## Does the spiritual benefit of SDA higher education Justify the Financial investment?

Seventh-day Adventists are known for many things: blue zones, healthy lifestyles, state-of-the-art healthcare systems, and more types of meat substitutes than the typical person could imagine. The Adventist education system is another prominent feature of the denomination. Adventist scholars have stated that the American Adventist education system is the largest Protestant education system in the country and the second largest Christian education system (behind the Catholic church). ∫

Another differentiating characteristic of Adventists is their strong preference for only being educated in Adventist institutions. Having been raised in the church, I feel confident in the validity of this observation. Adventists place a higher premium on denominational higher education than other Christian religions. While many Adventists attend non-Adventist universities, a large percentage that pursue post-secondary education opt to go straight from Adventist academy (K-12) to Adventist college. My father actually went from Adventist academy to Adventist college to Adventist medical school.

Adventists derive more utility (in a microeconomic sense of the word) from attending Adventist institutions than non-Adventist ones. There are many reasons for this, including worship services, emphasis on Adventist values, pedagogical curricula that conform to Adventist theology, gateways to careers with Adventist companies, and the familiarity of Adventist institutions. While these intangible benefits are undeniable, they’re difficult to capture in an econometric model.

In this post, I seek to answer the following question with data from the 2021 U.S. News & World Report rankings:

**“Is an Adventist college education a good financial investment?”**

Obviously, the econometric models I employ to help answer this question do not account for the spiritual and religious benefits of an Adventist education.

## 2021 u.S. News & World Report Rankings

There are a variety of college ranking systems that parents and high-school students use to help make college selection decisions, but the most prominent one is administered by U.S. News & World Report (*USNWR*). All ranking systems have significant flaws; in my opinion, we place far more stock in them than we should. For example, twenty percent of a university’s *USNWR* ranking is based on “Expert Opinion,” which refers to how administrators at peer universities perceive a college’s quality of education. Some critics say expert opinion is weighted too heavily in the ranking. Nevertheless, the reality is that college ranking systems (*USNWR, The Wall Street Journal, Forbes*, etc.) are closely watched by college administrators and consumers alike and do provide useful criteria to evaluate educational quality.

Before we dive into technical analysis, I’ll briefly explain the *USNWR* rankings methodology and show how each Adventist university and college compares to their peers. For a more granular explanation of the *USNWR* rankings calculations, click **here. **

There are ten *USNWR* college classifications: ∝

**National Universities***“Offer a full range of undergraduate majors, plus master’s and doctoral programs, and emphasize faculty research or award professional practice doctorates.”*

**National Liberal Arts Colleges***“Focus almost exclusively on undergraduate education and award at least 50% of their degrees in the arts and sciences.”*

**Regional Universities – North Region***“Offer a broad scope of undergraduate degrees and some master’s degree programs but few, if any, doctoral programs.”*

**Regional Universities – South Region****Regional Universities – Midwest Region****Regional Universities – West Region****Regional Colleges – North Region***“Focus on undergraduate education but grant fewer than 50% of their degrees in liberal arts disciplines. They sometimes predominantly award two-year associate degrees.”*

**Regional Colleges – South Region****Regional Colleges – Midwest Region****Region Colleges – West Region**

**Within each of these categories, colleges/universities are ranked according to these criteria:**

There are twelve Adventist colleges and/or universities in the United States. θ Two of them, *AdventHealth University* and *Kettering College*, are operated by SDA-affiliated hospital systems and are not stand-alone academic institutions. Furthermore, these institutions are exclusively focused on healthcare-related professional training and don’t report statistics to *U.S. News & World Report*. As a result, they are not classified as a **National University**, **National Liberal Arts College**, **Regional University** or **Regional College** by *U.S. News & World Report,* and are not analyzed in this study due to a lack of data.

None of the American Adventist institutions are classified as a **National Liberal Arts College, Regional College – North Region, **or **Regional University – Midwest Region.** Only one institution, *Andrews University*, is classified as a **National University**.

