1. Introduction

Much has been made of the impending shortage of school leaders. Analysis of the substantial increase in demand due to the retirement of the baby-boom generation of school leaders suggest as many as 50 percent will have left within the next six years. Meanwhile there is increasing evidence that schools are much more demanding places in which to work and lead others. Heightened standards, visibility and school accountability all make the jobs of school leaders more difficult, leading some to believe that the supply of potential administrators will be reduced. Anecdotal reports of small applicant pools for school leadership openings gives some credence to the shortages that might result from increased demand and reduced supply of school leaders. However, there has been very little systematic evidence on which to base these assertions and even less research as to how policymakers might respond.

Evidence from three recent studies suggests that there may be many more individuals certified to be school administrators than will be required to meet the increased demand (Gates et. al., 2002; Hill et. al., 2002 and Papa, Lankford and Wyckoff, 2002). Another recent study indicates that most of those certified for administrative positions are willing to serve under the right conditions, even though over half have stopped looking for an administrative position (Lankford, O’Connell and Wyckoff, 2003)

Ultimately, we believe the issue of recruiting and retaining effective school leaders is much more nuanced than a simple story of numbers and shortages. Rather, it involves understanding and identifying the attributes of effective school leadership across very diverse environments, as well as how to attract effective leaders to the schools where student performance is worst. Such an analysis is well beyond the scope of this research or any large-scale analysis of which we are aware. Our goals are much more modest.

This paper examines the issue of recruiting school leaders from the perspective that we need a far better understanding the decisions that individuals make to put themselves in a position to apply for school leadership positions and ultimately to take those positions. Which teachers are most likely to be interested in administrative certification? How has that changed over time? Of those having the required certification, which are most likely to take administrative positions? By better understanding these questions we believe that educational policymakers will be in a better position to understand the depth of the so-called administrative shortage and how to begin to address it.

2. Data and Methods

Our database links several administrative datasets, including information for every teacher and administrator employed in a New York public school at any time from 1983-84 through 2001-02. The core data comes from the Personnel Master File (PMF), part of the Basic Education Data System of the New York State Education Department. We have linked these annual records through time, yielding detailed data characterizing the career history of each individual. These data are linked with information from the New York State Teacher Certification database concerning whether individuals are provisionally certified to serve as principals or in other administrative positions, dates when such certification was awarded as well as the academic backgrounds of those individuals. The information regarding the institutions from which individuals earned degrees was matched with Barron’s rankings characterizing the selectivity of those institutions. Together these data contain a range of information characterizing prospective and current school leaders.[1]

We are interested in what factors are important when an individual transitions among various career options, including movement from teaching to lower or higher-level administration and leaving the public school system as well as movement from lower level administration to either higher level administration, teaching or leaving the public school system. Unlike the decision to become certified as an administrator, many of these career transitions (e.g., becoming a principal) are two-sided decisions—an individual must want to assume an administrative position and an employer must choose that applicant. Modeling this matching process is complex.[2] Rather than attempt to estimate such a model, we have chosen to employ a multinomial-logit model to examine transitions across an individual’s job status. Technically, the model is a discrete-time, competing-risk hazard model which, in the case of teachers having the certification needed to work as a principal, models the probabilities of the individual switching to lower-level administration, moving into higher-level administration (e.g., become a principal) and leaving the NYS workforce. The model can be viewed as a reduced-form analysis of the variables that affect an individual’s decision to move across job states, e.g., move from teaching to lower level administration. Such an analysis ideally would include individual, school and district variables and would allow us to examine the effect of changes in variables under the control of policymakers, e.g., the effect of an increase in the salaries of principals on the likelihood of individuals becoming certified for administrative positions and their willingness to move from lower-level administration to the principalship.

Developing the data for such a model is very demanding in two important respects. First, to correctly identify the effect of a particular school or district variable, e.g., salary, requires that all other relevant school and district variables be included in the estimation. Otherwise, we would incorrectly attribute the effects of the omitted variables to those variables included, to the extent these variables are correlated. Second, collecting data for correctly specifying variables such as salary in the context of school leaders is very difficult. Unlike teachers who are ostensibly paid according to a salary matrix, few districts have such rote procedures for compensating school leaders. Appropriately, the compensation of school leaders is subject to substantial discretion. Thus, trying to measure what a school leader would be paid in other districts becomes very difficult in general and virtually impossible with available public data. We have attempted to construct such variables and will continue to do so, but to date we are not satisfied that we have adequately captured the underlying concepts sufficiently well to include them in our models.

