Teacher Supply, Demand, and Quality: Policy Issues, Models, and Data Bases (1992)

ERLING E. BOE AND DOROTHY M. GILFORD

Since instruction is delivered primarily by teachers, the size, composition, and distribution of the teaching force are vital to the effectiveness of U.S. public education.¹ In turn, public education is widely regarded as central to the nation's social and economic well-being and to its international competitiveness. Consequently, the adequacy of the supply of teachers (in terms of numbers, composition, and distribution) in relation to the demand for them has been a matter of continuing concern among educators and policy makers responsible for ensuring the effectiveness of public schools.

During the current phase of educational reform commencing in the early 1980s, the teaching force has been widely characterized, on the whole, as inadequate to meet current national requirements. In contrast with earlier decades, we no longer define the inadequacy of the teaching force in terms of either its size or the general level of its members' formal credentials. Instead, we are concerned with the ability of the teaching force to deliver high-quality classroom instruction as ultimately measured by student learning outcomes. Numerous recent reports by blue ribbon commissions have concluded that education as a whole, and teachers in particular, are not producing public school students with the level of knowledge, skill, and discipline required in today's complex and competitive economic climate.

As a result of this broad and critical conclusion about the condition of U.S. public education, numerous policy responses have been, and continue to be, made. Among the policy responses is the recent specification of national education goals by President Bush and state governors, coupled with procedures to monitor progress toward these performance targets. As a major initiative for achieving these goals, the federal administration launched its America 2000 program, a massive effort to develop and test new and better ways to organize and operate schools and to measure student outcomes. Similarly, a number of policy initiatives are under way to upgrade the quality of the nation's teaching force. Among many that could be cited are policies that create alternative entry routes for new teachers, provide federal support for teacher preparation in several fields, and establish a national system for certifying teachers at a high level of teaching competence.

Given the current size (about 2.5 million) and complexity of the nation's teaching force in public education, the development of effective, manageable, and affordable policies to improve the overall quality and distribution of this work force is a formidable task. This task surely would benefit from reliable, detailed information about teacher supply, demand, and quality (TSDQ).

In response to the need for such information, the National Center for Education Statistics (NCES) asked the National Research Council (NRC) to convene a Conference on Teacher Supply, Demand, and Quality: Policy Issues, Models, and Data Bases. The proceedings published here are based on this conference, which was convened by the NRC's Committee on National Statistics (CNSTAT) and Division of Education. Training, and Employment (DETE) in 1991. The overarching purpose of the conference was to contribute to the development and retention of a teaching force of the highest possible quality for U.S. elementary and secondary schools through three activities:

• Identification of the major issues that policy makers can expect to face in the decade to come in developing and retaining a teaching force of the highest possible quality, and of their information needs in doing so;

• Review of the adequacy of leading TSDQ projection models in relation to these information needs and delineation of specific directions in which such models should be developed further to meet these needs more completely; and

• Review of the adequacy of state, regional, and national data bases relevant to TSDQ with respect to the basic data required to generate the information needed by policy makers and delineation of specific directions in which data bases should be developed further to meet these data requirements more completely.

Given this focus on policy issues, projection models, and data bases, con-

ference presentations naturally do not provide a comprehensive review of current knowledge about teacher supply, demand, quality, and shortage. Such reviews are available in the literature. The conference was designed to generate a variety of ideas about the topics addressed rather than to reach a consensus expressed as formal recommendations. These proceedings contain background papers prepared for the conference, invited discussions of these papers, and summaries of open discussion of the ideas presented. Accordingly, the proceedings represent the individual views of conference participants and do not necessarily reflect the views or policies of CNSTAT, DETE, or NCES.

The conference was designed to stimulate discussions and suggestions (a) concerning information needed by policy makers and others in addressing current and emerging TSDQ issues, and (b) for further development of models and data bases needed to generate this information. From the perspective of NCES, the conference was also intended to provide information about the utility and role of its existing data bases and programs in analyzing TSDQ issues and to provide suggestions for enhancing its contributions to this topic.

The conference brought together a broad spectrum of individuals with interests in TSDQ from the perspectives of policy making, modeling, data base development, practice, and research. Sessions were organized around the papers prepared on teacher quality issues, projection models, and data bases, along with prepared discussions of these papers followed by open discussion among conference participants. The conference agenda and the list of conference participants appear in Appendix A.

This volume is organized for the most part in accordance with the agenda: each of the three main topics is followed by a summary of the open discussion among participants, entitled "general discussion." Finally, summary descriptions of national data bases containing information relevant to TSDQ are presented in Appendix B, and the specific TSDQ data elements (i.e., variables) recorded in these data bases are organized and presented in tabular form in Appendix C.

This summary chapter provides a synopsis of the entire conference including TSDQ information needs and suggestions for action made by participants during open discussions of each major topic. Following an overview of teacher supply, demand, and shortage, the remainder of this chapter provides sections on (a) a major policy issue: teacher quality; (b) TSDQ projection models; and (c) teacher data bases at the state, regional, and national levels. Particular information needs and suggestions for action associated with each of these topics are included in these sections. The summary concludes with several general suggestions for action.

The information needs and suggestions for action reported here are not formal recommendations adopted by vote of conference participants and

therefore are not referred to as recommendations. Instead, such needs and suggestions emerged during the course of numerous open discussions by participants occurring during the course of the conference. The editors of this volume organized them in the form reported in this summary.

The national demand for public school teachers is defined operationally, and in the aggregate, as the total number of teaching positions funded by local education agencies (LEAs), i.e., the number that LEAs are able and willing to employ at a given time. Total demand thus defined is the end result of a number of considerations leading to the establishment of teaching positions. The main factors determining teacher demand in any particular year are the number of students enrolled in public schools, policies pertaining to curriculum and teacher-pupil ratios, prior commitments to employed teachers, LEA funding capacity, and the prices that must be paid for various types and qualities of teachers.

Aggregate demand, however, is of little use in understanding the dynamics of demand for the teaching force or in designing policies to ensure an adequate supply of teachers. For these purposes, total demand must be specified in greater detail, i.e., disaggregated by teaching assignment and geographic distribution of the teaching positions.

More specifically, computations of disaggregated teacher demand should be stratified by subject matter, grade level, preparation for serving the special needs of students (especially handicapped students and those with limited English proficiency, region of the country, and urbanicity of schools within which teaching positions have been established. In addition, demand should be specified by the attributes of teachers desired, especially teacher qualifications (their training, degree level, licensure, and experience) and race/ethnicity. When specified at this level of detail, teacher demand can be compared with information about teacher supply to examine supply-demand relationships.

The national supply of public school teachers in any year is defined, in the aggregate, as the number of eligible individuals available from all sources who are willing to supply their services under prevailing conditions. The supply includes qualified individuals who (a) currently hold teaching positions, (b) seek to enter the profession by applying for open positions, and

(c) would apply for positions if suitable openings existed. The main factors determining who is available to teach are considered to be the availability of teaching positions relative to the availability of positions in other occupations, teacher wages relative to wages in competing occupations, and working conditions in teaching relative to conditions in other occupations. Unfortunately, no sources of data are capable of providing adequate information about the total supply of teachers thus defined (Gilford and Tenenbaum, 1990).

What is known with reasonable precision is the annual number of teachers hired from among those available through several sources of supply. That is, the number of individuals continuing in public school teaching from one year to the next is known, as is the number of individuals entering public school teaching annually. The former group is often called continuing teachers, and the latter group is often called entering teachers or new hires. Collectively, continuing and entering teachers constitute the cohort of individuals employed as teachers (in short, the teaching force²), a group representing an unknown proportion of the potential total supply of teachers.

Aggregate information about the size of the teaching force is of only modest value for understanding teacher supply. In practice, it is virtually the same as aggregate demand. To be useful in understanding the teaching force, information is needed about various sources of supply of individuals hired as teachers, as well as about the composition and distribution of the teaching force. Information at this level of detail could then be related to comparable information about teacher demand in efforts to understand the degree to which teacher demand is being met by qualified individuals, as well as the sources of teachers that might be manipulated by policy in order to provide a more adequate supply.

In practice, the term supply (as in teacher supply and demand) is typically used imprecisely. Instead of referring to total potential supply, the expression teacher supply is used loosely to refer to the composition of the actual teaching force, to potential sources of entering teachers such as recent graduates of teacher preparation programs, and to teacher supply shortages that occasionally occur in some subject matter fields at various geographic locations. The total potential supply of hireable individuals almost always equals or exceeds the number of available teaching positions. Therefore, in the aggregate, the size of the teaching force is usually determined by the demand for teachers as defined by the number of funded teaching positions, not by supply constraints.

As previously mentioned, the teaching force is composed of two large groups—employed teachers continuing from year to year and entering teachers

in any year. Both are broad categories drawn from more specific sources. Continuing teachers typically have the option of remaining in the same position from one year to the next. Nonetheless, many practicing teachers choose to apply for teaching positions in other schools, in other subject matter fields, or both. Furthermore, some employed teachers may be reassigned to different teaching assignments within a school or reassigned to a different school within the same LEA. Thus, the flows of practicing teachers within the public education system constitute a major source of teachers hired into, or reassigned to, open teaching positions. Their transfer within the system creates openings in positions they vacate, assuming such positions continue to be funded by the LEA. Due to attrition of teachers from the profession and gradual expansion of the total number of teaching positions, a large number of additional individuals are also hired by the public education system each year (Rollefson, 1992). Such entering teachers are drawn from four sources:

• A reserve pool of qualified teachers composed of:

1. experienced former teachers and

2. graduates of teacher preparation programs from prior years (sometimes called delayed entrants);

• Recent graduates of teacher preparation programs (some of whom are also experienced teachers);

• College graduates who have not completed a teacher preparation program and who have not previously taught (sometimes referred to as entrants via alternate routes); and

• Teachers employed in private schools who migrate to teaching positions in public schools.

In view of this complexity in the sources of employed teachers, detailed information about flows of teachers into and within the profession is vital to understanding the relative importance of these sources of teacher supply.

The composition of the teaching force is a multifaceted concept that includes teacher qualifications (especially in relation to specific teaching assignments); subject matter specialty; grade level; preparation for serving special needs of students (especially handicapped students and those with limited English proficiency); and teacher characteristics such a race/ethnicity, gender, and age. Existing teacher data bases can provide good information about these and other attributes of the teaching force. This information makes it possible to determine how well the supply of teachers yields a teaching force that corresponds to the demand for teachers with desired characteristics, for various subject matters, grade levels, etc. It would also

be very desirable to have similar information about the composition of applicant pools for open teaching positions and, in fact, about the total potential supply of teachers.

