Abstract

Policy makers and school leaders are perennially concerned with the capacity of the nation's public schools to recruit and retain highly skilled teachers. Over the past two decades, policy strategies including the federal No Child Left Behind Act and alternative pathways to teaching, as well as changes in the broader labor market, have altered the context in which academically skilled college graduates choose whether to enter teaching, and, if so, where to teach. Using data from 1993 to 2008, we find that schools nationwide are recruiting a greater share of academically skilled college graduates into teaching, and that increases in teachers’ academic skills are especially large in urban school districts that serve predominantly nonwhite students. On the other hand, the increase in the share of academically skilled teachers coincides with the lower likelihood of nonwhite teachers being employed. Once teaching, nonwhite teachers report substantially lower job satisfaction than other teachers. The issue of how to recruit and support an academically skilled and diverse teacher workforce remains pressing.

1.  Introduction

The capacity of the nation's public schools to attract highly skilled college graduates into teaching is a perennial concern of both policy makers and school leaders. Although research provides evidence that teachers’ academic qualifications are only modestly aligned with measures of their instructional effectiveness (Murnane and Steele 2007; Goldhaber 2008), the academic ability of individuals entering teaching is an indicator of the attractiveness of the profession, and may reflect the schools’ capacity to recruit talented individuals along other, unobserved dimensions. Comparative analyses of highly successful school systems such as Finland, Korea, and Japan frequently contrast those nations’ success at recruiting “top talent” into teaching with the comparatively low status of the teaching profession in the United States (Auguste, Kihn, and Miller 2010; Darling-Hammond 2010; Tucker 2011; OECD 2012). In 2009, the former U.S. Secretary of Education, Arne Duncan, called for America's colleges of education to increase the selectivity of students and improve their programs to better prepare new teachers. More recently, the U.S. Department of Education, in partnership with the nation's two largest teacher unions, Teach for America, and several other organizations, launched the TEACH campaign aimed at recruiting America's best to teach (Rich 2013). These recent policy initiatives further elevate the national discussion of recruiting “top talent” into the teaching profession.

Over the last two decades, significant federal legislation including the No Child Left Behind (NCLB) Act of 2001 contain features intended to promote the recruitment of more “high quality” teachers, such as higher minimum standards for teachers’ certification (Loeb and Miller 2006). Available evidence shows a trend from 1993 to 2007 that teachers after NCLB are working longer hours, perceive greater control in their own classrooms, and feel greater support among peers, administrators, and parents (Grissom, Nicholson-Crotty, and Harrington 2014). This change may influence new graduates’ views of the teaching profession and their job satisfaction once they enter into this profession.

Spurred by both the NCLB legislation and by the difficulty of attracting teachers in many schools and subject areas, alternate pathways into teaching also expanded rapidly. In 1995, 6,932 teachers were prepared through alternative certification programs. That number jumped to 59,000, or about one-third of newly hired teachers, in 2007 (Grossman and Loeb 2008, p. 4). Many alternate pathways, including such prominent programs as Teach for America and Troops to Teachers, focus on filling teaching assignments in hard-to-staff schools in particular. The increase in the diversity of pathways into the teaching profession might be an important source of the increase in academic talent in the profession (Johnson, Birkeland, and Peske 2005).

Alongside political efforts and evolving pathways into teaching, substantial fluctuations in the nation's economic conditions and accompanying impacts on labor markets and public financing of education likely influenced the attractiveness of the teaching profession to new college graduates. The tech boom of the early 2000s and the aftermath of the financial collapse in 2008 represent sharp contrasts in this respect. The latter corresponded to substantial reductions in state education spending in many states as well as reductions in private-sector jobs (Leachman and Mai 2013). There is some evidence of an increase in the instructional effectiveness of teachers entering the profession during the Great Recession, potentially because of these labor market conditions and the associated increase in the relative attractiveness of teacher salaries relative to salaries in alternative occupations (Nagler, Piopiunik, and West 2015).

Previous research describing the academic qualifications of incoming teachers indicates that higher-scoring college graduates had become increasingly less likely to enter teaching since the 1950s, but that this trend may have reversed in recent years. As of the 1990s, higher-scoring college graduates were significantly less likely to enter teaching, capping a decline over several previous decades as changes in the labor market opened up more alternative career opportunities for women (Corcoran, Evans, and Schwab 2004; Bacolod 2007). More recent evidence, however, indicates that the academic quality of new teachers may have rebounded since then, with increases in the aggregate SAT scores, grade point average, and college competitiveness of prospective and new teachers (Goldhaber and Walch 2013; Lankford et al. 2014).

Although the recent uptick in incoming teachers’ academic competitiveness nationwide has been highlighted as an encouraging development (Goldhaber and Walch 2013), extant research has not examined the other characteristics of teachers that are contributing to this trend, such as race/ethnicity, nor where teachers with higher academic ability are recruited to teach. In addition, it remains unclear whether improvements in schools’ ability to recruit more academically skilled teachers corresponds to more positive job experiences for new teachers after they are employed so that they would more likely stay. Both teacher job and pay satisfaction have been shown to be predictive of teacher retention decisions, yet we know little about changes in such satisfaction of new teachers (Stockard and Lehman 2004; Guarino, Santibañez, and Daley 2006).

Along with concerns about the quality of new teacher recruits in the aggregate, a substantial body of research has documented disparities in the qualifications and academic ability levels of individuals teaching in “hard-to-staff” schools that serve more poor, black, and Hispanic students, particularly in urban contexts (Lankford, Loeb, and Wyckoff 2002; Peske and Haycock 2006; Clotfelter, Ladd, and Vigdor 2007). Teachers in these schools historically have been more likely to be provisionally or emergency certified, and typically scored lower on ability tests and licensure exams. Moreover, despite research indicating that being taught by a teacher of the same race improves students’ academic outcomes (Villegas and Lucas 2004; Dee 2005), there has been a persistent and large gap between the percentage of nonwhite students and the percentage of nonwhite teachers in the U.S. education system. For example, about 32 percent of students enrolled in 1993–2004 and 41 percent of students enrolled in 2007–08 were not white, while the percentage of nonwhite teachers was 13 percent in 1993–94 and 17 percent in 2007–08 (Ingersoll and May 2011). Another challenge of building a diverse teaching workforce is that minority teachers frequently teach in more challenging environments, tend to be less satisfied with their jobs, and are correspondingly less likely to remain in teaching (Ingersoll and May 2011). In light of these challenges, improving the capacity of schools to recruit and retain an academically skilled and diverse workforce may be critical to addressing the persistent academic achievement gaps of minority students.

In this study, we build on prior research by examining the improvement in new teachers’ academic qualifications in greater detail, and by considering the work experiences of new teachers after they began teaching. Using the Baccalaureate and Beyond studies in 1993, 2000, and 2008, we describe the subgroups of incoming teachers that have seen the most change in academic background, as well as the characteristics of schools where academically skilled teachers are increasingly employed. We supplement data on the academic ability scores of new teacher hires with survey data on new teachers’ job satisfaction over time, both overall and related to specific aspects of the job. These additional data provide information regarding the relative attractiveness of teaching in comparison to other professions, over time. The observed patterns offer insight into the potential drivers of recent changes.

We specifically ask the following five research questions:

  1. How has the academic ability of recent college graduates entering teaching changed across the graduating classes of 1993, 2000, and 2008?

  2. How have new teachers’ academic ability varied over time as a function of teachers’ characteristics, including their demographics and the grades and subjects in which they teach?

  3. How has the distribution of academically skilled new teachers to schools changed over this period, particularly for schools that serve disproportionately poor and minority students?

