Retention and Graduation Rate Analysis Prepared for Clarion University of Pennsylvania April 2013 In the following report, Hanover Research analyzes the retention and graduation rates of students entering Clarion University of Pennsylvania between 2006 and 2011. More specifically, we investigate various academic, institutional, and demographic factors that significantly influence retention and graduation rates. © 2013 Hanover Research | District Administration Practice 1 Hanover Research | April 2013 TABLE OF CONTENTS Executive Summary and Key Findings ............................................................................... 3 Introduction ...........................................................................................................................3 Discussion ..............................................................................................................................3 Structure of the Report..........................................................................................................4 Key Findings ...........................................................................................................................4 Factors Promoting Retention.............................................................................................4 Factors Promoting Retention for One Additional Year......................................................5 Factors Correlated with Graduation in Four Years ............................................................5 Factors Correlated with Graduation in Five Years .............................................................6 Section I: Methodology .................................................................................................... 7 Regression Analysis................................................................................................................7 Dependent (Outcome) Variable ........................................................................................8 Independent (Explanatory) Variables ..............................................................................12 Section II: Regression Analysis ........................................................................................ 19 Regression Results: Set 1 Models ........................................................................................19 Key Findings Across Set 1 Models....................................................................................20 Model 1 (2nd Year Retention) ...........................................................................................20 Model 2 and 3 (3rd and 4th Year Retention) .....................................................................20 Model 4 (5th Year Retention) ...........................................................................................21 Model 5 (6th Year Retention) ...........................................................................................21 Regression Results: Set 2 Models ........................................................................................23 Key Findings Across Set 2 Models....................................................................................23 Model 6 (3rd Year Conditional Retention)........................................................................23 Model 7 (4th Year Conditional Retention)........................................................................24 Model 8 (5th Year Conditional Retention)........................................................................24 Model 9 (6th Year Conditional Retention)........................................................................24 Regression Results: Set 3 Models ........................................................................................26 Model 10 (Graduation within Four Years) .......................................................................26 Model 11 (Graduation within Five Years) ........................................................................27 Appendix A: Retention Rates .......................................................................................... 29 Appendix B: Graduation Rates ........................................................................................ 32 © 2013 Hanover Research | District Administration Practice 2 Hanover Research | April 2013 EXECUTIVE S UMMARY AND KEY FINDINGS INTRODUCTION In the following report, Hanover Research investigates the factors contributing to bachelor’s degree-seeking students’ retention and graduation rates at Clarion University of Pennsylvania (Clarion). In particular, we use the data supplied by Clarion to assess which demographic, institutional, and academic variables proved the most useful predictors of four-year and five-year graduation and second, third, fourth, fifth, and sixth year retention for students entering in fall 2006 through 2011. In our analysis of Clarion’s data, we developed three sets of models. The first set analyzes factors associated with higher retention, with a focus on factors Clarion could know before a student begins their studies at the university. Next, the second set of models analyzes factors that can help a student be retained for an additional year, given that they were retained through the previous year. Lastly, the third set of models analyzes factors associated with increased graduation rates within four years and five years of enrollment at the university. DISCUSSION In terms of the most notable outcomes of our analysis, we find that a student’s high school ranking is a strong predictor of retention and graduation. A higher high school rank leads to higher retention, and ultimately a higher likelihood of graduation within four or five years of enrollment. This finding suggests that as Clarion seeks to boost retention and graduation rates, the university should strive to increase its admission rates of higher ranked students. As Clarion explained to Hanover, the university has found high school GPA to be a helpful predictor of retention and graduation in the past. In the datasets provided to Hanover, an extremely large number of students had missing values for the high school GPA variable (listed as “0” in the files) and therefore high school GPA was excluded from our analysis. Nevertheless, the finding that high school ranking (a reasonable proxy for high school GPA) is a strong predictor of retention aligns well with Clarion’s experience. Unfortunately, we also find that minorities, particularly black students, have a lower likelihood of retention and graduation within four or five years at Clarion. This finding holds true even after controlling for high school ranking, age, gender, and other observable variables. Based on this result, Hanover recommends a review of why black students are not being retained at higher rates at the university, with the goal of uncovering strategies to better support these students. Other notable takeaways from our analysis include the finding that being a student athlete increases a student’s likelihood of retention in their third and fourth year, as well as © 2013 Hanover Research | District Administration Practice 3 Hanover Research | April 2013 improves their chances of graduating within five years at Clarion. Perhaps unsurprisingly, having a higher first year GPA also boosts a student’s chances of retention and graduation. With regard to coursework, we find that if a student attempts a higher number of credits during the spring semester, they are more likely to be retained the following year. Additionally, if a student fails particular courses – ENG 111, MATH 050, MATH 110, and MATH 112 – they are less likely to be retained by the university