Heading for new shores? Longitudinal participation patterns in teacher professional development

ABSTRACT

Recent research indicated the advantages of diversified teacher professional development participation patterns (PDPP) to foster teachers’ professionalisation. This study investigated teachers’ PDPP from both cross-sectional and longitudinal perspectives using data from the National Educational Panel Study (NEPS) utilising information on teachers’ professional development enrolment across three years (N = 3,539). Results from latent class and multiple group analyses indicated that teachers showing higher professional engagement and teachers who were granted paid temporary leave from teaching demonstrated more diversified PDPP. Multiple logistic regression models revealed that the same variables seemed to prevent changes to less diversified PDPP over time. If our study results are confirmed in replications and studies that allow to detect cause-effect relationships, implications for educational policy and practice could be to raise teachers’ professional engagement and grant engaged teachers a paid temporary leave from teaching to participate in professional development activities.

KEYWORDS:

• Teacher professional development
• professional development participation patterns
• longitudinal

Introduction

Teachers are critical to the success of education systems and improving student learning (Blömeke and Olsen 2019). Empowering teachers to provide high-quality learning environments across teachers’ professional lives is often achieved through high-quality in-service teacher professional development (PD; Darling-Hammond Hyler, and Gardner 2017). PD can be defined as formal (e.g. structured face-to-face PD programmes; Kleickmann et al. 2016) and informal learning (e.g. social media; Fütterer et al. 2021) opportunities to improve teachers’ competency (D. Richter et al. 2011; Fütterer et al. 2023; for a systematic and comprehensive overview of PD definitions, see; Sancar, Atal, and Deryakulu 2021).

As demands of teaching change over time (Hendriks et al. 2010), PD becomes even more important after teachers enter their profession (Darling-Hammond et al. 2009; Darling-Hammond, Hyler, and Gardner 2017; OECD 2019). In particular, combinations of various PD activities (e.g. workshops, coaching, observation visits) are viewed as effective and sustainable for learning (Borko, Jacobs, and Koellner 2010; Darling-Hammond, Hyler, and Gardner 2017; Stein, Smith, and Silver 1999). This led to a call for teachers to pursue diversified PD (Borko, Jacobs, and Koellner 2010; Desimone and Garet 2015; Darling-Hammond, Hyler, and Gardner 2017; Powell and Bodur 2016). As diversified PD participation, we define the combination of different PD activities teachers participate in within a specific period (e.g. 12 months; referred to as PD participation pattern [PDPP] in the following).

Notably, previous research often focused on teachers’ participation in a specific PD activity; only a few studies examined diversified PD participation. As a result, little is known about changes in teachers’ PDPP over time. However, insights may help understand (a) which teachers show more effective PDPPs (i.e. a PDPP in which teachers learn more or better) and (b) how teachers may be encouraged to switch from less effective to more effective PDPPs, which this study intends to understand.

Theoretical framework

Changing professional development paradigms

The structure of PD offers is affected by ideological (e.g. different pedagogical approaches), practical (e.g. availability of funding, time, and other resources), cultural (e.g. cultural beliefs about teaching and learning), and historical (e.g. historical perceptions of the teaching profession) factors (Borko, Jacobs, and Koellner 2010; Darling-Hammond 2021; Guskey 1986; Hargreaves 2000; Stein, Smith, and Silver 1999). This study focuses on PD paradigms from an evidence-based perspective regarding effective learning processes, which is also fed by the various impinging factors.

In past decades, different PD paradigms emerged based on different underlying theoretical assumptions about learning processes (Borko, Jacobs, and Koellner 2010; Darling-Hammond and McLaughlin 2011; Darling-Hammond, Hyler, and Gardner 2017). Within traditional PD paradigms, learning is often assumed to be transmissive with knowledge passed from one person to another. Teacher educators take more active roles and teachers more passive roles (Avalos 2011; OECD; Blömeke and Olsen 2019), and teachers mostly learn alone instead of in a community (Borko, Jacobs, and Koellner 2010). As underlying theoretical assumptions have changed, also PD paradigms were modernised. Current PD paradigms are based on cognitive constructivist understandings of learning as active, (co-) constructive, and self-controlled processes (Borko, Jacobs, and Koellner 2010; Darling-Hammond, Hyler, and Gardner 2017; Powell and Bodur 2016).

These paradigms are reflected in teachers’ PDPP, i.e. the composition of types of PD formats teachers used within a specific period. In traditional PD paradigms, teachers’ PDPP is characterised as less diversified as teachers often participate exclusively in single-day PD activities like workshops or courses (Borko, Jacobs, and Koellner 2010; Darling-Hammond, Hyler, and Gardner 2017; Desimone and Garet 2015; Lindvall and Ryve 2019; OECD 2019). Moving from participation in a single PD format like one-off workshops to more diversified PDPP increases the probability of the inclusion of more intensive PD activities with longer duration or more community-based learning (Borko, Jacobs, and Koellner 2010; Darling-Hammond, Hyler, and Gardner 2017; Desimone and Garet 2015; OECD 2019). Notably, prior studies suggest that more diversified PDPPs are more sustainable for teachers’ learning (Borko, Jacobs, and Koellner 2010; Darling-Hammond, Hyler, and Gardner 2017; Desimone and Garet 2015; Hendriks et al. 2010; Jensen et al. 2016, OECD 2019; Powell and Bodur 2016), for example, regarding changes in teaching practices, such as identifying strategies to address students’ needs (Kraft, Blazar, and Hogan 2018).

