The Promise of Older Novices: Teach For America Teachers’ Age of Entry and Subsequent Retention in Teaching and Schools

by Morgaen L. Donaldson - 2012

Background/Context: Increasing teacher retirements and persistently high turnover have ignited interest in teacher recruitment and retention in recent years. Nowhere is the need to understand these issues more urgent than in low-income schools. These schools face especially large challenges in attracting and retaining competent and committed teachers. Such schools often experience high levels of turnover and, as a result, are staffed disproportionately by inexperienced teachers who are generally less effective than their more experienced counterparts. In addition to this educational cost, elevated turnover also generates substantial organizational and financial losses for low-income schools. Professionals working in other careers constitute one potential source of teachers for low-income schools. The proportion of midcareer entrants among first-year teachers nationwide nearly doubled in recent years. Although analysts assert that midcareer entrants bring great commitment to teaching, scant research examines whether they stay in teaching longer than individuals who enter teaching directly after college.

Purpose/Objective/Research Question/Focus of Study: The primary purpose of this study is to examine whether older entrants to teaching are more likely than younger recruits to voluntarily remain in low-income schools and the teaching profession as a whole. I also explore whether older new teachers’ backgrounds differ from those of younger new teachers and whether they are more likely than their younger counterparts to remain in K–12 school-based jobs (i.e., teaching, working as a specialist or administrator). In addition, I examine whether older new teachers who leave the profession have different reasons for doing so than younger new teachers and whether they enter different occupations after teaching.

Research Design: I investigate the questions described above using a sample of over 2,000 Teach For America (TFA) teachers who began their careers in schools serving high proportions of low-income and minority children. The sample for my study is drawn from a census of all teachers enrolled in the 2000, 2001, and 2002 TFA cohorts. These teachers would have accumulated 4, 5, or 6 years of teaching experience if they had taught continually. From 3,283 TFA enrollees in these cohorts, 2,029 individuals (62%) responded to an online survey that gathered data on teachers’ individual characteristics (e.g., subject matter preparation and assignment; demographic information) and, where relevant, the timing of their first departure from their school and the teaching profession. I used discrete-time survival analysis, logistic regression, and chi-square analysis to analyze the data.

Findings/Results: I found that older TFA entrants to teaching had a lower risk than did younger entrants of leaving low-income schools, the teaching profession, and broader school-based roles. I further found that older entrants’ backgrounds differed from younger entrants. Older entrants were significantly more likely than their younger counterparts to be male, to be African American, and to have lived in the locale where they were placed by TFA. Among respondents who left teaching, older entrants’ reasons for doing so differed significantly from those noted by younger entrants. Older entrants to teaching were significantly more likely than younger entrants to cite family or health matters as a very or extremely important factor in their decision to leave. Last, older entrants who left the profession also entered significantly different types of professions than did younger entrants. Most notably, older entrants to teaching were significantly more likely than younger entrants to become a K–12 specialist or administrator after they left the classroom.

Conclusions/Recommendations: Viewed broadly, these findings suggest that older entrants to teaching may prove a promising source of teachers for low-income schools. On all measures, older entrants demonstrated more commitment to low-income schools and the teaching profession than did their younger counterparts. Notably, even though they began their careers in challenging, low-income schools, they left the teaching profession at rates (61.3% in 3 years) that are not that distant from some estimates of attrition of teachers who began their careers in urban schools, some of which were likely less challenging than those where TFA teachers typically work. More broadly, districts seeking to develop human capital across multiple levels of the system might also consider targeting older individuals as a source of new teachers. Not only do these people appear to teach longer, but if they leave teaching, they are more likely than younger entrants to remain in schools in roles other than classroom teacher.

Increasing teacher retirements and persistently high turnover have ignited interest in teacher recruitment and retention in recent years (Guarino, Santibañez, Daley, & Brewer, 2004; National Commission on Teaching and Americas Future, 2003). Nowhere is the need to understand these issues more urgent than in low-income schools. These schools face especially large challenges in attracting and retaining competent and committed teachers. Such schools often experience high levels of turnover (Ingersoll, 2001) and, as a result, are staffed disproportionately by inexperienced teachers (Peske & Haycock, 2006) who are generally less effective than their more experienced counterparts (Rockoff, 2004). In addition to this educational cost, elevated turnover also generates substantial organizational and financial losses for low-income schools (Birkeland & Curtis, 2006; Neild, Useem, Travers, & Lesnick, 2003). It is clear that low-income schools need good teachers who are committed to working in these settings over time.

Professionals working in other careers constitute one potential source of teachers for low-income schools. Between 1997 and 2004, the proportion of midcareer entrants among first-year teachers nationwide nearly doubled, from 20% to 39% (Marinell, 2009). State-level research suggests that one third to one half of all new teachers enter teaching from another profession (Johnson, Kardos, Kauffman, Liu, & Donaldson, 2004). Recent evidence from Illinois, for example, indicates that 50%60% of new teachers are over 25 years old when they begin teaching (DeAngelis & Presley, 2007).¹ This shift toward older and second-career entrants to teaching is not unique to the United States. Researchers have observed such trends in both Ireland (Heinz, 2008) and England (General Teaching Council for England, 2006).

Although analysts assert that midcareer entrants bring great commitment to teaching, scant research examines whether they stay in teaching longer than individuals who enter teaching directly after college (Resta, Huling, & Rainwater, 2001). The purpose of this study is to examine whether older entrants to teaching are more likely than younger recruits to voluntarily remain in low-income schools and the teaching profession as a whole. I also explore whether older new teachers are more likely than their younger counterparts to remain in K12 school-based jobs (i.e., teaching, working as a specialist or administrator). In addition, I examine whether older new teachers who leave the profession have different reasons for doing so than younger new teachers and whether they enter different occupations after teaching.

I investigate these questions using a sample of over 2,000 Teach For America (TFA) teachers who began their careers in schools serving high proportions of low-income and minority children. TFA plays an increasingly important role in supplying teachers to hard-to-staff schools in some of our nations largest cities, including New York City, Houston, Philadelphia, and New Orleans. It also places teachers in struggling rural schools such as those in the Rio Grande Valley and the Mississippi Delta. With high standardized test scores and diplomas from selective universities, TFA teachers also represent the kind of candidate some, including Secretary of Education Arne Duncan, think should constitute a growing proportion of the teaching force, particularly in these schools. And, in some ways, support for TFA appears to be growing. TFA won one of the federal governments highly competitive i3 (Investing in Innovation) grants in 2010, guaranteeing the organization $50 million from the government and $10 million in matching funds.

Using discrete-time survival analysis (DTSA), I found that older TFA entrants to teaching did indeed have a lower risk than did younger entrants of leaving low-income schools, the teaching profession, and broader school-based roles. I further found that older entrants backgrounds differed from younger entrants. Among respondents who left teaching, older entrants reasons for doing so differed significantly from those noted by younger entrants. Last, older entrants who left the profession also entered significantly different types of professions than did younger entrants.

THE COSTS OF TEACHER TURNOVER

Research shows that teacher turnover creates a multitude of costs for districts, schools, and students. Such costs include the financial costs of recruiting, hiring, and training replacement teachers (Barnes, Crowe, & Schaefer, 2007), achievement costs if the new teacher is less effective than the one she replaces (Rivkin, Hanushek, & Kain, 2005), and cultural costs when the churn of teachers in and out of a school impedes the development of a strong, positive faculty culture of collaboration and coordination (Guin, 2004; Johnson, Berg, & Donaldson, 2005). Moreover, evidence suggests that low-performing schools serving high proportions of low-income students of colorjust the sort of setting in which TFA teachers workpay a particularly high price for teacher turnover (Barnes et al.).

Nationwide, the cost of teacher attritionteachers leaving the profession entirelyhas been estimated at $2.2 billion (Alliance for Excellent Education, 2005). Including the costs of teachers transferring out of one school and into another, these analysts estimate that the total annual cost of teacher turnover is $4.9 billion.

In a study conducted in five districts, Barnes et al. (2007) found that the financial costs of teacher turnover were substantial. The authors estimates of the annual average cost of turnover (recruiting, hiring, and training teachers to replace leavers) ranged from $4,366 per leaver in rural Jemez Valley, New Mexico, to $17,872 per leaver in Chicago. This range is consistent with other studies (see, e.g., Milanowski & Odden, 2007; Shockley, Guglielmino, & Watlington, 2006). The Chicago total includes $9,501 in costs borne by the district and $8,371 borne by individual schools. With 4,844 teachers leaving the district in the data collection year, the authors estimated the total cost to Chicago Public Schools at over $86 million for that year alone.

Barnes et al. (2007) also found that low-performing, high-minority, and high-poverty schools were more likely to experience turnover and therefore incurred higher financial costs than comparison schools. For example, based on their estimates, the authors stated that a low-performing (16th percentile or below within the district), average-sized Milwaukee school would spend, on average, $67,000 more annually on turnover than its high-performing (84th percentile or above within the district), average-sized counterpart.

In additional to the financial costs, the achievement costs of turnover are sizeable. Several studies have found that teachers effectiveness rises rapidly in their first few years in the classroom but then levels out (Rivkin et al., 2005; Rockoff, 2004). Staiger and Rockoff (2010), summarizing this research, stated that turnover that results in the replacement of a more experienced teacher (two or more years of experience) with a first-year teacher leads students assigned to that new teacher to lose approximately .10 standard deviations of achievement compared with similar students assigned a more experienced teacher. Staiger and Rockoff estimated the impact of this one-year achievement loss on those students future earnings at $10,000$25,000 per student.

Schools experiencing high teacher turnover also incur cultural costs, with implications for student learning. In her study of five schools in the same district, Guin (2004) found that schools with chronically high teacher turnover were less likely to have high levels of trust and collaboration among teachers (p. 19). Both trust among teachers (Bryk & Schneider, 2003) and teacher collaboration (Goddard, Goddard, & Tschannen-Moran, 2007; Jackson & Bruegmann, 2009) are associated with higher student achievement.

