The development of student teachers’ teacher self-efficacy before and during the COVID-19 pandemic

Abstract

This study used latent growth curve models to examine the impact of the COVID-19 pandemic on the development of teacher self-efficacy in student teachers. Results indicated that the teacher self-efficacy of student teachers taught during the first COVID-19 lockdown increased significantly less across a semester compared to student teachers taught prior to the pandemic, who gained practical experience in schools. There may be a cohort of student teachers at risk of entering the profession with lower self-efficacy than is typical. Universities and schools may wish to provide additional practical experiences to compensate for the missed opportunities during the COVID-19 pandemic.

No potential conflict of interest was reported by the authors. Materials and analysis code for this study are available by email from the corresponding author. This study was not preregistered.

1. Introduction

Teacher self-efficacy has been linked to a number of important outcomes in both in-service (Lauermann & ten Hagen, 2021; Zee & Koomen, 2016) and student teachers (Scarparolo & Subban, 2021), including instructional quality, wellbeing, burnout, and commitment to the teaching profession. Early practical experience in schools is posited to have a positive impact on the development of teacher self-efficacy in student teachers, since it provides them with opportunities to experience success, known as ‘mastery experiences’ (Berg & Smith, 2018; Clark & Newberry, 2019; Pfitzner-Eden, 2016a). The COVID-19 pandemic, however, resulted in widespread school closures and, consequently, student teachers across the globe were unable to complete the practicums that typically form an important part of their training (Kidd & Murray, 2020). Germany, where the current study was conducted, was no exception, with schools and universities delivering their teaching either fully or partially online for almost 12 months (Frendl et al., 2021). The aim of the present study was to examine how the lack of practical experiences in authentic classrooms due to school closures during the first COVID-19 lockdown in Germany might have affected the development of student teachers' self-efficacy for instructional strategies, classroom management, and student engagement. To answer our research question, we focused on the development of teacher self-efficacy in student teachers taught during the first COVID-19 lockdown, and compared this with the development of teacher self-efficacy in student teachers taught in the preceding two semesters. In addition, we examined whether individual characteristics of student teachers moderated the expected differential development of teaching self-efficacy. Specifically, we focused on student teachers' gender, since female teachers typically report lower teacher self-efficacy than male teachers (Skaalvik & Skaalvik, 2007), as well as prior pedagogical experience, which has been shown to be a potentially protective factor in the development of teacher self-efficacy (Römer et al., 2018).

1.1. Teacher self-efficacy: definition and importance for student teachers

According to Social Cognitive Theory (Bandura, 1997), human behaviour shapes, and is shaped by, interactions between personal (e.g. beliefs, abilities) and social/environmental (e.g. reinforcement, feedback) factors. One personal factor that has drawn considerable attention for its influence on behaviour are self-efficacy beliefs. Self-efficacy beliefs are the expectations or perceptions that one can successfully perform a behaviour (Bandura, 1977). In the school context, self-efficacy of teachers has been conceptualized as ‘ … the teacher's belief in his or her capability to organize and execute courses of action required to successfully accomplish a specific teaching task in a particular context’ (Tschannen-Moran et al., 1998, p. 233). Researchers in the field typically distinguish between different components of teacher self-efficacy, the most common of which include self-efficacy for instructional strategies (e.g. perceived ability to use a range of strategies to increase and assess student understanding); self-efficacy for classroom management (e.g. perceived ability to manage student behaviour, especially disruptive behaviour); and self-efficacy for student engagement (e.g. perceived ability to motivate and engage students, particularly those who lack interest) (Tschannen-Moran & Hoy, 2001).

Studies indicate that both teachers' general (Holzberger et al., 2013; Praetorius et al., 2017) and task-specific teacher self-efficacy, including self-efficacy for instructional strategies (Lazarides & Schiefele, 2021), self-efficacy for classroom management (Dicke et al., 2014; Lazarides et al., 2020), and self-efficacy for student engagement (Fackler et al., 2021), positively relate to indicators of effective teaching in in-service teachers. Analysis of data from the 2018 Teaching and Learning International Survey (TALIS 2018; OECD, 2019) found that teachers reporting higher teacher self-efficacy also reported using higher quality teaching strategies (Holzberger & Prestele, 2021), which in turn could lead to greater student motivation and learning through their effects on students’ perceptions of teaching quality (Lauermann & Berger, 2021). In addition, relations between lower teacher self-efficacy and negative outcomes have been found in student teachers. For example, lower levels of self-efficacy for instructional strategies and classroom management have been linked to higher burnout during a teaching practical (Fives et al., 2007), and lower self-efficacy for classroom management has been found to predict less commitment to the teaching occupation, and a greater intention to quit, irrespective of the level of instruction (Klassen & Chiu, 2011). It therefore seems important to identify early sources of teacher self-efficacy to ensure that newly qualified teachers are entering the workforce best placed to deliver sustainable, high-quality teaching, not least because teacher self-efficacy is argued to be at its most malleable early on (Pfitzner-Eden, 2016b). Structured and supportive placements in schools may be particularly beneficial in this regard (Clark & Newberry, 2019).

1.2. Factors influencing the development of teacher self-efficacy in student teachers

The 2018 Teaching and Learning International Survey (TALIS 2018; OECD, 2019) reported positive associations between the teacher self-efficacy of in-service teachers and a range of personal and environmental factors, including working with more experienced teachers and professional collaboration. Opportunities to work with and learn from other teachers have also been linked to the development of teacher self-efficacy in student teachers (Tschannen-Moran et al., 1998; Woolfolk Hoy & Burke Spero, 2005), and practical experiences in schools are a key element of teacher education programmes throughout Europe and internationally (Clark & Newberry, 2019; Kidd & Murray, 2020). These practical experiences can vary in duration, intensity and focus, but all will typically involve students participating in ‘teaching’ activities such as observation, lesson planning, teaching and/or managing classroom behaviour (Clark et al., 2014). In addition, student teachers may also gain practical experiences in more informal ways as part of their wider training programme, such as working on case studies, receiving feedback from peers or their tutors, and group work.

The positive link between placements in authentic classrooms and the development of teacher self-efficacy has been supported by a range of naturalistic studies investigating the impact of intensive school practicums (e.g. practical experiences lasting a full semester or year) on the development of student teacher self-efficacy. For example, Klassen and Durksen (2014) followed 150 Canadian student teachers over the course of a practical semester, and found that general teacher self-efficacy increased significantly week on week. In a study involving American student teachers, Fives et al. (2007) reported increases in teacher self-efficacy for instructional strategies, classroom management and student engagement following a practicum semester. Furthermore, Pfitzner-Eden (2016b) compared teacher self-efficacy development during a university-based semester with a semester involving a school placement in a large sample of beginner student teachers in Germany. They found that self-efficacy for instructional strategies and student engagement did not significantly increase, whilst self-efficacy for classroom management significantly decreased during the university-based semester, whereas self-efficacy for instructional strategies, classroom management, and student engagement increased following the practicum semester, although the increase in self-efficacy for student engagement was not significant. Interestingly, the pattern of results was different for more advanced student teachers. For students coming to the end of their degree programme, self-efficacy for instructional strategies and classroom management significantly increased during the university-based semester, and only self-efficacy for classroom management significantly increased following the school placement. Berg and Smith (2018), however, reported a significant change in all three facets of teacher self-efficacy in a smaller sample of student teachers from New Zealand completing their final school practicum at the end of their three-year teaching programme. The nature of the school placement undertaken by student teachers has been linked to the extent to which teacher self-efficacy develops (Clark et al., 2014), offering a possible explanation for these contradictory findings. Taken together, the available evidence suggests that practical experiences in schools, particularly at the beginning of teacher training programmes, seems to have a clear positive influence on the development of student teachers' self-efficacy.

