Looking at Our STEM Teacher Workforce: How to Model Self-Efficacy

Abstract

There is currently a severe shortage of teachers in the U.S. workforce. The problem is especially acute among science, technology, engineering, and mathematics (STEM) teachers and exacerbated by high turnover among new teachers—those with less than 5 years of teaching experience. In this article, the authors investigate one piece of the puzzle. The authors model a social cognitive approach to understanding self-efficacy, a key precursor to job performance and retention. Their interactionist approach accounts for both demographic (i.e., gender and age) and relational variables (i.e., social networks). The authors test their ideas on a sample of 159 STEM teachers across five geographic regions in the United States. Their analysis reveals patterned differences in self-efficacy across gender that are contingent on the communities of practice in which the teachers are embedded. Together, their theory and findings highlight the value of taking a holistic, interactionist view in explaining STEM teacher self-efficacy.

Keywords

STEM teachers gender age social networks self-efficacy

Right now, there is a severe shortage of science, technology, engineering, and mathematics (STEM) teachers in the U.S. workforce that appears to be worsening. As shown in Figure 1 (Sutcher et al., 2016), the projected shortage is expected to widen considerably over the next 5 years. This is a serious problem for one of the largest and most important occupational groups in the United States. Instability in a school’s teacher workforce negatively affects student achievement, diminishes teacher effectiveness and quality, and consumes economic resources that could be deployed elsewhere (Garcia & Weiss, 2019; Ronfeldt et al., 2013). Filling a vacancy costs $21,000 a year, on average, costing an estimated $8 billion a year (Garcia &Weiss, 2019). This shortage is exacerbated by high turnover among new teachers and is particularly salient in high-needs districts. Teachers are less likely to leave their jobs if they are confident in their abilities to perform well in these schools (Boyd et al., 2005; Boyd et al., 2011; Ingersoll & Strong, 2011; Pedota, 2015; Rodriguez, 2019).

Figure 1.

Teacher shortage.

In this article, we aim to advance knowledge on the social conditions that facilitate such confidence. We address the following research question: What factors contribute to perceived professional mastery, measured by self-efficacy, for STEM teachers? In doing so, we model a social cognitive approach that takes into account both demographic (i.e., gender and age) and relational variables (i.e., social networks). We test our ideas on a sample of 159 STEM teachers across five geographic regions. Our analysis reveals patterned differences in self-efficacy across gender that are contingent on the communities of practice in which the teachers are embedded. Our approach models an interactionist view of self-efficacy that accounts for the person and the environment. Accurately accounting for both is essential for good science and good policy.

Theory and Hypotheses

Social cognitive theory (SCT) is a paradigm within psychological research that explains human behavior as a function of three broad classes of sociocognitive influences (i.e., person, behavior, and environment). Unlike dualistic views of human nature that explain human behavior strictly in terms of reactions to the environment or innate drives of individuals, SCT contends that human behavior is best explained with a transactional perspective in which the person, environment, and behavior reciprocally influence one another within a unified and triadic causal structure (Bandura, 1986; see Figure 2). In this view, human behavior cannot be understood solely by looking at structural factors or psychological factors. People are both architects of and subjects to the social worlds they inhabit. Human action, exercised through agency, can shape the environment that in turns affects the person and subsequent behavior. Thus, SCT treats structural and psychological factors as interacting cofactors shaping human action.

Figure 2.

Triadic reciprocal determinism.

As such, perceived self-efficacy, defined as beliefs in one’s capabilities to organize and execute the courses of action required to produce given attainments, is a central explanatory element within the SCT framework (Bandura, 1997). Self-efficacy beliefs come from four sources: mastery experiences, social modeling, social persuasion, and physiological and affective states (Bandura, 2013). Mastery experiences, which is the doing of performances, are often the most influential source of efficacy information because they provide authentic information about one’s ability to perform a given task. Successful performances can strengthen self-efficacy, while failures can weaken it. Or, failures can result in more robust efforts to practice and correct performances to achieve success. Hence, enactive mastery experiences can serve as a training ground for the honing of skills to competently execute the duties of a given job. However, a tremendous power of the human intellect is its capacity to learn without direct experience. People can also learn complex skills through modeling, where they observe the actions of others and run simulations of that activity in their minds in a way that enhances their own confidence and capacity to execute the modeled attainment. Thus, social modeling, which operates through the tool of observational learning, also serves as a primary source of self-efficacy. Social persuasion affects self-efficacy beliefs through social influences processes. People can be persuaded verbally that they have what it takes to succeed, and such persuasion can indeed boost their self-efficacy. If significant others have confidence in one’s ability to accomplish a given task, one is more likely to adopt that self-affirming belief. In judging their capabilities, people also rely on physiological and affective states that convey somatic signs of the vulnerability to dysfunction in carrying out taxing or stressful tasks (Bandura, 1997, 2012).

These four sources of self-efficacy provide diagnostic information that is cognitively processed to determine the level of self-efficacy for a given performance. According to SCT and its triadic system of reciprocal causation, the relative weight given to diagnostic information from each of those sources and the manner and extent of integration between them is a function of the person and the environment.

In this study, we focus on two important characteristics of the person, gender and age, and an important environmental feature of the workplace, communities of practice, to explore individual differences in self-efficacy in the context of STEM teachers.

Gender and Self-Efficacy

Appraisals of self-efficacy are socially conditioned and influenced by social constructions of identity (Bandura, 1997). Gender, the social construction of sex that underlies notions of femininity and masculinity, is constantly created and recreated through social interaction, and is one of the fundamental ways that humans organize their lives (Justo et al., 2015; Lorber, 2014; Marlow & Patton, 2005; Oakley, 2016). Since organizations are relational entities, demographic similarities and differences between organizational members provide observable information that employees use in assessing their job satisfaction and other metrics of individual performance (Fields & Blum, 1997; Pfeffer, 2013). From this lens, there are strong forces of homophily that operate in the workplace, whereby people prefer to work with similar people along salient dimensions of personal identity (McPherson et al., 2001). The gender composition of a workforce, in particular, is likely to be an important expression of its social structure.