Characteristics associated with professional development participation

Internationally, a relatively homogeneous picture emerges as almost all teachers participate in at least one PD activity within a year (Darling-Hammond et al. 2009; OECD 2019). However, few teachers engage in PD on a regular and extended basis across their professional careers (Dede and Eisenkraft 2016). Most teachers tend to participate in singular, short-duration PD activities like workshops or courses (Darling-Hammond et al. 2009; Darling-Hammond, Hyler, and Gardner 2017; Desimone, Smith, and Ueno 2006; Murray 2012; OECD 2014; 2019) rather than enrol in intensive PD activities of longer duration (e.g. coaching, mentor programmes, observation visits; for an international comparison see OECD 2009, 2019). These PDPPs seem relatively stable over several years (i.e. more than a decade, if findings from the OECD studies in 2009 and 2019 are considered), and the extent to which teachers are more likely to participate in workshops compared to other PD activities is internationally comparable (OECD 2019). Studies that investigated teachers’ PDPP within a specific period (e.g. teachers are often asked about their PD activity in the past 12 months) revealed that the range of PD activities is limited, with teachers attending on average two to four different PD activities in a given year (Darling-Hammond et al. 2009; Kwakman 2003; OECD 2019).

Darling-Hammond et al. (2009) argued that such a less diversified PDPP may indicate an insufficiently diversified availability of PD activities. However, as many teachers are responsible for selecting their PD activities – besides potentially obligatory PD activities—(Desimone, Smith, and Ueno 2006), information on key levers – beyond the structure of PD offers – to encourage teachers to participate in different PD activities is crucial. Teachers’ actual participation in PD activities is the result of a complex interplay of the availability of PD activities and their quality, teachers’ individual needs (e.g. prior knowledge) and wants (e.g. their interests), teachers’ self-regulation abilities (e.g. ability to assess one’s strengths and weaknesses), teachers’ motivation and will to engage in PD, current working conditions (e.g. release from teaching to attend PD), or whether participation in PD is mandatory or voluntary.

Teacher characteristics

D. Richter et al. (2010) showed that teachers with stronger constructivist beliefs regarding learning processes participate in more PD. Transmissive beliefs of teachers are based on traditional learning theories, viewing students as passive receivers of information. Constructivist beliefs are based on assumptions about the importance of active and constructive learning principles in social contexts. Learning theory beliefs are important as students in classes of teachers who have strongly transmissive beliefs seem to be less cognitively challenged and less supported in their learning process than students in classes of teachers who have less transmissive beliefs (Dubberke et al. 2008). It is assumed that teachers who view student learning as a discursive process also have similar ideas about their own professional learning and take a more active approach to the learning opportunities offered (D. Richter et al. 2010).

Zhang et al. (2021) found that teachers’ self-efficacy is associated with teachers’ motivation to participate in learning activities. Furthermore, Krille (2020) reported associations between teachers’ perceived job workload and their PD activity, and D. Richter et al. (2013) found associations between the degree of collaboration of teachers with colleagues and their engagement in PD. In addition, teachers with higher work engagement and teachers holding more responsibilities (e.g. service or management responsibilities) tend to be more active in PD (Krille 2020; OECD 2014; D. Richter et al. 2011; E. Richter, Richter, and Marx 2018).

Studies focusing on teachers’ teaching experience revealed mixed findings. Whereas D. Richter et al. (2011) found evidence that teachers seem to participate in PD activities most often in the middle of their career, Birman et al. (2007) showed that teachers with less professional experience invest more time in PD than teachers with more years of professional experience (Hauk et al. 2022) and Desimone et al. (2006) found that more experienced teachers are more likely to participate in sustained PD (i.e. PD that takes 16 hours or more).

Contextual characteristics

Darling-Hammond et al. (2009) argued that less diversified participation in PD activities may indicate an insufficiently diversified PD supply structure (i.e. availability of only short-term activities; see also Hill 2009; Yoon and Kim 2022). This argument is plausible as costs are high when teachers are absent from class teaching to participate in long-duration PD activities; hence, PD activities tend to be arranged as short-term opportunities (Powell and Bodur 2016). This argument is supported by findings that PD providers spend the most money on workshops (OECD 2014) and that many teachers are not given the opportunity to invest more than two days per year in PD (Darling-Hammond et al. 2009). Moreover, recent studies indicated that the price and proximity of PD are important contextual barriers for teachers to participate in PD activities (McCoy et al. 2020). Therefore, contextual characteristics like support and financial compensation for teachers are important when encouraging teachers’ PD participation (OECD 2019).

However, previous research also suggests that time constraints (e.g. PD that fits teachers’ schedule) and whether teachers are granted paid temporary leave from teaching (i.e. ensuring that teachers can participate in PD without losing pay and without having to cancel classes; we call this in our study PD leave) are more important for teachers’ participation in PD than financial incentives (Hendriks et al. 2010; Krille 2020; OECD 2014; 2019). In particular, schedule conflicts pose barriers to PD participation (Mitchell and Peters 1988) and are becoming an increasingly important factor in many countries (visible by comparing findings from Teaching and Learning International Survey (TALIS) 2013 and 2018: OECD 2019).

In this context, the importance of school leaders who actively support teachers’ professional learning becomes evident (Desimone 2009; Postholm 2012). Besides shaping a school culture that encourages teachers to participate in PD, school leaders can be more or less successful in creating working conditions that allow opportunities for participation in professional development (Bredeson 2000; Clement and Vandenberghe 2001; Whitworth and Chiu 2015). School leaders are important because they ultimately decide if teachers are allowed to participate in PDs or not if they take place during working hours.