in the next year. All of these findings offer insight into concrete predictors of retention and graduation at the university and should assist in identifying groups of students who may need additional support while progressing toward graduation. STRUCTURE OF THE REPORT The report is organized as follows: In Section I, we provide a description of the methodology and data used to evaluate the factors affecting graduation and retention. In Section II, we discuss and present our findings from the regression models that quantify the effects of individual academic, institutional, and demographic factors on students’ likelihood of retention and graduation. Further, in Appendix A and B, we present cross-tabulations of retention and graduation rates broken down by students’ gender, ethnicity, and high school class rank. Before proceeding to the body of the report, below we offer a more detailed breakdown of the key findings of our analysis. KEY FINDINGS FACTORS PROMOTING RETENTION The following factors were found to be key predictors of retention: 1) Ranking in the top quarter of a high school class – For example, a student ranked in the top 10 percent of his or her class is 19 percent more likely to be retained in their second year at Clarion compared to a student whose ranking is within the 50-75 percent range of their high school class.1 2) Achieving higher SAT scores – While higher SAT scores were linked to an increased likelihood of being retained, the magnitude of this effect was fairly small. For instance, for every 100-point increase in total SAT score (Math, Reading, and Writing combined), a student’s probability of being retained in his or her second year is expected to increase by less than 1 percent. Additionally, minorities, particularly black students, are found to have lower retention rates. Holding other factors equal, our models indicate that a black student would be 12 to 19 percent less likely to be retained than a white student. 1 Note that the high school class rank measure calculates a given student’s rank as a percentage of the number of nd students in his or her high school class. For example, a student who is ranked 32 out of a class of 100 would have a value of 32 percent on this measure. © 2013 Hanover Research | District Administration Practice 4 Hanover Research | April 2013 FACTORS PROMOTING RETENTION FOR ONE ADDITIONAL YEAR Described as “conditional retention” in our report, the following factors are associated with an increased likelihood of being retained for an additional year if a student has completed their previous year of study at Clarion. 1) Achieving a higher first year GPA at Clarion – A one-point increase in first year GPA (e.g., 2.0 to 3.0) is associated with a roughly 3-4 percent increase in the likelihood of being retained for an additional year. 2) Being a student athlete in the previous year – The chances of an athlete being retained in the third or fourth year at Clarion is approximately 3-6 percent higher than that of a non-athlete. 3) Attempting a higher number of credits in the previous semester – For example, a one-credit increase in the number of credits attempted in the previous spring semester is associated with a 4 percent increase in the likelihood of being retained in the third year. Further, a three-credit increase would be associated with a 12 percent boost in the likelihood of being retained. Additionally, failing the following subjects is associated with a decreased likelihood of retention for an additional year (if the student has completed the previous year). The estimated magnitude of this decline in likelihood of retention is 8-18 percent depending on the course/year of retention. 1) Failing MATH 110 adversely affects third year retention. 2) Failing ENG 111 or MATH 050 adversely affects third and fourth year retention 3) Failing MATH 112 adversely affects third and fifth year retention. FACTORS CORRELATED WITH GRADUATION IN FOUR YEARS The following factors are associated with a higher likelihood of graduating in four years. 1) Having a higher high school ranking – For example, a student ranked in the top 10 percent of his or her high school class is roughly 12 percent more likely to graduate in four years than a student whose ranking falls between 25-50 percent of the class. 2) Having a higher first year GPA – For every one-point increase in first year GPA, a student’s probability of graduating within four years increases by 15 percent. Further, being black was found to decrease a student’s likelihood of graduation in four years by 11 percent, as compared to being white. © 2013 Hanover Research | District Administration Practice 5 Hanover Research | April 2013 FACTORS CORRELATED WITH GRADUATION IN FIVE YEARS The following factors are associated with a higher likelihood of graduating in five years. 1) Being female – Increases a student’s likelihood of graduating in five years by 10 percent, as compared to being a male. 2) Ranking in the top half of a high school class – For example, a student whose rank falls between 50 and 75 percent of the high school class is 17 percent less likely to graduate within five years than a student in the top 10 percent. 3) Being an athlete – Student athletes are nearly 9 percent more likely to graduate within five years, as compared to non-athletes. Finally, similar to our previous findings on retention and four-year graduation, being black decreases the probability of a student graduating within five years by 13 percent, as compared to being white. © 2013 Hanover Research | District Administration Practice 6 Hanover Research | April 2013 SECTION I: METHODOLOGY In this section, we offer an overview of the data and methodology used in our analysis. Clarion provided Hanover with seven different datasets (a separate dataset for each year), containing demographic, academic, and institutional information of students between the years 2006 and 2012. More specifically, the data included the following:   Demographic data – Students’ gender, age, and race  Institutional data – Students’ starting year, graduation year, and athlete status Academic data o Pre-university (high school) – Students’ high school class ranking and SAT scores o University – Students’ first GPA, performance in specific courses (ENG 110, ENG 111, MATH 050, MATH 110, MATH 112, and MGMT 120), number of credits earned in each fall semester, and number of credits attempted in each spring semester. Our analysis is based on a master dataset that was created by combining the seven datasets provided by Clarion. From each of the datasets, we selected students who enrolled at Clarion in that year (either freshmen or transfer students). This provided us with a list of 9,131 unique students, who enrolled at Clarion between 2006 and 2011. For each of these students we checked whether they appeared in the subsequent year’s dataset. Students who appeared in the subsequent year’s dataset were marked as retained while students who did not graduate and did not appear in the subsequent year’s dataset were marked as not retained. Note that 299 students were dropped from our combined dataset as these students continued to appear in the Clarion datasets even though they had been identified as graduates of the previous year. Our final dataset available for analysis therefore contains 8,832 unique students. REGRESSION ANALYSIS In order to examine the impact of the various demographic, academic, and institutional factors on retention and graduation rates, we constructed a series of linear probability models (LPMs). Overall, we estimated 11 models grouped as Sets 1, 2, and 3, based on the dependent (i.e., outcome) variable used.2 LPMs represent a variation of the standard ordinary least squares (OLS) regression model used to analyze dichotomous dependent variables. Dichotomous variables assume one of two values; in the context of the present analysis, we assign a value of 1 to our dependent variable in cases where students meet a specific criterion (e.g., having been retained or having graduated) and a value of 0 otherwise (having not been retained or having not graduated). 