Xipei
Guo
a,
Xuemin
Hao
ab,
Jun
Ma
b,
Hongyan
Wang
c and
Weiping
Hu
*ad
aMOE Key Laboratory of Modern Teaching Technology, Shaanxi Normal University, Xi’an, Shaanxi, China. E-mail: Weipinghu@163.com
bSchool of Education, Shaanxi Normal University, Xi’an, Shaanxi, China
cXi’an International Studies University, Xi’an, Shaanxi, China
dShaanxi Normal University Branch, Collaborative Innovation Center of Assessment toward Basic Education Quality at Beijing Normal University, China
First published on 30th May 2022
Although there are numerous chemistry-related careers within the STEM fields, chemistry-related careers are not well regarded. High school is a critical time for developing students’ career choices. Previous studies suggest that anxiety and identity may be predictors of career choice. Therefore, the purpose of this study was to investigate the influence of high school students’ chemistry anxiety (learning anxiety and test anxiety) and chemistry identity (competence/performance beliefs, interest, external recognition, and holistic impression on identity) on chemistry career choices. Guided by the possibility of different hindrances to chemistry career choice for males and females, the study further detected gender-specific patterns of relations between variables. The results of multigroup structural equation modeling firstly showed that different constructs of chemistry identity were positive and significant predictors of chemistry career choice but varied by gender. Specifically, competence/performance beliefs and holistic impression on identity were significantly associated with females’ chemistry career choices. In contrast, interest, external recognition, and holistic impression on identity motivated males’ chemistry career choices. Secondly, the effects of chemistry learning anxiety and test anxiety on chemistry career choice were completely mediated by chemistry identity, whereas the pathways and strength of mediation differed between females and males.
One reason is that many students don’t choose to study chemistry in high school, further contributing to the decline in the number of graduate students who continue to pursue chemistry in college (Ardura and Pérez-Bitrián, 2018; Avargil et al., 2020). Specifically, in many countries, science-related courses are not required in high school, and students can choose whether to learn science-related subjects by themselves. This has an impact on high school students’ science learning, with the most pronounced impact on chemistry and physics (Ardura and Pérez-Bitrián, 2018). However, the subject learning in high school is the foundation and a necessary prerequisite for the selection of a student's major in college. In addition, researchers have noted that adolescence is a critical stage in forming career choices (Dudovitz et al., 2017). Drawing on research associated with social psychology, the career choices formed during adolescence play a decisive role in their career trajectory as adults (Bandura et al., 2001; Riegle-Crumb et al., 2011). Therefore, enhancing high school students’ propensity for chemistry career choice is critical to retaining chemistry talents within STEM fields.
Understanding what factors influence students’ willingness to choose a chemistry career is key to improving talent retention within the chemistry field. According to the previous studies, identity plays a large role in a student's career choice (Robinson et al., 2019), while both identity and career choice are domain-specific. Despite the increasing attention on the relationships between identity and career choice, few studies have investigated the role of chemistry identity in chemistry career choice for high school students (Nauta and Kahn, 2007; Cass et al., 2012; Li et al., 2015; Godwin et al., 2016). Therefore, the first aim of this study was to test whether chemistry identity is related to students’ chemical career choices so that we can influence their career choices by enhancing their chemistry identity.
Furthermore, relations between anxiety and either identity or career choices have been well reported in previous literature, and anxiety is similar to identity and career choice in that they are domain-specific concepts (Pekrun, 2006). Nevertheless, there is a lack of research on the relationship between chemistry identity, chemistry anxiety, and chemistry career choice, especially in the high school student population. More importantly, both chemistry identity and chemistry anxiety contain multiple dimensions, whereas few studies have examined intertwined relations between them and chemistry career choices in a single model. Thus, the second purpose of this study was to explore whether it is possible to increase students’ chemistry identity in multiple dimensions by reducing their different dimensions of chemistry anxiety and further influencing their willingness to choose a career in chemistry.
In addition, studies have shown a gender gap in the chemistry field (Huryn et al., 2017). Drawing on expectancy-value theory and related research, the gender gap in domain-specific related career choices may be explained by gender differences in identity and anxiety (Wang and Degol, 2013). For example, the level of math anxiety and the effect of math anxiety on science-related career interests differed between males and females (Huang et al., 2018). Based on this, we sought to further test whether the relationship between the different variables differed by gender so that we could make valid recommendations for gender-specific students.
Eccles’ expectancy-value theory (EVT), as a leading theory of human motivation, provides a theoretical framework for linking identity and career choices (Eccles et al., 2015; Gottlieb, 2018). According to EVT, the interaction of expectancy × value could predict students’ achievement-related choices (e.g., university major choice) and career choices (Nagengast et al., 2011). Expectancy refers to the individual's expected beliefs about the successful completion of a task, involving self-efficacy and competence judgments. Value refers to the individual's subjective judgment of the significance or importance of the outcome of the behavior, including the importance of completing the task, interest, the link between the task and the individual's future goals, and so on (Eccles, 2009). Based on EVT, identity as a motivational structure that consists of self-concept and subjective value perceptions together influences individuals’ career choices by informing individuals of the importance they place on the task and their expectations of success (Eccles, 2009; Wang and Degol, 2013).