In summary, both teacher and context characteristics are important to explain teachers’ participation in PD. These characteristics are not independent (Lipowsky and Rzejak 2015). If teachers are engaged in their school, for instance, by voluntarily taking on additional tasks, then it can be assumed that they also perceive a higher job workload and are more likely to have scheduling conflicts. This further emphasises the importance of the respective support structure for PD participation.

Research questions

This study aims to address the following research questions (RQs):

(RQ1) What are teachers’ PD participation patterns?

(RQ2) What teacher and contextual characteristics are associated with specific PD participation patterns?

(RQ3) How do teachers’ PD participation patterns change, and what are teacher and school contextual characteristics related to these changes?

Method

This study used data from the National Educational Panel Study (NEPS) Starting Cohort Grade 5 and Starting Cohort Grade 9 (Blossfeld, Roßbach, and von Maurice 2011), a large-scale study in Germany to study the effects of education on the life course and describe educational processes across the lifespan. NEPS is based on a multicohort longitudinal sequence design in which multi-stage cluster sampling is utilised to obtain a representative sample of students from the general educational school system. Questionnaires and competence tests are used to assess a broad set of educational constructs. In this study, we considered data from the first three waves of the NEPS survey (2010–2012; referred to as measurement points t₁-t₃) in each of the two starting cohorts in Grade 5 and Grade 9. During the annual surveys, German language and Mathematics teachers and their school principals were invited to complete written paper-pencil questionnaires. Teachers provided information about themselves and the classes they taught, whereas school principals provided information about their respective schools.

Sample

We considered data from N = 3,539 teachers from Germany. Notably, most German states do not require teachers to take a mandatory number of PD courses so that teachers are not compelled to participate in PD activities to renew their teaching licences (D. Richter, Kleinknecht, and Gröschner 2019; Rzejak and Lipowsky 2020). Therefore, teachers often freely choose whether to participate in PD based on the PD that is provided by institutions of the states, by schools, or by private PD providers. The range of PD courses in Germany has not yet been compiled and analysed on a large scale. A first study for in one of the largest German states of Baden-Württemberg from 2019 showed that almost 7,500 PD courses were offered in a single year (Cramer, Johannmeyer, and Drahmann 2019).

Data were available from n_t1 = 2,238 teachers at the first measurement point (t₁), n_t2 = 1,701 teachers at the second measurement point (t₂), and n_t3 = 785 teachers at the third measurement point (t₃). Of the 2,238 teachers who participated at t₁, 914 teachers also participated at t_2, and of these, 202 at t₃. About 63% of teachers were female, and the average age of teachers was M = 44.28 years (SD = 11.33). Teachers reported having, on average, M = 18.31 (SD = 12.07) years of prior teaching experience.

Measures

Dependent variables

The main variables of our analysis were teachers’ participation in PD activities (courses/workshops, educational conferences/seminars, qualification programmes, observational visits, working groups, research work, mentor programmes/training programmes; full description in the online supplement Table S1). These seven items encompass those PD activities that were also surveyed in prominent other studies on PD, such as the Schools and Staffing Survey (SASS), the Teacher Questionnaire 1990–2012 (Cox et al. 2016), and the TALIS (OECD 2019). On these seven dichotomous items, teachers stated whether they had ( = 1) or had not ( = 0) participated in different PD activities during the past 12 months (excluding any initial teacher training if it occurred within the past 12 months) for each PD activity.

Teacher characteristics as independent variables

Demographics

Teacher demographic variables include gender, age, teaching experience, high school Grade Point Average (GPA), and university GPA. Grades were measured as a numeral with one decimal place, and valid values ranged from 1.0 (very good) to 4.0 (sufficient), whereby higher scores represented lower achievement. Table S2 provides descriptions of these variables and their descriptive statistics.

Learning theory beliefs

Learning theory beliefs describe how teachers perceive teaching and learning processes including constructivist and transmissive beliefs (Kunter et al. 2013). We assessed both learning theory beliefs with a single item, each on a 4-point scale ranging from 1 (completely disagree) to 4 (completely agree).

Teacher self-efficacy

According to Skaalvik and Skaalvik (2007), teacher self-efficacy is a multidimensional construct, whereby this study focused on the Instruction dimension (i.e. teachers’ beliefs about the impact of teaching on students’ academic achievement; for a classification alongside other dimensions, see Fackler, Malmberg, and Sammons 2021). We used a scale consisting of three items on a 4-point scale ranging from 1 (very unimportant) to 4 (very important) to assess teachers’ self-efficacy regarding teaching-related factors (e.g. teaching quality) on students’ academic achievement. The internal consistency (Cronbach’s alpha) was satisfactory (α = .67).

Workload

We measured the perceived workload during the preparation of school lessons with a single item on a 5-point scale ranging from 1 (not stressful at all) to 5 (very stressful).

Collaboration

We measured the extent of collaboration among colleagues with a scale consisting of three items that assessed different working activities on which teachers can typically either work alone or cooperate (e.g. preparing teaching/learning materials). Teachers were asked to rate the degree of collaboration on a scale ranging from 1 (never) to 6 (weekly). The internal consistency was fairly high (α = .83).

Work engagement

To measure teachers’ general professional engagement, we used a scale composed of seven items assessing how often teachers participated in professional activities (e.g. meetings and discussions regarding the school’s perspectives and mission). Teachers rated each item on a 6-point scale ranging from 1 (never) to 6 (weekly). The internal consistency was satisfactory (α = .70).

Demand of PD

We assessed teachers’ demand of PD with a scale consisting of four items asking for the extent of one’s personal need for PD in different topics (e.g. classroom management, expert knowledge, assessment methods). Teachers rated each item on a 4-point scale ranging from 1 (no need at all) to 4 (great need). The internal consistency was satisfactory (α = .71).

