Do Mentoring and Induction Programs Have Greater Benets for Teachers Who Lack Preservice Training?

61

4
DO MENTORING AND INDUCTION PROGRAMS HAVE GREATER BENEFITS FOR TEACHERS WHO LACK PRESERVICE TRAINING?
Laura Duke Adam Karson Justin Wheeler
The purpose of this paper is to examine the impact of mentoring and induction programs on teacher retention, as measured by teachers commitment to their profession. Using data from the 1999-2000 Schools and Stafng Survey, we perform logistic regression analyses to model the effect of induction and its different components on teacher commitment, and compare the marginal impact of induction programs on teachers with and without degrees in education. Our results show that teachers who have had mentors or gone through induction programs in their rst year of teaching are more likely to be committed to the teaching profession. Moreover, mentoring and induction programs have a greater marginal benet for teachers without education degrees than for those with education degrees. Based on our results, we recommend that districts (1) provide mentoring and induction programs for all teachers, and (2) allow school-level exibility in tailoring induction and mentoring programs. 1

INTRODUCTION
In the landmark No Child Left Behind school accountability legislation,
Laura Duke, Adam Karson and Justin Wheeler are Master in Public Policy candidates at the Terry Sanford Institute of Public Policy, Duke University (laura.duke@duke. edu, [email protected], [email protected]). 7

62

Laura Duke, Adam Karson, Justin Wheeler

the U.S. government acknowledged the importance of highly qualied teachers in promoting high student achievement. As teachers from the baby boom generation begin to retire, schools must nd new teachers to ll these vacancies. Unfortunately, over the last decade, new teachers (i.e., those with fewer than four years of full-time teaching experience) are more likely to leave teaching or move to another school than any other experience level (Tabs 2004, 9). Many cite lack of support or poor preparation as justications for leaving their current teaching positions. This turnover costs schools and teacher preparation programs time and money as they continually must nd and prepare new educators. Some schools and districts attempt to stem the tide of new teacher attrition through comprehensive induction programs that include mentoring. These programs seek to address the support and preparation issues cited by departing teachers as the cause of their dissatisfaction. In the 1999-2000 school year, approximately 232,000 individuals entered the teaching profession. One year later, about 287,000 teachers left the occupation, the majority of them far short of retirement age (Ingersoll 2003). In part because of the necessity created by these unprecedented classroom vacancies, many school systems have established emergency or alternative certication programs designed to place participants in the classroom as soon as possible. Unlike traditional education degree programs at colleges and universities, many of these lateral entry programs require fewer hours of preservice coursework and student teaching. Instead, they require observation and seminar meetings after the participant begins full-time teaching. Given the rising popularity of these lateral entry programs and the increasing evidence that perceptions of poor preparation and support lead novice teachers to leave the profession at higher rates, one might expect that comprehensive induction programs would have a positive effect on teacher commitment and thus retention. This may be particularly true among teachers lacking signicant preservice coursework in educational theory and practice (i.e., teachers lacking a bachelors degree in education). To examine this theory, we test three hypotheses:
1. Mentoring and induction will have a positive effect on teacher commitment. 2. The positive effects of mentoring and induction on retention will be greater for teachers who have not had the extensive preservice training required to complete a bachelors degree in education.

Do Mentoring and Induction Programs Have Greater Benets for Teachers Who Lack Preservice Training?

63

3. More comprehensive mentoring and induction programs (i.e., those that incorporate numerous induction strategies concurrently) will yield stronger positive effects on teacher retention for teachers without degrees in education.

Our research goes beyond previous efforts by attempting to link induction programs with the amount of preservice education training. Previous studies (including Smith and Ingersoll 2004; Moir 2003) nd that induction programs can reduce teacher attrition by helping teachers feel more prepared, but they fail to specically target those teachers whose lateral entry to teaching provided them with less preparation. Previous studies (including Gitomer, Latham and Ziomek 1999; Goe 2002) also nd evidence of deciencies in teachers who pursue these alternative paths to the classroom without considering whether induction programs may make up for a lack of preservice training. Using data from the teacher and school portions of the 1999-2000 Schools and Stafng Survey, we performed logistic regressions to test these hypotheses. Based on our results, we conclude that mentoring and induction programs positively affect teachers commitment to the profession and that this positive effect is stronger on teachers who do not have bachelors degrees in education than on those who do. We also isolated particular features of induction programs as more effective than others in reducing the likelihood of attrition among teachers with and without education degrees. Consequently, we recommend the following two policies for districts to implement in order to improve turnover rates: (1) increase access to mentoring and induction programs for all beginning teachers, and (2) allow campus-level discretion in tailoring induction and mentoring to meet local needs

LITERATURE REVIEW
Comparing Teacher Training Programs
The increasing popularity of lateral entry teacher preparation programs raises a philosophical question about how new teachers best learn to teach. Proponents of lateral entry programs argue that teaching skills can be picked up on-the-job as long as new teachers possess subject matter expertise. In contrast, supporters of traditional, university-based education degree programs defend the importance of preservice professional training in theory and practice (Stoddart and Floden 1995, 9). In a meta-analysis of various measures of teacher preparation, Wilson, Floden and Ferrini-Mundy (2002) nd education coursework a better