In addition, although “I see myself as a domain-specific person” is considered a single indicator of the holistic impression on identity and is closely related to career choices, educational researchers believe that identity should be conceptualized as a more complex construct (Carlone and Johnson, 2007; Godwin et al., 2016). Specifically, a domain-specific identity has been framed around three key shaping constructs in previous studies: competence/performance beliefs, interest, and external recognition (Carlone and Johnson, 2007; Hosbein and Barbera, 2020b; Verdín, 2021).
The three identity shaping constructs also have been considered when exploring the factors that influence students’ decisions to pursue domain-related careers (Hazari et al., 2010). Competence/performance beliefs refer to students’ beliefs in the ability to understand content knowledge and perform well in applying competencies to solve tasks (Chen et al., 2021). It influences students’ career pursuits primarily by enhancing their self-efficacy (Fouad et al., 2002). Interest represents students’ attitudes toward learning and their desire to learn more and engage in relevant learning activities (Chen and Wei, 2020), and research studies show that higher interest is positively related to student's career choice and persistence (Adams et al., 2006). External recognition refers to teachers’, parents’, and peers’ perceptions of whether a student is a science or chemistry person, which influences the student's self-perception and self-expectations, and in turn, has an impact on the student's career choices (Hazari et al., 2010).
As mentioned before, EVT treats the “value” as one of the key elements in determining career choice, while “value” depends in part on the judgment of the cost (e.g., time, effort, etc.) to complete the task. When individuals have high anxiety about completing a task, they will perceive that they need to invest a significant amount of cost and thus will be reluctant to choose that task (Eccles, 2009; Degol et al., 2018). Indeed, expanding the focus of the study from career choices to general decisions, researchers similarly point out that higher levels of anxiety lead to negative expectations of outcomes, overestimating and tending to avoid the risks associated with choices. Thus, the higher the level of anxiety, the more it promotes an individual's decision-making avoidance (i.e., not making a choice or postponing making a choice) (Hartley and Phelps, 2012; Arbona et al., 2021).
Although there are few studies conducted to examine the relationship between chemistry anxiety and chemistry career choice, studies in other fields have shown that anxiety is associated with a person's career choices. For example, studies conducted in math fields showed that math anxiety has an impact on students’ math and STEM career choices (Ahmed, 2018; Cribbs et al., 2021). Given that chemistry anxiety is closely related to math anxiety (Eddy, 2000), this study provides information to examine the relationship between chemistry anxiety and career choices.
While the above studies on the relationship between anxiety or math anxiety and career choice provide the empirical and theoretical basis for this study, it is worth noting that chemistry anxiety consists of three dimensions (i.e., evaluation anxiety, learning anxiety, and handling chemicals anxiety) (Senocak and Baloglu, 2014). Whether the chemistry career choices of gender-specific students are influenced by different chemistry anxieties also needs to be further explored.
Based on Marcia's research, Crocetti et al. (2009) further investigated the causal relationship between anxiety and identity in Western adolescents through a five-wave longitudinal study. The results suggested that high-anxiety individuals’ uncertainty level of the initial identity commitment was higher than for low-anxiety individuals. Meanwhile, individuals with high anxiety levels experience a gradual decrease in their commitment levels over time. The reason is that highly anxious individuals are prone to doubt, rethinking, and re-exploring alternative identities and outcomes, so it is difficult for them to make a firm commitment to an identity. In contrast, the identity commitments of adolescents with low levels of anxiety become more solid over time (Crocetti et al., 2009). In short, high levels of anxiety is an important factor that hinders adolescents’ identity development. In addition, Cribbs et al. (2021) also demonstrated that mathematical anxiety was a significant predictor of mathematical identity.
1. How do high school students’ chemistry anxiety and chemistry identity relate to chemistry career choice?
2. Do the relations between chemistry anxiety, chemistry identity, and chemistry career choice differ by gender?
Given that both chemistry anxiety and chemistry identity are multidimensional variables, we further define the relationship between the different dimensions of each variable. The case of chemistry anxiety mainly includes evaluation anxiety, learning anxiety, and handling chemicals anxiety (Eddy, 2000). However, the above three dimensions of chemistry anxiety mainly target college students, and 10th grade students in China are less exposed to chemistry experiments, so this study only focuses on chemistry learning anxiety and chemistry test anxiety. For chemistry identity, it is recognized as including three shaping constructs (i.e., competence/performance beliefs, interest, and external recognition) and a single indicator (i.e., I see myself as a chemistry person) (Hosbein and Barbera, 2020; Verdín, 2021). The single indicator represents students’ overall perceptions of chemistry identity. Previous research in other fields (e.g., engineering) has demonstrated a positive relationship among the four aspects of identity, i.e., competence/performance beliefs influence students’ overall perception of identity through the mediating role of interest and external recognition, respectively (Godwin et al., 2016; Dou and Cian, 2021; Verdín, 2021) (as shown in Fig. 1). Based on this, we expand our work on how the different sub-constructs of chemistry anxiety and identity are associated with chemistry career choice, with one hypothetical model further presented in Fig. 2.