  4. How has new teachers’ reported job satisfaction changed over time, both overall and with regard to job pay and job security in particular?

  5. How have changes in teachers’ reported job satisfaction varied by teacher characteristics and school contexts?

2.  Data Description

Our analyses rely on data from three Baccalaureate and Beyond (B&B) studies conducted by the National Center for Education Statistics. Each study includes a nationally representative survey of graduates from postsecondary institutions, and in each case we use survey data from graduates collected in the year immediately after they earned their bachelor degrees (i.e., 1994, 2001, and 2009). These survey data are linked to extensive information about graduates’ educational experience and test scores, demographic backgrounds, and post-graduation work experiences.

In the case of teachers, the B&B studies include data about the schools at which they reported working, drawn from the U.S. Department of Education's Common Core of Data (CCD). When linking teachers to schools, we identify the most recent school they taught at following graduation. We summarize our key measures below.

Teacher Characteristics and Job Experiences

The B&B studies include a range of information about recent graduates and about incoming teachers. We specifically utilize data on graduates’ gender, race, and academic ability. When comparing teachers with other graduates, we include only teachers who are employed full time in grades K–12, excluding substitute teachers and teacher aides. We derive the measure of graduates’ academic ability from individuals’ SAT and ACT test performance. Within our nationally representative sample, we estimate individuals’ relative percentile ranking on SAT and ACT scores separately, both for their combined scores and for verbal and math separately. Our academic ability measures are thus individuals’ test score percentile relative to that of their peer graduates from postsecondary institutions. We use ACT percentile ranks as a substitute measure for those individuals who did not take the SAT or for whom SAT data are unavailable.1 In order to examine teachers’ on-the-job experiences, we consider three survey questions about degree recipients’ job satisfaction that were identical in the 2000 and 2008 cohort surveys.2 These questions, which were given to all employed degree recipients, asked whether graduates were satisfied with their jobs overall, with their job pay, and with their job security in the year after graduation. As a benchmark, we compare responses from teachers with those from other professionals within each cohort.

School and Role Characteristics

In order to evaluate the distribution of teachers by academic ability level and reported job satisfaction, we include several characteristics of teachers’ roles and the schools where they taught. Our role characteristics include a measure of the highest grade level taught by the teacher as well as the subject matter taught. We use teachers’ grade level assignment to distinguish between secondary (grades 7–12) and elementary (grades K–6) school teachers, and we use subject area assignments to distinguish secondary teachers who teach science, math, and technology (i.e., STEM subjects) from secondary teachers in other subject areas.

Our available data on school characteristics include whether each teacher's school is public or private, the school's locale (urban, rural, or other), and information about the demographic composition of the students, including race/ethnicity and eligibility for lunch subsidies as a proxy for family income. Available lunch eligibility data from the 2000 cohort include the percent of students eligible for free lunch. This differs slightly from the 1993 and 2008 cohorts, where we have data on the combined percent of students eligible for free or reduced-price lunch. To standardize our measures of student poverty across cohorts, we calculate a relative proportion of low-income students by ranking schools within each cohort according to the poverty indicator available for that cohort. Finally, we create indicator measures for schools that serve predominantly nonwhite students (>66 percent), and for schools in the top third of the distribution of the proportion of students eligible for subsidized lunch.

3.  Analytic Strategies

To address our research questions, we first provide descriptive summaries of the academic ability scores and job satisfaction rates of teachers and other graduates. In each case, we include information regarding the confidence intervals around our point estimates from the nationally representative samples. We also examine patterns in the distribution of teacher academic ability scores as a function of teacher and school characteristics.

We then use multivariate regression analyses to predict (1) the academic ability score percentiles of new teachers across different teacher and school characteristics; and (2) the job satisfaction rates of new teachers across a range of teacher and school characteristics. When predicting job satisfaction, we utilize logit models to estimate the likelihood that teachers are satisfied with their jobs across different teacher and school characteristics, and report the results in the form of odds ratios. Our analyses utilize the sampling weights and the Balanced Repeated Replication weights appropriate to each of the individual B&B samples in question.

Some individual and school characteristics data related to the B&B study design and implementation are missing for individual survey respondents. Missing data occur, for instance, when a surveyed teacher did not provide sufficient information to identify the school where they taught, or when CCD datasets were missing information about a school that was successfully identified. We impute data in cases of missing values for each of our independent variables of interest using a multiple imputation strategy. We also include additional information related to the instances of missing data via two descriptive tables in Appendix A. Table A.1 details the extent of missing data for each of the variables in our analyses, and table A.2 examines whether observations with and without any missing school-level data appear significantly different in terms of observable characteristics of teachers. By and large, they do not. Nevertheless, since these data are not missing at random, it is possible that our results may be biased in some instances because of missing data. We provide additional technical details of our approach to handling missing data in Appendix B.

4.  Findings

How Has the Academic Ability of Recent College Graduates Entering Teaching Changed across the Graduating Classes of 1993, 2000, and 2008?

In line with recent research about the improving academic qualifications of incoming teachers (Goldhaber and Walch 2013), figure 1 shows a drop in teachers’ average academic ability from the 1993 to 2000 cohort. Although statistically insignificant, there is a statistically significant increase in teachers’ average academic test scores from the 2000 to 2008 cohort.3 College graduates entering the teaching profession from the 2000 cohort ranked around the 42nd percentile in SAT and ACT scores relative to their peers, on average, whereas teachers from the 2008 cohort ranked in the 48th percentile. Moreover, as shown in figure 2, although the peak of the 2008 cohort distribution was still among low performers (i.e., in the 20th percentile), there were fewer very-low-scoring college graduates and more very-high-scoring graduates entering teaching in 2008 relative to the 2000 cohort. The uptick in incoming teachers’ test scores—documented in several studies (Corcoran, Evans, and Schwab 2002; Bacolod 2007)4—represents an interruption of the downward trend in schools’ capacity to attract top performers in prior decades.

Figure 1.

SAT/ACT Percentile Ranks of Recent College Graduates Entering Teaching, Overall and by Subject Area

Notes: The overall includes the average of verbal and math. The error bars indicate the 95 percent confidence intervals. Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. Sample of 1993 cohort graduates with available by-subject data is a subset of the sample for overall scores, due to missing data on by-subject ACT in 1993. Data points in all cells represent >10 observations.

Figure 1.

SAT/ACT Percentile Ranks of Recent College Graduates Entering Teaching, Overall and by Subject Area

Notes: The overall includes the average of verbal and math. The error bars indicate the 95 percent confidence intervals. Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. Sample of 1993 cohort graduates with available by-subject data is a subset of the sample for overall scores, due to missing data on by-subject ACT in 1993. Data points in all cells represent >10 observations.

Figure 2.

Distribution and Mean of SAT/ACT Percentiles for Recent College Graduates Entering Teaching

Notes: Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. Data points used to estimate means all include >10 observations.

Figure 2.

Distribution and Mean of SAT/ACT Percentiles for Recent College Graduates Entering Teaching

Notes: Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. Data points used to estimate means all include >10 observations.

The improvement in average new teacher test scores between the 2000 and 2008 cohorts is driven primarily by schools’ ability to recruit teachers with higher math ability. As shown in figure 1, math scores reached their lowest point in 2000 before rebounding strongly in 2008. Incoming teachers’ verbal test scores increased only modestly, with no statistically significant differences across the three cohorts.5

How Have Trends in These New Teachers’ Academic Ability Varied by Teachers’ Demographics?

Trends in the academic ability of incoming teachers vary by gender. Although roughly a quarter of incoming teachers are men in each cohort, the test scores trends of men and women differ over time. Among women, ability scores were flat between 1993 and 2000 and then increased between 2000 and 2008, though the latter difference is not statistically significant. The test scores of men who enter teaching, on the other hand, have fluctuated substantially and significantly, reaching their nadir in the 2000 cohort before rebounding in 2008.

