Download presentation
Presentation is loading. Please wait.
Published byNelson Holmes Modified over 6 years ago
1
SOCIAL ANIMAL HOUSE: THE ECONOMIC AND ACADEMIC CONSEQUENCES OF FRATERNITY MEMBERSHIP
Jack Mara, Lewis Davis and Stephen Schmidt ASSA, Chicago IL, January
2
Fraternity membership is stereotypically linked to indulgent, non-academic behaviors
Associated with lower college grades (Grubb 2006, Routon and Walker 2014) Associated with increased drinking and increased binge drinking (Alva 1998, DeSimone and 2009, and more) But it may also form social capital Increases the probability of obtaining a job after graduation (Marmaros and Sacredote (2002), Routon and Walker 2014) Can signal student qualities to employers (Popov and Bernhardt 2012)
3
Question: What are the long-run consequences of fraternity membership on earnings?
Survey of alumni aged from a single Northeastern college, indicating earnings and Greek status Use changes in residential policies and social environment to measure the causal effect of fraternity membership on grades and income: Alternative housing systems (themes, Minervas) - Coeducation - Changes in the number of fraternities and sororities on campus These change the incentive to join a fraternity, providing a (plausibly) exogenous shock that we can use to identify outcomes
4
Answer: Fraternity membership substantially increases post-college income despite its negative effects on grades Social capital formation in fraternities is important enough to offset the harmful effects of fraternity membership on human capital Almost all of this social capital formation occurs through channels other than alcohol consumption. Policies towards fraternities should recognize their role in the formation of social capital as part of the broader education taking place
5
Outline Survey data used in the analysis Identification strategy
OLS and 2SLS estimation results Decomposition into alcohol, human capital, and social capital channels Interpretation and conclusion
6
The data Survey of alumni of a single college in the Northeastern United States College has a long and strong history of fraternities 3762 responses, 25.8% response rate Analysis limited to respondents who are male, age 23-65, employed full time, and have complete responses for variables used in the analysis Total of 1667 observations in our sample. Most drops are female or are not working full time.
7
Observed variables on students
Outcomes: college GPA (0-4 scale), log(earnings) at time of survey Treatment: Greek membership Demographics: age (quadratic), ethnicity Controls: SAT, self-rated college appearance, unemployment rate at time of graduation Alcohol use: Nights drinking per week in college, drinking intensity Alcohol use variables are potentially endogenous due to probable relationship to Greek membership
8
Identification strategy
Students decide whether to join fraternities or not Decisions about whether to join may be correlated with unobserved aptitude for academics or earning income post-college, creating an endogeneity bias in OLS estimates of the relationship between Greek membership and grades/post-college earnings Valid instruments are ones that are uncorrelated with student aptitude (exogenous), relevant to the student’s decision about Greek membership, and have no other influence on grades or post-college income Variables related to students are unlikely to be uncorrelated with student aptitude We derive instruments from college policies that affect housing policies but plausibly have no other effects on outcomes
9
Instrumental variables from college policies
Themes – indicator variable for the presence of theme houses on campus (1 in and later) Minervas – number of Minerva houses on campus (7 in 2004 and later, 0 before) Coeducation – indicator variable for women in classes (1 in 1971 and later) Number of housed fraternities and sororities on campus
10
Instrument sets We use four sets of instruments
Set A: Minervas and Themes. Most plausibly exogenous, but also not very influential on Greek housing decisions, so perhaps not very powerful Set B: Add Coed. Reduced the role of fraternities in connecting male students to compatible females. Set C: Number of fraternities and sororities, plus Minervas. Number of fraternities directly relevant to selecting one. Sororities are a social complement to fraternities. Set D: All instruments together, to test overidentification restrictions.
