Teacher education and early teaching career

ABSTRACT

This paper investigates the early working careers of graduates from teacher education using Norwegian register data. The longitudinal data follows the individuals up to 12 years after graduation. The analysis is on net attrition since mobility both out of and into teacher positions is taken into account. The findings reveal a gradual net attrition from teacher positions of about 1.3 percentage points yearly, starting 1–3 years after graduation. Some of the attrition is to leadership positions in education, but the main attrition is to jobs outside education. The attrition is mainly among males and graduates with relatively high measured academic ability, and it differs greatly across teacher education types. The differences in the probability of teaching related to gender and academic ability are low shortly after graduation, but increase over time. Teacher education graduates for secondary schooling work as teachers to the lowest degree. These systematic patterns are related to both the demand and supply side of the teacher labour market.

KEYWORDS:

• Attrition
• teacher education
• teacher labour market
• gender

Acknowledgments

I thank Ole Henning Nyhus for help with preparing the data and Bjarne Strøm for numerous discussions. The data availability is part of the project “Quality indicators in higher education”, no. 662406. The analyses and views in the paper are solely my responsibility.

Disclosure statement

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

Notes

1. The limited number of studies from the Nordic countries includes Falch and Strøm (2005) and Karbownik (2020), which are quantitative studies of teacher turnover, Skaalvik and Skaalvik (2011) and Räsänen et al. (2020), who study teachers’ intentions, Smith and Ulvik (2017), who use interviews to investigate motives and resilience, and Flores and Niklasson (2014) on student teachers motivation.

2. The data sources are described in Chapter 3.

3. The register is for employees and thus does not include self-employed persons. Teaching tasks as self-employed can only take the form of private tutoring, which is very rare. In the analyses below, the group denoted ‘not working’ includes self-employment. Notice that while yearly about 7–10% of the graduates are not registered with work (see, Figure 3 below), only 1.3% of the sample are not registered to work at all during the first five years after graduation

4. Notice that due to the sample design, there are relatively few observations with high experience levels. Only graduates from 2001 are observed 12 years after graduation (observed in 2013), only graduates from 2001 and 2002 are observed 11 years after graduation (observed in 2012 and 2013, respectively), etc. This feature of the data implies that 6.6% of the sample have 10–12 years of potential experience, while 36.1% have 0–2 years of potential experience.

5. In the empirical sample, 29.4% and 94.8% of the graduates have at least one grade from the old and the new system, respectively, while 4.6% do not have any grade information. The latter group consists mainly of graduates from one specific institution only providing ECEC teacher education. The regression analyses below add control variables in order to take these complications into account. Dummy variables for missing grade information, having at least one grade from the old system and having at least one grade from the new system are used. The effects of the average grade reported do not, however, depend on whether these three control variables are included or not.

6. Alternative model specifications can explicitly ensure that probabilities are between zero and one by assuming specific functional forms of the model, such as the probit and logit models. In these types of models, the estimated relationships can be interpreted outside the mean values of the included variables, under the assumed distributional function. However, such models give approximately the same effects for mean values, which is the only interest in the present paper.

7. When there are several observations for each individual, the precision of the estimated coefficients will most likely be overestimated without corrections. To account for this feature, all standard errors are estimated by allowing for clustering. I cluster at the municipal level, defined as the residential municipality at the age of 16. There are 475 municipalities in the data. I use a higher level than the individual level in the adjustment of the standard errors in order to take other potential common factors into account.

8. In 2016, 92.1%, 74.8% and 53.8% of the teachers were females in ECEC, compulsory schooling and upper secondary schooling, respectively, which is close to the female shares among the graduates presented in Figure 5a.

9. The number of observations is 50,243 for 2 years since graduation, 34,645 for 6 years since graduation, 13,912 for 10 years since graduation and 4,661 for 12 years since graduation.

10. The difference in the dynamic attrition pattern of male and female teachers is interesting in light of the overall gender imbalance in teaching. Teacher gender and student-teacher gender matching do not seem to affect student achievement (Cho, 2012), which casts doubt of the need for male teachers as role models and father figures (Heinz et al., 2021). However, some scholars argue that female dominance might have other important impacts (Heinz et al., 2022).

Additional information

Notes on contributors

Torberg Falch

Torberg Falch is a professor in Economics at the Norwegian University of Science and Technology. He works on various aspects of the teacher labour market and do empirical analyses on different educational policies and reforms.