A higher proportion of women teachers coupled with ingrained gender stereotypes regarding gender-typic occupational matching may make male teachers identify less with the teaching profession than their female coworkers. This expectation is supported by recent evidence finding that men entering a stereotypically masculine occupation (e.g., marine commando recruits, surgical trainees) were found to be less motivated than those they perceived to be more masculine (Ellemers, 2018; Peters et al., 2015). Similarly, in the context of STEM teaching, we theorize that the weaker identification with the teaching profession among men and the stronger identification among women may result in a gender difference in self-efficacy in the STEM teacher context.

In addition, teaching is a “caring” profession, which places a premium on tenderness, nurturance, and a relational orientation (Basow, 2000; Cancian & Oliker, 2000; O’Connor, 2008). Male socialization patterns, which tend to emphasize emotional restrictiveness and stoicism (Levant, 2011; Riggs, 1997; Yarnell et al., 2015), may limit the range and depth of emotional resources male teachers can harness in their dealings with students. Since effective teaching takes place in the socioemotional management of a messy human relationship (Lowman, 1995), female teachers may have a relational advantage interacting with students in the classroom. This relational advantage is likely associated with a greater frequency of successful mastery experiences that heighten the self-efficacy beliefs of women relative to men in the STEM teacher context, possibly resulting from more effective classroom management and control strategies (Woolfolk et al., 1990).

In sum, occupational gender typing of occupationally appropriate norms, a higher proportion of women in the teacher workforce, and a greater (albeit, socially conditioned) relational orientation among female STEM teachers are likely to create variance in self-perceptions of professional mastery, measured through self-efficacy, across genders in the STEM teacher context. Thus, we hypothesize the following:

Hypothesis 1: Gender is associated with self-efficacy, such that female STEM teachers will report higher overall levels of teacher self-efficacy.

Moderating Role of Age

According to SCT, gender’s effect on self-efficacy is likely to depend on other characteristics of the individual. Of those characteristics, prior experience within the task domain, unsurprisingly, has most commonly been identified as a key antecedent to self-efficacy within that domain. Prior experience is directly related to the number and quality of mastery experiences that can enhance self-efficacy (Bandura, 1997; Boyd et al., 2005).

In this study, since our focus is on new STEM teachers, we extend this research by examining the related but distinct concept of a worker’s chronological age. While related to prior experience since older workers are often more likely to have more work experience within a given profession, we view age as a distinct sociological construct. In accordance with recent studies that take a more nuanced view of age (Baù et al., 2017; Lévesque & Minniti, 2011), it is not only used as an indicator of human capital, but also a reflection of membership in population cohorts socialized by their common experience of similar significant events and norms at important life stages.

From this generational lens, individuals raised within the same historical and sociocultural context share collective memories and norms that lead to a common generational consciousness, a set of shared attitudes, values, and behaviors (Joshi et al., 2010; Lyons & Kuron, 2014). Age, as an indicator of membership in these separate generational cohorts, is likely to be associated with different socialization trajectories that alter the relationship between gender and self-efficacy. In the context of STEM teaching, in particular, teachers’ age is likely associated with the degree of socialization into stereotypical gender roles and cultural sex-typing of occupational domains. Older teachers are more likely to exhibit propensities associated with gender role socialization and younger teachers are less likely to do so. Thus, we expect the gender gap in teaching self-efficacy to widen with age.

Hypothesis 2: Age moderates the effect of gender on self-efficacy, such that the gender gap in teacher self-efficacy widens with increasing age.

Moderating Role of Social Networks

Like all socially conditioned cognition, self-efficacy beliefs arise and are shaped in social interaction. People interact, through imposition or preference, with others in ways that determine the type of cognitive skills, behavioral competencies, and emotional propensities that are repeatedly observed (Bandura, 1997). SCT views cognition as a malleable and generative human capability that is deeply influenced by the social milieu.

We theorize that the effect of gender is moderated by teachers’ communities of practice, which are widely viewed as enablers of professional development and as a primary mechanism through which self-efficacy is enhanced (Cross et al., 2006). Formal and informal networks within organizations regulate knowledge, influence, and peer modeling that can shape an employee’s perceptions of professional mastery. Thus, in this study, we define a community of practice (CoP) as the overall ego network of professional interactions that revolve around a shared craft. The ego is the teacher, and the CoP is the ego network of professional relationships that revolve around teaching content and/or pedagogy.

We distinguish between two types of network properties: Network composition refers to the nature of ties (e.g., tie strength, informational flow) or characteristics of alters (i.e., colleague, mentor), while network structure refers to the overall structure of the ties (e.g., dense or open). In this study, we posit that network composition and structure are likely to have distinct moderating influences on the relationship between gender and self-efficacy. Furthermore, we posit that the composition and structure of STEM teachers’ CoP may have differential moderating effects on the relationship between gender and teacher self-efficacy.

With respect to network composition, we expect network strength, measured by the total number of role models, the overall frequency of interaction, the number of colleagues that contribute to feelings of teacher effectiveness, and/or the overall level of positive affect in a teachers’ CoP, as likely to positively moderate the relationship between gender and teacher self-efficacy such that women derive more value from their social ties. If our preceding logic is correct regarding the organizational demography of the workplace (i.e., overrepresentation of women), women are more likely to have a CoP characterized by more women. Similarity in gender between a teacher and her colleagues and mentors facilitates richer exchange, trust, and deeper social modeling (Ensher & Murphy, 1997). In contrast, male teachers, who are less likely to have an equivalent number of male colleagues, are more likely to have a CoP characterized by higher levels of gender heterophily. Thus, female teachers are more likely to derive more value from their strong ties, broadly defined, in enhancing self-efficacy than men, who have relatively less homophilous ties that serve as richer conduits for social persuasion.

Hypothesis 3: Network strength moderates the effect of gender on self-efficacy, such that the female teachers derive more benefits to self-efficacy from a one-unit increase in network strength than male teachers.

With respect to network structure, on the other hand, we theorize an opposite effect. Network density, which is the degree to which an ego’s alters are all connected, is associated with strong normative pressures due to the “audience effect” and increased monitoring of more embedded structures (Coleman, 1988). Dense CoPs, in particular, are likely to result in more advantageous information exchange, leading to higher levels of trust, cooperative norms, and reciprocity between a teacher and his or her CoP. These denser structures, characterized by higher levels of trust and a shared identity from stronger social norms, are likely to benefit male teachers more than women, who are more likely to already have a stronger identification with the profession. In other words, men are more likely to derive more self-efficacy gains from a one-unit increase in network density than women. We posit that dense networks make male teachers feel more part of the teacher community, making them more receptive to social persuasion from their CoPs that enhance self-efficacy.