In sum, this research suggests that the costs of turnover borne by districts, schools, and students are considerable. Identifying which teachers stay in schools, especially those that are low-income and low-performing, is a critical first step toward building a stable and effective teaching force in our nations most challenged schools. For this reason, it is important to ask whether teachers who are older when they enter teaching are more likely to stay in low-income schools and the teaching profession.

CONCEPTUAL AND EMPIRICAL UNDERPINNINGS

Two theories derived from psychology suggest different relationships between age at entry and teacher turnover. Levinson, Darrow, Klein, Levinson, and McKees (1978) model of the human life cycle posits that individuals pass through psychological stages over a lifetime that are approximately related to ages. Levinson and his colleagues suggested that four life eras, childhood (age 020 years), early adulthood (2040), middle adulthood (4060), and late adulthood (over 60), characterize human development over the life span. Although Levinson and his colleagues first developed this model based on a study of White males, its applicability to females and other racial groups has been verified through subsequent research (see, e.g., Levinson, 1986; Lim, 1986; Minter & Samuels, 1998).

Within each era, these scholars identified distinct substages with characteristics that have implications for workplace turnover. Specifically, Levinson and colleagues (1976) identified a substage experienced between age 23 and 28 that they called Early Adult Transitions. In this stage, individuals are developing a sense of themselves and evaluating their choice of work. Their dedication to their job and their career is tentative and conditional because this is a time of exploration and provisional commitment to adult roles, memberships, responsibilities, and relationships (p. 22). Individuals then experience what Levinson and colleagues (1976) called the Age 30 Transition, typically between ages 28 and 33. This period may occasion considerable turmoil, confusion, and struggle with the environment and within oneself; or it may involve a more quiet reassessment and intensification of effort (p. 23). Exiting the Age 30 Transition, individuals enter what Levinson and colleagues termed the Settling Down substage, experienced between ages 34 and 39. In contrast to early adult transitions, in this period, one makes deep commitments to career and family. These researchers have found that the individual invests more of himself in his work, family, and valued interests; and within the framework of this life structure, makes and pursues more long-range plans and goals (p. 23). Some who have adopted Levinsons model have identified this substage as the period of greatest organizational commitment, job involvement, satisfaction, and performance (Ornstein, Cron, & Slocum, 1989, p. 120).

This theory suggests that younger workers will be more likely to leave their profession than older workers. However, if older individuals do in fact leave their profession, as the Age 30 Transition suggests they might, the model suggests that they may be more likely to commit to their new job than are younger entrants. Applied to the case of teachers, this suggests that older entrants to teaching will be less likely to leave the profession than their younger counterparts.

A second theory, developed by Donald Super (1957, 1984), focuses explicitly on career development rather than the life cycle and holds that individuals move through stages that are not necessarily related to age. He identified four career stages: trial, establishment, maintenance, and decline. In the trial stage, individuals work to define their interests and skills and assess the fit between themselves and their work. This is naturally a time of self-assessment and tentative commitment. During the establishment stage, by contrast, individuals establish commitments to career and professional growth. During the maintenance stage, they cultivate the accomplishments of other stages. During decline, individuals emphasis shifts from career to other aspects of their lives.  

Although some have suggested that these stages correlate with age ranges, others have argued that the stages align with tenure in ones career. For example, some scholars have suggested that the trial stage occurs in an individuals first 2 years in a career; the establishment stage takes place in years 310; and the maintenance stage after year 10 (see Ornstein et al., 1989).

According to this application of Supers theory, a new teacher would reside in the same career stage regardless of whether she entered the profession at age 22 or age 42. All new teachers would question teaching as a long-term career and assess its fit with their skills and interests. By this logic, midcareer entrants to teaching would be no more likely than first-career entrants to remain in the classroom. Thus, the Levinson and Super theories suggest different hypotheses regarding whether older entrants to teaching will have a lower probability of turnover.

TEACHER RETENTION AND AGE AT ENTRY

The literature on teacher recruitment and retention largely ignores the question of whether the career paths and retention rates of younger and older entrants to teaching differ. Although considerable research has examined how variables such as a teachers overall age, gender, race, college major, teaching assignment, effectiveness, or school level are related to her probability of turnover, few studies have explored age at entry.²

It bears noting that researchers differ in the way in which they define older and younger. Despite these variations, researchers have generally found that older entrants to teaching possess qualities that may prompt them to remain in the classroom longer than their younger counterparts. Crow and colleagues (1990) and Serow and Forrest (1994) found that late entrants to teaching exhibited a strong and serious commitment to the teaching career. In their sample, Crow et al. identified homecomers who had initially wanted to teach but had not done so, and converted individuals, constituting the majority of the sample, who had been drawn to teaching by a major life event such as having a baby. The deep commitment to teaching expressed by both of these groups may make them especially likely to remain in teaching over time. Based on multiple qualitative studies with older entrants to teaching, Freidus (1994) arrived at similar conclusions, stating, These men and women become teachers by choice not by default. . . . They see teaching as an important and desirable career choice (p. 3). This volition and determination may make this subgroup of new teachers especially likely to remain in teaching over the long term.

A recent, anecdotal inquiry into the phenomenon of older individuals entering teaching through Teach For America echoed these findings (Simon, 2009). Paula Lopez Crespin, the 50-year-old first-year teacher featured in the article, followed her daughter into urban teaching. Crespin, who just couldnt stomach anymore her work in a bank, said that she welcomed the opportunity to do something meaningful with my life by working in a Denver public school. She is not alone. Applications to Teach For America from older individuals who had changed careers or were in graduate school, as opposed to undergraduate, increased 81% in 2008. Although some of this increase may reflect the downturn in the economy, TFA believes this is a larger trend and has created an emerging markets team specifically to recruit older individuals for the program. Moreover, analyses of nationwide data from the Institute for Education Sciences Schools and Staffing Survey indicate that older individuals constitute a growing proportion of all new teachers (Marinell, 2009).

The few studies that have directly examined the relationship between teachers age at entry and turnover suggest that older entrants to teaching have a lower probability of leaving the profession, thus confirming the assumptions inherent in Levinsons model. For instance, Kirby, Berends, and Naftel (1999) examined the turnover of Texas teachers between 1979 and 1996. They found that the second-year attrition rate of older new teachers (age 30 or older) was lower than that of younger teachers. Specifically, they found that half of teachers aged 2024 left within 6 years, and half of those aged 3034 left in 9 years. Fifty percent of individuals over the age of 35 when they began teaching left the profession within 11 years.

Similarly, in their analysis of teacher retention in North Carolina and Michigan, Murnane, Singer, Willett, Kemple, and Olsen (1991) found that, compared with males, older female entrants (age 31 and older) to teaching had a lower probability of leaving teaching, and younger female entrants (age 30 and younger) had a higher probability. Based on data on White teachers who began teaching in Michigan in 1972 or 1973, the researchers estimated that 50% of younger female entrants left teaching within 4.6 years, which was 28 years less than the median career length of younger males and older males and females (Murnane et al.).

My study of retention among 636 graduates of Harvard Graduate School of Educations masters level teacher education program between 1985 and 2006 yielded similar results (Donaldson, 2008). Overall, I found that teachers who were age 30 or older when they entered the profession were less likely to leave teaching than were teachers who entered the classroom at age 29 or younger. The odds that older female entrants to teaching would exit were .62 times the odds of younger females. The negative effect of entering teaching at age 30 or above was even larger for men. The odds that males who began teaching when they were older would exit were .30 times the odds of younger male entrants. These differences compounded such that 50% of female first-career entrants who had taught in urban schools left teaching within 5 years, and 50% of younger males left within 7 years. By contrast, 50% of midcareer female entrants left within 11 years, and 50% of older males had not left at the time when data were collected. Among those respondents who left teaching, first-career leavers were more likely to cite child-rearing as a reason for departure; midcareer leavers were more likely to leave because of retirement or health concerns.

Increasing the complexity of the research reviewed in the preceding paragraphs, a recent study determined that individuals who entered teaching between the ages of 26 and 34 in Illinois were slightly more likely to leave during their first 5 years than those who starting teaching at age 25 or younger (43% vs. 42%; DeAngelis & Presley, 2007). Individuals who entered teaching at age 35 or older were the least likely to leave, however (35% left within 5 years).

Adamss (1996) study of 2,452 teachers in one large Texas school district from 1985 to 1991 found similar results using exit from the district as the outcome. He determined that younger new teachers (under age 40) were 43% more likely to depart the district than were older new teachers (age 40 and older). He further found that 50% of younger entrants to teaching left the district in approximately 3 years, and half of older entrants exited within approximately 5 years.

SIGNIFICANCE OF THE STUDY

Overall, the research suggests that younger entrants tend to experience shorter careers in districts and the teaching profession as a whole compared with their older counterparts. However, research to date is limited. Most studies have documented turnover patterns in only one state and have not examined teachers careers in low-income schools specifically.

Addressing these and other limitations, this study adds to the current knowledge base about younger and older entrants to teaching. None of the prior research investigating age at entry has used a national data set or focused on teacher retention in some of our nations most challenged schools. This is the first study to ask whether patterns of teacher turnover in low-income schools nationwide differ by teachers age at entry.

Moreover, none of the prior quantitative studies of this topic has been able to focus exclusively on voluntary exits. Most quantitative research on teacher retention has used state data sets that are unable to distinguish voluntary from involuntary leavers. In this data set, I have been able to isolate teachers voluntary career moves, which map more directly onto implications for policy and practice than do findings that are unable to differentiate whether teacher turnover is voluntary or involuntary.

In addition, very few of the prior studies on this subject have tested whether the reasons that older and younger entrants to teaching leave the profession vary, whether the occupations they subsequently enter differ, and whether, if they leave teaching, their probability of leaving school-based roles diverges. I have been able to gather data on these important questions. In sum, the unique characteristics of this studyits national data set, focus on low-income schools, ability to isolate voluntary turnover, and attention to why teachers leave teaching, what occupations they subsequently enter, and whether they leave school-based rolesprovide critical information to inform efforts to retain teachers in the schools that arguably most need strong and committed teachers.