The fluctuation in test scores among men parallels the decline in average math scores in the 2000 cohort, and may be a function in part of the prevailing economic conditions that college graduates faced in the year following graduation. Low unemployment around the time of the 2000–01 tech boom (i.e., 4.4 percent in the winter following graduation, compared with 6.6 percent in the same period for the 1993 cohort, and 7.7 percent for the 2008 cohort6) corresponds to lower average math scores for all teachers, particularly, lower scores for men going into teaching.7 Scores rebounded strongly in 2008–09 with the onset of the Great Recession. This pattern suggests that male graduates and graduates with greater math ability may be more sensitive to alternative job opportunities when choosing whether or not to enter teaching. These findings align with Nagler and colleagues’ study on the effectiveness of teachers who entered the profession during recession (Nagler, Piopiunik, and West 2015). They were more effective—as captured by teacher value added to student test scores—than teachers who entered the profession during non-recessionary periods. This pattern is also consistent with the supply of workers for public sector jobs in the United States more generally during economic downturns (e.g., Krueger 1988; Borjas 2002). Moreover, Nagler, Piopiunik, and West (2015) observed that male teachers seemed to be more affected by the recession than female teachers, which might suggest that the career options of men were more strongly influenced by recessions than those of women.

Over the same time frame, we observe changes not only in average test scores but also in the proportion of black and Hispanic college graduates who enter teaching and in the test scores of this subgroup of teachers. As shown in figure 3, incoming black and Hispanic teachers’ average test scores were lowest in the 2000 cohort, with a significant difference relative to scores in the 2008 cohort. The proportion of black and Hispanic graduates entering teaching, at 21.2 percent, was higher, however, than the proportion of black and Hispanic college graduates overall, at 17.0 percent. In contrast, in the 2008 cohort the average ability test scores of incoming black and Hispanic teachers increased, but minority representation among incoming teachers, at 15.7 percent, was less than their share of the college graduate population at 18.5 percent.8 These results provide evidence that the increase in academic selectivity of new teachers coincides with the reduction in the diversity of the teaching force of the 2008 cohort.

Figure 3.

SAT Percentile Ranks of Graduates Entering Teaching, by Gender and Race

Notes: The bar chart shows the trends in teachers' SAT percentile ranks by gender and race. The averages are labeled on the top of bars and the error bars indicate the 95 percent confidence intervals. ACT scores are used in cases where SAT test score data were unavailable. The table underneath the bar chart shows the percentage of graduates working in teaching by gender and race. We also compare the ethno-racial composition of teachers to that of all professions. Data points in all cells represent >10 observations.

Figure 3.

SAT Percentile Ranks of Graduates Entering Teaching, by Gender and Race

Notes: The bar chart shows the trends in teachers' SAT percentile ranks by gender and race. The averages are labeled on the top of bars and the error bars indicate the 95 percent confidence intervals. ACT scores are used in cases where SAT test score data were unavailable. The table underneath the bar chart shows the percentage of graduates working in teaching by gender and race. We also compare the ethno-racial composition of teachers to that of all professions. Data points in all cells represent >10 observations.

How Has New Teachers’ Academic Ability Varied Over Time by What They Teach and Where They Teach?

Although incoming teachers’ academic ability in the aggregate is one measure of the health of the profession, what academically skilled graduates end up teaching and where they teach are also of significant interest to policy makers and school leaders. Elementary and secondary school teachers might differ in terms of their responsiveness to policy and economic changes as a function of the skills required to fill those different roles, particularly new hires in STEM subject areas. Moreover, the NCLB legislation has an explicit focus on the recruitment of high-quality teachers in schools serving more nonwhite and poor students. We, therefore, analyze the distribution of teacher academic ability scores across schools as a function of teachers’ grade level, subject area, and school sector. Table 1 provides these results.

Table 1.
Predicting New Teachers’ SAT Percentiles as a Function of Their School Type, Grade, and Subject Area
1993 Cohort2000 Cohort2008 Cohort
Model 1Model 2Model 1Model 2Model 1Model 2
Secondary grades 8.308** 7.680* 8.150*** 6.232* 6.378** 5.049* 
 (3.319) (3.393) (2.645) (2.796) (2.704) (2.927) 
STEM secondary 5.424 5.435 6.945* 8.998** 9.758*** 9.512*** 
 (4.288) (4.282) (3.207) (3.182) (3.399) (3.422) 
Private schools 9.348*** 7.09 7.160* −0.805 0.217 −4.585 
 (3.406) (5.427) (4.725) (6.214) (2.92) (3.594) 
Private secondary schools  4.88  16.952*  8.638 
  (9.013)  (7.001)  (5.742) 
No. of teachers 710 1,150 1,160 
1993 Cohort2000 Cohort2008 Cohort
Model 1Model 2Model 1Model 2Model 1Model 2
Secondary grades 8.308** 7.680* 8.150*** 6.232* 6.378** 5.049* 
 (3.319) (3.393) (2.645) (2.796) (2.704) (2.927) 
STEM secondary 5.424 5.435 6.945* 8.998** 9.758*** 9.512*** 
 (4.288) (4.282) (3.207) (3.182) (3.399) (3.422) 
Private schools 9.348*** 7.09 7.160* −0.805 0.217 −4.585 
 (3.406) (5.427) (4.725) (6.214) (2.92) (3.594) 
Private secondary schools  4.88  16.952*  8.638 
  (9.013)  (7.001)  (5.742) 
No. of teachers 710 1,150 1,160 

Notes: ACT scores are used in cases where SAT test score data were unavailable. Analysis excludes teachers who did not take either the SAT or ACT or for whom test score data were unavailable. SAT/ACT percentile ranks are relative to all college graduates in each cohort. Missing values of exploratory variables were imputed using a multiple imputation procedure detailed in Appendix B. STEM = science and math teachers in secondary grades 7—12. Our analyses utilize the sampling weights and the Balanced Repeated Replication weights appropriate to each of the individual B&B samples. Sample sizes rounded to the nearest 10. Standard errors are included in parentheses.

*p < 0.05; **p < 0.01; ***p < 0.001.

We find that teachers in secondary grades (7–12), and particularly STEM teachers who teach in secondary grades, have consistently higher ability scores than elementary teachers. This disparity is consistent across cohorts. For instance, in the 2008 cohort secondary teachers’ SAT percentiles were, on average, roughly 6 percentage points higher than elementary teachers when controlling for other factors. Secondary school teachers who teach STEM subjects tend to score even higher than other secondary school teachers (around 10 percentage points higher, on average, than other secondary school teachers in 2008). This differential is not surprising, given that graduates who have majored in STEM subjects typically have higher SAT scores than their non-STEM peers (Goldhaber and Walch 2013).

To further understand the differences between elementary and secondary schools, we contrast the relative ability level of teachers entering public schools with that of teachers entering private schools. The average SAT percentile rank of teachers employed by private elementary schools was around 4 percentage points higher than their public school peers in 1993, but by 2008 their scores were roughly 5 percentage points lower. Thus, whereas new teachers at private secondary schools retained a sizeable ability advantage over those entering public secondary schools, the overall public--private school ability gap had shrunk from a more than 7 percentage point gap in 1993 to virtually none as of 2008. As shown in figure 4, this change is driven primarily by a substantial reduction in the average academic ability of new teachers entering private elementary schools over time.

Figure 4.

SAT Percentile Ranks of New Teachers at Public and Private Elementary Schools, by Cohort

Notes: Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. ES = Elementary Schools.