To understand how and why school practicums impact the development of teacher self-efficacy, we draw on the model proposed by Tschannen-Moran et al. (1998), which provides a comprehensive account of the sources, types, and consequences of teacher self-efficacy. In line with Social Cognitive Theory (Bandura, 1977), the model identifies four key influences on teacher self-efficacy: mastery experiences (situations in which an individual successfully performs a behaviour), vicarious experiences (opportunities for observing more experienced others), verbal persusion (receiving feedback on teaching ability from other teachers, mentors, and university staff), and physiological arousal (e.g. excitement or nervousness), with mastery experiences posited to have the strongest impact on the development of teacher self-efficacy. These assumptions have been supported empirically with both in-service teachers and student teachers. Quasi-experimental research suggests that for in-service teachers, professional development training that includes opportunities for mastery experiences (alongside vicarious experiences and verbal persuasion) have the greatest impact on general teaching self-efficacy (Tschannen-Moran & Mcmaster, 2009). Research involving student teachers has reported that whilst all four sources of teacher self-efficacy predicted changes in teacher self-efficacy during a practicum semester, mastery experiences explained the most variance in change scores (between 12% and 16%) (Pfitzner-Eden, 2016a). Similar findings have been reported for student teachers at the end of their teacher training, which included practical experience in schools (Clark & Newberry, 2019).

In summary, practical experiences in schools appears to have a positive impact on the development of teacher self-efficacy in student teachers, and this is likely due to the opportunities for mastery experiences it provides, alongside the other three sources of teacher self-efficacy. In addition, the research presented in this section suggests that self-efficacy for classroom management may be particularly influenced by a school practicum. This is possibly because student teachers may lack knowledge or awareness of classroom management strategies before undertaking a placement (Pfitzner-Eden, 2016b).

1.3. The potential impact of COVID-19-related school lockdowns on the development of teacher self-efficacy in student teachers

Measures introduced to control the spread of COVID-19 in March 2020 led to widespread school closures across the world including in Germany (König et al., 2020), where the current study takes place. Germany has a federal political system, and education and teacher training are the responsibility of individual states. In common across the states is the imbedding of practical experiences in teacher education programmes. In Brandenburg, the state in which this study was conducted, teacher training usually involves 3 years of bachelor's and 2 years of master's study, followed by a year-long placement in a school. Student teachers typically undertake three short school internships during the bachelor's phase of teacher training, and one in the master's phase, which is an intensive, 6-month school practicum at the end of the final year. The aim of these placements is to provide student teachers with the opportunity to apply the theoretical and pedagogical content from their university courses in practice, and to develop professional skills such as working collaboratively with key stakeholders (e.g. other education practitioners). The first school placement on the bachelor's programme (the “Orientierungspraktikum”), for example, which involves student teachers (typically in their first year of study) spending approximately 40 hours over two weeks in a school observing and analysing a specific aspect of a lesson, is designed to complement a course focusing on teacher quality. For the second school placement (“Praktikum in Pädagogisch-Psychologischen Handlungsfeldern”), student teachers (usually in their second year of study) spend two weeks in a non-school setting with children, such as a nursery, with the aim of developing their self-reflection and ability to derive and answer questions in alternative pedagogical settings. Finally, in the third school placement (“Schulpraktische Übungen”), which is usually undertaken in the final year of the bachelor's degree, student teachers spend approximately one day per week in a school throughout the academic year. During this placement they observe and support classroom teachers, and gain their first ‘real’ experience of teaching, delivering between three to six lessons independently. Alongside these more formal practical experiences, student teachers also have the opportunity to develop their teaching skills in more informal ways throughout their studies, such as through discussions with peers, and through guided activities in their university courses such as group-work, role-play, micro-teaching and reflection.

Due to COVID-19, schools and universities in Germany were closed state-wide from March 13, 2020, and did not return to full, in-person teaching until early 2021, although there was a period of predominately in-person teaching in schools from September to December 2020 (Frendl et al., 2021). These school and university closures left a ‘practicum vacuum’ for student teachers (Kidd & Murray, 2020, p. 545), meaning that many student teachers lost the opportunity to practice and develop their teaching skills (la Velle et al., 2020), through both formal practicums and informal experiences in their university courses. Where practicums and university lessons did continue, they tended to take place online, using new approaches to teaching and learning (König et al., 2020; la Velle et al., 2020), and resulting in more vicarious (and fewer mastery) learning experiences (Kidd & Murray, 2020).

Given the positive relation between practical experiences in schools and gains in teacher self-efficacy, it is likely that the school closures impacted the development of teacher self-efficacy in student teachers, and this impact may have been compounded by the move to online teaching. Support for this assumption comes from Pressley and Ha (2021), who surveyed 361 in-service teachers teaching either fully online, in a hybrid model, or fully face-to-face during the pandemic. They found that, in general, teachers reported lower levels of self-efficacy for instructional strategies and student engagement than in previous studies, with teachers teaching wholly online reporting the lowest levels of self-efficacy, irrespective of years of teaching experience. Interestingly, the authors chose not to measure self-efficacy for classroom management, as they felt that this would vary too much between the different teaching models. The current study builds on the Pressley and Ha study in three ways. Firstly, it examines the impact of school closures on the development of teacher self-efficacy in student teachers, rather than in-service teachers. Secondly, it explores the effect of the school closures on self-efficacy for classroom management in addition to self-efficacy for instructional strategies and student engagement. Thirdly, it includes a comparative sample of student teachers pursuing the same course of study in Germany, taught prior to the first COVID-19 lockdown.

1.4. Individual protective and risk factors in the development of teacher self-efficacy during the first COVID-19 lockdown

In addition to the factors discussed above, there may have been demographic (e.g. gender) or personal (e.g. prior pedagogical experience) factors that increased or decreased the extent to which the development of teacher self-efficacy was disrupted during the COVID-19 pandemic. In terms of gender, female teachers typically report lower levels of some facets of teacher self-efficacy than male teachers (Skaalvik & Skaalvik, 2007). It is important to note that there is limited research exploring gender differences in the development of teacher self-efficacy in student teachers. One study that did examine gender differences found that gender did not predict initial levels of general teacher self-efficacy, nor the rate of development in teacher self-efficacy, in student teachers undertaking a practical semester (Klassen & Durksen, 2014). However, since women were more likely than men to have seen a greater increase in childcare responsibilities during the first COVID-19 lockdown (Sevilla & Smith, 2020), it is possible that this may have resulted in gender differences, with women reporting lower levels of teacher self-efficacy due to fewer opportunities to engage with their studies. Secondly, with regard to prior experience, student teachers with greater prior pedagogical experience have been found to perceive themselves as more competent in a range of teaching activities (Römer et al., 2018). However, the relation between prior pedagogical experience and perceived competence can decrease following practical experience (Römer et al., 2018), perhaps as increased experience leads student teachers to identify gaps in their teaching ability. Thus, students who missed out on the school practicum, but who already had previous experience working with students, may have been more likely to begin the COVID-19 semester with higher initial teacher self-efficacy, and been at less risk of it stagnating or declining during the online semester, than students starting the semester with lower prior pedagogical experience.