Days of PD participation

We used a single item to assess the number of days teachers participating in PD activities in the last 12 months.

School contextual characteristics as independent variables

We used three dichotomous items to assess school contextual characteristics, related to the structure to support teachers’ PD activities. The first item assessed if costs for PD activities were reimbursed ( = 1) or not ( = 0) during the past 12 months. The second item measured whether teachers got financial compensation for PD activities ( = 1) or not ( = 0) during the past 12 months. The third item assessed if teachers were granted PD leave during the past 12 months ( = 1) or not ( = 0). Table S3 lists all items of all scales used.

Statistical analyses

Latent class analysis (RQ1)

We applied latent class analyses (LCA) to examine whether we are able to identify the non-diversified and highly diversified PDPP of teachers (Geiser 2013; Porcu and Giambona 2017). We examined the frequencies at t₁ of teachers who reported participating in a specific PD activity in the past 12 months. Therefore, we considered all available PD information on activities listed in Table S1. We determined the number of latent classes following the recommendations of Nylund-Gibson and Masyn (2016) and Clark and Muthén (2009) to conduct the LCA without covariates. Next, we categorised manifest response patterns at t₂ and t₃ based on the findings of the LCA for t₁ to ensure comparability of classes across the time points. The nominal coded variable CLASS was generated for the assignment of a teacher to the latent PD participation class. To ensure comparability of class assignments across all measurement points, we decided to categorise t₂ and t₃ similarly to t₁ (i.e. we assigned teachers from t₂ and t₃ using the same categorisation, which resulted from LCA using t₁ data).

Multiple group models and multivariate multiple linear regressions models (RQ2)

We used multiple group models to compare the means of all assessed teacher characteristics between the PD participation classes. We estimated effect sizes of mean differences between latent classes using Cohen’s d. Then, we specified multivariate multiple linear regression models with class probabilities (CLnPB)¹ as dependent variables to predict teachers’ PD participation classes using teacher characteristics and school contextual characteristics as predictor variables (Clark and Muthén 2009; Porcu and Giambona 2017). Teachers were assigned three class probabilities (one for each class respectively). We addressed the unreliability of scales to measure teachers’ beliefs and work practices (i.e. self-efficacy, collaboration, work engagement, demand of PD) using latent variable modelling. Differences in the predictive power of regression coefficients when predicting the three class probabilities were tested using the delta method as implemented in the model constraint option in Mplus 8 (Muthén and Bengt 1998-2017).

Multiple logistic regression models (RQ3)

We generated a dichotomous variable (CLASS_CHANGE) by comparing the values of the variable CLASS across all measurement points to examine teachers’ individual changes in classes across the three measurement points. CLASS_CHANGE indicated whether a teacher changed ( = 1) or did not change ( = 0) PD participation class membership at least once across measurement points. The direction of change was examined using two dichotomous variables: ASCENT and DESCENT. ASCENT indicated whether a teacher changed from a lower to a higher class or whether a teacher was assigned the highest class across all measurement points ( = 1) or not ( = 0). DESCENT indicated whether teachers changed from a higher to a lower class or whether they were assigned the lowest class across all measurement points ( = 1) or not ( = 0). These variables were used as dependent variables in separate multiple logistic regression models that included all teacher and school contextual characteristics as predictor variables.

The analyses included only teachers who participated in the questionnaire survey, either at only two of the three measurement points or at all three measurement points. Time points at which the teachers provided data were controlled by including the two dichotomous variables DATA12 and DATA23. Data preparation was conducted in $\mathit{R}$ version 3.6.0 (R Core Team 2017). Latent class, multiple group, and regression analyses were run in Mplus 8 (Muthén and Muthén 1998–2017).

Nested data

We calculated intraclass correlation coefficients (ICCs) to estimate the nesting of teachers within schools. ICCs of dependent variables (CLnPB) ranged from .002 to .044. As even low ICCs can lead to biases in the results of significance tests in regression analyses (Geiser 2013) and the multi-level structure was merely a nuisance factor, we estimated cluster-robust standard errors to account for the nesting in our models (McNeish, Stapleton, and Silverman 2017).

Missing values

Missing values existed in the cross-sectional latent class and regression analyses (RQ1, RQ2) due to item non-response. For instance, at t_1, approximately 8% of the values were missing. Missing values also existed in the longitudinal regression analyses (RQ3) due to panel attrition (see sample sizes in Figure 1). To address the missing data, we employed full information maximum likelihood (FIML) estimation in all our analyses. This approach typically outperforms traditional missing data methods, such as listwise deletion (Graham 2012; van Buuren 2018).

Figure 1. Teachers’ intraindividual changes between PDPP.

Note. PD = teacher professional development. The numbers at each path indicate the number of teachers who changed their PD participation pattern (PDPP). The information on sample sizes has the following meaning: sample size at each measurement point (cross-sectional)/size of the sample of teachers who participated at all previous measurement points (longitudinal)/size of the sample of teachers who participated at all previous measurement points (longitudinal) and for which a latent class could be assigned (bold).

Results

Identifying teachers’ professional development patterns (RQ1)

Preliminary analyses and LCA

First, we examined the frequencies of teachers who reported participating in specific PD activities across all measurement points (Table S4). For instance, at t_1, 76% of teachers reported participating in courses/workshops, and 43% reported participating in educational conferences/seminars in the past 12 months.

Next, we conducted latent class analyses to determine teachers’ PDPP. We evaluated and compared statistics of four different latent class models using Pearson χ², likelihood ratio statistics, and bootstrap likelihood ratio χ² difference tests. Table S5 shows that the degree of misfit between all latent class solutions with more than two classes seems negligible as the differences between x² statistics and degrees of freedom (df) were small.

