D. RESULTS
V. LINEAR PROBABILITY MODEL (LPM) OF DEP ATTRITION USING
3. Men and Women LPM Regression Analysis
Presented in Table 33 are variable descriptive statistics for men and women combined, including 474,752 observations in the sample. The data show an overall attrition rate of 21.12 percent. Most observations fall within the AFQT category II at 35.31 percent, followed by category IIIB at 31.84 percent. The most populous age category is the 18-to-20 year olds, which make up 53.51 percent of the sample, followed by those less than 18 years of age at 21.91 percent.
Table 33. Summary of LPM Variables for Men and Women
Name Variable Description Obs Mean Std. Dev. Min Max
Did not pass state exit
exam = 1 if true, 0 false 474752 0.3402 0.4738 0 1
Name Variable Description Obs Mean Std. Dev. Min Max
asian_pi Asian/Pacific Islander = 1 if true, 0 false 474752 0.0407 0.1976 0 1 black Black = 1 if true, 0 false 474752 0.1802 0.3843 0 1 Source: Derived from data provided by Defense Manpower Data Center (DMDC), 2009.
Table 34 presents the male and female LPM regression results for all Tiers combined and individual education credentials grouped into respective Tiers. The R-squared value is 0.0527, and the probability of attrition for the base individual is 0.2476.
The data show that the probability of attrition decreases by 0.0400 for AFQT category I individuals and decreases by 0.0224 for those in AFQT category II. The probability of attrition for persons in category IIIB increases by 0.0029. Category I and II variables are both statistically significant at the 1 percent level while category IIIB is significant at the 10 percent level. All age variables are significant at the 1 percent level. The less-than-18 variable decreases the probability of attrition by 0.0232, while all other age variables increase the probability of attrition as the age of the individual increases. The data show that the probability of attrition for Tier II individuals increases by 0.0082 and the probability of attrition for Tier III individuals increases by 0.0216. Both Tier II and Tier III variables are statistically significant at the 1 percent level. The data show that the probability of attrition increases for women by 0.0949 and is significant at the 1 percent level.
Table 34. Binary LPM, Tiers, Regression Results for Men and Women
Variable Coefficient Robust Std. Error p-Value
I -0.0400 0.0027 0.0000
II -0.0224 0.0015 0.0000
IIIB 0.0029 0.0016 0.0640
female 0.0949 0.0016 0.0000
und_18 -0.0232 0.0016 0.0000
age21_23 0.0129 0.0017 0.0000
age24_26 0.0242 0.0026 0.0000
age27_29 0.0283 0.0039 0.0000
ovr_30 0.0340 0.0045 0.0000
ave_tid 0.1499 0.0013 0.0000
tierII 0.0082 0.0026 0.0020
tierIII 0.0216 0.0043 0.0000
ave_mxwt -0.0228 0.0012 0.0000
youth -0.1874 0.0021 0.0000
metro -0.0006 0.0016 0.7230
mar -0.0595 0.0036 0.0000
dep 0.0132 0.0031 0.0000
asian_pi -0.0427 0.0030 0.0000
Variable Coefficient Robust Std. Error p-Value
race_oth 0.0080 0.0025 0.0010
ave_mep -0.0031 0.0014 0.0250
ave_rtc 0.0041 0.0037 0.2670
fy2000 -0.1041 0.0040 0.0000
fy2001 -0.0954 0.0040 0.0000
fy2002 -0.1026 0.0040 0.0000
fy2003 -0.1040 0.0041 0.0000
fy2004 -0.0940 0.0041 0.0000
fy2005 -0.0901 0.0042 0.0000
fy2006 -0.1059 0.0042 0.0000
fy2007 -0.1131 0.0041 0.0000
fy2008 -0.1209 0.0043 0.0000
Intercept 0.2476 0.0122 0.0000
R-squared Number of obs. Prob > F
0.0527 474,752 0.0000
Source: Derived from data provided by Defense Manpower Data Center (DMDC), 2009.
Base individual is Tier I white male with category IIIA AFQT who entered DEP in 1999.
Table 35 presents the LPM regression results for men and women from all Tiers, divided by each education credential. The R-squared value is 0.0549 and the probability of attrition for the base individual is 0.2335. The data show that the probability of attrition decreases by 0.0369 for persons in AFQT category I and decreases by 0.0204 for those in AFQT category II. The probability of attrition for persons in AFQT category IIIB increases by 0.0021. AFQT category I and II variables are both statistically significant at the 1 percent level, while AFQT category IIIB is not significant. All age variables are significant at the 1 percent level. The less-han-18 variable decreases probability of attrition by 0.0366, while all other age variables increase the probability of attrition as the age of the individual increases. The data show that the probability of attrition for individuals with a GED increases by 0.0227 and the probability of attrition for individuals with no education credential increases by 0.0336. All education variables are statistically significant at the 10 percent level, with the exception of I_aed, I_pnd, I_hsnx, II_opc, and II_csd. The data show that the probability of attrition increases for women by 0.0971 and is significant at the 1 percent level.
Table 35. Binary LPM, Education Codes, Regression Results for Men and Women
Variable Coefficient Robust Std. Error p-Value
I -0.0369 0.0027 0.0000 Source: Derived from data provided by Defense Manpower Data Center (DMDC), 2009.
Base individual is high school diploma graduate white male with category IIIA AFQT who entered DEP in 1999.
When analyzing men and women together, Figures 39 and 40 show that for men and women, the probability of attrition increases with age, with the greatest increase between the less-than-18 group and 18-to-20 groups. After that, the difference in the probability of attrition with age declines as the groups increase in age. The data also show that the probability of attrition decreases as AFQT category increases. The decrease in the probability of attrition with an increase in AFQT category is the smallest between category IIIB and category IIIA. From AFQT category IIIA to category II, the decrease in the probability of attrition with increase in AFQT category is the greatest. As with the other probability curves calculated from the LPM, there is no variation between men and women in the slopes of different age groups, as shown in Figure 40.
II I
IIIB IIIA
0.15 0.17 0.19 0.21 0.23 0.25 0.27
AFQT Category
Probability of Attrition >30
27_29 24_26 21_23 18_20
<18
Source: Derived from data provided by Defense Manpower Data Center (DMDC), 2009.
Figure 39. Men, all Gender, Tier I LPM of Attrition by Age and AFQT Category
II I
IIIB IIIA
0.15 0.17 0.19 0.21 0.23 0.25 0.27
AFQT Category
Probability of Attrition >30
27_29 24_26 21_23 18_20
<18
Source: Derived from data provided by Defense Manpower Data Center (DMDC), 2009.
Figure 40. Women, all Gender, Tier I LPM of Attrition by Age and AFQT Category