2 As described in greater detail below, the dependent variables examined in this analysis include retention (corresponding to Set 1), “conditional” retention (Set 2), and graduation (Set 3). © 2013 Hanover Research | District Administration Practice 7 Hanover Research | April 2013 For each of our regression models, we provide coefficient estimates, as well as an indication of which coefficients proved statistically significant. The coefficients of our models reveal how much we expect the dependent variable to change when the independent variable increases by one unit, holding all of the other independent variables in the model constant. In the context of LPMs, the coefficients capture the change in the likelihood of retention or graduation, depending on the model estimated. More specifically, multiplying the coefficients by 100 percent indicates the extent to which the likelihood changes due to a one-unit increase in the associated independent variable. For example, as seen in Figure 2.2, a one-unit increase in first GPA leads to a 4.21 percent rise in the probability of first year retention in Model 6. Lastly, when presenting our results, we also include the R-squared value, which reveals the percentage of variation in the dependent variable accounted for by the model. Using Figure 2.2 and Model 6 as an example once again, the R-squared value of 0.3284 indicates that the model can explain 32.84 percent of the variation in retention. DEPENDENT (OUTCOME) VARIABLE As discussed previously, we used three types of dependent, or outcome, variables in our regression analysis. We describe the dependent variables as follows:  Set 1 (Models 1, 2, 3, 4, and 5): The dependent variable in this set of models indicates whether or not a student (who did not yet graduate) was retained in a particular year of study. For instance, the second year retention of a student who initially enrolled at Clarion in 2006 indicates whether or not the student was retained in 2007. Similarly, the third year retention of that same student indicates whether he or she was retained in 2008. Please note that a particular year’s retention, as defined by this outcome variable, is independent of the students’ retention in the immediate previous year.3 Figure 1.1 below helps understand what a particular year’s retention means for students who enrolled at Clarion between 2006 and 2011. We estimate separate models for different years of retention. STARTING 3 Figure 1.1: Retention Years 2 YEAR 3RD YEAR 4TH YEAR 5TH YEAR ND 6TH YEAR YEAR RETENTION RETENTION RETENTION RETENTION RETENTION 2006 2007 2008 2009 2007 2008 2009 2010 2008 2009 2010 2011 2009 2010 2011 2012 2010 2011 2012 -- 2011 2012 --- In other words, if a student initially enrolled in 2006 but did not re-enroll in 2007, they would be marked as retained under “second year retention.” However, if this same student re-enrolled in 2008, they would then be marked as retained under “third year retention.” As discussed in greater detail below, the Set 2 models focus on “conditional retention,” a measure that takes into account whether a student had been enrolled in the previous year at Clarion in addition to whether they were enrolled in the current year. © 2013 Hanover Research | District Administration Practice 8 Hanover Research | April 2013 STARTING 2ND YEAR 3RD YEAR 4TH YEAR 5TH YEAR 6TH YEAR YEAR RETENTION RETENTION RETENTION RETENTION RETENTION 2010 2011 2011 2012 2012 -- --- --- --- In Figure 1.2, we provide descriptive statistics of the dependent variables used for each model in Set 1. For each of the student cohorts (e.g., students beginning in fall 2006, students beginning in fall 2007, etc.), we provide the percentage of students who were retained and not retained. The table indicates that for all cohorts, the retention rate declines over the years. Please note that we have also provided a breakdown of retention rates by various subgroups in Appendix A. Figure 1.2: Distribution of Dependent Variables in Set 1 – Descriptive Statistics STARTING TERM Fall 2006 Fall 2007 Fall 2008 Fall 2009 Fall 2010 Fall 2011  STATUS Not Retained Retained Count Not Retained Retained Count Not Retained Retained Count Not Retained Retained Count Not Retained Retained Count Not Retained Retained Count Total 2ND YEAR 3RD YEAR 4TH YEAR 5TH YEAR 6TH YEAR RETENTION RETENTION RETENTION RETENTION RETENTION 32% 68% 1,436 29% 71% 1,437 31% 69% 1,416 31% 69% 1,582 30% 70% 1,527 31% 69% 1,418 8,816 44% 56% 1,414 39% 61% 1,421 43% 57% 1,388 41% 59% 1,561 42% 58% 1,495 ---7,279 50% 50% 1,346 44% 56% 1,365 49% 51% 1,326 48% 52% 1,499 ------5,536 69% 31% 1,013 67% 33% 979 71% 29% 989 ---------2,981 92% 8% 792 92% 8% 744 ------------1,536 Set 2 (Models 6, 7, 8, and 9): The dependent variable for the models in Set 2 assumes a value of 1 whenever a student who was retained in a particular year is enrolled in the following year and a value of 0 whenever a student who was retained in a particular year is not enrolled in the following year. This is described as “conditional” retention, as being retained in a given year is conditioned on whether a student was retained in the previous year. For instance, in the case of a student who started in 2006, fourth year conditional retention refers to that student being © 2013 Hanover Research | District Administration Practice 9 Hanover Research | April 2013 retained in 2009 given that he/she was also retained in 2008. The table below provides brief clarification of what a particular year’s conditional retention indicates. Figure 1.3: Conditional Retention STARTING YEAR 2006 2007 2008 2009 2010 RD 3 YEAR CONDITIONAL 4TH YEAR CONDITIONAL 5TH YEAR CONDITIONAL 6TH YEAR CONDITIONAL RETENTION RETENTION RETENTION RETENTION Students retained in 2007, did not graduate, and were retained again in 2008 Students retained in 2008, did not graduate, and were retained again in 2009 Students retained in 2009, did not graduate, and were retained again in 2010 Students retained in 2010, did not graduate, and were retained again in 2011 Students retained in 2011, did not graduate, and were retained again in 2012 Students retained in 2008, did not graduate, and were retained again in 2009 Students retained in 2009, did not graduate, and were retained again in 2010 Students retained in 2010, did not graduate, and were retained again in 2011 Students retained in 2011, did not graduate, and were retained again in 2012 Students retained in 2009, did not graduate, and were retained again in 2010 Students retained in 2010, did not graduate, and were retained again in 2011 Students retained in 2011, did not graduate, and were retained again in 2012 Students retained in 2010, did not graduate, and were retained again in 2011 Students retained in 2011, did not graduate, and were retained again in 2012 -- -- -- -- -- -- As the above table illustrates, conditional retention only looks at the group of students who had been retained in the previous year. For example, if a student is marked as “not retained” in the third year, they would not be included in the calculation of fourth year retention (only students who had been retained in the third year would be included). This differs from the dependent variable used in the Set 1 models, where the measure of retention is not conditioned on whether the student