1.5. The current study

The purpose of the current study was to investigate the impact of the loss of practical experiences as a result of the first COVID-19 lockdown on the development of self-efficacy for instructional strategies, classroom management, and student engagement in a group of student teachers studying in Germany. With this aim, it addressed the following two research questions:

• 1.How did the different facets of teacher self-efficacy, specifically self-efficacy for instructional strategies, classroom management, and student engagement, develop from the beginning to the end of the semester in student teachers training during the first COVID-19 lockdown, and how did this compare to the development of teacher self-efficacy in student teachers taking the same course prior to the first COVID-19 lockdown?
• 2.How did the development of the different facets of teacher self-efficacy from the beginning to the end of the semester during the first COVID-19 lockdown differ for certain groups of student teachers, such as women or those with greater prior pedagogical experience?

Addressing these research questions, we tested the following hypotheses

H1

Student teachers studying during the first COVID-19 lockdown would show less of a change (or even a decline) in their teacher self-efficacy, particularly self-efficacy for classroom management, over the course of a semester than student teachers taught during the previous two semesters including practical experiences.

H2

The development of teacher self-efficacy in student teachers studying during the first COVID-19 semester would be lower, or more negative (relative to students taught in the previous two semesters) in women compared to men, and higher, or more positive, in student teachers with greater prior pedological experience.

2. Method

2.1. Sample and Procedure

Data for this study came from N = 201 student teachers (54.7% female; age = 24.75, SD = 4.84) studying at a large German university in the state of Brandenburg. The study sample was drawn from a larger sample of N = 352 (52% female; age = 24.21, SD = 4.21) student teachers at the same university participating in a longitudinal study examining the effectiveness of an intervention designed to enhance student teachers' professional development through the use of micro-teaching experiences and video-based self-reflections. Participants were included in the current study if they had been taught either exclusively face-to-face or online, and if they had teacher self-efficacy data for at least two of the three waves of data collection. We ran a Little's MCAR test to compare student teachers from the complete sample who were included in this study with those who were not. The Little's MCAR test was significant (p < .001), suggesting that the data were not missing completely at random. We also found positive correlations between inclusion status and initial self-efficacy for classroom management (r = .26, p < .001), initial self-efficacy for student engagement (r = .18, p < .01), and prior pedagogical experience (r = .16, p < .01). Specifically, the sample in our study had higher initial self-efficacy for classroom management and student engagement and greater prior pedagogical experience than the remaining student teachers whose data were not included.

Participating student teachers attended courses in one of three semesters: the summer semester of 2019 (n = 41), the winter semester of 2019/20 (n = 39) or the summer semester of 2020 (n = 121), each 15 weeks in length. Student teachers in the first two semesters were taught on campus in a traditional, face-to-face format, and data from these two groups were combined to form the ‘face-to-face’ group (n = 80; 56.3% female; age = 23.70; SD = 3.39). Student teachers in the summer semester of 2020 were taught entirely online due to the COVID-19 pandemic, and they comprised the ‘COVID-19’ group (53.7% female; age = 23.79; SD = 4.17).

All students included in the current study participated in courses on the module ‘School-Related Education Research: Theories and Research Approaches’, which focus on topics related to teaching quality and enhancing students' motivation. Student teachers generally participate in this module in the fifth semester of study (i.e., the beginning of the third year of their bachelor's degree), although they can opt to take this module in an earlier or later semester if necessary. The courses on this module typically include some form of practical teaching experience such as receiving feedback from course leaders, fellow students and schoolteachers on lesson plans; micro-teaching in schools; reflecting on their own, their peers' practice or on a teachers' practice; and engaging in case studies or group-work to solve teaching-related ‘problems’.58 student teachers in the face-to-face group were participating in the course that was delivering the intervention aimed at enhancing professional development though reflective micro-teaching opportunities, whilst the remaining 22 students participated in other courses. Due to country-wide school and university closures as part of the emergency measures to contain the spread of COVID-19, none of the participants in the COVID-19 group had the opportunity to engage in the usual practical experiences in these courses, or if they did take place, they were carried out online.

The majority of students were either in the second (n = 84) or third (n = 57) year of their bachelor's programme, and on average students were in their fourth or fifth semester of study (M = 4.66, SD = 2.00). Therefore, most students had at least some previous experience of working in schools, obtained through the first school placement (“Orientierungspraktikum”) undertaken in their first year of study. The practical experiences participating students gained (or missed, in the case of the COVID-19 group) in the semester in which they took part in our study were most likely either the second (“Praktikum in Pädagogisch-Psychologischen Handlungsfeldern”) or third (“Schulpraktische Übungen”), school placements, alongside the informal experiences that form part of the content of their university courses, such as group-work, micro-teaching, and case studies. More detailed information regarding the school placements can be found on p.8.

Participants were invited to take part in the intervention study at the start of the semester by their course instructors. Participation was voluntary, and involved completing an online questionnaire at multiple points throughout the semester. The data included in this study were collected during week one (T₁), nine (T₂), and 13 (T₃) of the semesters.

2.2. Measures

2.2.1. Teacher self-efficacy

Teacher self-efficacy was measured using an adapted German version of the Ohio State ‘Teachers’ Sense of Efficacy Scale’ (Tschannen-Moran & Hoy, 2001), developed by Pfitzner-Eden et al. (2014). The questionnaire measures perceived self-efficacy in three domains: Instructional Strategies (4 items, e.g. ‘I am convinced that I can provide an alternative explanation or example when students are confused’), Classroom Management (4 items, e.g. ‘I am convinced that I can control disruptive behaviour in the classroom’), and Student Engagement (4 items e.g. ‘I am convinced that I can help students value learning’). Participants responded on a 6-point scale from 1 (not correct at all) to 6 (that is completely right), with a higher score indicating higher perceived self-efficacy. The original version of the questionnaire uses a 9-point Likert scale, but a 6-point scale was adopted here to allow for ease of completion (e.g. to allow the options to fit on the screen of a mobile phone). The measure has good psychometric properties (Pfitzner-Eden et al., 2014). Psychometric data for the present study, reported in Table 1 below, were also good.

Table 1.

Descriptive statistics for each facet of teacher self-efficacy and prior pedagogical knowledge.

