Grit and Graduation

PAGE 1

GRIT AND GRADUATION : A META THESIS BY M. FAITH BENAMY A Thesis Submitted to the Division of Social Science New College of Florida in partial fulfillment of the requirements for the degree Bachelor of Arts Under the sponsorship of Heidi E. H arley Sarasota, Florida May 2012

PAGE 2

ii Dedication To my Bubbe, Marion Benamy, who passed away this year. My sister once said as this may have been true. Thank you B ubbe, for instilling in me the social skills to tackle a project which not only required me to network with people, but for them to like me enough to let me rummage through their filing cabinets and use up all their paper.

PAGE 3

iii Acknowledgements It takes a vil lage to raise a thesis, especially one that involves institutional research. First and foremost, I'd like to thank my supportive thesis sponsor, Professor Heidi Harley, who not only supplied me with feedback and edits, but also was my advocate when my thes is encountered logistical setbacks. I'd also like to thank Professor Angela Duckworth for providing the inspiration for my study and laboratory skills to manage it. Additional acknowledgements go out to my thesis committee members, Professor Michelle Barto n and Professor Gordon Bauer for their time, energy, and edits as well as Professor Duff Cooper for providing me with stacks of statistics literature. Beyond those who provided the intellectual support for my thesis, I'd like to thank those who made the actual data collection possible. Thank you to the entirety of the Office of the Registrar for letting me occupy the photocopier for several weeks. Kathy Allen and Marylin Brislin deserve special thanks; without their help, I would have had to change my t opic. I would be remiss if I did not acknowledge my programmer, Stefan Brown, who helped me to access critical information and saved me countless hours. Thank you to the Office of Admissions as well, especially, Molly Robinson. A great many thanks to my wonderful coder Nicole Noujaim, for her time commitment and social support. Additional thanks go out to my lovely roommates and best friends, Sherry Haber, Aidan Bailey, Mallory Fenn, and Zoe Posner for putting up with me when stacks of papers flooded the common room, Francesca Levya for being my solace during later nights in the SSRL, and my parents, Cindy and Richard Benamy, as well as my sister, Alanna, and my dog, Shayna, for being the least dysfunctional family I know.

PAGE 4

iv Table of Contents DED ICATION . . . . . . . . . . . ii ACKNOWLEDGEMENTS. . . . . . . . . . iii TABLE OF CONTENTS . . . . . . . . . iv LIST OF TABLES . . . . . . . . . . v LIST OF FIGURES. . . . . . . . . . . vi ABSTRACT . . . . . . . . . . . . vi i INTRODUCTION. . . . . . . . . . . 1 Academic and Systemic Factors and College Graduation. . 3 HSGPA and College Entrance Exam Scores . . 3 Socioeconomic Status . . . . . . . 6 Rac e . . . . . . . . . . 9 Gender . . . . . . . . . . 9 Individual Differences . . . . . . . . 11 Grit . . . . . . . . . . . . 13 The Present Study . . . . . . . . . 23 METHOD . . . . . . . . . . . . 2 5 RESULTS . . . . . . . . . . . . 28 DISCUSSION . . . . . . . . . . . 31 REFERENCES . . . . . . . . . . . 38 TABLES . . . . . . . . . . . . 4 5 FIGURES . . . . . . . . . . . . 5 4 APPENDIX . . . . . . . . . . . . 5 7

PAGE 5

v List of T ables TABLE 1: Profiles of Student Grit Scores TABLE 2 : Means and Standard Deviations by Entering Class TABLE 3 : Frequencies of Gender and Race by Entering Class TABLE 4: Race Frequencies by Gender TABLE 5: Means and Standard Deviations by Race TABL E 6: Grit Score Means and Standard Deviations by Entering Class TABLE 7: Grit Score Means and Standar d Deviations by Gender and Race TABLE 8 : Pearson Correlations for All Continuous Variables TABLE 9 : Summary of Regression Model Predicting Graduation

PAGE 6

vi Lis t of Figures FIGURE 1 : Proportion of Graduates Produced by Entering Class FIGURE 2: Proportion of Graduates by Gender FIGURE 3: Proportion of Graduates by Race

PAGE 7

vii GRIT AND GRADUATION : A META THESIS M. Faith Benamy New College of Florida, 20 12 ABSTRACT Recent statistics show that a large portion of students who enter college in the United States do not graduate. Although there is a wealth of information exploring how demographic factors like gender, race, and socio economic status and high school ac ademic achievement affect graduation rates, there is a scarcity of research on how individual traits affect the likelihood of graduation. The present study explores the trait of grit, perseverance and passion for pursuing long term goals, as a predictor of graduation at a small liberal arts college for the entering classes of 2002, 2003 and 2004. Grit was measured by coding the extracurricular portion of admissions applications for both long term commitment and ach ievement (e.g., multi year participation an d leadership positions held). De mographic variables, SAT scores and weighted high school grade point average (WHSGPA) were collect ed as well. Grit was mo destly predicative of graduation, but WH S GPA was a stronger predictor. Because WHSGPA includes the dif ficulty of classes taken, it is hypothesized that a coding schem e examining grit specifically in academic pursuits graduating. _______________________ Professor Heidi Harley Division of Social Scienc e

PAGE 8

1 Grit and Graduation : A Meta Thesis Americans are concerned about college graduation. When introducing the 2010 national budget, President Obama declared, By 2020, America will once again have the highest proportion of college graduates in the world. In a similar tone, The Gates Foundation's initiative on post secondary education states: We have set an ambitious goal for ourselves and the nation: double the number of young people who earn a post secondary degree or certificate with value in the marke tplace by the time they reach age 26. To accomplish this goal, we must connect the millions of young Americans who have the will to get the education they need with a way to get there: helping them get further, faster and with less cost in terms of time an d money (Bill & Melinda Gates Foundation, 2009) The United States currently ranks 10 th internationally in post secondary degree rates for people between the age s of 25 35 with only 39% percent obtaining a first post secondary degree (Organization for Ec onomic Co operation and Development [OECD], 2010). A pproximately half of the people who attempt a degree in the United States obtain it; a percentage that is considerably lower than most other developed countries where 70% of the people who start their bachelor's or associate's degree graduate (OECD, 2010). Lagging behind other nations has economic implications as well as social ones. There has not been a significant increase in the number of high school graduates receiving a ba chelor's degree since the 1970 s; however, the wage gap between those with a college degree and those with a high school diploma alone has almost doubled in the

PAGE 9

2 last 40 years (Goldin & Katz, 2008). The rate at which college graduates are produced not only a ffects the personal wealth of its citizens, but also the overall economic growth of a nation. In addition to a ffecting entire n ations, graduation rates also a ffect individual institutions of higher education. Colleges and universities spend money to recr uit students to their programs; if a student does not graduate, the institutions suffer financial los s es as it takes fewer resources to retain a student than to recruit another to take his or her place ( Horn & Berger, 2005; Levitz, Noel, & Richter, 19 99 ). Graduation rates also are considered to be an indicator of institutional quality and are often used to rank college s (Levitz et al 1999). Because graduation rates affect the perceived quality of an institution, a higher rated institution will attract hig her quality students as well as higher quality faculty members which in turn will increase graduation rates (Long, 2002). The United States has been a leader in providing access to higher education with a policy emphasis on initial enrollment (Goldin & K atz, 2008), but this has not led to sustained growth in the rate s of graduation from post secondary degree programs The focus on initial access led to a dismissal of the claims that the United States was falling behind other nations in terms of higher edu cation. However, completion is the key to success. In a recent interview with the Christian Science Monitor (2008), Pat Callan, the President of the National Center for Public Policy and Higher Education was quoted: Historically, our strength has been a ccess [to higher education] and our weakness has been completion. We've always said the reason we can't be expected to do so well on completion is because we're generous on access. But now, we see

PAGE 10

3 countries catching up to us and surpassing us on access and completion. In order to catch up with other countries and meet the lofty goals set by both President Obama and the Gates F oundation, it is necessary to examine w hich factors are associated with college graduation in the United States higher education s ystem. The current literature concerning college graduation primarily examines either individual academic factors ( or standardized test score) systemic factors ( such a s race, gender institution type and socioeco nomic status ) or evaluations of experimental interventions designed to increase student retention However, there is a scarcity of research concerning how individual student trait differences affect the likelihood of graduation. Because of the paucity of research surrounding how trait differences mediate graduation rates, the present study focuses on a no n cognitive personality trait, grit as a predictor of graduation at a small honors college. Grit is the quality of having perseverance a nd passion for pu rsuing long term goals and has been shown to be predictive of both retention and success in a variety of situations ( Duckworth, Peterson, Matthews, & Kelly 2007 ; Duckworth & Quinn, 2009; Duckworth, Quinn, & Seligman, 2009 ). Academic and Systemic Factors a nd College Graduation HSGPA and College Entrance Exam Scores High school grade point average (H SGPA) and college entrance exam scores on either the SAT or the American College Test ( ACT ) are common ly used by colleges to select potential students to admit In fact, the SAT was developed as a way to help admissions departments make predictions about based on academic aptitude ( Lawrence, Rigol, Van