was enrolled in the previous year. Figure 1.4 on the following page provides descriptive statistics of the dependent variables used in the Set 2 models. In general, between the third year and fifth year, the conditional retention rate increases slightly. Note that we have also provided a breakdown of the conditional retention rates by various sub-groups in Appendix A. © 2013 Hanover Research | District Administration Practice 10 Hanover Research | April 2013 Figure 1.4: Distribution of Dependent Variables in Set 2 - Descriptive Statistics STARTING YEAR Fall 2006 Fall 2007 Fall 2008 Fall 2009 Fall 2010  STATUS Not Retained Retained Count Not Retained Retained Count Not Retained Retained Count Not Retained Retained Count Not Retained Retained Count Total 3RD YEAR 4TH YEAR 5TH YEAR 6TH YEAR CONDITIONAL CONDITIONAL CONDITIONAL CONDITIONAL RETENTION RETENTION RETENTION RETENTION 21% 79% 956 17% 83% 1,002 19% 81% 952 18% 82% 1,074 19% 81% 1,039 5,023 11% 89% 726 9% 91% 808 10% 90% 729 10% 90% 853 ---3,116 15% 85% 347 18% 82% 381 17% 83% 336 ------1,064 42% 58% 96 45% 55% 95 ---------191 Set 3 (Models 10 and 11): The dependent variable in this set of models indicates whether or not a student graduated either within four years of enrollment (Model 10) or five years of enrollment (Model 11). Figure 1.5 on the following page provides descriptive statistics of this dependent variable. As expected, the graduation rate within five years is higher than the graduation rate within four years, as students have had more time to complete their studies. We have also provided a breakdown of the graduation rates by subgroups in Appendix B. Please note that we do not have complete four-year data for students who enrolled in 2009, 2010, and 2011. Furthermore, we also do not have five-year data for students who enrolled in 2008. For most of these students, we do not know whether they graduated within these years. For students in these years, we take into account the students who are listed as graduates only. Other students are excluded from the analysis as they are missing full four- and five-year data. © 2013 Hanover Research | District Administration Practice 11 Hanover Research | April 2013 Figure 1.5: Distribution of Dependent Variables in Set 3 – Descriptive Statistics STARTING YEAR Fall 2006 Fall 2007 Fall 2008 Fall 2009 Fall 2010 Fall 2011 STATUS GRADUATION WITHIN 4 YEARS GRADUATION WITHIN 5 YEARS Did not graduate Graduated Count Did not graduate Graduated Count Did not graduate Graduated Count Graduated Count Graduated Count Graduated Count Total 70% 30% 1,438 68% 32% 1,437 70% 30% 1,418 100% 131 100% 57 100% 11 4,492 55% 45% 1,438 52% 48% 1,437 -100% 542 100% 131 100% 57 100% 11 3,616 INDEPENDENT (EXPLANATORY) VARIABLES We included a series of independent, or explanatory, variables in the models in order to control for factors affecting the dependent variables of interest (retention, conditional retention, and graduation). We provide a complete list of the independent variables used in Figure 1.6. The table defines each independent variable, indicates the model(s) in which the variable appears, and the variable type. Note that in general, if a variable was not found to have a statistically significant effect on the outcome variable of interest (retention or graduation), it was excluded from the model. For example, the “Female” variable did not have a significant relationship with retention/graduation in Models 1-10 but did have a significant effect on likelihood of graduation within five years. Therefore, the “Female” variable was only included in our final version of Model 11. Figure 1.6: Summary of Independent Variables VARIABLES Female SUMMARY Gender of students: (0) Male (1) Female Age Age of students Age squared Ethnicity The squared value of the age of the students. This allows us to examine whether age has a non-linear effect on the dependent variable. Ethnicity of students, selected from three categories: Black, White, and “other or two or more races.” © 2013 Hanover Research | District Administration Practice MODEL Model 11 Models 3, 5, 7, and 10 VARIABLE TYPE Categorical Models 3 and 10 Continuous Models 1, 2, 3, 4, 6, 8, 10, and 11 Categorical Continuous 12 Hanover Research | April 2013 VARIABLES High School Rank SAT score total SAT Math, Reading, Writing First GPA ENG 111* MATH 050* MATH 110* MATH 112* SUMMARY Student’s high school rank as a percentage of class size. In other words, if a student was ranked 29th out of a high school class of 100, their value for this variable would be 29 percent. A student with a lower ranking in the same high school class, say 45th out of 100, would have a value of 45 percent. As such, higher values for this variable represent lower class ranks (e.g., 45th) while lower values represent higher class rankings (e.g., 29th). Student’s total SAT scores (includes math, reading and writing). We divided SAT score by 100 so that our models show the effect of a 100 point increase in SAT rather than a single point increase. In cases where total SAT score was not found to a statistically significant effect on retention or graduation, we entered individual Math, Reading, and/or Writing scores into the models. These scores remained in the final models if they were found to have a statistically significant effect on the outcome of interest. Student's first GPA at Clarion. Whether student passed, failed, or did not attempt this course. Whether student passed, failed, or did not attempt this course. Whether student passed, failed, or did not attempt this course. Whether student passed, failed, or did not attempt this course. Athlete Whether students is an athlete: (0) No (1) Yes Credits attempted in previous Spring Credits earned in previous Fall Number of credits attempted in the previous Spring semester Number of credits earned in the previous Fall semester MODEL VARIABLE TYPE Models 1, 2, 3, 4, 9, 10, and 11 Categorical/Continuous Models 1, 2, 3, 4, 6, 7, 8 , 9, and 10 Continuous Models 5, 10, and 11 Continuous Models 6, 7, 8, 10, and 11 Continuous Models 6 and 7 Categorical Model 6 and 7 Categorical Models 6 and 7 Categorical Model 6, 7, and 8 Model 6, 7, and 11 Models 6, 7, 8, and 9 Model 6 Categorical Categorical Continuous Continuous *Note that ENG 110 and MGMT 120 were excluded from our models as an extremely high percentage of students in our dataset (90 percent or more for most cohorts) did not attempt these courses. In general, when selecting variables for inclusion in our models, Hanover sought to use variables for which a substantial number of students had non-zero values. Figures 1.7-1.12 present descriptive statistics for each independent variable used in the models. While Figure 1.7 shows the distribution of students’ gender, ethnicity, and high school ranking, Figure 1.8 and 1.9 shows descriptive statistics for some measures of academic performance. In our models, we used two forms of the independent variable that indicates whether a student is an athlete. For the models in Set 2, the athlete variable refers to a student being © 2013 Hanover Research | District Administration Practice 13 Hanover Research | April 2013 an athlete in the year immediately prior to the year in which retention is measured. For instance, Model 7, which analyzes fourth year conditional retention, indicates the effect of a student being an athlete in his/her third year only. By contrast, the athlete variable used in the Set 3 models (Model 11), indicates whether the student was an athlete at any point within the four or five years taken into account. Descriptive statistics of these alternate forms of the athlete variable are provided in Figures 1.10 and 1.11. Finally, Figure 1.12 provides the