The teaching force is distributed among public schools that vary by type, grade level, and location. A major concern is that teachers are maldistributed among schools in terms of qualifications, experience, race/ethnicity, and other dimensions of the teaching force. For example, high schools in large urban areas usually attract a teaching force that is less experienced, younger, and less well prepared to teach high school subjects than teachers hired at nearby suburban schools. Therefore, information about the distribution of the teaching force needs to be presented in terms of teacher variables (such as qualifications) to understand fully how well the supply of teachers meets the demand for teachers at schools of various types, levels, and locations. Such analyses of the teaching force are possible with existing teacher data bases. Little is known, however, about the characteristics of applicants (from which entering teachers are selected) as a function of school location. Unless information about applicants is known, it is not possible to determine whether the supply of teachers available to various schools is adequate, or whether difficulty in hiring qualified teachers is due to hiring practices or other factors. This distribution problem stems from teachers' behavioral response to school location, one of the many variables affecting the supply of teachers available to a school. Supply obviously can vary from school to school since supply is a relationship between the number of qualified individuals who would be willing to teach and such incentives as the salary, working conditions offered, the location of the school, and other alternative career opportunities.

Policy makers responsible for ensuring the effectiveness of public education are necessarily concerned that the supply of qualified teachers is sufficient for all schools. Consequently, the topic of actual and potential teacher shortages is being addressed continually. Only occasionally is there concern about teacher surpluses, although that, too, can be a significant policy issue (especially for state governments) if the capacity for preparing teachers in certain fields is well in excess of demand.

In recent decades, there has been little or no shortage of individuals

available to fill open teaching positions; in 1987–88, only one percent of U.S. teaching positions in the public sector were unfilled (NCES, 1991a). The supply of teachers—in gross numbers—has generally been well in excess of demand. During the early 1980s, however, it appeared that general teacher shortages could develop in the late 1980s and 1990s.³ The following trends seemed to appear in several variables that affect teacher supply and demand in a direction that would increase demand and reduce supply:

• High teacher attrition from the profession partly due to low salaries and poor working conditions,

• Increasing teacher retirement rates due to an aging teacher force,

• Rising public school enrollments due to the echo of the baby boom,

• Continuing decline in teacher-pupil ratios,

• Falling enrollments in teacher preparation programs,

• Decreasing interest among women in teaching due to more lucrative opportunities in other professions, and

• Constriction in the numbers of entering teachers because of more stringent entry standards including entry-level teacher tests and early performance assessments.

Most of these trends did not develop, nor did teacher shortages materialize. Both attrition and retirement rates were lower than expected, a behavioral response on the part of teachers, which for some teachers was probably due to the desire for a high quality of life combined with financial conditions that required a family to have two incomes. Consider the actual trend for each of the projected trends.

High Teacher Attrition. In the early 1980s, attrition rates for public school teachers were estimated to be 8 percent based on an old study (Metz and Fleischman, 1974). More recent data show that the annual teacher attrition from 1987–88 to 1988–89 was 5.6 percent in the nation as a whole (Bobbitt et al., 1991). This lower attrition rate is one of the main reasons why the projected teacher shortages have not materialized. There are several reasons for the lower attrition rate. First, the proportion of young teachers was considerably smaller in the late 1980s than in the early 1970s, and middle-aged teachers have a lower attrition rate than young teachers. Second, women are leaving teaching at all-time low rates, and when they do leave they more often return and take shorter breaks from teaching (Grissmer and Kirby, 1991). This shift has little to do with salary or working conditions, but rather is due to the increasing importance of women's salaries as part of family income over the last 20 years. Third, new teachers are more often drawn from the 30–45 age group (Murnane and Olsen, 1989; Kirby et al., 1991; Murnane and Schwinden, 1989), and individuals who enter at a later age have lower attrition than those entering at younger ages. Fourth,

teacher salary levels began to rise in the early 1980s, and much recent work has shown that attrition rates decline as salary increases (Murnane and Olsen, 1989, 1990; Murnane et al., 1989; Grissmer and Kirby, 1991).

Increasing Teacher Retirement Rates. Because the average age of the teaching force has been gradually increasing for a number of years, teacher retirement rates have been increasing and are expected to continue to increase during the next 15 years. Since the average age of teachers is approximately 42, half of all teachers are within 13 years of retirement eligibility. However, most teachers do not retire at age 55; instead, retirement in the 62–65 age range is becoming more common. This slower rate of retirement means that new teacher demand is increasing more slowly than previously thought, and it is more likely that large numbers of teachers will retire during the 2000–2010 period than during the 1990s.

Rising Enrollments and Falling Teacher-Pupil Ratios. Expected higher public school student enrollments nationally have materialized and probably have been higher than originally predicted due to immigration. Teacher-pupil ratios have continued to decline, especially for earlier grades and for mathematics and English courses in high schools. However, further declines in teacher-pupil ratios appear unlikely, at least in the near future, since many states have severe budget problems. In some cases, increases in teacher-pupil ratios may occur.

Falling Enrollments in Teacher Preparation Programs. Although falling enrollments in teacher preparation programs were projected, actual enrollments have been increasing fairly strongly in recent years. Assumptions made in the early 1980s did not take into account student reaction to the perceived increased job opportunities in teaching. Perhaps even more important is recent research showing that new graduates from teacher preparation programs usually fill one-half or less of teacher vacancies each year (Murnane et al., 1989; Kirby et al., 1991). Other sources (former experienced teachers, previously trained but inexperienced teachers, and migrating teachers from private schools) meet a large share of the annual demand for entering teachers. Therefore, new teacher education graduates, instead of being needed to fill most of the demand for entering teachers, are needed to fill only part of this demand.

Decreasing Interest Among Women in Teaching. With increasing opportunities for women to enter medicine, law, business, and other professions during recent decades, forecasters anticipated a decline in the proportion of women in the teaching profession. Instead, the increasing proportion of women in the overall labor force and the increasing role they play in providing family income seems to have resulted in stronger interest by women in teaching, at least as measured by increases in the proportion of women in this profession. In Indiana, for example, women have become a larger percentage of the teaching force during the past 20 years (Kirby et

al., 1991; Grissmer and Kirby, 1991). A similar trend occurs in national statistics. In 1966, 69.0 percent of public school teachers were women. This percentage dipped to a low of 65.7 in 1971, but then gradually rose to a new high of 70.2 percent in 1988 (NCES, 1991a). Although it is true that more women are entering other professions, it is also true that more women are entering teaching. The large increase in the population of women ages 20–25 and their greater labor force participation has resulted in more women in almost all professions.

More Stringent Entry Standards. Little evidence exists to support the hypothesis that fewer teachers enter and continue in the profession due to either fear or failure of entry tests or of performance assessments. The evidence from Indiana is that attrition rates of young teachers have declined—not increased—over the last five years, roughly the period when testing was introduced. Most teacher testing may simply delay entry into teaching. While individuals may fail a teacher qualifying test the first time, the failure rates after multiple chances at passing are generally very low, Furthermore, few states or LEAs have adopted testing programs that could serve as barriers to aspiring teachers.

Although most people now recognize that the supply of teachers is generally adequate to fill the ranks of the nation's teaching force, a major concern remains about a shortage of teachers who are able to deliver high-quality instruction in the classroom. In other words, the match between the composition of the teaching force (in terms of subject matter knowledge, instructional skills, fluency in multiple languages, and demographic characteristics) and the demand for teachers with such abilities and characteristics is far from optimal.

In short, there are too few teachers who are able to perform at a high level in their particular teaching assignments. For example, there are shortages of highly qualified teachers in certain subject matters such as science and mathematics (Gilford and Tenenbaum, 1990) and in teaching fields such as bilingual education (Schmidt, 1992). Likewise, there is a shortage of minority teachers (American Association of Colleges for Teacher Education, 1987). Thus, the problem of teacher shortages is now construed in terms of certain inadequacies in the qualifications and characteristics of the teaching force, but not in its size in relation to gross demand.

Just as the problem of teacher shortage is now construed in terms of inadequacies in the composition of the teaching force, it is similarly con-

strued in terms of inadequacies in the distribution of the teaching force. Teachers are needed, of course, in schools that vary by grade level and by location in urban, suburban, and rural areas. While almost all teaching positions are filled in all these schools, the qualifications of teachers are unequally distributed among them (Oakes, 1990). For example, the shortage of highly qualified teachers in many urban schools is considered to be greater than the overall shortage of such teachers. In this instance, the concern about teacher shortages is defined in terms of the distribution of qualified teachers among schools of different characteristics, not about a general shortage of teachers.

Since there is a shortage of highly qualified teachers in some teaching assignments at some locations, it might be expected that a significant proportion of teaching positions would remain unfilled. Yet that is not the case. Hiring practices in the field of education ensure that ''teachers'' are present to staff almost all classrooms. Only one percent of teaching positions nationally were unfilled in 1987–88 (Hammer and Gerald, 1991).

In particular, three strategies are used to equilibrate supply and demand, two of which work on the supply side and one on the demand side. The main mechanism used is to relax qualification requirements during hiring. If a highly qualified applicant is not available to fill an open teaching position, a less qualified applicant typically will be hired. In fact, many teachers are hired on emergency certificates shortly before a school year begins, a strong indication that a fully qualified candidate was not available. Other ways to compromise on applicant qualifications include hiring experienced teachers with poor performance records or hiring otherwise qualified teachers out of their fields of competence. These compromises with qualifications, which are made to equilibrate teacher supply and demand, are counterproductive to efforts by policy makers and administrators to improve the quality of teaching practice.

Another supply side mechanism used to equilibrate teacher supply and demand is to offer financial incentives for teachers to enter and continue in the profession, i.e., to enhance teacher supply. Some of these incentives entail bonuses offered to teachers in shortage teaching fields, such as bilingual education, and at shortage locations, such as schools in large urban centers. Other financial strategies for enhancing teacher supply are to raise overall teacher salaries so as to make the profession more competitive with other occupations and to raise entry-level teacher salaries substantially so as to attract more novice teachers to the profession.