Ultimately, we would like to estimate a fully specified model of job transition that accounts for individual, school and district variables. We view the current paper focusing on the effects of the attributes of potential leaders as a first, important step towards such a more complete model.

3. Who Becomes Certified as an Administrator?

An individual’s decision to become provisionally certified as an administrator likely depends on a number of factors, some of which are readily observable such as gender, education, and experience. Others are much more difficult to observe, e.g., career aspirations, but possibly can be proxied using some of the observable attributes. We believe an understanding of how the likelihood of becoming certifiied varies with individuals’ attributes is important for a couple of reasons. First, certification is required of school leaders. By only examining who assumes leadership positions several different decisions lumped together. Because individuals choose to be certified, modeling that process can provide insights regarding who is attracted to administrative positions.

Our analysis starts with the full New York State education workforce database made up of individual-year records where a record reflects a particular person working in the NYS system during a particular school year, 1990-2002. Based on information from the Personnel Master File, the data typically reflects an individual’s place of employment and work assignments at the start of the school year. An individual’s record for a particular year was dropped from the database if the person already was provisional or permanently certification to be a principal/superintendent or was employed in such positions. For the individual-year records remaining, a variable was added indicating whether the individual over the next 12 months gained provisional certification to be a school principal. Using this data, we are able to analyze how the likelihoods of individuals subsequently gaining provisional certification vary with a range of individual attributes, holding other factors constant.

Administrative certification experienced a long and substantial decline from 1987 through 1997 (Figure 1). From its peak in 1987, when 1600 individuals received administrative certification, to its trough in 1997, the number of individuals receiving administrative certification declined by more than a third. In 1997 that trend began to dramatically reverse itself until by 2001 there were 1600 new administrative certifications.

Those obtaining administrative certification have become a much more diverse group over time. In 1984, 12 percent of those receiving provisional certification were Black or Hispanic. By 2002, this group accounted for 25 percent of newly certified administrators (Figure 2). In 2002, approximately 65 percent of those receiving administrative certification were female, an increase from 1984 when females accounted for about 58 percent of the newly certified administrators (Figure 3). Over the last ten years the average experience of newly certified administrators has fallen by nearly 25 percent (Figure 4). Nonetheless, on average newly certified individuals have more than 10 years of total experience. The decrease in experience mirrors a reduction in the age of individuals receiving certification. This decrease is not unexpected as the experience composition of the teacher workforce has declined in recent years as the increasing number of retiring baby boomers are replaced by novice teachers. In summary, the pool of certified candidates from which administrators are being drawn has become substantially more racially and ethnically diverse, more female and somewhat less experienced.

Estimating the demand for administrative certification. We now turn from a summary of descriptive statistics to an analysis of how various factors affect the likelihood of individuals obtaining administrative certification. Variables included in the estimated model are shown in Table 1, with some requiring further explanation. The variable female equals zero or one depending upon whether the individual is male of female. The degree variables bachelors degree, masters degree and more than masters equal one if the individual has no more than a bachelors degree, has a masters degree and has at least an MA+30, respectively, with the variables equal to zero otherwise. Each Barron’s variable equals one if the institution awarding the individual’s first bachelor’s degree was ranked by Barrons as indicated, zero otherwise. The job assignment variables assistant principal through science equal an individual’s FTE in each job category. The regional dummy variables, New York City through North Country, equal one if an individual is employed in that region and zero otherwise. The urban dummy variable equals one if an individual is employed in an urban district other than New York City. Finally, dummy variables are included indicating the year for each person-year record.

Employing these data, we estimated a Logit specification to analyze how the likelihood of obtaining provisional certification to serve as a principal varies with each of the variables, holding the other variables constant.[3] Using the parameter estimates for the model, shown Table 2, it is possible to consider a person having a particular set of attributes and working in a particular location and calculate the estimated probability that he or she over the next year would obtain provisional certification to serve as an administrator. For example, evaluating all the variables for an individual at the sample means shown in Table 1, such an individual is estimate to have a 0.0055 probability of gaining provisional certification over the next year. In turn, we can investigate how the probability varies with changes in the explanatory variables. For example, evaluating all the other variables at their sample means, the estimated probabilities for females and males are 0.0051 and 0.0067, respective, as shown in Figure 5. Other things equal[4], the estimated probability for a male is over 30 percent higher than that for a female.

