Sie sind auf Seite 1von 35

Business School Rankings and Business School Deans*

C. Edward Fee Michigan State University fee@msu.edu Charles J. Hadlock Michigan State University hadlock@msu.edu Joshua R. Pierce Michigan State University piercej7@msu.edu First Draft: September 19, 2003 This Draft: October 6, 2003

* The most recent version of this paper can found at http://nebula.bus.msu.edu:8080/feehadlock. All errors remain our own.

Business School Rankings and Business School Deans

ABSTRACT We examine the relation between dean turnover and changes in rankings in a comprehensive sample of business schools with ranked MBA programs from 1990-2002. We find little evidence that dean departures are related to changes in a school's overall rank in the U.S. News & World Report rankings. However, dean turnover does appear to be elevated when a school drops in the Business Week rankings and when a school deteriorates on the placement dimension as measured by U.S. News & World Report. These results are significant in both a statistical and economic sense. Our findings suggest either that schools respond to changes in rankings or that rankings reflect information schools use in their personnel decisions. JEL Classification: G30; J41; L20; M10, M51 Key words: Business schools, deans, rankings, turnover, performance evaluation

1. Introduction

An event that attracts substantial attention from constituents of leading business schools is the public release of rankings of MBA programs. The attention these rankings garner and the impact rankings may have on students, recruiters, faculty, and administrators has generated a lively discussion in the business and academic press. Some of this discussion centers on understanding what the various rankings actually measure and debating what the rankings should measure. Other discussion considers whether the focus on rankings has any effects on how business schools operate and whether or not these effects are welfare improving. In this paper we contribute to these discussions by providing systematic empirical evidence on the relation between rankings and the career of a school's most visible decision maker -- its dean. In particular, we examine whether changes in rankings are associated with the likelihood that a dean departs from his/her school. Our analysis allows us to examine the joint hypothesis that (a) deans are held accountable for school performance, and (b) that rankings are a useful indicator of a dean's performance. By using both the Business Week (BW) and U.S. News and World Report (USN) rankings along with the components of these rankings, we are able to assess which performance metrics appear to be most heavily relied on in evaluating deans. Our sample includes the deans of all schools ranked by BW or USN from 1990-2002. We use news articles to determine the circumstances surrounding each dean's departure and estimate logit models predicting whether dean turnover is related to rankings changes. Our estimates reveal no significant evidence that overall performance in the USN rankings affects the likelihood of dean turnover. However, poor performance in the BW rankings is associated with a significant increase in the likelihood of a dean leaving. Thus, while the relative information

content of the USN and BW rankings may be debated, schools are behaving as if they are evaluating deans based on factors that are captured in the BW ranking measure. To further explore these results, we estimate models where dean turnover is a function of each individual component of the respective rankings. Interestingly, while performance in the overall USN ranking is not associated with turnover, we find that a school's performance on the USN placement measure is significantly related to turnover. The three other USN component measures are insignificant. A possible explanation for this evidence is that placement success is the ultimate market measure of whether a school is effectively managing its operations. When we turn to the two component measures of the BW ranking, we find no evidence that a single component is driving the overall importance of the BW ranking. Finally, we find little evidence of systematic changes in rankings or program characteristics following dean changes. Taken as a whole, our evidence suggests that business school deans' preoccupation with rankings is well founded. The BW overall rankings and the USN placement metric are both related to dean turnover just as stock returns and earnings performance are related to CEO turnover. This evidence is consistent with schools believing both that deans can affect rankings related metrics and that these metrics are important. The rest of our paper is organized as follows. In section 2 we review the specific literature on business school rankings and the broader academic literature on performance evaluation and management turnover. In section 3 we discuss our sample construction and present summary statistics. Section 4 reports our main results on the relation between rankings and dean turnover. In section 5 we conclude.

2. Rankings, business schools, and turnover

2.1 The information content of rankings

As interest in business education has grown over time, publishers have expended considerable effort towards measuring the quality of MBA programs. Five major news organizations currently publish rankings, with the Business Week (BW) and U.S. News and World Report (USN) rankings being the oldest and most established.1 Business school rankings aggregate subjective and objective assessments of a school's inputs (quality of incoming students), outputs (quality of graduating students), and the process of turning inputs into outputs (the school experience). Each ranking methodology uses different measures and different weights on each type of measure, thus creating at times substantial differences between rankings. Two recent academic studies have attempted to understand the information content of business school rankings. Tracy and Waldfogel (1997) report that the BW and USN rankings are both heavily influenced by the quality of a program's students (inputs) and the program's placement success (outputs), but not necessarily by the value-added by the program (outputs minus inputs). Dichev (1999) studies the time series properties of the BW and USN rankings. For both rankings he finds that the change in a school's ranking exhibits strong negative firstorder autocorrelation. In addition, he reports that changes in BW rankings are not significantly correlated with changes in USN rankings. The Dichev (1999) evidence suggests that there is significant noise in these highly publicized ranking measures. Certainly school rankings have many shortcomings, many owing to the fact that school quality is a multidimensional and imprecise notion. Nevertheless, these rankings may contain or
1

The other rankings are published by Forbes, the Financial Times, and the Wall Street Journal. These rankings have not existed for a long enough time period to use in our study.

reflect useful information, or at least market participants may behave as if they do. By examining whether schools' choices are related to rankings changes, we can assess whether the information in rankings, despite the flaws, is useful or important to schools in assessing performance.

2.2 Business school objectives

Business school constituents have many diverse and, at times, conflicting objectives. The choice of objectives and the relative weights placed on each objective will depend on the school's governance structure and the identity of the individuals associated with the school. A magazine ranking creates a highly visible performance metric that may place more emphasis on certain areas than a school would choose on its own. If a school cares about the rankings, this may cause a change in a school's organizational objectives. Along these lines, Brickley and Zimmerman (2001) discuss how pressure caused by rankings resulted in a shift towards rewarding teaching performance at the University of Rochester business school. They report that faculty at the school appear to have responded to the altered incentive scheme. Zimmerman (2001) discusses how a fixation on rankings has affected business school strategies and the detrimental effect this has had on business school research activities. There is little systematic evidence indicating how much business schools and/or their deans care about rankings, although anecdotal evidence abounds. Some schools and deans conspicuously advertise their rankings, particularly when they are high, while others purport not to use rankings as a measure of success. Some deans are publicly overt about their desire to increase rankings. For example, after being appointed dean at the University of Maryland

business school, Howard Frank is quoted as saying "we want to be in the top 10 or top 15" (Washington Post, 7/13/1998, page F10). While not all deans are so blunt and some may truly not care about rankings, certainly many business faculty believe that their deans care about rankings and that this has had an impact on their school's activities. Why should deans care about rankings? One possibility is that they derive inherent satisfaction from superior performance in rankings. Alternatively, it could be that the labor market or the governance system of business schools is oriented towards rewarding excellence in rankings or in school characteristics that are highly correlated with rankings (quality of students, placement success, student satisfaction, corporate reputation, etc.). If schools are utilizing this type of reward structure, it presumably reflects the objectives of the principals who evaluate the deans. If schools do not use ranking related information to evaluate deans, this would suggest that schools either find these measures to be uninformative or, alternatively, that they have weak evaluation and governance systems.

