Special Education Teacher Attrition and Retention: A Review of the Literature

Definition of Attrition

We included empirical studies in which one or more of the primary research questions or the abstract addressed special education teacher leaving, moving, transferring, or intent to stay or leave, and excluded studies in which attrition arose from a secondary inductive analysis (e.g., Mitchem, Kossar, & Ludlow, 2006). In addition, we included studies of intent to stay or leave since it has been correlated with actual leaving (Gersten, Keating, Yovanoff, & Harniss, 2001), and teachers who want to leave, but cannot, may put forth less effort (Jones, Youngs, & Frank, 2013). We included Jones et al.’s (2013) study of commitment, since they used a measure consistent with intent (e.g., “I would prefer to continue teaching in this school”; Jones et al., 2013, p. 371). We excluded studies that purported to be about attrition but lacked attrition findings. For example, Gehrke and Murri’s (2006) purpose was focused on attrition, but their findings were about special educators’ experiences, not attrition.

Special Education Teacher

We defined special educators as teachers with assignments in special education, including speech-language pathologists. Some studies examined attrition among other educators (e.g., paraprofessionals, general educators), and we included these only if special educators were the majority of the sample (Albrecht, Johns, Mounstevens, & Olorunda, 2009), or if they disaggregated special educators’ findings. For example, we excluded Clotfelter, Glennie, Ladd, and Vigdor (2005), because they aggregated special educators and math/science teachers; in contrast, we included their follow-up study (i.e., Clotfelter, Glennie, Ladd, & Vigdor, 2008) as it disaggregated special educators. We also included studies of other stakeholders’ (e.g., administrators) perspectives on attrition (e.g., Berry, Petrin, Gravelle, & Farmer, 2011).

Types of U.S. Schools

We only included studies in U.S. schools, given variability in national policy contexts and approaches to special education. Countries differ in (a) the characteristics of students served in special education, (b) where students are taught (inclusive vs. segregated schools; UNESCO, 2018), and (c) how teachers are recruited and compensated (Darling-Hammond, Wei, Andree, Richardson, & Orphanos, 2009). These different contexts and factors likely influence teacher attrition. In the United States, we included teachers in public, private, special education, and charter schools, but not preschools, residential, hospital, wilderness, or juvenile justice settings (e.g., Houchins, Shippen, & Cattret, 2004).

Methods and Rigor

We included studies regardless of methodology and rigor, but we identify strengths and weaknesses in our critique, and we take these into account when drawing conclusions. We excluded studies that did not report sufficient methodological detail to judge reliability and generalizability (quantitative studies), or trustworthiness and credibility (qualitative, mixed methods studies). For example, we excluded a qualitative study by Nance and Calabrese (2009) because there was insufficient detail about sampling and analytic processes.

Type of Literature

We focused on peer-reviewed studies, but to reduce publication bias (Gage, Cook, & Reichow, 2017), we included grey literature (e.g., foundation reports) if data collection, analytic processes, and findings were included (i.e., Feng & Sass, 2017). We excluded unpublished work, except for Gilmour and Wehby (in press), as it received a national award.

Electronic Search

We conducted a systematic search of the following databases in September 2017: Academic Search Complete, Business Source Premier, Education Full Text, ERIC, PsycINFO, and the Psychology and Behavioral Sciences Collection. Search terms included all combinations of the terms: (a) attrition (and related terms: retention, commitment, satisfaction, career intentions, career plans, intent to leave, intent to stay, mobility) and (b) special educat* (and related terms: students with disabilities). This yielded more than 18,000 articles that used some combination of the specified terms, after exact duplicates were removed.

We exported the title and abstract of each article into an Excel file, coding each article as either possibly or not meeting criteria. We erred on the side of obtaining false positives; for example, studies examining teacher attrition were coded as possibly meeting criteria, even if abstracts did not mention special educators. This yielded 166 studies that possibly met criteria. Each author independently examined the abstract and methods sections of the 166 articles. We made final decisions by consensus, referring to our inclusionary and exclusionary criteria to discuss and resolve discrepancies. This step yielded 25 studies that met inclusion criteria.

Additional Search Efforts

Attrition Factors Studied

As Table 2 shows, the most common attrition factor studied was special educators’ working conditions (e.g., demands, social contexts). The vast majority (80%; n = 24) of studies addressed the relationship between one or more working condition(s) and attrition. Sixteen (53%) investigated aspects of preparation (e.g., initial preparation) and qualifications (e.g., certification). Four studies (13%) addressed the relationship between teacher demographics (e.g., race/ethnicity) and attrition, while 7 (23%) examined special educators retrospective nonwork reasons for leaving or moving. Few examined relationships between school characteristics, such as school demographics and urbanicity, and attrition (n = 3; 10%).

Theories and Frameworks

Only a third of the studies (n = 10) referenced a specific theoretical or conceptual framework. Three used Billingsley’s (1993) model of special education teacher attrition (Lesh et al., 2017; Prather-Jones, 2011a, 2011b), two relied on conservation of resources theory (Bettini et al., 2017; Gilmour & Wehby, in press), and other frameworks were used once, including affective events theory (Jones & Youngs, 2012), ecological systems theory (Lesh et al., 2017; combining two conceptual frameworks), House’s theory of social support (Cancio et al., 2013), human capital theory (Connelly & Graham, 2009), and sense-making theory (Jones et al., 2013).