average number of credits earned (fall) and attempted (spring) for each year of our dataset. © 2013 Hanover Research | District Administration Practice 14 Hanover Research | April 2013 Figure 1.7: Gender, Ethnicity, and High School Ranking – Descriptive Statistics GENDER FALL 2006 FALL 2007 FALL 2008 FALL 2009 FALL 2010 FALL 2011 Female 57% (n=803) 55% (n=789) 56% (n=793) 58% (n=924) 62% (n=941) 60% (n=850) Male 43% (n=604) 45% (n=648) 44% (n=625) 42% (n=661) 38% (n=589) 40% (n=574) Black Other or two or more races4 6% (n=85) 3% (n=37) 7% (n=98) 3% (n=40) 9% (n=125) 4% (n=63) 9% (n=135) 4% (n=55) 9% (n=130) 5% (n=70) 9% (n=122) 6% (n=84) White 91% (n=1281) 90% (n=1287) 87% (n=1225) 88% (n=1354) 86% (n=1276) 85% (n=1198) 0% to 10% 8% (n=108) 9% (n=111) 9% (n=114) 8% (n=104) 9% (n=117) 8% (n=98) 10% to 25% 19% (n=243) 15% (n=189) 19% (n=229) 19% (n=254) 20% (n=260) 18% (n=214) 25% to 50% 35% (n=445) 35% (n=443) 36% (n=443) 35% (n=476) 34% (n=446) 35% (n=413) 50% to 75% 27% (n=349) 29% (n=371) 27% (n=333) 26% (n=349) 26% (n=335) 25% (n=292) 75% to 100% 10% (n=132) 12% (n=156) 9% (n=109) 12% (n=162) 11% (n=145) 13% (n=157) ETHNICITY HIGH SCHOOL RANKING Figure 1.8: Age, High School Rank, SAT Score, and First GPA- Descriptive Statistics VALUES Average of Age Average of Rank as Percentage of Class Size Average of SAT Total* Average of SAT Math* Average of SAT Reading* Average of SAT Writing* Average of First GPA FALL 2006 18.93 43% 1381 462 466 455 2.44 FALL 2007 19.13 45% 1419 483 477 462 2.49 FALL 2008 19.20 42% 1412 479 474 460 2.51 FALL 2009 19.60 44% 1407 477 471 458 2.64 FALL 2010 19.70 43% 1419 478 477 464 2.66 FALL 2011 19.76 44% 1402 478 475 459 2.63 TOTAL 19.39 43% 1408 477 474 460 2.56 *Note that all SAT scores listed as “0” were recoded as missing and are not reflected in these averages. 4 Includes students who are listed as Asian, Hispanic, Native American, Pacific Islander, and from two or more races. © 2013 Hanover Research | District Administration Practice 15 Hanover Research | April 2013 Figure 1.9: University Course Performance – Descriptive Statistics COURSE ENG 110 ENG 111 MATH 050 MATH 110 MATH 112 MGMT 120 VALUES Did not attempt Did not pass Passed Did not attempt Did not pass Passed Did not attempt Did not pass Passed Did not attempt Did not pass Passed Did not attempt Did not pass Passed Did not attempt Did not pass Passed FALL 2006 99% (n=1429) 0% (n=) 1% (n=9) 27% (n=393) 10% (n=137) 63% (n=908) 79% (n=1135) 5% (n=75) 16% (n=228) 83% (n=1200) 4% (n=64) 12% (n=174) 76% (n=1099) 4% (n=63) 19% (n=276) 92% (n=1327) 1% (n=15) 7% (n=96) © 2013 Hanover Research | District Administration Practice FALL 2007 99% (n=1425) 0% (n=3) 1% (n=9) 21% (n=304) 11% (n=161) 68% (n=972) 81% (n=1165) 4% (n=62) 15% (n=210) 81% (n=1168) 4% (n=64) 14% (n=205) 77% (n=1112) 4% (n=57) 19% (n=268) 89% (n=1277) 2% (n=27) 9% (n=133) FALL 2008 91% (n=1288) 2% (n=22) 8% (n=108) 25% (n=351) 10% (n=148) 65% (n=919) 80% (n=1141) 5% (n=64) 15% (n=213) 83% (n=1176) 5% (n=71) 12% (n=171) 76% (n=1075) 5% (n=72) 19% (n=271) 88% (n=1251) 1% (n=18) 11% (n=149) FALL 2009 95% (n=1499) 1% (n=15) 4% (n=71) 26% (n=407) 8% (n=129) 66% (n=1049) 80% (n=1263) 4% (n=60) 17% (n=262) 87% (n=1375) 4% (n=62) 9% (n=148) 73% (n=1159) 4% (n=71) 22% (n=355) 92% (n=1458) 1% (n=13) 7% (n=114) FALL 2010 85% (n=1304) 2% (n=37) 12% (n=189) 31% (n=481) 8% (n=120) 61% (n=929) 78% (n=1193) 5% (n=78) 17% (n=259) 89% (n=1361) 4% (n=54) 8% (n=115) 72% (n=1105) 5% (n=75) 23% (n=350) 90% (n=1375) 2% (n=31) 8% (n=124) 16 FALL 2011 84% (n=1192) 1% (n=20) 15% (n=212) 32% (n=459) 7% (n=105) 60% (n=860) 84% (n=1191) 3% (n=47) 13% (n=186) 89% (n=1263) 4% (n=51) 8% (n=110) 76% (n=1081) 4% (n=63) 20% (n=280) 90% (n=1281) 2% (n=26) 8% (n=117) TOTAL 3,495 675 4,662 8,137 97 598 2,395 800 5,637 7,543 366 923 6,631 401 1,800 7,969 130 733 Hanover Research | April 2013 Figure 1.10: Athlete in the Year Prior to Retention – Descriptive Statistics 2006 2007 2008 2009 2010 ATHLETE STATUS Athlete prior to 2nd year retention Athlete prior to 3rd year retention Athlete prior to 4th year retention Athlete prior to 5th year retention NO YES TOTAL NO YES TOTAL NO YES TOTAL NO YES TOTAL NO YES TOTAL 92% 8% 978 92% 8% 1,018 91% 9% 980 92% 8% 1,095 92% 8% 1,071 94% 6% 792 93% 7% 864 91% 9% 788 93% 7% 914 93% 7% 861 96% 4% 675 94% 6% 766 91% 9% 670 93% 7% 780 -- -- -- 98% 2% 311 97% 3% 322 96% 4% 289 -- -- -- -- -- -- © 2013 Hanover Research | District Administration Practice 17 Hanover Research | April 2013 Figure 1.11: Athlete Before Graduation – Descriptive Statistics Fall 2006 Fall 2007 Fall 2008 Fall 2009 Fall 2010 Fall 2011 Total GRADUATION WITHIN 4 YEARS YES NO TOTAL 8% 92% 1,024 8% 92% 1,063 10% 90% 1,003 8% 92% 1,134 8% 92% 1,089 ---8% 92% 5,313 GRADUATION WITHIN 5 YEARS YES NO TOTAL 26% 74% 322 25% 75% 342 30% 70% 323 12% 88% 760 10% 90% 863 0% 100% 1,424 11% 89% 4,034 Figure 1.12: Average Credits Earned and Attempted 2007 2008 2009 2010 2011 2012 Fall Credits Earned 14.74 14.81 14.68 14.48 14.31 14.11 Spring Credits Attempted 13.22 13.42 13.34 13.18 12.89 12.32 © 2013 Hanover Research | District Administration Practice 18 Hanover Research | April 2013 SECTION II: R EGRESSION A NALYSIS In this section, we present the results of our regression models that help us understand the factors that affect student retention and graduation rates at Clarion. As noted previously, the models included in Set 1 and Set 2 examine students’ retention rates between the second and sixth years at the university, while the models in Set 3 examine students’ fouryear and five-year graduation rates. The results of each set of regression models are presented in Figures 2.1, 2.2., and 2.3. Each set of models focuses on a different combination of variables. More specifically,  The models included in Set 1 (Figure 2.1) were constructed to examine the effects of a variety of factors Clarion would be aware of prior to a student’s initial enrollment at the university (e.g., gender, age, ethnicity, high school rank, and SAT). The outcome variable for these Set 1 models is retention.  The Set 2 models (Figure 2.2) analyze the effects of a range of demographic, preuniversity, and university characteristics (e.g., age, gender, high school rank, SAT, first GPA at Clarion, passage/failure of specific courses, athlete status, credits attempted and earned) on conditional retention. As mentioned previously, conditional retention takes into account (or is “conditioned on”) whether a student was retained in the previous year. In other words, it only measures the retention of students in a given year if they were also retained in the previous year.  Finally, the Set 3 models (Figure 2.3) analyze the effects of demographic, preuniversity, and university characteristics on graduation within four years and graduation within five years. REGRESSION RESULTS: SET 1 MODELS We recall that Set 1, which covers Models 1 to 5, included a dependent variable showing the proportion of students who re-enrolled at Clarion in each of the five years. We developed separate models for the different years of retention (e.g., second year retention, third year retention, fourth year retention, etc.) with only demographic and pre-university data included in the models. Further, in general, the models only include variables that have a statistically significant effect on our dependent variables5 and in the points below, we only highlight findings that are statistically significant. 