If the supply of hireable teachers is still not sufficient to fill open teaching positions, then the demand for teachers can be reduced and brought

into line with available supply by increasing the workloads of employed teachers. This can be done by increasing class sizes and by increasing the average number of classes assigned to teachers. Both approaches increase the teacher-pupil ratio. While these strategies can be effective in equilibrating teacher supply and demand and thereby ensuring that all classrooms are staffed with a teacher, they do little to enhance the quality of teaching practice. On the contrary, increasing workloads can have an adverse effect on the quality of teaching and the morale of teachers.

Currently the principal policy issue relevant to teacher supply and demand is how to improve teacher quality while simultaneously maintaining a sufficient supply of teachers to meet demand. This issue is particularly important because it is widely presumed that higher-quality teachers will engage in higher-quality teaching practices in their classrooms, which will lead directly to improved student learning outcomes—the prime objectives of all stakeholders in public education. While this presumption may be intuitively obvious, education researchers have found it difficult to demonstrate robust relationships between potential indices of teacher quality, on one hand, and student learning outcomes, on the other (Hanushek, 1986).

The design of effective policies to improve teacher quality has been impeded by three problems. First, there is little general agreement about what specific characteristic of teachers indicates quality. Second, existing models of teacher supply and demand do not address the subject of teacher quality and therefore offer no guidance. Third, data on variables that might indicate teacher quality are very limited, a circumstance that restricts research that might lead to practical measures of quality.

In spite of these problems in defining teacher quality, education policy makers, administrators, and researchers have given considerable attention to five dimensions presumed to indicate quality in education. Two of these dimensions, teacher qualifications and tested ability, are characteristics of individual teachers; they pertain specifically to "teacher quality." The third dimension, the demographic matching of teachers and students, pertains to the quality of a teaching force such as the faculty of a particular school. The fourth dimension, teacher professionalism, involves the degree to which teachers are given responsibility for and authority over their work; it represents a quality dimension of LEA policy. Finally, the fifth dimension, classroom teaching practice, refers specifically to teaching quality, as distinguished from teacher quality. Each of these dimensions is considered in turn in the following sections.

Traditionally, teacher quality has been defined by a teacher's formal qualifications, which include completion of a teacher preparation program, accreditation status of the program completed, course work or earned degree(s) beyond the baccalaureate, subject matter majors and minors, credentials by subject matter (state certification/licensure status), and prior teaching experience. A standard or full credential (as contrasted with probationary, temporary, emergency, etc.) in some particular subject matter is intended to serve as a summary indicator of at least minimally adequate training and experience as defined by the particular state issuing the certificate. Thus, an individual holding, or eligible for, a standard certificate is presumed to be of higher quality for teaching in the subject matter of certification than an individual without a certificate.

There are at least three problems with using teacher certification as an indicator of teacher quality. First, standards and procedures for teacher certification vary widely by state. To qualify for standard certification, some states require a master's degree, a qualifying examination, participation in a teacher induction program, and/or a major in the subject matter of certification; others do not. These variations cause problems in equating credentials in conducting cross-state analyses and research on teacher quality. Second, a teacher who holds a standard certificate in two subject matters (e.g., biology and chemistry) but not a third (e.g., mathematics) may be assigned to teach classes in all three—a not uncommon event in a small high school. In this circumstance, it is not clear whether the teacher should be considered to be qualified. Third, past research has not demonstrated reliable or substantial associations between any of the dimensions of teacher qualifications and student learning outcomes (Gilford and Tenenbaum, 1990: Hanushek, 1986).

Nonetheless, teacher certification serves the symbolic function of defining professional qualifications for the teaching profession, a necessary condition for defining any recognized profession. Furthermore, it is argued by some that defining teacher quality in terms of credentials is all that can be done reasonably, and that this is precisely the practice used in other professions. A license to practice medicine, for example, is defined by completing an accredited program of medical education and the passing of a qualifying test. This defines, at the very least, a minimum level of quality of a person meeting these standards. Just as physician quality is not defined formally by the morbidity or mortality statistics of patients treated, so the argument goes, teacher quality should not be defined by associating teacher credentials with student learning outcomes.

In recent years, education policy makers have attempted to strengthen teacher credentials in several ways. Several states have moved to regulate teacher education curricula by increasing course requirements in subject matter fields, while decreasing or limiting courses in professional education. However, these policies, even if effective, will affect only newly prepared teachers, not practicing teachers or individuals in the reserve pool of potential entering teachers. A related, but partially contrary, development comes from The Holmes Group, a consortium of education deans of research universities, which strongly advocates a more demanding credential for teachers. This group maintains that a four-year teacher preparation program at the baccalaureate level is insufficient, that instead the education of prospective teachers should be concentrated in subject matter disciplines at this level. They further propose that the professional preparation of all beginning teachers should be provided by some means, e.g., a five-year baccalaureate or a master's degree.

Another policy initiative to strengthen teacher credentials has been the formation of the National Board of Professional Teaching Standards (NBPTS). The board plans to conduct examinations that highly capable teachers can take to qualify for board certification as master teachers. This plan, still in the development phase, uses the board certification system in the medical profession its it model.

Finally, another initiative, the use of alternative routes to certification, has gained momentum in recent years. A total of 33 states now have provisions for some form of alternative route to certification (Feistritzer, 1990). Through alternative routes, individuals who have earned baccalaureate (or higher) degrees but who have not completed a traditional teacher preparation program are allowed to short-circuit normal requirements for obtaining a teaching credential. Instead, some teacher training is typically provided by a variety of means, such as intensive summer programs before teaching commences, inservice training after teaching commences, and/or close supervision by experienced teachers once teaching begins. The use of alternative routes into the teaching profession was first proposed as a means of increasing the supply of math and science teachers, but proponents of this approach argue that it will also enhance teacher quality by attracting new teachers who are brighter, better educated in subject matter fields, and more experienced than the typical teacher education graduate. The general effectiveness of alternative route programs in achieving these objectives remains to be demonstrated.

There is evidence that college students who enter teacher preparation programs tend to score lower on tests of academic aptitude than students

entering other career tracks; that, among teacher education graduates, lower-scoring students are more likely to enter the teaching profession; and that, among employed teachers, those who score higher on the National Teacher Examination (NTE) tend to leave the profession sooner (Murnane et al., 1989; Murnane and Olsen, 1990). Thus, throughout the supply pipeline, lower-scoring individuals tend to be selected for and retained in the teaching profession.

While these trends appear to be contrary to the objective of securing a teaching force of high quality, the general relevance of tested ability of teachers to the task of teaching and to the quality of student learning outcomes has not been demonstrated.⁴ One reason for this may be that tests used to measure teacher abilities are much too narrow and have been limited to the multiple-choice format. The ability to select the correct answer on an aptitude or achievement test may not indicate anything about a teacher's ability to solve open-ended problems, to explain solutions to students, or to create problems for teaching students. While intellectual ability is obviously relevant to teaching, it is not known how low ability test scores must be before teaching performance is adversely affected.

Nonetheless, a substantial majority of states include some form of teacher assessment in the certification process for new teachers. Yet no state has been able to demonstrate the job relevance of these test scores. To avoid lawsuits, cutoff scores are set very low and ample opportunities are provided for those who fail to be retested. Even if test performance were demonstrated to be relevant to teaching performance, the use of tested ability in initial teacher certification would affect only new teachers, not teachers certified before the tests were introduced.

Two national initiatives are currently under way to develop better teacher ability tests. The process being developed by NBPTS for certifying master teachers involves a test, and the Educational Testing Service is developing a new version of the NTE. In addition to measuring subject matter knowledge, the tests being developed by both sources will include major performance components that will require teachers to demonstrate teaching practice skills.

Demographic matching refers to the similarity between the distribution of demographic characteristics (such as race/ethnicity, gender, and age) of the teachers in a school and those of the students. It has been argued that the distribution of the teaching force by race/ethnicity constitutes a significant dimension of quality relevant to the outcomes of schooling for minority students who need role models. There is concern about the teaching force in this respect because it is largely white, suburban, middle class, and

female. By contrast, demographic trends in the nation as a whole (and especially in the population of public school students) are clearly moving dramatically in the direction of a larger proportion of minority populations (especially of Hispanic background) and of an increasing proportion of students from young families with substandard incomes.

Even if the demographics of the teaching force matched the demographics of public school students on a school-by-school basis, it would not be possible to match teachers and students on a class-by-class basis, since classes of students are usually heterogeneous. Given this difficulty, perhaps the main concern should be with improving the quality of classroom teaching practice because of its centrality to academic outcomes, rather than teacher-pupil demographic matching.

Nonetheless, students are generally aware of the demographic characteristics of teachers in their schools, not just their own classroom teachers. Representative diversity of the teaching force is considered to be important because minority students will be exposed to role models of like characteristics who are achieving in the teaching profession. Moreover, diversity within a school's teaching force may increase knowledge and understanding of different cultural groups in both students and teachers and improve students' chances of interacting successfully with different cultural groups that they are likely to encounter later in the workplace. For these reasons, demographic matching of teachers and students is considered to be a quality dimension of education.

Policy initiatives to increase demographic matching have concentrated on the supply of new teachers by attempting to increase the numbers of minority students entering college, with the hope that these increased numbers will yield more minority students who elect teaching as a career. The main inducement used for increasing minority enrollment in college has been the provision of financial assistance. To be effective, this strategy must be sustained over many years because it addresses only the production of new teachers, not the demographic mix of the active teaching force and the reserve pool of potential teachers.

The fourth dimension of quality of interest to policy makers is the professional lives of teachers. In view of the complexity and context-specificity of teaching, it is widely advocated that teachers should have professional autonomy and responsibility similar to that typically accorded members of other professions. Advocates of conferring professional autonomy on teachers argue that it will enhance the attractiveness of the profession as a career choice and will improve the quality of classroom teaching practice. Though these arguments are often made and are cogent, there

is little research evidence to support them, and so there is ample room for skepticism.

Several major trends in education policy promote teacher professionalization. The rapidly proliferating practice of decentralizing authority over major aspects of school operations from district offices to school councils (under the rubric of school-based management) usually places teachers in key decision making roles as council members. Likewise, programs involving merit pay, career ladders, and/or peer evaluation typically entail enhanced professional responsibility for teachers. These programs are often resisted by teachers because they are difficult and time-consuming. They require special training for decision making and management and, in addition, require released time from instruction—resources that all too often are not made available. In addition, policies that require teachers to collaborate with each other and to observe and evaluate peers are counter to the traditional culture of teaching. Not surprisingly therefore, policies promoting the professionalization of teaching often do not originate with teachers, but instead with educational administrators and policy makers.