		Face-to-face					COVID-19
		M	SD	α	Skew	Kurtosis	M	SD	α	Skew	Kurtosis
Self-efficacy for instructional strategies (Range 1–6)	T1	4.78	0.53	.70	−0.27	−0.64	4.76	0.63	.78	−0.43	1.57
	T2	4.96	0.58	.70	−0.46	0.82	4.80	0.59	.78	−0.77	1.62
	T3	4.92	0.56	.82	−0.74	0.28	4.71	0.63	.85	−1.06	2.60
Self-efficacy for classroom management (Range 1–6)	T1	4.50	0.56	.75	−0.03	0.29	4.60	0.68	.85	−0.33	1.24
	T2	4.73	0.72	.86	−0.21	−0.25	4.63	0.67	.86	−0.09	−0.08
	T3	4.80	0.65	.85	−0.50	0.61	4.66	0.61	.70	0.12	−0.32
Self-efficacy for student engagement (Range 1–6)	T1	4.70	0.60	.72	−0.44	−0.20	4.86	0.65	.80	−0.25	−0.28
	T2	4.86	0.63	.80	−0.60	−0.01	4.81	0.64	.83	−0.21	−0.47
	T3	4.93	0.56	.82	−0.09	−0.25	4.73	0.64	.82	−0.70	0.85
Prior pedagogical experience (Range 0–3)	T1	1.06	0.99		0.52	−0.81	0.97	0.94		0.64	−0.54

2.2.2. Prior pedagogical experience

Prior pedagogical experience, measured at T₁, was included as a covariate in the current study. It was measured using the 5-item ‘Pedagogical Experience Scale’ (König et al., 2013). The questionnaire assesses the types of prior pedagogical experiences an individual has (e.g. ‘Care of children - siblings, own children, babysitting, au pair’). Participants indicate whether they have a particular type of experience on a dichotomous scale of ‘yes’ or ‘no’. A sum score is computed, with a higher score indicating a greater variety of prior pedagogical experience. In our analysis, we included just the three activities most relevant to teaching, namely ‘Tutoring/homework supervision as one-to-one tuition’, ‘Tutoring/homework support for study group/school class’, and ‘Own teaching activity at a school outside of study programme (e.g. substitute teacher, assistant teacher)’. The measure has been used previously with student teachers in Germany, and has good validity (Römer et al., 2018). Psychometric data for the present study, reported in Table 1 below, were good.

2.3. Data analysis

We used latent growth curve models (LGCMS) with an additional growth factor (AGF,Muthen & Curran, 1997) to examine the development of individual facets of teacher self-efficacy across a semester, and to test whether this development differed for students in the COVID-19 group as compared to the development in the face-to-face group. In the first step, we ran a series of Confirmatory Factor Analyses (CFA) to determine configural, loading, and intercept measurement invariance for each of the three facets of teacher self-efficacy. Adopting the guidance for sample sizes of N ≤ 300 from Chen (2007), loading noninvariance was assumed if there was a change in CFI of ≥ −0.005, accompanied by a change of ≥ 0.010 in RMSEA or ≥ 0.025 in SRMR. Intercept noninvariance was assumed if there was a change in CFI of ≥ −0.005, accompanied by a change of ≥ 0.010 in RMSEA or ≥ 0.005 in SRMR. If noninvariance was found, model constraints were systematically released until a partially invariant model was achieved. Partial invariance was established if the parameters of at least two indicators were equal across groups and/or time (Byrne et al., 1989). All three facets of teacher self-efficacy had full or partial measurement invariance. Partial invariance is considered sufficient for comparisons across groups to be made (Byrne et al., 1989). Model fit statistics, and the results of the tests of measurement invariance, are presented in the appendix.

Once measurement invariance had been established, we ran a series of first-order, linear LGCMs for each facet of teacher self-efficacy. First, we ran separate LGCMs for the face-to-face and COVID-19 groups. As mentioned in the Sample and Procedure section, the data analysed here were drawn from a larger project examining the impact of an intervention designed to enhance student teachers’ professional development. To rule out any potential intervention effects on teacher self-efficacy, we included intervention group status as a predictor of slope in the face-to-face group. Intervention group status did not significantly predict the slope for any of the three facets of teacher self-efficacy, and therefore it was not included in the remaining models.

After modelling the development for each group separately, a multi-group analysis was run, in which the intercept was set to zero for both groups, and the slope for the face-to-face group was held equal across groups. The AGF was modelled for the COVID-19 group only. An AGF significantly greater than zero suggested that there was a difference in the amount and/or direction of development between the two groups. In the next step, the interaction between the intercept and AGF was modelled for the COVID-19 group. A significant interaction implied that the rate of additional growth in a particular facet of self-efficacy changed as a function of an individual student's initial level of that facet of self-efficacy.

To explore the group differences further, the interaction model was compared to two comparison models to determine whether it fit the data best. It was first compared to a model in which the intercept in the COVID-19 group was allowed to differ from zero. If the intercept mean was significantly greater than zero, this suggested that there were differences between the intercepts in the two groups. Next, it was compared to a model in which the AGF mean and variance were set to zero in the COVID-19 group. If this comparison model showed a better fit with the data, this implied no difference in the development between the two groups. Finally, in addition to the steps proposed by Muthen and Curran (1997), we ran a latent class analysis that included two time-invariant covariates as predictors of the intercept and AGF, namely gender and prior pedagogical knowledge. At each stage of the analysis, model fit was regarded as adequate if CFI = ≥ 0.93, RMSEA ≤0.08 or SRMR = ≤ 0.1 (Browne & Cudeck, 1992; Byrne, 1994), and good if CFI ≥0.95, RMSEA ≤0.06 or SRMR ≤0.08 (Hu & Bentler, 1999).

3. Results

3.1. Descriptive statistics

Descriptive statistics are shown in Table 1. All three scales of teacher self-efficacy showed good internal consistency across the three waves of data collection. Data were generally negatively skewed, but skewness and kurtosis values fell within acceptable limits for latent growth curve models (Stull, 2008). However, the maximum likelihood parameter estimates with standard errors and the chi-square test statistic that were applied in the main analyses are robust to non-normality (Benson & Fleishman, 1994). Mean scores tended to increase over the course of the semester for students in the face-to-face group and to decline (or show a small increase, in the case of self-efficacy for classroom management) in the COVID-19 group.

3.2. Self-efficacy for instructional strategies

The first-order, linear latent growth curve models (LGCM) showed an acceptable fit to the data in the face-to-face group: χ²(1) = 1.37, p = .24, CFI = 0.924, RMSEA = 0.068, SRMR = 0.086, and a good fit to the data in the COVID-19 group: χ²(3) = 4.12, p = .24, CFI = 0.991, RMSEA = 0.056, SRMR = 0.178. The slope mean in the face-to-face group was positive, but small and non-significant (M = 0.01 p = .07), indicating that self-efficacy for instructional strategies did not change substantially across time in this group. The slope mean in the COVID-19 group was zero and non-significant (M = 0.00, p = .45), indicating that self-efficacy for instructional strategies remained the same across time in this group. The multi-group model with the AGF did not fit the data well: χ² (7) = 16.51, p < .05, CFI = 0.927, RMSEA = 0.116, SRMR = 0.489. The AGF mean was negative and significant (M = −0.01, p < .05), indicating that student teachers in the COVID-19 group experienced significantly less change in self-efficacy for instructional strategies across time, relative to the student teachers in the face-to-face group. Model fit improved when the AGF was regressed on to the intercept factor: χ² (7) = 11.94, p = .10, CFI = 0.962, RMSEA = 0.084, SRMR = 0.347. The interaction between the intercept and AGF was positive, but small and non-significant (β = 0.03 p = .06), indicating that initial levels of self-efficacy for instructional strategies in the COVID-19 group did not influence the extent to which the development of this facet of self-efficacy differed from the face-to-face group across time.