Some might argue that because this study uses a sample of Teach For America teachers who leave their initial schools and the teaching profession at high rates, its importance is diminished. However, the findings of this study indicate that older TFA teachers in this sample have turnover rates that are generally comparable with those of individuals who began their careers at the same time and in the same types of schools as the TFA teachers.

Moreover, although they constitute a small part of the national teaching force, TFA teachers are an important subpopulation of new teachers today. As noted, TFA plays a key role in providing new teachers in some of the nations largest districts and in some of the nations most challenged schools. They possess the few specific characteristicshigh test scores and diplomas from selective collegesshown to make a difference in student achievement (Ehrenberg & Brewer, 1994, 1995). TFAs influence is also expanding rapidly. Fueled by federal funds and private donations, TFA has recently opened sites in Boston, Dallas, Seattle, Appalachia, and South Carolina. Its cohort size doubled from 2,181 in 2005 to 4,485 in 2010, and this number is likely to grow considerably in years to come. Thus, although the TFA sample is unique and limited in some ways, it is an important subset of teachers who work in schools serving high proportions of children of color and low-income children. Thus, this study provides important information that lays the initial groundwork for subsequent research on a broader sample of teachers.

RESEARCH QUESTIONS

In this study, I examine TFA teachers conditional probabilities of voluntarily leaving their initial placement school and voluntarily leaving the teaching profession in every year following their entry. I also inquired as to why respondents who left did so and what occupations they entered subsequently. Last, I was interested in whether they left school-based roles, broadly defined, in the time period observed. Given my interest in the role that age at entry might play in these decisions, my research questions were:

1.

Do the backgrounds of older entrants to teaching differ from their younger counterparts?

2.

(a) Are older entrants to teaching at lower risk for voluntarily leaving their initial, low-income placement school than are younger entrants to teaching?

(b) Are older entrants to teaching at lower risk for voluntarily leaving teaching than are their younger counterparts?

3.

Do the factors that influence older entrants decision to leave teaching differ from those cited by younger entrants?

4.

If they leave teaching, do older entrants take up different occupations than do younger entrants?

5.

Are older entrants to teaching at lower risk for leaving school-based roles than are their younger counterparts?

THE SAMPLE: TEACH FOR AMERICA

To investigate these and other questions, I surveyed three entire cohorts of TFA teachers. Founded in 1990, TFA selects and places high-achieving individuals in low-income urban and rural classrooms after 5 weeks of preparation. Although they formally agree to teach for only 2 years, some TFA teachers remain longer (Donaldson & Johnson, 2011). Since 1990, TFA and programs like it have proliferated as alternatives to traditional, university-based pathways to teaching. TFA currently attracts large numbers of applicants from the nations most selective colleges (Azimi, 2007).

RESEARCH DESIGN

The sample for my study is drawn from a census of all teachers enrolled in the 2000, 2001, and 2002 TFA cohorts.³ These teachers would have accumulated 4, 5, or 6 years of teaching experience if they had taught continually.⁴ From 3,283 TFA enrollees in these cohorts, 2,029 individuals (62%) responded to our survey. Of these, 71.4% of respondents are female; 11.54% identified as Black/African American, 6.73% identified as Latino/Hispanic, and 77.53% identified as White/Caucasian. A total of 57.3% of the respondents reported that they were related to a teacher. The TFA organization provided information about the census of 3,283 enrollees. Although TFAs records are incomplete, I compared the sample and the census and found few statistically significant differences.⁵

PROCEDURES

I collected most of the data for this study during an online survey administered between January and March 2007 to the census of teachers. The survey requested information on teachers individual characteristics (e.g., subject matter preparation and assignment; demographic information) and, where relevant, on the timing of their first departure from their school and the teaching profession. Into this data set, I incorporated data from TFA placement records, which specify the districts in which individuals were placed.

MEASURES

I created my survey instrument by drawing on the School and Staffing Survey, earlier questionnaires designed by Kardos (2004) and Liu (2004), research into new teacher retention, and literature on survey question design (Dillman, 1978, 1991; Fowler, 2002; Payne, 1951) and reducing recall error (Sudman & Bradburn, 1982). Before administering this instrument, I piloted it with 30 TFA teachers drawn from cohorts who entered the profession immediately prior to 2000 or after 2002 (and thus were ineligible for this sample), and I tested specific survey questions and the online survey process with 812 teachers who were demographically similar to TFA teachers.

The piloting process improved the survey instrument and process considerably. I administered a draft version of the entire survey to the 30 TFA teachers and then conducted interviews or focus groups with all these individuals. This process enabled me to identify vaguely or incorrectly worded items and discern whether important options or items pertaining to turnover had been omitted. Testing the online process with 812 teachers enabled me to discern which Internet browsers best supported the online survey program I had chosen. Similar to the first piloting phase, interactions with respondents also allowed me to identify confusing, incomplete, or extraneous items on the survey.

I modified the survey based on feedback from the piloting process and then administered it to the TFA population. Once I had collected these retrospective survey data, I constructed a teacher-year data set to record important elements of the TFA teachers experience in each year they taught. Because of the challenges of respondent recall in retrospective research, most of the measures whose values I collected were time-invariant (Kelly, 2004; Taris, 2000). All analyses reported here were obtained using this dataset.

DTSA Outcomes

For analyses of whether teachers moved out of their school, the profession, or a school-based role, there are three related outcomes that correspond with Research Questions 2a 2b, and 5. These variables document the teachers (a) first voluntary exit from the initial placement school by transfer or resignation from teaching (VEXITSCHL); (b) first voluntary resignation from the teaching profession (VEXIT); and (c) exit from a school-based role (exitEDU). VEXITSCHL and VEXIT describe the time-varying outcome behavior of each teacher. I recorded the dichotomous values of each of these outcomes in each year of the profession in a separate row of the data set for each teacher, up until the year in which she left teaching or was censored by the end of data-collection (1 = if event was experienced in this year, given that it had not been experienced in an earlier year; 0 = otherwise). These outcomes are restricted to voluntary job movements. Individuals who indicated that they had left their schools or the professional involuntarily were not classified as having experienced that event. ExitEDU is not time-varying and is simply recorded 1 = if individual left and stayed out of all school-based roles (i.e., teaching, specialist, or administrator; 0=otherwise). It is not restricted to voluntary movements because I did not collect data about whether individuals exits from these roles were voluntary.  See the appendix for the names and definitions of all variables.

Question predictor

My principal question predictor is over25, a dichotomous variable that represents respondents age at entry to teaching (1 = respondent was over age 25 when she began teaching; 0 = respondent was age 25 or younger at that time). I chose this cut-off for theoretical and empirical reasons. First, there is a precedent for using age 25 to separate groups of entrants to teaching (see, e.g., Bullock & Scott, 1993; DeAngelis & Presley, 2007; and Kirby et al., 1999). Second, when I grouped respondents by age of entrance, I identified a substantial decline in numbers of respondents above the age of 25. Thus, I chose to divide older and younger entrants to teaching at this age. Unfortunately, small sample sizes did not permit me to further divide these groups.⁶ This is discussed at greater length in the limitations section.

Control Predictors

My principal control predictors represent the effect of time and are represented by its most general specification: six dichotomous predictors (T1-T6), each of which represents one of the up to 6 years in which respondents could have taught (T1=1 in the respondents first year in teaching, 0 otherwise, and so on).

In discrete-time hazard models, I also controlled for selected design and substantive covariates. To account for the fact that the 20002002 TFA teachers entered teaching in three different cohorts, I included a system of three time-invariant, dichotomous variables distinguishing the cohorts. I also controlled for variables well known to make a difference to teacher turnover, including: (a) gender (Ingersoll, 2001; Murnane et al., 1991), (b) race (Murnane et al., 1991), (c) college major (Donaldson & Johnson, 2010; Murnane et al.), (d) proximity of hometown to school placement site (Boyd, Lankford, Loeb, & Wyckoff, 2005), (e) familiarity with urban/rural areas similar to placement site (Johnson et al., 2005), (f) whether they were assigned to teach in an urban or rural region in their first year (Luekens, Lyter, Fox, & Chandler, 2004), and (g) school level (Luekens et al.).

DATA ANALYSIS  

To investigate the hypothesis that older entrants to teaching would remain in schools and teaching at greater rates than their TFA counterparts who entered teaching at a younger age, I employed discrete-time survival analysis (Singer & Willett, 1993). This method allows researchers to analyze longitudinal data to determine whether and when an event is likely to occur. Using logistic regression, it is designed to estimate the unbiased probability or risk that an event (i.e., leaving teaching) will occur in each discrete time period (i.e., year), given that it has not occurred up to that point. As such, this method is particularly well suited to studies of teachers careers and has been used often for this purpose (see, e.g., Murnane et al.s 1991 study of teachers career paths in Michigan and North Carolina). I used this method to derive answers for Research Questions 2a and 2b.

I present the most complex model here, which answers Research Question 2a. I used discrete-time survival analysis (Singer & Willett, 1993) to answer Research Question 1 by estimating the following baseline “time-only” no-intercept model:

p(VEXITSCHL=1) =

1+ e ^–[(α₁^T1_ij ^{+ α}₂^T2 _ij ^{+… α}₅^T5 _ij^{) + [ß}₁^over25 _i^{]+ [γ}₁ ^Z _i ^{+ γ} ₂^Z _ij^)]

where i represents the individual respondent, the α parameters represent the risk a teacher will leave teaching for the first time in each year, and γ₁ and γ₂ are vectors of parameters that represent the impacts of time-invariant controls (Z_i) and time-varying controls (Z_ij), respectively. The coefficient of interest is ß₁, which represents the effect of having entered teaching after age of 25 on teachers’ probability of leaving their initial school. From the α parameters, I can estimate the probability that a respondent will leave teaching in a particular year, given that she has continued to teach up to that point.

