« PreviousContinue »
Figure 10. Default rate by school enrollment in 1993 and 1994 school years
Multiple Regression Results
The figures in this section present the school characteristics that have had statistically significant relationships with the outcomes for three or more of the five years for which we have data. The height of the trend lines in the figures reflect the size of the net effects of the characteristics, and these effects can be negative as well as positive. Negative effects are shown by lines that go below the zero point on the vertical axis. No figures are entered on the trend lines for years when a school characteristic did not have a significant relationship with an outcome. A trend line that does not extend the full width of the figure indicates that the characteristic was not significant in the years that are not shown.
The measures of school characteristics shown in the figures are of two kind: categorical and continuous. Categorical variables reflect either or conditions. A school is a main or a branch campus. A school has accreditation by an agency in addition to the Commission or it does not. Categorical variables are interpreted in comparison to the opposite condition.
Continuous variables can have many values. Most of the continuous variables reflect the percentage of students with certain characteristics, such as having received different types of financial aid. These percentages are based on the total enrollment of the individual schools. Continuous variables are interpreted as the rate of change in the outcome measure for a unit change in the measure of a school characteristic. The trend lines for the continuous variables show how much the outcomes change when the school characteristics change 10 units.
Characteristics Related to Graduation
Figure 11 presents six school characteristics that have consistently been found to have major impact upon the graduation rates calculated from the annual total data.
Enrollment. The top two lines in Figure 11 reflect enrollment groupings of 300 or less and 301 to 600. These groupings are categorical variables of a special sort. They are interpreted in comparison to schools in the largest enrollment group-901 or more students. For all five years, schools with enrollments of 300 or less have had graduation rates 8 to 13 percentage points higher than schools with enrollments of 901 or more. Schools with enrollments of 301 to 600 have had graduation rates 5 to 9 percentage points higher than schools with enrollments of 901 or more. Schools with enrollments of 601 to 900, however, did not have significantly higher rates than those in the largest enrollment group.
NOTE: Years without data entries on the trend lines indicate that the characteristics did not have statistically significant net effects on the outcome in the missing years.
Figure 11. Net effect of selected school characteristics on graduation rate
It bears repeating that these estimates of the effects of size of enrollment are independent of the other variables that also influence graduation rates. Stated another way, in 1994 when the effect of other school characteristics, such as the percent of Pell recipients and the average length of programs, were controlled, schools with enrollments of 300 or less had graduation rates that averaged 12 percentage points higher than schools with enrollments of 901 or more.
Main or branch campus. Four of the five school years main campuses had graduation rates 4 percentage points higher than branch campuses. There is no entry on the trend line for 1993 indicating that this characteristic was not statistically significant that year. These findings suggest a combination of factors, such as facilities, equipment, and instruction, are likely to increase the holding power of a main campus in comparison to a branch.
Other accreditation. Each of the five years, about one-sixth of the schools reported they had or were a candidate for accreditation in addition to that from the Commission. Schools that reported such additional accreditation had graduation rates about 3 to 4.5 percentage points higher than schools accredited solely by the Commission. Holding more than one accreditation is a consistent indicator of higher graduation rates.
Percent receiving Pell grants. The financial aid variables indicate the percentage of enrollment at a school that received different types of aid. These variables range from 0 to 100 percent. Across all schools, half of all students received Pell grants. The percentage increased from 1990 to 1992, held steady in 1993, and dropped to its 1990 level in 1994.
The trend line in Figure 11 indicates that for a 10 point increase in the percentage of enrollment that received Pell grants, the percentage of graduates decreased 1 to almost 2 percentage points.
The results for the Pell variable do not mean that receiving Pell grants make it less likely that students will graduate. Pell grants are available only to those students whose own or family income is below the level defined by the federal government as poverty. High percentage of Pell recipients at a school reflect a high percentage of students from poverty families. Students from such families traditionally are the most difficult to serve in educational settings.
Average program length. The trend in average program length (measured in weeks) has been to longer programs. In 1994, the average length at Commission schools was almost 47 weeks, in comparison to 34 weeks in 1990, a 38 percent increase. When other characteristics are held constant, schools with shorter programs have higher graduation rates than schools with longer programs. For every 10 week increase in program length, graduation rates decrease by about 2 to 3.5 percentage points.
