Index

Abstract

This paper considers ethnicity variables and financial variables to study what impact they have on alumni salaries six and ten years after graduation, for colleges and universities in the United States. We find that certain ethnic groups earn less salary both six years out and ten years out, while other ethnic groups earn more salary both six and ten years out. The same is true for selected financial variables, such as median household income. We perform our study primarily using stepwise-regression analyses. Since many colleges and universities use future salary of alumni as a marketing tool, this study cautions that there is a danger to the diversity of a college campus, and a corresponding harm to students on such a campus, if the college or university places too much emphasis on future alumni salary and makes admissions decisions in a way that emphasizes that criterion.

Keywords: Salary of college/university alumni, Campus diversity, Stepwise-regression analysis, Ethnicity variables, Financial variables.

Received: 27 March 2018 / Revised: 16 April 2018 / Accepted: 20 April 2018/ Published: 23 April 2018

Contribution/ Originality

This study contributes to the existing literature by pointing out a potentially unintended consequence of an economically-motivated decision made by colleges and universities that is, arguably, bad for society. The danger is that of reducing the diversity of the undergraduate student-body, and the corresponding harm done to said student body.


1. INTRODUCTION

In this paper, we examine the impact of university marketing – specifically the use of average alumni salary metrics in university advertising - and the effect that it has on the university community and the students who create that community. It should be noted that we do not focus on the effectiveness of the marketing for recruiting purposes, but, rather, on the potential impact of diversity on campus; one might argue that we are examining at best the impact of “unintended consequences.”

Examples of universities that include salary data of alumni in their web marketing include, surely among others:

We are particularly interested in this topic from the marketing perspective, because society often focuses on marketing crises that are considered “scandals” within organizations, resulting in bad publicity. Occasionally, what is neglected is that the more important story about these marketing issues is the negative impact on the organization and society, overall, due to the result of small lapses, or oversights, that are embedded in our systems. These oversights can often be caused by how an organization chooses to evaluate itself, both as an organization and as a set of individuals within the organization, represented by the key performance indicators that the organization uses both to monitor its successes and to share with its customers and prospects.

While universities often advertise the amount of money that former students can earn once they start their careers, one might wonder how accurate these numbers are (after all, not all alumni keep in touch with a school within a few months after graduation, never mind after 6 or 10 years later [the time periods used in this paper, based on collected data]; one can argue that there is easily the possibility of bias, especially if “higher earning” alumni are pleased to report (brag??) about their salaries, while “lower earning” alumni may not wish to report their earnings, due to embarrassment, or the feeling that they are “letting the school down.” Not reporting salaries is a very common issue on surveys as a routine part of the demographic section. It is well-known that salary-data response is “missing” more than the response to any other routine demographic question.

One might also wonder about the impact of the desire to continually “improve” (i.e., increase) future salary figures - on individual students admitted to the university, and the overall university community. Universities are not the only ones sharing these metrics; there is also a variety of lists available online that are ranking colleges with the highest-paid alumni (Dale and Krueger, 2011 ). The characteristics that make up high-future-earning university-populations are the key focus of our analysis. We are seeking to determine the predictors of salary based on the characteristics of undergraduate students on campuses nationwide in the United States.

Essentially, we are examining what our universities would look like if striving for increased alumni salaries became priority number 1. It is worrisome to consider the consequences of a university’s attributing too high a priority to high future salaries; assuredly, as will be seen later in this paper, the demographic mix of the student body at such a university would not be what most people truly want.

2. LITERATURE REVIEW

3. METHODOLOGY

The data used for the following study is from the College Scorecard published by the U.S. Department of Education in September 2017. These data are available at https://collegescorecard.ed.gov/data/ (College Scorecard Data from US Department of Education, 2017). The College Scorecard data are collected to give students who are making choices about college access to information to help them in their decision making. The Department of Education collects the data through reporting from federal institutions, financial-aid reports, and tax information. The College Scorecard website states that “many data elements are drawn directly from, or derived from, data reported to the IPEDS [Integrated Postsecondary Education Data System].” The College Scorecard provides data from over 7,500 U.S. undergraduate degree-granting colleges and universities and the sample size in this study is n = 7,593. The data were last updated on September 28th, 2017.

The two dependent variables in this study are median annual income six years after college, and median annual income ten years after college, for former undergraduate students at each university. The independent variables in our study can be grouped into two categories: 1) ethnicity, and 2) family income when admitted to the university.