Hypothesis 4: Network density moderates the effect of gender on self-efficacy, such that the male teachers derive more benefits to self-efficacy from a one-unit increase in network density than female teachers.

Method

Data Collection

The study sample was drawn from a pool of teachers with recent involvement in a teacher preparation program from five higher education institutions awarded a Robert Noyce Teacher Scholarship Program grant by the National Science Foundation. The Noyce programs are designed specifically for preparing teachers to work in high-needs districts that have more serious problems with recruitment and retention (Kirchhoff & Lawrenz, 2011). The pool of teachers in our sample completed Noyce programs that spanned institutions in the Midwest, Northeast, and Southeast United States.

We designed an online survey to capture teacher demographics, network characteristics, and self-efficacy. To collect the network variables, we use a name generator, followed by more detailed questions about the nature of each teacher’s professional contacts (see Borgatti & Ofem, 2010, for an overview of social network analysis). On the basis of Coburn and coauthors’ (2013) finding that interactions based on teaching expertise are a sustainable feature in teacher networks, we ask each teacher, “Who do you interact with on matters pertaining to teaching content and/or pedagogy?” This question creates the basis of a teacher’s CoP. In ego network terminology, the ego refers to the respondent and the alter refers to the contacts of ego.

For the network composition variables, respondents assess the extent to which each contact is a role model, contributes to feelings of teacher effectiveness, is energizing, and/or is frequently interacted with. Network structure (i.e., density) was measured with Mehra et al.’s (2014) Visual Network Scale (VNS). This is an efficient way to capture perceptions of network structure and provides a visual scale of connectedness in which the respondent approximates the overall density.

After crafting the survey, we tested the instrument with a small group of teachers at a participating institution that resulted in a few revisions to improve survey design and item clarity. We then distributed surveys through email to approximately 431 teachers who went through these various teacher preparation programs, which generated 166 responses for a completion rate of 38.5%. A total of 159 responses were used in the analysis due to missing data in seven survey responses.

Measures

Self-Efficacy

Self-efficacy is measured with the Teacher Efficacy Belief System–Self (TEBS-Self), a 31-item instrument that assesses “teachers’ individual beliefs about their own abilities to perform specific teaching and learning related tasks within the context of their own classrooms” (Dellinger et al., 2008, p. 751). Since our interest is in the perception of teachers regarding their perceived professional mastery, which includes the overall task domain of teaching, we use a unidimensional operationalization of TEBS-Self. This is consistent with the best practice of specificity matching, whereby self-efficacy is operationalized at the level of the particular task domain under investigation (Bandura, 2006; Rigotti et al., 2008). Cronbach’s α for this measure is .96, indicating a high degree of reliability.

Gender

Gender is measured dichotomously, with 0 representing female and 1 representing male.

Age

Age is measured in years as a count variable of the length of life.

Network Composition

Network composition is operationalized through the following four dimensions:

Role models is the total number of alters identified as role model.

Frequency of interaction is the total level of interaction with alters.

Effectiveness as a teacher is the total degree to which the alters contribute to ego’s feelings of effectiveness as a teacher.

Positive affect refers to the total degree of positive affect, measured by the degree to which alters are evaluated as “energizing” on a 5-point Likert-type scale.

Each of these measures was collected after the name generator portion of the survey. After respondents defined their set of alters with whom they interact with on matters of teaching content and/or pedagogy (i.e., the name generator), they then provided more detailed information on the nature of each of their relationships (e.g., role models, frequency of interaction, contribution to teacher effectiveness, energizing) with that initial set of identified alters.

Network Density

We measured network density using Mehra et al.’s (2014) VNS. This methodological innovation in social network measurement makes it more efficient, and less burdensome on respondents, to collect network data on perceived ego network structure. It consists of stylized depictions of network properties. In our case, we used a scale that depicted increasing levels of network density, where respondents saw images of ego networks of increasing connectedness and chose the one that best approximates their network. Figure 3 depicts the scale we used.

Figure 3.

Visual Network Scale.

Controls

Career Changer

Career changer refers to whether the respondent had a previous career prior to becoming a teacher. Prior professional experience, due to the competencies acquired, could translate into higher self-efficacy in the teacher context.

Prior Teaching Experience

Prior teaching experience is a count measure of the total number of years teaching prior to the current school year. Prior experience is directly related to enactive mastery experiences that can boost self-efficacy.

High Needs

Since the teaching self-efficacy is likely contingent on the challenges in the classroom, we control for teachers working in “high needs” districts. These districts pose additional challenges since students are more likely to deal with social stressors that might interfere with learning and the efforts of the teachers.

Site

To control for fixed effects across the five sites that provided teacher training, we created four dummy variables to represent the four universities. This controls for differences in teacher preparation, geography, and other unobservable fixed effects.

Analysis

Our analysis begins by calculating descriptive statistics and bivariate correlations between all the study variables. We examined the normality of the variable distributions to ensure parametric statistics could be employed. Then, we conducted multiple ordinary least squares (OLS) regression to test the four hypotheses. To model the interactions within the multiple regression framework (Aiken & West, 1991), we created interaction terms for gender and age, gender and network composition, and gender and network density, respectively. For the interactions between gender and network composition, we ran four separate models that use a different operationalization of network composition. Our model specification is shown below.

\begin{array}{l} Self-efficacy = controls + β_{1} x_{1} + β_{2} x_{2} + β_{3} x_{3} + β_{4} x_{4} \\ + β_{5} (x_{1} * x_{2}) + β_{6} (x_{1} * x_{3}) + β_{7} (x_{1} * x_{4}) + ε_{1} \end{array}

In this specification, controls are the set of control variables, β is the regression coefficient, x₁ is gender, x₂ is age, x₃ is network composition, x₄ is network density, and ε₁ is the error term. In the four separate models of the main analysis, x₃ is operationalized using the four different measures of network composition (i.e., total role models, total frequency of interaction, total effectiveness as a teacher, and total positive affect).