The interaction model was used to examine alternative models. Firstly, it was compared to the model in which the intercept of the COVID-19 group was allowed to differ from zero. The model fit was similar: χ²(6) = 11.91, p = .06, CFI = 0.954, RMSEA = 0.099, SRMR = 0.343, but, importantly, the intercept mean for the COVID-19 group was small and non-significant (M = – 0.01, p = .87), suggesting that initial levels of self-efficacy for instructional strategies were similar across groups. Secondly, it was compared to the model in which the AGF mean and variance were set to zero. The model showed a poor fit to the data: χ² (9) = 22.29, p < .01, CFI = 0.897, RMSEA = 0.121, SRMR = 0.559, suggesting that the development of self-efficacy for instructional strategies across time differed between the two groups. The interaction model was therefore accepted.

The interaction model was subsequently used to perform the latent class analysis with the two time-invariant covariates. Gender did not predict the intercept (β = 0.00, p = .98) or AGF (β = 0.00, p = .97), nor did it interact with the intercept to predict the AGF (β = – 0.03, p = .06). Similarly, prior pedagogical experience did not predict the intercept (β = 0.02, p = .47) or AGF (β = 0.00, p = .85), nor did it interact with the intercept to predict the AGF (β = 0.00, p = .70). This implies that the development of self-efficacy for instructional strategies was lower in the COVID-19 group relative to the face-to-face group, irrespective of gender or prior pedagogical experiences.

With the covariates in the model, however, the interaction between the intercept and AGF became significant (β = 0.05, p < .05). For every one-unit increase in initial self-efficacy for instructional strategies, the AGF increased (became less negative) by 0.05. This indicates that students in the COVID-19 group with higher initial self-efficacy for instructional strategies showed more of an improvement in this facet of self-efficacy over the semester than student teachers with lower initial self-efficacy for instructional strategies.

Taken together, the results suggest that self-efficacy for instructional strategies did not significantly increase or decrease in either the face-to-face or COVID-19 group across a semester. However, the amount of change between the two groups differed significantly. Student teachers in the COVID-19 group showed less of an improvement in their self-efficacy for instructional strategies over the semester relative to the students in the face-to-face group. The initial level of self-efficacy for instructional strategies, the AGF, and the interaction between the intercept and AGF were not predicted by gender or prior pedagogical experiences, but, with these covariates in the model, the interaction between the intercept and AGF became significant. Student teachers in the COVID-19 group starting the semester with higher initial self-efficacy for instructional strategies showed more of an improvement in this facet of self-efficacy across time.

3.3. Self-efficacy for classroom management

The data showed a good fit to the model in the face-to-face group: χ²(1) = 0.56, p = .46, CFI = 1.000, RMSEA = 0.000, SRMR = 0.033, and the COVID-19 group: χ²(1) = 0.14, p = .70, CFI = 1.000, RMSEA = 0.000, SRMR = 0.007. The slope mean in the face-to-face group was positive and significant (M = 0.02, p < .01), indicating that there was a significant increase in self-efficacy for classroom management across time for this group. The slope mean in the COVID-19 group was zero and non-significant (M = 0.00, p = .27), indicating that self-efficacy for classroom management remained the same across time in this group. The multi-group model with the AGF fit the data well: χ² (5) = 9.52, p = .09, CFI = 0.970, RMSEA = 0.095, SRMR = 0.322 and the AGF mean was negative and significant (M = −0.01, p < .05), indicating that student teachers in the COVID-19 group experienced significantly less change in self-efficacy for classroom management across time relative to the student teachers in the face-to-face group. When the AGF was regressed on to the intercept factor, the data fit the model well: χ² (5) = 3.40, p = .64, CFI = 1.000, RMSEA = 0.000, SRMR = 0.093. The interaction between the intercept and AGF was positive and significant (β = 0.04, p < .01). For every one-unit increase in self-efficacy for classroom management, the AGF increased (became less negative) by 0.04. This indicates that students in the COVID-19 group with higher initial self-efficacy for classroom management showed more of an improvement in this facet of self-efficacy over the semester than students with lower initial self-efficacy for classroom management.

The interaction model was then compared to the model in which the intercept of the COVID-19 group was allowed to differ from zero. The model fit was similar: χ² (4) = 2.20, p = .70, CFI = 1.000, RMSEA = 0.000, SRMR = 0.113, but the intercept mean for the COVID-19 group was small and non-significant (M = 0.10, p = .27), suggesting that initial levels of self-efficacy for classroom management were similar across groups. Next, the interaction model was compared to the model in which the AGF mean and variance were set to zero. The model fit was acceptable, but significantly worse than the interaction model: χ² (7) = 14.48, p < .05, CFI = 0.950, RMSEA = 0.103, SRMR = 0.369, suggesting that the change in self-efficacy for classroom management across time differed between the two groups. The interaction model was therefore accepted.

The subsequent latent class analysis with the two time-invariant covariates showed that gender did not predict intercept (β = 0.10, p = .23) or the AGF (β = −0.01, p = .28), nor did it interact with the intercept to predict the AGF (β = 0.00, p = .94). Prior pedagogical experiences did not predict the intercept (β = 0.06, p = .87) or the AGF (β = 0.00, p = .87), nor did it interact with the intercept to predict the AGF (β = 0.00, p = .56). This implies that the development of self-efficacy for classroom management was lower in the COVID-19 group (relative to the face-to-face group), irrespective of gender or prior pedagogical experiences. However, the interaction between the intercept and AGF became non-significant with the covariates in the model (β = 0.02, p = .26).

Taken together, the results indicate that self-efficacy for classroom management increased, in the face-to-face group over the course of a semester but stayed the same in the COVID-19 group. There was, however, a significant difference in the strength of development between the two groups, with the students in the COVID-19 group showing significantly less of an improvement in their self-efficacy for classroom management than students in the face-to-face group. Without the covariates in the model, the AGF was predicted by intercept, indicating that student teachers in the COVID-19 group with higher initial self-efficacy for classroom management showed more of an increase in their self-efficacy over the semester than student teachers with lower initial self-efficacy for classroom management. The initial level of self-efficacy for classroom management, the AGF, and the interaction between the intercept and AGF were not predicted by gender or prior pedagogical experiences, but with these covariates in the model, the interaction between the intercept and AGF became non-significant.

3.4. Self-efficacy for student engagement

The model showed a good fit to the data in the face-to-face group: χ²(1) = 0.75, p = .39, CFI = 1.000, RMSEA = 0.000, SRMR = 0.030, and the COVID-19 group: χ²(1) = 0.01, p = .93, CFI = 1.000, RMSEA = 0.000, SRMR = 0.001. The slope mean in the face-to-face group was positive and significant (M = 0.02, p < .01), indicating that there was a significant increase in self-efficacy for student engagement across time for this group. The slope mean in the COVID-19 group was negative and significant, albeit small (M = −0.01, p < .05), indicating that self-efficacy for student engagement declined across time in this group. The multi-group model with the AGF fit the data well: χ² (5) = 9.60, p = .09, CFI = 0.973, RMSEA = 0.096, SRMR = 0.265 and the AGF mean was negative and significant (M = −0.02, p < .01). The model fit for the interaction model was good: χ²(5) = 5.08, p = .41, CFI = 1.000, RMSEA = 0.012, SRMR = 0.051. The interaction between the intercept and AGF was positive and significant (β = 0.03, p < .05). For every one-unit increase in self-efficacy for student engagement, the AGF increased (became less negative) by 0.03. This indicates that students in the COVID-19 group with higher initial self-efficacy for student engagement experienced less of a decline in this facet of self-efficacy over the semester than student teachers with lower initial self-efficacy for student engagement.