Methodological Approaches

Studies varied in their definitions/measurement of attrition and retention, the inferences their analyses permit, and their samples’ generalizability (Table 2).

Initial Preparation

The nature of teacher preparation varies considerably in the United States. For example, some programs are lengthier and more comprehensive (often traditional university programs), while others may have far fewer requirements, such as alternative programs (Billingsley & Bettini, 2017). Only three studies considered the relationship of initial preparation to teacher retention. Connelly and Graham (2008) analyzed SASS and TFS data using logistic regression to examine whether characteristics of initial preparation predicted the odds of 168 early career special educators remaining in teaching. They found that new teachers with 10 or more weeks of student teaching were more likely to be teaching 1 year later, with a 21% probability of leaving or moving, while those with less than 10 weeks had a 37% probability of leaving or moving (odds ratio = 2.18, p < .05). Almost 80% of special educators with more than 10 weeks of student teaching were still teaching after 1 year, compared to 63% of those with fewer than 10 weeks. Other aspects of preparation, such as coursework did not significantly predict retention.

In two studies, researchers reported their graduates stayed in special education at high rates, with the inference that quality preparation may enhance retention (Burstein et al., 2009; Edgar & Pair, 2005). Edgar and Pair (2005) conducted phone interviews with 149 of 161 graduates (93%) from seven cohorts of their special education program. They reported that attrition was lower than typical; 78% were still special educators, though 70% changed schools (consistent with high moving rates among new special education teachers; DeAngelis & Presley, 2011). Graduates of the dual certificate program had the highest attrition rate (28%), primarily transferring to general education. In a similar study, Burstein et al. (2009) surveyed graduates of Accelerate Collaborative Teacher, a full-time, year-long, graduate credential program. They reported an average 5-year retention rate of 74%, and a special educator retention rate of 71%, which are similar to Edgar and Pair’s findings. However, the response rate was low (45%) and they did not include systematic comparisons with nonrespondents. No conclusions can be drawn from these two studies given the small, unrepresentative populations and the absence of comparison groups.

Researchers should investigate the extent to which specific aspects of initial preparation (e.g., program comprehensiveness; quality of field experiences) are associated with retention. Furthermore, some special educators are prepared separately from general educators, while others graduate from dual or merged programs in which special and general education teacher candidates complete some or all of the same program components and may earn certification in both general and special education (Blanton et al., 2017). Edgar and Pair (2005) found higher transfer attrition from the dual program, suggesting the need to consider how more than one certification type influences retention. Future research should use administrative data sets for more nuanced investigations, such as Ronfeldt’s (2012) analysis of New York City’s data, which found preservice general educators’ practicum in a low-attrition school predicted longevity, regardless of whether the preservice placement matched in-service school demographics.

Mentoring and PD

Mentoring improves new general educators’ retention (Guarino et al., 2006; Ingersoll & Strong, 2011), but less evidence supports this relationship for special educators (Billingsley, 2004). Most special educators have mentors, but mentors may not be available in the same school or teach in the same area (Billingsley, Griffin, Smith, Kamman, & Israel, 2009). Fewer researchers have studied the relationship between PD opportunities and attrition.

Analyzing SASS/TFS data, Connelly and Graham (2009) found mentoring did not predict attrition among 168 early career special educators; however, they caution that most teachers in the sample reported working closely with mentors, and low variability may limit power to detect effects. Qualitative studies found special educators valued mentorship (e.g., López-Estrada & Koyama, 2010). For example, Gehrke and McCoy (2007) surveyed and interviewed seven stayers and three movers in their first years of teaching. They analyzed data thematically, though the coding process is not clearly described, and they omit key details (e.g., comparing respondents vs. nonrespondents). They reported movers made no references to mentor support, whereas stayers emphasized the importance of formal or informal mentorship.

Examining PD, Cancio et al. (2013) surveyed 408 special educators serving students with EBD, conducting group difference tests to examine factors differentiating those who planned to stay versus leave. They reported that administrators’ support for professional growth (e.g., workshops and opportunities to learn) was significantly higher for those who planned to stay (t = 2.089, p = .037). Similarly, Albrecht et al. (2009) surveyed 776 special educators serving students with EBD, also using group difference tests to determine differences between those planning to stay and leave. They found that “stayers” rated PD opportunities higher than “leavers” (t = 2.922, p < .01). Both of these studies are limited by the use of multiple significance tests without corrections, which can magnify Type I error rates, and by their sampling strategy (members of Council for Children with Behavioral Disorders), as membership in a professional group may indicate greater commitment than other teachers. However, their findings are consistent with qualitative studies, in which special educators reported valuing PD (e.g., Hagaman & Casey, 2018). For example, in their comparison of special educators planning to leave versus stay, Gehrke and McCoy (2007) found “stayers” referenced the value of PD on behavior management, induction focused on best teaching practices, conference attendance, opportunities to participate in content area curriculum development, and meetings with other new teachers; in contrast, only one mover discussed PD.

These studies suggest special educators value mentoring and PD, but the evidence of a relationship between these factors and retention is weak, as no studies measured PD duration or quality, and only two studies examined the significance of relationships between PD and intent (Albrecht et al., 2009; Cancio et al., 2013). Future research should address how specific characteristics of mentorship and PD relate to attrition and intervention research that permits causal inferences would be useful. For example, researchers could include measures of intent to stay in PD studies, to determine if PD supports plans to stay. Propensity score matching and/or regression discontinuity could also be used to make causal inferences using large data sets. For example, Ronfeldt and McQueen (2017) used propensity score matching to analyze SASS/TFS data, examining the causal impact of induction on attrition and found significant effects, permitting causal inferences. Similar studies in special education would be useful.