The quality of classroom teaching practice has been widely criticized in recent years because of its emphases on basic skills instead of higher-order thinking, on discrete facts instead of problem solving, on teaching for exposure instead of for understanding, and on competition among students instead of cooperation. In general, teaching has been watered down and too little has been expected of students.

Several recent reports on U.S. education argue that schooling must better prepare students for the world of work they will be entering, which will require problem-solving skills, adaptability, and the ability to work in teams. Not only must the content and strategy of teaching change, but teachers must also become increasingly prepared to teach a different kind of student, one that has been alienated traditionally from schools and that is unlikely to be similar to the teacher demographically.

Of the five quality dimensions reviewed here, the quality of teaching practice is most directly connected logically to the objective of increasing the quality of student learning. This is what students experience every hour of the school day. However, there are no clear indicators of the quality dimension of classroom teaching that correlate with student learning outcomes (Hanushek, 1986), and therefore it is difficult to assess strategies intended to change practice.

Improvements in teaching practices are also very difficult to implement, particularly since research indicates that there is a strong tendency for teachers to emulate the teachers they observed as students. We are caught

in a vicious cycle of mediocre practice modeled after mediocre practice, of trivialized knowledge begetting more trivialized knowledge. But traditional teaching is no longer good enough.

Improved teacher training is a critical component of improving teaching practice. Because the teaching force is so large and the vast majority of teachers continue from year to year, inservice training for practicing teachers is needed if rapid change is to occur. Yet there is no broad-based policy initiative (such as school-based management, for example) specifically addressed to upgrading the quality of classroom practice by active teachers. To be sure, there is some inservice training for active teachers, but most state and locally sponsored programs do not incorporate the strategies for influencing teaching practices that have been demonstrated to be effective during two decades of research.

Some policy actions have been initiated to improve the quality of teaching practice of new teachers. For example, several states prescribe a probationary period of a year or two for new teachers during which their teaching practice is assessed. The research on which these assessments are based was conducted at the elementary level, where the content is weighted toward basic skills and therefore has not much relevance to the teaching of higher-order thinking. Another strategy for improving the practice of new teachers is an induction program now required by 13 states. Under these programs, LEAs must provide first-year teachers with some sort of guidance, usually by experienced teachers. However, to the extent that experienced teachers themselves do not engage in the kind of quality classroom practice that is needed today, new teachers are socialized into traditional, but increasingly inadequate, teaching practices through such induction programs.

An alternative model for improving teaching practice currently under development is the professional development school (PDS), in which student teachers receive their practical experience and active teachers have opportunities to learn more about teaching. In contrast to traditional lab schools operated for decades by many colleges of education, PDSs are genuine collaborations between university-based teacher education programs and public schools. So far, although PDSs are intended to improve both the preparation of new teachers and the practice of the existing teaching force, they have emphasized the preparation of new teachers and have not developed means by which benefits will ripple to other schools within a LEA. Consequently, the potential of this model to upgrade teaching practice on a substantial scale remains to be demonstrated.

Finally, major policy efforts are being made in many states and by the federal government to restructure schools, the basic intent of which is to improve the quality of educational outcomes. Although these efforts may lead to significant improvements in teaching practice, as well as other as-

pects of school organization and operation, they are not targeted specifically at improving the quality of classroom practice. Hard data are not yet available about the educational benefits of restructured schools, including improvement in teaching practice.

As stated earlier, it is widely presumed that higher-quality teachers will engage in higher-quality teaching practice in their classrooms, which will lead directly to improved student learning outcomes. The general validity of this presumption is examined critically here in light of available knowledge and in terms of prospects for attaining further improvements in teacher quality.

If we recognize improved student learning outcomes as the prime objective of public education yet to be attained, then it is certainly important to identify factors that can be manipulated by educational policy that will yield progress toward achieving this objective. Improving teacher quality further is one major possibility—one that has been manipulated by policy for more than a century.

However, it should not be assumed that improvement in teacher quality (i.e., credentials or tested ability) will lead automatically to improvement in the quality of teaching practice or in other aspects of educational quality (i.e., demographic matching of teachers and students and teacher professionalism). For example, the selection of teachers who have earned master's degrees or who have scored high on ability tests will not result automatically in better teaching, unless they are taught how to teach more effectively. Likewise, improving the match between teachers' credentials and their teaching assignments will not necessarily improve practice, nor will reducing demographic disparity between the teaching force and the student population or adding programs to professionalize teaching. Conversely, improving classroom practice will not necessarily improve teacher credentials or reduce demographic diversity. Improvement in any one of the five dimensions of educational quality reviewed here will not necessarily improve any of the other dimensions. To the extent that these dimensions of quality are considered to be important, then improvement of each should be addressed directly, and any intended impact on teaching practice should be assessed instead of assumed.

An important question is whether the potential for further improvement in teacher quality is sufficient to affect student outcomes favorably. There are practical limits on gains that might be possible in indices of teacher quality. Given the sheer size of the teaching force in public education—approximately 2.5 million individuals employed as teachers (NCES, 1991a), the task of raising overall teacher quality is indeed formidable. About 10

percent of college-educated women and 4 percent of college-educated men are employed as teachers in the United States; no other profession draws such a large proportion of college-educated adults (Lanier, 1986). If entering the teaching profession were made so attractive in terms of compensation and working conditions that many more of the best and brightest went into teaching, then we would soon be worrying about a tested-ability problem in other professions in the private sector such as medicine, law, and engineering. Salaries in these professions would no doubt escalate to compete with the higher salaries for teachers. Under any likely scenario, the size of the teaching force would simply be too large to be filled by individuals from the upper levels of tested ability.

As previously stated, research on measures of teacher credentials and tested ability has demonstrated only very limited associations with student learning outcomes. There are several plausible reasons for this lack of association. First, available measures of teacher quality may have little relevance to the quality of teaching practice, though this cannot be determined with certainty because independent measures of the quality of teaching practice do not exist. Second, policy efforts over many decades to improve teacher credentials may have so restricted the range of differences among teachers that the remaining variability is irrelevant to the student learning outcomes. For example, a minuscule 0.6 percent of public school teachers have not earned a baccalaureate degree (NCES, 1991a). Thus, for all practical purposes, the range of variability on the degree credential is restricted to the difference between baccalaureate and master's degrees. Third, variables other than teacher quality, such as school policies, student readiness, and parents' expectations, may be so strongly associated with student learning that the contribution of the remaining variability in teacher quality is obscured. Finally, current measures of student learning outcomes may not capture critical aspects of student learning and therefore may not reflect adequately the contribution of classroom practices. These measures have been predominantly paper-and-pencil tests of the multiple-choice type. which primarily measure basic skills. Dissatisfaction with such measures of student learning is leading to the development of alternative performance-based measures.

The several limitations reviewed here in understanding and improving teacher quality as currently measured do not suggest that teacher quality is irrelevant. Basic command of subject matter knowledge is obviously vital in order to teach mathematics or Russian, for example. What is not clear is the relative importance of variations in teachers' subject matter knowledge and teaching skill above some threshold level when compared with the importance of other variables related to student learning.

During the course of conference presentations and discussions, participants identified additional information needed to address teacher quality issues. Much of this information is relevant to planning improvements in TSDQ data bases, and some can be useful in designing future research with existing data bases. Conferees also made a number of suggestions for actions to enhance knowledge about teacher quality. These information needs and suggestions are reported as such below. However, they are not formal recommendations adopted by vote of conferees, and should not be interpreted as such.

Information Need 1: Teacher Quality Indicators. Since understanding the importance of teacher quality is severely limited by the lack of accepted measures of this attribute, measures are needed that give a truer and more useful indication of teacher quality. Information is also needed about what such indicators reveal with respect to teacher quality-quantity tradeoffs, quality-price relationships, and projections of quality and quantity of teacher supply.

Information Need 2: Teacher Credentials. Since states license teachers under widely differing systems, better information is needed for individual states about approved and rejected applications for teaching credentials by type (i.e., permanent, emergency, etc.) and source (e.g., out-of-state teachers, range of college curricula leading to initial entry to teaching, including alternative routes) and about the match between teaching credentials and assignments and about the importance of teaching credentials to teaching practice. Information is also needed about the qualifications of teachers who fill new vacancies. More information is needed about whether and how teachers' credentials influence school district hiring, salary, and promotion practices. Finally, information is needed about whether and how credentialing policies influence the attractiveness of teaching. Do course requirements influence a student's decision to enter teaching?

Information Need 3: Demographic Matching. Given the widespread interest in promoting minority representation in the teaching force and the presumed relevance of demographic diversity of this force to desirable educational outcomes, information is needed at the school level about demographic mix of the teaching force and about how well this mix matches the demographic mix of enrolled students. In addition, more information is needed about the race/ethnicity of members of the teacher supply.

Information Need 4: Teacher Professionalism. The large number of broad policies intended to promote teacher professionalism (some of which are opposed generally by teachers) make research into their effectiveness difficult. As an interim step, it would be useful to document teachers' attitudes about taking more professional responsibility. Data are needed

about teacher perceptions of their control over various aspects of their work and about the relationship of variations in such perceptions to teacher retention, transfer, and attrition.

Information Need 5: Programs to Improve Practice. Data should also be gathered about some of the actual programs designed to improve practice. More specific data are needed about selection of and characteristics of participants in induction programs, assessment programs, and professional development schools. Detailed data are also needed on the total volume of inservice activities provided and the character and duration of these activities.

Information Need 6: Assessment of Quality of Teaching Practice. It is difficult to develop and expensive to apply direct measures of the quality of teaching practice. One feasible approach to tapping an important aspect of teaching practice would be to document teachers' beliefs about what should be taught and about their role in facilitating learning. Information about teacher beliefs and their relation to student learning could be helpful in assessing the quality of teaching practice.

Suggestions for action on the topic of teacher quality centered around its definition and measurement. Without consensus on appropriate and useful measures of teacher quality, relevant components of data bases cannot be developed, and analyses of the quality dimension within teacher supply and demand are not possible. Specific suggestions made are presented in the following paragraphs.