The interaction model was then compared to the model in which the intercept of the COVID-19 group was allowed to differ from zero. The model showed a better fit to the data: χ²(4) = 1.07, p = .90, CFI = 1.000, RMSEA = 0.000, SRMR = 0.058. Furthermore, the intercept mean for the COVID-19 group was positive and significant (M = 0.18, p < .05), suggesting that initial levels of self-efficacy for student engagement were higher in the COVID-19 group than in the face-to-face group. The model with the interaction term between intercept and AGF was a better fit to the data than the model where the AGF mean and variance were set to zero: χ²(7) = 14.48, p < .05, CFI = 0.950, RMSEA = 0.103, SRMR = 0.369, suggesting that the development of self-efficacy for student engagement differed between the two groups across time. Given that initial self-efficacy for student engagement was higher in the COVID-19 group, but finished lower in this group than the face-to-face group, and given that the AGF model was a better fit to the data than when the AGF was set to zero, the interaction model was therefore accepted.

The latent class analysis with the two time-invariant covariates was carried out using the interaction model. Gender did not predict the intercept (β = −0.08, p = .39) or the AGF (β = 0.00, p = .70), but it interacted with the intercept to predict the AGF (β = −0.02, p < .05). Self-efficacy for student engagement declined less in women who had higher initial self-efficacy for student engagement in the COVID-19 group (β = 0.04, p < .05), but not in men (β = 0.00, p = .95). For every one-unit increase in self-efficacy for student engagement in women, the AGF increased (became less negative) by 0.04. Similarly, prior pedagogical experiences did not predict the intercept (β = 0.01, p = .80) or the AGF (β = 0.00, p = .59), but it did interact with the intercept to significantly predict the AGF (β = 0.01, p < .05). Self-efficacy for student engagement declined less in the COVID-19 group in student teachers with higher prior pedagogical experiences if they also had higher initial self-efficacy for student engagement (β = 0.03, p < .01), but not in student teachers with lower prior pedagogical knowledge (β = 0.01, p = .59). For every one-unit increase in self-efficacy for student engagement in student teachers with higher prior pedagogical experiences, the AGF increased (became less negative) by 0.03. The interaction between the intercept and AGF became non-significant with the covariates in the model (β = 0.02, p = .07).

Taken together, the results suggest that self-efficacy for student engagement significantly increased in the face-to-face group over the course of the semester, and significantly declined in the COVID-19 group. The change in the COVID-19 group was significantly less than the development in the face-to-face group. This finding needs to be interpreted with caution, however, as initial self-efficacy for student engagement was significantly higher in the COVID-19 group than the face-to-face group. However, because self-efficacy for student engagement declined in the COVID-19 group, and because the model with the AGF showed a better fit to the data than the model without it, this suggests that the development of self-efficacy for student engagement indeed differed between the face-to-face and COVID-19 groups, despite different initial levels of self-efficacy. The AGF was significantly predicted by the intercept, suggesting that student teachers in the COVID-19 group with higher initial self-efficacy for student engagement showed less of a decline over the semester. Initial level of self-efficacy for student engagement and the AGF were not predicted by gender or prior pedagogical experiences, but the interaction between the intercept and AGF was. Higher initial self-efficacy for student engagement protected women and student teachers with higher prior pedagogical experiences in the COVID-19 group from the biggest declines in self-efficacy for student engagement.

4. Discussion

The purpose of this study was to investigate how the development of teacher self-efficacy in student teachers training during the first COVID-19 lockdown (the ‘COVID-19 group’) differed from the development of teacher self-efficacy in student teachers who participated in courses before the lockdown (the ‘face-to-face group’) that included micro-teaching and reflective experiences. In addition, we examined two potential protective or risk factors – gender and prior pedagogical experiences – to see whether they moderated the development of teacher self-efficacy in the COVID-19 group. Our key finding was that student teachers in the COVID-19 group showed significantly less improvement in self-efficacy for instructional strategies, classroom management, and student engagement over the course of a semester than students in the face-to-face group. This difference was greatest for self-efficacy for student engagement. Gender and prior pedagogical experience effects were found for self-efficacy for student engagement only. In the following sections, we discuss our findings in more detail and consider the practical implications of these findings for teaching and teacher education.

4.1. Differential development of student teachers’ self-efficacy

In line with our first hypothesis (H₁), the students in the COVID-19 group showed significantly less of an increase in all three facets of teacher self-efficacy compared to the students in the face-to-face group, although there was no significant change in self-efficacy for instructional strategies in either group. Previous research has highlighted the positive impact school practicums can have on the development of teacher self-efficacy (Berg & Smith, 2018; Fives et al., 2007; Klassen & Durksen, 2014; Pfitzner-Eden, 2016b). Thus, a possible explanation for our findings is the removal of these opportunities due to the COVID-19 pandemic. The lack of opportunities to gain practical experiences in schools or on university courses might have caused the stagnation (or decline) in all three facets of self-efficacy in the COVID-19 group compared with the student teachers in the face-to-face group. Overall, the differences in teacher self-efficacy between the two groups were small, however. Furthermore, the amount of change in each facet of teacher self-efficacy across the semester was lower in the face-to-face group than has been found in previous research examining the impact of school practicums, although the majority of studies have reported only small to medium effect sizes (Berg & Smith, 2018; Fives et al., 2007; Klassen & Durksen, 2014; Pfitzner-Eden, 2016a). This is possibly because the student teachers in the face-to-face group were not undertaking a full practical semester (i.e., they did not spend the majority of the semester in a school setting), unlike in the aforementioned studies. However, the changes in the COVID-19 group from the start to the end of the semester were similar to the changes reported by Pfitzner-Eden (2016a) for students completing a university-based semester. It is possible that the differences in development between the two groups would have been larger if we had included students who were undertaking (or who would have been, in the case of the COVID-19 group) a more intensive school practicum. Furthermore, although students in the COVID-19 group showed less of an increase in their self-efficacy across the semester, it is not possible to determine whether this difference would be observable: that is, whether this cohort of students would ‘appear’ to be lower in self-efficacy than previous cohorts. Future research examining teacher self-efficacy may wish to collect information about the extent to which teacher training was disrupted by COVID-19 in order to examine the longer-term effects of school and university closures.

It is interesting to note that the three self-efficacy facets were differently affected by group membership. We uncovered a particularly strong group difference in student teachers’ self-efficacy for student engagement and classroom management, whereby student teachers training before the COVID-19 pandemic experienced an increase in self-efficacy for classroom management and student engagement across the semester, whilst they remained the same and decreased, respectively, in the COVID-19 group. However, self-efficacy for instructional strategies seemed to be unaffected by group membership. Our results are thereby partially in line with those of previous research, which found improvements in all three facets of teacher self-efficacy (Berg & Smith, 2018; Fives et al., 2007), or just in self-efficacy for instructional strategies and classroom management (Pfitzner-Eden, 2016b) following a school practicum, and no changes, or declines, in teacher self-efficacy following a university-based semester with no practical experiences (Pfitzner-Eden, 2016b).