PAGE 11

4 Essen & Jackson, 2002 ) Given their popularity as tools of selecti on, it i s unsurprising that HSGPA and entrance exam scores are predictive of undergrad uate year t o year retention. Robbins, Lauver, Le, Davis, Langley, and Carlstrom (2004) conducted a meta analysis of 104 studies which examined either or both college GPA and year to year college retention and their relations to HSGPA, entrance exam scores, interpersonal skills, and study skills. Using the aggregated data across the 104 studies, Robbins et al. found a moderate correlation between year to year retention and HSGPA, r =. 23 The relation between SAT/ACT scores and retention r = .12, was not as strong as that between HSGPA and retention. Entrance Exam scores and HSGPA were strongly positively correlated, r = .46, indicating that factors relating to achieving good grade s, other than test taking ability account for the differences in the strength of their correlations with graduation. In addition to being associated with year to year retention as individual predictors, HSGPA and SAT scores have a combin ed effect on the likelihood that a student will graduate Astin, Tsui, and Avalos (1996) analyzed data from 76,000 students entering 365 colleges and universities in 1985. Even though both HSGPA and SAT positively affected graduation rates, the largest contrasts were betwe en students with both a high HSGPA and high SAT scores and students with both a low HSGPA and low SAT scores. Eighty percent of students who scored above 1300 on their SATs and had a while only 17% of students who s cored The SAT scores

PAGE 12

5 was restandardized in 1996. W hile high school academic mea sures predict graduation rates, the relationship between the two is not uniform across all institution types Burton and Ramist (2001) found variability in the predictiveness of SAT scores on graduation, particularly when comparing data from different institution types. Willingha 9 small liberal arts institutions showed a significantly smaller relation between SAT scores and than the literature had previously reported A larger and more recent unpublished thesis also show ed that SAT scores were more predictive of graduation from research focused universities than from liberal arts institution s (Moore, 2008). Not only does the relation between SAT scores and graduation vary by institution type, but graduation rates alone vary by institution type. Selective colleges i.e., colleges that admit students wi th higher HSGPAs and SAT scores also have higher graduation rates than less selective colleges (Bound, Lovenheim & Turner, 2010). Bound et al. (2010) used data from the Nat ional Longitudinal Study of the High School Class of 1972 and the National Educational Longitudinal Study of 1988 to compare college graduation rates between the high school senior classes of 1972 and 1992. Between the two classes, 8 year college graduatio n rates decline d 4.6% from 50.5% to 45.9%. However, graduation rates did not uniformly decline; only less selective public universities and community colleges experienced a decline in graduation rates while elite public universities and private colleges sh owed a slight increase in graduation rates. Additional analysis showed that at less selective institutions, increases in student faculty ratios as

PAGE 13

6 well as decreases in per student expenditure accounted for a large portion of the decrease in graduation rate s between the two classes as opposed to changes in ability levels as measured by a standardized test This finding is consistent with the hypothesis of an institution quality feedback loop in which schools with low er graduation rates attract less prestige and therefore fewer resources which in turn lower s graduation rates. The intertwined nature of the success of an institution and individual high school academic achievement is also demonstrated by graduation rates at an in dividual institution are more predictive of a student of graduation than SAT scores. Therefore, t he current literature indicates that while traditional pre admissions measures of scholastic aptitude are somewhat predictive, they are entangled with institutional selectivity and do not fully explain why some students graduate and others do not. Socioeconomic Status Not only are the rates of college graduation in the United States stratified by institutional selectivity and prestige, but also by family income (Manski, 1992). Using data from the high school senior class of 1980 collected from the High School and Beyond survey, Manski examined the relation between family income and a half years a fter graduating from high school. Family income was split into quintiles and the rates of graduation were calculated. By 1986, approximately 50% of the students with family income in the lowest quintile had the middle quintiles, approximately 63% percent had graduated and in the highest quintile approximately 72% had earned a degree.

PAGE 14

7 Even when there is no tuition cost, family income is related to graduation (Stinebrickner & Stinebrickner, 2003b). Berea College is a 4 year institution dedicated to serving low income students in Appalachia. There is no tuition at Berea, and a large portion of room and board is subsidized. Despite the fact that the direct cost to the student is limited, half of the students fail to gra duate. Student data from between the fall semester of 1989 and the fall semester of 1997 were collected and the relation between semester by semester retention and family income was analyzed. Students who had transferred into Berea or declared themselves f inancially independent were excluded from the analysis, leaving a total of 2,821 students. Family income was divided into 3 income brackets and entered into a Kaplan Meier survivor function. Students in the highest income bracket were 16% more likely to complete at least 6 semesters at Berea than students in the lowest income bracket. The study also examined college entrance exam scores using the ACT and found that the scores did not vary by income bracket, indicating that ability was similar across all i ncome groups. While students at Berea may not have to worry about tuition costs, most students must confront the issue of how they will afford their degrees. Tuition costs of attending college have risen approximate 247% between the academic years of 197 6 19 7 7 and 2006 2007 for 4 year private institutions and 266% for 4 year public institutions while Pell Grant aid fell from $4953 in 1976 to $4050 in 2006 (Bound & Turner, 2011). These astronomic tuition increases have far outstripped any rise in family i ncome levels particularly for those below the median (Bound & Turner, 2011).

PAGE 15

8 Coinciding with the disproportionate increases in tuition is a rise in the average number of hours worked by college students in the United States. Current population survey dat a show between the year s of 1972 and 1992 the average number of hours work by enrolled 18 21 year old college students increased from 9.5 hours to 12.4 hours and, more recently, to 13.2 hours per week in 2005 (Bound & Turner, 2011). Because income has not kept pace with college tuition and the number of hours worked by college students has increased in recent decades, it may be that the increase in hours worked may be related to rises in tuition; as well as the slight decline in graduation rates. A secon d study at Berea College indicated that working while enrolled at college has a negative impact on academic performance (Stinebrickner & Stinebrickner, 2003a). Students at Berea are required to work at least 10 hours a week in exchange for financial aid. Estimates of the effect of hours worked on academic performance using a model based on the variation in job assignments showed a negative effect of the number of hours worked on college GPA. Even though graduation rates were not studied in relation to hour s worked, it would not be unreasonable to posit that working while enrolled has a negative impact on overall academic performance, which may cause a student to dropout of college. a d to dropping out of college due to financial woes while pursuing a degree, the educational achievement gap between the rich and poor starts at an early age and does not close (Reardon, 2011). When entering kindergarten, students in the poorest quintile score one standard deviation be low students in the richest quintile on standardized math and reading tests (Duncan & Magnuson,

PAGE 16

9 2011). In recent years, the achievement gap between the richest and the poorest in the United States has continued to increase. The educational disparities be tween high and low income families is between 30 40% larger for students born in 2001 compared to those born in 1976 (Reardon, 2011). Race. In the United States, race and socioeconomic status are related, with most racial minority groups being over rep resented in lower income brackets (Heaton, Chadwick, & Jacobson, 2000). People of color also face the additional challenge of living within a society with implicit racist attitudes (Quillian, 2006; 2008). The intersection of current attitudes and historic oppression has led to lower rates of college graduation for students of color than for White Students Since the 1970 s the college enrollment gap between W hite students and B lack and H ispanic students has closed significantly (Hudson et al., 2005); unfor tunately, the gap in graduation rates has not followed the same trend. Nationally, within 6 years of entering as freshme n in the 1995 1996 academic year, 67% of White students completed their bachelor's degrees while only 46% of African American students a nd 47% of Latino students completed their bachelor's degrees (Carey, 2004). Gender In 1970, men made up the majority of college students (58%), but by 2002, they were the minority (44%) (Freeman, 2004). A study commissioned by the U.S. Congress found that for students entering college in 1996, women were more likely to complete their degrees than men ( Bae, Choy, Geddes, Sable, & Snyder 2000). In the 2005 though men made up 46 % of freshmen in 2002 (Snyder, Dillow, & Hoffman, 2008 ). It is

