The rapid development of economy and society has resulted in an increase in not only workplace competition but also conflicts between work and family. Since the 1980s, work-family conflict has attracted the attention of scholars, gradually becoming a hot topic of research. Researchers believe that work-family conflict is a form of role conflict. When stress from work adds to the stress from family, or when work roles make family roles more difficult, work-family conflict emerges (9). Work-family conflict has been subdivided into two types: work that affects family (work-family conflict) and family that affects work (family-work conflict). Studies have also shown that among these two types, work-family conflict is more likely to occur (10). A meta-analysis revealed that the impact of work-family conflict is far greater than that of family-work conflict (1). A study on primary and secondary school teachers also proved that the interference of work with family is far greater than that of family with work (11, 12). This study also believes that in the context of Chinese culture, for primary and secondary school teachers, the main impact is work affects family, so it mainly discusses work-family conflict.

Turnover intention refers to the idea of individuals in an organization leaving their current workplace or job. Although it is only an idea rather than an actual behavior, it is an effective indicator that predicts turnover (12). Individuals with turnover intention may cause a series of negative impacts on their organizations (13). The impact of work-family conflict on turnover intention can be explained by the conservation of resource theory, which implies that work-family conflict is caused by the individual's excessive investment and consumption of resources in the work field, resulting in insufficient resources invested in the family field (14, 15). Work-family conflict causes trouble, stress, anxiety, and depression to individuals, and when the conflict reaches a certain level and exceeds individuals' ability to cope, the idea of quitting their job arises. A meta-analysis of turnover intention revealed that the correlation coefficient between work-family conflict and turnover intention is 0.20 (16). Several studies have shown that work-family conflict has a significant positive predictive effect on employee turnover intention (17). In summary, we concluded that work-family conflict plants the idea of leaving the profession among primary and secondary school teachers. Therefore, the following hypothesis is proposed:

H1: Work-family conflict positively influences turnover intention.

Psychological contract is an individual employee's subjective cognition or concept of mutual responsibility and obligation between himself and the organization (18). Although unlike an employment agreement or a labor contract, it does not have a clear legal effect, employees have internal expectations and evaluations of what responsibilities they should assume and how much they should be paid. Work-family conflict refers to the negative impact on personal family life due to the high workload and work pressure. Social exchange theory suggests that when individuals feel that they are giving more and receiving less at work, they will experience psychological distress and a sense of imbalance. In order to seek balance, the individual will reduce the effort they expend in their behavior, thereby also reducing the emotional investment and support for the organization (19). When the school allocates too much work and too many tasks to teachers such that their work interferes with their personal family life or rest, work-family conflict emerges. If teachers feel that their pay is not proportional to their investment, and they feel that the obligations and responsibilities of both sides are out of balance, they will reduce the psychological contract to seek balance. It can be inferred that work-family conflict is negatively correlated with and can negatively predict psychological contract.

Turnover intention is an individual's idea of leaving the job, and it is a kind of withdrawal behavior in the workplace (20). Psychological contract is a “psychological account” of the mutual responsibility between the organization and the individual, and it is the reciprocal behavior of both sides. Therefore, when teachers feel that the school is a responsible organization beneficial to their personal growth and development, their intention to leave is less because psychological contract includes not only teachers' current reciprocal behavior, but also their expectation of future rewards. When teachers subjectively believe that the school fails to fulfill its responsibilities or commitments, turnover intention increases (21). Accordingly, it can be speculated that psychological contract is negatively correlated with and can negatively predict turnover intention. Given that work-family conflict can negatively predict psychological contract, and psychological contract can negatively predict turnover intention, this study proposes the following hypothesis:

H2: Psychological contract plays a mediating role between work-family conflict and turnover intention.

Job satisfaction refers to the individual's subjective satisfaction with various factors related to work. It is an attitude variable reflecting the individual's positive emotional experience of work (2). All factors related to work can affect job satisfaction. The root cause of work-family conflict is that the work factor affects one's normal family life, which results in various troubles to individuals. According to the social exchange theory, negative attitudes toward work arise and subjective evaluation of work decreases because individuals believe that it is work that causes them unnecessary problems, thereby attributing the cause of the problems to their work (22). Subsequently, job satisfaction will decrease (23). A meta-analysis revealed that work-family conflict has a negative predictive effect on job satisfaction (24). This study suggests that work-family conflict is negatively correlated with and can negatively predict job satisfaction (25, 26).