5 In the figures, statistical significance is denoted with asterisks: *** denotes statistical significance at 1 percent – meaning that there is a less than 1 percent likelihood that the observed effect is due to chance. ** denotes statistical significance at 5 percent – meaning that there is a less than 5 percent likelihood that the observed effect is due to chance. * denotes statistical significance at 10 percent – meaning that there is a less than 10 percent likelihood that the observed effect is due to chance. © 2013 Hanover Research | Academy Administration Practice 19 Hanover Research | April 2013 KEY FINDINGS ACROSS SET 1 MODELS The following subsections highlight findings relative to each individual model. In terms of broader points that can be drawn across the Set 1 models, we find that in general, the following are associated with a higher likelihood of being retained:   Being ranked in the top quarter of a high school class Having higher SAT scores By contrast, minorities, and particularly black students, are less likely to be retained than white students. MODEL 1 (2ND YEAR RETENTION)  Relative to white students, black students and students from other racial/ethnic backgrounds are less likely to be retained in their second year at Clarion.  Students ranked below the top 25 percent of their high school classes are less likely to be retained in their second year.6 In fact, outside of the top 25 percent, the lower the student’s ranking, the less likely the student is to be retained (e.g., students whose ranking is within 50-75 percent of their high school class are less likely to be retained than students whose ranking is within 25-50 percent of their high school class).  Second year retention has a positive relationship with SAT scores. With every 100point increase in SAT scores, a student’s chances of being retained increase slightly (by 0.78 percent). MODEL 2 AND 3 (3RD AND 4TH YEAR RETENTION) 6  Black students and students from other ethnic backgrounds are less likely to be retained in their third and fourth years than white students.  Students ranked below the top 25 percent of their high school classes are less likely to be retained in their third and fourth years than students ranked in the top 10 percent. Similar to Model 1, outside of the top 25 percent, the lower a students’ high school rank, the less likely they are to be retained. Note that we make this conclusion based on a combination of results. First, high school rank was separated into a series of dummy variables where the reference category was a high school rank within the top 10 percent of a students’ high school class. For example, the negative coefficient in Model 1 for the variable listed as “High School Rank (25%-50%)” indicates that students whose ranking was among 25-50 percent of their high school class were significantly less likely to be retained than students in the top 10 percent. The same could be said for students in the 50-75 percent range and the 75-100 percent range. However, the variable measuring “High School Rank (10%25%)” was not found to a have a statistically significant effect (i.e., students in the 10-25 percent range were not significantly less likely to be retained than students in the top 10 percent). Taken together, these results suggest that students ranked within the top 25 percent of their high school class are more likely to be retained than students with lower rankings. © 2013 Hanover Research | Academy Administration Practice 20 Hanover Research | April 2013  Students with higher SAT scores are more likely to be retained in their third year at Clarion. Once again, a 100-point increase in SAT scores increases a student’s chances of being retained by a relatively small margin (i.e., a boost of 0.84 percent in their likelihood of being retained in their third year). This effect was not observed with regard to fourth year retention.  Older students are less likely to be retained in their fourth year at Clarion than younger students. This finding did not hold true with regard to third year retention. MODEL 4 (5TH YEAR RETENTION)  Similar to the previous models, black students and students of other ethnicities are less likely to be retained in their fifth year than their white classmates.  Students ranked in the lower half of their high school classes (50-100 percent) are less likely to be retained compared to students in the top half of their high school classes. MODEL 5 (6TH YEAR RETENTION)  Similar to Model 3 (fourth year retention), older students are less likely to be retained in their sixth year than younger students.  While total SAT scores were not found to have an effect on sixth year retention, higher SAT Reading scores were associated with a higher likelihood of retention and higher SAT Writing scores were associated with a lower likelihood of retention. Note, however, these effects appear fairly small in terms of magnitude. A 100-point increase in Reading scores corresponds to a 3.43 percent increase in the likelihood of being retained in the sixth year. A 100-point increase in Writing scores corresponds to a 4.42 percent decrease in the likelihood of being retained. © 2013 Hanover Research | Academy Administration Practice 21 Hanover Research | April 2013 Figure 2.1: Regression Output Results - Set 1 MODEL 1 ND 2 YEAR MODEL 2 RD 3 YEAR MODEL 3 TH 4 YEAR MODEL 4 TH 5 YEAR MODEL 5 TH 6 YEAR RETENTION RETENTION RETENTION RETENTION RETENTION Age Age squared -0.0584*** -0.0281* (0.0121) (0.017) 0.0008*** (0.0002) Black 7 Other or two or more races 7 High School Rank (10%-25%) 8 8 High School Rank (25%-50%) 8 High School Rank (50%-75%) High School Rank (75%-100%) SAT Total 8 -0.1362*** -0.1864*** -0.1961*** -0.1231*** (0.022092) (0.0261) (0.0268) (0.0329) -0.0681** -0.0711* -0.0766* -0.1194** (0.032796) (0.0409) (0.0459) (0.0588) -0.0124 -0.0218 -0.0250 0.0379 (0.023772) (0.0284) (0.0297) (0.0465) -0.0718*** -0.0950*** -0.1150*** -0.0096 (0.022793) (0.0272) (0.0272) (0.0424) -0.1857*** -0.2156*** -0.2487*** -0.0722* (0.024141) (0.0288) (0.0279) (0.0425) -0.2757*** -0.3569*** -0.3317*** -0.1442*** (0.028049) (0.0337) (0.0323) (0.0464) 0.0078** 0.0084** (0.003156) (0.0038) SAT Read 0.0343** (0.0158) SAT Write -0.0442*** (0.0165) Constant 0.7294*** 0.6387*** 1.5181*** 0.3791*** 0.6320** (0.055672) (0.0664) (0.1580) (0.0396) (0.3174) Number of observations 5,949 4,907 4,786 2,558 986 R-squared 0.05398 0.0671 0.06993 0.0211 0.009382 Coefficients estimated using Ordinary Least Squares with a linear regression model, with standard errors in parenthesis. Statistical significance indicator using asterisk next to the coefficients, with * = significant at 10%, ** = significant at 5%, and *** = significant at 1%. 7 8 Reference category for ethnicity is white Reference category for high school rank is 0% to 10% © 2013 Hanover Research | Academy Administration Practice 22 Hanover Research | April 2013 REGRESSION RESULTS: SET 2 MODELS Figure 2.2, which includes Models 6, 7, 8, and 9, illustrates the relationship of various factors on the conditional retention of students in their third, fourth, fifth, and sixth year at Clarion. Once again, recall that conditional retention only looks at the retention of students in a given year if they were retained in the previous year. KEY FINDINGS ACROSS SET 2 MODELS In terms of broad findings across the Set 2 models, we note that the following factors are associated with a higher likelihood of being retained for one additional year (if the student had been retained in the previous year):    Having a higher first year GPA Being a student athlete in the previous year Having attempted a higher number of credits in the previous semester Additionally, having failed specific courses is associated with a lower likelihood of retention for one additional year. These include ENG 111 (third and fourth year retention), MATH 050 (third and fourth year retention), MATH 110 (third year retention), and MATH 112 (third and fifth year retention). MODEL 6 (3RD YEAR CONDITIONAL RETENTION)  Compared to white students, black students who were retained in their second year at Clarion are less likely to be retained in their third year at Clarion.  