Certification

U.S. states often permit uncertified (unlicensed) teachers to be hired if qualified teachers are not available. Two studies compared certified and uncertified teachers, obtaining contradictory results (Albrecht et al., 2009; Conley & You, 2017), while Gilmour and Wehby (in press) compared teachers with different certifications, preventing comparison with the other two studies. Using structural equation modeling (SEM) to analyze SASS data, Conley and You (2017) found certified secondary special educators (those with any certificate) were less likely to intend to leave than uncertified teachers, controlling for social support. In contrast, Albrecht et al.’s (2009) study found licensure did not differentiate those planning to stay versus leave, though they do not report the analytic method (e.g., analysis of variance, chi-square). Conley and You’s (2017) study was methodologically stronger, with a nationally representative sample and controls for covariates.

Gilmour and Wehby (in press) compared special education and dual certified teachers to general education certified teachers, using logistic regression to examine North Carolina state data. Special education certification was associated with a 22% increase in odds of leaving (confidence interval [CI: 21%, 50%]). Dual certification was associated with a 14% increase in odds of leaving CI [11%, 33%]) compared to general education certification, controlling for teacher and school characteristics. It is not clear whether dual certified teachers were working as special or general educators, a major limitation. Thus, we cannot interpret whether the smaller attrition rate is due to certification or teachers’ assignments. Of note, this study describes how special versus dual versus general education certification relates to attrition, not the value of being certified versus uncertified.

Although the relationship between certification and attrition is mixed and future study is needed, Conley and You’s (2017) findings are consistent with prior research (Billingsley, 2004).

Experience

Past research suggests that attrition tends to be higher for less experienced teachers and then drops off until they reach retirement age (Guarino et al., 2006). Early career attrition is problematic as it contributes to the teacher shortage (Smith & Ingersoll, 2004) and teachers become more effective with experience (Henry, Bastian, & Fortner, 2011).

Current studies suggest that special educators with less experience moved at higher rates than general educators, but left teaching at comparable rates (Boe et al., 2008; DeAngelis & Pressley, 2011). Boe et al. (2008) analyzed three waves of SASS/TFS data, and differentiated among three types of attrition: leaving, moving, and transferring. For all three types, they found significant differences between special and general educators over their careers (exit attrition [leaving]: χ² = 25.04, p < .001; transfer attrition: χ² = 10.45, p < .01; school migration [moving]: χ² = 60.31, p < .001). For both groups, exit attrition (leaving) was highest early in their careers, and then declined, but special education teacher exit attrition did not decline as much as general educators. Overall, migration rates (i.e., moving) were substantially higher than exit or field transfer attrition rates (Boe et al., 2008); 19.8% of special educators moved to a different school in their first 3 years, compared to 13.1% of general educators, a significant difference (X² = 9.33, p < .01). Patterns differed slightly for teaching field transfers (Boe et al., 2008). Early career special and general educators were more likely to transfer, but general education transfer rates remained steady throughout their careers, while special educators transfer rates declined. Boe et al. (2008) did not report significance of these differences.

Other results were consistent with Boe et al. (2008), though none examined all three types of attrition. DeAngelis and Pressley (2011) analyzed administrative data in Illinois, examining leaving and moving among all teachers from 1987 to 2001, the longest time period included in any study. They found 75% of special educators exited in the first 5 years, including 42% who left teaching and 34% who moved schools within Illinois—substantially higher than the average 5-year attrition rate (40%) in other subject areas. Note that their study did not track teachers who stayed in teaching but moved out of state. Analyzing surveys of 99 special educators who left or moved from a large urban district over 3 years, Billingsley (2007) found 25% of those who left were in their first 4 years. A higher percentage of those with 10+ years of experience left “to pursue another career in education” and only those with 20 years of experience or more left to retire. Menlove et al. (2004) reported different patterns of attrition for early career special educators. Surveying all special education directors in Utah (100% response rate) about their district’s attrition, they found early career teachers (i.e., those with less than 5 years of experience) moved at higher rates, whereas their more experienced counterparts left to retire. Early career and experienced special educators transferred to general education at similar rates (15.23% early career, 19.54% experienced) over the 3-year period.

Although several studies examined relationships between experience and intent, none used models that accounted for evidence that special educators are more likely to leave at the beginning and end of their careers (i.e., none modeled a nonlinear relationship), potentially biasing results. Among special educators serving students with EBD, Albrecht et al. (2009) found experience differentiated those planning to stay versus leave; 84.4% of teachers with 10+ years of experience planned to stay, compared to 70.7% of those with 2 to 5 years of experience (they reported no results for teachers with 1 or 6 to 9 years of experience, without explanation). In contrast, Berry (2012) analyzed data from a national survey of 203 rural special educators, examining correlations with intent; as experience increased, intent decreased (r = −.15, p value not reported).