PAGE 17

10 commonly accepted that gender parity was achieved because of shif ting societal norms in the 1970s and 1980 s that allowed women to complete college; however, there are very few empirical studies examining why women are outperforming men in post secondary education. D iprete and Butchmann (2006) explored gender difference in the value of college education and found evidence that a college degree influences the likelihood of favorable outcomes for w omen. Women who earn a college degree have a higher standard of living, and a lower probability of being poor than women who do not graduate from college. Diprete and Butchmann found t he difference in the likelihood of favorable outcomes between those with a degree and those without a degree is much smaller for men. Jacob (2002) used data from the National Educational Longitudinal Study (NELS) to examine how factors related to enrolling in college interact with gender to explain the achievement gap. The N ELS tracked 8 th grade students starting in 1988 and administered follow up surveys every two years after. Jacob used an Ordinary Least Squares regression to model how certain clusters of variables account of gender gap in college enrollment. Family demogra phics and cognitive ability as measured by standardized test scores explained very little of the difference in enrollment rates, however, the cluster of non cognitive abilities: middle and high school grades, self report of behavioral problems in middle school and self report of hours spent studying per week in middle school accounted for nearly 3% of the 5% achievement gap between men and women in college enrollment Although this study only looked at initial enrollment in college, the finding that contr ollable behavioral differences explains a large portion of the gender gap at the

PAGE 18

11 post secondary level is critical to understanding what factors cause a student to complete their degrees. Conger and Long (2010) found similar results when st udying the gend er gap in within college achievements. The study analyzed data from students from 11 Florida publ ic schools and 5 private Texas u niversities. In both da ta sets, w omen had higher first semester GPAs than men The difference in first semester GPA was found t o explain a large portion of the gender gap in retention until the 6 th semester at Florida schools and graduation at Texas schools when controlling for standardized test scores. An increase in the number of difficult courses taken in high school by women also secondary education. Peter and Horns (2005) compared the academic intensity of the graduating high school classes of 1982 and 1992. High school a cademic intensity was a composite measure based on the total number of Advanced Placement classes taken, highest level of math taken as well as the total number of credits taken in a variety of academic subjects. In the 1982 cohort, women were less likely than men to be in the top 20% of the high school academic intensity distribution among those who enrolled in college, but in the 1992 cohort women were more likely than men to be in the top 20% of the distribution. These finding indicate that behavioral differences between men and women are driving the behavioral gap. Individual Differences Although HSGPAs, SAT scores and demographic factors duation, people are much more complex than their test scores and demographic variables Unfortunately, almost no resea rch exists on the relation between individual trait dif ferences and college graduation. However, there

PAGE 19

12 are several studies which examine individual differences and either year to year or semester to semester college retention. Robbins et al.'s (2004) 104 study meta analysis examined retention and individual differences by sorting the studies into psychosocial and study skill constructs such as Achievement Motivation, Academic Goals, Social Support, Social Involvement, Academic Self Efficacy, Academic Rela ted Skills and General Self Concept. Most measures were found to be moderately correlate d with college retention: Academic goals, r = .340, Social Support r = .204, Social Involvement, r = .216, Academic Self Efficacy, r = .359 and Academic Related Skills, r = .366. But, General Self Concept was found to be unrelated to college retention. Now curious, but not satisfied by the meta analysis, Robbins, Allen Casillas, Peterson, and Le (2006) created a 108 item Student Readiness Inventory (SRI) consisting of 10 subscales based on factors from their previous study. The SRI was designed to measure constructs of skill (Study Skills and Communication Skills), social engagement (Social Activity and Social Connection), self regulation (Emotional Control and Academi c Self Confidence) and motivation (Academic Discipline, General Determination, Commitment to College, Goal Striving). The researchers hypothesized that these constructs would tap into an underlying generalized motivational construct that fuels academic su ccess also potentially measured by Big Five Conscientious ness ; many of the subscales of the SRI were correlated with conscientiousness but still outperformed conscientious ness when predicting college GPA (Peterson, Casillas, & Robbins, 2006).

PAGE 20

13 Robbins et al. collected SRI surveys and both one semester and one year retention data from incoming freshman from 25 4 year institutions in the Midwest. All subscales of the SRI were incrementally predictive of one year retention, but only Academic Discipline, Commi tment to College, General Determination, Goal Striving and Social Connectio n showed more than a 15% increase in the likelihood of retention for students who scored one standard deviation higher than the mean on that subscale. The motivational con struct which Robbins et al. hypothesized motivated academic success and designed the SRI to examine is similar the recently defined trait of grit. In their original paper, Duckworth, Peterson, Matthews and Kelly (2007) proclaimed the construct of grit to be pers everance and passion for long term goals (p. 1087 ) Given that graduation from college is a long term pursuit and that Robbins et al. showed that discipline, commitment, determination and goal striving are incrementally predictive of college retention, gr it may be part of the college graduation puzzle. Grit In order to measure grit quality, Duckworth et al. (2007) created a 27 item preliminary questionnaire. The items were designed not to be domain specific and to gauge consistency of interest over time as well as sustained effort in the face of adversity. To validate the scale as a measure of grit, Duckworth et al. collected survey data from 1,545 participants over the age of 25 from the noncommercial website www.authentichappiness.com which provides fre e information about psychological research and allows users access to a variety of self report measures. In addition to taking the survey, participants reported their age and highest education level. After reviewing

PAGE 21

14 the internal reliability measures, the s cale was reduced to 12 items and used a two factor sub scale solution to measure stability of pursuits and persistence of effort which combine to form both the perseverance and passion elements of grit. The scale asks participants to rate how well statemen have overcome setbacks to conquer an important challenge them on a 5 point L ikert scale. The statements target persistence of interests, hard work and handling setbacks. Using the 12 item scale, the researchers examine d differenc es in self reported grit across age and education levels. Older people were grittier than younger people and participants with more education were grittier than those with less education. Post hoc comparisons showed that after controlling for education gr it increases monotonically with age and that higher in grit than most other education groups. D ue to the cross sectional design it is not possible to determine the directions of the positive a ssociations between grit and age and grit and education level, however, the findings suggest that as people age, they may realize that persistently pursuing long term goals results in more success than short term bursts of commitment or that pursuing a de gree requires sticking with a long term goal. Because cross sectional evidence is not enough to determine that a construct is a successful predictor of achievement outcomes, Duckworth et al. implemented 5 more studies examining grit in relation to both c ognitive and non cognitive factors. A second online survey study was conducted using the same recruitment methods as the first The survey included the 12 item Grit Scale, a version of the Big Five Personality Inventory,

PAGE 22

15 and asked participants t o identify the highest degree they held and the number of times they changed careers. Grit was correlated with all Big Five measures and was more closely related to conscientiousness ( r = .77) than any other Big Five trait. Despite the strong correlation between grit and consci enti ousness, grit was incrementally predictive for age and education level over and beyond conscientiousness, even when controlling for conscientiousness. This finding is similar to Peterson, study predicting college GPA from the SRI and Big Five conscientiousness and hints that grit may be a part of the underlying motivational construct that Robbins et al. (2006) theorize drives academic achievement. Because the number of career changes was not normally distributed a median sp lit was used to compare those who changed careers often and those who did not. When all continuous predictor variables were standardized and entered into a binary regression grit was the only significant predictor Individuals who were one standard deviation higher in grit than the mean were 35% less likely to be frequent career changers. Because the first 2 studies solidified the relation between grit and educational achievement in a ge neral sample of adults, the researchers became interested in the relation between grit and performance in high achievers. In their third study, Duckworth et al. (2007) examined the cumulative GPA of 139 undergraduate psychology students at an elite univers ity in relation to grit and SAT scores and found that grittier students outshined their less gritty peers: g rit was positively correlated with GPA, r = .25. The correlation between grit and GPA was even stronger when controlling for SAT scores,

PAGE 23

16 partial r = .34. SAT scores were also positively correlated GPA, r = 30, however they were negatively correlated with grit, r = .20, indicating that students with slightly less innate talent may be slightly more gritty than their more talented peers. In their fo urth study, Duckworth et al. examined grit in a different high achieving context: The United States Military Academy, West Point. Despite the rigorous admissions requirements, 1 in 20 cadets drop out during the first summer of training known as the Beast Barracks Duckworth et al. collected grit scores using the 12 point scale self control scores, Whole Candidate Scores (a score calculated by West Point admissions consisting of a weighted average of academic achievement, physical fitness and leadership a bility), freshman academic GPA Military Performance Scores (MPS) and summer retention data from 1,218 entering candidates. Grit was not related to Whole Candidate Scores or SAT scores, but was related to self control, r =.63. Grit, self control and Whole Candidate Scores were standardized and entered into a binary regression model predicting retention. Grit predicted retention better than any other variable. Cadets who were one standard deviation higher in grit than their peers were 60% more likely to make it through the Beast Barracks O dds R atio = 1.62, =.48. Candidates who were one standard deviation higher in self control were only 50% more likely to complete training. Whole Candidate Scores were not predictive of summer retention, but they were posit ively correlated with MPS, r = .48, and freshman GPA, r = .13. Self control was also a positive predictor of MPS and GPA. Grit predicted MPS, as well as self control, r =.19, but was not as strong a predictor of GPA, r = 06 The fact that g rit was not th e best predictor of MPS or GPA, but was extremely