64

Laura Duke, Adam Karson, Justin Wheeler

predictor of teaching success than subject matter major or GPA prior to entering the lateral entry program. Teachers credit their education coursework with providing essential instructional and disciplinary skills. However, critics argue that the considerable variation among teacher training programs renders evaluations of such programs questionable, if not impossible. Data limitations prevent most studies from directly linking the content of education coursework to student achievement. Instead, many studies use teacher certication status or teachers scores on standardized certication examinations as proxies for the degree of pedagogical training (Goldhaber and Anthony 2003, 11). Although many lateral entry programs do require professional training, most demand fewer hours of preservice formal educational coursework, and instead require more hours of supervised eld experience as a fulltime teacher (Stoddart and Floden 1995, 8). However, formal preservice coursework provides the information that most states deem essential for new teachers. Gitomer, Latham and Ziomek (1999, 24) nd that educational coursework improves teachers performance on the Educational Testing Services Praxis II tests, the most widely used licensure tests in the country. Several studies conclude that certied teachers elicit greater student achievement than uncertied teachers. Goe (2002) nds evidence that California schools with higher percentages of teachers with emergency permits display lower levels of student achievement as reected by the state Academic Performance Index (API), even after controlling for students socioeconomic status, racial identication, and parents educational backgrounds. These controls are essential, since most unlicensed teachers nd placements in low-performing, high-poverty urban schools. In another California-based study, Laczko-Kerr and Berliner (2003) consider undercertied teachers, including teachers with no education coursework and those with some coursework but not enough to fulll formal certication requirements. They nd that students of certied teachers outscore students of undercertied teachers on the Stanford-9 Achievement Tests. They conclude that students of certied teachers gain about two months of achievement on a grade-equivalent scale, translating into a 20 percent penalty in academic growth for every year with an undercertied teacher. Several studies evaluating the effect of teacher certication focus exclusively on mathematics classes, since this subject faces greater teacher shortages than other subjects assessed by standardized tests. Goldhaber and Brewer (2000) nd that having a certied math teacher results in at

Do Mentoring and Induction Programs Have Greater Benets for Teachers Who Lack Preservice Training?

65

least a 1.3 point increase on the state assessment test. Monk (1994) nds that additional education courses on teaching mathematics have a greater positive effect on student achievement than do additional college mathematics courses. He concludes that subject area competencethe chief qualication of many alternative route teachersis a necessary but not a sufcient condition for effective teaching (Monk 1994, 142). In addition to actual teaching practice, traditional certication programs provide teachers with a sense of condence in their abilities, which translates into greater teacher efcacy (Laczko-Kerr and Berliner 2003, 37). According to a survey of beginning teachers in New York City, more certied teachers feel adequately prepared than noncertied teachers (Darling-Hammond, Chung, and Frelow 2002). Uncertied teachers show a weaker sense of responsibility for student learning, as they are more likely to blame poor student performance on the students and their home environments. Teachers sense of preparedness is signicantly correlated to their perceived teaching efcacy. Teachers who feel poorly prepared are also more likely to teach only until something better comes along (Darling-Hammond, Chung, and Frelow 2002).

Importance of Induction
Alternate routes of teacher preparation assume that school staffs will support unprepared teachers as they begin their service. Unfortunately, limited resources on individual campuses often yield insufcient support for new teachers who emerge from lateral entry programs (Laczko-Kerr and Berliner 2003, 37). Because they work autonomously in individual classrooms, new teachers lack easy access to more experienced educators. To better address the concerns of and difculties faced by new teachers, schools can explicitly endorse induction programs to build a professional culture of collaboration and problem solving (Feiman-Nemser 2003, 25). Robinson (1998) used a meta-analysis of various induction models to identify the key components of an effective induction program. New teachers should have mentors within their teaching elds. Mentor-novice interactions should encompass teaching eld materials and techniques as well as school-specic policies. Schools should assign novices limited teaching responsibilities so that they have extra time to prepare curriculum and observe their mentors and other experienced teachers in their classrooms. The key political justication for investing in induction programs is to prevent teacher attrition. Schools require adequate stafng with qualied teachers to maintain and improve student achievement. Additionally, at-

66

Laura Duke, Adam Karson, Justin Wheeler

trition imposes signicant costs on the education system as it must prepare more teachers to ll classroom vacancies. Currently, after the rst year of teaching, 15 percent of new teachers move to another school, and 14 percent leave teaching altogether (Smith and Ingersoll 2004, 694). Attrition rates are even greater at high-poverty schools, where teachers must contend with fewer resources, poorer working conditions, and needier students (Darling-Hammond 2003, 7). Several studies nd that attrition rates are higher for teachers with less formal preservice preparation. Darling-Hammond (2003) reviews studies showing higher than average attrition for alternative route and uncertied teachers in California, Massachusetts, and Texas. According to the National Center for Education Statistics, new teachers whose preservice training included student teaching had a 15 percent attrition rate over ve years, compared to a 29 percent rate among those who lacked student teaching experience (Heinke, Chen, and Geis 2000, 49). Growing evidence indicates that comprehensive induction programs can reduce attrition rates among new teachers. Induction can incorporate a variety of supports for new teachers, from a single orientation meeting at the beginning of a school year to a highly structured program involving multiple activities and frequent meetings over a period of several years (Smith and Ingersoll 2004, 683). Beyond nding that induction programs in general reduce attrition, Smith and Ingersoll (2004) measure how various induction activities individually affect new teacher retention. When considering these activities, Smith and Ingersoll nd that assigning new teachers mentors from the same teaching eld, scheduling new teachers extra time for collaboration or planning, and reducing new teachers teaching schedule all signicantly reduce the relative risk that new teachers would attrite by more than half. Rather than looking at induction activities individually, Moir (2003) considers all elements in a comprehensive induction program as a single induction variable. Moir tracks participants in the Santa Cruz New Teacher Project (SCNTP), a comprehensive induction curriculum with a core of structured mentoring created at the University of California at Santa Cruz in 1988. After seven years, 88 percent of California teachers who had participated in SCNTP remained in teaching, a higher than expected retention rate. Applying the same induction model, schools in CharlotteMecklenburg, North Carolina reduced rst year attrition rates from 32 percent for non-participants to 17.5 percent for teachers at participating schools (Moir 2003, 11).