For Research Question 2b, the outcome is VEXIT regardless of whether the individual left the teaching profession. For Research Question 5, the outcome is exitEDU. In this case, I also used logistic regression but did not include the dichotomous variables representing time because I did not collect data on when respondents permanently exited school-based roles. The coefficient of interest is ß₁, again associated with over25.

In fitting models, I made judgments about whether to retain predictors using post-hoc general linear hypothesis (GLH) tests based on differences in the -2 log likelihood goodness-of-fit statistics, comparing differences to a χ² distribution with the appropriate degrees of freedom.

To answer Questions 1, 3, and 4, I conducted chi-square analysis to ascertain whether older and younger entrants to teaching differed on key background characteristics (Research Question 1); had different reasons for leaving teaching (Research Question 3); and entered different occupations upon leaving (Research Question 4).  

FINDINGS

Overall, I found that Teach For America corps members who began teaching after age 25 had different demographic and experiential profiles than their younger counterparts. They also had a higher probability of staying in their initial, low-income placement school, remaining in teaching altogether, and remaining in school-based educational roles than those who began teaching at age 25 or younger. I further determined that, compared with their younger counterparts, older entrants who left teaching did so for different reasons and entered different occupations. In the section that follows, I examine the composition of each subgroup and discuss implications for inferential analysis. I then discuss findings regarding the relationship between age of entry and turnover, and I compare older and younger entrants reasons for leaving teaching and subsequent occupations. I conclude by discussing the findings as a whole and their broader implications.

YOUNGER AND OLDER ENTRANTS TO TEACHING DIFFERED IN IMPORTANT WAYS

TFA teachers who entered the classroom when they were over 25 years old differed in important ways from their younger counterparts. As a group, the older entrants exhibit characteristics that may make them more likely to remain in their initial, low-income placement schools and in the teaching profession overall. As shown in Table 1, older entrants were significantly more likely than younger entrants to be male (35.59% vs. 27.90%) and Black/African American (16.48% vs. 10.94%), and significantly less likely to be Asian American (3.41% vs. 8.38%). Males have been shown to be significantly more likely than females to stay in their schools (Ingersoll, 2001; Luekens et al., 2004). African Americans are generally more likely to stay in the profession (Ingersoll: Murnane et al., 1991) and in low-income schoolssuch as those where TFA teachers workthan are White teachers (Hanushek et al., 2004). Last, Asian American teachers are less likely to stay in the teaching profession than any other racial group (Luekens et al.).

Table 1. Demographic Characteristics of Teach For America Teachers who Entered Teaching Over the Age of 25 and Under 26 (n = 2,029)

	percent
	> age 25	£ age 25
Overall sample	9.04	90.96
Male	35.59	27.90*
Black/African American	16.48	10.94*
Asian American	3.41	8.38*
White	77.27	77.67
Latino	7.07	6.21
Related to a teacher	51.11	58.02~
Applied to a job in addition to applying to TFA	9.50	6.77
Had lived in placement city/town before being placed there by TFA	24.16	12.32***
Had lived in location at least as rural/urban as placement city/town previously	72.78	49.03***
Science, technology, or mathematics major	18.33	20.43
Humanities major	77.22	76.59
Education or social work major	4.44	3.04
Rural placement	23.33	20.32
2000 cohort	21.67	24.52
2001 cohort	24.44	26.89
2002 cohort	53.89	48.59

~p < .10. *p < .05. ** p <.01. ***p < .001.

Older TFA entrants to teaching were also significantly and substantially more likely to have lived in the town where their TFA placement was (24.16% vs. 12.32%) and to have had exposure to a region as rural or urban as their placement (72.78% vs. 49.03%). Boyd et al. (2005) have found that teachers who work near where they grew up are more likely to stay in the profession. It stands to reason that those who spent their youth in a similar type of rural or urban setting may also be more likely to remain. Overall, then, the TFA teachers who entered the profession at an older age display characteristics that may make them more likely than their younger counterparts to stay in their initial, low-income schools and the profession altogether.

OLDER TFA ENTRANTS HAD A LOWER RISK OF LEAVING THEIR LOW-INCOME SCHOOL

Overall, inferential analysis bore out my prediction that the older entrants would be more likely to remain in their schools. Using discrete-time survival analysis, I determined that older entrants odds of voluntarily leaving their low-income, initial placement school were not significantly different from those of younger entrants in their first year of teaching (odds ratio = 1.096) but were significantly lower than (.604 times) the odds of their counterparts in Years 26 (see Table 2). Importantly, these statistics are derived from the final model that controls for the factors, noted earlier, that might influence turnover: gender, race, proximity of their hometown to the placement, and familiarity with the urbanicity/rurality of the placement site. I also controlled for cohort, college major, school level, and school urbanicity.

Table 2. Parameter Estimates (Standard Errors) and Goodness-of-Fit Statistics From Selected Discrete-Time Hazard Models in Which the Risk That a Teacher Will Voluntarily Leave Her Initial, Low-Income Placement School Is Predicted by Age of Entry, Controlling for Key Covariates (n = 2,029)

Model	(1)	(2)	(3)	(4)	(5)
Predictor					final model
Year 1	-2.294**	-2.325**	-2.154**	-2.182**	-2.216**
	(0.094)	(0.098)	(0.112)	(0.170)	(0.178)
Year 2	-0.037	0.002	0.200*	0.189	0.155
	(0.074)	(0.075)	(0.094)	(0.158)	(0.167)
Year 3	-0.124	-0.085	0.133	0.123	0.090
	(0.089)	(0.090)	(0.109)	(0.167)	(0.175)
Year 4	-0.750**	-0.705**	-0.462**	-0.455*	-0.487*
	(0.118)	(0.120)	(0.134)	(0.185)	(0.193)
Year 5	-0.372*	-0.361*	-0.105	-0.108	-0.138
	(0.175)	(0.178)	(0.189)	(0.229)	(0.235)
Year 6	-0.727*	-0.659*	-0.397	-0.357	-0.382
	(0.305)	(0.306)	(0.314)	(0.338)	(0.342)
2001 cohort	0.132	0.131	0.107	0.120	0.119
	(0.089)	(0.090)	(0.091)	(0.091)	(0.091)
2002 cohort	0.112	0.123	0.121	0.128	0.128
	(0.081)	(0.082)	(0.082)	(0.083)	(0.083)
Over age 25		-0.444**	-0.476**	-0.482**	-0.488**
		(0.125)	(0.127)	(0.129)	(0.129)
Over age 25 × Year 1		0.590*	0.563~	0.603*	0.603*
		(0.280)	(0.291)	(0.292)	(0.292)
Female			-0.169*	-0.135~	-0.134~
			(0.074)	(0.075)	(0.076)
Black			-0.447**	-0.440**	-0.433**
			(0.103)	(0.106)	(0.107)
Latino			-0.239~	-0.243~	-0.239~
			(0.129)	(0.130)	(0.130)
Science, technology, or mathematics major				0.306*	0.302*
				(0.120)	(0.124)
Humanities major				0.009	0.009
				(0.105)	(0.105)
Had lived in location at least as rural/urban as placement city/town				0.041	0.046
				(0.070)	(0.070)
Proximity of home				-0.036	-0.030
				(0.031)	(0.032)
Secondary school placement					0.000
					(0.071)
Rural placement					0.073
					(0.087)
-2LL	5607.489	5527.411	5426.139	5376.47	5375.751
D -2LL		80.078***	101.272***	49.669***	0.719
df		2	3	4	2

Note: Covariates include time, cohort, gender, race, college major, proximity of hometown to placement site, familiarity with urban/rural context of placement site, school level, and placement urbanicity.

Standard errors are in parentheses.

 ~p < .10. *p <. 05. **p < .01. ***p < .001.

Figure 1 plots the fitted hazard of leaving ones school in each year for an older entrant to teaching and a younger entrant to teaching. This figure presents the fitted hazard function describing the effect of entering teaching at an older age on teachers risk of voluntarily leaving their initial placement school, given that they had not resigned previously. This plot is based on the pertinent final model in Table 2, with all covariates held at their sample mean.

In Year 1, older and younger entrants have a similar risk of leaving their school, as indicated by the fact that the lines representing each group nearly overlap at 9%10% at Year 1. However, in Years 2 and beyond, younger entrants have a significantly higher risk of leaving their low-income placement school, as reflected in Figure 1 by the elevated line representing this group. In Years 2 and 3, more than 50% of younger entrants still teaching in their initial school are estimated to leave that school. By contrast, approximately 40% of older entrants still teaching in their initial placement schools are estimated to leave that school in each year. Over time, these differences compound, such that 50% of younger entrants to teaching leave their initial, low-income placement schools within 1.83 years, and half of older entrants to teaching leave these schools within 2.11 years. A total of 31.4% of older entrants to teaching remain in their initial, low-income school more than 3 years, compared with 20.2% of younger entrants.

Figure 1. Fitted hazard functions describing the risk of voluntary exit from initial placement school conditional on not having experienced event previously, for younger and older entrants to teaching. Based on final fitted model (Model 5) from Table 2 with covariates held at their sample means (n = 2,029)

How do these figures compare with retention rates for other groups of teachers? Research on teachers beginning their careers in New York City at roughly the same time (20002003) as these TFA teachers revealed that 46% remained in their initial schools more than 3 years (Boyd et al., 2008). Turnover was higher in low-performing schools such as those where TFA teachers are placed.⁷ Boyd and colleagues found that 56% of teachers in low-performing elementary schools and 45% in low-performing middle schools taught for more than 2 years (Boyd et al., 2009). By comparison, 52.3% of older TFA entrants to teaching and 42.0% of their younger counterparts stayed in their initial school longer than 2 years. Thus, although the TFA school-based exit rates are high, there is some evidence that they are roughly similar to general teacher turnover rates in sites like those where TFA teachers are placed: low-performing schools serving high percentages of students of color and low-income students.