Turnover intention is a subjective idea generated by individual dissatisfaction with work. When individuals are satisfied with their work, their turnover intention will naturally decrease. A good working atmosphere, humane management system, reasonable salary and benefits, and broad scope for career development will increase primary and secondary school teachers' satisfaction with their jobs. In return, teachers will experience a sense of belonging and trust in schools, thereby reducing their turnover intention. Therefore, this study suggests that job satisfaction is negatively correlated with and can negatively predict turnover intention. Because work-family conflict can negatively predict job satisfaction, and job satisfaction can negatively predict turnover intention, this study proposes the following hypothesis:

H3: Job satisfaction plays a mediating role between work-family conflict and turnover intention.

Both psychological contract and job satisfaction are subjective. When the school is perceived to fulfill its responsibility and obligation, it has fulfilled psychological contract, teachers' job satisfaction will gradually improve. According to the theory of social exchange, the essence of social exchange is reciprocity (27). A well-established psychological contract, in which teachers' internal expectations are respected and met, leads to a sense of belonging to the school and satisfaction with their work, which in turn influences their attitudes and behaviors toward the school (28). It can be inferred that psychological contract can positively predict job satisfaction (29), and job satisfaction playing a mediating role between psychological contract and turnover intention. Because the mediating roles of psychological contract and job satisfaction between work-family conflict and turnover intention have been discussed, respectively, in previous research, the following hypothesis is proposed:

H4: Psychological contract and job satisfaction play a serial mediating role between work-family conflict and turnover intention.

Questionnaire survey was used in this study, which is a popular and extensively used method to collect data from a target population (30). There are two main reasons for adopting this research approach. First, the purpose of this study is to explore the direct and indirect influence of work-family conflict on turnover intention, the questionnaire survey is helpful to measure variables. Second, the questionnaire survey is fast and low cost to collect data, it is commonly and widely used by researchers (31).

This study aims to explore how work-family conflict direct and indirect influence turnover intention. With psychological contract and job satisfaction as mediating variables, this study constructed a serial mediation effect model. Four questionnaires used in this study were derived from existing literature and have been used several times in published academic articles, showing good reliability and validity. All the questionnaire items were rated on a five-point Likert scale, with 1 implying “completely disagree” and 5 implying “completely agree.” Three academic professors and seven primary and secondary school teachers review the questionnaires and the four questionnaires were regarded as concise and easy to understand. All questionnaires that we used in this study are given in Appendix A in the Supplementary Material.

In this study, 567 questionnaires were distributed in primary and secondary schools in Henan, Anhui, and Jiangsu provinces. Of these, 505 valid questionnaires were recovered, the effective response rate being 89.07%. During the period when we distributed the questionnaire (August 2021), COVID-19 outbreaks occurred in different provinces in China, and primary and secondary schools took strict prevention and control measures against the epidemic. Therefore, questionnaires could only be distributed through the combination of online survey and on-site survey. The network survey was carried out through professional platform what is named “Wenjuanxing,” and the on-site survey was carried out by sending questionnaires to the participants and taking them back after answering them. In order to ensure the data quality, we took three measures: First, select high-quality questionnaires for the survey, and the total number of items in the four questionnaires is 20, so as to avoid boredom or fatigue of the participants. Second, emphasizes that this survey is only for academic research, so the participants should fill in according to their real situation to reduce their defensive and social approval effect. Third, after collecting the questionnaire, review the answers, delete more than 3 missed answers and linear, wavy answers of the questionnaire.

In this study, work-family conflict is used an independent variable, psychological contract and job satisfaction are used as mediating variables, turnover intention is used as dependent variable.

Table 1 presents the demographic characteristics of the sample. A total of 505 participants in this study, included 198 (39.21%) males and 307 (60.79%) females; 184 (36.43%) participants were aged 30 years and below, 268 (53.07%) were aged 31–45 years, and 53 (10.50%) were aged 46 years and above. With regard to their education, 27 (5.34%) participants had a college degree or lower qualification, 305 (60.40%) had an undergraduate degree, and 173 (34.26%) had a master's degree or higher qualification. There were 286 (56.63%) participants with junior titles, 142 (28.12%) with intermediate titles, and 77 (15.25%) with senior titles. With regard to the school site location, 226 (44.75%) participants were from urban areas and 279 (55.25%) were from non-urban areas.

In SPSS25, descriptive statistics is used to calculate the mean and standard deviation of each variable, correlation analysis is used to calculate the correlation coefficient between the variables (40). Confirmatory factor analysis is used in AMOS25 to evaluate the discriminant validity and common method bias between the four variables (41). Model 4 of PROCESS3.5 macro for SPSS was applied to examine the mediating effect of the data, and Model 6 was applied to examine the serial mediating effect of the data. All variables were standardized in Model 4 and Model 6 before data analyses (42).

The mean and standard deviation of each variable are shown in Table 3, and these data were used to evaluate the discriminant validity and common method bias between the four variables.