11
Equations to estimate GPAit = b0 + b1*Greekmemi + bj*Xji + ei (1) log(Incomeit) = g0 + g1*Greekmemi + gj*Xji + ei (2) With the first stage regression Greekmemit = d0 + dj*Xji + lk*Zik + ei (3) X = demographic and control variables Z = instruments (four sets)
12
Results – First Stage – Instrument Set A
(1) (2) (3) (4) VARIABLES greekmem Minervas *** *** ** ** (-2.708) (-2.647) (-2.560) (-2.562) Themes 0.295*** 0.305*** 0.287*** 0.300*** (6.064) (6.571) (6.037) (6.517) nightsdrinking 0.0816*** 0.0738*** (9.756) (8.781) binge 0.0856*** 0.0643** (2.747) (2.068) collegegrade2 -0.198*** -0.130*** (-8.567) (-5.579) T-statistics in parentheses Intercept and control variables not shown Asterisks indicate statistical significance: *** p<0.01, ** p<0.05, * p<0.1. Bivariate correlations of instruments: Minervas, , Themes, Based on column 1: Partial R2 for instruments: F-statistic for exclusions: (10% critical value for 2SLS size – 19.93)
13
Results - First Stage – Other Instruments
Instrument Set B Instrument Set C Instrument Set D VARIABLES greekmem Themes 0.228*** 0.182*** (4.407) (3.169) Coed -0.233*** (-3.689) (-1.075) Minervas ** * * (-2.081) (-1.844) (-1.891) Fraternities 0.0356*** (3.282) (0.623) Sororities 0.0797*** 0.0688*** (3.578) (3.180) Partial R2 F-test for exclusion 23.67 28.07 18.21 Critical value: 5% size 13.91 18.37 Critical value: 10% bias 22.30 26.87 T-statistics in parentheses Intercept and control variables not shown Asterisks indicate statistical significance: *** p<0.01, ** p<0.05, * p<0.1. Bivariate correlations of instruments: Minervas, , Themes, 0.085, Coed, Fraternities, 0.198, Sororities, 0.072
14
Results – Grades – Instrument Set A
(1) (2) (3) GPA VARIABLES OLS IV greekmem -0.213*** -0.183 -0.167 (-8.536) (-1.359) (-1.290) binge -0.150*** (-4.303) nightsdrinking *** (-3.493) Overid p-value 0.583 0.692 First stage F-stat 28.49 31.99 Column 1: Robust t-statistics in parentheses Columns 2-3: z-statistics in parentheses Asterisks indicate statistical significance: *** p<0.01, ** p<0.05, * p<0.1. Intercept and control variables not shown Instruments: Minervas, Themes
15
Results – Grades – Instrument Set B
(1) (2) (3) GPA VARIABLES OLS IV greekmem -0.213*** -0.259** -0.236** (-8.536) (-2.139) (-2.003) binge -0.143*** (-4.138) nightsdrinking *** (-3.245) Overid p-value 0.374 0.415 First stage F-stat 23.67 26.12 Column 1: Robust t-statistics in parentheses Columns 2-3: z-statistics in parentheses Asterisks indicate statistical significance: *** p<0.01, ** p<0.05, * p<0.1. Intercept and control variables not shown Instruments: Minervas, Themes, Coed
16
Results – Grades – Instrument Set C
(1) (2) (3) GPA VARIABLES OLS IV greekmem -0.213*** -0.423*** -0.385*** (-8.536) (-3.711) (-3.333) binge -0.127*** (-3.631) nightsdrinking ** (-2.322) Overid p-value 0.327 0.295 First stage F-stat 28.07 28.58 Column 1: Robust t-statistics in parentheses Columns 2-3: z-statistics in parentheses Asterisks indicate statistical significance: *** p<0.01, ** p<0.05, * p<0.1. Intercept and control variables not shown Instruments: Minervas, Fraternities, Sororities
17
Results – Earnings – Instrument Set A
(1) (2) (4) (5) (6) VARIABLES OLS IV greekmem 0.183*** 0.397** 0.415** 0.450** 0.468*** (5.826) (2.261) (2.386) (2.511) (2.685) binge (-0.999) (0.0134) nightsdrinking 0.0195 0.0342** (1.087) (2.031) collegegrade2 0.286*** 0.316*** (6.003) (7.963) Overid p-value 0.863 0.843 0.957 0.915 First stage F-stat 28.49 31.99 27.59 31.15 b=0.183 = 20.1% increase in earnings, b=0.397 = 48.7% increase in earnings Column 1: Robust t-statistics in parentheses Columns 2-6: z-statistics in parentheses Asterisks indicate statistical significance: *** p<0.01, ** p<0.05, * p<0.1. Intercept and control variables not shown Instruments: Minervas, Themes
18
Results – Earnings – Instrument Set B