Estimates of the model present a somewhat different picture of who becomes certified as an administrator than the descriptive statistics presented earlier. Males are about 30 percent more likely to choose to be certified than females, all else held constant (Figure 5). Even if women apply, are made offers and accept administrative positions with equal likelihood as their male counterparts, their choosing to obtain administrative certification at a lower rate would result in them being underrepresented as administrators compared to men having otherwise similar attributes,. Those with a masters degree and 30 credit hours or a Ph.D. are more than twice as likely to be certified as are those with only a masters degree, all else held constant.

Surprisingly, individuals who attended less highly ranked undergraduate colleges are more likely to seek administrative certification than those from more competitive colleges, other things held constant (Figure 6). An individual who graduated from the lowest ranked colleges are more than 10 percent more likely to obtain administrative certification than someone who graduated from a college in the most competitive rank. This is not a large effect but does suggest, based on this one measure of qualifications, that districts are not attracting the most able individuals into the pool of certified candidates.

The role of age and experience in the certification process are interesting. As depicted in Figure 7, the likelihood of being certified decreases steadily as age increases, holding all else, including experience, constant. However, it is more useful to examine the likelihood of an individual gaining certification as age and experience increase together. The figure shows an individual who begins teaching at age 25 and follows her over the next 35 years, increasing her age as well as district and total experience. The likelihood of teachers and other employees gaining administrative certification increases as experience increases over the first seven years, but then begins to decline. The probability of obtaining administrative certification is more than twice as high in the seventh year of experience as it was in the first year or will be 17^th year. This suggests that recruiting people to begin preparing for an administrative career is likely to be most effective relatively early in their careers.

As shown in Figure 1, over the last two decades there have been marked changes in the annual number of individuals awarded provisional certification to serve as principals. This could be explained by a combination of factors. First, with some types of individuals having higher likelihood of gaining certification (e.g., males and those having 5-10 years of prior experience) than other types, changes in the demographic composition of the NYS workforce could change the overall average propensity. Second, holding the demographic composition of those lacking principal certification constant, it is possible that the likelihood of individuals going on to gain principal certification has changed over time, due to other factors. For example, an overall increase in the number of first-time principals being hired could lead to an increase in the propensity for other individuals to gain certification, as they observe increasing opportunities to move into administrative positions. Third, even if the average propensity in the NYS workforce were unchanged, a change in the size of the workforce would likely lead to a change in the total number of individuals awarded principal certification.

Figure 8 was constructed to explore the importance of the first two possibilities – changes in the demographics of the NYS workforce and changes in the likelihood of individuals gaining certification holding demographics constant. The solid line shows the proportion of the workforce in each year that actually went on to gain provisional administrative certification over the next year. For example, of those in the New York State workforce and not having principal certification at the start of the 1986-87 school year, 7.5 in a thousand (0.00751) went on to gain certification that year. We see that the proportion gaining certification reached its maximum that year and then declined for a decade, followed by a partial rebound in recent years.[5] Again, these changes reflect a combination of demographic changes in the workforce and changes in the propensity to become certified holding demographics constant. To isolate the importance of each of these possible explanations, we estimated a model similar to that shown in Table 2 except that the years covered were extended to include data for the years 1983-84 onward. Those individuals in the NYS workforce for years 1985-86, 1990-91 and 2000-01 were then considered separately. The top dashed line in Figure 8 is for the 1985-86 workforce and shows the average estimated probability of individuals gaining principal certification for this group, allowing for how certification probabilities were estimated to vary over time, through the year effects. For example, for the 1985-86 workforce facing the environment as it existed in that year, we estimate that 0.729 percent of the individuals would have gone on to gain certification within a year. In contrast, in the environment as it existed in 1997-98 we estimate that 0.632 percent of those same individuals would have gained certification, a meaningful 15 percent difference.