2.3 Performance evaluation and turnover

The approach we take in this paper draws heavily from the empirical modeling in the literature examining CEO turnover [e.g., Coughlan and Schmidt (1985), Warner, Watts, and Wruck (1988), and Weisbach (1988)]. The basic approach is to estimate logit models where the dependent variable indicates whether a manager departs from his/her firm and the main independent variable of interest is performance. The magnitude of the relation between performance and turnover is then interpreted as either a measure of the strength of an

organization's governance system, or alternatively, as a measure of the degree to which the chosen performance metrics provide useful information concerning managerial ability. In adapting this methodology to our study, a few considerations are worth discussing. For both CEOs and deans, it is often not clear whether an individual was forced to depart or not. We will follow the management turnover literature in making the implicit assumption that the typical departure is forced unless there is compelling evidence to the contrary.2 However, it is important to note here that the distinction between voluntary and involuntary departures is often not important, as most job separations can be viewed as two-sided decisions where both parties agree that the firm and the manager are a poor match [e.g., Jovanovic (1979)]. Viewed in this way, we would interpret a relation between dean turnover and rankings downgrades as evidence of a poor dean-school match leading to a job separation. With regards to performance measurement, most studies of management turnover assume that executives are evaluated based on metrics that reflect shareholder objectives, primarily stock returns and/or earnings performance. Boards of directors presumably rely on these metrics because of their information content with regards to managerial ability. By using rankings as our performance metric, we are testing whether rankings provide or capture relevant information concerning the perceived ability of a school's dean. There is, however, a subtle identification issue as to the role of rankings in performance evaluation. It could be that the actual publication of a ranking provides useful information to evaluate a dean. Alternatively, it could be that the ranking simply reflects or publicizes other information that is known to school administrators and faculty about the quality of a school's programs. While it is impossible to completely
Supporting this assumption, Fee and Hadlock (2003a) present evidence indicating that most managerial departures that are not reported as forced appear similar to forced departures in many respects including post-separation labor market prospects. The one type of departure that appears quite different from others are cases where an individual jumps to take a more prestigious position elsewhere (see Fee and Hadlock (2003b)).
2

disentangle these two possibilities, we will make some attempts when we examine the timing of dean departures around rankings releases.

3. Data and sample selection

3.1 Sample construction

We begin by identifying all schools with MBA programs that are ranked by either USN or BW from 1990 to 2002.3 We include a school in a given year only if it was assigned a numerical ranking in either USN or BW that year. Over time the USN ranking has expanded from 25 schools to 50, while the BW ranking has expanded from 20 schools to 30. Thus, our later sample period contains more schools and these are the lower tier schools. More specific details on how the rankings are constructed are reported in the appendix. For each included observation, we identify the school's dean as of the start of the academic year (September 1) using data from AACSB membership directories, various Hasselback directories of business school faculty, and annual editions of Peterson's Annual Guide to Graduate Studs. We double-check all discrepancies between these sources and add to our information set by looking at histories on school websites, online vitas of the deans in question, and news searches on the Factiva electronic database (this source includes hundreds of local and national newspapers). This process allows us to determine who was in office every September 1 and whether there was a change in dean over the subsequent 1-year period.

We begin in 1990, as this is the first time that both magazines published rankings in the same year. USN first published a ranking in 1987 but began regular annual rankings in 1990. BW published its first biannual ranking in 1988.

For each dean in the sample, we collect several pieces of relevant information. An important control variable in many turnover studies is age. Unfortunately, however, systematic data on age was not available from our sources. As a reasonable alternative, we construct a proxy variable by using data from our various directories and news searches to identify the year the individual received his/her highest degree (usually a Ph.D.). We then construct an implied age variable by assuming an individual receives their doctorate at age 27, master's degree at age 26, and bachelor's degree at age 21. Since an individual's tenure on the job may affect the strength of his/her attachment to the position, we create a tenure variable indicating how many years the individual had served as dean as of the start of the observation year. Finally, we record whether an individual was hired as an insider or outsider and whether they had an academic or non-academic background. We report in Table 1 basic summary statistics for our entire sample where each deanschool-year from 9/1990 to 9/2002 is treated as a single observation and deans serving on an interim basis are excluded. To identify potential time trends or differences by school rank, we also report these statistics for selected subsets of observations. As the figures indicate, our sample includes 60 schools and 521 total observations. For the sample as a whole, there are approximately an equal number of public and private schools, but for schools ranked in the top 25, almost two thirds are private. The average dean is approximately 53 years old, with slightly over 5 years on the job. Deans at top 25 schools in the latter half of the sample exhibit a lower average tenure than those in the first half of the sample (4.37 years vs. 5.56 years), suggesting that deans may not last as long in office in the more recent time period. Below we will formally test for time trends in turnover. Finally, a slight majority of deans are outsiders and a little more than three quarters are traditional academics.

3.2 Characterizing dean departures

For every dean departure identified by our sample collection procedure, we review all articles referring to the departure. In addition, we search news articles and various directories for any evidence that the individual resurfaced at a new employer. In all cases we are able to find some news article announcing or confirming that an identified departure took place, but there is substantial variation in the information content of these articles. To each turnover event, we assign a decision date equal to the earliest date that our sources can confirm that the individual was leaving or had left. The year in which the decision date falls, measured on a September 1 to September 1 basis, is considered the turnover year for that event. Some basic summary statistics on dean turnover and career lengths is reported in Panel A of Table 2. Since we need information on the identity of the individual serving as dean at both the start and end of the year to calculate turnover, our sample of possible turnover school-years with non-interim deans drops to 471 from the 521 total school-years reported in Table 1. In this sample we find 61 dean departures, implying a turnover frequency of 12.95%. To gauge what this means in terms of typical career lengths, we treat each dean-school match as a single observation and estimate the distribution of the survival time of a dean's career at a school.4 As we report in the table, the estimated 25th percentile of a dean's stint at a school is 4 years, the 50th percentile is 8 years, and the 75th percentile is 13 years. Interestingly, these figures on turnover rates and job tenure are fairly similar to what others have reported for CEOs of large corporations in the 1990s.