One possible explanation for the increase in self-efficacy for student engagement and classroom management in the group that had the chance to gain practical experiences in authentic classrooms might be that these students encountered more possibilities for mastery experiences in engaging students and managing classroom disturbances – and that these two aspects of good teaching were easier to perceive and capture for the student teachers than the consequences of an activating instruction in class. Student teachers who were presenting their lesson plans in virtual courses to their peers, in turn, might not have had opportunities to perceive the consequences of motivation-enhancing teaching strategies, therefore resulting in lower self-efficacy for student engagement in the group that participated in courses during the COVID-19 semester. This assumption is supported by findings from Pressley and Ha (2021), who found that in-service teachers reported the lowest levels of self-efficacy for student engagement when teaching exclusively online during the COVID-19 pandemic. Interestingly, Pressley and Ha did not measure self-efficacy for classroom management in their study, as they argued that this would vary too much between teachers teaching online and those teaching face-to-face due to the different types of classroom management strategies needed in these different learning environments. This could also provide a possible explanation for why classroom management remained the same for the student teachers training during the COVID-19 semester in our study.

In addition to these findings, we also found a significant, positive interaction between initial self-efficacy and the additional growth factor (the difference in growth between the face-to-face and the COVID-19 group) for self-efficacy for classroom management and student engagement. Student teachers in the COVID-19 group who started the semester with higher levels of self-efficacy in classroom management and student engagement showed more of an improvement, and less of a decline, in these facets respectively, over the course of a semester, than student teachers who started the semester with lower levels of self-efficacy. Lower initial teacher self-efficacy has been found to place student teachers at risk of more negative outcomes. Dicke et al. (2014) found, for example, that low initial self-efficacy for classroom management placed graduate teachers at greater risk of emotional exhaustion, and this was mediated by greater classroom disturbances. Students with lower initial self-efficacy in the COVID-19 group may have encountered more difficulties when attempting to engage with the limited practical experiences that were on offer during COVID-19, and as a result of this, their self-efficacy was less likely to increase, or more likely to decrease, overtime. Conversely, students with higher self-efficacy may have been protected from stagnating or declining self-efficacy because their higher initial self-efficacy enabled them to more confidently approach the new online learning environment, and therefore be more likely to experience mastery. This in turn may have led to an increase in self-efficacy across the semester. Interestingly, the interaction between initial levels of self-efficacy for instructional strategies and the AGF only became significant when the covariates were added into the model, but neither gender nor prior pedagogical experience predicted the initial levels of this facet of self-efficacy, the AGF, nor their interaction, making this finding difficult to interpret. Overall, initial levels of self-efficacy seemed to act as a possible protective or risk factor in the development of self-efficacy in the COVID-19 group but the design of our study unfortunately limits us from drawing conclusions regarding the mechanism by which they did so. Given its clear impact on later outcomes, future research is needed to explore how initial levels of teacher self-efficacy interact with classroom experience to predict self-efficacy development in student teachers.

Our study not only provides insight into the potential impact of the COVID-19-induced school closures on the development of teacher self-efficacy in student teachers, but it also contributes to the field of teacher self-efficacy research more generally. First, it adds to the extant literature on teacher self-efficacy, demonstrating the positive impact of practical experiences in schools on the development of teacher self-efficacy in student teachers and, second, it uses a novel design to do so. Previous studies (e.g. Fives et al., 2007) have typically used pre-post designs, measuring the teacher self-efficacy of student teachers at the beginning and end of their school practicums. Thus, it has not been possible to determine whether it is the practical experience in schools that leads to increased teacher self-efficacy, or the fact that the students have been in their teacher training course for longer (although the link between mastery experiences and higher self-efficacy suggests the former). Although previous studies (see Pfitzner-Eden, 2016a) have addressed this to some extent by comparing the development of teacher self-efficacy during a university-based semester with a practicum semester, the practicum semester followed the university-based semester. This still makes it difficult to rule out whether more practical experiences in the teacher training programme overall, or the practicum itself, led to the increase in teacher self-efficacy of student teachers. This lack of clarity is increased by the finding that, in more advanced student teachers, teacher self-efficacy did increase during a university-based semester. In our study, we compared student teachers at similar points in their teacher education programmes, with the only difference being whether or not they were able to undertake practical experience in schools or on their university courses. The finding that teacher self-efficacy increased in the face-to-face group but not the COVID-19 group lends weight to the argument that it is school practicums, and not the length of time in a course, that impacts the development of teacher self-efficacy in student teachers.

4.2. Moderators of developmental trends in facets of student teachers’ self-efficacy

Our findings only partially confirmed our expectations about potential moderators of developmental changes in facets of student teachers’ self-efficacy (H₂). Contrary to the prediction, gender and prior pedagogical experiences did not significantly predict starting levels of self-efficacy, nor did they predict the additional growth factor for any of the three facets of self-efficacy. However, they did moderate the interaction between the initial level of self-efficacy and the additional growth factor for self-efficacy for student engagement. Women, but not men, who started the semester with higher levels of self-efficacy for student engagement showed less of a decline in this facet throughout the semester. This finding suggests, in line with Social Cognitive Theory (Bandura, 1977, 1997), that personal resources (such as teacher self-efficacy) offer important protective effects when encountering difficulties (such as a lack of practical experiences), but that these resources may be more salient for some groups than for others. As noted in the literature review, gender differences in the development of teacher self-efficacy in student teachers has received limited attention. Further research is needed to understand more about how and when gender differences in the development of teacher self-efficacy occur.

In terms of prior pedagogical experiences, students with high prior pedagogical experiences who also started the semester with higher self-efficacy for student engagement showed less of a decline in self-efficacy for student engagement across the semester than students with low prior pedagogical experiences. Previous research has suggested that prior pedagogical experiences could be a protective factor in the development of teacher self-efficacy (Römer et al., 2018). In our study, it appeared to have either no effect or a protective effect, depending on the starting level and the facet of teacher self-efficacy. It is worth pointing out that the Römer et al. (2018) study measured perceived competence (rather than teacher self-efficacy specifically), which may explain the differences between their findings and ours. Clearly more research is needed to examine the impact of prior pedagogical experiences on the development of teacher self-efficacy in student teachers.

4.3. Implications for teaching and teacher education

Our findings have some important implications for teaching and teacher education. Firstly, it seems that providing student teachers with formal and informal opportunities to engage in practical experience, such as through observing more experienced teachers, delivering short lessons, or engaging in role-play is important for the development of teacher self-efficacy. Teacher education programmes should therefore look for ways to provide meaningful practical experiences, alongside and including school practicums throughout their courses. Secondly, opportunities for gaining practical experience may be even more important now, given that there is a cohort of students who have missed out on these experiences, and may have lower teacher self-efficacy as a result. Student teachers having lower teacher self-efficacy is worrying not only because teacher self-efficacy is linked to important teacher and student outcomes, but also because our study findings indicate that higher initial levels of self-efficacy seemed to protect student teachers from stagnating or declining teacher self-efficacy for classroom management and student engagement. If student teachers may now be entering the profession with lower levels of self-efficacy than is typical, they may also deal with classroom challenges less effectively in the future or be less confident dealing with challenging situations in class. Thus, the third implication of our findings is that teacher training programmes may want to consider implementing targeted interventions for students with the lowest teacher self-efficacy. Interventions targeting women, and students with less prior pedagogical experiences, may also be particularly beneficial.