Do Mentoring and Induction Programs Have Greater Benets for Teachers Who Lack Preservice Training?

67

DATA
To test our hypotheses, we use data from the school and teacher portions of the 1999-2000 Schools and Stafng Survey (SASS) administered by the National Center for Education Statistics (NCES). The SASS is designed to facilitate research on teacher demand and shortage, teacher and administrator characteristics, school programs, and general school conditions. SASS also collects data on many other topics, including principals and teachers perceptions of school climate and problems in their schools, teacher compensation, district hiring practices and basic characteristics of the student population. We merged variables from the school and teacher surveys, resulting in a sample of 42,549 teachers. Because our data is cross-sectional, we are unable to measure teacher attrition directly. In order to gauge the likelihood of a teacher leaving the profession due to a lack of success in the classroom, we used the measure of teacher intent to remain in the profession as a proxy for likely attrition.2 Teachers who indicated that they intended to stay in the professional for as long as possible or until retirement were coded as committed to the profession; those who indicated that they would teach unless something better comes along, who intended to leave teaching as soon as possible, or who were undecided were coded as not committed to teaching. Our rst independent variables of interest are a set of indicator variables for a teachers participation in activities or receipt of supports typical of induction programs. Our second covariate of interest is a binary variable indicating whether the teachers undergraduate degree is in education or some other discipline. Beyond the teachers preparedness for the classroom and the presence of professional supports in the school, a number of factors might inuence a teachers decision to leave the profession. Among these are the availability of other jobs, the remuneration the teacher receives for her work, working conditions within the teachers school, and the degree of challenge presented in educating the schools students. We capture these factors imperfectly through control variables for characteristics of teachers and the schools in which they teach. Since teachers in urban schools are exposed to larger job markets, we control for the urbanicity of the area in which a school is located. High school teachers and teachers of science and math are likely to leave the profession at higher rates than other teachers, as are teachers who work in schools that serve large numbers of at-risk students. We use binary variables to control for a teacher teaching high school and teaching science or math and include measures of the percentage of limited-English-

68

Laura Duke, Adam Karson, Justin Wheeler

procient students, special education students, and students who qualify for free or reduced-price lunch at the teachers school. We are also able to control for the extreme circumstance of a teachers being threatened by a student. Other covariates include an indicator variable for large schools and schools receiving Title I funding and controls for teacher age, gender, race and earnings.

METHODS
We use logistic regressions to model the effect of induction and its different components on teacher commitment. Our model for this regression is as follows: teacher commitment = + 1education degree + 2induction program or practice + x + (model 1) In a second model we test the hypothesis that the positive effect of induction may be greater on teachers without bachelors degrees in education by interacting the indicator variable for a bachelors degree in education with the indicator variable for induction. teacher commitment = + 1education degree + 2induction program or practice + 3educ.degree*induction + x + (model 2) Part II of our analysis explores how these different groups of teachers react to three types of mentoring and induction. Following the logic of Smith and Ingersoll, we hypothesize that the increased intensity of a mentoring and induction program should have a greater positive effect on non-education degree teachers commitment to their job. Intensity, in the context of this study, is synonymous to the completeness or comprehensiveness of the induction program. Considering the limited resources of schools, this analysis is useful for identifying the most cost-effective methods of induction. Using our original variable for teacher commitment, we test three clusters of mentoring and induction programs: (1) the existence of a mentor and whether that mentor teaches in the same eld, (2) group induction activities, and (3) extra resources provided. As in Part I, we use data from the 1999-2000 SASS. Each cluster contains several variables, and we interact each of these variables with the education degree indicator to isolate the effects of mentoring and induction for teachers with and without bachelors degrees in education. This method results in nine separate models that have the following general equation:

Do Mentoring and Induction Programs Have Greater Benets for Teachers Who Lack Preservice Training?

69

teacher commitment = + 1(education degree) + 2(induction program type) + 3(interaction between induction variable and ed degree) + x + (models 3-11) All nine models in Part II control for the same teacher and school characteristics as in Part I, so that we are explaining only the effect of a particular cluster of mentoring and induction programs on teacher commitment. To account for the differential probabilities of sampling and the multi-stage design of the SASS, we weight observations in all models by their probability of selection and adjust standard errors for clustering at the school level.