SAT score was found to have a negative relationship with conditional retention. However, similar to our other findings surrounding SATs, the effect appears small. A 100-point increase in SAT score corresponds to a 1.05 percent decrease in the likelihood of third year retention.  Students with higher first year GPAs at Clarion are more likely to be retained in their third year than students with lower GPAs.  Students who failed ENG 111, MATH 050, MATH 110, and MATH 112 in the previous year are less likely to be retained than students who did not attempt these courses.  Students who were listed as athletes in their second year at Clarion are more likely to be retained in their third year than non-athletes.  Attempting more credits during the previous spring semester increases a student’s likelihood of being retained in the third year. A one-credit increase is associated with a 4.07 percent boost in the likelihood of retention (so a three-credit increase would translate to a 12 percent boost). By contrast, the number of credits earned in the previous fall semester was found to have a negative impact on likelihood of retention. However, this effect was small (a one-credit increase in the number of credits earned in the fall was associated with a 0.61 percent decrease in the © 2013 Hanover Research | Academy Administration Practice 23 Hanover Research | April 2013 likelihood of retention), and this finding did not hold for retention in the fourth, fifth, and sixth years (Models 7-9). MODEL 7 (4TH YEAR CONDITIONAL RETENTION)  Among students who were retained in their third year, older students are less likely to be retained in their fourth year than younger students.  SAT score also has a negative effect on the likelihood of retention, though once again, the magnitude of this effect appears small in practical terms. A 100-point increase in SAT scores is associated with a 0.63 percent decrease in the likelihood a student will be retained in their fourth year.  Students with higher first year GPAs are more likely to be retained in their fourth year.  Compared to students who did not attempt ENG 111 and MATH 050, students who failed these courses have a lower chance of being retained in their fourth year.  Students who passed MATH 110 and MATH 112 in the previous year are expected to have a higher retention rate than students who did not attempt these courses.  Students who were listed as athletes in their third year are more likely to be retained in their fourth year.  Students who attempted more credits in the previous spring semester are more likely to be retained in their fourth year. MODEL 8 (5TH YEAR CONDITIONAL RETENTION)   Black students are less likely to be retained in their fifth year than white students.  First year GPA has a positive impact on students’ fifth year conditional retention. As a student’s first GPA increases, he or she is more likely to be retained in the fifth year.  Students who failed MATH 112 are less likely to be retained in their fifth year than students who did not attempt the course.  Students who attempted more credits in the previous spring semester are more likely to be retained in their fifth year. Similar to Model 6, higher SAT scores are associated with a lower likelihood of being retained in the fifth year. A 100-point increase in SATs is associated with a 1.57 percent decrease in the likelihood of retention at this stage. MODEL 9 (6TH YEAR CONDITIONAL RETENTION)  An increase in high school rank as a percentage of class size is associated with a decrease in the likelihood that a student would be retained in the sixth year. In other words, lower ranked students (e.g., a student ranked 75th out of 100) are less © 2013 Hanover Research | Academy Administration Practice 24 Hanover Research | April 2013 likely to be retained in the sixth year than students with higher ranks (e.g., a student ranked at 25th out of 100).  Number of credits attempted in the previous spring semester has a positive impact on students’ sixth year retention. Students who attempted more credits in the previous spring are more likely to be retained in their sixth year. Figure 2.2: Regression Output Results - Set 2 MODEL 6 RD 3 YEAR CONDITIONAL MODEL 7 TH 4 YEAR CONDITIONAL MODEL 8 TH 5 YEAR CONDITIONAL MODEL 9 TH 6 YEAR CONDITIONAL RETENTION RETENTION RETENTION RETENTION Age -0.0286*** (0.0091) Black 9 Other of two or more races 9 SAT -0.0595*** -0.1337*** (0.021) (0.044) -0.0140 -0.0624 (0.0301) (0.073) -0.0105*** -0.0063** -0.0157** (0.003) (0.003) (0.0061) High School Rank as percentage of class size -0.4786*** (0.1566) First GPA ENG 111 in the previous year: Failed 10 ENG 111 in the previous year: Passed 10 MATH 050 in the previous year: Failed 11 MATH 050 in the previous year: Passed MATH 110 in the previous year: Failed 12 MATH 110 in the previous year: Passed MATH 112 in the previous year: Failed 11 12 13 0.0421*** 0.0281*** 0.0461*** (0.0082) (0.0077) (0.0162) -0.1564*** -0.1074*** (0.0289) (0.0331) -0.0184 0.0089 (0.0168) (0.0197) -0.1854*** -0.1779*** (0.0364) (0.0397) -0.0151 -0.0059 (0.0145) (0.0136) -0.1475*** -0.0367 (0.0283) (0.0259) -0.0186 0.0353*** (0.0155) (0.0135) -0.0837*** -0.0256 -0.1296** 9 Reference category for ethnicity is white Reference category for ENG 111 is did not attempt 11 Reference category for MATH 050 is did not attempt 12 Reference category for MATH 110 is did not attempt 13 Reference category for MATH 112 is did not attempt 10 © 2013 Hanover Research | Academy Administration Practice 25 Hanover Research | April 2013 MODEL 6 RD 3 YEAR CONDITIONAL MODEL 7 TH 4 YEAR CONDITIONAL MODEL 8 TH 5 YEAR CONDITIONAL MODEL 9 TH 6 YEAR CONDITIONAL RETENTION RETENTION RETENTION RETENTION (0.0302) (0.0282) (0.0549) 0.0173 0.0452*** 0.036 (0.0132) (0.0114) (0.0266) 0.0641*** 0.0355* (0.0182) (0.0182) 0.0407*** 0.0303*** 0.0294*** 0.0301*** (0.0012) (0.0011) (0.0026) (0.0057) 0.4165*** 1.0009*** 0.5613*** 0.5099*** (0.0588) (0.1801) (0.0917) (0.1067) Number of Observations 3,904 2,538 841 173 R-squared 0.3284 0.278 0.1793 0.1809 MATH 112 in the previous year: Passed Athlete in the previous year 13 14 Credits attempted in the last spring semester Credits earned in the last fall semester -0.0061** (0.0028) Constant Coefficients estimated using Ordinary Least Squares with a linear regression model, with standard errors in parenthesis. Statistical significance indicator using asterisk next to the coefficients, with * = significant at 10%, ** = significant at 5%, and *** = significant at 1%. REGRESSION RESULTS: SET 3 MODELS Finally, we present the results to the regression models that estimate the variations in students’ four-year and five-year graduation rates. MODEL 10 (GRADUATION WITHIN FOUR YEARS) 14  In contrast to the relationship of age with retention, older students are more likely to graduate within four years than younger students.  Black students are less likely to graduate within four years than their white classmates.  Students in the top 25 percent of their high school class are more likely to graduate within four years than other students. Once again, outside of the top 25 percent, the lower a student ranks, the less likely they are to graduate within four years.  First year GPA is positively associated with the likelihood of graduating. In fact, a one-point increase in GPA (e.g., 2.0 to 3.0) is associated with a 15.21 percent increase in the likelihood of graduating within four years.  