Survival time percentiles are calculated using the Kaplan-Meier product-limit estimate of the survivor function.

Using information on the turnover events, we assign each departure to one of five mutually exclusive categories, where the death category is self-explanatory. All departures where articles mention that the dean was pressured to leave by faculty or administrators are assigned to the forced category. Non-forced departures where an individual leaves and takes a position outside his school within 6 months of departure are assigned to either the "move up" or move down" category depending on the relative quality of the new position. Specifically, we assign a departure to the move down category for deans who take dean positions at schools that are ranked lower than the school they leave. All other external moves are assigned to the move up category. This category is composed entirely of cases where an individual assumes the deanship of a higher ranked business school, takes a university president position, or is hired as a CEO or CFO of a large corporation. The remaining category is the "generic departure" category. These are cases where little or no information is reported or where the dean departs for standard generic reasons such as "a desire to return to research and teaching." Typically (73% of the time), deans in the generic departure category stay on at their current schools as faculty members. The figures in Panel B of Table 2 report the fraction of departures in each category. Similar to CEO turnover, and not surprising given the delicate nature of these decisions, most departures (62.30%) are classified as generic departures. Approximately 10% of dean departures fall into the forced category.5 A fair number of departures (25%) are associated with the dean going elsewhere, with moves up being more common than moves down (18.03% vs. 6.56%). In Panel C of Table 2 we add figures on the identity of a departed dean's replacement. The propensity of schools to hire outsiders and non-academics is consistent with the sample-wide
5

Weisbach (1988) reports that only a handful of CEOs are overtly forced from office. In a more recent sample, Fee and Hadlock (2003a) find that only 15.6% of CEO departures can be categorized as overtly forced.

10

averages reported earlier in Table 1. It appears that it is quite common to initially appoint an interim dean after a dean departure, with 36.07% of all turnover events following this pattern.

3.3 Business school ranking data

The final necessary component for our empirical analysis is information on rankings performance. We construct these measures based on the change in a school's ranking, as changes have the potential to provide incremental information that may be used in updating beliefs concerning a dean's abilities. Since the USN rankings are annual rankings, we calculate USN rankings changes on a 1-year and a 2-year basis. The BW rankings are released every other year, so we calculate BW rankings changes only for 2-year windows. We report summary statistics on rankings changes in Panel A of Table 3. Both rankings appear to exhibit similar levels of variation over time. In particular, over 2-year windows both rankings series have a median absolute value change of 2.00, indicating that half of the schools move up or down by no more than 2 in a 2-year window. Figures for the 25th percentile and 75th percentile of the distribution of changes confirm this observation. For both USN and BW the mean of the absolute value of rankings changes is larger than the median (means of 3.187 and 2.822 respectively), indicating that there is some skewness in these distributions. Since we are interested in the relative information content of these two performance related measures, it is informative to reproduce and extend Dichev's (1999) evidence on the correlation structure of these series. As we report in Panel B of Table 3, if we match each BW ranking with the USN ranking that is published immediately after it (around 5 months later), we find a high correlation (.832). Thus, it appears that the level of a school's ranking in the BW and

11

USN listing is highly correlated across series. However, when we consider the change in BW and USN rankings over two-year periods, the correlation in changes is quite low (.007) and not significant (p-value of .949).6 Thus, using a significantly larger sample than Dichev (1999), we are able to confirm his finding of very little correlation in the updating of these series. Turning to the autocorrelation structure, we confirm the second main finding of Dichev (1999) that changes in rankings exhibit significantly negative first-order autocorrelation. This indicates that some part of a school's rankings change in either poll is predictable. Whether market participants filter out this predictability in their use of rankings changes is an open question.

4. Models of Dean Turnover

4.1 Empirical strategy

We now turn to examining the role of rankings in dean turnover. Our main dependent variable assumes a value of 1 if a dean leaves over the observation period for any reason except death or a move up, and 0 if the dean stays. This dependent variable should pick up most cases where the dean's departure is precipitated by dissatisfaction with performance. Following the existing literature on management turnover, our empirical approach is to estimate logit models predicting this dependent variable as a function of measures related to rankings performance and other control variables such as age and tenure. In all models we exclude deans who served for

We deviate from Dichev's (1999) treatment here in that we match each BW ranking with the subsequent USN ranking. This seems appropriate as both rankings are based on information collected primarily in the spring and summer of the previous year based on the most recent graduating class. If we follow Dichev's (1999) approach of matching each BW ranking to the average of USN rankings that straddle it, our results are very similar.

12

less than one year as of the start of the observation period, as any performance changes are unlikely to be attributable to them. Some subtle timing issues arise in our construction of performance measures. Because the natural way to organize data on dean turnover is based on the academic calendar, our dependent variable is coded using a September 1 to September 1 window, where we consider a turnover to occur on the event's decision date. Since the USN and BW rankings data are released at different times and with different frequencies, we have to tailor our empirical modeling to each ranking.

4.2 The USN rankings

The USN rankings are annual rankings that are released in March. Thus, one natural performance metric, which we call USN lagged performance, is the change in a school's USN rank in the March to March period that completely precedes the September to September observation year (e.g., rankings changes from March 1996 to March 1997 predicting turnover from September 1997 to September 1998). An alternative metric, which we call USN contemporaneous performance, is the change in a school's USN rank in the March to March period that ends during the September to September observation period (e.g., rankings changes from March 1997 to March 1998 predicting turnover from September 1997 to September 1998). We have no strong feel as to which performance measurement period is more appropriate. This depends on how quickly schools react to rankings changes and whether a school can predict its ranking before the actual ranking release date using internal information such as student quality and placement. A third performance metric, which is closely related to the other two, is the

13

change in a school's USN ranking between the March ranking that falls during the observation year and the March ranking two years prior (e.g., rankings changes from March 1996 to March 1998 predicting turnover from September 1997 to September 1998). We refer to this as the USN 2-year performance measure. We report in columns 1 to 3 of Table 4 estimates from logit models in which each column includes one of these USN based performance metrics. Our sign convention is such that increases in rankings are coded as positive performance numbers and decreases are coded as negative numbers. While the estimates on the performance variables in these three columns have the expected negative sign, none are significant at conventional levels. Moreover, the implied probabilities derived from these estimates and reported at the bottom of the table suggest a very small change in the probability of turnover for large changes in a school's USN ranking. Thus, there is little evidence of a simple relation between USN rankings changes and dean departures. It is also interesting to note that the year variable is in all cases negative and insignificant, thus lending little support to the hypothesis that dean turnover rates have increased over time.

