# Results

The results from the binary logit models are presented in Table 4.3. Starting with Friends Influence, the model implies that boys are 10.5 percentage points (ppts), or 120.0%, more likely to be influenced by their friends in making their decision about which HEI to attend, while students in a DEIS school are 7.8 ppts (50.0%) less likely to be influenced by their peers, all else equal. We also find that students living farTable 4.3 Binary logit models of friends, sibling and parent influence on choice of HEI

 Variable Friends Influence Sibling Influence Parent Influence Male 0.105 (2 99)*** 0.034 (0.57) -0.038 (1.29) Honours Subjects 0.037 (1.07) -0.235 (2.41)** 0.031 (0.63) Honours Subjects Sq -0.004 (1.21) 0.022 (2.59)*** -0.003 (0.75) Mother Higher Education -0.023 (1.02) -0.016 (0.34) 0.088 (2.85)*** Grant Eligibility = Yes -0.016 (0.58) -0.037 (0.52) -0.055 (1.94)* Grant Eligibility = Don't Know -0.042 (1.43) 0.018 (0.34) -0.044 (1.31) DEIS School -0.078* (1.93) -0.100 (1.30) 0.099 (2.92)*** Single Sex School 0.018 (0.76) 0.044 (0.82) -0.018 (0.49) Distance Chosen HEI -0.000 (0.63) -0.002 (4.78)*** 0.000 (0.68) Distance Nearest HEI 0.001 (1.98)** 0.005 (4.09)*** -0.001 (1.37) Region Y Y Y Wald x2 statistic 218.56 106.88 36.30 Prob > x2 0.000 0.000 0.016 Observations 855 421 931

Notes: The table presents estimated average marginal effects from three separate binary logit models of Friends Influence, Sibling Influence and Parent Influence. Absolute values of t-statistics are presented in parentheses. Y denotes region indicator variables included in the model and were found to be statistically significant. *** denotes statistically significant at 1%, ** denotes statistically significant at 5%, and * denotes statistically significant at 10%. Standard errors are clustered at the school level Source: Analysis of data from Walsh et al. (2017)

ther away from a HEI are more likely to take their friends’ choices into account. In terms of Sibling Influence, the evidence suggests a non-linear U-shaped association with number of honours subjects, while distance to chosen HEI is negatively associated with this influence and distance to nearest HEI is positively associated with it. Finally, we find Parent

Influence to be 8.8 ppts (13.0%) higher for students whose mother has completed higher education, 5.5 ppts (7.1%) lower for those eligible for a student grant and 9.9 ppts (14.0%) higher for students in DEIS schools, all else equal.

Overall these results suggest that boys are more likely than girls to be influenced by their peers, though there are no differences across gender for sibling and parent influences. Mother’s education is not associated with the extent to which students are influenced by their peers’ or siblings’ choices but strongly and positively correlated with parental influence. Being eligible for a student grant does not seem to affect the likelihood of peer or sibling influence but does matter for parent influence. In particular, grant eligibility is correlated with lower levels of parental guidance. Students in DEIS schools differ from students in non-DEIS schools in terms of both peer and parent influence, with the former less likely to be influenced by their friends and more likely to be influenced by their parents. Finally, geographic proximity to HEIs also appears to matter for peer and sibling influence, a finding that adds to the analysis in Chap. 3.

Turning to the multinomial logit model of planning to live at home, results are presented in Table 4.4. They suggest that students who are eligible for a grant are 7.7 ppts more likely to be unsure about this decision, while those who stated they would be influenced by their friends are 6.5 ppts less likely to live at home. Students influenced by their parents are 9.8 ppts less likely to live away from home, 4.1 ppts more likely to live at home and 5.7 ppts more likely to be unsure. School DEIS status is also a significant predictor, with students from a DEIS school 9.5 ppts less likely to live at home and 13.6 ppts more likely to be unsure. Students from single-sex schools are 10.9 ppts less likely to intend to live at home, while greater distances to a student’s chosen HEI are associated with an increased likelihood of living away from home and a decreased likelihood of living at home, as would be expected. Similar effects are found for students living greater distances from their nearest HEI, again as expected. Overall the model suggests that friend and parent influence is associated with the decision to live at home or not, while grant eligibility too is important, all else equal. We also find

Table 4.4 Multinomial logit model of planning to live at home while at higher education

 Variable Live Away from Home Live at Home Don't Know Male 0.020 (0.63) 0.013 (0.36) -0.033 (0.94) Honours Subjects 0.062 (1.37) -0.023 (0.70) -0.039 (1.03) Honours Subjects Sq -0.005 (1.03) 0.002 (0.61) 0.003 (0.72) Mother Higher Education 0.003 (0.14) -0.019 (0.55) 0.016 (0.43) Grant Eligibility = Yes -0.056 (1.63) -0.022 (0.79) 0.077 (2.34)** Grant Eligibility = Don't Know -0.050 (1.42) -0.005 (0.12) 0.055 (1.47) Friends Influence 0.007 (0.20) -0.065 (2.00)** 0.058 (1.91)* Parent Influence -0.098 (3.22)*** 0.041 (1.81)* 0.057 (1.69)* DEIS School -0.042 (0.89) -0.095 (2.71)*** 0.136 (3.02)*** Single Sex School 0.055 (1.28) -0.109 (2.62)*** 0.053 (1.37) Distance Chosen HEI 0.002 (11.19)*** -0.002 (10.45)*** 0.000 (1.05) Distance Nearest HEI 0.007 (5.98)*** -0.006 (5.34)*** -0.001 (1.15) Region Y Y Y Pseudo R2 0.367 Observations 842

Notes: The table presents estimated average marginal effects from a multinomial logit model of whether respondents plan to live at home while attending higher education. Absolute values of t-statistics are presented in parentheses. Y denotes region indicator variables included in the model and were found to be statistically significant. *** denotes statistically significant at 1%, ** denotes statistically significant at 5%, and * denotes statistically significant at 10%. Standard errors are clustered at the school level Source: Analysis of data from Walsh et al. (2017)

interesting differences across school characteristics, with students from DEIS schools and single-sex schools less likely to plan to stay at home while in higher education. As expected, proximity to HEIs plays a big role too, which is consistent with the analysis presented in Chap. 3.