PAGE 24

17 predictive of completion of the Beast Barracks shows that grit is related to long, arduous pursuits rather than the hour to hour regulation required for studying and earning good grades. The fifth study was a replication and expansion of (2007) fourth study. The predictive validity of grit for summer retention was examined against Big Five Conscientiousness. One thousand, three hundred and eight freshman cadets completed the G rit S cale and the c onscientiousness subscale of the Big Five Inventory. Whole Candidate Scores and summer retention data were obtained through official records. Whole Candidate Score was associated with conscientiousness, r = .12, but not with grit. Replicating the results of the second study, grit and conscientiousness were strongly related, r = .64. Despite their strong relation, summer retention was predicted better by grit than conscientiousness. When all three measures were entered into a binary regression toget her, only grit predicted summer retention, = .29, OR = 1.47. This study reinforces grit's uniqueness as a trait rather than a function of another personality factor. The sixth study (2007) was a prospective, longitudinal study involving finalists in the Scripps National Spelling Bee. Duckworth et al. were interested in examining the role of grit in avocational settings as well as the mechanisms of grit. One hundred and seventy five National Spelling Bee finalists agreed to participate in the study. All participants completed the Grit Scale and a measure of self control They also reported how many hours per day they studied on both weekdays and weekends. Seventy nine of the participants volunteered to complete a verbal IQ test via telephone. Prior competitions and final round completed before el imination were provided by the Scripps National

PAGE 25

18 Spelling Bee for each participant. G rit, self control and verbal IQ were standardized and entered into ordinal regression models to assess their predictiveness with regard to how far the participants advanced in the competition. Age was also entered as a co variate because older participants were more likely to rank higher in the competition. When all variables were entered simultaneously, only grit, = .62, OR = 1.86 and age, = .29, OR = 1.33, were predictive of how far a participant advanced in the competition. When only grit, age and self control were simultaneously entered as predictors of the highest round achieved in the competition, only gri t and age were predictors. When only grit, age and verbal IQ were entered into the model, grit was not predictive despite the fact that grit was unrelated to IQ; however, because only half of the participants completed the IQ measure, the statistical power of the test was reduced. Duckworth et al. suggest that with a larger sample, grit may be predictive. Duckworth, Kirby Tsukayama, Berstein and Ericsson (2010) continued to use the Scripps National Spelling Bee to study grit. One hundred and ninety Natio nal Spelling Bee Finalists completed an 8 item version of the Grit Scale as well as the Big Five Openness to Experience subscale, which was found to be negatively associated with highest round achieved (Duckworth & Quinn, 2009). The finalists also filled ou t a questionnaire pertaining to how much time they spent preparing using deliberate practice (i.e solitary studying) the time spent being quizzed by a coach or computer in the month prior to the survey, as well as the time spent on each activity for eac h year they had studied spelling. Time spent on word based leisure activities was measured by a self report of the number of books read in the past year. Attitudes towards each of the three

PAGE 26

19 types of verbal activities : deliberate practice, quizzing and word based leisure activities were measured using a 9 point Likert scale to rate the enjoyability, effort and relevance to preparing for the bee of activities. All measures of time spent practicing were transformed using log 10 because the data were skewed. Fin al round achieved was provided by Scripps for each participant. When standardized values of time spent on deliberate practice, quizzing and word based leisure activities were separately entered into ordinal regression models predicting highest round achie ved, both the number of hours spent on deliberate practice ( OR = 2.64) and the number of hours spent on quizzing ( OR = 1.61) proved to be predictive, but the number of hours spent on leisurely practice activities was not. When all three practice activities were entered simultaneously, only the number of hours spent on deliberate practice was predictive ( OR = 2.49). A repeated measures ANOVA was used to determine the differences in enjoyment, effort and relevance between the different modes of studying. Dif ferences between activities were found with regards to enjoyment. Post hoc tests showed that leisure practice activities were rated as more enjoyable than quizzing which was rated as more enjoyable than deliberate practice. The same pattern held for percei ved effort with leisure activities being considered to require the least amount of effort and deliberate practice the most. The bee finalists considered leisure activities to be the least relevant form of studying. Duckworth et al. (2010) used multi medi ator modeling software to examine how grit and openness to experience were mediated by time spent on different types of

PAGE 27

20 studying when predicting final round achieved in the bee. The inverse relation between openness to experience and final round was not me diated by the time spent on any type of studying. However, grittier spellers spent more time studying using deliberate practice than the ir less gritty counterparts. Spellers who spent more hours doing deliberate practice made it further in the competition than those who did not. In the case of the spelling bee, grit tier spellers spent more time doing the most effective, but least enjoyable preparation technique and attained a higher rank in the bee If grittier spellers eschew more enjoyable activities to push themselves towards a better ranking at the bee, perhaps grittier students eschew more social activities to push themselves towards completing a degree. In addition to predicting final round achieved in the National Spelling Bee, completion of the Bea st Barracks, GPA at elite universities, rates of career change and lifetime educational attainment, grit is predictive of teacher effectiveness (Duckworth & Robertson Kraft, Unpublished; Duckworth, Quinn & Seligman, 2009). While a variety of factors influ ence a student's achievement, an effective teacher can push students towards academic gains. However, effective teachers are hard to come by. Teaching is a difficult and rigorous job (Stanford, 2001) with a high attrition rate (Boser, 2000; Henke, Chen, Ge is & Knepper, 2000). Duckworth et al. (2009) partnered with a national teacher training program to examine predictors of teacher effectiveness. The program Duckworth et al. studied is an organization that pairs novice teachers with under resourced schools across the United States. The organization derives a measure of teacher effectiveness based on student academic gains and content mastery.

PAGE 28

21 Duckworth et al. (2009) recruited volunteers from the program via email to participate in their study. Three hundre d and ninety participants responded. The participants completed survey measures for grit, life satisfaction and explanatory style. The program provided demographic information and teacher effectiveness ratings. The researchers standardized all continuous v ariables prior to analysis to aid in the interpretation of odds ratios. When each variable was entered into an ordinal regression model predicting teacher effectiveness ratings, teachers who were one standard deviation higher in grit were 31% more likely to outperform their less gritty peers, teachers who were one standard deviation higher in life satisfaction were 43% more likely to out perform their peers and those one standard deviation higher in optimistic explanatory style were 20% more likely to ou tperform their peers. All three traits were positively correlated with each other, because the researchers entered the traits into the model simultaneously to see if the effects of one variable were explained by the others. In the simultaneous model, grit ( = 0.21, OR =1.23) and life satisfaction ( = .31, OR = 1.31) remained predictive, but explanatory style did not. Because grit is strongly associated with positive outcomes in difficult situations, admissions committees and hiring departments may want to examine the grittiness of their applicants. While surveys are a very common and useful way to measure traits, they are subject to desirability effects and thus ineffective applicant screening tools. To develop a way to measure grit that was not easily manipulated for desirability, Duckworth and Robertson Kraft (u npublished) established a way to code for grit from

PAGE 29

22 college accomplishments for evidence of leadership In their coding scheme, a person received 1 point for each activity in which they had participated for at least 2 years as listed on their rsums. A person received additional points for moderate (1 additional point) and high (2 additional points) achievements in multi year activities. Grit score was considered the sum of the two highest scoring activities on rsum. Duckworth and Robertson Kraft (unpublished) applied this coding scheme to rsums from the same teacher training organization used in the previous study The organization provided the rsums, SAT scores, college GPAs, teacher effectiveness ratings and first year retention data (whether or not a teacher dropped out) for two samples of teachers. The researchers conducted separate analyses for each sample. Teachers whose stu dents had made a t least a year's worth of academic progress of demonstrated mastery of the school district's content standards were considered to be effective, while teachers whose students had not made satisfactory progress were considered less effective. Continuous variables were standardized before entered into binary regression models. The first sample consisted of a stratified random sample of teacher effectiveness and retention of teachers from two low income, urban school districts. In this sample, there was a strong association between SAT scores and college GPA, r =. 40 but neither variable was related to grit. Teachers who completed the school year had higher grit scores ( M = 3.98, SD = 1.45) than teachers who resigned in the middle of the year ( M = 2.79, SD = 1.45 ). Teachers who were one standard deviation higher in grit were 234 %

