The dependent variable in this study is the level of depression, measured using items adapted from the Center for Epidemiologic Studies Depression Scale (CES-D), originally developed by Radloff (1977). The depression score was derived from nine questions in the CLASS survey assessing respondents’ mood over the past week, including items such as “Have you felt very sad in the past week?” These questions capture various dimensions of psychological well-being, including negative affect, loneliness, life satisfaction, sleep quality, perceived self-worth, and identity.

Each item was rated on a three-point scale: “never” (1), “sometimes” (2), and “often” (3). The total score ranges from 9 to 27, with higher scores indicating a greater tendency toward depression and poorer mental health.

The independent variable in this study is the status of retired Danwei leaders, as identified through respondents’ answers to the CLASS survey question: “What was your occupation before you stopped working for income?” Individuals who reported their former occupation as “national, enterprise, or institutional leaders” were coded as 1, representing retired Danwei leaders. Those who selected “professional technicians,” “general office staff,” or “general staff in commerce, service, or manufacturing sectors” were coded as 0, indicating retirees who did not hold Danwei leadership positions.

In selecting control variables, this study draws on the work of Wang et al. (43) and includes demographic characteristics such as gender, age, ethnicity, marital status, religious beliefs, and whether living with children. It also incorporates health-related variables, including basic functional ability and ability to perform daily activities, as well as social support, which serves as an indicator of social capital.

To evaluate the robustness of the association between retired work-unit identity and depression, this study employs “whether the pre-retirement employer belonged to the public sector” as an instrumental variable to address potential endogeneity. This variable satisfies both the relevance and exogeneity conditions.

In terms of relevance, leadership positions in China’s public sector organizations (party and government agencies, public institutions, and state-owned enterprises) are standardized and associated with explicit administrative ranks, whereas management positions in the non-public sector are more flexibly defined (44). Thus, pre-retirement sector type is a strong predictor of whether an individual held a leadership position.

In terms of exogeneity, any direct effect of sector type on depression would primarily operate through cumulative work environment effects, social support structures, and family patterns. However, these influences are mitigated by controlling for health, social capital, and demographic covariates. Institutionally, leaders in both public and non-public sectors enjoyed similar high status and resource advantages prior to retirement (45), and the core challenge they faced was identity transformation and status decline after retirement. Sector type therefore serves mainly as an identifier of “work-unit leadership” and does not directly affect depression.

The instrumental variable was constructed from respondents’ answers to the survey item on “type of pre-retirement work unit.” Those who reported “government departments/party and government agencies/people’s organizations,” “public institutions,” “state-owned or state-controlled enterprises,” “collective enterprises,” or “military” were coded as 1 (public sector before retirement). All other responses were coded as 0 (non-public sector before retirement).

Building upon the preceding discussion and drawing on previous studies (31, 32), this study conceptualizes the impact of retired Danwei leaders’ occupational identity on depression through two primary pathways: changes in physical health and changes in social relationships.

The extent of change in physical health is measured using a single-item scale, with response options ranging from “basically no change,” “very little change,” and “moderate change,” to “significant change,” coded as 1, 2, 3, and 4, respectively. Higher scores reflect greater deterioration in physical health after retirement compared to pre-retirement.

Changes in social relationships are assessed using three items that evaluate perceived changes in relationships with family, former colleagues, and friends or neighbors. Each item follows the same four-point response scale used for physical health. The total score is calculated by summing the three items, with higher scores indicating greater perceived disruption in social relationships after retirement.

Given that retired Danwei leaders’ subjective attitudes toward society may influence their mental well-being, this study incorporates “willingness to participate in social activities” as a moderating variable.

This variable is derived from two items in the CLASS survey: “If given the opportunity, I am willing to participate in some work of the village/residential committee” and “I often think about doing more for society.” Responses to both items are rated on a five-point Likert scale: “completely disagree,” “somewhat disagree,” “neutral,” “somewhat agree,” and “completely agree,” coded from 1 to 5, respectively. Higher scores reflect a stronger willingness to engage in social activities.

The results above suggest that holding the identity of a retired Danwei leader is significantly associated with elevated levels of depression. To enhance the scientific validity and credibility of these regression findings, this study conducted robustness checks using nearest neighbor matching, kernel matching, and radius matching methods.

Table 3 presents the results of the covariate balance test. After matching, the standardized differences of most covariates fall below 10%, and the differences between the treatment and control groups are statistically insignificant. This indicates that the propensity score matching (PSM) procedure has effectively reduced selection bias in the sample.

Table 4 reports the average treatment effect on depression levels among retired Danwei leaders. The nearest neighbor matching results show that, compared to retirees without Danwei leadership experience, retired Danwei leaders exhibit depression scores that are 0.404 points higher, significant at the 5% level—consistent with the baseline regression results. Similar findings are obtained using kernel matching and radius matching, further confirming the robustness of the main conclusions.