Other characteristics. In addition to the six school characteristics shown in Figure 11, three others had statistically significant relationships with graduation rates three of the five years: percentage of enrollment receiving Stafford loans, percentage receiving Supplemental Loans to Students, and turnover among part-time faculty. They are not shown in the figure because their influence has always been less than a change of 1 percentage point in graduation rate for a 10 point change in the variables.
Characteristics Related to Withdrawal
All students who enroll during a given school year do not graduate or withdraw during that school year. Each of the years for which we have data, about 40 to 45 percent of both full-time and parttime enrollment neither graduated nor withdrew. Because of these continuing students, it was necessary when analyzing the annual total data to establish a definition of graduation that did not penalize schools for continuing students.
The definition adopted was the number of students graduating divided by the number graduating plus the number withdrawing. Withdrawal rates were defined as the number withdrawing divided by the number enrolled in a school year. Since these definitions differ, withdrawal is not just the opposite of graduation. Some of the variables that influence graduation also influence withdrawal, but they are not identical. Table 1 presents six characteristics found to have significant relationships with withdrawal three of the five years.
Table 1. Net effect of selected school characteristics on withdrawal rates
Because of the difficulty in presenting the results for withdrawal in graphic form, the net effect of the variables on withdrawal are presented in tabular form in Table 1. It is noteworthy that three of the six characteristics in Table 1 were also included in Figure 11: enrollments of 300 or less, enrollments of 301 to 600, and other accreditation. In Figure 11, however, these lines were above the zero point on the vertical axis. This indicated that these variables had a positive relationship with graduation rate. In Table 1, these characteristics indicate a negative relationship. Withdrawal rates were lower in schools that had the characteristics charted than they were in schools with the opposite. characteristics. As in all the charts in this section, these lines show the independent effect of the characteristics, holding all other characteristics for which we had measures constant.
Note that in Table 1 there are some missing data. The years for which there are no data, the characteristics did not have statistically significant relationships with withdrawal rates.
There are three characteristics in Table 1 that were not present in Figure 11. One of these, separate facilities, is negative. The other two, default rate and percentage of Ability-To-Benefit (ATB) students are positive.
Separate facilities. A relatively small proportion, about one-sixth, of ACCSCT schools and colleges have reported separate facilities over the five school years. For the first three years, separate facilities had a consistent, negative relationship with withdrawal. The rates at schools with separate facilities were about 1.5 to 2 percentage points lower than at schools without such facilities. That relationship has not been significant during the past two years. In some years, separate facilities have also been found to have significant relationships with graduation, but this has not been consistent.
Ability to benefit. As noted in connection with Figure 4, the percent of enrollment classified as Ability-To-Benefit (ATB) has steadily declined over the five years. In 1994, only 9 percent of students were ATB. Despite this decline, the percentage of ATB continues to have a strong relationship with default rate, which will be discussed later. Prior to 1994, it also had a significant net effect on withdrawal. Figure 8 indicates that for every 10 point increase in the percentage of ATB students, withdrawal rates increased by 1.3 to 2.5 percentage points. In 1994, the relationship was not significant.
The declining number of ATB students may be weakening relationships previously observed. Schools are admitting fewer ATB students, and it is likely they are being more selective with those they admit.
Default rate. The percent of students who default on their Stafford loans obviously cannot be a cause of withdrawal. Default occurs after withdrawal. We included it as an explanatory variable because we thought default rates might reflect certain characteristics of schools not captured by the other variables in our analysis. Default rate had a consistent relationship with withdrawal independent of all the other school characteristics examined.
It will be recalled that default rates are for the students who left schools, either by graduation or withdrawal, two years prior to the school year analyzed. Measures for school characteristics in a given school year are correlated with default rates for the students who had left the school two years previously. If default rates were unstable, varying widely from year to year, we would not have found the consistent relationships shown in Table 1. Direct one-to-one correlation of default rates across school years typically yields r values in the .70s and 80s. Schools and colleges that have low default rates one year tend to have low rates the next year, and those with high rates one year tend to have high rates the next year.
Other variables. There were four other school characteristics that had significant relationships with withdrawal that are not shown in Table 1: turnover of both full-time and part-time faculty, average length of programs, and percentage of enrollment receiving Pell grants. With one exception, the estimates of the effects of these characteristics were 1 percentage point or less for a 10 point change in the measures of the characteristics.