Conclusions About Teacher Preparation and Qualifications

The most consistent finding is that special educators with less experience are more likely to leave, a well-established finding from prior literature reviews (Billingsley, 2004; Guarino et al., 2006). However, recent research adds little to our understanding of the preparation and qualification factors that predict attrition and retention. No studies used strong measures of teacher quality or effectiveness to examine whether their knowledge, skills, or self-efficacy are related to attrition. The lack of research on teacher preparation and qualifications is surprising given an expanded federal policy focused on improving teacher quality, effectiveness, induction, and evaluation (e.g., No Child Left Behind, 2001; Every Student Succeeds Act, 2015) and expanded options for preparation, such as varied types of alternative programs (Cochran-Smith et al., 2016). Stronger studies are needed to determine the relationship of teacher preparation to attrition. Although researchers should use the National Teacher and Principal Survey (NTPS), as it includes some aspects of preparation, more comprehensive databases are needed to study preparation and licensure variables important to special education teachers’ preparation and attrition/retention.

Conclusions About School Characteristics and Attrition

Few studies examined student race/ethnicity or poverty, but, consistent with research on general educators, all found that special educators were more likely to leave or plan to leave schools serving more students of color and more students living in poverty. Based primarily on studies of general educators, scholars have proposed two main explanations for this: (a) teachers’ racial/ethnic and socioeconomic biases may lead them to prefer teaching in schools with a larger proportion of White and socioeconomically privileged students (Hanushek, Kain, & Rivkin, 2004) or (b) demographics may be a proxy for other factors (e.g., working conditions, proximity of schools to teachers’ homes, safety) that covary with demographics (Johnson, Kraft, & Papay, 2012; Simon & Johnson, 2015). Studies of general educators provide empirical support for both explanations. For example, Johnson et al. (2012) found working conditions attenuated, but did not eliminate, a relationship between school poverty and student race/ethnicity and attrition. This suggests that working conditions may partially explain higher attrition in schools serving more students of color and students living in poverty, but still allows for the possibility that teacher preferences for Whiter and more affluent students may explain some proportion of the variance. Future research needs to determine why special educators’ attrition varies by school demographics, control for both working conditions and for teachers’ proximity to their schools and consider methods that would allow researchers to determine potential racial/ethnic biases in their decisions.

In addition, other school characteristics were barely studied and warrant further research; only one study compared urban versus rural schools (Prater et al., 2007), and no recent studies controlled for elementary versus secondary school level, despite prior evidence of higher attrition in secondary schools (Borman & Dowling, 2008).

Demands

As described above, special educators have many complex responsibilities, and they may experience frustration and be more likely to leave when their demands require more than they can reasonably provide (Bettini et al., 2017).

Social Contexts

Extant findings were consistent with prior reviews, finding that social contexts predicted teachers’ intent to stay or leave (Billingsley, 2004), and they extended prior research by identifying specific aspects of social contexts that may be especially important. They examined school culture, administrative support, collegial support, paraprofessional support, and the degree to which social contexts provided autonomy to make decisions.

Resources

Only, a few studies considered how materials (e.g., such as curricula) and time relate to special educators’ attrition.

Financial Compensation

Teacher pay in the United States is determined at the district level, such that teachers in two neighboring districts can make very different salaries (Dee & Wyckoff, 2013). Within districts, salaries are determined by a relatively rigid salary schedule, in which compensation is determined by a combination of years of experience and educational attainment, though some districts have recently begun incorporating measures of teacher effectiveness into their salary schedules (Dee & Wyckoff, 2013). Two studies took advantage of state policy changes to test whether financial benefits for teachers causally effected their attrition (Clotfelter et al., 2008; Feng & Sass, 2017), the first such studies in special education. Clotfelter et al. (2008) used regression discontinuity and difference-in-difference analyses to examine North Carolina’s 1999–2000 to 2003–2004 administrative data, testing effects of an $1,800 bonus awarded to secondary teachers in shortage areas in low-performing, low-income schools. The authors caution that the bonus program was poorly publicized and administered on a timeline that limited its potential for informing teachers’ decisions (see Clotfelter et al., 2005). Aggregating across all shortage areas, they found eligible teachers in eligible schools were 12% less likely to leave than other teachers in the school, controlling for relevant factors (e.g., race/ethnicity, gender, middle/high school). However, using an interaction to test differential effects across math, science, and special education teachers, they found effects were primarily accounted for by math teachers; effects on special educators were insignificant.

Feng and Sass (2017) examined Florida’s administrative data set, using difference-in-difference analyses to test effects of a program that provided certified early career secondary teachers in “high needs” subjects student loan repayment and/or a bonus if they stayed in eligible subjects and grades. They estimated effects of each program on retention, conditional on relevant teacher characteristics, class characteristics, and school characteristics. The loan forgiveness program provided up to $2,500/year for undergraduate loans and up to $5,000 per year for graduate loans, with a maximum of $10,000/person; on average, teachers received a benefit equal to 3.4% of net pay. Loan forgiveness significantly reduced special educators’ attrition, though effects on special educators were smaller than effects on math and science teachers. The bonus program paid eligible teachers up to $1,200 in 1999–2000 if they remained in the same district and content area the following year; in 2001–2002, the program offered special education and English as Second Language teachers a smaller bonus, up to $850 (the average recipient received $750). The $1,200 bonus resulted in a 32.2% reduction in the likelihood of leaving; this result was robust to alternate specifications of the model. For special educators, a $1,200 bonus (equal to, on average, a 5.7% pay raise) reduced likelihood of attrition significantly, by 10.1% to 12.5%; the smaller bonus (equal to, on average, a 2% pay raise) had no significant effects.