Figure 4.

SAT Percentile Ranks of New Teachers at Public and Private Elementary Schools, by Cohort

Notes: Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. ES = Elementary Schools.

We next examine the distribution of teachers with higher academic ability as a function of the students and locales which they serve. As shown in table 2, schools that serve predominantly minority students employed teachers with test-score percentile ranks approximately 4 and 9 points lower in 1993 and 2000, respectively, when controlling for the urbanicity of the schools and whether they serve a high-poverty student population.9 As of the 2008 cohort, however, high-minority schools on average employed teachers with academic ability levels approximately 3 points higher than other schools on average. In these models, our reference group is teachers worked in settings that are nonurban, nonrural, and neither high-poverty nor high-minority in terms of the student composition.

Table 2.
Predicting New Teachers’ SAT Percentiles as a Function of Their School's Locale and Students Served
1993 Cohort2000 Cohort2008 Cohort
Model 1Model 2Model 1Model 2Model 1Model 2
Urban schools 2.313 6.300 2.210 −1.852 1.364 −6.323* 
 (3.100) (3.574) (3.385) (3.295) (2.925) (3.164) 
Rural schools −3.302 −3.010 −2.839 −2.951 −3.219 −4.132 
 (3.109) (3.089) (2.697) (2.720) (2.733) (2.653) 
 (2.963) (2.950) (5.356) (5.350) (3.798) (3.479) 
High-minority schools −3.761 1.648 −8.894*** −15.948*** 3.311 −5.075 
 (3.775) (4.685) (3.973) (5.500) (3.527) (3.591) 
High-minority urban schools  −12.38  14.138*  19.355*** 
  (7.818)  (7.151)  (5.507) 
No. teachers 710 970 1,150 
1993 Cohort2000 Cohort2008 Cohort
Model 1Model 2Model 1Model 2Model 1Model 2
Urban schools 2.313 6.300 2.210 −1.852 1.364 −6.323* 
 (3.100) (3.574) (3.385) (3.295) (2.925) (3.164) 
Rural schools −3.302 −3.010 −2.839 −2.951 −3.219 −4.132 
 (3.109) (3.089) (2.697) (2.720) (2.733) (2.653) 
 (2.963) (2.950) (5.356) (5.350) (3.798) (3.479) 
High-minority schools −3.761 1.648 −8.894*** −15.948*** 3.311 −5.075 
 (3.775) (4.685) (3.973) (5.500) (3.527) (3.591) 
High-minority urban schools  −12.38  14.138*  19.355*** 
  (7.818)  (7.151)  (5.507) 
No. teachers 710 970 1,150 

Notes: ACT scores are used in cases where SAT test score data were unavailable Analysis excludes teachers who did not take either the SAT or ACT or for whom test score data were unavailable. SAT/ACT percentile ranks are relative to all college graduates in each cohort. Missing values of exploratory variables were imputed using a multiple imputation procedure detailed in Appendix B. High-minority schools are defined as schools with >66 percent minority students. High-poverty schools are defined as schools in the top tercile in each period in terms of the proportion of students eligible for subsidized lunches. Our analyses utilize the sampling weights and the Balanced Repeated Replication weights appropriate to each of the individual B&B samples. Sample sizes rounded to the nearest 10. Standard errors are included in the parentheses.

*p < 0.05; ***p < 0.001; p < 0.1.

The recent improvement in the ability level of new teachers entering high-minority schools is driven almost entirely by a shift in the competitiveness of teachers working in high-minority urban school settings. Although high-minority, nonurban schools continue to employ teachers of lower academic ability as of 2008, high-minority urban schools in the most recent cohort employed teachers whose academic ability scores were above average for incoming teachers. Urban high-minority schools have gone from employing teachers who ranked roughly 4 percentage points lower than nonurban, low-minority schools in 1993, to employing teachers who ranked approximately 8 percentage points higher than nonurban, low-minority schools in 2008, when controlling for school poverty levels.

The improvement in teacher scores in urban high-minority schools corresponds to a striking bimodal distribution. As shown in figure 5, these schools went from employing predominantly teachers with very low academic ability in 1993 and 2000, to employing from a mix of both teachers with very low and very high academic ability in 2008. This distinctive shift to a bimodal distribution of teacher academic ability scores suggests that the increase in test scores in urban high-minority schools may reflect a targeted outreach to very high-performing college graduates in particular.10

Figure 5.

SAT Percentile Ranks of New Teachers at High-minority Urban Schools, by Cohort

Notes: Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. High-minority schools are defined as schools serving >66 percent nonwhite students. Data points used to estimate means all include >10 observations.

Figure 5.

SAT Percentile Ranks of New Teachers at High-minority Urban Schools, by Cohort

Notes: Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. High-minority schools are defined as schools serving >66 percent nonwhite students. Data points used to estimate means all include >10 observations.

Although academic ability score improvements are apparent among teachers of all racial groups in the 2008 cohort, the distribution of these improvements differs between black and Hispanic teachers, and other teachers. Figures A.1 and A.2 in the Appendix illustrate teacher score distributions in high-minority urban schools separately within the nonblack and non-Hispanic teacher sample and the black or Hispanic teacher sample, respectively. Incoming nonblack and non-Hispanic teachers in these schools include a dramatically larger proportion of very high-scoring individuals. Black and Hispanic teachers’ scores have also improved substantially, but primarily through a marked reduction in the share of teachers with very low academic ability employed.

How Has New Teachers’ Reported Job Satisfaction Changed Over Time?

Improvements in new teachers’ academic ability scores over time provide evidence that schools are increasingly successful at recruiting academically talented college graduates. Nevertheless, recruiting more academically skilled individuals into teaching may be of limited utility if teachers at these schools are unsatisfied with their jobs. In this section, we examine the overall job satisfaction, job pay satisfaction, and job security satisfaction of college graduates who enter teaching, in comparison with graduates who are employed in other fields. It is worth noting that our two time points for which satisfaction data are available, 2000 and 2008, reflect divergent economic conditions that may have contributed to changes in teachers’ job satisfaction relative to their nonteacher peers. Figure 6 illustrates the results, comparing job satisfaction rates in each period for teachers, all nonteachers, and a potentially more comparable group of nonteachers who are employed in nonprofit or local, state, or federal government industries.

Figure 6.

Percent of Recent Graduates Reporting Satisfaction with their Job Overall, Job Pay, and Job Security, by Cohort and Job Type

Notes: The bars indicate the percentages of teachers and other professionals who reported being satisfied with their job overall, job pay, and job security. The error bars indicate the 95 percent confidence intervals. Data points in all cells represent >10 observations.

Figure 6.

Percent of Recent Graduates Reporting Satisfaction with their Job Overall, Job Pay, and Job Security, by Cohort and Job Type

Notes: The bars indicate the percentages of teachers and other professionals who reported being satisfied with their job overall, job pay, and job security. The error bars indicate the 95 percent confidence intervals. Data points in all cells represent >10 observations.

Recent college graduates’ overall job satisfaction declined between the 2000 and 2008 graduating cohorts for teachers, nonteachers, and nonprofit and government workers. As of the 2008 cohort, however, teachers remained more satisfied (84.7 percent) with their jobs overall than either nonteachers overall (71.0 percent) or government and nonprofit employees (77.6 percent), and in both cases this advantage in relative job satisfaction represents a slight increase between the two periods. Teachers’ satisfaction with job pay also increased slightly (though not significantly) between these two periods, while nonteachers saw a significant decline. Thus, whereas teachers were less likely to be satisfied with their pay than nonteachers in 2000, they were more likely to be satisfied with their pay in 2008. This reversal may stem from the relative inelasticity of teacher pay scales. Public school teacher pay is likely to be slower to change in response to economic conditions, whereas nonteacher pay may have a greater reduction as a result of the onset of the recession in 2008. In keeping with these trends, nonprofit and government workers similarly appear less impacted in terms of job pay satisfaction than nonteachers overall, with only a modest decline between the two periods.