To verify the robustness of the baseline regression, this study further employed the entropy balancing method (Entropy Balancing), which is considered more accurate for handling high-dimensional data. This method, first proposed and applied by Hainmueller (46), assigns optimal weights to covariates by constraining the first, second, and third moments of the covariates in the treatment and control groups to achieve balance. In doing so, the weighted distributions of the treatment and control groups are made as similar as possible, thereby attaining exact matching. A simple regression is then conducted on the matched sample, and the coefficient of the explanatory variable represents the average treatment effect.

In this study, the treatment group comprised retired Danwei leaders, while the control group consisted of non-Danwei retirees. Tables 5, 6 report the means, variances, skewness, and standardized differences of covariates between the treatment and control groups before and after entropy balancing. The results show that covariate differences existed between the two groups prior to entropy balancing; however, after assigning the optimal weights, the covariates were largely balanced across groups. Table 7 presents the regression results based on entropy balancing. It can be observed that, after entropy balancing, compared with retirees from other occupational categories, retired Danwei leaders had a 38.7% higher likelihood of depression, which was statistically significant at the 5% level. This conclusion is consistent with the previous PSM results and provides further support for Hypothesis 1.

To further assess the robustness of the regression results, this study employs an ordinary least squares (OLS) regression model as a substitute for the mixed-effects model. The results are presented in Table 8. Column (1) shows that, after controlling for demographic characteristics, health status, and social capital variables, the depression level of retired Danwei leaders is 0.387 points higher than that of non-Danwei leaders, and this effect is statistically significant at the 1% level. This finding is consistent with the results from the mixed-effects model, thereby confirming the robustness of the primary conclusion.

In addition, this study applies Pseudo-Poisson Maximum Likelihood (PPML) regression and Negative Binomial regression as further robustness checks. The results in Column (2) using the PPML model and in Column (3) using the Negative Binomial model are both significantly positive at the 1% level, aligning with the baseline regression results and further reinforcing the reliability of the findings.

To address potential reverse causality in the analysis, this study employs an instrumental variable (IV) approach. Table 9 presents the estimated association between being a retired Danwei leader and depression levels. The first-stage regression results in Column (1) show a significant positive association between having worked in the public sector prior to retirement and the likelihood of being a Danwei leader. The F-statistic exceeds the conventional threshold of 10, indicating that the selected instrumental variable is sufficiently strong and not weakly correlated with the endogenous variable.

The second-stage regression results in Column (2) demonstrate that the depression level of retired Danwei leaders is 0.031 points higher than that of non-Danwei leaders, and this effect is statistically significant at the 1% level. This result aligns with the baseline regression findings and supports the conclusion that Danwei leadership identity in retirement is associated with elevated levels of depression.

To ensure the validity and robustness of the findings, this study adopted several methodological strategies.

Internal validity. This study adopted multiple strategies in research design and methodological application to ensure the validity and robustness of the findings. First, with respect to internal validity, demographic characteristics, health awareness, and social capital variables were included as controls to systematically reduce potential confounding effects on depression levels. In addition, to address possible endogeneity arising from reverse causality, an instrumental variable approach was employed to strengthen the robustness of the findings and mitigate concerns about endogeneity. Nevertheless, unobserved confounders remain a potential threat. For instance, individual personality traits such as neuroticism may simultaneously influence both retirement decisions and vulnerability to depression, leading to omitted variable bias. Therefore, the internal validity of this study should still be interpreted with caution.

External validity. With respect to external validity, the study focused on retired Danwei leaders, a group with typically strong occupational identity and higher social status. The role transition and loss of social identity after retirement may exert particularly pronounced effects on their mental health. As such, the findings can be reasonably generalized to senior managers or professional and technical personnel with similar occupational characteristics, but caution is warranted when extending them to ordinary employees or other occupational groups.

Construct validity. In terms of construct validity, depression was measured using the well-established short form of the Center for Epidemiologic Studies Depression Scale (CES-D), ensuring adequate psychometric properties of the measurement instrument. To address potential conceptual overlap between “willingness to participate in social activities” and depression, a correlation test was conducted. The results revealed an extremely low correlation coefficient, indicating no substantive construct confusion and confirming discriminant validity between the two variables.

Robustness checks. Finally, multiple robustness checks further reinforced the credibility of the findings. In addition to propensity score matching (PSM) and entropy balancing methods that address sample selection bias, an instrumental variable (IV) approach was employed to mitigate potential reverse causality. The first-stage results confirmed a strong association between prior employment in the public sector and the likelihood of becoming a Danwei leader, with the F-statistic well above the conventional threshold of 10. The second-stage regression showed that retired Danwei leaders reported significantly higher depression levels compared to their non-Danwei counterparts, consistent with the baseline estimates. Furthermore, alternative estimation models—including ordinary least squares, Poisson regression, and negative binomial regression—were applied, and the results remained robust across specifications. Together, these robustness checks confirm that Danwei leadership identity in retirement is significantly associated with elevated depression levels.

