Examining Lottery Scholarships in Tennessee
Charles E. Menifield
Maxine Smith Fellow
University of Memphis
Division of Public and Nonprofit Administration
901-678-5527
Paper prepared for presentation at the ABFM Conference, Chicago IL October 2008.
Abstract
Scholarship retention is one of the most significant problems with the Tennessee Lottery. Evidence to date suggests that the problem is chronic to certain populations in the state’s colleges and universities. This research examines the lottery scholarship retention data to determine what factors affect scholarship retention. The data in the research was taken from Tennessee Higher Education Commission data sets for the fall 2005, fall 2006, and the fall 2007 semesters. The data shows that African Americans are the most likely to loose their scholarship than any other group at every undergraduate level. In addition, the data shows that high school grade point average, undergraduate grade point average, gender, and adjusted gross family income are highly correlated with scholarship retention.
Introduction
Given budget shortfalls in states around the country, many states have begun to use alternative means to financing primary, secondary and higher education. Despite an increase in funding for higher education scholarships, retention has become increasingly important as more and more students from disadvantaged backgrounds loose their scholarships.
Using data from the Tennessee Higher Education Commission, a model is developed to assess factors that affect scholarship retention. Special interest is paid to demographic, education, and economic factors. The analysis test the null hypothesis that there is no relationship between scholarship retention and race, gender, ACT scores, high school grade point average, undergraduate grade point average, pell grant eligibility, estimated family contribution, and adjusted gross family income. Ideally, this analysis will provide lottery officials with evidence that can be used to create programs that will give lottery recipients the greatest opportunity to maintain their scholarship.
Background on Lottery Scholarships in Tennessee
The Tennessee Education Lottery Scholarship program was created with the passage of the Tennessee Education Lottery Implementation Law in 2003 (Tennessee Code Annotated §§ 4-51-101 et.seq). The program is managed by the Tennessee Lottery Scholarship Corporation.
The Tennessee Education Lottery Scholarship Program currently has five scholarship programs with different monetary awards, grade point requirements, minimum ACT scores, minimum grade point averages, and maximum income requirements (limited to the ASPIRE and Access Awards). These include the HOPE Award, General Assembly Merit Scholarship (GAMS), ASPIRE Award, Access Award, and the Wilder-Naifeh Technical Skills Grant. The data in Table 1 summarizes the various requirements and award levels. As shown the ASPIRE award has the highest monetary value followed by the GAMS, HOPE, Access and Wilder-Naifeh respectively. The GAMS award has the highest minimum grade point average requirement at 3.75 followed by the HOPE, ASPIRE, and Access awards respectively. The Access award has the lowest minimum grade point average of 2.75.
Table 1 Tennessee Lottery Scholarships[1]
HOPE GAMS ASPIRE Access Wilder-Naifeh
Amount (4-yr) $4000 $5000 $5500 $2750 N/A
Amount (2-yr) $2000 $3000 $3500 $1750 $2000
Minimum HS GPA 3.00 3.75 3.00 2.75 N/A
Minimum ACT Comp. or 21 & 29 or 21 & 18-20 N/A
Family Adj. Gross Inc. N/A N/A $36k or less $36k or less N/A
College Retention GPA Cumulative 2.75 at 24 hours, Cum. 2.75 at Satisfactory
Cumulative 3.0 at 48 hours, 24 hours allows Academic
72, and 96 hours qualification Progress
For HOPE
The ASPIRE and Access awards are the only two scholarships that require a maximum family income level (adjusted) at $36,000. The minimum grade point average for retention of the HOPE, GAMS, ASPIRE, and the Access scholarships is 2.75 at the 24 hour thresholds and 3.0 at the 48, 72 and 96 hour threshold. The Wilder-Naifeh scholarship requires satisfactory progress towards completing the degree.
Summary of the Literature
Data assessing the impact of lottery scholarships on education budgets and policies is fairly well developed. Policy implementation, converting intent to action, is a dynamic and evolutionary process. To evaluate the effectiveness of policy implementation, researchers focus on the degree or level of achievement of outcomes, adherence to statutory requirements, accomplishment of specific goals and objectives, and/or impact of stakeholders and actors (Mazmanian & Sabatier1983; O’Toole 2000). From a policy accountability perspective, researchers have questioned whether state lotteries act as an implicit regressive tax (Clotfelter and Cook, 1987; Hansen, Miyazki, & Sprott, 2000; Pirog-Good and Mikesell 1995; Price & Novak 2000, Price & Novak 1999; Borg & Mason 1990); Brinner & Clotfelter 1975), anticipated net revenues actually occur (Davis, Filer, & Moak 1992); how the lottery structure affects demand and revenue (Vrooman 1976); lottery revenue earmarked for education actually results in positive effects on education expenditures (Spindler 1995; Dee 2004; Campbell 2003; Land & Alsikafi 1999; Stanley & French 2003; Miller & Pierce 1997; McCrary & Condrey 2003) or lottery revenues become more of a legislative “shell game” (Perlman1998; Allen 1991); and whether the states’ purported purpose to maximize revenue occurs, especially over time (Garrett 2001; Laschober 1989; Mikesell 1994; Mikesell & Zorn 1986; Mikesell & Zorn 1988; McCrary & Condrey 2003). This important body of research addresses important political and economic issues. However, programmatic-level accountability of state lottery program administration, processes and procedures, has received limited research attention.