Therefore, we considered all class solutions with more than two classes trustworthy. However, we could not identify a model that outperforms the other models solely based on model fit indices. Thus, our modelling decision for a three-class solution was guided by five additional considerations: (a) reliability of mean class allocation probability; (b) Bayesian Information Criterion (BIC); (c) the entropy; (d) interpretability of classes; and (e) the model parsimony (Geiser 2013; Nylund, Asparouhov, and Muthén 2007).

Characterization of the latent classes

To assess differences between non-diversified and highly diversified PDPP, we examined (at t₁) class-related response probabilities of chosen PD activities in the past 12 months. Figure 2 illustrates that teachers assigned to different latent classes differed in their diversification of PDPPs.

Figure 2. Item response probabilities conditional upon latent class membership.

Note. n = 2,238 teachers. PDPP = professional development participation pattern; PD1 = courses/workshops; PD2 = educational conferences/seminars; PD3 = qualification programmes; PD4 = observational visits; PD5 = working group for PD; PD6=research work; PD7=mentor programmes/training programmes.

Teachers assigned to the first latent class (n = 664) had the highest response probability (45%) for courses/workshops, which was substantially higher compared to the other six PD activities. Teachers assigned to the second latent class (n = 1,532) differed by overall higher response probabilities for all seven PD activities. These teachers had the highest probability of participating in courses/workshops (87%), educational conferences/seminars (54%), and working groups (48%). Teachers assigned to the third latent class (n = 42) showed an overall higher probability of participating in all PD activities. When comparing teachers who were assigned to the third latent class with teachers assigned to the second latent class, differences related to the probabilities of participating in qualification programmes (65% Class 3 vs. 10% Class 2) and mentor programmes/training programmes (47% Class 3 vs. 14% Class 2).

Response probabilities of the three latent classes were also reflected in manifest response patterns and their assignment to the three latent classes (Table S6).

We designated teachers who were assigned to the first latent class as teachers with a non-diversified PDPP, to the second latent class as teachers with a low diversified PDPP, and to the third latent class as teachers with a highly diversified PDPP.

Teacher and contextual characteristics and teachers’ PDPP (RQ2)

To better understand teacher and contextual characteristics across the three different PDPPs, we investigated interindividual mean differences of central teacher and contextual characteristics (Table S7). We found a variety of statistically significant differences between teachers with different PDPPs. Regarding teacher characteristics, teachers with different PDPPs differed, for instance, regarding self-efficacy, perceived workload, collaboration, and general work engagement. Regarding contextual characteristics, teachers with different PDPPs differed in all assessed characteristics (i.e. cost absorption, financial compensation, PD leave). Compared to teachers with a low or high diversified PDPP, teachers with a non-diversified PDPP showed lower self-efficacy (|d| = 0.12, p = .009; |d| = 0.44, p = .007), had lower values on collaboration (|d| = 0.21, p < .001; |d| = 0.64, p < .001), and exhibited lower general work engagement (|d| = 0.43, p < .001; |d| = 1.15, p < .001). In addition, teachers with a non-diversified PDPP experienced more workload compared to teachers with a highly diversified PDPP (|d| = 0.34, p = .015).

Compared to teachers with a high diversified PDPP, teachers with a low diversified PDPP had lower values on collaboration (|d| = 0.45, p = .010) and exhibited lower general work engagement (|d| = 0.73, p < .001).

Overall, teachers with a more diversified PDPP invested more days in PD activities than teachers with a less diversified PDPP. We conducted multivariate multiple linear regression analyses to examine the association between teacher and school contextual characteristics with teachers’ PDPPs. First, we examined the statistics of the probabilities. For all three PDPPs, the range was almost completely exhausted ranging from 0.00 to 1.00. In addition, we found that teachers were most likely to be assigned to the low diversified PDPP (CL1PB: M = 0.26, SD = 0.32; CL2PB: M = 0.71, SD = 0.32; CL3PB: M = 0.03, SD = 0.13). Second, we investigated the resulting regression coefficients for teacher and school contextual characteristics (Table 1).

Table 1. Regression coefficients of class probabilities on teacher characteristics and school contextual aspects.

Predictors	Dependent variable
	CL1PB			CL2PB			CL3PB
Teacher	β	SE	p	β	SE	p	β	SE	p	Δp12	Δp23	Δp13
Gender^a, b	−.019	.021	.375	.030	.021	.156	−.028	.022	.200	.237	.096	.769
Age^b	−.007	.047	.873	−.013	.048	.780	.052	.045	.246	.950	.534	.588
High School GPP^b	.068	.027	.011	−.045	.027	.103	−.057	.027	.034^†	.034^†	.507	.003
University GPA^b	−.018	.025	.485	.013	.026	.611	.011	.024	.640	.539	.773	.432
Teaching experience^b	−.014	.047	.767	.035	.049	.474	−.053	.047	.258	.602	.329	.885
Transmissive Belief ^b	.016	.020	.445	−.001	.021	.948	−.035	.024	.147	.675	.623	.218
Constructivist Belief ^b	−.024	.021	.256	.016	.021	.440	.019	.022	.388	.329	.724	.201
Self-Efficacy^b	.015	.030	.625	−.028	.031	.361	.034	.032	.297	.471	.257	.976
Workload^b	−.006	.021	.774	.023	.021	.281	−.042	.019	.027^†	.485	.100	.639
Collaboration^b	.071	.038	.059	−.056	.037	.127	−.037	.032	.255	.083	.306	.044^†
Work Engagement^c	−.284	.048	<.001	.207	.050	<.001	.191	.049	<.001	<.001	.031^†	<.001
Demand of PD^c	−.007	.025	.774	.018	.026	.483	−.027	.028	.336	.610	.354	.897
Days of PD^c	−.212	.027	<.001	.141	.029	<.001	.175	.040	<.001	<.001	.106	<.001
School Contextual
Cost Absorption^a, c	.054	.030	.067	−.044	.030	.145	−.026	.040	.223	.096	.309	.040^†
Financial Compensation^a, b	−.040	.022	.070	.024	.023	.308	.041	.025	.093	.150	.803	.026^†
PD Leave^a, c	−.218	.028	<.001	.203	.028	<.001	.038	.019	.038^†	<.001	<.001	<.001
R^2	.198	.021	<.001	.117	.019	<.001	.091	.018	<.001
R^2adj	.192			.111			.084

