Assessing the mental health implications of carbon trading policies: evidence from urban China

Baseline results

The results of the implementation of the CET policy on depression among middle-aged and older adults are shown in Table 2. In column (1), we included the dummy variable DID (Treat × Time) and individual fixed effects, year-fixed effects, and province-fixed effects to control fixed and unchanging characteristics. Notably, region-year fixed effects were also included to capture unaffected by time and region. The coefficient of DID was 1.114, statistically significant at the 1% level, indicating that the policy implementation increased the risk of depression symptoms among adults in the trial cities.

In column (2), we retained year and province-fixed effects while excluding individual-fixed effects. The DID dummy variables remained significant and positive, although the coefficient was slightly lower than column (1). Moreover, females were more likely to have depression than males. Education emerged as a significant protective factor against depression, while smoking was identified as a detrimental factor for mental health. In column (3), individual-fixed effects were introduced from column (2), with the DID dummy variable maintaining its significant and positive coefficient. In column (4), we included additional confounding variables from column (1). The DID dummy variable continued to display a significant and positive coefficient, albeit slightly reduced. Furthermore, marriage was found to act as a protective measure against depression. Poor memory was linked to increased depression risk, while active participation in social activities reduced the likelihood of experiencing depression. Higher life satisfaction was also associated with reduced depression levels.

Parallel trend testing

To ensure consistency in DID estimation results, the treatment and control groups must conform to the parallel hypothesis. This implies that the depression scores of both groups should exhibit a relatively stable trend without the CET policy intervention. While Fig. 4 depicts the variation in depression scores between the treatment and control groups spanning from 2011 to 2018, additional empirical examinations are necessary to corroborate these findings. Therefore, this study utilized the event study methodology outlined by Jacobson et al. (1993) to affirm the discrepancies in CET policy implementation across various time frames. The specification was set as follows:

Analyzing the coefficients of the parallel trend test illustrated in Fig. 5, a noticeable pattern emerges. Prior to the implementation of the CET policy in 2013, the coefficients showed a downward trajectory, indicating a declining trend. However, following the policy enactment (2015–2018), there was a clear shift as the coefficients started to ascend. This suggests that the introduction of the CET policy in China had a substantial effect, with its influence becoming more pronounced post-implementation and sustaining a positive trajectory. Consequently, these observations furnish supplementary evidence bolstering the reliability of the principal findings within this study.

Robust checks

Placebo test

The placebo test was conducted to assess the robustness of the impact of the CET policy on depression. We randomly assigned “false CET policy cities” as treatment groups and re-estimated the model from column (4) of Table 2, repeating this process 500 times. Figure 6 shows the kernel density of the estimated coefficients, which exhibit a near-normal distribution with a mean close to 0, contrasting sharply with the baseline regression results. This confirmed the robustness of the effect of the CET policy on the depression of dwelling residents.

Placebo test (Distribution for coefficients of treat).

Table 3 displays the other types of the placebo test. In column (1), the pilot cities were substituted with eight adjacent provinces: Hebei, Zhejiang, Hunan, Guangxi, Anhui, Jiangxi, and Inner Mongolia. Here, the DID coefficient was insignificant, suggesting that the policy implementation did not influence depression. Moving to column (2), observations from 2015 were excluded. Despite this adjustment, the coefficient of DID remained positive and significant, indicating that policy implementation increased depression in the pilot sites. Finally, in column (3), the policy implementation year was redefined at the beginning of 2013. The positive and significant coefficient of DID was inconsistent with the hypothesis. The policy was announced in 2011 and had a significant effect on depression in the second wave. However, the CHARLS data set only includes survey data from four waves, with the first wave occurring in 2011.

PSM-DID

The PSM-DID approach was employed to tackle the issue of non-random selection of CET policy pilot cities. Given that these selected CET policy cities do not strictly adhere to a natural experiment, which potentially biased results. Initially, all control variables were designated as matching variables, The nearest neighbor method was utilized to match treatment and control groups, which ensured the fulfillment of common support conditions. Finally, the DID method was applied to estimate the potential impact of policy implementation on depressive symptoms. The findings, presented in column (4) in Table 3. indicated that the CET policy tended to elevate the risk of depression, consistent with the baseline results.