OLDER TFA ENTRANTS HAD A LOWER RISK OF LEAVING THE TEACHING PROFESSION

Again, I found that older TFA entrants to teaching have a lower probability of voluntarily leaving the teaching profession, compared with their younger counterparts. Controlling for race, gender, college major, cohort, proximity of their hometown to the placement, familiarity with the placement site, school urbanicity, and school level, I found that older entrants odds of leaving teaching are about half those of younger entrants in their second and third years of teaching (see Table 3). In Years 1 and 46, their odds of exit are slightly lower than (.860 times) those of their counterparts.

Table 3. Parameter Estimates (Standard Errors) and Goodness-of-Fit Statistics From Selected Discrete-Time Hazard Models in Which the Risk That a Teacher Will Voluntarily Leave Teaching Is Predicted by Age of Entry, Controlling for Key Covariates (n = 2,029)

Model	(1)	(2)	(3)	(4)	(5)
Predictors					final model
Year 1	-2.990**	-3.010**	-2.809**	-2.980**	-3.093**
	(0.115)	(0.118)	(0.130)	(0.181)	(0.189)
Year 2	-0.686**	-0.643**	-0.391**	-0.550	-0.661**
	(0.074)	(0.076)	(0.093)	(0.156)**	(0.164)
Year 3	-1.028**	-0.963**	-0.688**	-0.851	-0.958**
	(0.085)	(0.087)	(0.104)	(0.162)**	(0.170)
Year 4	-1.554**	-1.554**	-1.280**	-1.429**	-1.535**
	(0.103)	(0.106)	(0.121)	(0.174)	(0.182)
Year 5	-1.378**	-1.393**	-1.114**	-1.272**	-1.376**
	(0.135)	(0.138)	(0.151)	(0.196)	(0.203)
Year 6	-1.945**	-1.935**	-1.633**	-1.780**	-1.868**
	(0.245)	(0.246)	(0.253)	(0.282)	(0.287)
2001 cohort	0.126	0.134	0.099	0.096	0.095
	(0.087)	(0.088)	(0.089)	(0.090)	(0.090)
2002 cohort	0.127	0.146~	0.153~	0.153~	0.153~
	(0.080)	(0.081)	(0.082)	(0.083)	(0.083)
Over age 25		-0.086	-0.125	-0.132	-0.151
		(0.188)	(0.194)	(0.195)	(0.195)
Over age 25 × Year 2		-0.521*	-0.521~	-0.544*	-0.543*
		(0.265)	(0.272)	(0.274)	(0.274)
Over age 25 × Year 3		-0.718*	-0.739*	-0.736*	-0.735*
		(0.315)	(0.323)	(0.324)	(0.325)
Female			-0.334**	-0.317**	-0.313**
			(0.072)	(0.073)	(0.073)
Black			-0.272**	-0.275*	-0.250*
			(0.105)	(0.108)	(0.109)
Asian			0.282*	0.257*	0.273*
			(0.119)	(0.121)	(0.122)
Science, technology, or mathematics major				0.270*	0.244*
				(0.116)	(0.120)
Humanities major				0.284**	0.278**
				(0.104)	(0.104)
Had lived in location at least as rural/urban as placement city/town				0.156*	0.171*
				(0.069)	(0.069)
Proximity of home				-0.071*	-0.054~
				(0.031)	(0.032)
Secondary school placement					0.023
					(0.070)
Rural placement					0.220**
					(0.084)
-2LL	6072.233	5966.896	5838.289	5784.057	5776.606
∆-2LL		105.337***	128.607***	54.232***	7.451*
df		3	3	4	2

Note: Covariates include time, cohort, gender, race, college major, proximity of hometown to placement site, familiarity with urban/rural context of placement site, school level, and placement urbanicity.

Standard errors are in parentheses.

~ significant at 10%. *significant at 5%. **significant at 1%.

Figure 2 depicts these differences, presenting the fitted hazard function describing the effect of entering teaching at an older age on teachers risk of voluntarily resigning from teaching, given that they had not resigned previously. This plot is based on the pertinent final model in Table 3, with all covariates held at their sample mean.

Figure 2. Fitted hazard functions describing the risk of voluntary exit from the teaching profession conditional on not having experienced event previously, for younger and older entrants to teaching. Based on final fitted model (Model 5) from Table 3 with covariates held at their sample means (n = 2029)

Older and younger entrants to teaching experience similar risk of leaving the profession in Years 1 and 46, as indicated in Figure 2 by the fact that the lines nearly overlap at each of these years. For example, about 4.5% of older entrants are estimated to leave teaching in their first year, and nearly the same percentage (5.2%) of younger entrants are estimated to do the same. However, the groups risk of exit is markedly different in Years 2 and 3, when the bulk of the sample as a whole left teaching. This is shown in Figure 2 in the large gap between the lines representing each groups risk of exit. In Year 2, for instance, 23.6% of older entrants to teaching are estimated to leave, compared with 38.2% of younger entrants. Older entrants significantly and substantially lower risk of leaving teaching in Years 2 and 3 leads to a sizeable divergence in the career paths of these two groups. Fifty percent of younger entrants to teaching leave this profession within 2.46 years, whereas 50% of older entrants to teaching leave teaching within 4.02 years. A total of 61.3% of older entrants to teaching but just 40.1% of younger entrants remain in the profession for more than 3 years.

These figuresespecially that for younger entrants to teachingare lower than retention rates for comparable groups of teachers. Boyd et al. (2009) found that at least 72% of teachers who began teaching in New York City schools between 2000 and 2003 remained in teaching beyond 3 years⁸. From this source, it appears that TFA teachers, especially those who are younger when they enter the classroom, remain in the teaching profession at lower rates than teachers who begin their careers in similar settings.

OLDER TFA ENTRANTS REASONS FOR LEAVING TEACHING WERE DISTINCT

Although TFA corps members who begin teaching over the age of 25 are less likely to leave their school or the profession than their younger counterparts, they still leave teaching in considerable numbers. An important question is: Do older entrants who leave teaching do so for reasons that differ from those cited by younger leavers?

As shown in Table 4, the factors that have the greatest impact on older and younger entrants decision to leave teaching differ in important ways.⁹ Older entrants to teaching are significantly more likely to cite family or health matters as a very or extremely important factor in their decision to leave. Specifically, they are significantly more likely to rate pregnancy/child-rearing, other family reasons, and health as having a large influence on their decision to exit. Older entrants to teaching also were significantly more likely to say that their opposition to new reform measures played a major role in their exit. Last, they were more likely to report that a school staffing action (e.g., layoff or involuntary transfer) played a very or extremely large role in their exit. This makes it especially important that the above analyses of turnover are restricted to voluntary job movements.

Table 4. Respondents Ratings of Factors That Caused Them to Leave Teaching (n = 1,407)

	% responding very or extremely important
Reason for leaving	> age 25	£age 25
Change in residence	22.86	25.83
Pregnancy/child-rearing	8.57	2.78**
Other family reasons	12.38	10.17*
Health	9.53	5.55~
School staffing action (e.g., layoff or involuntary transfer)	9.52	4.78~
Better salary or benefits outside teaching	20.00	14.88
Desire to pursue a position other than K12 teacher	52.34	65.39**
Desire to take course to improve career opportunities within education	22.86	24.44
Desire to take courses to improve career opportunities outside education	25.71	38.47*
School received little support from the community	10.47	7.63
Dissatisfied with job description or responsibilities	21.90	15.80
Did not feel prepared to implement new reform measures	10.47	7.01
Did not agree with new reform measures	11.43	8.41~
Isolation and lack of collaboration at school	16.19	17.34
Lack of discipline at school	20.00	21.51
Lack of sufficient materials at school	11.43	12.88
Poor administrative leadership at school	30.48	31.92
My TFA commitment ended	18.10	25.68~

~p < .10. *p<. 05. **p < .01. ***p < .001.

By contrast, younger entrants to teaching were significantly more likely than older entrants to say that their exit was strongly influenced by their desire to pursue a new job outside K12 teaching, to take classes to improve career opportunities outside education, and by the fact that their TFA commitment ended.

Viewed together, these results suggest that older entrants to teaching in this sample left the profession for reasonsfamily concerns and school staffing actions that some have classified as involuntary (see Dolton & von der Klaauw, 1999) and not related to the substance of their work as classroom teachers or the school as a workplace. In contrast, younger entrants to teaching in this sample appear to leave teaching for reasonsdesire to pursue a new job or studies outside education, the end of the TFA commitmentthat suggest that they are voluntarily selecting alternatives to classroom teaching and possibly K12 schools as workplaces. Given these patterns, are older entrants to teaching who leave the profession likely to remain in schools? Are they likely to reenter teaching, as many individualsespecially mothers of young children (Beaudin, 1993)eventually do?

OLDER TFA ENTRANTS WHO LEFT TEACHING REMAINED IN SCHOOLS

The answer to this question lies in data regarding the occupations that TFA corps members who left teaching subsequently entered. As hypothesized, I found that older entrants to teaching who left the profession entered different occupations than did their younger counterparts (Table 5). Most notably, older entrants to teaching were significantly more likely than younger entrants to become a K12 specialist or administrator after they left the classroom. They were also more likely to work as craftsmen; editors, writers, or artists; business people; or higher education administrators. In contrast, younger entrants to teaching were more likely than their older counterparts to enter graduate school. These findings reinforce the notion, surfaced in leavers reasons for exiting, that older entrants leave teaching but remain engaged with education whereas younger entrants leave for different sectors altogether. However, these findings are uncontrolled. A key question is whether older entrants to teaching are more likely to stay in school-based roles (i.e., teacher, specialist, or administrator) than are younger entrants, controlling for key covariates.