Research Design, Data and Methodology
The data for this paper was derived from data sets collected by the Tennessee Higher Education Commission. The data was examined at the individual level and includes an array of demographic, economic, and education variables for the fall 2005, fall 2006 and fall 2007 semesters[2]. The three original data sets contained a total of 149,616 students.[3] However, after removing the students with missing values, the final pooled data set had 46,355 cases in the logistic regression analysis and 62,143 cases in the bivariate analysis.[4]
The analysis begins by providing descriptive data on the percentage of scholarships retained by race, gender, ACT scores, adjusted gross family income, high school grade point average and undergraduate grade point average. The second set of analysis provides the summary statistical result for the crosstabs examining the relationship between scholarship dispositions and several dependent variables. Last, the analysis uses a logistic regression model to determine the affects of numerous dependent variables on the independent variable scholarship disposition. Class level was used as a control variable since this is a natural break among students. Hence, there are four logistic regression models, one for each undergraduate class group.
Y = Scholarship Disposition (Retained or Lost)
Demographic Variables
X1 = Race (Dummy Variable: White = 1, Nonwhite = 0)
X2 = Gender (Male, Female)
Economic Variables
X3 = Family Contribution to Education ($)
X4 = Pell Grant Eligible (Yes = 1 or No = 0)
X5 = Adjusted Gross Family Income ($)
Education Variables
X6 = High School Grade Point Average (0->4.0)
X7 = ACT Score (0-36)[5]
X8 = Undergraduate Grade Point Average (0-4.0)
Control Variable
X9 = Class Level (Freshman, Sophomore, Junior, & Senior)
Findings
The data in Table 2 shows Tennessee lottery scholarship recipients by demographic, educational, and economic factors. As seen, African Americans are most likely to loose their scholarship than any other racial group at every level. However, the percentage of African Americans that loose their scholarship over the four levels diminishes as they continue to matriculate. Overall, student at higher levels are a lot more likely to retain their scholarship than students at lower levels. The data also shows that women are more likely to retain their scholarship than men. However, this gap decreases dramatically by the senior year. In fact, there is no statistical difference between the two groups by the senior year.
The data also shows that ACT scores are highly correlated with scholarship retention rates. Students who score in the top all 25% of test takers are much more likely to retain their scholarship than the students who make a lower score. As students move from one class level to the next, the ACT score becomes less important. Essentially, the scholarship retention rates for all of the students increases at comparable rates as they increase their level.
Table 2 Lottery Recipients by Demographic, Edu. and Eco. Factors (Fall Sems 2005-2007)
Freshmen Sophomore
Scholarship Retention Scholarship Retention
Lost Retained Lost Retained
Race
Black 57.9% 42.1% 49.3% 50.7%
White 40.6 59.4 32.3 67.8
Hispanic 44.1 55.9 40.2 59.8
Asian 45.6 54.4 28.4 71.6
Am. Indian 33.3 66.7 23.3 76.7
Alaskan 37.5 62.5 12.0 88.0
Gender
Male 48.8% 51.2% 37.4% 62.6%
Female 39.2 60.8 31.3 68.7
ACT Scores
<20 50.5% 49.5% 53.6% 46.4%
21-22 46.3 53.7 40.0 60.0
23-25 35.6 64.4 26.6 73.4
>25 27.3 72.7 13.6 86.4
H.S. Grade Point Average
<3.10 61.9% 38.1% 58.1% 41.9%
3.11-3.48 45.6 54.4 43.6 56.4
3.49-3.71 28.9 71.1 24.7 75.3
>3.71 16.3 83.7 10.5 89.5
Undergraduate Grade Point Average
<2.81 78.7% 21.3% 89.3% 10.7%
2.81-3.23 18.5 81.5 24.8 75.2
3.24-3.60 15.8 84.2 7.0 93.0
>3.60 14.2 85.8 4.2 95.8
Adjusted Gross Family Income
<$35,000 44.9% 55.1% 41.2% 58.8%
$35,000-$67,000 41.0 59.0 34.4 65.6
$67,001-$97,000 37.7 62.3 31.4 68.6
>$97,000 34.7 65.3 23.7 76.3
N= 4435 5862 8873 17358
Table 2 Lottery Recipients by Demographic, Edu. and Eco. Factors (Fall Sems 2005-2007)-cont
Junior Senior
Scholarship Retention Scholarship Retention
Lost Retained Lost Retained
Race
Black 45.3% 54.7% 36.2% 63.8%
White 19.3 80.7 15.2 84.8
Hispanic 30.5 69.5 26.6 73.4
Asian 16.6 83.4 12.7 87.3
Am. Indian 16.7 83.3 16.7 83.3
Alaskan 0 100 22.2 77.8
Gender
Male 25.0% 75.0% 18.0% 82.0%
Female 19.9 80.1 16.2 83.8
ACT Scores
<20 44.9% 55.1% 32.7% 67.3%
21-22 28.4 71.6 23.5 76.5
23-25 17.9 82.1 14.9 85.1
>25 9.3 90.7 10.0 90.0
H.S. Grade Point Average
<3.10 46.7% 53.3% 24.2% 75.8%
3.11-3.48 33.4 66.6 26.9 73.1
3.49-3.71 15.6 84.4 17.0 83.0
>3.71 6.9 93.1 9.9 90.1
Undergraduate Grade Point Average
<2.81 90.6% 9.4% 87.8% 12.2%
2.81-3.23 25.8 74.2 29.3 70.7
3.24-3.60 2.9 97.1 9.4 90.6
>3.60 1.7 98.3 6.7 93.3
Adjusted Gross Family Income
<$35,000 28.5% 71.5% 21.7% 78.3%
$35,000-$67,000 23.1 76.9 16.8 83.2
$67,001-$97,000 19.7 80.3 16.2 83.8
>$97,000 16.9 83.1 14.1 85.9
N= 3170 11251 1893 9301
Source: Data sets provided by the Tennessee Higher Education Commission.