Unlike most research in this review, researchers who examined financial compensation used sophisticated methods permitting causal inferences; these were the strongest in this review. These are the only studies of special education teacher attrition to date that permit causal inferences, providing a high degree of confidence in their findings (Clotfelter et al., 2008; Feng & Sass, 2017); furthermore, their findings are supported by three analyses of the TFS, in which special educators identified salary and benefits as important reasons for leaving teaching (Berry et al., 2011; Boe et al., 2008; Carver-Thomas & Darling-Hammond, 2017; Luekens et al., 2004). For example, in Carver-Thomas and Darling-Hammond’s analysis of the 2012–2013 TFS, 16% of special educators and 13% of general educators who left teaching indicated a better salary was a reason for leaving, while 8% of both groups reported leaving for better benefits.

One limitation to the extant research is that the effects of financial compensation are likely to vary across labor markets (e.g., availability of higher paying positions in teaching and other fields, local economic conditions; Imazeki, 2005) and based on teachers’ marketability, but no studies examined these nuances; for example, a teacher who believes she or he has few other employment options might choose to stay in teaching even with low compensation. In addition, further research is needed to calibrate what amount of compensation is necessary to have an effect, and what delivery mechanisms (e.g., bonuses, loan forgiveness) maximize the efficiency of increased compensation. Future studies could contribute value by testing whether relationships between financial compensation and attrition vary (a) based on the size and distribution mechanism (e.g., cash bonus vs. loan forgiveness vs. base salary increase), (b) across different labor markets, (c) longitudinally as economic conditions change, (d) across teachers whose qualifications are more/less marketable, and (e) across stages of teachers’ careers.

Affective Responses and Coping Strategies

Teachers experience positive and negative emotional responses to their work, and they deploy strategies for dealing with these responses (Sutton & Wheatley, 2003). These responses are shaped by both their working conditions as well as their emotional responses and coping strategies (e.g., their stress management strategies, emotional regulation; Cooley & Yovanoff, 1996). Several studies examined how these emotional responses and coping strategies contributed to attrition/intent.

In descriptive studies, researchers reported that 13% to 27% of special educators (or other informants) identified stress and/or burnout as a contributor to attrition (Berry et al., 2011; Billingsley, 2007; Hagaman & Casey, 2018; Kaff, 2004). Jones and Youngs (2012) used affective events theory, examining whether emotional responses (i.e., positive affect, negative affect, perceptions of skill, fatigue) predicted intent to stay among early career general (n = 26) and special educators (n = 14) in urban districts. They measured intent in fall and spring, and they examined emotional responses using the experience sampling method, a tool through which teachers respond at random times (averaging once/2 hours) during the workday, to questions about what they are doing and how they feel at that moment; items came from the well-validated Positive and Negative Affect scale. Although their special education sample included only 16 teachers, they found that mean positive affect (R² = .46, p < .05) predicted teachers’ commitment to stay. Controlling for affective responses, special educators were significantly less likely than general educators to plan to stay (controlling for: positive affect, R² = −.61, p < .10; negative affect, R² = −.66, p < .05; perceptions of skill, R² = −.64, p < .05; fatigue, R² = −.66, p < .05). This implies that differences in affective responses cannot fully explain differences in special versus general educators’ intent, though caution is warranted given the small sample and the omission of working conditions from analyses.

Bettini et al. (2017), using the same survey, found emotional exhaustion (a component of burnout) in fall predicted both special educators’ (R² = −.430, p = .021) and general educators’ (R² = −.683, p = .000) spring intent; emotional exhaustion mediated relationships between workload manageability and intent for special educators (R² = .242, p = .050) and general educators (R² = .591, p = .000). Of note, Bettini et al. (2017) was the only study in the present review to consider affective responses as mediators of relationships between working conditions and intent. More research is needed to understand which working conditions and resulting affective responses contribute to attrition as well as identifying coping strategies that retain teachers.

Only one study identified coping strategies special educators used to address work demands. In a qualitative analysis of experienced special educators serving students with EBD, Prather-Jones (2011b) found that special educators felt it was essential to not take student behavior personally; for example, one teacher shared, “You can’t get bent out of shape when they say something derogatory towards you . . . ” (p. 184). Another strategy included “acceptance of limitations” as teachers said that, to remain over the long term, special educators must “learn not to blame themselves for students’ failures” (Prather-Jones, 2011b, p. 185), since they cannot solve all of students’ problems. Finally, teachers stated the need for flexibility, as teaching students with EBD is often unpredictable. The focus on a purposeful sample of experienced special educators was a strength of this study, but it was difficult to discern whether responses were related to retention or to other questions the authors asked the participants (e.g., what they like about their jobs).

These studies considered teacher responses to their working conditions (e.g., burnout, coping strategies), including those factors that made their work more or less attractive. Although the relationship between affective responses and attrition have been documented (Billingsley, 2004), coping strategies have received little attention. Working conditions and teacher qualifications likely shape their affective responses to their work in important ways (e.g., by presenting challenges and tools to deal such challenges), and affective responses may mediate relationships between these other factors and attrition/intent. Interventions that support teachers’ development of coping strategies may reduce negative affective responses and improve commitment (Cooley & Yovanoff, 1996) and future studies should explore these possibilities.