Table 5. Current Occupations of Teach For America Teachers Who Left Teaching During the Time Period Observed (n=1,411)

Occupation	> age 25	£age 25
Clerical	1.94	.62
Craftsman/laborer (e.g., farmer, carpenter, chef)	1.94	.39*
Law enforcement/military	.97	.088
Business/financial services employee	10.68	4.63**
Legal professional	1.94	3.93
Medical professional	.97	1.77
Engineer/architect/scientist	1.94	1.00
Computer/technical worker	2.91	.62*
Editor/writer/artist	6.80	1.93**
Sales or service worker	.97	1.08
Community/social service employee	3.88	3.01
K12 teacher	14.56	10.64
K12 specialist (including guidance counselor)	8.74	3.78*
K12 administrator	9.71	5.47~
Educational consultant or researcher	3.88	4.55
Educational nonprofit worker (outside TFA)	11.65	12.17
TFA employee	5.83	5.78
Professor or instructor in higher education	1.94	2.16
Higher education administrator	3.88	1.46~
Urban planner	.00	.39
Government employee	0.00	1.39
Stay-at-home parent	2.91	1.31
Researcher	.97	1.77
General nonprofit employee	1.94	2.08
Educational for-profit employee	.00	.85
Unemployed	.00	.62
Graduate student	13.59	37.32***

~p < .10. *p < .05. **p < .01. ***p <. 001.

OLDER ENTRANTS HAD A LOWER PROBABILITY OF LEAVING SCHOOL-BASED ROLES

Controlling for race, college major, cohort, proximity of their hometown to the placement site, familiarity with the urbanicity/rurality of the placement site, school urbanicity, and school level, I found that older entrants to teaching had a significantly lower probability of leaving school-based roles than did younger entrants to teaching (Table 6). However, I found that these results differed for males and females. Younger male entrants to teaching had a 81.09% probability of leaving school-based roles within the data collection period, whereas older male entrants had a 56.7% chance of doing so. Among females, younger entrants had a 72.6% chance of leaving schools, whereas older entrants to teaching had a 66.5% probability of doing so. The odds that an older male entrant to teaching would leave a school-based role were .306 times the odds of their younger male counterparts. This difference is significant (chi-square = 9.89, p = .002). The odds that an older female would stop working in schools were .749 times the odds that a younger female would do so, but this difference is not significant (chi-square = .83, p = .37) On the whole, we see that older male entrants to teaching have a significantly higher probability of remaining at work in schools than do younger male and female (chi-square = 3.88, p = .05) entrants to teaching. This finding echoes earlier research on age at entry and turnover conducted on high-achieving entrants to teaching (Donaldson, 2008). Older females are significantly more likely than younger male entrants to continue to work in schools (chi-square = 5.10, p = .02).  Figure 3 depicts the probability that each of these groups will leave school-based work in any given year. Older males probability of leaving school-based roles is lowest, followed by the risk of older females, younger females, and younger males.

Table 6. Parameter Estimates (Standard Errors) and Goodness-of-Fit Statistics From Selected Logistic Regression Models in Which the Probability That a Teacher Will Permanently Leave A K-12 School-Based Role (i.e., Serving as a Teacher, a Specialist, or an Administrator) Within the Time Period Observed Is Predicted by Age of Entry, Controlling for Key Covariates (n = 2,029)   

	(1)	(2)	(3)	(4)	(5)
Predictor					Final Model
2000 cohort	1.154**	1.183**	1.580**	1.373**	1.239**
	(0.123)	(0.126)	(0.180)	(0.335)	(0.354)
2001 cohort	1.339**	1.400**	1.749**	1.544**	1.405**
	(0.124)	(0.127)	(0.176)	(0.331)	(0.351)
2002 cohort	1.568**	1.612**	1.989**	1.782**	1.646**
	(0.103)	(0.106)	(0.166)	(0.328)	(0.347)
Over age 25		-0.581*	-1.179**	-1.185**	-1.185**
		(0.226)	(0.363)	(0.375)	(0.377)
Female			-0.518**	-0.516**	-0.482**
			(0.166)	(0.170)	(0.171)
Female × Over age 25			0.925~	0.921~	0.896~
			(0.473)	(0.488)	(0.490)
Black				-0.272	-0.233
				(0.220)	(0.222)
Latino				0.448	0.455
				(0.324)	(0.325)
Science, technology, or  mathematics major				0.161	0.001
				(0.265)	(0.273)
Humanities major				0.019	0.008
				(0.236)	(0.236)
Had lived in location at least as rural/urban as placement city/town				0.197	0.204
				(0.145)	(0.146)
Proximity of home				0.032	0.021
				(0.068)	(0.069)
Secondary school placement					0.385**
					(0.147)
Rural placement					-0.116
					(0.172)
Goodness of fit
-2LL	1411.124	1392.357	1377.905	1335.94	1332.216
∆-2LL		18.767***	14.452***	41.965***	3.724
df		1	2	6	2

Note: Key covariates include time, cohort, gender, race, college major, proximity of hometown to placement site, familiarity with urban/rural context of placement site, school level, and placement urbanicity.

Standard errors are in parentheses.

~p < .10. *p < .05. **p < .01. ***p < .001.

Figure 3. Fitted hazard functions describing the risk of permanent exit from a school-based job, for younger and older entrants to teaching. Based on final fitted model (Model 5) from Table 6 with covariates held at their sample means (n = 2,029)

DISCUSSION

In sum, I found that Teach For America corps members who began teaching after age 25 had a lower probability of leaving their initial, low-income placement school, exiting the teaching profession, and leaving school-based roles than did younger entrants to teaching. I further found that older entrants in this sample who left did so for reasons that differed from those selected by their younger counterparts. Older entrants who left teaching were significantly more likely than younger entrants to cite family, child-bearing, and health reasons to explain why they left. They were also more likely to say that they left because of a school staffing action than were younger entrants. By contrast, younger new teachers who left teaching were significantly more likely than their older counterparts to enter graduate school. Finally, older entrants to teaching who left the profession were significantly more likely than younger entrants to take nonteaching school-based jobs, such as that of a literacy specialist, guidance counselor, or administrator.

Viewed broadly, these findings suggest that older entrants to teaching may prove a good source of teachers for low-income schools. On all measures, older entrants demonstrated more commitment to low-income schools and the teaching profession than did their younger counterparts. Notably, even though they began their careers in challenging, low-income schools, they left the teaching profession at rates (61.3% in 3 years) that are not that distant from some estimates of attrition of teachers (72%) who began their careers in urban schools, some of which were likely less challenging than those where TFA teachers typically work (see Boyd et al., 2006, 2009).

Because older entrants to teaching in this sample differed substantially and significantly from younger entrants to teaching, these findings run counter to Supers (1957) model of career development, which posits that entrants to a profession go through the same stages of career development, regardless of age of entry. Although the division of this sample into older and younger entrants to teaching does not align exactly with Levinsons framework, the findings echo his model of the human life cycle. Within the TFA sample, it appears that the older group demonstrates some of the dedication to work and career that, according to Levinson et al. (1978), solidifies in ones 30s. Younger entrants to teaching in this sample, by contrast, seem to exhibit some of the tentative and conditional commitment to the career of teaching that Levinsons model suggests characterize the Early Adult Transition phase. Thus, although not a perfect fit, these findings lend greater empirical support to Levinsons framework than Supers model.

It is important to note that older entrants to teaching were more likely to remain in their jobs, the teaching career, and school-based work even when many of the factors known to correlate with turnover were controlled. Controlling for gender, race, distance of hometown to teaching placement, familiarity with the urbanicity/rurality of the placement, and other observable characteristics, older entrants to teaching still had a lower probability of turnover. Outside these observable characteristics, there appears to be something in the older entrant or her experience in teaching that predisposes her to stay. This warrants investigation. Do older entrants to teaching have more financial obligations than younger entrants and are thus less likely to leave their jobs or careers for fear of losing a much-needed paycheck? Do older entrants possess greater psychological resources than do younger entrants and are thus more able to handle the stress of learning to teach in low-income schools? Do older entrants, by virtue of their age and prior experience, integrate more easily into the professional culture of low-income schools, thereby gaining access to resources and social capital that help them remain in their job?

A related finding pertains specifically to older entrants higher probability of being African American, male, and a prior resident of the setting in which they taught through TFA, and lower probability of being Asian American. Is there something about having these qualities and being older that helps such teachers continue to work in low-income classrooms? Does their gender, race, and familiarity with the city or town in which they teach interact with their age to uniquely equip them to persist as teachers in these settings? Beyond their retention, are they more successful in teaching students in these settings than are their younger counterparts?

Much remains to be learned about the relationship between a teachers age at entry and her probability of staying in her school and the profession as a whole. One intriguing avenue for further research centers on the reasons individuals selected to explain why they left. Older entrants to teaching cited factorsfamily, health, school staffing actions that suggest they left teaching involuntarily. By contrast, younger entrants to teaching appear to leave voluntarily because they have decided that alternative career paths outside of education better fit with their goals. Why do younger entrants dismiss teaching and education more readily than do older entrants in this sample? Were the older entrants more committed to teaching or working in schools from the beginning, or did something happen in their placement that increased their desire and commitment to work in schools but decreased their younger counterparts inclination to do so?

Research should also build on prior work that examines individuals relationship with work in smaller age increments. For example, DeAngelis and Presley (2007) divided respondents over age 25 into a group aged 2634 and one age 35 and above. They found that the 2634 age groups turnover patterns were more similar to those of the under 25 group than those of the over 35 group. It would be helpful to similarly subdivide my sample to see whether the oldest entrants to teaching are, in fact, driving my findings.   

LIMITATIONS

This study is limited in a number of ways. First, as noted, my decision to adopt age 25 as the cut-off between older and younger teachers does not align perfectly with Levinsons framework, which posits an early career transition from age 23 to 28, followed by the Age 30 Transition. Although adopting a cutoff between 28 and 30 would have been the best choice to test Levinsons theory directly, it was impossible to do with my data set given small sample sizes over age 30. This is a limitation of this study that further research should investigate.