Taken together, these strategies substantially bolster the validity and robustness of the findings, increasing confidence in the conclusion that occupational identity plays an important role in shaping the mental health of retired Danwei leaders.

Several mechanisms may help explain why retired Danwei leaders are more vulnerable to depression.

Health pathway. Better physical health is generally associated with lower levels of depression (24). During their careers, Danwei leaders often maintained regular routines and relatively active lifestyles, which benefited their health (28). However, once retired, these health-promoting habits may be disrupted, leading to deterioration in physical health and subsequently increasing depressive symptoms (29).

Social relationship pathway. Changes in social networks also play an important role (31). Compared with other workers, Danwei leaders were more likely to build broad social circles and enjoy higher social status during their tenure. Retirement, however, often brings a sharp contraction of networks and a rapid decline in status, which may result in psychological maladjustment, a sense of isolation, and insufficient social support (36, 37). While these losses increase vulnerability to depression, some retirees may attempt to offset them through alternative forms of social engagement—a compensatory mechanism that will be discussed in the next subsection.

Contextual differences. Beyond these mechanisms, heterogeneity analysis revealed that retired Danwei leaders in highly urbanized areas displayed significantly higher levels of depression. This may be associated with higher living costs, greater daily stress, and weaker neighborhood support in urban environments (51–53).

Taken together, these mechanisms highlight the vulnerability of retired Danwei leaders to depression. At the same time, they point to the potential importance of protective factors—most notably social activity participation—which will be elaborated in the next subsection.

As noted above, retirement often leads to a contraction of social networks and diminished social support, increasing the risk of depression among retired Danwei leaders. However, our findings reveal that active participation in social activities can buffer these negative effects. Specifically, willingness to engage in community or volunteer work significantly weakens the adverse association between retired Danwei leadership identity and depression. Retirees who remain socially active report lower levels of depressive symptoms, suggesting that social participation functions as a protective factor for psychological well-being.

This moderating role can be understood through its capacity to sustain a sense of purpose, maintain interpersonal networks, and reconstruct social identities that may otherwise erode after retirement. In this sense, social participation alleviates loneliness and mitigates the identity discontinuity caused by the abrupt withdrawal from institutional authority and prestige.

More broadly, the finding underscores the interplay between structural constraints and individsual agency in shaping late-life mental health. While the occupational identity of Danwei leaders reflects structural dependence on political and organizational authority, their willingness to remain socially active demonstrates agency in adapting to post-retirement life. This interaction indicates that even groups vulnerable to sharp role loss can effectively reduce psychological risks by engaging in meaningful social activities.

Taken together, the moderating effect highlights the importance of fostering social participation opportunities for retired leaders, aligning with broader evidence on active aging and suggesting a promising avenue for intervention in community-based mental health strategies.

This study adds to the long-standing debate on the mental health consequences of retirement. Some prior research has argued that retirement can relieve stress and improve well-being, particularly among those leaving highly demanding or stressful jobs (9, 12, 54, 55). In contrast, other studies emphasize the strong connection between occupational identity and psychological well-being in later life, showing that the loss of high-centrality roles increases vulnerability to depression (16, 18, 56). Our findings align more closely with the latter perspective, demonstrating that retired Danwei leaders—a group with highly central roles—are particularly prone to depression due to the abrupt loss of authority, prestige, and institutional support.

In addition, our results highlight the moderating role of social activity participation, which resonates with international evidence stressing the protective effects of productive and volunteer roles for mental health in later life. Previous studies have shown that maintaining active social engagement helps older adults sustain social integration and a sense of purpose, thereby reducing depressive symptoms (57–59). Consistent with this, we find that willingness to participate in social activities significantly weakens the adverse association between retired Danwei leadership identity and depression. This suggests that even for subgroups facing heightened risks in retirement, active social participation can provide an important pathway to resilience.

In summary, the evidence reveals a potential general pattern: the greater the power, autonomy, and symbolic recognition embedded in one’s occupational role prior to retirement, the greater the psychological costs of losing that role may be afterward. At the same time, our findings indicate that these negative effects are not irreversible; active social participation can partly buffer the psychological risks associated with role loss. In other words, post-retirement mental health outcomes depend not only on the extent of authority and status held before retirement but also on the ability to reconstruct purpose and social connections through engagement afterward. While similar tendencies have been observed in some Western managerial and professional groups, the case of Chinese Danwei leaders underscores the importance of institutional context, showing that post-retirement depression is shaped both by the scope and constraints of pre-retirement authority and by the capacity to maintain social participation afterward.