![]() | ||
Fig. 1 Relationships between recognition, performance/competence beliefs, interest, and holistic impression on identity based on previous identity framework. |
The reasons for selecting the 10th-grade students in China include the following three points. First, these students have already experienced two years of systematic chemistry learning and had some basic understanding of chemistry. Second, these students are about to choose which subjects they would continue to learn in 10th grade. Third, compared to other high school grades, students in grade 10 are under less academic pressure and thus have enough time to fill out the questionnaire carefully. Among them, 249 were males, accounting for 43.0%, and 330 were females, accounting for 57%.
Permissions were obtained from all appropriate authorities prior to the implementation of the survey. In addition, we explained the purpose of the survey to teachers, parents, and students. All students participated voluntarily with the consent of their parents and teachers, and no additional fees or costs were provided to students. We committed to protecting the privacy of our students.
First, the questionnaires used in the present study were adapted from the well-established questionnaire. Therefore, the reliability and validity of the adapted questionnaires need to be examined first. All samples were randomly divided into two groups. An exploratory factor analysis (EFA) was conducted firstly using SPSS 21.0 for sample 1 (Lee et al., 2008). After further revision based on the results of the EFA, sample 2 was subjected to CFA using AMOS 23.0 to verify the validity of the potential variables (Velayutham and Aldridge, 2012). In addition to this, Cronbach's Alpha (α) was performed using SPSS 21.0 to detect the internal consistency of the questionnaires (Hair, 2006). For the EFA results, the number of factors extracted was determined based on parallel analysis. According to O’Conno (2000), if the eigenvalue explained by a factor drawn from the actual data is larger than that explained by the corresponding factor drawn from the eigenvalue from the random data, the factor should be retained. In addition, the factor loading scores for each item should be higher than 0.40 (Wei et al., 2020). The CFA results indicate an acceptable model fit when the following indicators are shown: χ2/df < 5, RMSEA values < 0.08, SRMR < 0.08, as well as CFI and TLI values are greater than 0.9 (Opperman et al., 2013). An excellent model fit is indicated when χ2/df < 3, RMSEA and SRMR values are below 0.05, and CFI and TLI are greater than 0.95 (Brown and Cudeck, 1992).
Second, the direct and indirect effects in the hypothesized model were tested using multigroup structural equation modeling (Multi-group SEM) with AMOS 23.0. Indirect effects refer to the effect of the independent variable on the dependent variable through one or more mediating variables. Hence it is also called the mediating effect. If both mediating and direct effects are significant, the effect of the independent variable on the dependent variable is incompletely mediated; when the mediating effect is significant while the direct effect is insignificant, it means that the relationship between variables is completely mediated (González and Paoloni, 2015). The significance test for the effect was performed using the bias-corrected bootstrap method. It has been shown that the bootstrap method has higher test power than the Sobel test and is not affected by the pattern of data distribution (Edwards and Lambert, 2007; Preacher and Hayes, 2008). 5000 bootstrap samples along with 95% confidence intervals were used to determine the significance of the effect. If the 95% confidence intervals do not include 0, the effect is significant (Hayes, 2015). The fit of the entire hypothetical model was evaluated by the χ2/df, RMSEA, SRMR, GFI, CFI, and TLI.
Compared to the SEM, the advantage of multi-group SEM is that it not only tests the relationship between variables but more importantly, it also examines whether the relationship between variables is equal across sample groups or whether the parameters have invariance (Lee and Whittaker, 2021). Specifically, we constructed the following four nested models. An unconstrained model was constructed in the first step, which assumed a different model for each gender group. In the second step, we gradually added the constraints of equal structural weights, equal structural covariances, and equal measurement residuals to obtain three constrained models in turn (Orkibi and Ram-Vlasov, 2019). Then, the changes in CFI and chi-square values between the different nested models were used to determine whether the path relationships in Model 1 differed by the gender of the students. If the change in chi-square value is not significant (Δχ2p-value > 0.05) and ΔCFI < 0.01, it indicates that the relationships in the model remain consistent across groups (Cheung and Rensvold, 2002; Kang et al., 2018).
Although 10th grade students in China also perform chemistry experiments, the number of experiments is much less compared to college students. Their chemistry learning is mainly in the classroom, where the teacher will conduct a chemistry experiment demonstration. Thus, only two subscales from the DCARS (learning-chemistry anxiety and chemistry-evaluation anxiety) were used in this study. In addition, Eddy (2000) did not provide adequate reliability and validity statements for the DCARS, and Hopko (2003) demonstrated that the two-factor model fit of RMARS cited by DCARS was poor (Derek and Hopko, 2003). Furthermore, Hopko et al. (2003) censored the RMARS and constructed the Abbreviated Math Anxiety Scale (AMAS) with nine items. The two-factor model of AMAS has been shown to have good reliability and validity, as well as cross-gender measurement invariance in adolescent populations by researchers in several countries, including Germany, Spain, and Italy (Primi et al., 2014; Schillinger et al., 2018; Martin-Puga et al., 2022).