Second, the external validity of this study is limited. This research has focused on Teach For America teachers, who are a distinct subset of all new teachers today. With high test scores and diplomas from selective colleges, they differ from the average new teacher in U.S. schools in a number of ways that limit the generalizability of these findings. Again, further research should examine whether the trends identified here hold for a more typical population of teachers.

Third, the internal validity of this study is circumscribed. Although the relationship between age at entry and turnover was strong and consistent in this study, there was no mechanism at my disposal to establish a causal relationship between a teachers age at entry and subsequent turnover. Age, in and of itself, may not have caused these individuals to stay in or leave their initial placement schools or the profession overall. It is possible, for example, that younger new teachers in this sample were systematically assigned to more difficult schools or teaching assignments, which then drove their turnover. This points to the need for further research examining whether a teachers age does in fact cause her to remain in or leave teaching.

IMPLICATIONS

This study has implications for efforts to recruit and retain promising teachers in low-income schools. Given these findings, if low-income school leaders want to secure a stable and committed faculty, they would be wise to hire older new teachers. As discussed, older novices higher retention rates could translate into lower financial, achievement, and cultural costs for schools. More broadly, districts seeking to develop human capital across multiple levels of the system might also consider targeting older individuals as a source of new teachers. Not only do these people appear to teach longer, but if they leave teaching, they are likely to remain in schools in other roles. Thus, if a district wants a good return on the professional development and other resources it invests in teachers, older novices appear to be a safer bet than younger new teachers. Districts might focus recruiting efforts on older entrants who are also male, African American, and residing in a city or town where TFA offers teaching placements.

Districts would also be wise to examine the role of involuntary turnover in a larger analysis of district retention. Who leaves involuntarily? Who leaves voluntarily? What are the implications for teacher quality at the school site? Given our current economic conditions, involuntary turnover is likely to rise in the coming months. Districts would be wise to attend to this and, to the extent possible, avoid laying off individuals who are promising teachers and want to remain in their schools and the profession.

An additional implication of this study is that more research needs to focus on teacher retention in low-income schools. Besides this examination of Teach For America retention, I know of no nationwide, longitudinal data set that tracks in detail teacher retention in low-income schools.  If we are serious about addressing teacher recruitment and retention in these settings, this sort of resource is sorely needed.

It is clear that low-income schools face large challenges in attracting and retaining skilled and committed teachers in the coming years. By focusing on older new teachers, these schools may slowly build a cadre of strong teachers who lend their talents, energy, and enthusiasm to the important work that takes place in their classrooms.

Appendix

Variables Included in Statistical Models

Variable	Description
Outcomes
Voluntarily leave the placement school	Time-varying dichotomous outcome variable indicating whether the teacher voluntarily left her school for the first time; measured repeatedly during each of the up to six academic years beginning in the first year in which respondent i taught (coded 1 if the teacher experienced the event of interest during year j, 0 otherwise).
Voluntarily leave the teaching profession	Time-varying dichotomous outcome variable indicating whether the teacher voluntarily left the teaching profession for the first time; measured repeatedly during each of the up to six academic years beginning in the first year in which respondent i taught (coded 1 if the teacher experienced the event of interest during year j, 0 otherwise).
Leave a school-based role	Dichotomous outcome variable indicating whether the teacher left school-based roles permanently during the time period observed (coded 1 if the teacher left school-based roles permanently, 0 otherwise).
Predictors
Question Predictors
Over age 25	Dichotomous variable indicating whether the teacher was over age 25 when she began teaching (coded 1 if the teacher was over 25, 0 otherwise).
Time specifications
Year 1Year 6	System of six dichotomous variables that distinguish each of the up to six years during which respondents could have taught.
Covariates
2000 cohort, 2001 cohort, 2002 cohort	System of three dichotomous variables corresponding to the year in which a respondent entered teaching: C1=1 if respondent was in the 2000 cohort, C2=1 for 2001 cohort, etc.
Female	Time-invariant dichotomous predictor indicating respondents self-reported gender (coded 1 if female, 0 if male).
Black/African American	Time-invariant dichotomous predictor indicating respondents self-reported race (coded 1 if respondent identified as African American or Black, 0 if respondent did not).
Latino	Time-invariant dichotomous predictor indicating respondents self-reported race (coded 1 if respondent identified as Latino or Hispanic, 0 if respondent did not).
Asian American	Time-invariant dichotomous predictor indicating respondents self-reported race (coded 1 if respondent identified as Asian or Pacific Islander, 0 if respondent did not).
Science, technology, or mathematics major	Time-invariant dichotomous predictor indicating whether respondent majored in the sciences, technology, engineering, or mathematics (coded 1 if respondent majored in one of these subjects, 0 otherwise).
Humanities major	Time-invariant dichotomous predictor indicating whether respondent majored in English, social studies/social science, or the arts (coded 1 if respondent majored in one of these subjects, 0 otherwise).
Had lived in location at least as rural/urban as placement city/town previously	Dichotomous variable indicating whether respondent grew up in a region as urban/rural as her TFA placement site (coded 1 if yes, 0 if no).
Proximity of home	Ordinal variable indicating how close respondent lived to TFA placement site prior to being placed there (coded 1 if respondent lived more than a 5-hour drive away; 2 if respondent lived between a 2- and 5-hour drive away; 3 if respondent lived within a 2-hour drive away; 4 if respondent lived in the town/city where she was placed.
Secondary school placement	Time-varying dichotomous variable that captures whether the teacher taught at the elementary level (Grades 712 check) in Year 1 (coded 1 if the teacher taught these grades during year 1, 0 otherwise).
Rural placement	Time-invariant dichotomous predictor indicating respondents self-reported teaching location (coded 1 if respondent indicated a region classified by researcher as rural, 0 if respondent indicated a region classified as urban).

Notes

1. Although this is not the case for Illinois, the shift toward older entrants to teaching in some states may be due in part to some states requirement that individuals obtain a masters degree to qualify for teacher certification. It may also be due to some universities requirement that teacher education students complete an internship (e.g., Michigan State) or masters degree (e.g., University of Connecticut) after completing a bachelors degree to conclude the universitys teacher education program.

2. For comprehensive reviews of factors associated with teacher turnover, see Borman & Dowling, 2008; Guarino et al., 2004; Guarino, Santibañez, & Daley, 2006; and Johnson et al., 2005.

3. Most studies of teachers careers indicate that attrition from the teaching profession declines substantially after the fifth year (Kirby et al., 1999; Murnane et al., 1991; Stinebrickner, 2001), which suggested we should follow teachers for at least 5 years. However, because this is a retrospective study, I wanted to focus on relatively recent TFA cohorts to reduce recall errors (Taris, 2000) and take advantage of TFAs more reliable contact information for members of the more recent cohorts.

4. I included three cohorts in my data analyses to improve statistical power. A priori power analysis suggested that the resulting sample size should provide sufficient statistical power to detect small effects (Cohen, 1977).

5. Each contained similar proportions of females, Latinos, Asians, American Indians, members of each cohort, and teachers assigned to elementary, middle, and high school. The two statistically significant differences between sample and census were in the percentage of those who identified as only Black/African American (14.53% of the census vs. 10.43% of the sample) and those who identified as only White (67.59% of the census vs. 73.73% of the sample). However, only 90.6% of individuals in the census provided information about their race. By contrast, 97.4% of respondents reported their race on my survey. Thus, it is possible that the reported racial composition of the census is not a good standard against which to measure the representativeness of our sample. Nonetheless, my sample may not be representative of Blacks and Whites.

6. Overall sample size and sample size within subgroups (e.g., individuals who entered teaching at age 28 or older) may have been increased by extending the sampling frame to TFA cohorts prior to 2000. I chose not to do this because of (a) research indicating that recall becomes less accurate over time (i.e., the longer ago an event, the more difficult it is to recall it with accuracy) and (b) discussions with TFA officials that indicated the organization had made substantive changes in TFAs summer preparation program and in-service supports in the late 1990s, making the TFA experience for those who entered TFA in the 1990s quite different than it was for those who began in the early 2000s.

7. In a different analysis, Boyd, Grossman, Lankford, Loeb, and Wyckoff (2006) found that Teach For America teachers in New York City schools were more likely to teach poor and low-performing students than teachers who entered the system through other routes. Across the system, 82% of first-year teachers students qualified for free lunch. A total of 92% of TFA teachers students qualified for this service. Prior test scores were -0.26 on average, but -0.51 for TFA teachers.

8. Because they based their analysis on a New York state data set, the actual percentage of teachers continuing to teach could be higher if individuals who left the state continued to teach elsewhere.

9. The analysis in this section includes both voluntary and involuntary leavers.

References

Adams, G. (1996). Using a Cox regression model to examine voluntary teacher turnover. Journal of Experimental Education, 64, 267286.

Alliance for Excellent Education. (2005). Teacher attrition: A costly loss to the nation and to the states. Retrieved from http://www.all4ed.org/files/archive/publications/TeacherAttrition.pdf

Azimi, N. (2007, September 30). Why Teach For America? New York Times Magazine.

Barnes, G., Crowe, E., & Schaefer, B. (2007). The cost of teacher turnover in five school districts: A pilot study. Washington, DC: National Commission on Teaching and Americas Future. Retrieved from http://www.eric.ed.gov/PDFS/ED497176.pdf

Beaudin, B. Q. (1993). Teachers who interrupt their careers: Characteristics of those who return to the classroom. Educational Evaluation and Policy Analysis, 15(1), 5164.

Birkeland, S. E., & Curtis, R. (2006). Ensuring the support and development of new teachers in the Boston Public Schools: A proposal to improve teacher quality and retention. Boston: Boston Public Schools.

Borman, G. D., & Dowling, N. (2008). Teacher attrition and retention: A meta-analytic and narrative review of the research. Review of Educational Research, 78, 367409.

Boyd, D., Grossman, P., Lankford, H., Loeb, S., & Wyckoff, J. (2006). How changes in entry requirements alter the teacher workforce and affect student achievement. Education Finance and Policy, 1, 176216.

