly larger than for men. The results also show that older persons in urban areas, particularly urban males, have higher untapped work capacity than older persons in rural areas. Moreover, urban–rural differences in the results are not fully explained by the different types of pension schemes. We also find that the types of work may explain part of the urban–rural difference, as those engaged in farmwork tend to retire later, regardless of the type of pension coverage.

Both methods find that there is significant additional work capacity among older persons, but the results differ in the relative magnitude for men versus women (unless a young reference group for women is used in the CMR analysis). Both methods find that older persons in urban areas have more additional health capacity to work than older persons in rural areas and that rural men work the most and have the least untapped work capacity.

Our results show that older persons in the PRC have the potential to work more. Realizing this untapped potential could help address labor force shortages associated with rapid population aging. Policymakers may consider a variety of ways to encourage older persons to work more by, for example, increasing the mandatory retirement age, designing pensions systems to avoid disincentives to work beyond the retirement age, making work arrangements and wage-setting more flexible for older workers, and investing in work environments and human capital to make older workers more productive, among others. Our findings also suggest that additional health capacity to work varies across different subpopulations, which can provide guidance to policymakers on how to target their efforts to encourage older persons to work more. Specifically, urban residents, women, and more educated workers not working in agriculture have higher untapped work potential. In contrast, older persons in rural areas who are working in agriculture, especially rural men, already work to their health limit. A policy priority could be to provide better social security to allow them to retire earlier.

ORCID

Zeyuan Chen  https://orcid.org/0000-0002-2468-3585

Albert Park  https://orcid.org/0009-0007-3323-120X

Notes

¹ To view all appendixes, please refer to the supplemental materials that are available at: https://www.worldscientific.com/doi/suppl/10.1142/S0116110524400055.

² A public pension is available to employees working in the public sector.

³ This estimate derives from the authors’ calculations based on the National Bureau of Statistics of China (2022).

⁴ The 2020 census did not release the employment rate by age. The 1990 and earlier censuses used a different definition of employment. Therefore, they are not used in our analysis.

⁵ CHNS is an international collaborative project between the Carolina Population Center at the University of North Carolina at Chapel Hill and the National Institute for Nutrition and Health (formerly the National Institute of Nutrition and Food Safety) at the Chinese Center for Disease Control and Prevention. It records rich information on health and nutrition. CHARLS is managed by the Institute of Social Science Survey at Peking University and surveys a nationally representative sample of Chinese residents aged 45 years and older (Zhao et al. 2013). CHARLS includes information on demographic, socioeconomic, and health characteristics of older persons, as well as information on living arrangements and intergenerational transfers. The two surveys use slightly different versions of a self-assessed health questionnaire. CHARLS asks: “I have some questions about your health. Would you say your health is very good, good, fair, poor, or very poor?” Those who respond “poor” or “very poor” are classified as having poor health. CHNS asks: “Right now, how would you describe your health compared to that of other people your age? Excellent, good, fair, poor?” People who answer “fair” or “poor” are considered as having poor health.

⁶ Follow-up surveys followed the baseline sample from 2011 and periodically added “younger” households to keep the sample nationally representative for those aged 45 years and older.

⁷ The increase in additional work capacity is particularly large for urban men aged 70 years and above, and it is due to the joint effect of decreases in actual employment and either an increase or only slight decrease in potential work capacity (Table 10). We follow Zhan et al. (2022) to decompose the changes in potential work capacity into three parts: (i) changes due to differences in health status, (ii) changes due to differences in a job’s sensitivity to health status, and (iii) changes owing to other factors. For urban men aged 75–79 years, the changes in potential work (5.7%) are mainly driven by other reasons (Appendix 13).