Results

Table 1 reports the descriptive statistics and bivariate correlations. Participants are predominantly early career teachers with 5 or less years of teaching completed at the time of the survey (87%) or greater than 5 years (13%); identify as science (78%) or mathematics (22%) teachers; spanned ages 20 to 65 years; were female (59%); and were White/non-Hispanic (94%). Seventy-three percent of the participants work in a “high needs” district. Regarding the correlations, there a few noteworthy observations. First, self-efficacy, as expected, has a positive and statistically significant correlation with prior experience (r = .28) and teacher identity (r = .54). Second, there is a very strong correlation between all the network composition variables (r > .79), justifying our decision to operationalize these separately in the regression models to avoid multicollinearity.

Table 1.

Descriptive Statistics and Bivariate Correlations.

		M	SD	1	2	3	4	5	6	7	8	9	10	11	12	13	14
1	Self-efficacy	58.10	17.20
2	Age	33.85	9.27	.06
3	Career changer	0.48	0.50	−.04	.58
4	Prior experience	3.40	3.49	.28	.19	−.08
5	High needs	0.74	0.44	−.14	−.17	−.03	.01
6	Site 1	0.18	0.38	−.22	−.27	−.12	−.15	−.08
7	Site 2	0.27	0.44	.01	.67	.49	.00	−.20	−.27
8	Site 3	0.17	0.38	.05	.07	.18	.28	.19	−.21	−.27
9	Site 4	0.19	0.39	.12	−.29	−.29	−.12	−.03	−.22	−.28	−.22
10	Gender	0.41	0.49	−.14	.03	−.04	.05	−.11	.04	−.01	−.06	.06
11	Role models	14.89	10.8	.14	−.09	.01	.27	.05	.05	−.14	.20	−.04	−.12
12	Frequency	33.67	19.98	.16	−.15	.01	.18	.07	.06	−.22	.22	.02	−.13	.86
13	Effectiveness	12.47	8.43	.12	−.13	−.06	.16	.09	.08	−.15	.11	.02	−.06	.85	.82
14	Positive affect	39.89	26.75	.20	−.11	.05	.28	.07	.06	−.18	.34	−.08	−.17	.88	.93	.79
15	Identity	0	1.03	.54	−.02	−.03	.13	−.01	−.16	.02	.04	.15	−.16	.21	.17	.16	.23

Note. p < .05 if |R| > .16.

Table 2 reports six models that test the direct and interactive effects of gender, age, and CoPs. Model 1 includes only the controls, Model 2 adds gender, and Models 3 through 6 use the four operationalizations of network composition. As Model 2 reveals, there is a strong gender effect on self-efficacy (β = −6.37; p < .01), supporting Hypothesis 1. However, the complete models, Models 3 through 6, indicate that gender is moderated by age (β = 0.65; p < .05), all four operationalizations of network composition (i.e., role models, frequency, effectiveness, and positive affect; β = −0.61, −0.31, −0.78, and −0.63, respectively; p < .05) and network density (b > 9.06; p < .01). Together, these results broadly support Hypotheses 1, 2, 3, and 4.

Table 2.

OLS Regression Results.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Age	−0.11 (0.23)	−0.08 (0.22)	0.36 (0.27)	0.39 (0.27)	0.35 (0.27)	0.39 (0.27)
Career changer	2.27 (3.69)	2.04 (3.64)	0.81 (3.54)	0.28 (3.55)	1.52 (3.53)	0.26 (3.55)
Prior experience	2.36*** (0.55)	2.43*** (0.54)	2.40*** (0.54)	2.26** (0.52	2.30*** (0.52)	2.24*** (0.54)
High needs	−6.14* (3.16)	−7.23* (3.12)	−5.81* (2.99)	−5.99* (2.98)	−5.88* (2.99)	−6.09* (2.97)
Site 1	−9.51* (4.28)	−9.07* (4.21)	−9.85* (4.06)	−9.97* (4.05)	−10.29* (4.07)	−10.39* (4.05)
Site 2	−2.76 (5.08)	−3.16 (4.99)	−5.67 (4.84)	−5.05 (4.81)	−5.90 (4.83)	−5.53 (4.79)
Site 3	−3.44 (4.63)	−3.86 (4.56)	−4.25 (4.37)	2.52 (3.94)	−4.52 (4.36)	−5.60 (4.44)
Site 4	3.61 (4.17)	4.15 (4.11)	2.43 (3.95)	2.53 (3.94)	2.62 (3.94)	2.37 (3.92)
Gender		−6.37** (2.59)	−4.99 (14.07)	−1.32 (14.39)	−3.97 (14.10)	−2.31 (14.17)
Network composition			0.22 (0.15)	0.18* (0.18)	0.34 (0.20)	0.15* (0.07)
Network density			−0.05 (2.08)	−0.36 (2.08)	−0.15 (2.07)	−0.53 (2.08)
Gender × age			−0.64* (0.27)	−0.66* (0.27)	−0.66* (0.27)	−0.65* (0.27)
Gender × network composition			−0.61* (0.26)	−0.31* (0.14)	−0.78* (0.32)	−0.26* (0.11)
Gender × network density			9.35** (3.33)	9.04** (3.30)	9.52** (3.34)	9.06** (3.28)
Constant	60.63 (7.61)	62.54 (7.53)	45.60 (11.06)	43.42 (11.19)	45.26 (11.04)	44.18 (11.04)
R ²	.19	.22	.32	.32	.32	.33
Adjusted R²	.14	.17	.25	.25	.25	.26

Note. N = 159. Value in parenthesis are standard errors. OLS = ordinary least square.

p < .05. ^**p < .01. ^***p < .001.