4.3 The BW rankings

The BW rankings are bi-annual rankings released in October of every even year (i.e., 1990, 1992, and so on). Given the timing of these rankings, our natural window for predicting turnover when using these rankings is a 2-year period. There are again some subtle timing issues with regards to measuring a program's performance. One possibility is to use the change in the BW ranking in the October to October twoyear period that ends just after the start of the 2-year turnover observation period (e.g., rankings

14

changes from October 1996 to October 1998 predicting turnover from September 1998 to September 2000). We will refer to this variable as Lagged BW performance. If most turnover occurs in response to an actual rankings downgrade and if it takes up to 2 years for deans to depart after the downgrade, this measure would be the most appropriate. An alternative possibility is to use the change in the BW ranking over the 2-year October to October period that approximately overlaps the 2-year turnover observation period (e.g., rankings changes from October 1998 to October 2000 predicting turnover from September 1998 to September 2000). This performance measure, which we call Contemporaneous BW performance, would not be appropriate if deans depart only in response to a rankings change, since the pubic release of the ranking change actually comes after the end of the turnover period. However, this measure is appropriate if deans depart not in response to an actual rankings change, but rather in response to the underlying forces that cause a rankings change or in anticipation of a rankings change. For example, if a school knows that it has unhappy students or a deteriorating reputation amongst recruiters, the dean may depart either because the school anticipates a rankings drop or because the school is evaluating the dean using factors that happen to be reflected in the rankings. A compromise variable, which we refer to simply as BW 2-year performance, uses the change in ranking between October of year t-1 and October of year t+1 to predict turnover between September of year t and September of year t+2 (e.g., rankings changes from October 1998 to October 2000 predicting turnover from September 1999 to September 2001). This measure will match rankings downgrades with turnover events that happen in the period shortly (approximately 1 year) before the downgrade and shortly (approximately 1 year) after the downgrade. If dean departures are driven by a combination of leaving in response to or in anticipation of an actual rankings downgrade, or if they are driven by information that a school

15

receives in the window around a rankings change, then this variable will be the most appropriate measure. Since it seems reasonable to us that some combination of these possibilities will govern dean turnover behavior, we will use this BW 2-year performance measure as our baseline BW performance measure. In column 4 of Table 4 we present logit estimates using this performance metric. The estimated coefficient on the performance variable is negative and significant at the 5% level (t= -2.13). The implied probabilities from the estimation indicate that the role of performance in turnover is also economically significant. If a dean goes from the 75th percentile of performance (an increase in ranking of 2) to the 25th percentile of performance (a drop in ranking of 2), the model implies that the probability of his/her departure over the 2-year turnover period increases from 15.5% to 31.7%. The magnitudes are more dramatic for larger ranking changes. This evidence indicates that the BW ranking itself, or factors that are reflected in the BW ranking, are strongly related to the dean turnover decision. In results that we omit for brevity, we experiment with using the other possible BW performance variables in place of BW 2-year performance in the specification of column 4 of Table 4. When we use the Lagged BW performance variable, the coefficient is negative but not significant (t=-1.01). When we use the Contemporaneous BW performance variable, the coefficient is negative but very small in magnitude and not even close to significant (t=-0.40). A reasonable interpretation of this evidence is that most of the dean turnover that is related to rankings changes occurs in the 2-year window centered around the release of the information that the ranking has changed. This suggests that schools are reacting somewhat to the release of the ranking itself, and somewhat to either the anticipation of a ranking drop or a change in school characteristics that ultimately are reflected in their ranking.

16

We report in column 5 of Table 4 estimates from a regression including both USN and BW performance measures. For purposes of comparison, we measure both turnover and performance on a 2-year basis. The results here are similar to our earlier findings. The coefficient on the BW measure remains negative and significant, while the coefficient on the USN ranking measure is small and insignificant (t=.052). The difference between the estimated coefficients on these two rankings variables is significant at the 15% level (p-value = .146). Taken as a whole, the evidence seems fairly convincing that the BW rankings are related to dean departures and the USN rankings are not.

4.4 Robustness and extensions

To explore the robustness of our findings, we consider several modifications to our basic empirical framework. First, because of concerns about outliers, we experiment with using coarser measures of performance that assume a value of 1 when a school's ranking falls over a performance measurement period and a value of 0 if the ranking increases or remains constant over the period. The results using this type of variable in place of the corresponding performance variables in the specifications of Table 4 are very similar to what we report above. In particular, we find no evidence that USN rankings changes are related to turnover, while the relation between BW rankings changes measured in this up/down manner and turnover is significant at the 5% level. Since we document earlier that rankings changes are somewhat predictable because of first-order autocorrelation, we also experiment with using the unexpected component of rankings changes instead of the raw rankings change. Specifically, for each performance variable used in

17

Table 4, we regress the variable against its lagged value and use the residuals from the regression as our measure of performance. In general, turnover/performance sensitivities tend to be slightly stronger using the residual measures. Specifically, in the specification corresponding to column 4, the coefficient on BW performance is 0.381 (as opposed to 0.232) and is significant at the 1% confidence level. The coefficient on USN performance remains insignificant in the specification corresponding to column 2, but is significant at the 10% (5%) level in the specification corresponding to column 1 (3). Thus it appears that turnover does respond to new information reflected in rankings, rather than just the predictable component of changes. Our finding of greater turnover/performance sensitivity for BW rankings than for USN rankings is unaltered by the alternate performance definitions. When we include both USN and BW residual performance in a specification analogous to column 5, we again find a significant (1% level) coefficient on BW performance and an insignificant coefficient on USN performance. As a final alternative performance metric, we consider the possibility that dean turnover is related to major gifts, as fundraising is often one of a dean's major responsibilities. Our source of information here is the list of business school naming gifts reported by Burch and Nanda (2003) as well as AACSB Internationals list of the largest business school donations from a single source7. We use this information to create variables measuring total gifts received and total non-naming gifts received by a school over the performance measurement period corresponding to our rankings variables. When we include each of these variables, in turn, in the specifications of Table 4, the estimated coefficients on these variables are in all cases insignificant and inclusion of these variables has no appreciable effect on any of the performance coefficient estimates. We should caution that the data on gifts is quite sparse and only includes

This list, which we downloaded from http://www.aacsb.edu/members/communities/interestgrps/donors.asp, provides information on 97 gifts ranging in size from $4 million to $62 million.