The 12 independent variables used to indicate ethnicity are:

We acknowledge that there is a high correlation/multi-colinearity among certain of these variables.
The 3 independent variables used to indicate family income are:

3.1. Hypotheses

Based on the previous research included in the above literature review, we hypothesized, for a stepwise-regression model, the following:

These hypotheses come from the unfortunate reality that Black and Hispanic employees earn less, on average, than White employees (Patten, 2016). Additionally, as mentioned in the aforementioned literature review, students whose families have lower household incomes when they attend college tend to earn less as workers in the future (Morrison, 2016).

4. ANALYSIS

4.1. Initial Stepwise Regressions

We will run a series of stepwise regressions. First, we will run four stepwise regressions, two for each dependent variable (salary six years out, salary ten years out); the two per dependent variable will be one for each set of independent variables. In other words, the four initial stepwise regressions will be:

Table 1 shows the final step of the stepwise-regression analysis for (Y, X’s) combination 1.

Table-1. Salary ten years out, 12 ethnicity variables

9
(Constant)
27998.69
294.843
94.961
0
UGDS_NRA
83351.79
3987.917
0.276
20.901
0
UGDS_WHITENH
6913.382
465.113
0.194
14.864
0
UGDS_ASIAN
34718.72
2568.784
0.185
13.516
0
UGDS_BLACKNH
-5875
982.144
-0.078
-5.982
0
UGDS_AIAN
-12986.3
2158.201
-0.078
-6.017
0
UGDS_HISP
-112.213
21.164
-0.069
-5.302
0
UGDS_2MOR
21112.06
4725.984
0.058
4.467
0
UGDS_UNKN
7089.218
1732.858
0.053
4.091
0
UGDS_NHPI
-18737.1
5221.629
-0.046
-3.588
0

a. Dependent Variable: MD_EARN_WNE_P10

It can be seen in Table 1 that 9 of the 12 variables were contained in the final stepwise-regression model. The multiple R2 for the model is .200 (adjusted R2 = .199). All of the variables have a p-value (to 3 digits) of .000 (last column in the table). Results indicate that, holding other variables in the model constant, the larger the percentage non-resident alien students, White-non-Hispanic students, Asian students, students who indicated 2 or more races, and students whose race was not indicated (i.e., unknown), the larger the median salary of students ten years out. The larger the percentage of Black-non-Hispanic students, American Indian/Alaskan Native students, Hispanic students, and Native Hawaiian/ Pacific Islander students, the smaller the median salary of student ten years out. It might be surprising to the reader that the variables, percent of White students and percent of Black students, did not enter the final model; undoubtedly, that is because the variables are each sufficiently co-linear with other variables in the final model, as to not provide incremental value to predicting Y.

Table 2 shows the final step of the stepwise-regression analysis for (Y, X’s) combination 2.

Table-2. Salary six years out, 12 ethnicity variables

10
(Constant)
23577.54
237.56
99.249
0
UGDS_NRA
51158.98
3153.229
0.215
16.224
0
UGDS_WHITENH
6163.185
379.096
0.214
16.258
0
UGDS_ASIAN
27715.7
2066.547
0.183
13.412
0
UGDS_BLACKNH
-5908.5
798.909
-0.097
-7.396
0
UGDS_AIAN
-10896.2
1809.646
-0.078
-6.021
0
UGDS_HISP
-92.567
17.499
-0.069
-5.29
0
UGDS_NHPI
-15452.7
4333.943
-0.047
-3.566
0
UGDS_UNKN
4618.039
1419.146
0.042
3.254
0.001
UGDS_2MOR
9167.641
3602.351
0.033
2.545
0.011
UGDS_WHITE
28.52
13.592
0.027
2.098
0.036

a. Dependent Variable: MD_EARN_WNE_P6

It can be seen in Table 2 that 10 of the 12 variables were contained in the final stepwise-regression model. The multiple R2 for the model is .165 (adjusted R2 = .163). Seven of the variables have a p-value (to 3 digits) of .000, and two others have p-value < .011. Results indicate that, holding other variables in the model constant, the larger the percentage non-resident alien students, White-non-Hispanic students, Asian students, students whose race was not indicated (i.e., unknown), students who indicated 2 or more races, and percentage of White student, the larger the median salary of students ten years out. The larger the percentage of Black- non-Hispanic students, American Indian/Alaskan Native students, Hispanic students, and Native Hawaiian/ Pacific Islander students, the smaller the median salary of students six years out. The results for the two dependent variables are not identical but very similar. 

Table 3 shows the final step of the stepwise-regression analysis for (Y, X’s) combination 3.