To better interpret the significant interaction, we create six visualization plots using factor notation and the margins command in Stata (StataCorp, 2015). The plots include 95% confidence intervals for each predicted value at the various levels of each covariate. This enables us to see the statistical uncertainty associated with the predicted values and provides a better sense of the effect size of each interaction. Figure 4 plots the interaction between gender and age, which shows that the gender gap in self-efficacy is larger for older teachers. This supports the notion that older teachers may be more likely to exhibit propensities associated with gender role socialization. Figure 5 shows four interaction plots between gender and the four operationalizations of network composition in predicting self-efficacy. These plots indicate that female teachers derive more positive returns to self-efficacy from a one-unit increase in their network composition (i.e., strong/positive relationships). Surprisingly, male teachers report lower self-efficacy as our measures of network composition increase. This finding corroborates the notion that female teachers, due to a greater likelihood of having ego networks characterized by homophily, may be more likely to experience deeper social modeling and efficacy gains from their networks. In contrast, male teachers, due to a greater likelihood of having ego networks characterized by heterophily, may not reap such gains. The interaction plot suggests that stronger ties for male teachers may result in less identification with the profession, lowering self-efficacy. However, Figure 6 indicates that male teachers experience higher self-efficacy gains from a one-unit increase in network density than female teachers. This means that when male teachers perceive their networks to be more cohesive, they reap higher gains to self-efficacy than female teachers, who may already have a stronger identification (i.e., sense of belonging) with the profession.

Figure 4.

Interaction between gender and age.

Figure 5.

Interaction between gender and network composition.

Figure 6.

Interaction between gender and density.

Discussion

In this study, we direct attention to our STEM teacher workforce. Since turnover is especially high among new teachers, we model a social cognitive approach to investigating teacher self-efficacy, a key precursor to job performance and retention of new STEM teachers in the workforce. We document a gender effect whereby women report higher self-efficacy than men. We explain this observation using workforce composition logic, occupational gender typing, and socialized gender differences in relational competencies. However, we find this effect is contingent on the teachers’ age and features of their CoP. Our findings indicate that female teachers generally derive more value from age and the compositional properties of their CoP than male teachers. However, we find that the gender gap in self-efficacy narrows when accounting for the structural properties of their CoP. A dense CoP, which may impose stronger norms and a shared sense of teacher identity, appears to moderate the relationship between gender and self-efficacy, such that male teachers derive more value from closed structures.

Our study offers valuable midrange theory for educational institutions that employ teachers. We demonstrate that SCT and the social network perspective are useful complementary perspectives to understand gender effects on teacher self-efficacy in the context of STEM high school teachers, an organizational population of critical importance to workforce training and local economic development, particularly in high-need areas where there is a severe shortage of qualified science teachers. The majority of our sample of teachers work in high-needs districts, so our findings are particularly relevant for administrators and teachers working in economically disadvantaged districts. The overall effectiveness of the national teacher workforce depends on teacher self-efficacy, and our study highlights its salient, socially conditioned sources. In addition, quality instruction requires gender diversity, which is why it is important to better understand the sources of efficacious teaching across the gender divide (Hansen et al., 2018). Thus, we contribute to the sparse work that provides explanatory models of teacher self-efficacy by accounting for teacher characteristics, particularly gender, age, and CoP (Klassen & Chiu, 2010; Siciliano, 2016; Tschannen-Moran & Hoy, 2007).

Our theory and findings also provide promising avenues of inquiry to explore similar questions in other occupational sectors, particularly those under explored in empirical workforce development research. Occupational gender typing with respect to scientific, technological, and mathematical fields is a constraining social pressure that limits the talent, opportunity set, and career trajectories of women. Our findings add nuance to the literature since they suggest that female STEM teachers, who serve as key modeling influences and as a source of scientific efficacy for female students, tend to report higher efficacious than male teachers, an opposite result in prior studies conducted in scientific vocations (Eccles, 1994; Klassen & Chiu, 2010; Scherer & Siddiq, 2015). We hope our work stimulates additional inquiry into gender research in other workforce domains. Knowledge of how to reduce the gender gap in occupational sectors characterized by a higher incidence of ingrained stereotypes will aid in their deconstruction.

Limitations and Future Research Directions

Our study has limitations that provide additional avenues of inquiry. First, we note that our study focuses on one particular occupation, STEM teaching, which may limit the generalizability of our findings. However, we think our theoretical and empirical focus is also a strength. Self-efficacy is shaped by contextual factors that are a function of the particular workforce, and we focus on a workforce that is instrumental in equipping the next generation of scientifically literate knowledge workers. Thus, a better understanding of the drivers of perceived professional mastery among STEM teachers cannot only result in better teacher outcomes but also result in better student outcomes that enhance career trajectories. Nonetheless, we encourage future research to use our approach to investigate the sources of self-efficacy in qualitatively distinct (e.g., nature of work, demographic composition) workforce populations.

Second, although rich and an advancement over prior studies that minimize the importance of professional networks on self-efficacy, the social network data used in our analysis do not include information on demographic characteristics of the alters. We had to make trade-offs between measurement and response burden and did not collect more detailed demographic information regarding each alter within a teacher’s CoP. While our logic is supported by the higher proportion of women in our sample (i.e., 59% female) and studies showing a greater representation of women in this profession (Polizzi et al., 2015; Rushton et al., 2014), we did not empirically measure the gender composition of the ego networks. Thus, we encourage future work to empirically test the theory we use to explain the gender disparity observed in teacher self-efficacy in our data set.

Managerial and Policy Implications

Despite the focus on one particular occupation, our findings have implications for workforce development more generally. Most important, we contend that SCT and its self-efficacy element should be more commonly leveraged as an analytical lens among leadership development practitioners. Bandura’s (1986) exposition of how individual behavior is a function of the person, behavior, and environment in a triadic system of reciprocal causation is the bedrock of some of our best psychological science (Bandura, 2018; Fiedler, 2018). Leaders should be equipped with the knowledge that SCT is one of social science’s most fertile and empirically validated frameworks for capturing psychological phenomena, including the drivers of perceived professional mastery, measured through self-efficacy, in the workplace (Bandura, 2012).

Knowledge of how social and workplace conditions affect self-efficacy can enable managers to more effectively and efficiently facilitate employee growth. More specifically, our study shows that socialization processes, occurring through gendering, generational, and social network effects, codetermine the self-efficacy of teachers in the STEM teacher context. This means that managers could benefit from deliberate efforts to create tailored communities of practice that fit the needs of the particular demographics of their workforce. For example, managers might incentivize the formation of denser CoP through the assignment of collaborative projects that reward more frequent interactions among select employees. Or, they might design job enrichment activities or other organizational interventions that enhance the self-efficacy of the workforce through constructive feedback (Parker, 1998). Ample evidence indicates that organizations that provide their new employees with guided mastery experiences, effective coworkers as models, and constructive performance feedback enhance employee’s self-efficacy, emotional well-being, and job satisfaction (Bandura, 2000).