Note. n = 2,238. All variables are assessed at t1. CL1PB = probability to be assigned to a class of teachers with a non-diversified PDPP; CL2PB = probability to be assigned to a class of teachers with a low diversified PDPP; CL3PB = probability to be assigned to a class of teachers with a high diversified PDPP.a Dichotomous variable. b Predictors were measured at t₁.c Predictors were measured at t₂. ^†Not statistically significant after Benjamini and Hochberg (1995) correction.

Statistically significant predictors for assignment to a non-diversified PDPP included high school GPA (β = .07, p = .011), work engagement (β = -.28, p < .001), and PD leave (β = -.22, p < .001). Teachers with a better high school GPA, showed less general work engagement, and were not granted PD leave were more likely to be assigned to the class of teachers with a non-diversified PDPP. Statistically significant predictors for assignment to a low diversified PDPP included general work engagement (β = .21, p < .001) and PD leave (β = .20, p < .001). Teachers who showed more general work engagement and were granted PD leave were more likely to be assigned to the low diversified PDPP. Statistically significant predictors for assignment to a highly diversified PDPP included higher general work engagement (β = .19, p < .001).

Changes in teachers’ professional development participation patterns over time (RQ3)

To describe temporal changes in teachers’ PDPP across the three measurement points, we examined the number of teachers per PD activity at each measurement point. From a cross-sectional perspective – which was taken in most prior studies – it is reasonable to assume that the ratio between the number of teachers who did and number of teachers who did not participate in a specific PD activity remained stable for all PD activities across all measurement points (Table S4). At all measurement points, teachers participated most frequently in courses/workshops (t₁: 76%; t₂: 75%; t₃: 73%), and least frequently in qualification programmes (t₁: 10%; t₂: 13%; t₃: 13%) and mentor programmes and/or training programmes (t₁: 11%; t₂: 11%; t₃: 9%). Similarly, the relative distribution of teachers among the three classes also remained relatively stable over the three measurement points (Figure S1). Therefore, from a cross-sectional perspective, it seems reasonable to assume that overall, teachers may change their PDPP over time only to a small extent.

However, when examining teachers’ intraindividual changes in PDPP, we found that many teachers changed their PDPP over time (Figure 1). About 34% of teachers changed their PDPP from the first to the second measurement point (n = 299; 147 teachers changed from a low diversified to a non-diversified PDPP). About 38% of teachers also changed their PDPP from the second to the third measurement point (n = 56; 24 teachers changed from a low diversified to a non-diversified PDPP). Notably, a change of two classes (e.g. from the third to the first class) was only observed for only three teachers.

Next, we examined variables associated with these changes by conducting multiple logistic regression analyses (Table 2). Besides the number of days participating in PD (β = -.44, p = .001), teachers’ general work engagement was significantly associated with changes in PD patterns. Teachers who stated that they were generally more engaged regarding work (β = -.39, p < .001) tended to change their PDPP more frequently from a non-diversified to a low or high diversified PDPP and less frequently from a low to a high diversified PDPP. Even though statistical significance was absent after correction, according to Benjamini and Hochberg (1995), considering equivalent findings from previous studies using cross-sectional data (e.g. Hendriks et al. 2010; OECD 2014; 2019), it is worth mentioning that PD leave may also be an important contextual characteristic for changes of teachers’ PDPP. For instance, when analysing the changing of classes independent of the direction of change only PD leave was a significant predictor (Table S8).

Table 2. Standardised logits of class changes between the three measurement points on teacher characteristics and school contextual aspects.

Predictors	Dependent variable
	ASCENT			DESCENT
Teacher	β	SE	p	β	SE	p
Gender^a, b	.051	.050	.304	−.020	.038	.587
Age^b	.069	.102	.498	−.047	.096	.625
High School GPP^b	.007	.069	.915	.060	.041	.170
University GPA^b	−.057	.059	.330	.018	.044	.682
Teaching experience^b	−.144	.103	.160	.083	.096	.390
Transmissive Belief ^b	.042	.047	.373	.002	.041	.966
Constructivist Belief ^b	−.017	.057	.765	−.007	.040	.868
Self-Efficacy^b	−.085	.072	.240	.042	.053	.421
Workload^b	.033	.050	.504	−.041	.040	.311
Collaboration^b	−.113	.072	.117	.008	.052	.872
Work Engagement^c	.130	.079	.100	−.392	.077	<.001
Demand of PD^c	−.004	.054	.944	.001	.045	.986
Days of PD^c	.035	.052	.501	−.435	.133	.001
School Contextual
Cost Absorption^a, c	−.008	.057	.893	.009	.044	.847
Financial Compensation^a, b	−.033	.064	.599	−.006	.044	.894
PD Leave^a, c	−.102	.052	.051	−.104	.044	.018^†
Control of Measurement Point
DATA12	−.027	.057	.631	−.042	.049	.386
DATA23	.111	.051	.028^†	−.060	.047	.207
R^2	.067	.030	.023	.389	.096	<.001

Note. n = 914. ASCENT indicates whether a teacher changed from a lower to a higher class or stayed in the highest class across all three measurement points (=1) or not (=0). DESCENT indicates whether a teacher changed from a higher to a lower class or stayed in the lowest class across all three measurement points (=1) or not (=0). a Dichotomous variable. b Predictors were measured at t₁. c Predictors were measured at t₂. ^†Not statistically significant after Benjamini and Hochberg (1995) correction.