Table-3. Salary ten years out, 3 family-income variables

2
(Constant)
3349.97
747.746
4.48
0
MEDIAN_HH_INC
0.498
0.013
0.509
39.687
0
PCTFLOAN
1789.422
576.887
0.04
3.102
0.002

a. Dependent Variable: MD_EARN_WNE_P10

It can be seen in Table 3 that 2 of the 3 variables were contained in the final stepwise-regression model. The multiple R2 for the model is .268 (adjusted R2 = .268). One of the variables has a p-value (to 3 digits) of .000, and the other has a p-value = .002. Results indicate that, holding other variables in the model constant, the larger the median household income of the students, and the larger the percentage of students receiving a federal loan, the higher the median salary ten years out.

Table 4 shows the final step of the stepwise-regression analysis for (Y, X’s) combination 4.

Table-4. Salary six years out, 3 family-income variables

2
(Constant)
4994.462
607.549
8.221
0
MEDIAN_HH_INC
0.364
0.01
0.465
35.619
0
PCTFLOAN
2335.093
472.814
0.064
4.939
0

Dependent Variable: MD_EARN_WNE_P6

It can be seen in Table 4 that the same two variables as in Table 3 were contained in the final stepwise-regression model. The multiple R2 for the model is .232 (adjusted R2 = .232). Both variables have a p-value (to 3 digits) of .000. The same as indicated in Table 3, results indicate that, holding other variables in the model constant, the larger the median household income of the students, and the larger the percentage of students receiving a federal loan, the higher the median salary six years out. For the dependent variable, salary six years out, the income coefficient is somewhat smaller than that for median salary ten years out, but the coefficient for percentage of students receiving a federal loan is somewhat higher than that for salary ten years out.

It might also be noted that the models using the two-family income variables provided higher R2 values (.268, .232) than those provided by the models based on the 9 or 10 ethnicity variables (.200, .168).

4.2. Final (Stage 2) Stepwise Regressions

In each of the initial stepwise regression analyses, the variables in the final model are not highly correlated with one another; the nature of the stepwise-regression process guarantees this. Redundant variables are “not allowed” to enter the model, and if redundancy occurs at a later stage, variables are accordingly deleted. However, there may be “overlap,” or correlation between some of the variables in the ethnicity-based models and the family-income-based models. That is why we now perform our “final stage” stepwise-regression analyses that, for each dependent variable, have as eligible variables, the “winners” from each of the two initial analyses. So, for the salary ten years out dependent variable, the set of eligible independent variables will be the 9 variables shown in Table 1 and the 2 variables shown in Table 3. For the salary six years out dependent variable, the set of eligible independent variables will be the 10 variables shown in Table 2 and the 2 variables shown in Table 4.

Table 5 shows the final stepwise-regression analysis for salary ten years out. Table 6 shows the final stepwise-regression analysis for salary six-years out.

Table-5.  Salary ten years out – FINAL MODEL

8
(Constant)
6198.747
739.195
8.386
0
MEDIAN_HH_INC
0.382
0.014
0.39
28.138
0
UGDS_NRA
61931.47
3755.118
0.209
16.493
0
UGDS_WHITENH
4507.698
438.354
0.127
10.283
0
PCTFLOAN
4301.95
575.639
0.096
7.473
0
UGDS_BLACKNH
-5551.84
920.37
-0.074
-6.032
0
UGDS_ASIAN
16856.87
2516.639
0.091
6.698
0
UGDS_NHPI
-21316.4
5539.678
-0.047
-3.848
0
UGDS_HISP
-56.227
19.663
-0.036
-2.86
0.004

a. Dependent Variable: MD_EARN_WNE_P10

Table-6. Salary six years out – FINAL MODEL

9
(Constant)
4732.441
603.497
7.842
0
MEDIAN_HH_INC
0.247
0.011
0.316
21.827
0
UGDS_WHITE
7203.365
521.232
0.195
13.82
0
UGDS_NRA
41601.75
2986.564
0.178
13.93
0
UGDS_ASIAN
22967.39
2128.544
0.154
10.79
0
PCTFLOAN
4087.124
467.236
0.113
8.747
0
UGDS_WHITENH
2191.89
378.787
0.077
5.787
0
UGDS_HISP
-59.025
16.048
-0.046
-3.678
0
UGDS_UNKN
5101.498
1603.794
0.04
3.181
0.001
UGDS_NHPI
-12788.1
4554.394
-0.034
-2.808
0.005