In contrast with the results for teachers’ satisfaction with pay, teachers’ satisfaction with their job security decreased more than that of nonteachers over the same periods. In 2000, incoming teachers were significantly more likely to report being satisfied with their job security (90.1 percent versus 83.3 percent, respectively), whereas in 2008 both teachers and nonteachers reported roughly the same levels of job security. The steeper decline in teacher job security may reflect the degree of uncertainty in public school budgets as a result of growing deficits around the peak of the recession or the public emphasis on teacher accountability that was taking hold at this time.11 Nonprofit and government workers experienced less of a decline in job security than either teachers or nonteachers overall.

How Has Teachers’ Reported Job Satisfaction Varied over Time for Different Types of Teachers and across Different School Contexts?

Finally, in keeping with our exploration of teacher academic ability scores, we examine whether teachers of different ability levels, demographic groups, roles, or school contexts saw differential trends in job satisfaction during this period. As shown in table 3, we find that black and Hispanic teachers were less likely to report job satisfaction overall in both the 2000 and 2008 cohorts, when controlling for teacher gender, ability levels, and school levels. The magnitude of this difference is apparent in a descriptive comparison of the two groups. Whereas the majority of black and Hispanic teachers report being satisfied with their job in each cohort (89 and 78 percent, respectively), they are less likely to be satisfied than white teachers in the same cohort (95 and 86 percent, respectively).12,13 We do not, however, observe that teachers’ academic skills predict either higher or lower job satisfaction in any period.

Table 3.
Predicting the Likelihood of a Teacher Being Satisfied with His Job Overall, Job Pay, or Job Security, as a Function of School Type, Teacher Role, and Teacher Characteristics
2000 Cohort2008 Cohort
Job OverallJob PayJob SecurityJob OverallJob PayJob Security
Secondary teachers 1.188 0.96 1.603 0.748 1.075 0.646* 
STEM secondary 0.864 0.916 0.749 1.231 0.903 1.606 
Private schools 1.541 1.113 1.198 0.390* 0.193*** 1.57 
Private Secondary 0.888 0.794 0.86 1.162 5.043*** 0.594 
Teacher SAT / ACT percentile 0.993 1.005 0.995 0.999 1.000 0.998 
Male teachers 0.598 1.343 1.123 1.242 1.374 1.403 
Black or Hispanic Teachers 0.358*** 0.803 0.715 0.539* 0.937 0.576* 
No. teachers 1,410 1,330 
2000 Cohort2008 Cohort
Job OverallJob PayJob SecurityJob OverallJob PayJob Security
Secondary teachers 1.188 0.96 1.603 0.748 1.075 0.646* 
STEM secondary 0.864 0.916 0.749 1.231 0.903 1.606 
Private schools 1.541 1.113 1.198 0.390* 0.193*** 1.57 
Private Secondary 0.888 0.794 0.86 1.162 5.043*** 0.594 
Teacher SAT / ACT percentile 0.993 1.005 0.995 0.999 1.000 0.998 
Male teachers 0.598 1.343 1.123 1.242 1.374 1.403 
Black or Hispanic Teachers 0.358*** 0.803 0.715 0.539* 0.937 0.576* 
No. teachers 1,410 1,330 

Notes: The table presents the results of logit estimations in which the dependent variable is an indicator for being satisfied (1 = satisfied, 0 = dissatisfied). The results are expressed in odds ratios. STEM = science and math teachers in secondary grades 7—12. Missing values of exploratory variables were imputed using a multiple imputation procedure detailed in Appendix B. Our analyses utilize the sampling weights and the Balanced Repeated Replication weights appropriate to each of the individual B&B samples. Sample sizes rounded to the nearest 10.

*p < 0.05; ***p < 0.001; p < 0.1.

In line with our previous finding of reduced teacher ability levels in private elementary schools, we see a marked decrease in the likelihood that private elementary teachers will be satisfied with their job overall and with job pay between the two periods (table 4). Although these teachers were similarly likely to be satisfied in the 2000 cohort, incoming private elementary teachers in the 2008 cohort were significantly less likely to be satisfied with their job pay and significantly less likely to be satisfied with their job overall compared with their peer teachers in public elementary schools.

Table 4.
Predicting the Likelihood of a Teacher Being Satisfied with His Job Overall, Job Pay, or Job Security, as a Function of School Context
2000 Cohort2008 Cohort
Job OverallJob PayJob SecurityJob OverallJob PayJob Security
Urban schools 0.846 0.769 0.792 0.552 0.608* 1.110 
Rural schools 0.689 0.875 0.652 0.680 0.984 1.472 
High-minority schools 1.308 1.034 1.160 0.519* 0.767 1.125 
High-minority urban schools 0.696 1.065 1.210 4.088* 2.278* 1.333 
Teacher characteristic controls Yes Yes Yes Yes Yes Yes 
No. teachers 1,190 1,230 
2000 Cohort2008 Cohort
Job OverallJob PayJob SecurityJob OverallJob PayJob Security
Urban schools 0.846 0.769 0.792 0.552 0.608* 1.110 
Rural schools 0.689 0.875 0.652 0.680 0.984 1.472 
High-minority schools 1.308 1.034 1.160 0.519* 0.767 1.125 
High-minority urban schools 0.696 1.065 1.210 4.088* 2.278* 1.333 
Teacher characteristic controls Yes Yes Yes Yes Yes Yes 
No. teachers 1,190 1,230 

Notes: The table presents the results of logit estimations in which the dependent variable is an indicator for being satisfied (1 = satisfied, 0 = dissatisfied). The results are expressed in odds ratios. High-minority schools are defined as schools with >66 percent minority students. Teacher characteristic controls include indicators for race and gender. Missing values of exploratory variables were imputed using a multiple imputation procedure detailed in Appendix B. Our analyses utilize the sampling weights and the Balanced Repeated Replication weights appropriate to each of the individual B&B samples. Sample sizes rounded to the nearest 10.

*p < 0.05; p < 0.1.

In a similar vein, we see a substantial difference in teacher satisfaction in high-minority schools in urban versus nonurban settings, as of the 2008 cohort. Urban high-minority school teachers appear to be trending higher in both their job and pay satisfaction. In contrast, teachers in nonurban high-minority settings report lower job satisfaction than teachers in schools that serve fewer minority students. Teachers in urban high-minority schools report overall job satisfaction with odds roughly twice that of either teachers in high-minority nonurban schools or teachers in urban schools that serve fewer minority students. These results suggest that the changes we observe in urban high-minority schools’ capacity to recruit more talented teachers correspond to other material changes in the job that have simultaneously buttressed teachers’ satisfaction with their work.

5.  Discussion and Conclusions

The ability of K–12 schools to recruit and retain an academically skilled teacher workforce is a barometer of the health of the teaching profession in the United States. In this study, we utilize cross-sectional data over a fifteen-year period to examine the career decisions and job experiences of recent college graduates entering teaching. We highlight changes over time in who chooses to teach, where they teach, and how satisfied they are once they enter the profession, and consider these results in light of factors including the prevailing economic conditions and evolving K–12 education policies. We expand upon previous research on new college graduates entering teaching by considering measures of both teacher academic quality and job satisfaction, and by identifying substantial heterogeneity in our results across teacher demographics, types of school organizations, and teaching assignments.