18

large gifts, so it could be that we do not have sufficient power to detect the role of fundraising on dean turnover. In addition to these robustness checks, we also examine whether there is significant variation in the dean-turnover mechanism across schools or types of deans. In particular, we experiment with adding dummy variables and dummy variables interacted with the performance metric to the specifications of Table 4. The dummy variables we employ here, one at a time, are based on (a) the school's status as a private or public school, (b) whether the school was ranked in the top 25, (c) whether the dean was an insider or an outsider, and (d) whether the dean was an academic or non-academic. In all cases the estimated coefficient on the dummy variable and the dummy variable interacted with performance is insignificant. Thus, we find little evidence of substantial heterogeneity in the turnover mechanism by school or dean characteristics. Finally, we consider some alternative treatments of the dependent variable. First, because we are concerned that some dean departures are natural retirements, we re-run the regressions in Table 4 for all deans under the age of 63. The results here are qualitatively very similar to what report in the table. Second, because we suspect that rankings increases may cause deans to move up in the dean labor market, we create a dependent variable that assumes a value of 1 for a "move up" departure and a 0 for all other observations. When we use this dependent variable in the specifications of Table 4, all of the estimates on the explanatory variables are insignificant. It may be that these variables play no role in upward moves, but given the small number of these types of job movements, our power here may be quite limited.

4.5 Rankings components

19

Our results to this point indicate that the raw USN ranking is not significantly related to dean turnover while the BW ranking appears to be significantly related to dean turnover. To further understand what information is used in evaluating deans, we consider the role of the major components of the rankings in dean turnover. The BW ranking has two major components, a graduate poll and a corporate poll.8 We experiment with including variables based on a school's change in these components in the specification of column 4 of Table 4 in place of the overall BW performance variable. The estimates from these models reveal no significant difference between these two components when they are both included, and neither has a significant coefficient when included individually. Thus, it appears that the BW ranking aggregates several pieces of information that are collectively used in evaluating deans. The USN poll has consistently included four major components: placement, selectivity, corporate reputation, and academic reputation. We experiment with using variables based on a school's change in each of these ranking components, included one at a time, in specifications (1)-(3) of Table 4 in place of the overall USN rank. In no cases are the coefficients on selectivity, corporate reputation, or academic reputation significant at the 5% confidence level (the coefficient on corporate reputation was significant at the 10% level in one specification but of small magnitude). The one component that is ever significant at the 5% confidence level is the change in a school's placement score, and we report results for this variable in Table 5. When we include the lagged version of this variable (see column 1), the coefficient is negative but not significant at conventional levels (t=-1.28). However, in column 2 the contemporaneous version of this variable displays a negative coefficient that is significant at the 10% level (t=-1.77). As we report in column 3, the estimated coefficient on the 2-year version of this
8

In the more recent polls BW has also placed a small weight on a third component, intellectual capital. For more details on the rankings, see the appendix.

20

placement variable is negative and highly significant (t=-2.40). The estimates in column 3 imply that a dean at a school that displays a strong increase in placement success (75th percentile) has only a 7.5% chance of departing over the course of the year. This almost doubles to 13.4% for a dean at a school with poor performance (25th percentile) on this dimension. Thus, while the overall USN ranking is not strongly associated with dean turnover, there does appear to be some relevant information in a school's placement success as reported by USN in their rankings.

4.6 What happens after dean departures?

The fact that dean departure rates are elevated around poor performance in the BW rankings and the USN placement rankings suggests that schools are behaving as if deans can actually affect rankings or rankings related measures. If this is the case, we might expect to observe large changes in a school's ranking or in other program characteristics after a new dean is put in charge.9 To examine this possibility, we collect data on what happens to a school's rankings, enrollment, and tuition in the period after a new permanent dean starts leading a school. We calculate these changes over the two-year period starting with the first BW or USN rankings release after the new deans appointment. We report in Table 6 statistics on median changes in the above measures and, as a measure of variance, the median absolute values of these changes. We report these statistics separately for schools with a new dean appointed immediately prior to the observation and for schools with no new dean appointment. As we report in the table, there is little evidence of an

Along similar lines, Denis and Denis (1995) find evidence of significant changes in firm performance after top management dismissals. See also Bertrand and Schoar (2003) on how the identity of a manager can have a significant effect on firm characteristics.

21

abnormal increase or decrease in rankings in the two-year period following the appointment of a new dean. For schools with new deans, the median change in the USN (BW) ranking is 0.25 ( 0.50). These figures do not differ significantly from the median ranking change of 0 in both polls for schools without a new dean. The figures in Table 6 also reveal no significant tendency towards an abnormal increase or decrease in placement success, enrollment, or tuition following the appointment of a new dean. When we examine the median absolute value of changes in these variables, we find little evidence of large increases in the variance of subsequent program changes following a new dean appointment. In particular, only one measure (the median change in the absolute value of a school's USN ranking) is significantly elevated around dean replacements, and this difference is significant at only the 10% level. Taken as whole, the evidence of major changes in rankings or program characteristics around dean changes is weak. However, given the long windows and limited data, our power to detect such effects may be low.

5. Conclusion

Using a comprehensive sample of business schools with ranked MBA programs from 1990-2002, we find that departures of deans are related to some rankings related metrics and not to others. We find little evidence that dean departures are related to the change in a school's overall rank in the U.S. News & World Report (USN) rankings. However, dean turnover does appear to be elevated when a school drops in the Business Week (BW) rankings and when a school deteriorates on the placement dimension as measured by USN. These results are

22

significant in both a statistical and economic sense. We find little evidence of systematic changes in rankings or program characteristics after a dean is replaced. The timing of dean departures around rankings releases indicates that these departures are concentrated in the period shortly before and shortly after the release of a rankings downgrade. This suggests that schools are reacting both to rankings downgrades themselves and to the anticipation of a downgrade. Alternatively, it may be that schools are not responding to rankings or anticipated rankings, but rather to information that is reflected in the rankings. In either case, it appears that market participants are behaving as if rankings are either informative or important when it comes to measuring a dean's success. From the perspective of deans, our results indicate that a strong focus on rankings or rankings related metrics by these individuals is a rational response to some sort of evaluation scheme. Poor performance in the rankings tends to lead to a dean stepping down from his/her position for reasons other than taking a more attractive position elsewhere. Some deans probably leave voluntarily (perhaps with schools who have fallen in rankings doing little to dissuade them), while others are likely to be either gently nudged out or more forcefully pressured to leave. Whatever the scenario, our results indicate that deans and schools use ranking related information to evaluate the individual's ability and/or the quality of the match between the individual and the school. An increase in rankings will reflect well on the dean and a decrease will do the opposite. Aside from the issue of a dean's incentives, our results are relevant to some other issues concerning how business schools function. First, our findings suggest that there is an active governance system in business schools where deans are evaluated based on performance. Second, our results tell us something about business school objectives. In particular, our results