Taken together, we selected the corresponding nine items from the DCARS according to the AMAS to constitute the instrument for measuring chemical anxiety in the present study. Five items measure students’ anxiety about learning chemistry (e.g., having to use the tables in a chemistry book, which was translated into Chinese “”) and four items measure students’ chemistry-evaluation anxiety (e.g., thinking about an upcoming chemistry test one day before, which was translated into Chinese as “
”). All items were scored on a 5-point Likert-type scale (1 = “not at all”, 2 = “a little bit”, 3 = “moderately”, 4 = “quite a bit”, and 5 = “extremely” anxiety). As for the EFA, the results showed that the revised DCRAS had a two-factor structure (KMO = 0.905, Bartlett's spherical test showed χ2(36) = 1618.148 and p < 0.001) and the two factors explained 70.834% of the total variance. Parallel analysis results showed that the actual data eigenvalues of the first two factors were greater than the random data eigenvalues 95th percentile. The factor loadings of each item in the EFA ranged from 0.442 to 0.909. According to the CFA, the two-factor model fits well: χ2/df = 2.208, RMSEA = 0.065, SRMR = 0.039, CFI = 0.974, and TLI = 0.964.
Variable | Minimum | Maximum | Mean | SD | Skewness | Kurtosis |
---|---|---|---|---|---|---|
Notes. CLA chemistry learning anxiety, CTA chemistry test anxiety, Com/Per competence/performance, INT interest, E-Rec external recognition, HICI holistic impression on chemistry identity, Car-Cho chemistry career choice. | ||||||
CLA | 1 | 5 | 2.17 | 0.85 | 0.56 | −0.10 |
CTA | 1 | 5 | 2.98 | 1.03 | 0.07 | −0.65 |
Com/per | 1 | 5 | 3.06 | 0.82 | 0.24 | 0.08 |
INT | 1 | 5 | 2.08 | 0.93 | 0.26 | 0.96 |
E-Rec | 1 | 5 | 3.60 | 0.84 | 0.53 | 0.36 |
HICI | 1 | 5 | 2.31 | 1.08 | 0.37 | 0.56 |
Car-Cho | 1 | 5 | 2.56 | 1.06 | 0.21 | 0.40 |
1 | 2 | 3 | 4 | 5 | 6 | 7 | |
---|---|---|---|---|---|---|---|
Notes. CLA chemistry learning anxiety, CTA chemistry test anxiety, Com/Per competence/performance, INT interest, E-Rec external recognition, HICI holistic impression on chemistry identity, Car-Cho chemistry career choice. | |||||||
(1) CLA | — | ||||||
(2) CTA | 0.58*** | — | |||||
(3) Com/per | −0.46*** | −0.43*** | — | ||||
(4) INT | −0.35*** | −0.27*** | 0.61*** | — | |||
(5) E-Rec | −0.31*** | −0.35*** | 0.52*** | 0.38*** | — | ||
(6) HICI | −0.33*** | −0.34*** | 0.51*** | 0.45*** | 0.74*** | — | |
(7) Car-Cho | −0.28*** | −0.28*** | 0.48*** | 0.44*** | 0.57*** | 0.60*** | — |
Next, we performed tests of measurement invariance to ensure that the differences between groups were not due to measurement differences (Wigfield and Harold, 1997). Before the measurement invariance test, we first performed separate CFA analyses on the male and female data, and the results showed that all models fit well (see Table 3). Then, the measurement invariance test is completed in four steps. In the first step, a configured model (baseline model) is built to ensure that the survey items are related to each other in a similar way in all groups. In the second step, the metric model is built based on the configured model with the constraint of equal factor loadings. If the difference between the metric model and the unconstrained model is not significant, it means that there is weak invariance. In the third step, the scalar model is built by restricting the item intercepts to be equal based on the invariance of the metric model. If the intercept invariance holds, the measurement tool has strong measurement invariance. Finally, limiting the error covariances to be equal, the model has strict measurement invariance if the change from the third step is not significant (Robinson et al., 2020; Rocabado et al., 2020). As displayed in Table 3, the results support weak invariance (Δχ2p-value > 0.05, ΔCFI < 0.01, ΔRMSEA < 0.015, ΔSRMR < 0.03) and strong invariance across gender groups (Δχ2p-value > 0.05, ΔCFI < 0.01, ΔRMSEA < 0.015, ΔSRMR < 0.01). The significant change in the chi-square values of the conservative model compared to the scalar model indicates that the two measurement instruments didn’t reach a strict level of measurement invariance (Rocabado et al., 2020).