RESULTS PART I
Our analysis nds that teachers who do not have bachelors degrees in education differ systematically from those who do across a number of variables (see Table 1). Most importantly for the purposes of our hypotheses, we nd that teachers with education degrees are 5.1 percent more likely to express a commitment to the profession than teachers with non-education degrees. Several other trends in the data are worth noting. As we would expect, the variable across which the two groups of teachers most differ is the type of school in which they teach: Fully 68 percent of teachers without education degrees are employed in high schools, whereas only about 45 percent of those with education degrees are. Teachers without education degrees are also more likely to work in large schools and urban or suburban schools and less likely to be female. Several studies have shown that teachers in these types of environments tend to earn higher salaries. For example, in New York pupil population density and district enrollment were found to have positive and signicant effects on teacher earnings. When incorporating these ndings into predictive teacher wage indices, the same study found that it costs urban districts between 33 percent and 83 percent more than rural districts to attract teachers of similar qualications (Duncombe 2002, 32). A comparable study in Oregon found that it costs urban districts between 17 percent and 29 percent more to attract similar teachers (Karson forthcoming). Consistent with these ndings, teachers without education degrees have higher earnings on average as well. Additionally, though both sets of teachers work in schools with similar percentages of limited-English-procient students and students who qualify for federal special education protections and/or modications, teachers without education degrees are signicantly less likely to work in schools that serve high percentages of students who receive free or reduced-price

70

Laura Duke, Adam Karson, Justin Wheeler

lunch. It is also interesting to note a trend in the relative ages of the two groups of teachers: Among teachers with bachelor of arts in education, 34 percent are between the ages of forty and forty-nine, as compared to only 29 percent of teachers without bachelor of arts in education. We posit that this difference may be due to the galvanizing effect of the 1981 publication of A Nation at Risk, which emphasized the discipline-specic training of teachers.3 With the exception of their higher average earnings and reduced likelihood of teaching in schools serving poorer students, these trends would lead us to expect higher attrition among teachers without education degrees for reasons unrelated to their preparation for teaching. As noted above, we expect that younger teachers in more urban areas are more likely to leave the profession due to exposure to a large job market. Some evidence also suggests that men are more responsive than women are to the salary difference between teaching and other professions and that men leave teaching at higher rates than women in general (Tabs 2004, 9). Taken together, these trends suggest the importance of controlling for these relevant characteristics in determining the differential effect of induction on teachers without education degrees and those with education degrees. The results of our logistic regressions are presented in Table 2 below. In all models, we nd that salary earnings above $50,000, suburban or rural school location (as opposed to an urban location), and a free or reduced-lunch population of 5 to 19 percent (as compared to free or reduced-lunch population of less than 1 percent) predict a statistically signicant increase in the probability of a teacher being committed to the profession. Conversely, we nd that any percentage of special education students greater than 5 percent is associated with a statistically signicant decrease in teacher commitment, as is, not surprisingly, a teacher having been threatened by a student. The rst two columns of Table 2 contain the results of the regression of teacher commitment on a bachelors degree in education, mentoring, and the interaction of the two. The mentoring variable coefcients in the second row of the rst and second columns indicate that the effect of a mentor in the rst year of teaching is positive and marginally statistically signicant to statistically signicant. The interpretation of the coefcient on mentoring in the rst column is that having a mentor in the rst year of teaching is associated with a 16 percent greater likelihood of being committed to the teaching profession. The latter two columns contain the results of the same models with induction replacing mentoring. The coefcient on the induction variable

Do Mentoring and Induction Programs Have Greater Benets for Teachers Who Lack Preservice Training?

71

in the third row of the third column indicates that the effect of induction in the rst year of teaching is positive and statistically signicant at the 0.05 level. The interpretation of this coefcient is that participating in an induction program in the rst year of teaching is associated with a 20 percent greater likelihood of being committed to the teaching profession. As the coefcient on induction in the fourth column shows, the effect of an induction program is even greater once the interaction term has been introduced into the regression. Interpreting the interaction terms in these models requires some care. In nonlinear models the interpretation of coefcients on interaction terms differs from that of the coefcients on other terms. Rather than an odds ratio, the interaction term is the ratio of odds ratios. The statistically signicant coefcient of 0.67 on the interaction of the induction and bachelors education degree variables can be interpreted in two ways: rst, that the positive effect of induction is greater on teachers without education degrees than on those with education degrees; second, that the positive effect of a bachelors degree in education is lower on teachers who have been through induction programs than on those who have not. The temporal precedence of the bachelors degree recommends the rst interpretation. The following gure presents a method for quantifying the interaction effect. It displays the predicted probability of a teacher being committed to the profession for each combination of induction and bachelors degree, setting the cumulative effect of the control variables at its mean (0.38), and shows how an interaction effect in percentage points is derived: Figure 1: Bachelors Degree in Education
Induction Program No Yes No 0.60 (A) 0.69 (C) Yes 0.67 (B) 0.67 (D) (B-A) = (D-C) = 0.07 -0.02

Interaction effect = (D-C) - (B-A) = -0.09

The gure indicates that the probability of a teacher without a bachelors degree in education being committed to the profession increases by 9 percentage points if that teacher went through an induction program. It also provides evidence for the validity of the analysis in that it suggests that the interaction effect is entirely due to the positive effect of induction on teachers without education degrees and not to a negative effect of induction on the commitment of teachers with education degrees.