Discussion

Prior research suggested that non-diversified PDPP, such as the sole participation in one-shot workshops of short duration, were ineffective for teacher learning (e.g. Darling-Hammond, Hyler, and Gardner 2017). Instead, teachers were encouraged to engage in more diversified PD activities, including intensive and long-term programmes (Borko, Jacobs, and Koellner 2010; Desimone and Garet 2015; Darling-Hammond, Hyler, and Gardner 2017). However, research on aspects that may support teachers in applying more diversified PDPP is limited. This study contributes to this research base by investigating teachers’ PDPP from both a cross-sectional and a longitudinal perspective.

On an individual PD level, we found that most teachers (more than 70%) participated in courses/workshops, whereas teachers participated less frequently in more intensive and time-consuming PD activities such as qualification programmes or mentor programmes/training programmes aligned with prior research (Darling-Hammond et al. 2009; Darling-Hammond, Hyler, and Gardner 2017; Desimone, Smith, and Ueno 2006; OECD 2014; 2019). As these findings are based on single PD activities, this study goes beyond what is traditionally analysed in the teacher education literature by analysing the entire range of PD activities teachers participate in each year. We were able to show that around 30% of teachers applied a non-diversified PDPP, whereas only around 2% applied a highly diversified PDPP, indicating that intensive and long-term PD activities are rarely part of teachers’ PDPP. That mirrors studies that examined simultaneous participation in PD activities (Darling-Hammond et al. 2009; Kwakman 2003). From a cross-sectional perspective, the distribution of teachers across the three PDPPs remained stable for each measurement point. However, this study also contributes to the limited research base that utilises longitudinal data revealing that 30% to 40% of teachers change their PDPP within three years at least once. Notably, regarding teacher and contextual characteristics, we found that teachers’ general work engagement and their opportunity for PD leave (an important contextual characteristic) were of particular importance for both the PDPP at a specific time and changes in PDPP over time. The higher teachers’ general work engagement and if teachers were granted PD leave, the more likely teachers were to show a more diversified PDPP. Furthermore, these two characteristics also prevented teachers from downgrading their PDPP from more diversified to less diversified over time. The importance of teacher general work engagement as an influencing characteristic for teacher participation in PD aligns with findings from previous studies (OECD 2019; E. Richter et al. 2021). The reasons for our findings are manifold and difficult to ultimately trace back to specific causes as our study does not allow causal claims. For instance, it seems reasonable that teachers do not show a highly diversified PDPP because the requirements (offering PD leave) are not fulfilled. PDPP requires a greater time investment in PD activities and, thus, planning by the school how PD leave can be compensated. Notably, teacher shortages might foster schools to very selectively offer teachers to participate in multiple different PD programmes. In addition, specific geographical regions might decide not to offer the whole range of PD activities required for diversified PD. Hence, challenging structural characteristics might reciprocally influence one another.

It seems reasonable that teacher characteristics are at least as important as contextual characteristics in explaining why teachers rarely participate in diversified PD. We found that teacher engagement is a key characteristic for predicting diversified PD participation. This finding might suggest that teachers’ attitudes towards a diversified PDPP under challenging conditions (e.g. teacher shortage, reduction in teaching time per student, family involvement) might be of similar relevance when explaining the small number of teachers showing a diversified PDPP. Notably, we did not assess all potentially relevant variables in this regard (e.g. personality, attitudes towards PD). Considering a more extensive set of characteristics in explaining teachers’ PDPPs constitutes a fruitful avenue for future research.

Contrary to our expectations, some teacher characteristics we included in our study to predict teachers’ PDPPs were not statistically significant. For instance, some studies suggests that teaching experience (e.g. Hauk et al. 2022; D. Richter et al. 2013; Zhang, Admiraal, and Saab 2021) and teachers’ gender (Siddiq, Scherer, and Tondeur 2016) relate to teachers’ PD behaviour. In our study, neither teaching experience nor gender was a statistically significant predictor of the diversification of PD behaviour and the change in diversification over time. However, other recently published studies found these teacher characteristics also unimportant (e.g. regarding teachers’ intention to participate in technology-related PD, see Fütterer et al. 2023).

Implications for education policy and practice

First, we found that highly diversified PDPPs are applied by very few teachers (42 of 2,338 teachers). This raises the question of how educational stakeholders can encourage teachers to apply for more diversified PDPPs. Some might argue for favour mandatory requirements to participate in PD activities to retain teacher licences (D. Richter et al. 2014). However, sustainable participation in PD can seldomly be enforced (Kennedy 2016). As PD participation is always a trade-off between costs and benefits, it seems more promising– based on previous research on aspects that influence participation in PD activities– to establish an effective incentive system for PD participation (see Mitchell and Peters 1988; OECD 2019). For instance, one-shot workshops of short duration are preferred because they do not require substitute teachers to take over teaching for a longer period (Darling-Hammond, Hyler, and Gardner 2017). Furthermore, short duration, timing, and even a short distance to PD location are key aspects for participating in workshops (McCoy et al. 2020; Krille 2020), and schedule conflicts are increasing barriers for teachers to participate in PD activities (Hendriks et al. 2010; OECD 2019).