23

suggest that schools care more about their BW ranking than their USN ranking and further suggest that schools emphasize placement success. The emphasis put on the BW ranking may reflect the fact that many consider this to be the most prestigious ranking. Alternatively, emphasis on the BW ranking may indicate that schools care more about qualitative measures of student and recruiter perceptions that are weighed heavily by BW. The fact that schools care about rankings in general and this type of ranking in particular suggests that deans and schools will tend to try to improve on dimensions that will earn them higher BW rankings. Whether this is a good or bad thing depends on one's perspective, but certainly catering to the rankings is not without costs, as Zimmerman (2001) emphasizes.

24

Appendix In this appendix we provide a brief description of the construction of the BW and USN rankings and how they have changed over time. We also discuss our construction of the placement success variable. Before providing these details, we first note that when a group of schools is tied for a certain rank, we assign all schools in the group a rank equal to the midpoint of the rank of the schools immediately above and immediately below the group. Thus, if two schools were tied with a rank of 10, we would assign them both a rank of 10.5. The first BW ranking was published in November of 1988 and subsequent rankings have been released every even year in the month of October. BW ranked 20 schools from 1988-1994, 25 schools from 1996-1998, and 30 schools from 2000-2002. BW also publishes a list of "runner-up" schools. We ignore this list, as it does not rank schools within the runner-up category. Before 2000, the BW ranking was based entirely on a corporate poll (poll of recruiters) and a graduate poll (poll of recent graduates). The graduate poll rank from 1992 onwards is a weighted average of the most recent poll and the two preceding polls. It is not clear exactly how BW weighs the corporate poll rank and the graduate poll rank to construct an overall rank, but it appears that they weigh these approximately equally. Starting in 2000, BW explicitly weighs the two polls 45% each, with the remaining 10% being based on an intellectual capital score that is based on scholarly output. The first USN ranking was published in 1987. Regular annual rankings began in 1990 and have been published every year since in the months of either March or April. USN ranked 25 schools in 1990, 1992, and 1993, 30 schools in 1991, and 50 schools from 1994 onwards. A school's USN rank is based on placement, selectivity, reputation by recruiters/executives, and reputation by academics. For a brief period (1990-1994) the USN rank also placed a small weight (5%) on a school's graduation rate. Each of the four major components of the USN rank has evolved slightly over time. The placement rank is based primarily on objective statistics concerning whether graduates obtain jobs and the salaries they obtain. The selectivity rank is based on objective statistics such as GMAT scores, undergraduate grades, etc. The recruiter reputation rank is based on a poll of executives or recruiters who hire MBAs. The academic rank is based on surveys of business school deans and MBA program directors. The weightings on the four major components has changed over time, with the placement rank and selectivity rank always being weighed 25% or more, and the other two components being weighed 25% or less. To construct our measure of USN placement success, we began with each school's USN rank on the placement dimension. From 1999 to 2002 USN did not report the placement rank, but they did report data on the components of the placement rank and the weights they assign to each of these components in calculating a placement score. Using this data and the reported weights, we are able to construct a school's USN placement rank within the top 50 for these years. Given the way we create these variables, no school could have a rank worse than 50 on the placement dimension after 1998. In earlier years, some schools with an overall USN rank in the top 25 or top 50 have USN placement ranks that are lower than 25 or 50. To make the figures from the earlier and later years comparable, we assign each USN ranked school a placement score equal to its relative rank on the placement dimension within the set of ranked schools. Thus, if a school with a placement rank of 82 has the 49th worst placement score within the top 50 schools, we would assign the school a placement score of 49. Our measures of placement performance are then based on the change in a school's placement score over the relevant time window. We set placement performance equal to missing for all measurement periods that begin in the last year that U.S. News reported placement scores for only 25 schools, as the change in coverage from 25 schools to 50 schools causes placement data from the two regimes to be non-comparable. Inclusion of these observations has no material effect on our reported results. 25

References AACSB Membership Directory, various years, AACSB International, St. Louis. Bertrand, Marianne and Antoinette Schoar, 2003, "Managing with Style: The Effect of Managers on Firm Policies," Quarterly Journal of Economics 118-4. Brickley, James and Jerold Zimmerman, 2001, "Changing Incentives in a Multitask Environment: Evidence from a Top-Tier Business School," Journal of Corporate Finance 7, 367-396. Burch, Timothy R. and Vikram Nanda, 2003, "What's in a Name? Hotelling's Valuation Principal and Business School Namings," Journal of Business (forthcoming). Business Weeks Guide To The Best Business Schools, various years, McGraw-Hill, New York. Coughlan, Anne T. and Ronald M. Schmidt, 1985, "Executive compensation, management turnover, and firm performance: an empirical investigation," Journal of Accounting and Economics 7, 43-66. Denis, David and Diane Denis, 1995, "Performance changes following top management dismissals," Journal of Finance 50, 1029-57. Dichev, Ilia, 1999, "How Good Are Business School Rankings?" Journal of Business 72-2, 201213. Fee, C. Edward and Charles J Hadlock, 2003a, "Management Turnover across the Corporate Hierarchy," Journal of Accounting and Economics (forthcoming). Fee, C. Edward and Charles J. Hadlock, 2003b, "Raids, Rewards and Reputations in the Market for Managerial Talent," Review of Financial Studies 16-4, 1311-1353. Hasselback, James R., various years, Prentice Hall Guide To Finance Faculty, Prentice Hall, New Jersey. Jovanovic, Boyan, 1979, "Job matching and the theory of turnover," Journal of Political Economy 87-5, 972-990. Petersons Annual Guide to Graduate Study, various years, Petersons, New Jersey. Tracy, Joseph and Joel Waldfogel, 1997, "The Best Business Schools: A Market-Based Approach," Journal of Business 70-1, 1-31. U.S. News and World Report: Americas Best Graduate Schools, various years, Washington D.C.: U.S. News and World Report.