The other dashed lines reflect average probabilities for the 1990-91 and 2000-01 workforces. The vertical differences (e.g., shifts) in the three dashed lines reflect the estimated effects of changes in the composition of the workforce. For example, as noted above, in the counterfactual situation where the 1985-86 workforce faced the environment as it existed in 1997-98, we estimate that 0.632 percent of the workers would have gained certification over the following year. In contrast, considering the workforces for 1991 and 2001 facing the 1997-98 environment, we estimate that the percentage of the workforce gaining certification would be 0.532 and 0.469, respectively.

Reconsidering the solid line in Figure 8, we see that the proportion of the workforce actually gaining provisional administrative certification in the next year fell from 0.751 percent in 1986-87 to 0.498 percent in 1997-98. As the dashed lines show, demographic changes and a decline in the likelihood of gaining certification, holding demographics constant, both contributed to the overall decline. Calculations indicate that roughly 60 percent of the change from 0.751 to 0.498 was a result of the change in the composition of the workforce.

Since 1998 there has been an increase in the proportion of the workforce going on the gain provisional principal certification (0.498 to 0.603 in 2000-01). We estimate that this is completely the result of an increase in the likelihood of gaining certification holding demographics constant. We view this as clear evidence that a meaningful number of individuals who previously would not have sought such certification are now doing so – likely as a result of increasing recruitment efforts and more job opportunities.

4. Transitions to Administration Positions

We now turn to an analysis of which of those certified to be school leaders actually take up such positions. Employing the full NYS workforce database, we identified those individuals having the certification required to serve as a principal or superintendent but were not so employed. For these individual-year records a variable was created indicating whether each individual remained in the NYS workforce the following year and, if so, whether the person worked as a teacher, lower-level administrator or higher-level administrator (principal or superintendent). Splitting the observations by whether the individuals currently were teachers or lower-level administrator, two separate models were estimated – one for the teacher sub-sample and a second for the lower-level administrator sub-sample. In each case the model allowed for four outcomes for the following year. For example, a teacher could (1) remain in teaching, (2) switch to lower-level administration, (3) switch to higher-level administration or (4) leave the NYS workforce. Similarly, a lower-level administrator could (1) remain a lower-level administrator, (2) switch to teaching, (3) switch to higher-level administration or (4) leave the NYS workforce.

Transition to leadership positions has changed rather dramatically in recent years, especially among those recently certified for administrative positions. Figure 9 follows cohorts of teachers who became certified for administrative positions through their careers following certification to examine the percentage of them who became principals or superintendents. For example, the line labeled 1994-96 is for those teachers who gained certification for administration in those years and shows the percentage of this cohort remaining in the NYS workforce that went on to obtain positions as principals or superintendents. Four years after certification, about 10 percent of that cohort had become a principal or superintendent. As is evident, the 1994-96 and 1999-01 cohorts are much more likely to have assumed senior leadership positions than their colleagues from earlier cohorts after a comparable period of time. For example, seven percent of the members of the 1999-01 cohort had become principals or superintendents within two years of certification. It took the cohort of 1989-91 a little over five years for a comparable proportion to have taken up principal and superintendent positions. A very similar picture emerges when we examine the transition for those who occupy lower level administrative positions into higher-level administration (Figure 10). Individuals in more recent cohorts are much more likely to assume principal and superintendent positions than earlier cohorts. There may be several reasons for the increased movement to higher administrative positions by recent cohorts, including the increased demand for administrators to fill the vacancies created by those retiring. Whatever the reason, the increased rate of transition to higher-level administration will likely send signals to others interested in administration that there is increased mobility. This is likely one of the reasons behind the dramatic increase in administrative certification that has occurred since 1997 (Figure 1) and the increasing propensity for administrative certification holding demographic variables constant (Figure 8).

Estimates of a Model of Leadership Transitions.

We estimated two separate models – one for the teacher sub-sample and a second for the lower-level administrator sub-sample. In each case the model allowed for four outcomes for the following year: serve as a teacher, a lower-level administrator, a higher level administrator or leave the New York State public school system. Variables included in the multivariate analysis of job transitions by teachers having provisional certification to serve as a principal, and their descriptive statistics, are shown in Table 3. Each of the variables included in the analysis of the incidence of certification is included here as well. In addition, the dummy variable prior permanent certification equals one if the individual already had permanent certification in some other area and the variable years since permanent certification measures the number of years since that certification was received. The descriptive statistics for the variables included in the analysis for individuals working as lower-level administrators are shown in Table 5.