Replace the dependent variable

The dependent variable was replaced to validate that the baseline was robust. Column (1) of Table 4 substituted the CES-D score with a binary indicator for moderate to severe depression. The analysis revealed that individuals in the eight pilot cities significantly exhibited a higher propensity for experiencing moderate to severe depression post-policy enforcement. Moving to column (2), self-reported health was utilized as an independent variable. Here, the negative and statistically significant coefficient of DID suggested a decrease in self-reported health following policy implementation in the pilot cities.

Considering the interference of the Fujian Province

The staggered time treatment group of Fujian province was introduced to validate the robust result. In Table 4, column (3) excluded the observations from Fujian province, yet the statistically significant and positive DID coefficient persisted. Finally, column (4) introduced the interaction term between DID and Fujian, employing the triple difference method. The positive and statistically significant coefficient for DID suggested that policy implementation elevated depression likelihood among adults in Fujian compared to other provinces.

Excluding other policy interference

Alongside the implementation of the CET policy, China also introduces other environmental-related policies to promote low-carbon development. Focusing solely on the CET policy in our estimation cannot capture whether the depression effect is due to the carbon emission trade policy or others (Zhao et al. 2023). To this end, we included a dummy variable for the low-carbon city policy in our model to estimate the net effect. As shown in column (5) in Table 4, the coefficient of implementing the CET policy was positive and significant after accounting for the potential influence of the concurrent low-carbon city policy.

Staggered did biased diagnosis

The staggered DID method may cause biased estimates due to the different timing exposures in the two treatment groups. Fortunately, this issue can be solved by decomposing the estimation into three groups based on traditional 2 (stage) × 2 (group) DID estimation (Bacon, 2021). The results of Bacon decomposition are shown in Table 5, which validated our baseline was robust.

Heterogeneity

According to the relative deprivation theory (Ragnarsdóttir et al. 2013; Smith and Pettigrew, 2014), individuals don’t assess their life circumstances in isolation but rather in comparison to others. Even when objectively their outcomes or social standing seem satisfactory, they may still feel dissatisfied. This can lead to emotions like anger, sadness, and dissatisfaction, as well as an increase in psychological disorders (Beshai et al. 2017; Osafo Hounkpatin et al. 2015). Socioeconomic status influences how people perceive relative deprivation; those with higher status may perceive social transitions as riskier due to possessing more resources and money, which could lead to greater financial loss or a decrease in status (Oyenubi and Kollamparambil, 2023). Consequently, the implementation of the CET policy may have a more pronounced and negative impact on adults with greater social resources and financial capital among higher socioeconomic status groups.

The results of the heterogeneity analysis are outlined in Table 6. Columns (1) and (2) display the findings for different age cohorts. In column (1), the coefficient of the DID term was 0.444, significant at the 10% level, while in column (2), it was 0.979, significant at the 1% level. Notably, policy implementation had a more pronounced adverse impact on depression among adults aged 60 years and above. At this stage of life, older adults face unique circumstances making them more vulnerable to the effects of new policies and societal changes, leading to exacerbated depressive symptoms. Columns (3) and (4) present the results for different genders. In column (3), the coefficient of the DID term was 0.818, significant at the 5% level, while in column (4), it was 0.870, significant at the 1% level. This suggests that the negative impact of policy implementation on depressive status is similar for males and females, albeit slightly more pronounced among males. Despite traditional cultural norms linking males to higher socioeconomic status in East-Asian society, the impact of policy implementation across gender cohorts was found to be similar. Columns (5) and (6) demonstrate the results for different educational levels, categorized as low (primary school and below) and high (junior high school education and above). The coefficients in columns (5) and (6) were 0.668 and 1.005, respectively, significant at the 5% level. This indicates that policy implementation had a more significant effect on adults with higher educational attainment. Higher education groups, typically associated with higher socioeconomic status and more resources, are more sensitive to welfare losses and social transitions following the CET policy implementation.