Suggestion 1: Teacher Quality Indicators. Sustained research should be undertaken on various indicators of teacher quality with the intent of developing consensus about valid indicators of teacher quality that can be measured and included in teacher data bases. The new indicators of teacher quality that will be generated by NBPTS and the revised NTE, as well as by efforts of states (such as Vermont) to develop new performance measures of teacher effectiveness, should be included in the NCES data systems. More generally, it was also suggested that for teachers, as for physicians and lawyers, credentials become the prime indicator of quality.

Suggestion 2: Tested Ability of Teachers. Emphasis should be placed on research on test development. Tests of pedagogical subject matter knowledge that are directly relevant to effective teaching need to be developed and used in place of present tests, which emphasize general ability and general subject matter knowledge and have little demonstrable relevance to teaching practice. Such test development should be undertaken under the auspices of specialized test development organizations rather than by NCES. However, NCES should play a role in the validation of these new assessments by adding some items from such tests to the Teacher Questionnaire for the National Assessment of Educational Progress (NAEP) so the teachers' responses could be correlated with student achievement scores.

Suggestion 3: Teaching Practice. NCES should develop items about teaching practice for inclusion in its Teacher Questionnaire for NAEP. Items might include teachers' beliefs about subject matter content that should be taught and about the role of teachers in facilitating student learning. Teacher responses to such items could be correlated with student learning outcomes.

Suggestion 4: Demographic Matching. Efforts to improve minority representation in the teaching force so as to better match the proportion of minority pupils could be enhanced by better tracking and projections of the race/ethnicity attribute of the teachers. Specifically, a race/ethnicity item should be incorporated routinely in all national surveys of teachers. As indications of potential new minority teachers in the pipeline, NCES should consider making routine projections of (a) the total pool of minority high school pupils who are eligible for college and (b) the total pool of minority college graduates.

Suggestion 5: Successful Schools of Education. Schools of education vary in the degree to which they are successful in producing teacher graduates that are actively recruited by LEAs. As an approach to understanding the nature of teacher training leading to LEA satisfaction with teaching practices, it was suggested that characteristics of successful schools of education be compared with those of schools considered to be poor in this respect.

The continuing concern of educational policy makers over the adequacy of the prospective supply of teachers in relation to demand has stimulated efforts over the years to produce data that could be used in projections of future supply-demand balance. Many of these efforts have focused on producing data bases fundamental to teacher supply and demand analyses (the topic of the following major section). Simultaneously, other efforts have focused on developing the analytic tools required to use these data to provide information needed by policy makers and others.

Prominent among these tools are teacher supply-demand projection models that have been developed to predict trends in the teacher labor market, especially to estimate imbalances or deficiencies in time for policy makers to take corrective action. An assessment of the adequacy of TSDQ projection models with respect to predicting trends in the teacher labor market is presented in the following sections. In short, useful teacher demand models—which depend on student enrollment and pupil-teacher ratio—have been developed. Little has been done. however, to create behavioral demand models that will show how the pupil-teacher ratios will change with future changes in causal factors such as economic, demographic, or fiscal conditions. At the national level, NCES has moved to develop such a model.

By contrast, teacher retention/attrition models are more advanced, and recent research findings and improved data bases can be expected to lead to further progress in the development of such models. Finally, models for projecting the supply of entering teachers from multiple sources have not been developed because of lack of data about the number of eligible individuals willing to enter the teaching profession at any given time. People enter teaching for a variety of reasons: for economic reasons, to have a schedule similar to that of their own children, to work with children, because of an outstanding teacher who served as a role model, or because of the satisfaction of helping others. Teachers also leave teaching for a variety of reasons—to engage in full-time child rearing or to follow a spouse whose job has been relocated. Although it is known that teacher supply, like the supply of skilled labor for other occupations, is sensitive to financial incentives (Freeman,1971), data are not available to model the effect of other determinants of teacher supply.

Almost all of the existing models deal only with numbers of teachers and not with teacher quality. One exception is a model (Manski, 1987) consisting of two parts: one which explains occupation choice (between teaching and nonteaching) as a function of the earnings and nonmonetary characteristics associated with alternative occupations. The second explains occupation-specific earnings as a function of academic ability and other factors. This behavioral model is of interest because, in principle, it could be used to answer policy questions about improving the average ability of the teaching force by maintaining minimum ability standards in combination with sufficient salary increases. As Manski notes, the model is dependent on the intuitive assumption that SAT score is a measure of teacher quality, although there is no research that supports this assumption.

A teacher supply-demand projection model consists of a set of mathematical relationships that can be used to estimate future levels of teacher supply and demand. Ideally, these projections are linked to future economic and educational conditions and policies. A complete model consists of three main components, or submodels, for projecting: (a) the demand for teachers, (b) the supply of teachers retained from year to year (or, equivalently, a model of teacher attrition), and (c) the supply of potential entrants into the active teaching force. The latter two components, taken together, constitute projections of total teacher supply, which (ideally) can be compared with projections of total demand. In contrast with projection models, system models provide only a cross-sectional identification and organization of major variables involved in understanding teacher supply-demand:

they do not involve other variables and equations needed to estimate future trends in supply and demand.

The types of information that can be generated with a supply-demand projection model depends on several model characteristics. One such characteristic is the extent of disaggregation of teachers by subject matter specialty and other variables. Some models deal only with teachers in the aggregate; others disaggregate in ways that make it possible to respond to detailed policy questions.

Another key characteristic is whether a teacher supply-demand projection model is mechanical or behavioral. Mechanical models estimate only what will happen in the future if established trends continue. Behavioral models, in contrast, link demand and supply estimates to changes in pertinent conditions and policies. Only behavioral models can be used to address ''what if'' questions about the effects of hypothetical changes in circumstances on teacher supply and demand.

A third critical attribute of teacher supply-demand projection models is whether and how they deal with teacher quality. Unfortunately, nearly all projection models avoid the quality dimension and focus only on numbers of teachers. Nonetheless, teacher quality is considered here because of the growing urgency of this topic.

The total national demand for public school teachers is defined by the sum of the number of teaching positions funded by all LEAs, i.e., the number of teachers that LEAs are able and willing to employ at a given time. The size of the employed teaching force is virtually the same (99 percent) as total demand (NCES, 1991b). Projections of demand less projections of the supply of retained teachers estimate the number of individuals that will have to be hired into the teaching force to replace those who leave, to adjust for changes in student enrollment and to respond to changes in educational policy such as in teacher-pupil ratio.

The concept of teacher demand as the number of funded teaching positions is an economic definition that treats teacher demand as contingent on how much teachers cost and how much money LEAs are able and willing to spend on teachers. An alternative concept of teacher demand is framed in terms of fixed requirements for teachers, such as set by predetermined teacher-pupil ratios. The significance of this difference for modeling is that the requirements notion is expressed in mechanical projection models based on teacher-pupil ratios (or trends therein); whereas the economic definition points toward behavioral models in which estimates are derived from multivariate equations linking demand to various economic and demographic

variables. Thus far, mechanical models dominate the field of teacher demand projections.

The prevailing method for projecting demand is the mechanical method of multiplying projected student enrollments by current or extrapolated teacher-pupil ratios. The simple, standard demand model of this type can be expressed as follows: (Projected number of teachers demanded in year t) = (Projected student enrollment in year t) × (Projected teacher-pupil ratio in year t). This method yields no information on how the teacher-pupil ratios themselves, and hence the numbers of teachers demanded, can be expected to change in the future in response to changes in school policy, course enrollment patterns, and economic and demographic conditions, This does not mean that such projections are necessarily inaccurate, but it minimizes their utility in policy analysis. Although there do not appear to have been any state-level efforts to create behavioral demand models, NCES has moved in the direction of such a model at the national level by shifting from a mechanical to a regression-based projection method. Although the present NCES model needs to be developed further, it represents an important step in the right direction and a point of departure for further national and state-level modeling efforts.

Models incorporating disaggregated projections of teacher demand by level of education and subject specialty have become relatively common, but the methods used to disaggregate by subject are crude. The standard approach is simply to assume that the distribution of enrollment among subjects at each level of education will remain constant.

An important implicit assumption underlying all the current demand projection models is that the teacher market has been and is in a condition of excess supply. This assumption is reasonable in the aggregate and for most subject areas, but currently it may be wrong for certain fields, such as science, mathematics, and special education. Wherever the excess-supply assumption is violated (i.e., where unmet demand exists), projections based on actual numbers of teachers employed in the past will understate the numbers of teachers likely to be wanted in the future.

Ideally, behavioral demand models should be developed in which projections are based on multivariate econometric models relating numbers of teachers demanded to economic and demographic factors such as per student expenditure, per capita fiscal capacity, student/population ratio, the makeup of the student population, and salary levels both inside and outside teaching. Such models will probably have to be constructed at a level of aggregation one rung below the level at which they are to be used; that is, state-level demand models will have to be constructed with district-level data, and a national model with state-level data. Since necessary data appear to be available nationally and for at least some states, such modeling

seems eminently feasible and could proceed if necessary resources were available.

Apparently no existing teacher demand projection model incorporates teacher quality variables. The fact that teacher salary schedules typically are scaled by experience and training indicates that qualifications matter to LEAs. These variables define dimensions of teacher quality, just as number of funded positions define demand for teacher quantity, so it is meaningful to speak of a demand for quality. Other aspects of teacher quality that warrant investigation are quality-price relationships (how much extra LEAs are willing to pay for valued attributes of teachers) and quality-quantity (the extent to which LEAs would be willing to accept lower teacher-pupil ratios to gain higher quality but more expensive teachers) tradeoffs. Conceivably these relationships might be included in future projection models, but there are formidable obstacles to designing models incorporating both teacher quantity and quality, and it is unlikely that this will be accomplished soon.

The primary source of supply of total teachers in any year is the active teaching force during the prior year. Typically, 92 to 96 percent of a state's teaching force continues from one year to the next. The remainder constitute annual attrition from the active teaching force. The simple, standard retention model can be expressed as follows: (Projected number of teachers retained from year t to year t + 1 = (Number of teachers employed in year t) × (Projected teacher retention percentage for year t). The difference between total teacher demand and retained teachers represents the number of entering teachers that need to be hired to satisfy demand fully. Hence, it is very important to be able to estimate teacher retention (or attrition) accurately because even a small percentage error in estimating teacher retention translates into a large percentage error in estimating the number of entering teachers to be hired.