(1) (2) (4) (5) (6) VARIABLES OLS IV greekmem 0.183*** 0.309** 0.328** 0.384** 0.403** (5.826) (1.976) (2.098) (2.369) (2.549) binge (-0.828) (0.130) nightsdrinking 0.0263 0.0388** (1.567) (2.429) collegegrade2 0.273*** 0.307*** (6.069) (8.043) Overid p-value 0.506 0.473 0.668 0.644 First stage F-stat 23.67 26.12 22.22 24.96 b=0.309 = 36.2% increase in earnings Column 1: Robust t-statistics in parentheses Columns 2-6: z-statistics in parentheses Asterisks indicate statistical significance: *** p<0.01, ** p<0.05, * p<0.1. Intercept and control variables not shown Instruments: Minervas, Themes, Coed
19
Results – Earnings – Instrument Set C
(1) (2) (4) (5) (6) VARIABLES OLS IV greekmem 0.183*** 0.316** 0.321** 0.435*** 0.441*** (5.826) (2.186) (2.144) (2.796) (2.840) binge (-0.817) (0.0627) nightsdrinking 0.0269 0.0361** (1.637) (2.280) collegegrade2 0.283*** 0.313*** (6.425) (8.198) Overid p-value 0.978 0.976 0.987 0.980 First stage F-stat 28.07 28.58 24.58 26.26 b=0.316 = 37.2% increase in earnings Column 1: Robust t-statistics in parentheses Columns 2-6: z-statistics in parentheses Asterisks indicate statistical significance: *** p<0.01, ** p<0.05, * p<0.1. Intercept and control variables not shown Instruments: Minervas, Fraternities, Sororities
20
Instrument set D – all five instruments
Because of concerns about bias from weak instruments, results not shown Broadly similar to what is found with the first three instrument sets: - Fraternities lower grades by Fraternities raise post-college incomes by b=0.324 = 38.3% P-value for overidentification restrictions is 0.841, suggesting that all instruments are plausibly exogenous Robustness of results to choice of instrument sets makes us more confident in findings than we would with in any single set of results
21
Interpretations Fraternity membership offers a substantial increase in post-college earnings, despite the damage it does to grades. This can help explain why students choose to do it – it’s rational! The effect is very similar for students with different drinking behaviors, suggesting that learning to hold booze is not the cause of the gain Fraternities may be forming social capital in several ways: 1) Developing soft skills (people skills) with applications in teamwork, recruiting, organization, persuasion, etc. 2) Building a social network that allows members to obtain opportunities and connect people to positions where they can be effective 3) Gaining insider status in networks that favor their own members (not socially beneficial)
22
Interpretations continued
Fraternities attract members, on the margin, with less potential for post-college earnings (because bOLS < b2SLS, and the LATE nature of 2SLS estimates) This may occur because students with particular aptitude for forming social capital are self-selecting into fraternities Some evidence that fraternities also, on the margin, attract members with greater academic potential (because bOLS > b2SLS) This is probably because marginal members think carefully about whether they can tolerate academic distractions. Members committed to fraternity life regardless of campus policies may be not have the same academic strengths as the marginal members whose behavior we capture
23
Policy implications College housing and social policies are within the control of administrators who may wish to alter the effects of fraternity systems on their campuses The effects of policy changes on marginal members are the right ones to think about for many decisions – and those are the effects our results are measuring College administrations making policies towards fraternities should recognize effects of those policies on social capital formation and post- college graduates of incomes (future donations?) However, social capital formation doesn’t appear to depend on alcohol use, so reducing alcohol use within fraternities may be the most efficient policy to pursue
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.