72

Laura Duke, Adam Karson, Justin Wheeler

RESULTS PART II
The results of the logistic regressions testing our third hypothesis are presented in Table 3. Each column shows the separate results for Models 3 through 11. The interpretation of the coefcients is the same as in Part I; coefcients above 1 show a greater positive effect for teachers with education degrees, and vice versa. Overall, Part II of our analysis does not support our third hypothesis. Of our 9 models, only Models 5 and 7 show signicant results. Moreover, Model 5 shows greater positive effects for teachers with education degrees and Model 7 shows greater positive effects for teachers without education degrees. The inconsistency of our results and lack of signicance on seven out of nine models prohibit us from drawing strong conclusions about the individual effects of various types of mentoring and induction programs. The coefcient for the interaction term between supportive communication and the education degree indicator variable (Int: supportive * Ed degree) found in the column for Model 5 is 1.72, and is statistically signicant at the 0.05 level. The interpretation of this coefcient is that a teacher with an education degree who gets supportive communication from her principal is more likely to be committed to the teaching profession compared to a teacher without an education degree who also gets supportive communication from his principal. The coefcient for the interaction term between collaborative planning time and the education degree indicator variable (Int: collaborative * Ed degree) found in the column for Model 7 is 0.79, and is statistically signicant at the 0.05 level. The interpretation of this coefcient is that a teacher without an education degree who has time for collaborative planning is more likely to be committed to the teaching profession compared to a teacher with an education degree who does get time for collaborative planning. There are, however, two ways to interpret the results for these interaction terms. Model 5 may also be interpreted as indicating that the positive effect of an education degree is greater for teachers who receive supportive communication than those who lack supportive communication. Similarly, Model 7 may be interpreted to mean that the negative effects of a non-education degree are reduced when teachers are given time for collaborative planning when compared to non-education degree teachers who are not given time. As in the case of our results for Part I, since the decision of college major precedes any teacher induction program, these interpretations are not useful for our analysis.

Do Mentoring and Induction Programs Have Greater Benets for Teachers Who Lack Preservice Training?

73

LIMITATIONS
Our use of the public-use version of this data entails a few important limitations. First, the surveys complex stratied design necessitates the use of either replicate weights or Taylor Series approximations for calculating standard errors accurately. Unfortunately, the public-use version of SASS omits the variables required for the latter method in order to protect the privacy of specic schools and individuals. As a result, though we weight observations according to their probability of selection and adjust standard errors for clustering at the school level, the estimated standard errors are likely to be too small, resulting in an increased probability of Type I errors. Another concern relates to use of the bachelors degree in education as a measure of comprehensive preservice training in teaching. We work from the supposition that a bachelors degree in education provides new teachers with more preservice hours of instruction in curriculum development, classroom management, and instructional methods than they would have received otherwise. However, new teachers increasingly become certied through minors in education, credentialing programs offered through universities or one-year master of arts in teaching degrees. Each of these means of certication may provide preservice instruction comparable to an education bachelors degree, drawing into question our use of the degree as a proxy for the extensiveness of preservice training. The effect of including these teachers with those who do not have bachelors degrees in education is likely to bias downward our estimates of the benecial effects of induction programs on teachers who enter the classroom without training. Thus, this problem in measurement attenuates rather than invalidates our ndings. In addition to these limitations in our data and model, there are alternate hypotheses that threaten the validity of our results. Teachers without degrees in education may be more likely to leave the profession because they have skills marketable outside of education rather than because they are not prepared for the classroom. We expect that mentoring and induction reduce attrition rates because they provide the necessary training. An alternate hypothesis consistent with our results is that mentoring and induction provide a socializing function that reduces the likelihood of teachers being drawn to other professions. If it is the greater availability of career options that drives the attrition of teachers without education degrees, then we would expect there to be a stronger relationship between dissatisfaction and intent to leave the profession among these teachers. To test the validity of this hypothesis, we

74

Laura Duke, Adam Karson, Justin Wheeler

performed a logistic regression in which an indicator variable for teacher satisfaction is included as a covariate in Model 1 and interacted with the education degree indicator. The results do not provide grounds to reject the hypothesis that satisfaction in teaching has no more of an effect on the commitment of teachers without education degrees than on those with education degrees. Nonetheless, the p-value (0.20) and magnitude of the interaction term (0.68) suggest the possibility that dissatised teachers with bachelors degrees in disciplines other than education may leave the profession more quickly than equally dissatised teachers with bachelors degrees in education.

POLICY RECOMMENDATIONS
Our results suggest that a comprehensive induction program has strong positive effects on teacher commitment, and that this effect is greater for teachers without education degrees than for those with education degrees. Consequently, despite the limitations to our analysis, we recommend that school districts implement the following policies to reduce teacher attrition: (1) Increase access to mentoring and induction programs for all beginning teachers. Our results concur with the body of research that nds mentorship and induction programs do succeed at preventing new teachers from leaving the classroom. Since induction programs have a positive effect on all teachers, they should be broadly implemented rather than targeted only toward those teachers who lack prior educational coursework and student-teaching experience. Widespread implementation of these programs may prove prohibitively expensive to some districts in terms of scheduling release time for teachers. However, we believe that the longterm benets to student achievement brought on through retention of more experienced teachers justify any short-term costs. (2) Allow campus-level discretion in tailoring induction and mentoring to meet local needs. Models 5 and 7 indicate that teachers with education degrees respond more favorably to supportive communication from their principals, while teachers without education degrees benet most from collaborative planning time. These results suggest that principals faced with budget constraints may still reap benets by encouraging those forms of mentorship and induction programs that operate within school culture. School administrators could use campus-level discretion to adapt individual induction programs to teaching staffs with differing levels of preservice training. As further research on mentoring and induction yields additional evidence of the effectiveness of particular practices, school leaders can adapt their programs accordingly.