As our findings reveal, self-efficacy is a malleable cognitive construct shaped by generational and social factors. This is actionable knowledge for managers because it means that theory and practice can be harnessed to strengthen self-efficacy in tailored ways across occupational domains. Professional mastery is within reach of all employees, and skilled managers create conditions that enable its grasp. For employees, knowledge of how socialization processes, occurring through gendering and generational conditioning, shape self-efficacy beliefs can be empowering because it can deflate the power of unwarranted self-doubt to constrain the actualization of professional potential. As we know from the burgeoning literature on psychological capital (Newman et al., 2014), organizations operate at their highest levels when their employees actualize their highest professional potentialities.

Regarding policy ideas for promoting retention in the new teacher workforce, in particular, we see promise in efforts aimed at facilitating a greater sense of belonging among new teachers with their schools and/or the teaching profession. These efforts could come in many forms, with content-specific induction and mentoring having strong research support (Luft et al., 2003). An exemplar we see in practice is the Minnesota STEM Teacher Induction Network (TIN; Roehrig et al., 2015). TIN is an online induction program that provides professional development and mentoring to beginning science and math teachers. Components of TIN include a mentor–mentee private chat area, small group learning communities to investigate case studies, reflective blogs/journals, action research through professional development inquiries, and critical reflections on videos of teaching through video annotation, with feedback from supervisors, mentors, and mentees. In a resource-constrained environment, creating virtual communities to support teachers is an efficient way to build communities of practice and reflective practice (Ellis et al., 2015; McFadden et al., 2014).

Conclusion

This article models a social cognitive approach to understanding professional mastery that highlights the importance of gender, age, and communities of practice, with implications for research on workforce induction, professional development, and retention. We focus on the STEM teacher workforce, in particular, since it is a population of critical importance to equipping the next generation of workers for the dynamic and rapidly evolving global economy, but one that faces a host of pressing challenges, including a teacher shortage, pervasive job dissatisfaction, and high turnover, problems largely attributable to working conditions within schools and districts (Garcia &Weiss, 2019; Ingersoll & Strong, 2011; Sutcher et al., 2016). Since self-efficacy is a key determinant of job satisfaction and retention (Bandura, 2000; Boyd et al., 2011), we advance knowledge on the individual, social, and interactive sources of variation in perceived professional mastery, measured through self-efficacy beliefs, that can inform the designing of managerial solutions to address these critical workforce challenges.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by National Science Foundation (NSF) Awards DUE-1035451, DUE-1660665, and DUE-1660736.

ORCID iDs

Brandon Ofem

Rebecca Konz

Author Biographies

Brandon Ofem is an assistant professor in the Global Leadership and Management Department in the College of Business at the University of Missouri–St. Louis. His research interests include interorganizational collaboration, social networks, entrepreneurship, and local economic development.

Samuel J. Polizzi is an assistant professor of chemistry in the School of STEM at Georgia Highlands College. His research interests include structure/function relationships, both at the organizational and biochemical levels.

Gregory T. Rushton is the director of the Tennessee STEM Education Center SEC and is a professor of chemistry at Middle Tennessee State University. His research interests include conceptual change in tertiary chemistry learning environments, science classroom discourse practices, policy reform in K-16 chemistry education, large-scale demographic analyses of K-12 STEM teaching populations, and science teacher leadership.

Michael Beeth is the director of alternative careers in teaching at the University of Wisconsin–Oshkosh. His research areas of interest include the role of the teacher in creating learning conditions necessary for students to engage in changing their misconceptions in science learning, and what, in addition to the science content, needs to change when students engage in learning for conceptual change.

Brock Couch is a graduate teaching assistant at Middle Tennessee State University. His research interests are ecology and evolution, biodiversity, conservation ecology, community ecology, and science education.

Jessica Doering is graduate research assistant at the University of Kentucky. Her research interests include trauma-informed education and the preschool-to-prison pipeline.

Rebecca Konz is a graduate research assistant at the University of Minnesota. Her research interests include experience-based projects, curriculum and content design, learning strategies, specialized learning materials, and STEM.

Margaret Mohr-Schroeder is a professor of STEM education at the University of Kentucky. Her research areas include informal learning in STEM education, preservice teacher preparation, and transdisciplinary professional development.

Gillian Roehrig is a professor of science education at the University of Minnesota. Her research and teaching interests are focused on understanding how teachers translate national and state standards into their classrooms. Of particular interest is how teachers, from preservice through induction and into the in-service years, implement inquiry-based teaching and how different induction and professional development programs can influence teachers’ knowledge, beliefs, and classroom practices.

Keith Sheppard is a core faculty member at Stony Brook University. His recent research has focused on the history of science education, particularly the late 19th and early 20th century development of science curriculum, exams, and laboratory work. Additional studies have examined student and teacher learning and understanding of various topics in chemistry. Current research includes studies of student access to science courses in New York State and the use of master teachers in teacher preparation.

References

Aiken

L. S.

West

S. G.

(1991). Multiple regression: Testing and interpreting interactions. Sage.

Bandura

(1986). Social foundations of thought and action. Prentice Hall.

Bandura

(1997). The exercise of control. Freeman.

Bandura

(2000). Cultivate self-efficacy for personal and organizational effectiveness. In Locke

(Ed.), The Blackwell handbook of principles of organizational behavior (pp. 120-136). Blackwell.

Bandura

(2006). Guide for constructing self-efficacy scales. In Pajares

Urdan

T. C.

(Eds.), Self-efficacy beliefs of adolescents (pp. 307-337). Information Age. https://www.uky.edu/~eushe2/Bandura/BanduraGuide2006.pdf

Bandura

(2012). On the functional properties of perceived self-efficacy revisited. Journal of Management, 38(1), 9-44. https://doi.org/10.1177/0149206311410606

Bandura

(2013). Social cognitive theory. In Kessler

(Ed.), Encyclopedia of management theory (Vol. 1, pp. 711-715). Sage.