At their best, state projection models for teacher retention differentiate extensively among various categories of teachers. The more sophisticated state retention models now project numbers of continuing teachers by applying age-specific attrition rates to age distributions of teachers, differentiating also by level of education and subject specialty. Nonetheless, many states still project the supply of retained teachers simply by applying statewide average attrition rates to all categories of teachers. However, recent research on teacher attrition has yielded findings that could translate into further improvements in models for projecting numbers of retainees. Among these are findings about male-female differences in attrition patterns, changes in attrition rates over time, and, perhaps most important, effects of salary levels, working conditions, and other economic factors on exit rates from

teaching. If the estimates of effects of salaries, working conditions, etc., can be refined, the present mechanical (albeit sophisticated) models could be transformed into behavioral models of teacher attrition.

Until recently, data limitations at the national level have constrained the number of separate average attrition rates used by NCES to two: rates for elementary teachers and rates for secondary teachers. National projections can be expected to improve, however, because of the availability of more accurate and detailed data from SASS and TFS questionnaires. It will soon be possible to project national as well as state teacher retention on the basis of disaggregated age-specific and subject-specific attrition rates. Research now under way may also lead to the incorporation of certain behavioral elements (e.g., responses to salary changes) into the national models.

Though it would be desirable to be able to estimate the quality of retained teachers, apparently no retention/attrition model has included teacher quality variables. However, a major concern that higher-quality teachers (to the extent that the NTE score signifies quality) are more likely to leave the profession has been borne out by some research (Murnane, Singer, and Willet, 1989; Murnane and Olsen, 1990). Although the problem of measuring teacher quality remains, it would be relatively easy to include a quality variable in retention/attrition models because there would be no need to estimate what quality would have been under hypothetical conditions.

In sum, retention/attrition projection models are the most advanced elements of teacher supply-demand models, and further progress using recent research findings can be expected, widening the sophistication gap between attrition models and those for other components.

To satisfy teacher demand fully, entering teachers usually must be hired to augment the number of teachers retained from the prior year. The four sources of supply of potential entering teachers—the reserve pool, recent graduates of teacher preparation programs, entrants via alternate routes, and private school teachers—yield the number of entering teachers that are hired annually into public education to augment the number of retained teachers. Estimates of future demand for entering teachers can be obtained by subtracting the projected number of retained teachers from the projected demand for teachers. Then comparisons between the projected numbers of entrants needed and the projected supply of entrants should lead to conclusions about the adequacy of supply and therefore the likelihood of future teacher shortages or surpluses. Although the number of teachers entering the profession annually can be measured and projected, the supply of potential entrants cannot be measured or projected, at the present time, for any of the sources of supply. Little credence should be given, therefore, to any

projections of teacher shortage or surplus based on comparisons between projections of demand and purported projections of supply.

The main problem in modeling entering teacher supply is that the number of eligible individuals willing to supply their services as teachers at a given point in time under the conditions prevailing at that time is not known with respect to any of the sources of supply. While estimates can be made, for example, of the number of experienced (but inactive) teachers and the number of recently trained teachers, no data are available about the proportion of these groups that are available to enter the teaching profession.

The lack of adequate methods, or even a sound conceptual framework, for projecting the supply of potential entrants into teaching continues to prevent the development of complete supply-demand models that can be used to assess the supply-demand balance. Many current models offer no projections at all of the supply of potential entrants into teaching. Those that do rely on one or the other of two unsatisfactory approaches. The first approach, equating the projected supply to the projected stock of persons certified to teach in a state, is clearly inadequate. Many certified persons are not in the supply (i.e., willing to teach under current conditions), and many persons in the supply may not (yet) be certified. The second approach, projecting supply by applying past entry rates to numbers of persons presumed to be in various supply pools, is invalid because past entry rates are more likely to have been determined by numbers of open teaching positions (demand) than by numbers available (supply). Most model developers have not yet come to grips with the fundamental problem that, if past hiring mainly reflects demand, then supply cannot be inferred from it. Consequently, data on employment and hiring cannot be used to measure current teacher supply nor to project future supply. Additional information is needed to estimate how many teachers would be available to take teaching jobs if hiring were not limited by demand.

Fortunately, some progress may be possible in modeling the reentry behavior of inactive experienced teachers by using essentially the same methods as have been used to model teacher attrition. These methods do not apply, however, to the other major sources of potential entrants. Although satisfactory methods of projecting the supplies of these other sources of entrants are unlikely to be developed soon, several strategies do seem worth pursuing. One is to estimate supply from information on numbers of applicants for teaching positions. Such information is collectable either from individuals or school districts. Serious problems would arise in interpreting it because a teacher may submit applications to multiple districts, but the resulting estimates, even if flawed, might be substantially better than what is now available. A related approach is to derive supply estimates from survey data on the job-seeking behavior and intentions of persons in selected groups. Although this method would be subject to all the

usual problems of using responses to hypothetical questions to infer labor market behavior, the issue is whether substantial gains could be made, compared with the present unsatisfactory situation.

We consider briefly the feasibility of introducing a quality variable into the supply model. The principal available strategy for improving teacher quality is to use a higher quality standard in hiring entering teachers, thereby gradually altering the quality of the teaching force. However, both the feasibility of improving teacher quality and the rate at which quality can be improved depend on the distribution of quality in the supply of prospective new hires. In principle, prospective entrants could be classified by quality, and separate entry rates or propensities to apply for positions could be determined for persons in the different quality strata. At present, the unsolved basic conceptual problems of supply modeling are so severe that taking the quality dimension into account is a remote prospect.

In addition to quality related topics, participants identified a number of information needs pertaining to various aspects of teacher demand, supply, and their equilibration. Conferees also suggested a number of actions that could be undertaken to improve projection models. These information needs and suggestions are reported as such below. However, they are not formal recommendations adopted by vote of conferees, and should not be interpreted as such.

Information Need 7: Subject-Specific Course Enrollment. With respect to projecting teacher demand by subject matter field, more refined information is needed about trends over time in subject-specific course enrollment rates. Methods will also have to be developed for projecting such rates.

Information Need 8: Reserve Pool. Although the reserve pool of certified teachers is the major source of entering teachers hired into public education and is probably in decline, very little is known about its size and composition. Information is needed about the current and projected size of this pool as a function of teacher quality and other characteristics. Adding a brief survey of applicants for teaching positions in the Demand and Shortage Questionnaire of the Schools and Staffing Survey would increase knowledge about characteristics of the active portion of the reserve pool and changes in this group over time.

Information Need 9: Teacher-Pupil Ratios. Information is needed about conditions under which teacher-pupil ratios change, because this ratio can serve as an equilibrating factor in supply-demand models and its an indicator of teacher workloads. More needs to be known about the impact of economic factors, changes in student demographics, and the introduction

of instructional technology on teacher-pupil ratios, as well as teacher receptivity to tradeoffs between teacher-pupil ratios and teacher salaries.

Information Need 10: Teacher Migration. Although teacher migration from one school to another is one of the major sources of supply to, and attrition from, schools, LEAs, and states, little is known about this phenomenon. Therefore, information is needed about patterns of migration as a function of teacher characteristics and attributes of schools and LEAs from which and to which teachers migrate.

Information Need 11: Teacher Support. Since support provided to teachers—especially novice teachers—is presumed to be related to the quality of teaching practice and to teacher retention, better information is needed about the usefulness of various kinds of support provided to teachers.

Information Need 12: Supply-Demand Analysis of Science and Math Teachers. In view of the importance of science and mathematics teachers to national education goals and the discrepant findings about their retention in the teaching profession, much better information is needed to assess the potential for shortages in these subject matter fields. More specifically, information is needed about sources of entering supply, migration and reassignment, attrition and reentry, and retirement for teachers in the various science disciplines and mathematics.

Information Need 13: Teacher Mobility. Since teachers have opportunities for employment in the general labor market as well as in the teaching profession, more extensive comparative data related to occupational choice are needed for teaching and other occupations. Information is also needed about factors related to occupational choices made by individuals who hold teaching credentials, whether or not currently employed as teachers.

Information Need 14: Teacher Retirement. Since the average age of teachers has gradually increased during recent decades, there is the clear prospect of accelerated attrition due to retirement. To project the magnitude of this attrition and to develop retirement policies, information is needed to project retirement rates as a function of teacher characteristics and tradeoffs between salary levels and various retirement options and benefits.

The present inadequate state of development of TSDQ projection models generated a number of suggestions about actions that could be taken to develop models more fully and about variables that need to be taken into account during the course of such development. These are identified in the paragraphs below.

Suggestion 6: Comprehensive TSDQ Model Development. A comprehensive model (or models) of teacher supply and demand needs to be developed to improve the accuracy and utility of estimates of supply-demand balances and imbalances, and to serve as a framework for determining the relevance of data to analyses of teacher supply-demand issues. In this

regard, it was also suggested that NCES work toward integrating their supply-demand model into a larger model that encompasses the entire economy, and that NCES explore the use of both cross-sectional and time-series approaches in their modeling.

Suggestion 7: Guidebook for TSDQ Model Development. Given the fairly rudimentary state of TSDQ models in most states and the many technical issues involved in model development, it was suggested that NCES or some other organization prepare a guidebook for states. The guidebook should incorporate basic TSDQ modeling concepts, practical step-by-step guidance on how to develop a functional TSDQ projection model, and guidelines for assembling data from state records in a form suitable for making supply-demand projections.

Suggestion 8: Theoretical Basis of TSDQ Modeling. Though discussion revealed that a considerable theoretical base (drawing on theories of human capital, career progression, and imperfect information in the labor market) exists for TSDQ modeling, it has not been systematically explicated. It was suggested that a statement of the theoretical underpinnings of TSDQ projection models would be useful for policy makers and others who are concerned about teacher supply-demand issues.

Suggestion 9: Teacher Demand Projections. It was suggested that certain changes in NCES teacher demand projection modeling can yield more useful estimates. As a major example, the use of pooled time-series, cross-sectional data from states for the past 10 years would permit development of a more detailed model than the current use of 30-year national-aggregate time-series data and avoid the assumption of a stable relationship between educational and fiscal variables over a 30-year period. It was also suggested that a few states might experiment with the development of econometric projection modeling similar to that used in the 1988 and 1990 NCES projections.