Do Mentoring and Induction Programs Have Greater Benets for Teachers Who Lack Preservice Training?

75

Table 1: Characteristics of Teachers with and without Bachelors Degrees and of the Schools in Which They Work Bachelors in Non-education Difference education bachelors eld Teacher Characteristics % % % Committed to teaching 74.0 68.9 5.1 High school teacher 44.5 68.3 -23.8 Female 70.1 61.5 8.6 Non-white 14.8 16.8 -2.0 Mentoring Programs Mentor in 1st year 58.1 55.6 2.5 Mentor from same eld 74.8 73.0 1.8 Induction Programs Induction program in 1st year 54.4 56.0 -1.6 Supportive communication 75.8 74.1 1.7 Teacher network 25.1 27.0 -1.9 Collaboration or planning time 37.6 35.4 2.2 Beginners seminars 53.2 56.2 -3.0 Teachers aide 25.4 23.4 2.0 Reduced schedule 6.3 6.7 -0.4 Reduced preparations 8.3 10.4 -2.1 Teacher age Less than 30 15.4 16.5 -1.1 30 to 39 21.5 23.2 -1.7 40 to 49 34.4 29.2 5.2 50 or older 28.7 31.1 -2.4 Base teaching salary Less than $25,001 15.4 12.8 2.6 $25,001 to $30,000 19.9 18.4 1.5 $30,001 to $35,000 18.5 19.7 -1.2 $35,001 to $45,000 27.1 26.8 0.3 $45,001 or more 19.1 22.4 -3.3 School Characteristics Title I School 46.0 33.6 12.4 Enrollment under 500 44.4 29.7 14.7 Number of LEP students Less than 1% 21.6 22.1 -0.5 1 to 4% 44.1 43.4 0.7 5 to 19% 22.1 21.0 1.1

76

Laura Duke, Adam Karson, Justin Wheeler

20% or more 12.2 Number of IEP students Less than 5% 6.6 5 to 9% 30.7 10 to 14% 35.6 15 to 19% 16.0 20% or more 11.1 Number of free- or reduced-lunch students Less than 5% 6.0 5 to 19% 25.1 20 to 49% 38.9 50% or more 30.0 Urbanicity of school Large or mid-size central city 21.0 Urban fringe of large or mid-size city 37.4 Small town/Rural 41.5 Source: School and Stafng Survey, 1999-2000

13.6 6.9 32.0 36.3 14.6 10.2 9.4 30.7 34.9 25.0 25.4 43.2 31.4

-1.4 -0.3 -1.3 -0.7 1.4 0.9 -3.4 -5.6 4.0 5.0 -4.4 -5.8 10.1

Table 2. Logistic Regression: Mentoring and Induction as Predictors of Teacher Commitment to Stay in the Profession, Interacted with Bachelors Degree in Education Variable Education degree Mentor Int: mentor * Ed degree Induction Int: induction * Ed degree School earningsa 30 to 39K 40 to 49K Effect of Mentoring Model 1 Model 2 1.05 1.26 -0.097 -0.18 1.16 1.37 (0.10)* (0.17)** 0.74 -0.13 1.16 -0.15 0.96 -0.14 1.15 -0.15 0.96 -0.14 Effect of Induction Model 1 Model 2 1.05 1.34 -0.097 (0.20)* 1.20 1.48 (0.10)** (0.20)*** 0.67 (0.12)** 1.15 -0.14 0.95 -0.14 1.15 -0.14 0.95 -0.14

Do Mentoring and Induction Programs Have Greater Benets for Teachers Who Lack Preservice Training?