Model | χ 2 | DF | χ 2/df | CFI | SRMR | RMSEA | Δχ2p-value | ΔCFI | ΔSRMR | ΔRMSEA |
---|---|---|---|---|---|---|---|---|---|---|
Notes. CLA chemistry learning anxiety, CTA chemistry test anxiety, Com/Per competence/performance, INT interest, E-Rec external recognition, HICI holistic impression on chemistry identity, Car-Cho chemistry career choice. | ||||||||||
Chemistry anxiety | ||||||||||
Male | 62.084 | 26 | 2.388 | 0.963 | 0.041 | 0.075 | ||||
Female | 44.547 | 26 | 1.713 | 0.987 | 0.042 | 0.047 | ||||
Configural | 106.647 | 52.000 | 2.051 | 0.977 | 0.041 | 0.043 | ||||
Metric (weak) | 111.891 | 59.000 | 1.896 | 0.978 | 0.043 | 0.039 | 0.630 | 0.001 | 0.002 | −0.004 |
Scalar (strong) | 121.848 | 68.000 | 1.792 | 0.977 | 0.047 | 0.037 | 0.354 | −0.001 | 0.004 | −0.002 |
Conservative (strict) | 178.206 | 77.000 | 2.314 | 0.957 | 0.047 | 0.048 | 0.000 | −0.020 | 0.000 | 0.011 |
Chemistry identity | ||||||||||
Male | 116.563 | 51 | 2.286 | 0.965 | 0.051 | 0.072 | ||||
Female | 139.281 | 51 | 2.731 | 0.962 | 0.051 | 0.073 | ||||
Configural | 255.850 | 102.000 | 2.508 | 0.963 | 0.0509 | 0.051 | ||||
Metric (weak) | 263.366 | 111.000 | 2.373 | 0.963 | 0.0505 | 0.049 | 0.584 | −0.0004 | 0.000 | −0.002 |
Scalar (strong) | 283.936 | 123.000 | 2.308 | 0.961 | 0.0505 | 0.048 | 0.057 | 0.0000 | 0.002 | −0.001 |
Conservative (strict) | 364.581 | 135.000 | 2.701 | 0.945 | 0.0521 | 0.054 | 0.000 | 0.0016 | 0.0016 | 0.006 |
Third, the present study investigated the mean differences across gender on each variable through an independent-samples t-test. As shown in Table 4, both chemistry learning anxiety and chemistry test anxiety were significantly higher in females than in males. Regarding the chemistry identity, females’ competence/performance, external recognition, and holistic impression on chemistry identity were lower than males, and the differences were statistically significant. Nevertheless, females’ interest in chemistry was slightly lower than males’, but the difference was not significant. In addition, females’ willingness to choose a chemistry-related career in the future was less likely than males.
Mean | β | t | df | p | ||
---|---|---|---|---|---|---|
Female | Male | |||||
Notes. CLA chemistry learning anxiety, CTA chemistry test anxiety, Com/Per competence/performance, INT interest, E-Rec external recognition, HICI holistic impression on chemistry identity, Car-Cho chemistry career choice. | ||||||
CLA | 2.27 | 2.05 | 0.13 | 3.15 | 577.00 | <0.001 |
CTA | 3.22 | 2.66 | 0.27 | 6.68 | 577.00 | <0.001 |
Com/Per | 2.88 | 3.29 | 0.25 | −6.21 | 577.00 | <0.001 |
E-Rec | 1.90 | 2.31 | 0.22 | −5.44 | 577.00 | <0.001 |
INT | 3.55 | 3.67 | 0.07 | −1.58 | 537.91 | 0.116 |
HICI | 2.12 | 2.57 | 0.21 | −5.03 | 577.00 | <0.001 |
Car-Cho | 2.42 | 2.75 | 0.15 | −3.72 | 577.00 | <0.001 |
Model | χ 2/df | TLI | CFI | RMSEA | SRMR |
---|---|---|---|---|---|
Theoretical model | |||||
All students | 5.956 | 0.939 | 0.997 | 0.093 | 0.016 |
Female | 5.848 | 0.884 | 0.994 | 0.121 | 0.021 |
Male | 1.721 | 0.980 | 0.999 | 0.054 | 0.013 |
Revised model | |||||
All students | 2.059 | 0.987 | 0.998 | 0.043 | 0.016 |
Female | 2.054 | 0.975 | 0.996 | 0.057 | 0.022 |
Male | 1.386 | 0.989 | 0.998 | 0.039 | 0.018 |
![]() | ||
Fig. 3 Revised model regarding relationships between chemistry anxiety, chemistry identity, and chemistry career choice. |
Model | χ 2 | DF | χ 2/df | TLI | CFI | RMSEA | SRMR |
---|---|---|---|---|---|---|---|
Unconstrained (A) | 10.320 | 6.000 | 1.720 | 0.981 | 0.997 | 0.035 | 0.215 |
Structural weights (B) | 30.700 | 23.000 | 1.335 | 0.991 | 0.995 | 0.024 | 0.271 |
Structural covariances (C) | 45.335 | 26.000 | 1.744 | 0.981 | 0.988 | 0.036 | 0.307 |
Structural residuals (D) | 53.250 | 31.000 | 1.718 | 0.982 | 0.986 | 0.035 | 0.293 |
Model | Δχ2 | ΔDF | Δχ2p-value | ΔCFI |
---|---|---|---|---|
Notes. Δχ2, Δdf, ΔTLI, ΔCFI, and ΔGFI represents the difference between constrained models and the unconstrained model in chi-square, df, Tucker–Lewis fit index (TLI), comparative fit index (CFI), goodness-of-fit index (GFI). Δχ2p-value refers to the significance of the difference between constrained models and the unconstrained model in chi-square. | ||||
Structural weights | 20.380 | 17.000 | 0.255 | −0.002 |
Structural covariances | 35.015 | 20.000 | 0.020 | −0.009 |
Structural residuals | 42.930 | 25.000 | 0.014 | −0.011 |
Table 7 further presented the variation between the constrained and unconstrained models. As indicated in Table 7, compared to the unconstrained model (Model A), the constrained models (Model B) with equal structural coefficients had insignificant changes in chi-square (Δχ2p-value = 0.255 > 0.05). This suggested that the path model works for both males and females. However, after adding the restriction of equal structural covariance and the structural weights, Model C changed significantly compared to model A (Δχ2p-value = 0.020 < 0.05). This indicated that the relationship between different variables is not identical for females and males. After further adding the condition of equal variance of structural residual variables in model C, model D also changed significantly compared to model A (Δχ2p-value = 0.014 < 0.05). Accordingly, we used the present accepted unconstrained model to further explore which direct or indirect effects have gender differences (Wu, 2018).