Columns (7) and (8) in Table 6 examine the results across different asset cohorts. A household holding less cash than the average is categorized as a low-asset cohort, while those with an amount equal to or above the average are considered part of the high-asset cohort. In column (8), the coefficient of the DID term was statistically significant at the 1% level, with a value of 0.971. However, in column (7), it did not reach significance. This indicates that the policy implementation tended to increase the risk of depression among adults in the high-asset cohort, while no discernible impact was observed among adults in the low-asset cohort. Therefore, individuals in the high-asset cohort demonstrate a heightened tendency towards risk perception and a more pronounced sense of relative deprivation. Moving on to columns (9), (10), and (11), the results are presented across different regions. In the eastern and western regions, the DID coefficients were 1.388 and 0.597, respectively, both significant at the 1% level. However, in the central region, while the DID coefficient was positive, it did not achieve statistical significance, likely due to the limited sample size. Remarkably, the impact of the CET policy appears to be more pronounced in the eastern region, which is situated in more developed economic areas and exhibits a higher wealth status among households within the six pilot cities.

Mechanism analysis

The research indicates a significant increase in depression risk among adults following CET policy implementation, yet the precise mechanism remains unclear. However, insights from related studies suggest potential explanations. Firstly, individuals with depression often exhibit memory deficits, which may manifest prior to depression onset, serving as crucial precursors. Secondly, common stressors such as retirement and child bereavement, particularly prevalent among individuals aged 45 and above, can exacerbate mental health challenges. In the context of policy implementation, these stressors may synergistically impact individuals’ mental well-being, triggering feelings of nervousness, sadness, and anxiety. Consequently, two key questions emerge: What types of memory deficits contribute to depression during new policy implementation? And, how do stressful life events influence depression during this period?

The mediating effect of memory deficits

The results of the medication effects are detailed in Table 7. While the policy implementation does not consistently have a direct impact on depression, there is evidence to suggest that memory deficits may exert a mediating effect. Immediate and delayed word recall tests were utilized to investigate this. In column (1), the coefficient of immediate word recall was −0.164, significant at the 1% level, indicating an increased risk of depression in adults. Column (2) revealed a coefficient of −0.261 at the 5% significance level, suggesting that policy implementation had a negative and significant effect on immediate word recall. This implies that the partial mediating effect of immediate word recall between policy implementation and depression was significant. However, in column (4), the coefficient of delayed word recall showed no significant effect on depression, indicating no mediating effect. Bootstrapping and Sobel tests were used to ensure the robustness of the mediating effect. Results confirmed that the CET policy and depression symptoms mediated immediate word recall, but not delayed word recall. These findings underscore the nuanced role of memory deficits in influencing the relationship between policy changes and depression, with immediate word recall exhibiting a notable mediating effect while delayed word recall does not.

The moderating effect of life stress events

The results of the moderating effect analysis are outlined in Table 8. Traumatic life experiences can inflict considerable emotional distress, including sadness and depression. Retirement, a significant life transition for middle-aged and elderly adults, as well as child bereavement, can profoundly impact individuals and exacerbate depressive symptoms. In column (1), the synergistic effect of depression was estimated using the interaction term of DID and retirement. The coefficient of the interaction term was 0.701, significant at the 1% level. This suggests that middle-aged adults were more susceptible to depression post-retirement, particularly in the eight pilot cities following policy implementation. Column (2) estimated the moderating effect of policy implementation on depression by introducing the interaction term of DID and child death. The coefficient for this interaction term was positive, with a value of 1.557, significant at the 1% level. This indicates that child bereavement significantly heightened the risk of depression in adults, particularly in the context of the CET policy implementation. Overall, the findings highlight the significant synergistic effects between stressful life events and the implementation of the CET policy, underscoring the importance of considering individual circumstances and life experiences when assessing the impact of policy changes on mental health outcomes.