Conclusions About Working Conditions and Attrition

Studies examined a wide array of working conditions, generally finding that special educators are more likely to leave or intend to leave when they experience more demands (e.g., Gilmour & Wehby, in press) and weaker social supports (e.g., Conley & You, 2017). Relevant dimensions of social contexts include their school culture, administrative support, collegial support, and paraprofessional support. Studies of administrative support were especially useful, if they articulated specific aspects that support retention, some of which are inexpensive to incorporate into daily interactions with teachers (e.g., communicating trust, respect, appreciation). However, there are several key areas in which further research is needed.

First, it is unclear which demands, or combination of demands, are most predictive of attrition and how they contribute. Likely, a combination of demands, coupled with insufficient resources, contributes to negative affective responses, and thereby contributes to attrition or intent to leave. Demands that interfere with teachers’ effectiveness, in combination with the absence of resources essential for effectiveness (e.g., curricular materials, collaborative time for special educators in inclusive settings), may be particularly problematic, as they may prevent teachers’ from experiencing a “sense of success,” an important factor in career decisions (Johnson & Birkeland, 2003, p. 581). More research is needed to understand which demands are most impactful, and how those demands interact with other working conditions to shape attrition.

In that vein, extant studies provide little insight into how different aspects of social contexts may interact with one another to contribute to attrition. For example, scholars hypothesize that administrators cultivate healthy school cultures (Billingsley, McLeskey, et al., 2017) and create conditions for collegial support (Youngs et al., 2011), yet no studies examined these kinds of complex relationships among various dimensions of special educators’ social contexts.

More research is needed to articulate, more specifically, the forms of social supports that contribute to retention. Studies of social supports largely focused on generic interpersonal factors (e.g., trust), rather than supports specific to special education (e.g., leading for an inclusive vision, scheduling for collaboration; Billingsley, McLeskey, et al., 2017). Future studies should articulate relational, leadership, and administrative actions that support special educators’ instruction, to determine how these actions are associated with attrition.

Interestingly, some qualitative studies suggest some special educators perceive that they and their students are misunderstood, unappreciated, and unsupported (e.g., DeMik, 2008; Gehrke & McCoy, 2007; Hagaman & Casey, 2018), and perhaps even marginalized in their schools. Lack of support for including students in general education is of particular concern (e.g., DeMik, 2008), given special educators’ need to collaborate with general educators to address students’ needs across settings. If special educators feel marginalized in their interactions, specific types of administrative support may be needed (e.g., facilitating an inclusive ethic and collaborative structures/norms), and the relationship of these supports to attrition warrant further study.

Teacher Demographics

Retaining male teachers and teachers of color is especially important, given that the special education teacher workforce is substantially more White and female than the population of students with disabilities; in 2011–2012, only 18% of special educators were people of color, compared to 47% of students with disabilities (Billingsley, Bettini, & Williams, 2017). Yet, few researchers investigated the relationship between demographic variables and attrition.

Nonwork Reasons for Leaving

Some teachers leave for reasons that are not malleable to intervention, such as child care, health, and family moves. Researchers identified these nonwork reasons for leaving using retrospective questionnaires (Billingsley, 2007; Boe et al., 2008; Menlove et al., 2004). We focus only on findings from nationally representative TFS studies (i.e., Boe et al., 2008; Carver-Thomas & Darling-Hammond; 2017; Luekens et al., 2004).

Not surprisingly, nonwork reasons play a substantial role in teachers’ decisions to leave, a reminder that many aspects of attrition may not be amenable to intervention. Carver-Thomas and Darling-Hammond (2017) combined items related pregnancy/child care, health, and caring for family, into “other personal life reasons” and found that 42% of special educators and 37% of general educators reported these as very important to their decisions to leave teaching. Boe et al. (2008) reported 14% of special educators and 16.9% of general educators left for family or personal reasons, 17.8% of special educators and 12.2% of general educators left for pregnancy/child rearing, and 1.7% of special educators and 5.2% of general educators left for health reasons. Although one might conclude that more teachers left for personal reasons in recent years, differences in response formats prevent such inferences about changes over time.

Some teachers leave for other careers. Using data prior to 2002, Boe et al. (2008) reported 14.5% of special educators and 9.3% of general educators left for other careers. In 2012–2013, pursuing another career was identified by a higher percentage of leavers, with 23% of former special educators and 28% of all former teachers indicating this was very important to their decision to leave (Carver-Thomas & Darling-Hammond, 2017). Although a substantial proportion of teachers pursue other careers, it is not clear whether these are careers within or outside of education.

The proportion of teachers leaving for retirement changed from earlier to more recent studies, perhaps due to changes in the TFS response options. Earlier versions of the TFS included selecting the top three reasons for leaving, while more recent versions included ratings of many different options. Boe et al. (2008) reported that a significantly lower percentage of special (16.5%) than general educators (28.8%) gave retirement as a reason for leaving (p < .02) and Luekens et al. (2004) reported similar results for special educators, with 18.9% leaving to retire. In contrast, more recently, Carver-Thomas and Darling-Hammond (2017) reported a higher proportion of special educators (37%) reported they left to retire, compared to 31% of all teachers.

Caution is needed when comparing results across these studies, as researchers (a) used TFS data with varied response formats (e.g., aggregating results across multiple versions of the TFS, Boe et al., 2008), (b) used different comparison groups (general educators vs. all teachers), (c) did not report all reasons for leaving (e.g., Boe et al., 2008), and (d) combined individual items into larger constructs, without reporting individual items (Carver-Thomas & Darling-Hammond, 2017). In the future, researchers should consider reporting both combined and individual items (e.g., Boe et al., 2008), to permit comparisons across studies and across different versions of the SASS/TFS and NTPS.