77

Greater than 50K Teacher age in yearsb 30 to 39

1.51 (0.26)**

1.50 (0.26)**

1.51 1.52 (0.26)** (0.26)**

0.96 0.95 0.98 0.97 -0.12 -0.12 -0.12 -0.12 40 to 49 1.08 1.08 1.09 1.10 -0.14 -0.14 -0.14 -0.14 50 or older 0.95 0.95 0.95 0.95 -0.12 -0.12 -0.12 -0.12 Math or science teacherc 0.89 0.89 0.90 0.91 -0.11 -0.11 -0.11 -0.11 Special education teacherc 0.98 0.97 0.98 0.97 -0.14 -0.14 -0.14 -0.14 ESL teacherc 1.01 0.99 1.00 1.01 -0.26 -0.25 -0.25 -0.25 Male teacherc 1.04 1.04 1.03 1.03 -0.10 -0.10 -0.10 -0.10 c Non-white teacher 1.04 1.03 1.03 1.03 -0.13 -0.13 -0.13 -0.13 Threatened by a student 0.64 0.64 0.64 0.64 (0.08)*** (0.07)*** (0.07)*** (0.07)*** 1.13 1.13 1.14 1.14 Students receive Title I services -0.12 -0.12 -0.12 -0.12 d Percentage of students limited-English-procient 1 to 4% 1.21 1.20 1.20 1.20 -0.14 -0.14 -0.14 -0.14 5 to 19% 0.98 0.98 0.98 0.98 -0.14 -0.14 -0.14 -0.14 20% or more 1.14 1.14 1.13 1.13 -0.20 -0.20 -0.20 -0.20 Percentage of students with Individual Development Planse 5 to 9% 0.62 0.62 0.61 0.60 (0.12)** (0.12)** (0.12)** (0.12)** 10 to 14% 0.64 0.65 0.64 0.64 (0.13)** (0.13)** (0.12)** (0.12)** 15 to 19% 0.54 0.55 0.53 0.53 (0.11)*** (0.11)*** (0.11)*** (0.11)***

78

Laura Duke, Adam Karson, Justin Wheeler

Table 2.(continued) Logistic Regression: Mentoring and Induction as Predictors of Teacher Commitment to Stay in the Profession, Interacted with Bachelors Degree in Education

Variable 20% or more

Effect of Mentoring Effect of Induction Model 1 Model 2 Model 1 Model 2 0.57 0.57 0.57 0.57 (0.13)** (0.13)** (0.13)** (0.13)** Percentage of students eligible for free- or reduced-price lunche 5 to 19% 1.60 1.59 1.62 1.64 (0.35)** (0.35)** (0.36)** (0.37)** 20 to 49% 1.28 1.27 1.30 1.30 (0.29) (0.28) (0.30) (0.30) 50% or more 1.45 1.45 1.48 1.49 (0.37) (0.37) (0.38) (0.39) Urbanicityf Suburban 1.34 1.34 1.35 1.36 (0.16)** (0.16)** (0.16)** (0.16)*** Rural 1.59 1.58 1.61 1.61 (0.21)*** (0.21)*** (0.21)*** (0.21)*** Enrollment>500 0.92 0.92 0.92 0.92 (0.10) (0.10) (0.10) (0.10) High school 0.91 0.91 0.91 0.92 (0.09) (0.09) (0.09) (0.09) Observations Pseudo-R2 4952 0.025 4952 0.026 4952 0.026 4952 0.027

Coefcient estimates in bold. Robust standard errors in parentheses, adjusted for clustering on school identifying variable. * signicant at 10% level; ** signicant at 5% level; *** signicant at 1% level a: Omitted category is less than $30K. b: Omitted category is under 30. c: Omitted category is all other teachers. d: Omitted category is under 1%. e: Omitted category is under 5%. f: Omitted category is central city/urban.

Do Mentoring and Induction Programs Have Greater Benets for Teachers Who Lack Preservice Training?

79

Table 3. Logistic Regression: Mentoring and Induction as Predictors of Teacher Commitment to Stay in the Profession, Interacted with Bachelors Degree in Education
These models include the same teacher and school control variables as in Models 1 and 2.

Model 3 Bachelors Degree Type Ed degree or other degree (0.14)** Mentoring Program Mentor in 1st year? Int: Mentor * Ed degree Mentor from same eld Int: Mentor from same eld * Ed degree Group Induction Program Supportive communication Int: supportive * Ed degree Teacher network Int: network * Ed degree Collaborative planning time Int: collaboration * Ed degree Beginners seminars Int: seminar * Ed degree Extra Resources Provided Teachers aide Int: aide * Ed degree Reduced Preparations Int: red. prep. * Ed degree Reduced schedule Int: red. Sch. * Ed degree Observations Pseudo-R2 4952 0.263 1.26 (0.18)

Model 4 1.21 (0.32)

Model 5

Model Model 6 7

Model Model Model Model 8 9 10 11

0.87 1.39 1.45 1.46 1.27 1.33 1.33 (0.16) (0.08)** (0.09)** (0.23)* (0.11)**(0.11)**

1.37 (0.17)* 0.74 (0.13) 0.97 (0.22) 1.10 (0.32) 0.82 (0.14) 1.72 (0.38)* 1.19 (0.11) 0.88 (0.10) 1.14 (0.10) 0.79 (0.85)* 1.20 (0.18) 0.85 (0.16) 1.11 (0.16) 1.16 (0.21) 1.31 (0.25) 0.86 (0.21) 1.23 (0.37) 0.78 (0.30) 2924 .0203 4952 0.023 20858 20858 0.013 0.013 4952 8160 8160 4952 0.0207 0.014 0.013 0.020

Robust standard errors in parentheses, adjusted for clustering on school identifying variable. * signicant at 10% level; ** signicant at 5% level; *** signicant at 1% level