26

Warner, Jerold, Ross L. Watts, and Karen H. Wruck, 1988, "Stock prices and top management changes," Journal of Financial Economics 20, 461-92. Weisbach, Michael S., 1988, "Outside directors and CEO turnover," Journal of Financial Economics 20, 431-60. Zimmerman, Jerold, 2001, "Can American Business Schools Survive?" Working paper, University of Rochester.

27

Table 1 Summary Statistics on Sample of Deans from Schools with Ranked MBA Programs
All Schools Number of schools Number of dean-school matches Number of dean-school-year observations Percentage of schools that are public Percentage of schools that are private Mean dean age (in years) Mean dean tenure (in years) Percentage of deans who are insiders Percentage of deans who are outsiders Percentage of deans who are academics 60 121 521 45.49% 54.51% 52.887 5.092 38.15% 61.85% 83.04% Top 25 Schools 36 76 313 34.19% 65.81% 53.195 4.907 43.27% 56.73% 77.24% Top 25 Schools 1990-1995 32 44 141 34.75% 65.25% 52.879 5.560 47.52% 52.48% 77.30% Top 25 Schools 1996-2002 33 54 172 33.72% 66.28% 53.453 4.372 39.77% 60.23% 77.19 %

Percentage of deans who are non-academics 16.96% 22.76% 22.70% 22.81 % Note: The table includes data on all deans who served at business schools with MBA programs ranked by U.S. News and World Report (USN) or Business Week (BW) in their annual ranking of programs from 1990 to 2002. For each school that is ranked in a given year, we report data for the dean who is leading that school in the September of the calendar year of the ranking. We exclude from the figures all observations where the dean was serving on an interim basis (totaling 40 dean-school-year observations). Data on deans and schools is collected from a variety of school directories/guides and by Factiva news searches. All dean characteristics are calculated over the entire set of dean-school-year observations. Age is inferred from the year an individual received their highest degree where we assume an individual was 21 when they earned a bachelor's degree, 26 when they earned a master's degree, and 27 when they earned a doctorate. Tenure is the number of years the individual had served as dean as of the observation date. Insiders are all individuals who were employed by the school immediately before becoming dean, with all other deans being considered outsiders. A dean's academic status is determined by examining whether he/she was employed by an academic institution immediately prior to becoming dean. A school's Top 25 status is determined by their USN ranking.

28

Table 2 Turnover Statistics for Business School Deans


All Schools Panel A: Frequency of turnover Number of observations Number of dean departures Annual turnover rate Implied length of career - 25th percentile Implied length of career - 50th percentile Implied length of career - 75th percentile Panel B: Classification of turnover Generic departure Forced Death Moves up Moves down 471 61 12.95% 4 8 13 62.30% 9.84% 3.28% 18.03% 6.56% Top 25 Schools 287 42 14.63% 4 8 11 64.29% 9.52% 0.00% 21.43% 4.76% Top 25 1990-1995 141 20 14.18% 4 8 11 55.00% 20.00% 0.00% 20.00% 5.00% Top 25 1996-2001 146 22 15.07% 4 8 11 72.73% 0.00% 0.00% 22.73% 4.55%

Panel C: Background on replacement Inside replacement 36.67% 38.10% 30.00% 45.45% Outside replacement 63.33% 61.90% 70.00% 54.55% Academic replacement 83.33% 76.19% 65.00% 86.36% Non-academic replacement 16.67% 23.81% 35.00% 13.64% Initially replaced by interim dean 36.07% 30.95% 35.00% 27.27% Note: The entire sample consists of all dean-school years for non-interim deans serving at schools ranked by USN or BW in that calendar year. The table reports figures on whether a dean departs from a school over the year measured on a September 1st to September 1st basis. Schools that are ranked 25 or better at the start of a given year are assigned to the Top 25 group. The 1990-1995 (1996-2001) column reports turnover behavior for observation-years that begin in calendar 1990 through 1995 (1996 through 2001). The implied career length data is calculated by treating each dean-school match as a single observation and estimating survival times in a hazard model that accounts for right censoring arising from the fact that for many deans we do not observe their completed tenure. For each dean departure we collect information from Factiva news searches and assign the departure to one of five mutually exclusive categories. Turnovers were classified as forced if there was evidence that the dean was pressured to leave because of disputes with university administrators or faculty. Non-forced departures where a dean takes a dean position at a school that is ranked lower than the school they leave is assigned to the "move down" category. All other external moves are assigned to the "move up" category. This category is composed entirely of cases where an individual assumed the deanship of a higher ranked business school, took a university president position, or was hired as a CEO or CFO of a large corporation. All departures that are not in these categories and were not deaths are assigned to the generic departure category. These departures are typically described as motivated by a desire to return to teaching or research or to pursue other interests. For cases in which an interim dean is appointed, the figures for replaced by an insider/outsider and academic/non-academic refer to the eventual permanent dean replacement.

29

Table 3 Summary Statistics on Changes in Business School Rankings


Panel A: Magnitudes of ranking changes Absolute value of 1-year change in USN ranking Absolute value of 2-year change in USN ranking 25th percentile of 2-year change in USN ranking 75th percentile of 2-year change in USN ranking Absolute value of 2-year change in BW ranking 25th percentile of 2-year change in BW ranking 75th percentile of 2-year change in BW ranking Panel B: Correlations and autocorrelations in rankings Mean/Level 2.908 3.187 -2.00 2.00 2.822 -2.00 2.00 Median 2.000 2.000 2.000 Number of Observations 476 419 419 419 129 129 129

Correlation/ p-value Number of Autocorrelation Observations Correlation of levels of USN and BW rankings .832 .000 167 Correlation of changes in 2-year USN and BW rankings .007 .949 100 First-order autocorrelation of changes in 2-year USN ranking -.200 .008 178 First order autocorrelation of changes in 2-year BW rankings -.380 .000 100 Note: All ranking data is collected from the 1990-2002 editions of U.S. News and World Report (USN) and Business Week (BW). The USN rankings are released annually in March/April, while the BW rankings are released bi-annually in October. The 1-year USN change is calculated as the change in rank between two successive annual editions. The 2-year USN change is calculated as the change in rank between the year t edition and the year t+2 edition. The 2-year BW change is the change in rank between two successive biannual BW rankings releases. In correlating the two sources of ranking data, we match the October of year t BW ranking for a given school with the March of year t+1 USN ranking of that school. In calculating autocorrelations for the 2-year USN data, we only use rankings releases from even years (i.e., 1990, 1992, 1994, etc.) For cases where a group of schools are tied for a certain rank, we assign them all a rank equal to the midpoint of the rank of the schools immediately above and immediately below the group. Thus, if two schools were tied with a rank of 10, we would assign them both a rank of 10.5.