Conclusions About Teacher Demographics and Nonwork Reasons for Leaving

Identifying patterns of attrition is crucial for workforce planning and to design strategies for addressing attrition. Although we have a shortage of special educators of color (Billingsley, Bettini, et al., 2017), we do not have data about their attrition rates, suggesting a need to disaggregate results. There are a lower proportion of males in special education, suggesting the need to understand whether male special educators leave at differential rates, and if so, why. In general, researchers should include demographic data in attrition studies.

It is surprising that researchers studying special education teacher attrition with TFS data used only reasons for leaving, but not moving, especially given the higher moving rates among special educators. Carver-Thomas and Darling-Hammond (2017) reported both types of attrition, however, they did not disaggregate special education movers in their analyses.

Focus on Studies of Actual Attrition Versus Intent

Less than half of the researchers studied teachers’ actual leaving, relying on intent to leave and stay. Intent is a useful variable, as it is relatively inexpensive to measure and is moderately correlated with actual attrition (Gersten et al., 2001). However, overreliance on intent as a proxy for attrition is problematic, as the degree to which it predicts actual attrition may vary over the school year and it is possible that some factors shaping intent may be not be powerful enough to predict attrition.

Few studies differentiated movers versus leavers. We need more studies that include leaving and moving, especially as moving rates are higher among special educators than among most other teachers (Carver-Thomas & Darling-Hammond, 2017). In addition, transfers from general to special education are often not included in attrition research (Boe et al., 2008, is an exception).

Define Stayer

Several studies defined stayers as teachers who had remained for a set time, ranging from a minimum of 5 years (López-Estrada & Koyama, 2010) to a minimum of 12 years (Lesh et al., 2017). Defining stayers based on years of experience is potentially problematic, as stayers may in fact be planning to leave soon. In addition, researchers did not consider whether stayers demonstrated strong skills and would be desirable to retain. Future researchers should consider focusing on effective stayers, considering a combination of the number of years they stayed, their plans to stay in the future, and factors influencing their decisions.

Include or Generate Theory to Inform Understanding of Attrition

Only a third of the studies used a conceptual or theoretical framework to inform their work, a clear weakness in this body of research. Strong theoretical or conceptual frameworks are needed in future studies. They help researchers build off one another’s work in more deliberate ways, as they illuminate how different constructs relate to one another. In addition, they provide insights into potential new directions for inquiry. For example, Cancio et al. (2013) used House’s theory to parse varied types of administrative support; the theory supported a more focused inquiry into the construct of administrative support. Other studies (e.g., Bettini et al., 2017) used conservation of resources theory to justify examining workload manageability, a promising construct that may be useful for examining complex interactions among varied demands and supports. This theory might inform the design of interventions to improve retention. For example, does reducing demands (e.g., nonteaching responsibilities) and increasing support (e.g., PD, curricula, collaboration) lead to increased workload manageability, reduced burnout, stronger effectiveness, and increased retention? Researchers should also consider other potentially useful frameworks (see Guarino et al., 2006, for a discussion about theories relevant to the study of teacher attrition and retention).

Study Interrelationships Among Factors

Past researchers have examined complex interrelationships among predictors of intent (e.g., Billingsley, 2004), but few current studies examined these kinds of interrelationships. We need research that explores complex relationships among malleable factors (i.e., working conditions, preparation/qualifications), affective reactions to work (e.g., stress, burnout, job satisfaction), and attrition. A combination of complex demands (e.g., caseloads, problematic student behavior), weak supports (e.g., resources, social support), and insufficient preparation may be more stressful, and more strongly associated with attrition, than any single factor alone. Articulating these relationships, and how they might be mediated by affective responses, would be useful. For example, are some supports (e.g., collegial support) more strongly associated with attrition when others (e.g., administrative) are weak, or when demands (e.g., instructional responsibilities) are high? What specific social contexts are most important for less experienced teachers? Are social contexts more predictive of attrition for teachers from minoritized groups? These questions could be answered using SEM with survey data (e.g., SASS/NTPS), or by combining surveys with administrative data.

Conduct Intervention Studies

Only two studies permitted causal inferences (Clotfelter et al., 2008; Feng & Sass, 2017). Given that research has consistently found some conditions predict intent (e.g., administrative support, demands), scholars should evaluate the effects of interventions to improve retention to document causal relationships. For example, scholars could develop interventions to improve special educators’ working conditions and use single case design methods to document the effects of those interventions on proximal outcomes related to attrition (e.g., workload manageability, burnout, job satisfaction). In addition, scholars could embed measures of attrition into other intervention research, to determine how PD initiatives and school reform efforts (e.g., PBIS) relate to changes in attrition. In addition, using extant state and district administrative data sets, scholars can use regression discontinuity and propensity score methods to take advantage of natural variability in the implementation of intervention efforts, and determine how those efforts are associated with changes in attrition rates.