PAGE 30

23 more likely to complete the full year than their less gritty peers. Similar results were found for teacher effectiveness. Effective teachers had higher grit scores ( M = 4.16, SD = 1.43) than less effective teachers ( M = 3.54, SD = 1.50), t( 119) = 2.24, d = .42. There were no differences between effective and less effective teachers on college GPA, SAT scores or any demographic variables. Teachers who were one standard d eviation higher in grit were 71% more likely to outperform their less gritty peers, = 54, OR = 1.71. The second sample was a random sample of 307 teachers from 6 low income schools. There was not enough variation in teacher retention in the second sam ple for analysis so the researchers only analyzed the effectiveness data. As in the first sample, there was a strong association between SAT score and college GPA, r =. 40 but neither was related to grit. Effective teachers ( M = 3.88, SD = 1.56 ) were grit tier than less effective teachers ( M = 3.20, SD = 1.48 ) Teachers who were one standard deviation higher in grit were 64% more likely to outperform their less gritty peers, = .50, OR = 1.64. There was no difference between effective and less effective teachers in terms of college GPA, SAT scores or demographic variables. The Present Study Despite rising rates of college matriculation, there have not been corresponding ri ses for college graduation. There is a wealth of res earch examining the relation between factors like gender (Bae et al., 2000; Jacob, 2002;Snyder et al.,2000), race ( Carey et al., 2004; Hudson et al., 2005) and socioeconomic status (Reardon, 2011, Stinebr ickner & Stinebrickner, 2003b) and graduation rates as well as a body of literature on academic

PAGE 31

24 factors (Astin et al., 1999; Robbins et al., 2004) However, there is much less literature exploring how individual differences influence graduation from colleg e. This research is especially important because graduation rates not only affect the health of the national economy ( Goldin & Katz, 2008 ), but the wealth of an individua l institution (Horn & Berger, 2005). If trait differences have an impact on a student s likelihood of graduation, interventions can be designed to help foster those traits and help both nations and colleges flourish. Grit has proven to be a reliable predictor of success in a variety of arenas, such as lifetime educational attainment, rete ntion through West Point's Beast Barracks, ranking in the National Spelling Bee, and teacher effectiveness (Duckworth et al., 2007; Duckworth & Quinn, 2009; Duckworth et al., 2009, Duckworth & Robertson Kraft, unpublished). Completing a college degree is a task that eludes many and requires that one pick a track of study and pursue it to completion. One is likely to receive bad grades and other se tbacks along the way and must persevere through difficult exams and projects Grit, perseverance and passion for long term goals, may facilitate this process. The present study s ought to examine if grit is able to predict graduation using admissions data at a small 4 year honors college. Unlike other non cognitive measures that may be associated with college success grit can be ascertained from non survey measure s (Duckworth & Kraft Robinson, u np ublished).The present study used a variation of the coding scheme from Duckworth & Kraft extracu rricular po applications and hypothesize d that grit would be

PAGE 32

25 predictive of graduation. Method Sample A small, public liberal arts honors college in the southeastern United States provided access to admissions applications and gradua tion data from traditional college aged (17 22 year old) students entering college for the first time in the fall of 2002 ( N = 159), 2003 ( N = 156), and 2004 ( N = 188). Between all three cohorts, there were 503 students in the sample. D emographic var iables did not differ significantly between the entering classes. Of the total sample, 83.3% identified as White and 16.7 % identified as persons of color; 62.6% were female and 37.4% were male. Average household income as determined by zip code was $ 49,674 with a standard deviation of $ 18,961. H ousehold income was not normally distributed. Procedures and Measures Researchers accessed admissions applications other information gathered during the admissions process and graduation data. From the admissions applications, the researcher obtained the portion of the application pertaining to extracurricular activities for each student. From an admissions database, the researcher obtained SAT scores weighted high school GPA (WHSGPA) zip code race and year of graduation for each student in the sample. Grit. Two researchers coded the extracurricular activity portion of admissions applications for evidence of grit. Coding was completed without knowledge of graduation

PAGE 33

26 status to avoid bias. This study used a 10 po int (0 9) grit scoring scale adapted from previous research (Willingham, 1985) by Duckworth and Robertson Kraft (unpublished) to quantify evidence of grit. Student grit was evaluated on this scale using the following procedure: first, one point was assign ed for extracurricular activities in which the student had participated for at least a total of two years. Activities which were not multi year or were unskilled labor positions received no points. If a moderate level of achievement (i.e., award within an activity or a leadership position, but not the highest form possible, e.g., the treasurer of a club) had been attained, the activity or experience was given an additional point. Activities in which a high level of achievement (i.e., the highest honor withi n an activity, e.g., Gold Award in Girl Scouts, or running a club or organization. e.g. president, vice president, or captain) had been attained the activity received two additional points. Given this coding scheme for any activity, students could receive from 0 points (i.e., involvement less than two years) to 3 points (i.e., involvement over two years with high achievement). To tabulate a final grit score, the grit scores of the highest 3 activities were added together. Thus, each student could receive a total score between 0 and 9. S ee Table 1 for sample profiles and Appendix A for specific examples of high and moderate achievements. After coding each application, the two coders reviewed discrepancies, if they existed, until they both agreed upon a fin al grit score. Before reaching consensus via discussion, inter rater agreement for the coders was r (482) = .963, p < 001. Ten students did not receive grit scores because their applications were missing an attached extracurricular sheet; they had attach ed rsums, or applied via alternate applications that

PAGE 34

27 did not follow the format of the standard applications. Students with missing grit scores did not have significantly different graduation rates from students with available grit data st p = .179 Because there was no d ifference in graduation rates, s tudents without grit scores were excluded from analyses involving grit, but included in general demographic analyses. Grit scores were normally distributed, M = 4.82, SD = 2.34. Income I n the present study, student family income was often missing To remedy this problem income was considered median household income as determined by zip code from the 2000 census. Because median household income was heavily skewed to the right, the natural log of median household income was used to normalize the data and used for analysis M = 10.75, SD = 0.35. Zipcode data were missing for 12 students. Students missing income data did not have significantly different graduation rate s students with availabl e data, p = .763, and were included in all analyses that did not involve income. Race. Students self reported race in the admissions applications. Fifteen students identified as Asian, 8 identified as Black, 48 identified as Hispanic 414 identified as White and 1 student identified as other. Because of the school's small size and limited racial diversity, students were either considered White (reported Caucasian or White) or a person of color (reported race other than White) in the ana lysis of the data. Race data were available for every student. High School Academic Achievement. The present study utilized SAT scores ( M = 1313.79, SD = 108.68) and weighted high school grade point averages ( M = 3.91, SD =

PAGE 35

28 .37) as measures of high scho calculated as the highest combined math and verbal SAT score attained in a single sitting for the test The Office of Admissions calculated weighted high school grade point averages (WHSGPA) by weighting e ach Advanced Placement or International Baccalaureate class with 1 additional point and each honors class with .5 extra points. Twenty two students did not have SAT score data and 6 did not have WHSGPA data. Students missing SAT scores did not have differ ent graduation rates tha n students with available data, 2 (1, N = 494) = .77, p = .497 Students missing WHSGPA data also did not have different graduation rates tha n students with available data, p = .672 Students missing high school academic achievement data were excluded from analyses using the variables containing the missing data, but included in all other analyses. Both measures were normally distributed. Graduation. Graduation data were attained from the published lists of grad uates from the years 2007 2011. In the present study graduation is defined as gradu ating from the college within 5 years of entering the institution. Students who graduated from the college in 6 or 7 years were excluded from the analysis leaving N = 494. O f the remaining students, only 64% received a degree. Results Entering Class The entering classes of 2002 2004 did not differ by age, F (2,491) = .04, p = .957; income, F (2,479) = .02, p = .979; or SAT scores, F (2,469) = .41, p = .662 However, they di d differ by of WHSGPA, F (2,485) = 4.25, p = .015 ( Table 2 ) Post hoc analysis