Bandura

(2018). Toward a psychology of human agency: Pathways and reflections. Perspectives on Psychological Science, 13(2), 130-136. https://doi.org/10.1177/1745691617699280

Basow

S. A.

(2000). Best and worst professors: Gender patterns in students’ choices. Sex Roles, 43(5-6), 407-417. https://doi.org/10.1023/A:1026655528055

10.

Baù

Sieger

Eddleston

K. A.

Chirico

(2017). Fail but try again? The effects of age, gender, and multiple-owner experience on failed entrepreneurs’ reentry. Entrepreneurship Theory and Practice, 41(6), 909-941. https://doi.org/10.1111/etap.12233

11.

Borgatti

S. P.

Ofem

(2010). Overview: Social network theory and analysis. In Daly

A. J.

(Ed.), Social network theory and educational change (pp. 17-31). Harvard Education Press.

12.

Boyd

Lankford

Loeb

Ronfeldt

Wyckoff

(2011). The role of teacher quality in retention and hiring: Using applications to transfer to uncover preferences of teachers and schools. Journal of Policy Analysis and Management, 30(1), 88-110. https://doi.org/10.1002/pam.20545

13.

Boyd

Lankford

Loeb

Wyckoff

(2005). Explaining the short careers of high-achieving teachers in schools with low-performing students. American Economic Review, 95(2), 166-171. https://doi.org/10.1257/000282805774669628

14.

Cancian

F. M.

Oliker

S. J.

(2000). Caring and gender. AltaMira Press.

15.

Coburn

C. E.

Mata

W. S.

Choi

(2013). The embeddedness of teachers’ social networks: Evidence from a study of mathematics reform. Sociology of Education, 86(4), 311-342. https://doi.org/10.1177/0038040713501147

16.

Coleman

J. S.

(1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95-S120. https://doi.org/10.1086/228943

17.

Cross

Laseter

Parker

Velasquez

(2006). Using social network analysis to improve communities of practice. California Management Review, 49(1), 32-60. https://doi.org/10.2307/41166370

18.

Dellinger

A. B.

Bobbett

J. J.

Olivier

D. F.

Ellett

C. D.

(2008). Measuring teachers’ self-efficacy beliefs: Development and use of the TEBS-Self. Teaching and Teacher Education, 24(3), 751-766. https://doi.org/10.1016/j.tate.2007.02.010

19.

Eccles

(1994). Understanding women’s educational and occupational choices. Psychology of Women Quarterly, 18(4), 585-609. https://doi.org/10.1111/j.1471-6402.1994.tb01049.x

20.

Ellemers

(2018). Gender stereotypes. Annual Review of Psychology, 69, 275-298. https://doi.org/10.1146/annurev-psych-122216-011719

21.

Ellis

McFadden

Anwar

Roehrig

(2015). Investigating the social interactions of beginning teachers using a video annotation tool. Contemporary Issues in Technology and Teacher Education, 15(3), 404-421.

22.

Ensher

E. A.

Murphy

S. E.

(1997). Effects of race, gender, perceived similarity, and contact on mentor relationships. Journal of Vocational Behavior, 50(3), 460-481. https://doi.org/10.1006/jvbe.1996.1547

23.

Fiedler

(2018). The creative cycle and the growth of psychological science. Perspectives on Psychological Science, 13(4), 433-438. https://doi.org/10.1177/1745691617745651

24.

Fields

D. L.

Blum

T. C.

(1997). Employee satisfaction in work groups with different gender composition. Journal of Organizational Behavior, 18(2), 181-196. https://doi.org/10.1002/(SICI)1099-1379(199703)18:2<181::AID-JOB799>3.0.CO;2-M

25.

Garcia

Weiss

(2019). US schools struggle to hire and retain teachers: The second report in “The perfect storm in the teacher labor market” Series. Economic Policy Institute. https://files.eric.ed.gov/fulltext/ED598209.pdf

26.

Hansen

Levesque

Valant

Quintero

(2018). The 2018 Brown Center report on American education: How well are American students learning? The Brookings Institution. https://www.brookings.edu/wp-content/uploads/2018/06/2018-Brown-Center-Report-on-American-Education_FINAL1.pdf

27.

Ingersoll

R. M.

Strong

(2011). The impact of induction and mentoring programs for beginning teachers: A critical review of the research. Review of Educational Research, 81(2), 201-233. https://doi.org/10.3102/0034654311403323

28.

Joshi

Dencker

J. C.

Franz

Martocchio

J. J.

(2010). Unpacking generational identities in organizations. Academy of Management Review, 35(3), 392-414. https://doi.org/10.5465/AMR.2010.51141800

29.

Justo

DeTienne

D. R.

Sieger

(2015). Failure or voluntary exit? Reassessing the female underperformance hypothesis. Journal of Business Venturing, 30(6), 775-792. https://doi.org/10.1016/j.jbusvent.2015.04.004

30.

Kirchhoff

Lawrenz

(2011). The use of grounded theory to investigate the role of teacher education on STEM teachers’ career paths in high-need schools. Journal of Teacher Education, 62(3), 246-259. https://doi.org/10.1177/0022487110397840

31.

Klassen

R. M.

Chiu

M. M.

(2010). Effects on teachers’ self-efficacy and job satisfaction: Teacher gender, years of experience, and job stress. Journal of Educational Psychology, 102(3), 741-756. https://doi.org/10.1037/a0019237

32.

Levant

R. F.

(2011). Research in the psychology of men and masculinity using the gender role strain paradigm as a framework. American Psychologist, 66(8), 765-776. https://doi.org/10.1037/a0025034

33.

Lévesque

Minniti

(2011). Age matters: How demographics influence aggregate entrepreneurship. Strategic Entrepreneurship Journal, 5(3), 269-284. https://doi.org/10.1002/sej.117

34.

Lorber

(2014). The social construction of gender. In Ore

(Ed.), The social construction of difference and inequality: Race, class, gender, and sexuality (pp. 112-119). McGraw-Hill Education.

35.

Lowman

(1995). Mastering the techniques of teaching (2nd ed.). Jossey-Bass.

36.

Luft

J. A.

Roehrig

G. H.

Patterson

N. C.

(2003). Contrasting landscapes: A comparison of the impact of different induction programs on beginning secondary science teachers’ practices and beliefs. Journal of Research in Science Teaching, 40(1), 77-97. https://doi.org/10.1002/tea.10061

37.