We observe teachers during a time of substantial increase in test-based accountability for schools, although the time points are also relatively unique economic periods of low unemployment (in 2000) and emerging recession (2008). Data from these periods can offer valuable evidence regarding employment patterns and working conditions for teachers under differing economic contexts, but may also affect our ability to infer broader trends in teacher recruitment and satisfaction due to policy changes.

Among the 1993, 2000, and 2008 cohorts, we find that incoming teachers’ test scores were highest for the graduating class of 2008, who entered teaching at a time of greater accountability as well as rapidly rising unemployment rates associated with early stages of the Great Recession, while scores were lowest for the graduating class of 2000, who entered teaching in the midst of the tech boom of the early 2000s. Consistent with prior research, the decisions of whether to enter teaching for men and individuals with higher test scores fluctuate more dramatically than do those of other college graduates (Guarino, Santibañez, and Daley 2006). The uptick in the 2008 class is in line with previous research (Goldhaber and Walch 2013; Lankford et al. 2014), and is promising, potentially reflecting a greater recruitment effort. The patterns may also reflect the role of the economy and the increasing relative appeal of teaching as other job opportunities become more difficult to find. Even if the economic changes drove the observed increases in the academic ability and relative satisfaction of new teachers, however, the results provide evidence that the substantial policy changes in accountability did not have a great enough negative effect—if their effect was negative at all—to lead to lower likelihood of attracting academically skilled college graduates to teaching.

The academic ability level of incoming teachers shifted substantially over the period of our study, yet teachers retained a relative advantage over nonteachers in terms of their job satisfaction rates. Teachers tended to be more satisfied with their jobs overall than graduates going into other careers, and this advantage persisted across both the 2000 and 2008 cohorts. These results are consistent with other research indicating robust teacher job satisfaction through the post-NCLB period (Grissom, Nicholson-Crotty, and Harrington 2014; Sun, Ye, and Saultz 2014). Moreover, whereas other recent graduates saw a decline in their job pay satisfaction between those two periods, teachers’ pay satisfaction was unchanged or even improved somewhat. The relative inelasticity of teacher pay scales may be an appealing attribute during challenging economic times. On the other hand, new teachers experienced a greater decline in their reported job security between the 2000 and 2008 cohorts than nonteachers, perhaps because of the decline in public financing of education during the recession period or an increase in accountability. Particularly given that the studied sample includes new (untenured) teachers, these teachers often feel less secure in their jobs than veteran teachers. This is due to collective bargaining agreements and the fact that when layoffs did happen as a result of the financial downturn, the “last hired, first fired” provisions made seniority the determining factor for which teachers were laid off (Goldhaber and Theobald 2013). The impact of education-specific policies may also help to explain why teacher job security dropped substantially in the early stages of the Great Recession, while nonprofit and government workers’ job security did not register the same decline.

Aggregate shifts in new teachers’ academic skills and job satisfaction rates mask substantial variation in where academically skilled new teachers are teaching and where they report being more satisfied with their jobs. We observe a large increase in the number of new teachers with very high academic ability teaching in urban schools that serve predominantly nonwhite students and have historically been among the hardest to staff (Lankford, Loeb, and Wyckoff 2002). The increasing competitiveness of teacher hiring in high-minority urban schools may be the result of intentional efforts by districts and educational organizations in the post-NCLB period. One example is the growth of alternative pathways, such as Teach for America, that recruited an increasing number of high-ability teachers into hard-to-staff districts during this period.14 Another example is The New Teacher Project that became active in partnership with large urban school districts to improve their hiring and recruiting process by filling vacancies earlier and working with district human resources staff to shape district policies. In addition, the observed patterns may have resulted in part from the expansion of charter schools, which are most prevalent in urban settings and which may be more likely than regular public schools to recruit recent college graduates into teaching (Epple, Romano, and Zimmer 2015).

The study results highlight the uneven distribution of teacher academic quality. Whereas urban high-minority schools’ incoming teachers are increasingly more academically skilled, we observe persistent and substantial disparities in new teachers’ academic ability and job satisfaction in nonurban high-minority schools. In addition, even as the ability scores of new teachers have improved, the diversity of the incoming teaching workforce, and in particular the hiring of black and Hispanic college graduates into teaching, has dropped off as of 2008. Moreover, once hired, minority teachers report substantially lower job satisfaction rates than other teachers, a trend that has been consistent over time. Preparing and supporting an academically skilled and diverse teacher workforce as a whole remains as a pressing issue.

This study is descriptive in nature. We do not explicitly evaluate potential drivers of change in teachers’ academic ability levels or job satisfaction and, thus, can only speculate as to the underlying causes of the trends that we observe in incoming teachers. In addition, the specific cohorts that we examine correspond to varying economic and labor market circumstances, therefore the trends we observe may be the result of economic as well as policy forces. We lack sufficient data to assess whether the shifts toward increased academic ability levels and relative job satisfaction among incoming teachers are likely to continue over time and across changing economic circumstances. Nevertheless, the patterns that we observe at both the aggregate level and within high-minority urban schools suggest that efforts to bolster the recruitment of academically skilled college graduates into teaching by policy makers and practitioners over the past two decades may in fact be bearing fruit.

Acknowledgments

The authors acknowledge the support of the Smith Richardson Foundation. Min Sun gratefully acknowledges financial support from the National Science Foundation under grant no. DRL-1506494. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. We owe a debt of gratitude to Mark Steinmeyer and Michelle Reininger for their substantive help with the research. We also thank the editor and two anonymous reviewers for their thoughtful suggestions.