30

Table 4 Logit Models of Business School Dean Turnover and Business School Rankings
(1) USN Lagged Performance USN Contemporaneous Perf. USN 2-year Performance BW 2-year Performance Dean's tenure Dean's age Year Constant Number of observations Pseudo R2 Log Likelihood 0.036 (0.035) 0.079** (0.038) -0.021 (0.060) 35.821 (120.16) 337 0.051 -114.57 0.055* (0.033) 0.070** (0.036) -0.006 (0.053) 5.608 (106.29) 376 0.055 -126.35 0.050 (0.036) 0.066* (0.038) -0.029 (0.061) 51.629 (121.04) 325 0.056 -110.63 -0.055 (0.043) -0.029 (0.039) -0.057 (0.037) -0.232** (0.109) 0.086 (0.065) -0.006 (0.071) -0.113 (0.103) 223.382 (204.81) 73 0.119 -36.88 -0.006 (0.115) -0.233** (0.109) 0.086 (0.066) -0.006 (0.072) -0.112 (0.103) 222.676 (205.199) 73 0.119 -36.88 (2) (3) (4) (5)

Implied probability of turnover at varying performance levels US BW 10th Percentile 0.128 0.115 0.133 0.541 0.230 0.541 25th percentile 0.113 0.108 0.115 0.317 0.226 0.317 50th percentile 0.103 0.102 0.104 0.226 0.224 0.226 75th percentile 0.093 0.097 0.094 0.155 0.222 0.155 90th percentile 0.082 0.091 0.082 0.084 0.219 0.084 Note: All estimates are logit estimates with asymptotic standard errors reported in parentheses under the coefficient estimates. In specifications (1), (2), and (3) turnover is measured over a 1-year window commencing on September 1st of every year. In specifications (4) and (5) turnover is measured over a 2-year window commencing on September 1st of every odd year. In all specifications the dependent variable assumes a value of 1 for all observations were a dean departs for reasons other than death or "moves up" and 0 when the dean does not turnover. The dependent variable for deaths and moves up is assigned a missing value. The USN Lagged performance measure is calculated as the change in the USN ranking in the 1-year period ending in the March issue immediately prior to the start of the observation year. The USN Contemporaneous performance measure is the change in the USN ranking in the 1-year period ending in the March that falls within the observation year. In column (3) the USN 2-year performance measure is the change in the USN ranking in the 2-year period ending in the March that falls within the observation year. In columns (4) and (5) where we consider turnover between September of year t and year t+2, we calculate the BW (USN) performance measure as the change in the school's BW (USN) rank between October (March) of year t-1 and October (March) of year t+1. Ranking ties are treated as described in Table 3 and all other variables are defined as in the earlier tables. The implied probabilities are calculated holding all other variables at their means and varying the level of the performance metric used in each respective column. All estimated models exclude deans who served for less than one year as of the start of the turnover observation period. We measure rankings performance as ranking at the start of the period minus ranking at the end of the period. Thus, improvements in rankings are indicated by a positive level of performance. * Significant at the 10% level, **Significant at the 5% level, *** Significant at the 1% level

31

Table 5 Logit Models of Business School Dean Turnover and Placement Success
(1) Lagged Placement Performance Contemp. Placement Performance 2-year Placement Performance Dean's tenure Dean's age Year Constant Number of observations Pseudo R2 Log Likelihood 0.055 (0.041) 0.052 (0.042) -0.015 (0.060) 25.04 (118.83) 245 0.050 -88.38 0.054 (0.040) 0.057 (0.041) -0.020 (0.055) 35.07 (110.36) 283 0.061 -93.79 -0.042 (0.033) -0.057* (0.032) -0.080** (0.034) 0.072 (0.044) 0.044 (0.044) -0.036 (0.062) 66.58 (123.22) 223 0.076 -75.35 (2) (3)

Implied probability of turnover at varying performance levels 10th Percentile 0.159 0.156 0.214 25th percentile 0.133 0.122 0.134 50th percentile 0.115 0.100 0.101 75th percentile 0.102 0.085 0.075 90th percentile 0.078 0.058 0.056 Note: All estimates are logit estimates with asymptotic standard errors reported in parentheses under the coefficient estimates. In all specifications turnover is measured over a 1-year window commencing on September 1st. The dependent variable assumes a value of 1 for all observations were a dean departs for reasons other than death or "moves up". All placement variables are based on a USN placement score we create based on the USN placement ranking and/or placement data. Details are provided in the appendix. The Lagged placement performance measure is calculated as the change in the placement score in the 1year period ending in the March immediately prior to the start of the observation year. The Contemporaneous placement performance measure is the change in the placement score in the 1-year period ending in the March that falls within the observation year. The 2-year placement performance measure is the change in the placement score in the 2-year period ending in the March that falls within the observation year. All other variables are defined as in the earlier tables. The implied probabilities are calculated holding all other variables at their means and varying the level of the performance metric used in each respective column. All estimated models exclude deans who served for less than one year as of the start of the turnover observation period. * Significant at the 10% level, **Significant at the 5% level, *** Significant at the 1% level

32

Table 6 Change in Program Characteristics After Dean Replacements


New Dean Median Change in USN Rank Median Absolute Value Change in USN Rank Median Change in BW Rank Median Absolute Value Change in BW Rank 0.25 3.25 -0.50 2.50 No New Dean 0.00 2.00 0.00 2.00 Difference 0.25 (0.407) 1.25 (0.071) -0.50 (0.822) 0.50 (0.430)

Median Change in USN Placement Score 2.00 -1.00 3.00 (0.236) Median Absolute Value Change in Placement Score 5.50 3.50 2.00 (0.101) Median Change in Enrollment 1.76% 0.93% 0.83% (0.327) Median Absolute Value Change in Enrollment 5.95% 6.29% -0.34% (0.743) Median Change in Tuition 11.06% 12.72% -1.66% (0.193) Note: Figures in the New Dean column are median changes over the two-year period starting with the first BW (for BW rankings and enrollment) or USN (for USN rankings and tuition) rankings release after the appointment of a new permanent dean. Figures in the No New Dean column are median changes over two-year periods following an academic year beginning in which no new dean had been appointed. Tuition almost always increases, so an absolute value calculation for tuition changes is not meaningful. Figures in parentheses in the Difference column are p-values from a Wilcoxon rank-sum test.

33

Das könnte Ihnen auch gefallen