Formally, the estimated models employed for teachers and lower-level administrators shown in Table 4 and Table 6, respectively, are discrete-time, competing-risk hazard models. For example, in the case of teachers having the certification needed to work as a principal, the model allows us to estimate the probabilities of the competing alternative outcomes – switching to lower-level administration, taking up higher-level administration or leaving the workforce. (Remain in teaching is the status-quo.) Similarly, the model for lower-level administrators allows us to calculate the probabilities that such an individual switches to teaching or higher-level administration or leaves the workforce. (Remain in lower-level administration is the status-quo.) Note that these are annual transition probabilities. Even if the probability that an individual becomes a principal in a particular year is low, the probability that an individual will make the transition in a given number of, say five, years will be larger.

The parameter estimates for the two models are shown in Tables 4 and 6. Using these parameter estimates, it is possible to compute the estimated probability that a person having a particular set of attributes and working in a particular location would make each of the transitions. In turn, we can investigate how changes in each of the explanatory variables, individually or in combination, lead to changes in the estimated probabilities. Clearly many of these attributes may affect both an employee’s interest in becoming an administrator and an employer’s interest in hiring that individual. In this sense the estimates presented here are reduced form estimates reflecting the likelihood that a person with a particular set of attributes becomes a principal or superintendent such that we cannot disentangle the extent to which variables are affecting the behavior of employees or that of employers, or both.

Teachers are less likely to become lower-level administrators as their educational attainment increases, although educational attainment has relatively little impact on the transition from teaching to higher administrative positions, holding other things constant (Figure 11A). Among lower-level administrators, the movement to higher administrative positions is not systematically affected by educational attainment.

The role of gender in these transitions is often as expected. Males are more likely to become lower-level administrators, but among lower-level administrators there is little difference in who moves on to higher-level positions, all else held constant (Figures 12A and 12B). Likewise, there is no difference between males and females in direct transitions from teaching to higher-level administration.

The ranking of the undergraduate college from which the individual graduated has no meaningful effect on transitions for either teachers or lower-level administrators.

Paths to administrative positions vary greatly by geographic regions as shown in Figures 13 to 14. For example, it is more common for teachers to transition to lower level administrative positions in metropolitan areas (e.g., New York City or Rochester) than rural areas (e.g., Southern Tier or North Country). By contrast, it is much more common for teachers to move directly to higher administrative position in rural areas (e.g., Southern Tier or North Country) than from largely urban regions (e.g., New York City or Rochester).

Time since administrative certification is a very important determinant of an individual’s likelihood of transitioning to the next level of administration. For example, teachers who receive administrative certification are twice as likely to assume a lower-level administrative position immediately following certification as they are four years after certification (Figure 15A). Similarly, teachers are most likely to transition to a higher-level administrative position right after certification, with the likelihood declining over time. Several explanations may account for this behavior. It may be that good matches are made soon and that as time passes either the individual cannot find a position that interests them or the individual does not have administrative skills that are attractive to employers. [6]

The pattern for transition from lower-level administration to higher-level administration is somewhat different (Figure 15B). In this case, the likelihood of moving to a higher administrative position steadily increases as time since certification increases. This is consistent with individuals gaining experience in lower-level administrative positions, preparing them for the more selective process associated with filling higher administrative positions.

It is also the case that the likelihood of a transition responds to changes over time in relative demand and supply. From 1995 through 2001 there has been a dramatic increase in the probability that a teacher with administrative certification would move to lower-level administration (Figure 16A). The likelihood of such a transition more than doubled during that period. Recall that this is occurring during a period during which there was a substantial increase in the number of individuals gaining certification (Figure 1). This is clearly the response to the increased demand for new administrators associated with the baby-boom retirements. A similar, but less dramatic, pattern emerges when the transition of lower-level administrators is considered (Figure 16B). Since 1995 there has been an increase of about 50 in the likelihood a lower-level administrator will transition to a higher-level administrative position. At the same time, the likelihood that a lower-level administrator returns to teaching declines by about 40 percent. Taken together the increased rates of transition to administrative positions suggest a substantial tightening of the labor market for leadership positions.