REFERENCES

Associated Press
.
2009
.
Budget crisis forces deep cuts at California schools
.
USA Today
,
22 June
.
Auguste
,
Byron
,
Paul
Kihn
, and
Matt
Miller
.
2010
.
Closing the talent gap: Attracting and retaining top-third graduates to careers in teaching
.
Chicago
:
McKinsey & Company
.
Bacolod
,
Marigee P.
2007
.
Do alternative opportunities matter? The role of female labor markets in the decline of teacher quality
.
Review of Economics and Statistics
89
(
4
):
737
751
. doi:10.1162/rest.89.4.737.
Borjas
,
George J.
2002
.
The wage structure and the sorting of workers into the public sector
.
NBER Working Paper No. 9313
.
Clotfelter
,
Charles T.
,
Helen F.
Ladd
, and
Jacob L.
Vigdor
.
2007
.
How and why do teacher credentials matter for student achievement?
NBER Working Paper No. 12828
.
Corcoran
,
Sean P.
,
William N.
Evans
, and
Robert S.
Schwab
.
2002
.
Changing labor market opportunities for women and the quality of teachers 1957–1992
.
NBER Working Paper No. 9180
.
Corcoran
,
Sean P.
,
William N.
Evans
, and
Robert M.
Schwab
.
2004
.
Women, the labor market, and the declining relative quality of teachers
.
Journal of Public Policy Analysis and Management
23
(
3
):
449
470
. doi:10.1002/pam.20021.
Darling-Hammond
,
Linda
.
2010
.
What we can learn from Finland's successful school reform
.
Washington, DC
:
National Education Association
.
Dee
,
Thomas S.
2005
.
A teacher like me: Does race, ethnicity, or gender matter
?
American Economic Review
95
(
2
):
158
165
. doi:10.1257/000282805774670446.
Epple
,
Dennis
,
Richard
Romano
, and
Ron
Zimmer
.
2015
.
Charter schools: A survey of research on their characteristics and effectiveness
.
NBER Working Paper No. 21256
.
Goldhaber
,
Dan
.
2008
.
Teachers matter, but effective teacher quality policies are elusive
. In
Handbook of research in education finance and policy
,
edited by
Helen F.
Ladd
and
Edward B.
Fiske
, pp.
146
165
.
New York
:
Routledge
.
Goldhaber
,
Dan
, and
Roddy
Theobald
.
2013
.
Managing the teacher workforce in austere times: The determinants and implications of teacher layoffs
.
Education Finance and Policy
8
(
4
):
494
527
. doi:10.1162/EDFP_a_00111.
Goldhaber
,
Dan
, and
Joe
Walch
.
2013
.
Rhetoric versus reality: Is the academic caliber of the teacher workforce changing?
CEDR Working Paper No. 2013–4
.
Grissom
,
Jason A.
,
Sean
Nicholson-Crotty
, and
James R.
Harrington
.
2014
.
Estimating the effects of No Child Left Behind on teachers’ work environments and job attitudes
.
Educational Evaluation and Policy Analysis
36
(
4
):
417
436
. doi:10.3102/0162373714533817.
Grossman
,
Pamela L.
, and
Susanna
Loeb
.
2008
.
Alternative routes to teaching: Mapping the new landscape of teacher education
.
Cambridge, MA
:
Harvard Education Press
.
Guarino
,
Cassandra M.
,
Lucrecia
Santibañez
, and
Glenn A.
Daley
.
2006
.
Teacher recruitment and retention: A review of the recent empirical literature
.
Review of Educational Research
76
(
2
):
173
208
. doi:10.3102/00346543076002173.
Ingersoll
,
Richard M.
, and
Henry
May
.
2011
.
The minority teacher shortage: Fact or fable
?
Phi Delta Kappan
93
(
1
):
62
65
. doi:10.1177/003172171109300111.
Johnson
,
Susan Moore
,
Sarah E.
Birkeland
, and
Heather G.
Peske
.
2005
.
A difficult balance: Incentives and quality control in alternative certification programs
.
Boston, MA
:
Project on the Next Generation of Teachers, Harvard Graduate School of Education
.
Krueger
,
Alan B.
1988
.
The determinants of queues for federal jobs
.
Industrial & Labor Relations Review
41
(
4
):
567
581
. doi:10.1177/001979398804100406.
Lankford
,
Hamilton
,
Susanna
Loeb
,
Andrew
McEachin
,
Luke C.
Miller
, and
James
Wyckoff
.
2014
.
Who enters teaching? Encouraging evidence that the status of teaching is improving
.
Educational Researcher
43
(
9
):
444
453
. doi:10.3102/0013189X14563600.
Lankford
,
Hamilton
,
Susanna
Loeb
, and
James
Wyckoff
.
2002
.
Teacher sorting and the plight of urban schools: A descriptive analysis
.
Educational Evaluation and Policy Analysis
24
(
1
):
37
62
. doi:10.3102/01623737024001037.
Leachman
,
Michael
, and
Chris
Mai
.
2013
.
Most states funding schools less than before the recession
.
Washington, DC
:
Center on Budget and Policy Priorities
.
Loeb
,
Susanna
, and
Luke C.
Miller
.
2006
.
A federal foray into teacher certification: Assessing the “highly qualified teacher” provision of NCLB
.
Stanford University CEPA Working Paper
.
Murnane
,
Richard J.
, and
Jennifer L.
Steele
.
2007
.
What is the problem? The challenge of providing effective teachers for all children
.
Future of Children
17
(
1
):
15
43
. doi:10.1353/foc.2007.0010.
Nagler
,
Markus
,
Marc
Piopiunik
, and
Martin R.
West
.
2015
.
Weak markets, strong teachers: Recession at career start and teacher effectiveness
.
NBER Working Paper No. 21393
.
Organisation for Economic Co-operation and Development(OECD)
.
2012
.
Preparing teachers and developing school leaders for the 21st century: Lessons from around the world
.
Available
www.oecd.org/site/eduistp2012/49850576.pdf.
Accessed 15 May 2017
.
Peske
,
Heather G.
, and
Kati
Haycock
.
2006
.
Teaching inequality: How poor and minority students are shortchanged on teacher quality
.
Washington, DC
:
Education Trust
.
Rich
,
Motoko
.
2013
.
Campaign seeks to recruit top students to become teachers
.
New York Times
,
20 November
.
Royston
,
Patrick
.
2004
.
Multiple imputation of missing values
.
Stata Journal
4
(
3
):
227
241
.
Rubin
,
Donald B.
1987
.
Multiple imputation for nonresponse in surveys
.
New York
:
Wiley
. doi:10.1002/9780470316696.
Stockard
,
Jean
, and
Michael Bryan
Lehman
.
2004
.
Influences on the satisfaction and retention of 1st-year teachers: The importance of effective school management
.
Educational Administration Quarterly
40
(
5
):
742
771
. doi:10.1177/0013161X04268844.
Sun
,
Min
,
Yincheng
Ye
, and
Andrew
Saultz
.
2014
.
Federal policy and the teacher labor market: Exploring the effects of NCLB on teacher turnover
.
Paper presented at the 39th Annual Conference of Association for Education Finance and Policy, San Antonio, TX
.
Tucker
,
Marc S.
2011
.
Surpassing Shanghai: An agenda for American education built on the world's leading systems
.
Cambridge, MA
:
Harvard Education Press
.
Villegas
,
Ann María
, and
Tamara F.
Lucas
.
2004
.
Diversifying the teacher workforce: A retrospective and prospective analysis
.
Yearbook of the National Society for the Study of Education
103
(
1
):
70
104
. doi:10.1111/j.1744-7984.2004.tb00031.x.

Notes

1. 

The correlations between the ACT and SAT percentiles among those who took both the SAT and ACT in our three samples are 0.86 in the 1993 cohort, 0.88 in the 2000 cohort, and 0.91 in the 2008 cohort. These high-correlation coefficients indicate that an individual who scored well on the SAT would most likely have scored well on the ACT if he or she were to take this test.

2. 

The job satisfaction survey questions of our interest were either unavailable or were framed in substantially different ways in the 1993 cohort B&B survey. Thus, we exclude this year's data from the analysis.

3. 

Approximately the same percentage of college graduates are either entering teaching or preparing to teach in each of the three cohorts (1993, 2000, and 2008). Comparable test score increases are apparent among both full-time teachers and those who are still just preparing to teach in the 2008 cohort.

4. 

Bacolod (2007) documented a decline in the quality of young female college graduates going into teaching between 1960 and 1990, using standardized test scores, undergraduate institution selectivity, and positive assortative mating (e.g., their husband's position in male wage distribution and their husband's education) as indicators of quality. This decline in teacher quality was significantly explained by the increase in alternative opportunities for young women in other professions. Corcoran, Evans, and Schwab (2004) depicted a similar trend about high school graduates spanning the classes of 1957 to 1992. The likelihood for a female from the top of her high school class to enter teaching fell dramatically during this period of time.

5. 

This increase in teachers’ math ability scores does not, however, correspond to a commensurate increase in the propensity for graduates from STEM majors to begin careers in teaching (Goldhaber and Walch 2013).

6. 

Unemployment rates are seasonally adjusted rates as reported by the U.S. Bureau of Labor Statistics for 1 January of the year following college graduation (i.e., 1994, 2001, and 2009).

7. 

Subject and gender trends appear to be independent. Math scores declined in 2000 among both men and women, and both math and verbal scores declined among male teachers in the 2000 cohort.

8. 

The over- and underrepresentation of black and Hispanic teachers in the 2000 and 2008 time points, respectively, are statistically significant at p-values of less than 0.1.

9. 