Use High-Quality Qualitative Methods

Some qualitative studies in this review contributed value by exploring areas that have seldom been examined in prior research (e.g., Prather-Jones, 2011b). However, some qualitative studies were unfocused, broadly examining factors teachers felt contributed to attrition, such that they provided little or no new information beyond what has long been known. We know that some factors relate to attrition, and simply reiterating that these conditions matter does not add to the literature. Qualitative research can add value if it is used to build stronger theory, illuminating what aspects of these factors are most important, why, for whom, and under what circumstances. For example, qualitative studies have unique potential to elaborate our understanding of collegial support, by, for instance, illuminating what dimensions of collegial support shape teachers’ plans.

Qualitative research could contribute particular value to understanding the process by which teachers choose to leave. Current studies of leavers and movers are largely quantitative, requiring teachers to respond to discrete, brief statements about their reasons for attrition. Thus, there are no studies that provide in-depth and contextualized accounts of teachers’ decisions to leave, a notable weakness (though Gehrke & McCoy, 2007, interviewed several teachers who left). Teachers do not decide to leave at a single time, rather, they consider their options over time with others in their social networks (Clandinin, Long, Schaefer, Downey, & Steeves, 2015). Yet, we have no studies investigating the process by which they decide to leave nor any studies following them from the time they enter teaching until they leave.

Examine the Role of Teacher Effectiveness

Only one study examined how teachers’ instructional practices related to intent (Albrecht et al., 2009). None examined how their effectiveness (or student achievement) related to attrition, despite strong research showing relationships between general educators’ effectiveness and attrition (Goldhaber et al., 2018). Ideally, retention efforts would be targeted to effective teachers. To target these teachers, we need research examining conditions under which more effective teachers leave. Furthermore, scholars have posited that teachers are more likely to leave when they do not experience success in serving students (Johnson & Birkeland, 2003). Several current studies suggested working conditions may influence attrition by reducing or promoting teachers’ capacities to effectively serve students (e.g., Kaff, 2004). For example, one teacher shared, “I have a great support system in my building. Until that system breaks down and I don’t have the support I need to run a positive, productive program, I plan to continue what I am doing” (Albrecht et al., 2009, p. 1014). Future research should consider including teachers’ perceptions of their own effectiveness as a mediator of relationships between various conditions and their attrition.

Consider Additional Comparisons for Specific Teacher Groups

There is a need for comparison studies for special educators at particular risk of leaving, as well as those teacher groups that have rarely been disaggregated in attrition/retention research. None of the research we reviewed was focused primarily on (a) teachers entering through varied pathways (e.g., alternative, for profit, on-line programs), (b) males, an important are of research given the lack of male special educators, (c) low-incidence teachers, a key shortage area (Levin et al., 2015), or (d) teachers working in varied service delivery models (e.g., co-teaching, self-contained). Several studies compared special and general educators, providing an understanding about how factors may differentially relate to their attrition (e.g., Jones et al., 2013) and research articulating such differences continues to add value.

It is particularly concerning that only one study addressed retention of Mexican American teachers, and none examined other teachers of color, despite an urgent need to retain teachers of color (Billingsley, Bettini, et al., 2017). We encourage researchers to move beyond reporting demographic identifiers for participants and begin considering how teachers’ sociocultural identities—their understandings of the key indicators of their own identity, including but not limited to their racial/ethnic identity—shape their attrition. Such work will require moving beyond the simplistic use of the racial/ethnic categories from the U.S. census, and will probably require qualitative and mixed methods approaches.

Use National Data Sets to Determine Reasons for Attrition

Several studies used TFS data to examine teachers’ retroactive reasons for why they actually left teaching (e.g., Boe et al., 2008; Carver-Thomas & Darling-Hammond, 2017), however, attrition data for some TFS years have not been reported. In addition, special educators’ reasons for moving or transferring to other fields were only explored by Boe et al. (2008). The TFS is readily available and (a) provides special educators’ perspectives about leaving and moving, allowing comparisons with general educators; (b) allows the comparison of reasons for attrition over time; (c) includes reasons that received little or no attention in other research (e.g., accountability Carver-Thomas & Darling-Hammond, 2017); (d) can be used to triangulate findings from studies of intent (Conley & You, 2017, administrative support); and (e) could include comparisons of teachers (e.g., urban vs. rural). The TFS is limited in that it cannot be used to study different groups of special educators (e.g., high vs. low incidence as it is not possible to assess these subgroups), so other surveys might be designed to make these comparisons. In addition, special educators’ reasons for leaving may be influenced by factors that they are unaware of; a teacher leaving because of student behavior may not realize that a mentor could have helped and is not likely to select “lack of mentorship” as a reason for attrition. We encourage those interested in surveying teachers about reasons for attrition to use the TFS data, which are currently underused or to extend the research by studying special education teacher groups and response options not included in the TFS.

Study the Consequences of Attrition

Some researchers have documented the deleterious financial and academic effects of general education teacher attrition (e.g., Ronfeldt et al., 2013). Yet, no studies have examined consequences of special education attrition. What effect does special education attrition have on their students, students’ families, and the teachers with whom they collaborated? How does attrition effect the achievement of students with disabilities?

Disseminate to Practitioner and Policy Audiences

Researchers need to disseminate findings about attrition/retention to a broader stakeholder group, those outside of special education (e.g., principals, district leaders). Only two of the present studies reached general education leadership or policy journals (Clotfelter et al., 2008; Feng & Sass, 2017), while one was published in a special education leadership journal (Billingsley, 2007). Scholars should consider improving dissemination efforts, especially to practitioner audiences responsible for retention, using varied types of material (e.g., briefs, newsletters, social media).