5. Summary

In this paper we have examined attributes of individuals that are important in their decision to obtain certification for administrative positions and in whether they transition to administrative positions. These findings provide some guidance to policymakers regarding the pool of eligible school leaders, how it has changed over time and the factors that influence an individual’s decision to obtain administrative certification. The findings also begin to explore how transitions to administrative positions have changed over time and the factors that influence whether those with administrative certification transition to increasing levels of administrative responsibility. We have a much better understanding of when during an individual’s career these transitions occur, how these relationships have changed over time and role that gender plays in these transitions.

There is much more to be learned about these and other relationships. Most importantly, we would like to better understand the role that variables like salary play in these decisions. As explained earlier, adequately measuring and modeling variables such as these is very difficult. We continue to work on these issues but are not sufficiently confident that we have adequately addressed the problems to include these variables in our models to date.

References

Boyd, D., H. Lankford, S. Loeb, and J. Wyckoff, “Analyzing the Determinants of the Matching of Public School Teachers to Jobs” Working Paper, 2003.

Gates, S., J. Ringel, L. Santibanez, C. Chung and K. Ross, Who is Leading Our Schools? An Overview of School Administrators and their Careers, RAND Corporation, 2002

Hill, P., M. Roza, M. Celio, J. Harvey, and S. Wishon, A Problem of Definition: Is There Truly a Shortage of School Principals? Draft, Center on Reinventing Public Education, University of Washington, 2002.

Lankford, H., R. O’Connell and J. Wyckoff, “Identifying the Next Generation of School Leaders” Report Submitted to the New York State Education Department, 2003.

Papa, F., H. Lankford and J. Wyckoff “The Career Paths of School Leaders” Working Paper, 2002.

Table 1: Descriptive Statistics for Variables Employed in Principal Certification Analysis

female	0.72	0.449	urban (except NYC)	0.065	0.247
age	43.226	9.908	suburban	0.507	0.500
(age)²	1967	853	rural	0.215	0.411
(age)³	93391	58170	New York City	0.212	0.409
bachelors degree	0.119	0.324	Capital District	0.059	0.235
masters degree	0.593	0.491	Buffalo	0.079	0.270
more than masters	0.288	0.453	Lower Hudson	0.072	0.259
total experience	14.829	9.879	Long Island	0.194	0.396
(total experience)²	317	336	Rochester	0.087	0.282
(total experience)³	7908	10971	Syracuse	0.059	0.235
district experience	12.721	9.416	Utica-Rome	0.022	0.146
(district experience)²	250	298	Mid-Hudson	0.079	0.270
(district experience)³	5885	9083	Southern-Tier	0.084	0.277
Barrons Most or highly competitive	0.113	0.317	North Country	0.052	0.222
Barrons Very competitive	0.220	0.415	year1990	0.077	0.267
Barrons Competitive	0.544	0.498	year1991	0.078	0.269
Barrons Less or Not competitive	0.122	0.327

Commonly Held Beliefs		Results from Our Analysis
Ø Females are disadvantaged in the hiring process for leadership positions		Ø Females are less likely to become certified as administrators than males. Conditional on being certified males are more likely to become lower level administrators but females are just as likely to become principals.
Ø Urban teachers are less likely to become leaders due to the extraordinary challenges facing urban school leaders		Ø In fact, urban teachers are more likely to become certified for administrative positions and once certified appear to be more likely to assume administrative positions.
Ø Individuals typically decide upon moving to administration early in their careers		Ø There is evidence that many individuals seek to become certified within the first few years of their careers.
Ø If someone does not secure a leadership position soon after becoming certified for one, they are unlikely to do so later on		Ø Yes, individuals are most likely to assume lower level administrative positions immediately following certification and move to higher level positions over time. Ø The implication of the preceding two bullets is that early efforts to identify and “tap” prospective administrators are likely to meet with the most success.
Ø The labor market for school leaders has become much tighter in recent years, making it increasingly difficult to recruit leaders		Ø We find some evidence for this with a substantial increase in the proportion of individuals moving into administrative positions earlier in their careers.