PAGE 36

29 WHSGPA ( M = 3.84, SD = 0.38) than the classes of 2002 ( M = 3.93, SD = .35) and 2004 ( M = 3.95, SD = .38). The classes did not have an unequal dis tribution of students by gender 2 (2, N = 494) = 4.13, p = .127, but did by race, 2 (2, N = 494) = 7.49, p = .024 The class of 2002 was much less diverse than the classes of 2003 and 2004. Only 9.7% of en tering students identified as people of color in 2002 compared to 19% in 2003 and 19.9% in 2004 (Table 3) Although there was a difference in diversity make up, there was no difference in proportion of graduates produced across years 2 (2, N = 494) = 2.61 p = .272 (Figure 1) Gender Male students were slightly older ( M = 18.04, SD = .554) than female students ( M = 17.88, SD = .584), t (492) = 3.16, p = .002. Men also had higher SAT scores (M = 1335.72, SD = 103.99) than women ( M = 1300.97, SD =109.48) t (470) = 3.39, p = .001), but women had higher WHSGPAs ( M = 3.96, SD = .35) than men ( M = 3.83, SD = .40), t (486) = 4.02, p < .001. Income did not differ between m ale ( M = 10.74, SD = .33) and female ( M = 10.77, SD = .37) students, t (480) = .96, p = .34 0. Racial break down did not vary between men and women 2 (1, N = 494) = .296 p = .615, nor d id the proportion of graduates, 2 (1, N = 494) = .09 p = .772. (Table 4; Figure 2) Race White students and students of color did not differ by age, t (492) = 1 .03, p = .302, or income, t (480) = 1.798, p = .073 However, they varied on measures of academic achievement. White students scored higher on the SAT ( M = 1323.31, SD = 100.57) than

PAGE 37

30 students of color ( M = 1266.46, SD = 133.28), t (96.63) = 3.59, p = .001. W hite students also had slightly higher WHSGPAs ( M = 3.93, SD = .37) than students of color ( M = 3.83, SD = .37), t (486) = 2.119 p = .035 .(Table 5) Although W hite students had higher SAT scores and WHSGPAs, White students and students of color graduated at comparable rates, 2 (1, N = 494) = .21 p = .704. (Figure 3) Grit Grit did not vary between the entering classes F (2, 481) = 1.65, p = .194. (Table 6). Grit also did not differ by gender t (482) = 1.07, p = .283 nor between White students and students of color, t (124.79) = .50, p = .621 (Table 7) St udents who graduated had higher grit scores ( M = 4.95, SD = 2.38) than students who did not ( M = 4.58, SD = 2.22) at the one tailed level of significance, t (482) = 1.70, p = .045 d = .14 See Table 6 fo r m eans and standard deviations by entering class and Table 7 for means and standard deviations by gender and race. High School Academic Achievement Students who graduated did not have higher SAT scores ( M = 1317.93 SD = 106.58 ) than those who did not ( M = 1306.57 SD = 112.21 ) t (470) = 1.09, p = .252. However, students who graduated had higher WHSGPAs ( M = 3.96, SD = .36) than those who did not graduate ( M = 3.84, SD = .36), t (485) = 3.33, p = .001 d = .30 Variable Correlation Students with high WHSGP A also had high SAT scores, r (465) = .233, p < .001. In addition to being positively correlated with WHSGPA, SAT scores were also moderately associated with higher income, r (450) = .139, p = .003. Despite its relation to

PAGE 38

31 SAT scores, WHSGPA was not correla ted with income, p = .152. Grit and age were not related to any other variables, p >.05. ( Table 8 ) Predicting Graduation To investigate incremental predictivity of grit in relation to other known predictors of graduation the researcher fit a binary logis tic regression predicting graduation from all collected variables All scale variables (WHS G PA, SAT score, income and grit) were standardized before entry into the model to assist in the interpretation of the o dds ratios. As shown in Table 9 of all the f actors entered, only WHSGPA and grit were incrementally predic tive of graduation. Students who had WHSGPAs one standard deviation higher than the mean were 38% more likely to graduate than their peers with lower WHSGPAs, = .322, OR = 1.38, p =.003. Stud ents who were one standard deviation higher in grit were 24% more likely to graduate than their less gritty peers, = .213, OR = 1.24, p = .041. (Table 9) Discussion Grit, WHSGPA and Graduation levels would pred ict graduation was confirmed. However, grit was not the strongest predictor of graduation. WHSGPA had a much stronger association with graduation than did grit While the literature supports a relation between HSGPA and retention, t he relation between WHSG PA and graduation may be related to the inclusion of course difficulty as well as individual course grade in WHSGPA as a measure of academic achievement. WHSGPA and SAT scores were positively correlated, but only WHSGPA was associated with graduation SAT scores

PAGE 39

32 have been considered a proxy for general intelligence in previous literature, and new data show a strong positive correlation between SAT scores and scores on general intelligence tests (Frey & Detterman, 2004). If SAT scores are considered to be a proxy for general intelligence, then the factors which cause a student to have both a high WHSGPA and graduate are related to more than innate general cognitive ability. It can be speculate d that taking more difficult courses and achieving good grades in dicates a drive to succeed with respect to academics and that drive accounts for why WHSGPA is related to grad uation and SAT scores are not. Although grit as coded from the applications had a weaker relation to graduation th an WHSGPA, connection w ith graduation may be explained by the general concept of grit rather than the way the present study measured it. When conducting research on the National in the bee was mediated by grittier finalists choosing to spend more time studying using less enjoyable but more effective preparation techniques. Perhaps students who choose to take more difficult courses in high school are taking a similar approach to college achi evement through those high school choices It may seem bold to claim that grit may be related to WHSGPA even though the two measures were not correlated; however, t he wa y grit was measured in the present study focuses on leadership and hobby achievemen ts r ather than academic ones. The present study speculate s that a more academically specific measure of grit would show a relation between WHSGPA, grit and graduation.

PAGE 40

33 Additional Findings In addition to the main findings of the study that WHSGPA predicts gra duation better than grit score, the present study also had several interesting findings that were inconsistent with the current literature. The present study found that SAT scores were not associated with graduation. While previous research has found that SAT scores are less predictive of graduation from liberal arts institutions than other types of institutions (Moore, 2008; Willingham, 1985), they are rarely reported to be entirely unrelated. A potential explanation of this result is that because the pres ent study was conducted at an honors college, SAT scores had a restricted range compared to larger schools, which often have both an honors college in addition to their general population (and therefore have a wider range of score), and hence did not show any relation to graduation rates. Another explanation is that because the college does not focus on standardized testing, the ability to score well on a standardized test does not reflect success at this particular school. Other findings which were incon sistent with the literature were that race, income and gender were unrelated to graduation rates. It is a positive finding that demographic variables had no effect on graduation at this particular school; however, the school is very small and not an accura te gauge of overall systemic inequality in the United States. The fin dings that students of color were more likely to have lower SAT scores and lower WHSGPAs than White students and that SAT scores were positively correlated with income are consistent with the current literature on race and socio economic achievement gaps (Heaton et al., 2000)

PAGE 41

34 Another result that is consistent with the current literature was that men were also more likely to have higher SAT scores than women, but women were more likely to have higher WHSGPAs than men. Men tend to score higher than women on the SAT (Schmidt & Camara, 2003), and wo men tend to have higher GPAs tha n men (Jacob, 2002). Limitations While the coding scheme used by Duckworth and Kraft Robinson (unpublished) was h ighly effective at ascertaining grit from the rsums of teachers and predicting positive teacher outcomes this may be because teaching is a leadership position When applied to college admissions applications, particularly at the college where the presen t study was conducted, the coding scheme may have been less effective The college in present study does not have grades and course credit is determined by a pass/fail system. A student whose application receives a high grit score may be highly motivated b y external awards and recognized leadership positions like getting the best grade in the class or graduating as valedictorian and hence attending a college which does not emphasize formal and visible achievement may cause a student who is extrinsically m otivated either to transfer or drop out and decrease the visibility of the relationship between grit and graduation. In addition, because t he college used in t he present study is very small (e ach entering class c o ntained fewer than 200 students), t here a re few on campus branches of national organizations and any formalized sports teams. The coding for grit rewards people who have shown commitment, achievement and leadership in the se types of activities, however, there are not many opportunities provided b y the college to continue

PAGE 42

35 to pursue such interests. A student who is committed and passionate a bout a specific organization or sport in high school may choose to leave a college that does not provide them with pre existing opportunities to participate. A nother limitation of the grit code is that it does not account for private study. Several students listed on their applications that they had participated in intense private study of subjects such as foreign languages. The students who reported their perso nal studies often did not report any awards or e xternal achievements within those activities Because they did not report any achievements other than ability, they did not receive high grit sub scores for independent mastery of a language, clearly a gritty activity. In addition to the grit code focus on types of achievement that are not particularly salient at the college used in the present study as well as excluding independent, but gritty, achievements it is also subject to the limitations of the va riability in actual achievement by title For example, a student who becomes captain of a highly competitive team because s/he has demonstrated leadership qualities obtains the same grit sub score as a student who becomes captain of a less competitive team because s/he is th e only senior on the team. The grit score is also limited by the arbitrariness of what is considered to be a moderate achievement versus what is considered to be a high achievement, especially with regard to placing in a competition. Whi le an effort was made to keep the boundary between moderate and high achievement equivalent across activities, it is impossible to know exactly how prestigious a specific position or award is simply by its title.