Total SAT score did not have a significant effect on graduation. However, individual tests did have such an effect. More specifically, SAT Math and SAT Writing scores Reference category is not an athlete © 2013 Hanover Research | Academy Administration Practice 26 Hanover Research | April 2013 had a positive relationship (a 100-point increase in Math is associated with a 2.34 percent increase in likelihood of graduation, while a 100-point increase in Writing is expected to yield a 3.96 percent increase in the likelihood of graduation). By contrast, a 100-point increase in SAT Reading scores is associated with a 6.25 percent decrease in the probability of graduating in four years. MODEL 11 (GRADUATION WITHIN FIVE YEARS)  Female students are 10.13 percent more likely to graduate within five years than male students. Notably, this was the only model in which gender was found to have a significant effect on graduation (or retention).   Black students are less likely to graduate within five years than white students.  First year GPA of students has a positive impact on the probability of five-year graduation. A one-point increase in first GPA is associated with an 8.27 percent increase in the likelihood of graduating within five years.  SAT Reading scores have a negative relationship with the likelihood of graduating within five years. A 100-point increase in SAT Reading scores is associated with a 4.71 percent decrease in the probability of graduation.  Athletes are 8.6 percent more likely to graduate within five years than non-athletes. Students in the top half of their high school class are more likely to graduate within five years compared to students in the bottom 50 percent of their class. Outside of the top 50 percent, the lower a student’s high school rank, the lower their likelihood of graduating within five years. Figure 2.3: Regression Output Results - Set 3 MODEL 10 MODEL 11 GRADUATION WITHIN 4 YEARS GRADUATION WITHIN 5 YEARS Female 0.1013*** (0.0332) Age -1.4003*** (0.377) Age squared 0.0388*** (0.0101) Black 15 Other or two or more races 15 High School Rank (10%-25%) 15 16 16 -0.1153*** -0.1331* (0.0316) (0.0699) -0.0445 -0.0582 (0.0491) (0.1141) -0.0507 -0.0794 Reference category for ethnicity is white Reference category for high school rank is 0% to 10% © 2013 Hanover Research | Academy Administration Practice 27 Hanover Research | April 2013 High School Rank (25%-50%) 16 High School Rank (50%-75%) 16 High School Rank (75%-100%) SAT Math 16 MODEL 10 MODEL 11 GRADUATION WITHIN 4 YEARS GRADUATION WITHIN 5 YEARS (0.0328) (0.0659) -0.1218*** -0.0945 (0.0317) (0.064) -0.1872*** -0.168** (0.0347) (0.0708) -0.2362*** -0.2406*** (0.0414) (0.0864) 0.0234* (0.0125) SAT Read SAT Write -0.0625*** -0.0471** (0.0154) (0.0219) 0.0396** (0.0153) First GPA 0.1521*** 0.0827*** (0.009) (0.0258) Athlete 0.086** (0.0371) Constant 12.7073*** 0.853*** (3.5365) (0.15) Number of Observations 2,839 604 R-squared 0.2112 0.1118 Coefficients estimated using Ordinary Least Squares with a linear regression model, with standard errors in parenthesis. Statistical significance indicator using asterisk next to the coefficients, with * = significant at 10%, ** = significant at 5%, and *** = significant at 1%. © 2013 Hanover Research | Academy Administration Practice 28 Hanover Research | April 2013 APPENDIX A: RETENTION R ATES Figure A.1: Retention Rates by Gender 80% 72% 67% 61% 60% 55% 54% 40% 50% 32% 31% 20% 9% 8% 0% Female 2nd year retention Male 3rd year retention 4th year retention 5th year retention 6th year retention Figure A.2: Conditional Retention by Gender 100% 90% 83% 90% 85% 81% 79% 80% 61% 53% 60% 40% 20% 0% Female Male 3rd year conditional retention 4th year conditional retention 5th year conditional retention 6th year conditional retention © 2013 Hanover Research | Academy Administration Practice 29 Hanover Research | April 2013 Figure A.3: Retention Rate by Ethnicity 80% 60% 71% 61% 60% 57% 55% 48% 40% 40% 41% 36% 33% 22% 21% 20% 10% 8% 9% 0% Black 2nd year retention Other or two or more races 3rd year retention 4th year retention White 5th year retention 6th year retention Figure A.4: Conditional Retention Rate by Ethnicity 100% 80% 65% 91% 89% 82% 77% 71% 83% 80% 85% 67% 59% 60% 56% 40% 20% 0% Black Other or two or more races White 3rd year conditional retention 4th year conditional retention 5th year conditional retention 6th year conditional retention © 2013 Hanover Research | Academy Administration Practice 30 Hanover Research | April 2013 Figure A.5: Retention Rate by High School Class Rank 100% 83% 80% 81% 74% 74% 72% 69% 68% 63% 58% 60% 36% 40% 63% 55% 51% 44% 41% 38% 34% 36% 29% 22% 20% 15% 9% 9% 5% 5% 0% 0% to 10% 2nd year retention 10% to 25% 3rd year retention 25% to 50% 4th year retention 50% to 75% 5th year retention 75% to 100% 6th year retention Figure A.6: Conditional Retention Rate by High School Class Rank 100% 94% 91% 86% 80% 95% 87% 87% 91% 84% 85% 84% 82% 76% 74% 67% 86% 65% 75% 62% 66% 60% 40% 23% 20% 0% 0% to 10% 10% to 25% 25% to 50% 50% to 75% 75% to 100% 3rd year conditional retention 4th year conditional retention 5th year conditional retention 6th year conditional retention © 2013 Hanover Research | Academy Administration Practice 31 Hanover Research | April 2013 APPENDIX B: GRADUATION RATES Figure B.1: Graduation Rate by Gender 80% 62% 60% 40% 52% 37% 29% 20% 0% Female Male Four year graduation rate Five year graduation rate Figure B.2: Graduation Rate by Ethnicity 80% 37% 40% 20% 59% 57% 60% 35% 33% 15% 0% Black Other or two or more races Four year graduation rate © 2013 Hanover Research | Academy Administration Practice White Five year graduation rate 32 Hanover Research | April 2013 Figure B.3: Graduation Rate by High School Class Rank 100% 81% 75% 80% 63% 61% 60% 50% 45% 37% 40% 28% 22% 20% 12% 0% 0% to 10% 10% to 25% 25% to 50% Four year graduation rate © 2013 Hanover Research | Academy Administration Practice 50% to 75% 75% to 100% Five year graduation rate 33 Hanover Research | April 2013 PROJECT EVALUATION FORM Hanover Research is committed to providing a work product that meets or exceeds partner expectations. In keeping with that goal, we would like to hear your opinions regarding our reports. Feedback is critically important and serves as the strongest mechanism by which we tailor our research to your organization. When you have had a chance to evaluate this report, please take a moment to fill out the following questionnaire. http://www.hanoverresearch.com/evaluation/index.php CAVEAT The publisher and authors have used their best efforts in preparing this brief. The publisher and authors make no representations or warranties with respect to the accuracy or completeness of the contents of this brief and specifically disclaim any implied warranties of fitness for a particular purpose. There are no warranties which extend beyond the descriptions contained in this paragraph. No warranty may be created or extended by representatives of Hanover Research or its marketing materials. The accuracy and completeness of the information provided herein and the opinions stated herein are not guaranteed or warranted to produce any particular results, and the advice and strategies contained herein may not be suitable for every partner. Neither the publisher nor the authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. Moreover, Hanover Research is not engaged in rendering legal, accounting, or other professional services. Partners requiring such services are advised to consult an appropriate professional. © 2013 Hanover Research | Academy Administration Practice 34 Hanover Research | April 2013 1750 H Street NW, 2nd Floor Washington, DC 20006 P 202.756.2971 F 866.808.6585 www.hanoverresearch.com © 2013 Hanover Research | Academy Administration Practice 35