4.4. Limitations and future directions

Despite the strengths of our longitudinal, quasi-experimental design, there are some important limitations to the current study that require consideration. Firstly, although the sample size was adequate for the multi-group latent growth curve analysis we conducted (Muthen & Curran, 1997), it is recommended that the group sizes be equal to ensure that there is enough power to adequately detect group differences (Curran & Muthén, 1999). In the current study, the group sizes were unequal, with the COVID-19 group being almost one-third larger than the face-to-face group. Future research should include larger and more equal group sizes. In addition, our sample was selected from a larger sample of student teachers, based on them having provided teacher self-efficacy ratings for at least two of the three data collection timepoints. Little's MCAR test suggested that the data were not missing at random, and that the student teachers in our study had higher initial self-efficacy for classroom management and student engagement, and greater prior pedagogical experiences. Although we chose not to include student teachers with only one data collection point (due to our focus on development overtime), future research should try to ensure that students with lower initial self-efficacy are also included.

Secondly, we did not collect data on the kind of experience the student teachers were, or were not, experiencing in schools or online throughout the semester. Future research should include this as a moderator of development to help further explain the change in self-efficacy development across a semester, and to strengthen the claim that practical experiences in schools is central to the development of teacher self-efficacy in student teachers. This is supported by research from Clark et al. (2014), who found that students in more intensive ‘internship’-type school placements tended to show less of an increase in teacher self-efficacy than students in more structured and supportive ones. Building on this, it would also be beneficial to collect more detailed information about student teachers' prior pedagogical experiences to understand how this interacts with experience gained in school practicums to impact the development of teacher self-efficacy.

Thirdly, it is possible that the stagnation or decline in teacher self-efficacy in the COVID-19 group was a product of the disruption to their learning caused by the COVID-19 pandemic more generally, rather than the lack of school practicums specifically. For example, research examining the impact of online teaching on university students has found that students perceive themselves as learning less online as compared to face-to-face teaching (Usher et al., 2021), as well as experiencing more negative emotions towards their learning (Madrigal & Blevins, 2021; Usher et al., 2021). For student teachers specifically, the removal of practical experiences in schools led to anxiety about how it would impact their own teacher development and ability to graduate (Kidd & Murray, 2020). Adapting to new forms of learning was compounded by university students feeling less capable of learning (Usher et al., 2021), or less motivated to do so (Madrigal & Blevins, 2021). Taken together, it is possible that the move to online learning negatively impacted student teachers’ teacher self-efficacy through reduced learning - individuals are unlikely to feel efficacious about performing behaviours they have not been taught to perform (Bandura, 1997). Future research should consider course-related factors besides practical experiences that might impact the development of teacher self-efficacy.

5. Conclusions

Our study found that student teachers training during the first COVID-19 lockdown showed less of an increase in student teachers' self-efficacy for instructional strategies, classroom management, and student engagement across the semester, relative to students taught face-to-face in the preceding two semesters. These findings suggest that a lack of practical experiences in schools due to widespread school closures may have negatively impacted the development of teacher self-efficacy in this cohort of student teachers. Student teachers in the COVID-19 group with higher initial self-efficacy for classroom management and student engagement reported a greater increase in these facets of self-efficacy across the semester, and this effect was moderated by gender and prior pedagogical experiences for self-efficacy for student engagement. Teacher educators should be aware of the potential negative impact of the COVID-19-related school closures on the development of student teachers’ teacher self-efficacy, and put in place interventions to help these students increase their teacher self-efficacy. Future research should examine the impact different types of school placements, including online teaching, have on the strength and direction of the change in the different facets of teacher self-efficacy, and identify further factors that protect student teachers from the greatest declines in teacher self-efficacy when encountering challenging situations.

Credit author statement

Wendy Symes: Conceptualization, Methodology, Formal analysis, Writing – Original Draft, Rebecca Lazarides: Methodology, Formal analysis, Resources, Writing – Review and Editing, Supervision, Project, Funding acquisition, Isabell Hußner: Investigation, Resources, Data Curation, Writing - Review and Editing, Project.

Appendix.

Results of tests of measurement invariance for each facet of teacher self-efficacy across groups (face-to-face and COVID-19) and time (T1, T2 and T3).

Table 1.

Results of Tests of Measurement Invariance for Self-Efficacy for Instructional Strategies, Classroom Management, and Student Engagement Across Groups and Time.

Model	χ2 (df)	CFI	RMSEA	SRMR	ΔCFI	ΔRMSEA	ΔSRMR	Decision a
Self-efficacy for instructional strategies
Configural invariance (group)	104.96∗ (78)	.969	.059	.075
Loading invariance (group)	108.33 (87)	.975	.049	.085	.006	−.01	−.004	Accept
Intercept invariance (group)	111.78 (96)	.982	.04	.086	.007	.009	.001	Accept
Loading invariance (group and time)	115.25 (102)	.985	.036	.087	.003	−.004	.001	Accept
Intercept invariance (group and time)	131.05∗ (106)	.971	.048	.099	−.014	.012	.012	Reject
Partial intercept invariance (group and time)	118.23 (104)	.984	.037	.089	−.001	.001	.002	Accept
Self-efficacy for classroom management
Configural invariance (group)	114.62∗∗ (78)	.970	.068	.074
Loading invariance (group)	125.81∗∗ (87)	.968	.067	.106	−.002	−.001	.032	Accept
Intercept invariance (group)	135.10∗∗ (96)	.968	.064	.119	0	−.003	.013	Accept
Loading invariance (group and time)	141.81∗∗ (102)	.967	.062	.125	−.001	−.002	.006	Accept
Intercept invariance (group and time)	150.99∗∗ (106)	.963	.065	.128	−.004	.003	.003	Accept
Self-efficacy for student engagement
Configural invariance (group)	121.29∗∗ (78)	.956	.074	.069
Loading invariance (group)	138.98∗∗∗ (87)	.947	.077	.120	−.009	.003	.051	Reject
Partial loading invariance (group)	130.06∗∗ (86)	.955	.071	.097	−.001	−.003	.028	Accept
Intercept invariance (group)	137.98∗∗ (95)	.956	.067	.101	.001	−.004	.004	Accept
Loading invariance (group and time)	150.03∗∗ (101)	.950	.069	.128	−.006	.002	.031	Reject
Partial loading invariance (group and time)	139.39∗∗ (99)	.959	.064	.107	.003	−.003	.006	Accept
Intercept invariance (group and time)	141.77∗∗ (103)	.960	.061	.108	.001	−.003	0.001	Accept

∗p < .05.∗∗p < .01.∗∗∗p < .001.

^aDecision to accept = measurement invariance, decision to reject = measurement noninvariance.

Data availability

Data will be made available on request.