High school grade point average is also correlated with retention rates. Students with higher high school grade point averages are more likely to retain their scholarship when compared to students with lower high school grade point averages. However, the level of significance for this variable diminishes over time. In fact, if a student with a lower high school grade point average stays in college until their senior year, the likelihood that they would retain their scholarship is 76% when compared to 38% as a freshman.
Undergraduate grade point average is highly correlated with scholarship retention. The relationship is consistently significant over time. With respect to income, students from higher income families are more slightly more likely to retain their scholarships. However, the percentage difference between the lowest income group and the highest income group diminishes considerably by the senior year.
The data in Table 3 shows that we can reject the null hypotheses that scholarship retention is not related to each of the dependent variables. That is, there is a statistically significant relationship between scholarship retention and each of the dependent variables at every level. However, the table does reveal some interesting finding. First, although the level of significance for ACT scores and high school grade point average are significant at the .00 level, the coefficient does decrease over time. This corresponds to the descriptive data found in Table 2. The same finding is true for Pell grant eligibility, gender, estimated family contribution and adjusted gross family income. The coefficients for the race dummy variable vacillate over the various levels. Nonetheless, the relationship is significant.
Table 3 Pooled Bivariate Analyses of Independent and Dependent Variables (2005-2007)[6]
Scholarship Retention
Freshmen Sophomore Juniors Seniors
Coefficients Coefficients Coefficients Coefficients
Race 291.9 *** 201.1 *** 293.9 *** 239.9 ***
Gender 110.3 *** 92.7 *** 40.2 *** 9.5 ***
Pell Grant Eligibility 115.5 *** 149.2 *** 48.8 *** 23.4 ***
High School G.P.A. 1831.6 *** 1470.3 *** 1125.8 *** 337.5 ***
ACT Scores 1547.7 *** 1537.8 *** 956.6 *** 442.1 ***
Adj. Gross Family Inc. 200.1 *** 246.1 *** 78.1 *** 49.2 ***
Undergraduate G.P.A. 8983.1 *** 6046.8 *** 5233.5 *** 1827.2 ***
Est. Family Contribution 212.98 *** 208.6 *** 66.7 *** 46.1 ***
Pearson Chi-Square Significance: * = .05, ** = .01, *** = .00.
Source: Data sets provided by the Tennessee Higher Education Commission.
The data in Table 4 shows provides a regression model for each undergraduate level. The data shows that gender, high school grade point average, undergraduate grade point average, and ACT score are consistent predictors of lottery scholarship retention at every academic level. Adjusted gross family income and estimated family contributions are sporadically significant in the models. Pell grant eligibility is significant in one of the models. Interestingly, the race variable is not significant in the sophomore model and only moderately significant in the freshman model despite the findings in the bivariate analysis indicating the importance of race.[7] It is also interesting to note that the explanatory power of the models decreases over time as the r squared diminishes to .261 in the senior scholarship model from a high of .536 in the freshman model.
Table 4 Pooled Logistic Regression Models by Class Levels^
Model 1 Freshman Scholarship Retention Model 2 Sophomore Scholarship Retention
Coefficient Coefficient
Race (Dummy) .223 (.073) * Race (Dummy) .028 (.079)
Gender .173 (.050) *** Gender .245 (051) ***
Est. Family Contribution .000 (000) *** Est. Family Contribution .000 (.000) *
Pell Grant .091 (.061) Pell Grant -.100 (.062) *
Adj. Gross Family Inc. .000 (.000) Adj. Gross Family Inc. .000 (000) *
High School G.P.A. .804 (.053) *** High School G.P.A. .736 (.058) ***
Undergraduate G.P.A. 2.002 (.043) *** Undergraduate G.P.A. 2.114 (055) ***
ACT Score .114 (.008) *** ACT Score .122 (.008) ***
Nagelkerke R-Square = .536, N=14,771 cases. Nagelkerke R-Square = .477, N=11,624 cases.
Model 3 Junior Scholarship Retention Model 4 Senior Scholarship Retention