Associations between high-minority schools and incoming teacher test scores are very similar in a specification that does not include controls for school percent poverty indicators. While school percent minority and percent low-income characteristics are fairly highly correlated (ρ = 0.56), the bulk of the associations we observe between teacher ability levels and students served stem from the percent of minority students in the teachers’ schools, rather than the percent of poor students served. As a consequence, we focus the bulk of our analyses on high-minority schools in particular.

10. 

Urban high-minority schools have also employed an increasing share of college graduates over time. They employed 9.5 percent of graduates from the 1993 cohort, 12.2 percent from 2000, and 15.5 percent from the 2008 cohort. This trend could relate to demographic and mobility changes among students and families over time, as well as to changes in these schools’ propensity to hire more teachers with college degrees after NCLB.

11. 

California, for example, experienced a particularly steep decline in new teachers’ job satisfaction in 2008–09, with job security satisfaction rates going from 91 percent in 2000–01 to 38 percent in 2008–09, as shown in Appendix figure A.3. This sharp decrease in reported job security among new teachers in California corresponds to the substantial shortfall in public school funds that educators and policy makers anticipated during the 2008–09 school year, in particular (Associated Press 2009).

12. 

These results are consistent with Ingersoll and May (2011), who also provide evidence that minority teachers are more often hired into more challenging work environments.

13. 

Results are virtually identical in a model that includes additional controls for school context—including urbanicity, high-minority, and high-poverty contexts—that we ran as an additional specification check.

14. 

Teach for America (TFA) recruits primarily teachers of very high academic ability and places many of their teachers in high-poverty urban school settings. Their efforts may have played a meaningful—but likely not the only—role in the increasing numbers of very high-scoring teachers entering high-minority urban schools. For instance, 3,600 new TFA teachers entered a mix of urban and nonurban schools in 2008, compared with our estimate of around 18,000 total new teachers recruited by high-minority urban schools nationwide from the 2008 cohort of college graduates.

Appendix A:  Additional Tables and Figures

Table A.1.
Proportion of Missing Observations for Each Independent Variable Used in Our analyses, by Cohort
1993 Cohort2000 Cohort2008 Cohort
Teacher SAT / ACT ability score 0.310 0.239 0.175 
School sector (public or private) 0.124 0.200 0.076 
School poverty (% eligible for lunch subsidies) 0.201 0.503 0.214 
School % minority students 0.099 0.196 0.081 
School locale (e.g., urban, rural, other) 0.222 0.198 0.077 
School level (elementary or secondary) 0.258 0.039 0.076 
1993 Cohort2000 Cohort2008 Cohort
Teacher SAT / ACT ability score 0.310 0.239 0.175 
School sector (public or private) 0.124 0.200 0.076 
School poverty (% eligible for lunch subsidies) 0.201 0.503 0.214 
School % minority students 0.099 0.196 0.081 
School locale (e.g., urban, rural, other) 0.222 0.198 0.077 
School level (elementary or secondary) 0.258 0.039 0.076 

Notes: Independent variables that had no missing data are not shown. Missing SAT scores include individuals who did not take the SAT or ACT and individuals for whom testing data are unavailable.

Table A.2.
Descriptive Statistics for Observations with and without Any Missing School Characteristic Data, by Cohort
1993 Cohort2000 Cohort2008 Cohort
Missing DataNon-Missingt-test p-valueMissing DataNon-Missingt-test p-valueMissing DataNon-Missingt-test p-value
Teacher proportion male 0.288 0.256 0.253 0.292 0.238 0.130 0.241 0.242 0.978 
Teacher proportion white 0.855 0.859 0.906 0.736 0.793 0.071 0.739 0.799 0.159 
Teacher proportion black or Hispanic 0.112 0.123 0.733 0.232 0.190 0.183 0.182 0.150 0.377 
Teacher SAT / ACT percentile 47.8 42.8 0.022 43.3 39.7 0.148 47.5 48.3 0.774 
1993 Cohort2000 Cohort2008 Cohort
Missing DataNon-Missingt-test p-valueMissing DataNon-Missingt-test p-valueMissing DataNon-Missingt-test p-value
Teacher proportion male 0.288 0.256 0.253 0.292 0.238 0.130 0.241 0.242 0.978 
Teacher proportion white 0.855 0.859 0.906 0.736 0.793 0.071 0.739 0.799 0.159 
Teacher proportion black or Hispanic 0.112 0.123 0.733 0.232 0.190 0.183 0.182 0.150 0.377 
Teacher SAT / ACT percentile 47.8 42.8 0.022 43.3 39.7 0.148 47.5 48.3 0.774 

Notes: For simplicity, observations are categorized as missing data if they were lacking data about school sector, school locale, school level, school proportion minority students, or school proportion of students eligible for lunch subsidies. t-tests examine the statistical significance of differences in means between individuals with and without missing data from any of these school characteristics in each cohort.

Figure A.1.

SAT Percentile Ranks of New Nonblack and Non-Hispanic Teachers at High-minority Urban Schools, by Cohort

Notes: Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. High-minority schools are defined as schools serving >66 percent nonwhite students. Data points used to estimate means all include >10 observations.

Figure A.1.

SAT Percentile Ranks of New Nonblack and Non-Hispanic Teachers at High-minority Urban Schools, by Cohort

Notes: Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. High-minority schools are defined as schools serving >66 percent nonwhite students. Data points used to estimate means all include >10 observations.

Figure A.2.

SAT Percentile Ranks of New Black or Hispanic Teachers at High-minority Urban Schools, by Cohort

Notes: Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. High-minority schools are defined as schools serving >66 percent nonwhite students. Data points used to estimate means all include >10 observations.

Figure A.2.

SAT Percentile Ranks of New Black or Hispanic Teachers at High-minority Urban Schools, by Cohort

Notes: Percentile ranks are relative to all college graduates in each cohort. ACT scores are used in cases where SAT test score data were unavailable. High-minority schools are defined as schools serving >66 percent nonwhite students. Data points used to estimate means all include >10 observations.

Figure A.3.

Satisfaction with Job Security, California Compared to Other States

Notes: The bars indicate the percentages of teachers and other professionals who reported being satisfied with their job security. The error bars indicate the 95 percent confidence intervals. Data points in all cells represent >10 observations.

Figure A.3.

Satisfaction with Job Security, California Compared to Other States

Notes: The bars indicate the percentages of teachers and other professionals who reported being satisfied with their job security. The error bars indicate the 95 percent confidence intervals. Data points in all cells represent >10 observations.

Appendix B:  Description of Multiple Imputation Approach to Dealing with Missing Data

Values for some measures of teacher characteristics and school context are missing in each wave because of nonresponse or challenges associated with linking extant datasets. The percentage of missing values on teacher characteristics averages around 5 percent in the 2008 cohort, 13 percent in 2000, and 15 percent in 1993. The percentage of missing values on school characteristics is, on average, about 20 percent across these three waves, except that the number of schools missing free and reduced-price lunch data was around 50 percent in the 2000 cohort. We use multiple imputation to fill in missing data, following the methods implemented in the Stata ice package by Royston (Royston 2004). This appendix describes the multiple imputation procedure that we use.

We impute five datasets, estimate the models separately for each set, and combine the estimates. When choosing variables to include in the imputation stage, we include any other variables that provide information either about the true values of the missing data or about their probability of being missing, including the dependent variables. We use logit models for binary variables and ordinary least squares regression for continuous variables. The estimation only uses imputed values of explanatory variables, not imputed values of the dependent variables (e.g., teachers’ SAT percentiles and their job satisfaction), in order to reduce the risks of overestimating model fit and regression estimates. We then combine the estimates across these five samples using Rubin's (1987) method. All imputation and estimation procedures account for complex survey data by applying both sampling weights and the Balanced Repeated Replication weights appropriate to each of the individual B&B samples.