80

Laura Duke, Adam Karson, Justin Wheeler

NOTES
1

The authors wish to thank Dr. Christina Gibson-Davis and Dr. Helen Ladd for their assistance in the development of this paper. 2 The SASS teacher survey includes four questions that might be used to gauge the likelihood of a teachers leaving the profession due to a lack of success in the classroom: the rst assesses the teachers satisfaction at the school; the second assesses the teachers sense of efcacy in her job; the third asks whether the teacher would become a teacher if she had the choice to make over again; and the last asks the teacher how long she intends to remain in teaching. Teacher responses to these items, which vary in the number of acceptable answers they allow, are not highly correlated. 3 A Nation at Risk would have rst shown an effect on future teachers choices of major in the mid-80s. A hypothetical 18-year-old who chose a discipline-specic major in 1981 would have been 37 in 2000. Teachers age 40 to 49 are thus too old to have been affected. Those over 50 may have attended college before schools of education became large. 4 An additional consideration in interpreting interactions in nonlinear models is that both the magnitude and the signicance of the interaction term can vary with the predicted probability of the model (in other words, depending on the effect of the covariates) and with the method used to calculate the coefcient (See Norton et al. 2004). Not surprisingly, as the predicted likelihood of a teachers being committed to the profession rises above 0.7, the differential effect of induction on teachers without education degrees and teachers with education degrees falls.

REFERENCES
Darling-Hammond, Linda. 2003. Keeping Good Teachers. Educational Leadership 60: 8. May 1. Darling-Hammond, Linda, Ruth Chung, and Fred Frelow. 2002. Variation in Teacher Preparation: How Well Do Different Pathways Prepare Teachers to Teach? Journal of Teacher Education 53: 4. September/October. Darling-Hammond, Linda and Peter Youngs. 2002. Dening Highly Qualied Teachers: What Does Scientically-Based Research Actually Tell Us? Educational Researcher 31: 9. December. Duncombe, William. 2002. Estimating the Cost of an Adequate Education in New York. Center for Policy Research Working Paper No. 44. February. Feiman-Nemser, Sharon. 2003. What New Teachers Need to Learn. Educational Leadership 60: 8. May 1.

Do Mentoring and Induction Programs Have Greater Benets for Teachers Who Lack Preservice Training?

81

Fetler, Mark. 1999. High School Staff Characteristics and Mathematics Test Results. Education Policy Analysis Archives 7: 9. March 26. Gitomer, Drew H., Andrew S. Latham, and Robert Ziomek. 1999. The Academic Quality of Prospective Teachers: The Impact of Admissions and Licensure Testing. Educational Testing Service Teaching and Learning Research Report Series. Goe, Laura. 2002. Legislating Equity: The Distribution of Emergency Permit Teachers in California. Education Policy Analysis Archives. 10: 42. October 14. Goldhaber, Dan and Emily Anthony. 2003. Teacher Quality and Student Achievement. ERIC Clearinghouse on Urban Education Urban Diversity Series: No. 115. May. Goldhaber, Dan and Dominic J. Brewer. 2000. Does Teacher Certication Matter? High School Teacher Certication Status and Student Achievement. Educational Evaluation and Policy Analysis 22: 2. Summer. Henke, Robin R., Xianglei Chen, and Sonya Geis. 2000. Progress through the Teacher Pipeline: 1992-93 College Graduates and Elementary/Secondary School Teaching as of 1997. National Center for Education Statistics Statistical Analysis Report. January. Ingersoll, Richard M. 2003. Is There Really A Teacher Shortage? Center for the Study of Teaching and Policy. University of Washington. September, 2003. Jorissen, Kathleen Topolka. 2002. Why They Stay: Alternate Route Teachers Perspectives on Their Decisions to Continue Teaching after the Induction Period. American Educational Research Association. April 2. Karson, Adam. Forthcoming. Estimating the Cost of an Adequate Education in Oregon. Terry Sanford Institute of Public Policy Masters Thesis. Laczko-Kerr, Ildiko and David C. Berliner. 2003. In Harms Way: How Undercertied Teachers Hurt Their Students. Educational Leadership 60: 8. May 1. Moir, Ellen. 2003. Launching the Next Generation of Teachers through Quality Induction. National Commission on Teaching and Americas Future. July. Monk, David H. 1994. Subject Area Preparation of Secondary Mathematics and Science Teachers and Student Achievement. Economics of Education Review 13: 2. June. Norton, Edward, Hua Wang and Chunrong Ai. 2004. Computing Interaction Effects and Standard Errors in Logit and Probit Models. The Stata Journal 4: 2. Robinson, Gary W. 1998. New Teacher Induction: A Study of Selected New Teacher Induction Models and Common Practices. Midwestern Educational Research Association. October. Smith, Thomas M. and Richard M. Ingersoll. 2004. What are the Effects of Induction and Mentoring on Beginning Teacher Turnover? American Education Research Journal 41: 3. Fall.

82

Laura Duke, Adam Karson, Justin Wheeler

Stoddart, Trish and Robert E. Floden. 1995. Traditional and Alternate Routes to Teacher Certication: Issues, Assumptions, and Misconceptions. National Center for Research on Teacher Learning Issue Paper 95-2. February. Tabs, E.D. 2004. Teacher Attrition and Mobility: Results from the Teacher Follow-up Survey, 2000-01. National Center for Education Statistics. August. Wenglinsky, Harold. 2000. How Teaching Matters: Bringing the Classroom Back into Discussions of Teacher Quality. Educational Testing Service. October. Wilson, Suzanne M., Rober E. Floden, and Joan Ferrini-Mundy. 2002. Teacher Preparation Research: An Insiders View from the Outside. Journal of Teacher Education 53: 3.