Boyd, D., Grossman, P., Lankford, H., Loeb, S., Wyckoff, J., & National Bureau of Economic Research, C. A. (2008). Who leaves? Teacher attrition and student achievement (NBER Working Paper No. 14022). Cambridge, MA: National Bureau of Economic Research.

Boyd, D., Grossman, P., Lankford, H., Loeb, S., Wyckoff, J., & Urban, I. (2009). Who leaves? Teacher attrition and student achievement (Working Paper No. 23). Washington, DC: National Center for Analysis of Longitudinal Data in Education Research.

Boyd, D. J., Lankford, H., Loeb, S., & Wyckoff, J. (2005). The draw of home: How teachers preferences for proximity disadvantage urban schools. Journal of Policy Analysis and Management, 24, 113132.

Bryk, A. S., & Schneider, B. (2003). Trust in schools: A core resource for school reform. Educational Leadership, 60(6), 4044.

Bullock, K. M., & Scott, W. H. (1993). The shortage of teachers of mathematics, physics and technology: Can mature entrants fill the gap? Research in Science and Technological Education, 11(2), 127139.

Cohen, J. (1977). Statistical power analysis for the behavioral sciences. New York: Academic Press.

Crow, G., & Others. (1990). No more business as usual: Career changers who become teachers. American Journal of Education, 98, 197223.

DeAngelis, K., & Presley, J. (2007). Leaving schools or leaving the profession: Setting Illinois record straight on new teacher attrition (Illinois Education Research Council Policy Research Report: IERC 2007-1). Edwardsville: Illinois Education Research Council.

Dillman, D. A. (1978). Mail and telephone surveys: The total design method. New York: Wiley.

Dillman, D. A. (1991). The design and administration of mail surveys. Annual Review of Sociology, 17, 225249.

Dolton, P., & van der Klaauw, W. (1999). The turnover of teachers: A competing risks explanation. Review of Economics and Statistics, 81, 543550.

Donaldson, M. L. (2008, March). The retention of academically talented teachers: Whether, when, and why graduates of one selective preparation program leave teaching. Paper presented at the annual meeting of the American Educational Research Association, New York, NY.

Donaldson, M. L., & Johnson, S. M. (2010). The price of misassignment: The role of teaching assignments in Teach For America teachers exit from low-income schools and the teaching profession. Educational Evaluation and Policy Analysis, 32, 299323.

Donaldson, M. L., & Johnson, S. M. (2011). Teach For America teachers: How long do they teach? why do they leave? Phi Delta Kappan, 93(2), 4751.

Ehrenberg, R. G., & Brewer, D. J. (1994). Do school and teacher characteristics matter? Evidence from high school and beyond. Economics of Education Review, 13(1), 117.

Ehrenberg, R. G., & Brewer, D. J. (1995). Did teachers verbal ability and race matter in the 1960s? Coleman revisited. Economics of Education Review, 14(1), 121.

Fowler, F. (2002). Survey research methods (3rd ed.). Thousand Oaks, CA: Sage.

Freidus, H. (1994, April). Supervision of second career teachers: Whats our line? Paper presented at the annual meeting of the American Educational Research Association, New Orleans, LA.

General Teaching Council for England. (2006). 2005 annual digest of statistics. London: Author.

Goddard, Y. L., Goddard, R., & Tschannen-Moran, M. (2007). A theoretical and empirical investigation of teacher collaboration for school improvement and student achievement in public elementary schools. Teachers College Record, 109, 877896.

Guarino, C., Santibañez, L., Daley, G., & Brewer, D. (2004). A review of the research literature on teacher recruitment and retention (No. TR-164-EDU). Santa Monica, CA: RAND.

Guin, K. (2004). Chronic teacher turnover in urban elementary schools. Education Policy Analysis Archives, 12(42).

Hanushek, E. A., Kain, J. F., & Rivkin, S. G. (2004). Why public schools lose teachers. Journal of Human Resources, 39, 326354.

Heinz, M. (2008). The composition of applicants and entrants to teacher education programmes in Ireland: Trends and patterns. Irish Educational Studies, 27, 223240.

Ingersoll, R. M. (2001). Teacher turnover and teacher shortages: An organizational analysis. American Educational Research Journal, 38, 499534.

Jackson, C. K., & Bruegmann, E. (2009). Teaching students and teaching each other: The importance of peer learning for teachers (Working Paper No. 15202). Cambridge, MA: National Bureau of Economic Research.

Johnson, S. M., Berg, J. H., & Donaldson, M. L. (2005). Who stays in teaching and why: A review of the literature on teacher retention. Washington, DC: NRTA.

Johnson, S., Kardos, S., Kauffman, D., Liu, E., & Donaldson, M. L. (2004). The support gap: New teachers early experiences in high-income and low-income schools. Education Policy Analysis Archives, 12(61).

Kardos, S. M. (2004). Supporting and sustaining new teachers in schools: The importance of professional culture and mentoring. Harvard University, Cambridge, MA.

Kelly, S. (2004). An event history analysis of teacher attrition: Salary, teacher tracking, and socially disadvantaged schools. Journal of Experimental Education, 72, 195220.

Kirby, S. N., Berends, M., & Naftel, S. (1999). Supply and demand of minority teachers in Texas: Problems and prospects. Educational Evaluation and Policy Analysis, 21, 4766.

Levinson, D. J. (1986). A conception of adult development. American Psychologist, 41, 313.

Levinson, D. J., Darrow, C. N., Klein, E. B., Levinson, M. H., & McKee, B. (1978). The seasons of a mans life. New York: Knopf.

Levinson, D. J., & Others. (1976). Periods in the adult development of men: Ages 18 to 45. Counseling Psychologist, 6(1), 2125, 76.

Lim, C. O. (1986). Assessment and characteristics of adult development of male Filipino college professors. Chicago: Loyola University of Chicago.

Liu, E. (2004). New teachers experiences of hiring in four states. Harvard University, Cambridge, MA.

Luekens, M. T., Lyter, D. M., Fox, E. E., & Chandler, K. (2004). Teacher attrition and mobility: Results from the teacher follow-up survey, 2000-01. Washington, DC: National Center for Education Statistics.

Marinell, W. (2009, April). Will mid-career entrants help avert a teacher shortage, reduce racial and gender imbalances, and fill vacancies in hard-to-staff subjects and schools? Paper presented at the annual meeting of the American Educational Research Association, San Diego, CA.

Milanowski, A. T., & Odden, A. (2007). A new approach to the cost of teacher turnover (Working Paper No. 13). Center on Reinventing Public Education, Seattle, WA. Retrieved from http://www.crpe.org/cs/crpe/download/csr_files/wp_sfrp13_milanowskiodden_aug08.pdf

Minter, L. E., & Samuels, C. A. (1998). The impact of the dream on womens experience of the midlife transition. Journal of Adult Development, 5(1), 3143.

Murnane, R. J., Singer, J. D., Willett, J. B., Kemple, J. J., & Olsen, R. J. (1991). Who will teach? Policies that matter. Cambridge, MA: Harvard University Press.

National Commission on Teaching and Americas Future. (2003). No dream denied: A pledge to Americas children: Summary report. Washington, DC: Author.

Neild, R. C., Useem, E., Travers, E. F., & Lesnick, J. (2003). Once and for all: Placing a highly qualified teacher in every Philadelphia classroom. Philadelphia: Research for Action.

Ornstein, S., Cron, L., & Slocum, J. (1989). Life stage versus career stage: A comparative test of the theories of Levinson and Super. Journal of Organizational Behavior, 10, 117133.

Payne, S. L. (1951). The art of asking questions. Princeton, NJ: Princeton University Press.

Peske, H., & Haycock, K. (2006). Teaching inequality: How poor and minority students are shortchanged on teacher quality. Washington, DC: Education Trust.

Resta, V., Huling, L., & Rainwater, N. (2001). Preparing second-career teachers. Educational Leadership, 58(8), 6063.

Rivkin, S. G., Hanushek, E. A., & Kain, J. (2005). Teachers, schools and academic achievement. Econometrica, 73, 417458.

Rockoff, J. (2004). The impact of individual teachers on student achievement: Evidence from panel data. American Economic Review, 94, 247252.

Serow, R., & Forrest, K. (1994). Motives and circumstances: Occupational-change experiences of prospective late-entry teachers. Teaching and Teacher Education [serial online].

Shockley, R., Guglielmino, P., & Watlington, E. J. (2006). A national crisis in teacher education: What are the costs? Retrieved January 15, 2012, from http://www.pearsonassessments.com/hai/images/NES_Publications/2006_11ShockleyGuglielminoWatlington_558_1.pdf

Simon, C. C. (2009, July 20). So you want to be a teacher for America? The New York Times (Education Life Suppl.), p. 8.

Singer, J. D., & Willett, J. B. (1993). Its about time: Using discrete-time survival analysis to study duration and the timing of events. Journal of Educational Statistics, 18, 155195.

Staiger, D., & Rockoff, J. (2010). Searching for teachers with imperfect information. Journal of Economic Perspectives 24(3), 97118.

Stinebrickner, T. R. (2001). A dynamic model of teacher labor supply. Journal of Labor Economics, 19, 196230.

Sudman, S., & Bradburn, N. M. (1982). Asking questions: A questionnaire design. San Francisco: Jossey-Bass.

Super, D. E. (1957). The psychology of careers. New York: Harper.

Super, D. E. (1984). Career and life development. In D. Brown & L. Brooks (Eds.), Career choice and development (pp. 192234). San Francisco: Jossey-Bass.

Taris, T. W. (2000). A primer in longitudinal data analysis. Thousand Oaks, CA: Sage.

Cite This Article as: Teachers College Record Volume 114 Number 10, 2012, p. 1-37 https://www.tcrecord.org ID Number: 16677, Date Accessed: 10/9/2021 1:27:29 AM Purchase Reprint Rights for this article or review