PAGE 43

36 Students also varied in the way they repo rted activities, which influenced how they were coded. Some simply listed a general category of activities (e.g ., dancing and martial arts) while others listed several specific activities wi thin a category of activities (e.g ., listing multiple forms of da nce and their achievements within them ). Because the ways students reported their activities were not consistent, grit scores were influenced by reporting style. The results of the present study are also limited by the uniqueness of the college where dat a w ere co llected. Most post secondary institutions are considerably larger and the vast majority use gra des to track academic progress. Both the uncommon social and academic structure of the institution used in the present study may lead to a different se t a college with a more normative social and academic climate. Future Research While the code the present study used for grit was not a particularly strong predictor of graduation at an extremely small honors liberal arts college, it would be unwise to discount it as a potential predictor of graduation at other, larger institutions. However, the present study suggests that future research should attempt to develop a code that focuses on grit employed specifically towards success in academic pursuits. Future research should also attempt to untangle the relation between the difficulty of high school courses and high school GPA and graduation from college. Unfortunately, because both unweigh ted HSGPA and the number of difficult courses taken in high school were unavailable in the present study WHSGPA c ould not be broken down into simpler

PAGE 44

37 components to examine if it is the difficulty of courses taken, average course grade or an interaction of the two factors which propagates its relation to graduation. Both at a national level and at an institutional level, raising graduation rates is critical On the institutional level increased retention helps individual universities save money and recrui t better quality professors ( Long, 2002 ). On the national level, even more is at stake as higher degree completion rates equate to both increased wealth for individual citizens and increased wealth for the nation ( Goldin & Katz, 2008 ). Grit is one trait th at increases the likelihood of a student graduating. F uture studies should examine other types of individual trait differences as well as grit that influence graduation rates in order to develop practical intervention programs for students There is strong evidence to suggest that personality traits are mutable across time and experience ( Duckworth, Grant, Loew, Oett ingen, & Gollwitzer, 2011 ) and even cross sectional evidence shows that grit increases with age (Duckworth et al., 2007) President Obama and the Gates Foundation have set lofty goals to increase to the rates of college graduation in the United States and further research examining what factors contribute to college graduation will help them achieve it.

PAGE 45

38 References Astin, A., Tsui, L., & Av alos, J. (1996). Degree attainment at American colleges and universities: Effect of race, gender, and institutional type Washington, DC: American Council on Education. Bae, Y., S. Choy, C. Geddes, J. Sable, & Snyder, T.(2000). Trends in Educational Equi ty of Girls and Women Washington, DC: National Center for Education Statistics. Bill & Melinda Gates Foundation. (2009). Postsecondary Success: Focusing on Completion [Brochure]. Retrieved from http://www.gatesfoundation.org/learning/Documents/ posts econdary education success plan executive summary.pdf Bound, J. Lov enheim, M.F., & Turner, S. (201 ). Why have college completion rates declined? An analysis of changing student preparation and collegiate resources. American Economic Journal: Applied Economics, 2 ( 3), 129 157. Bound J, & Turner, S. (2011). Dropouts and diplomas: The convergence in collegiate outcome. In E. Hanushek, S. Machin & L. Woessmann (Eds.), Handbook of the Economics of Education (Vol. 4.) (pp. 573 614). Walthan, MA: Elsevier B.V. Boser, U. (2000). A picture of the teacher pipeline: Baccalaureate and beyond. Education Week 19, 16 17. Burton, N.W. & Ramist, L. (2001). Predicting success in college: SAT studies of classes graduating since 1980 (No. 2001 2). New York, NY: Col lege Entrance Examination Board. Carey, K. (2004). A matter of degrees: Improving graduation rates in four year colleges

PAGE 46

39 and universities. Washington, DC: The Education Trust. Conger, D. & Long, M.C. (2010) Why are men falling behind? Gender gaps in co llege performance and p ersistence The ANNALS of the American Academy of Political and Social Science 627 ( 1 ), 184 214 doi: 10.1177/0002716209348751 Diprete, T. A., & Buchmann, C.(2006). Gender specific trends in the value of education and the emergin g gender gap in college completion. Demography 4 (1). 1 24. doi: 10.1353/dem.2006.0003. Duckworth, A., Grant, H., Loew, B., Oettingen, G. & Gollwitzer, P. M. (2011). Self regulation strategies improve self discipline in adolescents: benefits of mental contrasting and implementation intentions. Educational Psychology: An International Journal of Experimental Educational Psychology 31 (1) 17 26 Duckworth, A., Kirby, T., Tsukayama, E., Berstein, H., Ericsson, K. (2010). Deliberate practice spells succe ss: Why grittier competitors triumph at the National Spelling Bee. Social Psychological and Personality Science 2 174 181 Duckworth, A.L., Kraft Robinson, C. E. (Unpublished). True grit: Trait level perseverance and passion for long term goals predic ts effectiveness and retention among novice teachers. Duckworth, A.L., Peterson, C., Matthews, M.D., & Kelly, D.R.(2007). Grit: Perseverance and passion for long term goals. Journal of Personality and Social Psychology 92 (6 ) 1087 1101. Duckworth, A.L & Quinn, P.D. (2009). Development and validation of the Short Grit

PAGE 47

40 Scale (Grit S). Journal of Personality Assessment 91 166 174. Duckworth, A.L., Quinn, P.D., & Seligman, M.E.P. (2009). Positive predictors of teacher effectiveness. Journal of Positiv e Psychology 19 540 547 Duncan, G. & Magnuson, K. (2011). The nature and impact of early achievement skills, attention skills, and behavior problems. In G. Duncan & R. Munane (Eds.), Whither opportunity? Rising inequality, schools, and children's life chances New York: Russell Sage Foundation. 47 70. Freeman, C.E. (2004). Trends in the Educational Equity of Girls and Women: 2004 Washington, DC: National Center for Education Statistics. Frey, M.C., & Detterman, D.K. (2004). Scholastic assessment or g? The r elationship b etween the scholastic assessment test and general cognitive a bility Psychological Science 15 (6). 373 378 Goldin, C.D., & Katz, L.F. (2008). Mass Higher Education in the 20 th Century. The Race Between Education and Technology (p p. 247 286). United States: Harvard University Press. Heaton, T.B,Chadwick, B.A., & Jacobson, C.K. (2000). Statistical handbook on racial groups in the United States. Phoenix, Arizona: Oryx Press. Henke, R.R., Chen, X., Geis, S. & Knepper, P. (2000). P rogress through the teacher pipeline: 1992 93 college graduates and elementary/secondary school teaching as of 1997. Washington, DC: US Department of Education, Office of Educational Research and Improvement. Horn, L. & Berger, R. (2005). College persis tence on the rise? Changes in 5 year degree

PAGE 48

41 completion and postsecondary persistence rates between 1994 and 2000 (Report No. NCES 2005 156) U.S. Department of Education, National Center for Education Statistics. Hudson, L., Aquilino, S., & Kienzl, G. (2005). Postsecondary participation rates, by gender and race/ethnicity: 1974 2003 (Report No. 2005 025). Washington, DC: U.S. Department of Education, National Center for Educational Statistics. Jacob, B A. (2002) cognitive skills, returns to school and the gender gap in higher education. Economics of Education Review 21 (6), 589 5 98. Levitz, R.S., Noel, L.,& Richter, B. J. (1 999). Strategic moves for retention success. New Directions for Higher Education 1999 (108), 31 49. Lawrence, I., Rigol, G.W., Van Essen, T. & Jackson, C.A. (2002). A historical perspective on the SAT 1926 2001 (Report No. 2002 7). New York: College Ent rance Examination Board. Long, B.T. (2002). Attracting the best: The use of honors programs to comp lete for students Harvard Graduate School of Education: Spencer. Manski, C.F. (1992). Parental income and college opportunities (DSC Report Series). Democracy Study Center. Moore, B. ( 2008). Assessing SAT validity at liberal arts institutions (Undergr aduate Thesis ). Retrieved from NCF Digital Collections. As Prepared for Delivery