Second, this study also confirms the heterogeneity in teachers’ PD participation (see Yoon and Kim 2022) and the importance of the fit of PD opportunities into one’s schedule and the demand for PD as predictors of PDPP, whereas financial compensation was found to be less important, which is in line with recent findings (e.g. Kwakman 2003; D. Richter, Kleinknecht, and Gröschner 2019).

Given the importance of PD leave and the extent of a teacher’s general engagement, recommendations for educational policy and practice can be discussed. For instance, our study indicates that it is promising to give teachers who are willing to use a variety of different PD activities the support to do so by reducing their workload or by providing PD opportunities suitable to their time schedules. Similar to the systematic literature review by Krille (2020) reducing their workload or providing PD opportunities suitable to their time schedules seems as teacher are often encouraged to participate in PD outside their class time to avoid class cancellation (see also Sellen 2016). Furthermore, along with the development of new technologies in recent years, the potential of online PD becomes increasingly important (Dede and Eisenkraft 2016; Fishman 2016). Online PD can be used flexibly in terms of time and location suiting teachers’ schedules.

Finally, this study indicates that PD leave is more important than financial compensation or cost absorption. Indeed, the ineffectiveness of such financial incentives, which are considered effective in other professions, is well known in countries where teachers’ salaries or opportunities for advancement are restricted (Krille 2020).

Limitations

This study includes predominantly Mathematics and German language teachers in German secondary schools which potentially limits the generalisability. However, prior research examining PDPP across multiple subjects indicated similar results across subjects (Fischer et al. 2020). Although we did not use data from different countries, previous studies suggest that results from Germany may be transferable to other countries (D. Richter, Kleinknecht, and Gröschner 2019). Also, our models – consistent with similar studies (e.g. D. Richter, Kleinknecht, and Gröschner 2019) – are only to explain limited variance in teachers’ PDPP (e.g. 8% of the variance in teachers’ highly diversified PDPP at a specific time point) indicating that important influential variables are missing. For example, we did not have data on motivational variables (e.g. interest in specific PD topics, perceived quality of PD offered; Fütterer et al. 2023; Fütterer, Richter, and Richter 2024) or individual context variables (e.g. family commitments such as childcare). Similarly, we did not have access to variables describing the PD demand and corresponding PD supply structure.

Future research

Future research is encouraged to replicate this work. For instance, large data sets that capture teachers’ PD participation including the TALIS, the Schools and Staffing Survey, and the National Teacher and Principal Survey seem promising candidates for such replication studies. Similarly, future studies should capture broader ranges of variables that may potentially explain teachers’ PDPP. Furthermore, future research should validate the importance of the PDPPs highlighted. Also, it is unclear what a change from a more diversified PDPP in one year to a less diversified one in another year tells us about the overall quality of a teacher’s PDPP throughout a teaching career.

Declaration

This paper uses data from the National Educational Panel Study (NEPS): Starting Cohort 3 - 5th Grade, doi:10.5157/NEPS:SC3:8.0.1 and Starting Cohort 4 - 9th Grade, doi:10.5157/NEPS:SC4:10.0.0. From 2008 to 2013, NEPS data were collected as part of the Framework Program for the Promotion of Empirical Educational Research funded by the German Federal Ministry of Education and Research (BMBF). As of 2014, NEPS is carried out by the Leibniz Institute for Educational Trajectories (LIfBi) at the University of Bamberg in cooperation with a nationwide network.

Acknowledgments

This project is part of the ‘Qualitätsoffensive Lehrerbildung’, a joint initiative of the Federal Government and the Länder which aims to improve the quality of teacher training. The programme is funded by the Federal Ministry of Education and Research.

This research was supported by the Postdoctoral Academy of Education Sciences and Psychology of the Hector Research Institute of Education Sciences and Psychology, Tübingen, funded by the Baden-Württemberg Ministry of Science, Research, and the Arts.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/02619768.2024.2370891

Additional information

Notes on contributors

Tim Fütterer

Tim Fütterer is a postdoc at the Hector Research Institute of Education Sciences and Psychology at the University of Tübingen. His research focuses on effective (adaptive online) teacher professional development, teaching effectiveness, digitalization and digitality, applied artificial intelligence in the educational context, and teachers’ reflective practice.

Nicolas Hübner

Nicolas Hübner is an Assistant Professor at the Institute of Education at the University of Tübingen. In his current research, he is interested in characteristics of effective forms of student assessment, evaluation, and feedback and the comparability and meaning of written and oral grades. Beyond this, he is also interested in questions related to knowledge translation, implementation research, and effective teacher professional development.

Christian Fischer

Christian Fischer is an Assistant Professor at the University of Tübingen. His research is situated in the intersections of STEM education, teacher education, and educational technologies. He is interested in the role of digital technologies to transform teaching and learning. In some of his current projects, he examines teacher change in response to nationwide science curriculum reforms, and student success in higher education settings.

Kathleen Stürmer

Kathleen Stürmer is a professor of effective teaching and learning arrangements at the Hector Research Institute of Education Sciences and Psychology and the Tübingen School of Education (TüSE). Her research interests include professional competence building (in teacher education), visual expertise, teaching effectiveness research, teaching with digital media, and the use of reactive methods to capture teaching and learning processes.

Notes

1. CL = Class; n is a natural number representing the numbering of the latent class; PB = probabilities.