Suggestion 10: Teacher Retention/Attrition Projections. A number of suggestions were offered about improving projections of teacher retention/attrition. First, methods must be developed to project changes or trends in attrition rates, and then these projected rates should be used in computing projections of the size of the retained teaching force, instead of the present practice of assuming fixed attrition rates based on the rate observed in a prior year. Second, a teacher quality variable, e.g., teacher test scores, should be incorporated into teacher retention/attrition projections when feasible. Third, as more than one or two teacher characteristic variables are built into retention/attrition projection models the use of multivariate methods will be required to sort out the relationships of teacher variables with attrition rates and to estimate the effects on attrition of hypothetical or projected changes in particular teacher characteristics. Fourth, in modeling the supply of retained teachers, a clear distinction should be made between

voluntary attrition (a supply-side phenomenon that reflects the decisions of individual teachers) and involuntary attrition (a demand-side phenomenon that reflects decisions of employers). Fifth, it was suggested that modeling teacher flows at the school level should be attempted because of the many changes in assignment of retained teachers within schools and between schools. Finally, recent research on teacher retention/attrition has yielded a number of findings pertaining to attrition and reentry, such as attrition rate differences and differences in attrition patterns among categories of teachers, that have direct implications for the improvement of retention/attrition projection models and should be used to this end.

Suggestion 11: Entering-Teacher Projections. Since data are not available on the quantity and quality of individuals in supply pools of potential entering teachers, methods for modeling projections of teacher supply and entry are underdeveloped. Both relevant data bases and methods to model teacher supply and entry are needed to achieve a complete TSDQ model and to make meaningful projections of imbalances of teacher demand and supply. To this end, some useful data could be collected from applicants for teaching positions. Data could also be collected about the job-seeking behavior and intentions of persons in selected groups of the population—an approach subject to the problems inherent in inferring labor market behavior from responses to hypothetical questions.

Suggestion 12: Teacher Compensation and Working Conditions. Because variables pertaining to teacher compensation and working conditions are of particular relevance to teacher entry, retention, and attrition, it was suggested that such factors be incorporated in teacher supply-demand projection models. Suggested variables include teacher salary levels, teacher salaries relative to salaries in other occupations, and relative unemployment by occupation. In addition, various school and student variables can be used as indices of working conditions.

Suggestion 13: Differentiated Staffing. In addition to teachers, teacher aides and specialists in instructional technology are involved in providing instruction to students. It was suggested that the effect of prospective or hypothetical changes in the numbers and instructional roles of these auxiliary personnel on teacher demand and supply be examined.

The need by education policy makers and others for factual and reliable information about the teaching force, including future estimates of supply-demand relationships generated by projection models, has led to the development of teacher data bases at the state, regional, and national levels. The data must be of reasonable quality, timeliness, and accessibility, however, to be useful for projection models and decision making.

Fortunately, there are now several data bases that will support rather sophisticated analyses of a number of important aspects of teacher supply and demand. At the national level, teacher surveys are particularly rich sources of data about many supply-demand variables. A brief description and assessment of teacher data bases is presented here.

Much information about teachers practicing in a state is contained in state administrative records. Typically, relevant information is contained in different sets of records such as those pertaining to teacher certification (or licensure) and to retirement, each of which can be used separately for study of some aspect of teacher supply and demand.

The use of state data for investigating teacher supply and demand is especially promising because of the wealth of detailed, longitudinal information typically residing in state records. In practice, however, these records vary considerably in the breadth of teacher supply-demand data maintained and in the degree to which teacher supply-demand variables are linked to individual teachers. Though various state records often can be linked to create integrated or comprehensive teacher data bases, states do not ordinarily maintain such integrated data bases incorporating longitudinal information by teacher. Instead, the data must be assembled at considerable effort and cost by researchers interested in studying teacher supply-demand phenomena at the state level. In fact, it is common for researchers to spend several years creating usable data bases from state administrative records before beginning to conduct analyses of teacher supply-demand issues.

A number of variables relevant to teacher quality are contained in state records. For the most part, these are variables pertaining to teacher qualifications. Commonly (though not uniformly) found in state records is information about teachers' educational attainment, academic major, field of certification, type of certification, and years of teaching experience. A few states also record teachers' tested ability scores.

The quality of state records raises several major concerns about the use of data from these sources for studying teacher supply-demand issues. One concern is about incomplete data. Problems include incomplete reporting of data by teachers and LEAs and periodic purging of state records. Another concern is the inaccuracy of recorded data. Inputs from teachers and LEAs may contain errors, and state data management procedures can be inadequate. A third concern is inadequate documentation of the recorded state data, an important consideration for researchers attempting data analyses. The comparability of data among states is a serious issue in making cross-state comparisons. For example, a major incompatibility of data among states arises from their different categories and requirements for teacher

certification. Finally, the timeliness of data is of concern. All data files used should be reasonably current if analyses based on them are to be of maximum value to policy makers and others.

Nonetheless, the difficulties inherent in using state data for teacher supply-demand analyses can be managed, and experience has demonstrated that there are substantial benefits to be derived from research with teacher data bases. These sources of data, at their best, contain a remarkable degree of detailed, longitudinal information about teachers, students, and schools. In recent years, several research programs (e.g., Richard Murnane and associates from the Harvard Graduate School of Education; David Grissmer and associates from RAND; James Wilson and associates from the Massachusetts Institute for Social and Economic Research) have used state data successfully to investigate major supply-demand issues such as trends in teacher licensure, entering teachers, attrition, reentry, and retirement as a function of teacher characteristics such as age, gender, race/ethnicity, subject matter specialty, and tested ability. The long time-series on individual teachers available from state records is particularly valuable in studying the course of teacher career development over time and the changing trends in how such careers develop.

The only regional teacher supply-demand data base (the Northeast Regional Data Base) currently operational has been developed for New York and the New England states by the Massachusetts Institute for Social and Economic Research (MISER).⁵ In cooperation with these states, MISER has taken the lead in assembling useful data bases for seven states by extracting information available from multiple sources within each of the seven states in the Northeast region. However, MISER has not generated original teacher data through sample surveys or other means.

With funding from federal, state, and private sources, MISER worked closely with the seven states of the Northeastern region to improve and to expand the collection and management of teacher data within each state and has begun to create linkages among the emerging state data bases for analytic purposes. The development of these data bases goes beyond what the cooperating states would have implemented on their own initiatives.

Since the Northeast Regional Data Base is comprised of seven state data bases constituted from administrative records within each state, the virtues and limitations of state data bases described in the prior section likewise apply to this regional data base. Being a collection of data bases from all states of a region, however, it has the special advantage of permitting study of reciprocal cross-state migration of teachers within the region. In addition, cross-state comparisons of teacher data has gained the attention

of chief state school officers and has led to the emergence of policy changes such as regionalizing teacher credentialing.

This regionalized teacher supply-demand data base is a new phenomenon, which has generated a great deal of favorable attention. In fact, the Southern Regional Education Board is developing a comparable data base using the MISER model. Although the development of the MISER data base has been limited by the availability and quality of information in state records and by the availability of funding to develop and sustain the MISER data set, it has already produced some useful findings and benefits for the region. MISER plans to continue testing the feasibility and utility of a regional data base for the Northeast.

Several data bases useful for analyses of TSDQ issues now exist at the national level. With one or two exceptions, none of these are derived from data collected and reported by state or local education agencies. Instead, most national data relevant to TSDQ have been generated by sample surveys with questionnaires—the prime example being the Schools and Staffing Survey (SASS) and its longitudinal component the Teacher Followup Survey (TFS), conducted by NCES. SASS was administered during the 1987–88 and 1990–91 school years and is scheduled to be conducted every three years thereafter.

SASS was designed to collect extensive information especially relevant to teacher supply, demand, and quality, as well as the status of schooling generally. For schools in the public sector, SASS includes separate questionnaires designed to obtain information from teachers, from principals of schools to which the teachers are assigned, from schools to which the teachers are assigned, and from LEAs sponsoring these schools. With this design, teacher, principal, school, and LEA data are linked. TFS is administered in years following each SASS to three groups of teachers included in the prior SASS sample: (a) teachers who continued teaching in the same school; (b) teachers who continued in the profession but transferred to a different school; and (c) teachers who left the teaching profession, thereby permitting detailed study of teacher mobility and attrition.

More specifically, data generated by SASS and TFS serve the following five purposes: (a) to profile the nation's teaching force; (b) to improve estimates and projections of teacher supply and demand by teaching field, sector, level, and geographic location; (c) to allow analyses of teacher mobility and turnover; (d) to enhance assessment of teacher quality and qualifications; and (e) to provide more complete information on school policies and programs, administrator characteristics, and working conditions. Ac-

cordingly, SASS includes a great deal of information about representative samples of teachers from each state and from the nation as a whole.

NCES has established several other national data bases that provide data relevant to teacher supply and demand topics. Among them are the Common Core of Data, High School and Beyond, the National Longitudinal Study, the National Education Longitudinal Study (NELS:88), Surveys of Recent College Graduates, and the National Assessment of Education Progress (NAEP). Survey data from these sources and SASS typically include information about teacher qualifications. The Longitudinal Study of American Youth sponsored by the National Science Foundation (NSF) provides an opportunity to study the relationship between teacher qualification variables and student academic achievement. This is also possible with NELS:88 and NAEP. A complementary component of NELS:88 was the NSF sponsored Teacher Transcript Study (Chaney, 1991), which obtained transcripts of the academic records of science and mathematics teachers within the NELS:88 study. The analysis for this study focused on teachers' backgrounds in terms of their own college-level preparation in science and mathematics as a prediction of their students' achievement.

Several national data bases relevant to teacher supply and demand have been sponsored by organizations other than NCES. Two examples are the National Surveys of Science and Mathematics Education, sponsored by NSF, and the Status of the American Public School Teacher, sponsored by the National Education Association, both of which include extensive information about national samples of teachers. In addition, tested ability measures of prospective teachers and practicing teachers can be found in data bases of the Graduate Record Examination and the Scholastic Aptitude Test.

There are major differences between state and national data bases, as well as complementary strengths. In contrast with data from national sample surveys, state records contain a wealth of detailed data about populations of teachers, students, and schools. These records are maintained year by year, thereby permitting examination of trends over time and permitting cohort studies of teachers as their careers develop. In contrast, national surveys are usually based on samples of these populations, and limited sample sizes often do not permit detailed analyses of many variables and their cross-tabulations.