![]() | ||
Fig. 4 Structural relationships between variables in females. Notes: *p < 0.05, **p < 0.01, ***p < 0.001. |
![]() | ||
Fig. 5 Structural relationships between variables in males. Notes: *p < 0.05, **p < 0.01, ***p < 0.001. |
Paths | Standardized β | SE | Bootstrap 95% CI | p | |
---|---|---|---|---|---|
Lower | Upper | ||||
Notes. CLA chemistry learning anxiety, CTA chemistry test anxiety, Com/Per competence/performance, INT interest, E-Rec external recognition, HICI holistic impression on chemistry identity, Car-Cho chemistry career choice. | |||||
CLA → INT → HICI → Car-Cho | −0.007 | 0.005 | −0.022 | −0.001 | 0.014 |
CLA → Com/Per → Car-Cho | −0.089 | 0.030 | −0.158 | −0.039 | 0.001 |
CLA → Com/Per → HCSI → Car-Cho | −0.019 | 0.009 | −0.043 | −0.005 | 0.006 |
CLA → Com/Per → INT → HCSI → Car-Cho | −0.010 | 0.005 | −0.023 | −0.003 | 0.003 |
CLA → Com/Per → E-Rec → HCSI → Car-Cho | −0.041 | 0.012 | −0.071 | −0.022 | 0.000 |
CTA → Com/Per → Car-Cho | −0.039 | 0.018 | −0.084 | −0.012 | 0.002 |
CTA → Com/Per → HCSI → Car-Cho | −0.008 | 0.005 | −0.022 | −0.002 | 0.005 |
CTA → Com/Per → INT → HCSI → Car-Cho | −0.004 | 0.003 | −0.012 | −0.001 | 0.004 |
CTA → Com/Per → E-Rec → HCSI → Car-Cho | −0.017 | 0.007 | −0.037 | −0.006 | 0.001 |
Paths | Standardized β | SE | Bootstrap 95% CI | p | |
---|---|---|---|---|---|
Lower | Upper | ||||
Notes. CLA chemistry learning anxiety, CTA chemistry test anxiety, Com/Per competence/performance, INT interest, E-Rec external recognition, HICI holistic impression on chemistry identity, Car-Cho chemistry career choice. | |||||
CLA → Com/Per → INT → Car-Cho | −0.049 | 0.022 | −0.108 | −0.018 | 0.001 |
CLA → Com/Per → INT → HCSI → Car-Cho | −0.012 | 0.008 | −0.036 | −0.002 | 0.005 |
CLA → Com/Per →E-Rec → Car-Cho | −0.050 | 0.019 | −0.100 | −0.021 | 0.000 |
CLA → Com/Per → E-Rec → HCSI → Car-Cho | −0.026 | 0.014 | −0.064 | −0.006 | 0.006 |
CTA → E-Rec → Car-Cho | −0.073 | 0.032 | −0.153 | −0.023 | 0.003 |
CTA → E-Rec → HCSI → Car-Cho | −0.037 | 0.02 | −0.091 | −0.008 | 0.007 |
CTA → Com/Per → E-Rec → Car-Cho | −0.034 | 0.015 | −0.074 | −0.011 | 0.003 |
CTA → Com/Per → E-Rec → HCSI → Car-Cho | −0.017 | 0.01 | −0.048 | −0.004 | 0.007 |
CTA → Com/Per → INT → Car-Cho | −0.033 | 0.017 | −0.082 | −0.008 | 0.003 |
CTA → Com/Per → INT → HCSI → Car-Cho | −0.008 | 0.006 | −0.027 | −0.002 | 0.006 |
As for females, chemistry anxiety influenced chemistry career choice through nine indirect pathways, with a total effect of −0.234. The sum of significant indirect effects from chemistry learning anxiety on females’ career choices was −0.166; the total significant indirect effects from chemistry test anxiety on females’ career choices was −0.068. With regard to males, chemistry anxiety influenced chemistry career choice through ten indirect pathways, with a total effect of −0.339. The sum of significant indirect effects of chemistry learning anxiety on males’ career choices was −0.137. The sum of significant indirect effects of chemistry test anxiety on males’ career choices was −0.202.
Specifically, the effect of competence/performance beliefs on chemistry-related career choices was significant in the female group but not in the male group. This result can be partially supported by the study of Bubić and Ivanišević (2016), which showed that adolescents’ career self-efficacy (i.e., beliefs about completing the necessary tasks to achieve desired career outcomes) significantly predicted career decisions only for females, but not for males.
Interest and recognition, however, only significantly predict males’ career choices. According to previous studies, because of gender bias, discrimination, and lack of inclusion for females in STEM fields, females do not choose careers related to STEM fields even if they have a strong interest in them (Cardador et al., 2020). Alternatively, females have the ability and tendency to focus on multiple goals at the same time, as opposed to males who typically focus on one career choice (Maines, 1983; Eccles, 2009). In contrast, higher competence/performance beliefs usually mean that less time or effort may be required at work and they can devote enough time to caring for their families. Therefore, females are more likely to choose their career goals based on their ability/performance beliefs rather than their interests. In addition, the results of the relationship between external recognition and chemistry career choice could be corroborated by the investigation of Lee et al. (2020), which reported a significant parental influence on STEM career aspirations of adolescent males, but a smaller and non-significant influence on females.