Lyons

Kuron

(2014). Generational differences in the workplace: A review of the evidence and directions for future research. Journal of Organizational Behavior, 35(S1), S139-S157. https://doi.org/10.1002/job.1913

38.

Marlow

Patton

(2005). All credit to men? Entrepreneurship, finance, and gender. Entrepreneurship Theory and Practice, 29(6), 717-735. https://doi.org/10.1111/j.1540-6520.2005.00105.x

39.

McFadden

Ellis

Anwar

Roehrig

(2014). Beginning science teachers’ use of a digital video annotation tool to promote reflective practices. Journal of Science Education and Technology, 23(3), 458-470. https://doi.org/10.1007/s10956-013-9476-2

40.

McPherson

Smith-Lovin

Cook

J. M.

(2001). Birds of a feather: Homophily in social networks. Annual Review of Sociology, 27(1), 415-444. https://doi.org/10.1146/annurev.soc.27.1.415

41.

Mehra

Borgatti

S. P.

Soltis

Floyd

Halgin

D. S.

Ofem

Lopez-Kidwell

(2014). Imaginary worlds: Using visual network scales to capture perceptions of social networks. In Brass

D. J.

Labianca

Mehra

Halgin

D. S.

Borgatti

S. P.

(Eds.), Contemporary perspectives on organizational social networks (pp. 315-336). Emerald Group.

42.

Newman

Ucbasaran

Zhu

F. E. I.

Hirst

(2014). Psychological capital: A review and synthesis. Journal of Organizational Behavior, 35(S1), S120-S138. https://doi.org/10.1002/job.1916

43.

Oakley

(2016). Sex, gender and society. Routledge. https://doi.org/10.4324/9781315243399

44.

O’Connor

K. E.

(2008). You choose to care: Teachers, emotions and professional identity. Teaching and Teacher Education, 24(1), 117-126. https://doi.org/10.1016/j.tate.2006.11.008

45.

Parker

S. K.

(1998). Enhancing role breadth self-efficacy: The roles of job enrichment and other organizational interventions. Journal of Applied Psychology, 83(6), 835-852. https://doi.org/10.1037/0021-9010.83.6.835

46.

Pedota

(2015). How can student success support teacher self-efficacy and retention? Clearing House: A Journal of Educational Strategies, Issues and Ideas, 88(2), 54-61. https://doi.org/10.1080/00098655.2014.998600

47.

Peters

Ryan

M. K.

Haslam

S. A.

(2015). Marines, medics, and machismo: Lack of fit with masculine occupational stereotypes discourages men’s participation. British Journal of Psychology, 106(4), 635-655. https://doi.org/10.1111/bjop.12106

48.

Pfeffer

(2013). Organizational demography. In Kessler

(Ed.), Encyclopedia of management theory (Vol. 1, pp. 540-542). Sage.

49.

Polizzi

S. J.

Jaggernauth

Ray

H. E.

Callahan

Rushton

G. T.

(2015). Highly qualified or highly unqualified? A longitudinal study of America’s public high school biology teachers. BioScience, 65(8), 812-821. https://doi.org/10.1093/biosci/biv093

50.

Riggs

J. M.

(1997). Mandates for mothers and fathers: Perceptions of breadwinners and care givers. Sex Roles, 37(7-8), 565-580. https://doi.org/10.1023/A:1025611119822

51.

Rigotti

Schyns

Mohr

(2008). A short version of the occupational self-efficacy scale: Structural and construct validity across five countries. Journal of Career Assessment, 16(2), 238-255. https://doi.org/10.1177/1069072707305763

52.

Rodriguez

A. C.

(2019). Please don’t go: An examination of teacher retention in high needs schools (Publication No. 13895521) [Doctoral dissertation, California State University, Long Beach]. ProQuest Dissertations and Theses Global.

53.

Roehrig

G. H.

Donna

J. D.

Billington

B. L.

Hoelscher

(2015). Design of online induction programs to promote reform-based science and mathematics teaching. Teacher Education & Practice, 28(1-2), 286-303.

54.

Ronfeldt

Loeb

Wyckoff

(2013). How teacher turnover harms student achievement. American Educational Research Journal, 50(1), 4-36. https://doi.org/10.3102/0002831212463813

55.

Rushton

G. T.

Ray

H. E.

Criswell

B. A.

Polizzi

S. J.

Bearss

C. J.

Levelsmier

Kirchhoff

(2014). Stemming the diffusion of responsibility: A longitudinal case study of America’s chemistry teachers. Educational Researcher, 43(8), 390-403. https://doi.org/10.3102/0013189X14556341

56.

Scherer

Siddiq

(2015). Revisiting teachers’ computer self-efficacy: A differentiated view on gender differences. Computers in Human Behavior, 53(December), 48-57. https://doi.org/10.1016/j.chb.2015.06.038

57.

Siciliano

M. D.

(2016). It’s the quality not the quantity of ties that matters: Social networks and self-efficacy beliefs. American Educational Research Journal, 53(2), 227-262. https://doi.org/10.3102/0002831216629207

58.

StataCorp. (2015). Stata statistical software: Release 14. Author.

59.

Sutcher

Darling-Hammond

Carver-Thomas

(2016). A coming crisis in teaching? Teacher supply, demand, and shortages in the U.S. Learning Policy Institute. https://files.eric.ed.gov/fulltext/ED606666.pdf

60.

Tschannen-Moran

Hoy

A. W.

(2007). The differential antecedents of self-efficacy beliefs of novice and experienced teachers. Teaching and Teacher Education, 23(6), 944-956. https://doi.org/10.1016/j.tate.2006.05.003

61.

Woolfolk

A. E.

Rosoff

Hoy

W. K.

(1990). Teachers’ sense of efficacy and their beliefs about managing students. Teaching and Teacher Education, 6(2), 137-148. https://doi.org/10.1016/0742-051X(90)90031-Y

62.

Yarnell

L. M.

Stafford

R. E.

Neff

K. D.

Reilly

E. D.

Knox

M. C.

Mullarkey

(2015). Meta-analysis of gender differences in self-compassion. Self and Identity, 14(5), 499-520. https://doi.org/10.1080/15298868.2015.1029966