Unfortunately, state data bases are beset by a number of limitations. One is that teachers from any particular state, or from a combination of a few states, do not constitute a representative sample of the national teaching force. Furthermore, variations among states in definitions of variables and in policies pertaining to teachers limit the generalizability of state-level

findings. Another limitation of state data bases is their inability to distinguish between teachers that leave the teaching profession or simply migrate to out-of-state schools. Without this information, attrition studies lack precision and may yield misleading interpretations.

For the most part, the strengths of national data bases relevant to TSDQ are the limitations of state data bases, while the limitations of national data bases are the strengths of those at the state level. At their best, national data bases, such as SASS, include a great deal of information about probability samples of teachers for each state and for the nation as a whole, thereby providing (a) a national perspective of the teaching force, (b) opportunity for state-by-state comparisons, and (c) a basis for analysis of teacher migration among all states. In addition, the surveys on which they are based define variables uniformly across all units sampled and use a standardized procedure, thereby permitting interpretations of data that are generalizable nationally.

National data bases, however, are as yet very limited in the extent to which they can support time-series analyses for nationally representative samples of teachers. In addition, practical limitations on the length of survey questionnaires do not permit the level of detailed information gathering that is often possible with administrative records from which state data bases are derived. Finally, resources available for national surveys limit sample sizes, thereby restricting many important analyses.

Since the best currently available state and national data bases are complementary in their respective patterns of strengths and limitations, there is great advantage not only in having both types available, but also in being able to link data bases at these two levels. From the perspective of NCES, there is considerable potential for linking SASS with state data bases, and it is important to work toward achieving this dual strategy.

Major advances in the development of data bases relevant to TSDQ projections and research have been made in recent years at the national, regional, and state levels. Nonetheless during the course of open discussions occurring throughout the conference, participants suggested many further steps that could be taken to strengthen and expand such data bases. These suggestions are reported as such below.

Suggestion 14: State Record Management. Since state administrative records are the information source for state data bases relevant to TSDQ, it is important that these records be maintained so as to be suitable for TSDQ projections and research. Three related suggestions were made about management of these records: (a) that states design and maintain pertinent state records in ways that support TSDQ research as well as administrative

functions; (b) that the experience of recent researchers with state data bases be used to improve state record management; and (c) that guidelines for state data management be developed to assist states in responding to this need.

Suggestion 15: Funding State Data Base Development. Since most states do not have sufficient resources to produce high-quality data bases of the type described in Suggestion 11, it was suggested that NCES, or some other federal agency, provide states with capacity-building grants for this purpose.

Suggestion 16: Teacher Demand Data. NCES should develop a more elaborate behavioral model for teacher demand projections that includes economic and fiscal variables, especially at the state and local levels.

Suggestion 17: Longitudinal Supplements to SASS. Although SASS has a longitudinal supplement for teachers in the TFS, it was suggested that NCES give high priority to sustaining this component over time, that TFS content on teacher attitudes toward teaching and teaching styles be increased, and that ideally observations of direct classroom teaching be made in addition to using survey questionnaires.

Suggestion 18: Unused National Data Bases. There are a number of national data bases (e.g., the Longitudinal Study of American Youth, the Scholastic Aptitude Test, and the Graduate Record Examination) relevant to analyses of teacher quality that have not been used for this purpose. A suggestion was made that these data be used to study the quality of teacher supply starting in high school at the beginning of the pipeline.

Suggestion 19: Linking SASS and State Data Bases. Inasmuch as data bases derived from state administrative records and from national sample surveys with questionnaires are both extremely valuable for teacher supply-demand analyses but in distinctive and complementary ways, it was suggested that steps be taken to link SASS data with state data to enhance the range and depth of possible analyses.

Suggestion 20: Linking SASS with Work Force Data Bases. The teaching force is a component of the larger national labor force in which eligible teachers compete for jobs in a wide variety of occupations. Similarly, a very large, but unknown, number of educated adults can compete for jobs in the teaching profession and enter through widely available alternative routes, even though they have not undertaken teacher training. Consequently, it was suggested SASS be linked to larger work force data bases so that the transfer of workers among occupations (including teaching) can be analyzed, thereby providing a broader understanding of factors involved in teacher supply and shortage.

Suggestion 21: International Education Data. It was suggested that data from the United States included in international assessments of education should be used more extensively, especially data relating teacher characteristics to student outcomes.

In addition to the suggestions for action related to teacher quality, projection models, and data bases as described above, conference participants made the following suggestions pertaining to broader aspects of teacher supply-demand projections and research:

Suggestion 22: On-Line Classroom Computers. It is now technically possible to connect NCES with a national sample of teachers electronically through on-line classroom microcomputers. It was suggested that NCES explore the possibility and utility of having a standing national sample of teachers who routinely transmit data on classroom events to a central repository so that teacher behavior and student learning can be observed as it happens.

Suggestion 23: TSDQ Research Consortium. The value of SASS and other national data bases for analyzing teacher supply and demand issues is much greater than can be exploited by NCES and contract analysts. Accordingly, it was suggested that NCES create and fund a consortium of independent TSDQ researchers in order to capitalize more fully on the potential of these data bases.

Suggestion 24: Annual TSDQ Conference. As a means to stimulate the development and analysis of state data bases relevant to TSDQ, it was suggested that an annual TSDQ conference be held for key state representatives and TSDQ model developers and researchers.

Suggestion 25: Information Useful in Decision Making. Since the major purpose of generating information about TSDQ is to provide a factual basis for policy and administrative decisions in this area, TSDQ data base managers, model developers, and researchers need to understand better (a) the nature of current and emerging policy issues and (b) the attributes of information of practical value to education policy makers and others. NCES could convene focus groups on these topics periodically. A better understanding of information utilization is important as guidance for the production of technically sound but useful TSDQ data and analyses that minimize the risk of misinterpretation by consumers.

American Association of Colleges of Teacher Education 1987 Minority Teacher Recruitment and Retention: A Public Policy Issue. Washington, D.C.: author.

Bobbitt, Sharon A., Elizabeth Faupel, and Shelley Burns 1991 Characteristics of Stayers, Movers, and Leavers: Results from the Teacher Followup Survey, 1988–89. June. NCES 91–128. Washington, D.C.: National Center for Education Statistics.

Chaney, Bradford 1991 Teachers' Academic Preparation and its Connection to Teaching Methods and

Student Outcomes in Science and Mathematics: NSF/NELS:88 Teacher Transcript Study. December. Prepared for the National Science Foundation, Washington, D.C.

Feistritzer, Emily 1990 Alternative Teacher Certification: A State-by-State Analysis. Washington, D.C.: National Center for Educational Information.

Freeman, Richard B. 1971 The Market for College-Trained Manpower. Cambridge, Massachusetts: Harvard University Press.

Gilford, Dorothy M., and Ellen Tenenbaum, eds. 1990 Precollege Science and Mathematics Teachers: Monitoring Supply, Demand, and Quality. Panel on Statistics on Supply and Demand for Precollege Science and Mathematics Teachers, Committee on National Statistics. Washington, D.C.:National Academy Press.

Grissmer, David W., and Sheila Nataraj Kirby 1991 Patterns of Attrition Among Indiana Teachers, 1965–1987. Santa Monica, California: The RAND Corporation.

Hammer, Charles H., and Elizabeth Gerald 1990 Selected Characteristics of Public and Private School Teachers: 1987–88. July. NCES 90-087. Washington. D.C.: National Center for Education Statistics.

Hanushek, Eric 1986 The economics of schooling: Production and efficiency in public schools. Journal of Economic Literature 26:1141-1177.

Kirby, S.N., D.W. Grissmer, and L. Hudson 1991 Entering and Reentering Teachers in Indiana: Source of Supply. R-4049-LE. October. The RAND Corporation.

Lanier, Judith 1986 Research on teacher education. Pp. 527–569 in M.C. Wittrock, ed., Handbook of Research on Teacher Education, 3rd edition. New York: Macmillan.

Manski, Charles F. 1987 Academic ability, earnings, and the decision to become a teacher: Evidence from the National Longitudinal Study of the High School Class of 1972. In David A. Wise, ed., Public Sector Payrolls. Chicago: University of Chicago Press.

Metz, A. Stafford, and H.L. Fleischman 1974 Teacher Turnover in Public Schools, Fall 1968 to Fall 1969. DHEW Publication No. (OE) 74-1115. Washington, D.C.: U.S. Department of Health, Education, and Welfare.

Murnane, Richard J., and Randall J. Olsen 1989 The effects of salaries and opportunity costs on duration in teaching: Evidence from Michigan. The Review of Economics and Statistics 71:347-352. 1990 The effects of salaries and opportunity costs on length of stay in teaching. Journal of Human Resources 25(Winter): 106-124.

Murnane, Richard J., and Michael Schwinden 1989 Race, gender, and opportunity: Supply and demand for new teachers in North Carolina, 1975–1985. Educational Evaluation and Policy Analysis 11 (2)(Summer):93-108.

Murnane, Richard J., Judith D. Singer, and John B. Willett 1989 The influences of salaries and ''opportunity costs'' on teachers' career choices: Evidence from North Carolina. Harvard Educational Review 59(3)(August):325-46.

National Center for Education Statistics 1991a Digest of Education Statistics 1990. Washington, D.C.: U.S. Department of Education.

1991b E.D. TABS: Aspects of Teacher Supply and Demand in Public School Districts and Private Schools: 1987–88. August. Charles H. Hammer, and Elizabeth Gerald. NCES 91-133. Washington, D.C.: U.S. Department of Education, Office of Educational Research and Improvement.

Oakes, Jeannie 1990 Multiplying Inequalities: The Effects of Race, Social Class, and Tracking on Opportunities to Learn Mathematics and Science. July. R-3928-NSF. With Tor Ormseth, Robert Bell, and Patricia Camp. Santa Monica, California: RAND.

Rollefson, Mary R. 1992 Sources of Newly Hired Teachers in the U.S.: Results of the 1987–88 Schools and Staffing Survey. Paper presented at the NCES Annual Data Conference, Washington, D.C. July. National Center for Education Statistics, Washington, D.C.

Schmidt, Peter 1992 Shortage of trained bilingual teachers is focus of both concern and attention. Education Week February 12:14.