A direct and significant correlation between chemistry anxiety and chemistry career choice was not found in the present study. This finding is also in line with a study that there was no significant direct correlation between math anxiety and career choice (Cribbs et al., 2021). Regarding the relationships between chemistry anxiety and chemistry identity, the current research found that female and male students’ higher chemistry learning and test anxiety were linked with lower competence/performance beliefs, which are in line with Li et al.'s (2021) investigation. Chemistry test anxiety was not found to be significantly associated with interest in either females or males, which is consistent with previous research (Lohbeck et al., 2016). Chemistry learning anxiety significantly predicted interest only for females. This is similar to the findings of Huang et al. (2018) that math anxiety was associated with science career interests for females, but not for males. In addition, the present study found that chemistry test anxiety was significantly associated with external recognition for males only.
Based on the fact that chemistry anxiety was significantly correlated with chemistry identity, while chemistry identity was a significant predictor of students’ career choices, the present study further verified the fully mediating role of chemistry identity between chemistry anxiety and chemistry career choice. This finding supports the argument of social cognitive theory and social cognitive career theory (SCCT) that the relationship between anxiety and career choice may be mediated by self-efficacy, outcome expectations, and interest, which overlap with identity (Aydin et al., 2011; Mozahem, 2020; Luo et al., 2021).
Considering the gender differences in the direct correlations, the current research further found similarities and differences in indirect pathways and effect sizes from chemistry anxiety to chemistry career choice (see Tables 8 and 9). From the results of the mediation analysis, we found an interesting phenomenon is that the largest effects of either chemistry learning anxiety or chemistry test anxiety on career choices for females were mediated by competence/performance beliefs. These results suggest that females’ competence/performance beliefs play a relatively important mediating role between chemistry anxiety and chemistry career choice. One possible explanation is that females are less dependent on others’ evaluations (Mustafa Alpaslan, 2019), so their career choices more rely on their perceptions of chemistry learning rather than chemistry tests. In turn, the higher the anxiety experienced in learning chemistry, the more negative self-perception the individual will hold about their chemistry competencies and chemistry performance, ultimately leading to a stronger avoidance of chemistry learning and chemistry-related careers (Hembree, 1990; Ashcraft, 2002).
Nevertheless, the greatest indirect effect of chemistry test anxiety on chemistry career choice was mediated by external recognition of males. In addition, chemistry learning anxiety had a greater impact on males' career choices mainly through two chain mediating paths (competence/performance beliefs and external recognition, as well as competence/performance beliefs and interests). The above results suggest that chemistry anxiety is more likely to influence chemistry career choice through the mediating effect of external recognition in males than in females. This may be because the traditional impression is that males have higher talent in science than females and should perform better (Tenenbaum and Leaper, 2003). Therefore, when males hold a higher level of chemistry anxiety, they may believe that they fall short of others’ expectations and thus have difficulty perceiving others’ recognition and feel less confident in themselves.
In the classroom, teachers may be able to reduce students’ learning anxiety and increase their competence/performance beliefs through collaborative active learning. On the one hand, active learning satisfies the students’ need for autonomy in the learning process (Daniel, 2016). In such a learning environment, students are more likely to perceive their value in science learning, which further can lead to higher interest and competence beliefs (Cicuto and Torres, 2016; Hendrickson, 2021). Thus, active learning contributes to improving students’ chemistry identity. On the other hand, for high school students, active learning is more difficult than passively receiving knowledge, it may trigger students’ learning anxiety. According to the results of this study, elevated chemical anxiety affects students’ career aspirations by reducing their chemistry identity. Therefore, we need to integrate cooperative learning into active learning, so that students could create interdependent social cohesion by helping each other (Johnson and Maruyama, 1983; Slavin, 2015), which helps reduce students’ learning anxiety (Daniel and Awokoya, 2010). Also, the mutual encouragement among peers and the feeling of being needed by others is crucial to alleviate students’ anxiety and boost their self-confidence (Downing et al., 2020). Thus, teachers should be careful to guide students to respect and praise the efforts of others in the cooperative learning process.
In addition, the present study provides evidence that females’ interest in chemistry was not significantly different from males, but the effect of interest on females’ chemistry career choice intention was not as significant as that of males. In contrast, females’ competence/performance beliefs had a stronger impact on their chemistry career choice intention, but females’ competence/performance beliefs were significantly lower than males. At the same time, learning anxiety has a greater impact on females’ competence/performance beliefs and interests than test anxiety. This suggests that it is crucial to reduce females’ anxiety and enhance their competence/performance beliefs through appropriate strategies during the learning process. In other words, it is more important for females to have positive experiences and feelings about their chemistry learning process rather than chemistry test results. Therefore, teachers should pay special attention to giving positive feedback and evaluating female performance during the learning process so that they can enhance females’ competence/performance beliefs through emotional support (Lou and Noels, 2020). Besides, teachers could help students set appropriate and stage-based learning goals. On the one hand, a clear and explicit learning direction is of benefit to reduce students’ learning anxiety (Law et al., 2010). On the other hand, breaking down a difficult learning goal into stage-based and progressive learning goals allows students to gain confidence through stage-based success, further boosting their competence/performance beliefs to complete more difficult tasks (Chang et al., 2022).