a. Dependent Variable: MD_EARN_WNE_P6

For Table 5, the R2 = .339 (adjusted R2 = .338). For Table 6, the R2 = .319 (adjusted R2 = .317). Based on the coefficients provided by the analyses above, we can identify the characteristics of a university community which are likely to have an impact on median salary of alumni after ten and six years. In Table 5, (salary ten years out) 6 of the 9 ethnicity variables and both of the income variables made the final model. In Table 6, (salary six years out) 7 of the 10 ethnicity variables and both of the income variables made the final model. The ethnicity variables that entered the initial stepwise models but did enter the final model were highly correlated with the two income variables, and not in the final model due to redundancy. Figure 1A and 1B sum up the variables that are positively and negatively associated with median income, assuming that all other variables are held constant.

Figure-1A.  Characteristics of universities based on median salary level after six years

Figure-1B. Characteristics of universities based on median salary level after ten years

Some of the findings agree with our predicted hypotheses while others do not. Figure 2 displays the outcome of these hypothesis tests.

Figure-2. Hypotheses assessment

This study shows that in theory, universities who wish to increase their median alumni salary for marketing or other purposes could have an incentive to limit the percentage of Hispanic, American Indian/Alaska Native and Black non-Hispanic students who attend.

Universities would also have an incentive to increase the percentage of students who are receiving a federal student loan, who are non-resident aliens, Asian, of 2 or more races, race unknown, White, and are White non-Hispanic.

Therefore, based on these results, universities should closely monitor how they are using alumni salary in their marketing materials and the types of processes, traditions, and norms it is creating in the university culture. For example, universities should consider how employees at the university are rewarded or not rewarded for increasing this metric. Is it encouraging team members (perhaps, the admissions committee?) to limit diversity? If not careful, colleges could end up limiting the diversity of the school in pursuit of increasing the median alumni salary. Limiting diversity can impact student experiences in a variety of ways and has a negative social impact as well (Bowman, 2014).

5. LIMITATIONS AND FUTURE RESEARCH

One limitation of this study is that the data does not capture which universities use the alumni salary metric in their marketing materials. Therefore, we do not know for sure if schools which use this information in their marketing have an increased likelihood of building a university community with higher earning potential and, correspondingly, less diversity.

Another limitation is that we don’t know the impact of each of the variables on the graduation rate of students. Are certain variables associated with lower salary levels because students from that ethnic or financial background are less likely to graduate? Further research is possible on both topics.

Funding: This study received no specific financial support.   
Competing Interests: The authors declare that they have no competing interests. 
Contributors/Acknowledgement: Both authors contributed equally to the conception and design of the study.

REFERENCES

Babson College, 2017. Career outcomes. Undergraduate Career Outcomes, Babson College. Retrieved form www.babson.edu/student-life/career-development/undergraduate/Pages/career-outcomes.aspx .

Bowman, N.A., 2014. Conceptualizing openness to diversity and challenge: Its relation to college experiences. Innovative Higher Education, 39(4): 277-291. View at Google Scholar | View at Publisher

Bucknell University, 2017. Annual postgraduate report of activity. Retrieved from www.bucknell.edu/PostGradReport .

Caltech Undergraduate Admissions, 2017. Life after Caltech. Retrieved from www.admissions.caltech.edu/content/life-after-caltech .

College Scorecard Data from US Department of Education, 2017.

Dale, S. and A. Krueger, 2011. Estimating the return to college selectivity over the career using administrative earnings data. National Bureau of Economic Research. View at Publisher

FastWeb, 2011. Profile of pell grant recipients quick reference guide. Retrieved from www.finaid.org/educators/ProfileofPellGrantRecipients.pdf .

Fastweb Team, 2017. Income and financial aid eligibility. Retrieved from www.fastweb.com/financial-aid/articles/income-and-financial-aid-eligibility [Accessed 26 June].

Harvey Mudd College, 2017. Alumni outcomes. Retrieved from www.hmc.edu/admission/discover/alumni-outcomes/ .

Morrison, N., 2016. You'll never guess the biggest factor in graduate salaries. Forbes, Forbes Magazine. Retrieved from www.forbes.com/sites/nickmorrison/2016/03/31/youll-never-guess-the-biggest-factor-in-graduate-salaries/#411cdee25e41 [Accessed 31 Mar. 2016].

Patten, E., 2016. Racial, gender wage gaps persist in U.S. despite some progress. Pew Research Center. Retrieved from www.pewresearch.org/fact-tank/2016/07/01/racial-gender-wage-gaps-persist-in-u-s-despite-some-progress/ [Accessed 1 July 2016].