In the fitting index of confirmatory factor analysis, if χ²/df < 3, RMSEA < 0.08, SRMR < 0.08, TLI > 0.90, CFI > 0.90, which proves that the fitting effect of the model is good. In this study, five models were constructed, Table 2 reveals that the fitting index of the “four-factor model” (χ²/df = 2.97, RMSEA = 0.06, SRMR = 0.06, TLI = 0.93, CFI = 0.94) is best and other four competition models did not meet the basic requirements. This also shows that the discriminant validity between four variable in this study is good.

In order to control the bias effect of common methods, the questionnaire with good reliability and validity is used as the measuring tools. In the test process, the confidentiality of the results and the use of the results for academic research only were emphasized, and some questionnaire items were scored using the reverse scoring method. The common method bias was evaluated by control unmeasured single method- factor approaches (43, 44). Introduced the common method variance (CMV) in confirmatory factor analysis, Table 2 reveals that the fitting degree of the “four-factor model + CMV” model was not significantly improved (ΔRMSEA = 0.01 < 0.05, ΔSRMR = 0.02 < 0.05, Δ TLI = 0.02 < 0.1, Δ CFI = 0.03 < 0.1). Thus, there is no serious common method bias in this data (45).

The composite reliability (CR) of work-family conflict is 0.87, turnover intention is 0.77, psychological contract is 0.79, and job satisfaction is 0.90. The average variance extracted (AVE) of work-family conflict is 0.58, turnover intention is 0.48, psychological contract is 0.36, job satisfaction is 0.76. According to Fornell and Larcker (46) suggested, CR should exceed 0.6, and AVE should exceed 0.5 under ideal condition (47), while 0.36–0.5 are acceptable (46, 48, 49), so the CR and AVE of this study meet the standard.

Table 3 summarizes the mean, standard deviation, and correlation coefficient of work-family conflict, psychological contract, job satisfaction, and turnover intention. The mean values of four valuables ranger from 2.46 to 3.78, while the Standard Deviation ranged from 0.51 to 0.85. The results reveal that work-family conflict is significantly positively correlated with turnover intention (r = 0.29, p < 0.01) and significantly negatively correlated with psychological contract (r = −0.12, p < 0.01) and job satisfaction (r = −0.29, p < 0.01). Psychological contract is significantly positively correlated with job satisfaction (r = 0.63, p < 0.01) and significantly negatively correlated with turnover intention (r = −0.50, p < 0.01). Job satisfaction is significantly negatively correlated with turnover intention (r = −0.65, p < 0.01).

The input method was used for linear regression analysis in SPSS, and the results revealed that work-family conflict could positively predict turnover intention (β = 0.29, t = 6.082, p < 0.001). Therefore, Hypothesis 1 was verified.

In SPSS, the PROCESS macro was used to select Model 4 provided by Hayes for mediating effect analysis. When psychological contract is a mediating variable, work-family conflict can negatively predict psychological contract (B = −0.12), psychological contract can negatively predict turnover intention (B = −0.47), work-family conflict can positively predict turnover intention (B = 0.23), indirect effect is 0.06 (BootSE = 0.02, BootCI [0.01, 0.10]), the mediating effect accounts for 21% of the total effect, as shown in Table 4, so Hypothesis 2 was also verified.

When job satisfaction is a mediating variable, job-family conflict can negatively predict job satisfaction (B = −0.29), job satisfaction can negatively predict turnover intention (B = −0.62), job-family conflict can positively predict turnover intention (B = 0.11), indirect effect is 0.18 (BootSE = 0.03, BootCI [0.12, 0.25]), the mediating effect accounts for 62% of the total effect, as shown in Table 4, so Hypothesis 3 was also verified.

In SPSS, the PROCESS macro is used to select Model 6 provided by Hayes for serial mediating effect analysis. The results are shown in Figure 2, work-family conflict negatively predicted psychological contract (B = −0.12, t = −2.68, p < 0.01) and job satisfaction (B = −0.22, t = −6.61, p < 0.001), psychological contract positively predicted job satisfaction (B = 0.60, t = 17.83, p < 0.001), and work-family conflict positively predicted turnover intention (B = 0.12, t = 3.47, p < 0.01). Psychological contract can negatively predict turnover intention (B = −0.17, t = −3.94, p < 0.001) and job satisfaction can negatively predict turnover intention (B = −0.51, t = −11.53, p < 0.001). In the whole model, the total effect is 0.29, the direct effect is 0.12, and the total indirect effect is 0.17. There are three mediating paths that are obviously indigenous, as shown in Table 5. Path 1 is work-family conflict → psychological contract → turnover intention, and its effect value is 0.02. Path 2 is work-family conflict → job satisfaction → turnover intention, its effect value is 0.11; path 3 is work-family conflict → psychological contract → job satisfaction → turnover intention, and its effect value is 0.04; as a result, Hypothesis 4 was verified.