It is important to note that the above results do not imply that enhancing females’ interest and males' competence/performance beliefs are not important. Because females’ interest affects their career choice by enhancing their overall perception of identity, males’ competence/performance beliefs are favorable for enhancing interest and internalizing the external recognition into self-recognition.
Second, the measurements in this study were primarily derived from students’ self-reports, which may result in some errors. More diverse and objective measurement methods need to be considered in future studies.
Third, this study has a limited explanation for the differences between males and females in the model. Our interpretation is based on some inferences drawn from previous studies. For example, we conjecture that gender differences in the relationship between gender and career choice are caused by gender bias. However, ours did not investigate whether subjects were influenced by gender bias. Therefore, in future research, we need to further investigate the mechanisms that lead to gender differences through empirical studies.
Forth, this study used a single item for the survey of chemistry-related career interests. However, high school students may not have a complete and deep understanding of chemistry-related careers (Avargil et al., 2020). They may think of chemistry-related careers as chemists in the laboratory (Solano et al., 2011; Avargil et al., 2020). This may be a factor contributing to students’ low interest in chemistry careers. Therefore, educators should pay attention to enhancing career education for high school students to avoid stereotypes about chemistry-related careers.
Path | β | SE | Bootstrap 95% CI | p | |
---|---|---|---|---|---|
Lower | Upper | ||||
CLA → Com/Per | −0.367 | 0.057 | −0.475 | −0.252 | 0.000 |
CTA → Com/Per | −0.182 | 0.058 | −0.295 | −0.066 | 0.002 |
Com/Per → INT | 0.541 | 0.051 | 0.434 | 0.635 | 0.000 |
Com/Per → E-Rec | 0.419 | 0.046 | 0.324 | 0.507 | 0.000 |
CLA → INT | −0.143 | 0.064 | −0.271 | −0.019 | 0.022 |
CLA → E-Rec | −0.053 | 0.056 | −0.165 | 0.054 | 0.339 |
CTA → E-Rec | −0.099 | 0.058 | −0.211 | 0.016 | 0.101 |
CTA → INT | 0.032 | 0.052 | −0.073 | 0.132 | 0.543 |
INT → HCSI | 0.122 | 0.043 | 0.035 | 0.203 | 0.008 |
E-Rec → HCSI | 0.635 | 0.043 | 0.549 | 0.713 | 0.000 |
Com/Per → HCSI | 0.124 | 0.050 | 0.026 | 0.223 | 0.011 |
CLA → HCSI | −0.036 | 0.042 | −0.121 | 0.045 | 0.386 |
CTA → HCSI | 0.035 | 0.043 | −0.047 | 0.122 | 0.402 |
E-Rec → Car-Cho | 0.108 | 0.071 | −0.039 | 0.247 | 0.145 |
INT → Car-Cho | 0.098 | 0.053 | −0.005 | 0.202 | 0.065 |
HISI → Car-Cho | 0.351 | 0.072 | 0.212 | 0.495 | 0.000 |
Com/Per → Car-Cho | 0.207 | 0.062 | 0.084 | 0.326 | 0.001 |
Path | β | SE | Bootstrap 95% CI | p | |
---|---|---|---|---|---|
Lower | Upper | ||||
CLA → Com/Per | −0.272 | 0.075 | −0.418 | −0.125 | 0.001 |
CTA → Com/Per | −0.223 | 0.078 | −0.373 | −0.067 | 0.006 |
Com/Per → INT | 0.634 | 0.052 | 0.523 | 0.731 | 0.000 |
Com/Per → E-Rec | 0.428 | 0.061 | 0.303 | 0.545 | 0.000 |
CLA → INT | −0.037 | 0.061 | −0.158 | 0.082 | 0.546 |
CLA → E-Rec | 0.032 | 0.074 | −0.111 | 0.179 | 0.651 |
CTA → E-Rec | −0.204 | 0.073 | −0.347 | −0.058 | 0.005 |
CTA → INT | 0.014 | 0.064 | −0.109 | 0.141 | 0.810 |
INT → HCSI | 0.196 | 0.060 | 0.085 | 0.316 | 0.001 |
E-Rec → HCSI | 0.619 | 0.061 | 0.490 | 0.727 | 0.001 |
Com/Per → HCSI | −0.013 | 0.081 | −0.167 | 0.150 | 0.905 |
CLA → HCSI | −0.040 | 0.058 | −0.153 | 0.076 | 0.484 |
CTA → HCSI | −0.095 | 0.062 | −0.219 | 0.027 | 0.124 |
E-Rec → Car-Cho | 0.330 | 0.074 | 0.187 | 0.479 | 0.000 |
INT → Car-Cho | 0.217 | 0.074 | 0.079 | 0.368 | 0.001 |
HISI → Car-Cho | 0.270 | 0.101 | 0.060 | 0.464 | 0.012 |
Com/Per → Car-Cho | −0.006 | 0.094 | −0.192 | 0.174 | 0.943 |
This journal is © The Royal Society of Chemistry 2022 |