PAGE 49

42 Organization for Economic Co o peration and Devolopment (2010). Education at a Glance 2010: OECD Indicators OECD Publishing. doi: 10.1787/eag 2010 en Peter, K & Horn L. ( 2005 ) Gender differences in participation and completion of undergraduate education and how they have changed over time NCES 2005 1 69. Washington, DC: National Center for Education Statistics Peterson, C.H., Casillas, A., & Robbins, S.B. ( 2006). The Student Readiness Inventory and the Big Five: Examining social desirability and college academic performance. Personality and Individ ual Differences 41 (4) 663 673. doi:10.1016/j.paid.2006.03.006. Porter, O. F. (1990). Undergraduate completion and persistence at four year colleges and universities : Detailed Findings. Washington, DC: National Institute of Independent Colleges and Un iversities. Reardon, S. F. (2011). The widening academic achievement gap between the rich and the poor: New evidence and possible explanations. In G. Duncan & R. Munane (Eds.), Whither opportunity? Rising inequality, schools, and children's life chances New York: Russell Sage Foundation. 91 116. Robbins, S.B., Allen, J., Casillas, A., Peterson, C. H., & Le, H. (2006) Unraveling the differential effects of motivational and skills, social, and self management measures from traditional predictors of col lege outcomes. Journal of Educational Psychology 98 (3), 598 616. doi: 10.1037/0022 0663.98.3.598

PAGE 50

43 Robbins, S.B.,Lauver, K.,Le, H., Davis, D., Langley, R., & Carlstrom, A. (2004) Do psychosocial and study skill factors predict college outcomes? A meta analysis. Psychological Bulletin 130 (2), 261 288. doi: 10.1037/0033 2909.130.2.261 Schmidt A. E. & Camara, W. J. (2003).Group differences in standardized test scores and other educational indicators. In. R. Zwick (Ed.), Rethinking the SAT: The fu ture of standardized testing in university admissions New York: NY. Routledge Stanford, B.H. (2001). Reflections of resilient, persevering urban teachers. Teacher Education Quarterly 28 75 87. Stinebrickner, R., & Stinebrickner, T.R. (2003a). Working during school and academic performance. Journal of Labor Economics 21 (2), 473 491. doi: 10.1086/345565 Stinebrickner, R., & Stinebrickner, T.R. (2003b). Understanding educatio nal outcomes of students from low income families: Evidence from a liberal ar ts college with a full tuition subsidy program. Journal of Human Resources 38 (3), 591 617. doi: 10.3368/jhr.XXXVIII.3.591 Snyder, T., D., Dillow, S. A.,& Hoffman C. M.(2008) Digest of education statistics 2007 NCES 2008 022. Washington, DC: National Center for Education Statistics Quillian, L. (2006). New Approaches to Understanding Racial Prejudice and Discrimination. A nnual Review o f Sociology 32 299 328. doi:10.1146/annurev.soc.32.061604.123132 Quillian, L. (2008). Does Unconscious Racism Ex ist?. Social Psychology Quarterly

PAGE 51

44 71 (1), 6 11. doi:10.1177/019027250807100103 Willingham, W. (1985). Success in college: The role of personal qualities and academic ability New York: College Entrance Examination Board.

PAGE 52

45 Table 1 Profiles of Student Grit Scores Grit Score Sample Profile Explanation 0 No multi year involvement in any activities 1 Played violin in school orchestra for 3 year. Did not earn any awards; N o other multi year activities. 1 pt for multi year activity 2 Member of the cross co untry team for 2 year s and earned a varsity letter; N o other multi year activities 1 pt for multi year activity +1 pt for moderate achievement in that activity 3 Member of a youth group for 3 years but no leadership roles; Member of National Honors Socie ty for 2 years; Member of Latin Club for 4 years 1pt for multi year activity; 1pt for multi year activity; 1pt for multi year activity 4 Historian of Spanish Honors Society for 2 years; Player on the volleyball team for 3 years; Participated in choir for 4 years 1pt for multi year activity +1 pt for moderate achievement in that activity; 1pt for multi year activity;1 pt for multi year activity 5 President of the Political S cience club for 3 years; Member of Key Club for 2 years; Member of Beta Club fo r 3 years 1pt for multi year activity + 2pts for high achievement in that activity; 1 pt for multi year activity; 1 pt for multi year activity 6 Member of the varsity soccer team for 4 years and Captain senior year; Worked with school newspaper for 3 yea rs and was a section editor junior year. Member of Best Buddies club for two years 1pt for multi year + 2pts for high achievement in that activity; 1 pt for multi year +1 pt for moderate achievement in that activity; 1 pt for multi year. 7 Member of jaz z band for 4 years, 1 st chair flute junior and senior year; member of marching band for 4 years, lettered junior and senior year; Member contest 1pt for multi year + 2pts for high achievement in that activity ; 1 pt for multi year +1 pt for moderate achievement in that activity; 1 pt for multi year +1 pt for moderate achievement in that activity 8 Member of yearbook for 4 years and editor in chief senior year; Member of choir for 3 years and made All State junior and senior year; Member of art club for 2 years and was Treasurer senior year 1pt for multi year + 2pts for high achievement in that activity;1pt for multi year + 2pts for high achievement in that activity; 1 pt for multi year +1 pt for moderate achievement in that activity 9 Member of marching band for 4 years and dru m owl for 4 years and President junior and senior year; Founder an d Vice President of Environmental Action C lub for 2 years 1pt for multi year + 2pts for high achievement in that activity;1pt for multi year + 2pts for high achievement in that activity;1pt for multi year + 2pts for high achievement in that activity

PAGE 53

46 Tabl e 2 Means and Standard Deviations by Entering Class Age Income SAT WHSGPA Entering Class M SD M SD M SD M SD 2002 17.84 .52 10.76 .35 1322.98 109.21 3.95 .36 2003 17.93 .49 10.74 37 1308.19 106.23 3.84 .36 2004 17.91 .56 10.75 .33 1309.03 110.43 3. 95 .37 Note. p < .05

PAGE 54

47 Table 3 Note. p < .05 Frequencies of Gender and Race by Entering Class Gender Race Entering Class Men Women White People of Color 2002 47 108 140 15 2003 61 93 124 29 2004 74 112 149 37

PAGE 55

48 Table 4 Race Frequencies by Gend er Gender Race Men Women White 150 263 Person of Color 32 49

PAGE 56

49 Table 5 Means and Standard Deviations by Race Age Income SAT*** WHSGPA* Race M SD M SD M SD M SD White 17.95 .57 10.74 .34 1323.31 100.57 3.93 .37 Person of Color 17.88 .62 10.81 .39 1266.46 108.68 3.83 .37 Note. p < .05, *** p < .001

PAGE 57

50 Table 6 Grit Score Means and Standard Deviations by Entering Class Entering Class N M SD 2002 154 5.08 2.42 2003 150 4.79 2.35 2004 180 4.62 2.22

PAGE 58

51 Table 7 Grit Score Means and Standard Devi ations by Gender and Race Gender Race Men Women White Person of Color M 4.97 4.73 4.80 4.93 SD 2.35 2.32 2.38 4.82

PAGE 59

52 Table 8 Pearson Correlations for All Continuous Variables WHSGPA SAT Income Grit Age WHSGPA .23** .07 .00 .02 SAT .14* .09 .04 Income .02 .04 Grit .04 Note. ** p < .01

PAGE 60

53 Table 9 Summary of Regression Model Predicting Graduation Predictor SE Odds Ratio p WHSGPA .32 .11 1.38 .003 SAT .01 .11 1.00 .966 Income .08 .10 1.08 .460 Grit .21 .1 1 1.24 .041 Person of Color .19 .28 .82 .495 Male .22 .23 1.24 .337 2002 .35 .25 .71 .164 2003 .26 .25 .77 .301 Age .30 .20 .74 .129 Note. N = 446 after excluding students with missing data from the analysis.

PAGE 61

54 Figure 1

PAGE 62

55 Figure 2

PAGE 63

56 Figure 3

PAGE 64

57 Appendix List of Specific Moderate and High Achievement Activities Moderate Achievements Editor of Literary Magazine General Officer Position in an Organization Earning a Letter in either a Sport or Marching Band Playing a Varsity Sport Coordinator or Organizer Competing in a Group Sport or Music Group at the State Level Being in an Auditioned Music Group Participation in a Magnet Extracurricular Activity Founder of a Club Ranked Chair in Music Group General Board Position in an Organization Support ing Role in a Play High Achievements Editor in Chief or Editor of Section President or Vice President of an Organization Capitan of a Team Placing in a Group Sport or Music Group at the State Level Competing at the National Level in Any Field Achieving a Black Belt in Any Form of Martial Arts Teaching or Being at the Professional Level of a Performance Hobby Leading Role in a Play Director of a Play Section Leader in a Music Group