Many people (myself included) would consider *Loma Linda University* to be a **National University **due to its reputable medical school, dental school, school of public health, school of pharmacy, and its storied history of innovation and research in medicine. However, because there is only one school (School of Religion) not related to healthcare, it does not meet the criteria for a **National University**. While *Loma Linda* isn’t classified as a national university, it is classified as a **Global University** due to the strength of its medical programs and scientific research. Out of 1,678 global universities, *Loma Linda University *is ranked #971. Γ In all econometric analyses, I treat *Loma Linda University *as a **National University**. In doing so, I risk overestimating the financial return to American Adventist education as *Loma Linda *doesn’t offer low-paying humanities major options that lead to higher levels of student debt upon graduation (such as English/Creative Writing, Art, Photography, Graphic Design, History, Fashion, etc.). However, I believe the costs of excluding *Loma Linda* from analysis outweigh the benefits.

**Here is how American Adventist universities/colleges fare against their peers. The five numbers at the bottom of each graph, from left to right, represent the 99th percentile, 75th percentile, 50th percentile, and 1st percentile – in terms of rankings. **

##### Regional Colleges

Out of the thirty-three ranked institutions in the **Regional Colleges – West Region** category, *Southwestern Adventist University* is ranked eleventh (11/33), which is roughly the 63th percentile. *Pacific Union College *is unranked. It is unclear whether the university is unranked due to poor performance or simply because the institution did not report all data that *USNWR* requires to rank colleges.

Out of the eighty-six ranked institutions in the **Regional Colleges – Midwest Region** category, *Union College *is ranked thirty-eighth (38/86), which is around the 55th percentile.

Out of the ninety-one ranked institutions in the **Regional Colleges – South Region** category, *Oakwood University* is ranked forty-seventh (47/91), which places it just below the 50th percentile.

##### Regional Universities

Out of the one hundred and seventy-six ranked institutions in the **Regional Universities – North Region** category, *Washington Adventist University* is ranked one hundred and seventy-sixth – dead last (176/176).

Out of the one hundred and thirty-four ranked institutions in the **Regional Universities – South Region category**, *Southern Adventist University* is ranked sixty-eighth (68/134), which places it squarely in the middle of the pack.

Out of the one hundred and twenty-four ranked institutions in the **Regional Universities – West Region **category, *La Sierra University *is ranked sixty-eighth (68/124) and *Walla Walla University *is ranked seventy-fifth (75/124). So, they’re situated between the 45th and 40th percentiles.

##### National Universities

Out of the three hundred and eighty-nine ranked institutions in the **National Universities** category, *Andrews University *is ranked two hundred and ninety-ninth (289/389), which places it right around the 25th percentile.

## SDA University Outcome Metrics Performance & Data Description

I constructed this dataset directly from *USNWR* by purchasing a premium report subscription and manually inputting university statistics into a spreadsheet. While all variables comes from *USNWR*, their rankings are NOT part of the analysis.

The summary statistics (shown below) showcase the variables of interest. The sample size is 365 and the sample only includes **private, non-profit universities**; no public or private, for-profit universities are included. A broad array of institutions from all *USNWR* categories with an Adventist institution were included for a robust sample.* For a full list of all included universities, please visit the Appendices section below (A1).*

**There are three main outcome (output) variables that I examine in the analysis:**

**Median Starting Salary of Alumni ($)**- Abbreviated “MedianStartingSalary”

**Typical Debt After Graduation ($)**- Abbreviated “AverageDebtAfterGraduation”

**Average 6-Year Graduation Rate (0-100%)**- Abbreviated “Ave6YearGradRate”

The remaining variables are “input” variables. If we think of higher education as a pedagogical “production process” that leads to educated professionals (as illustrated in the graphic below), there are two classes of variables “input” variables and “output” variables. The input measures capture quality of instruction, student engagement, and academic rigor.

**The “input” variables are:**

**Enrollment (#) –***number of undergraduate students***Tuition ($) –***cost of undergraduate tuition***Room & Board ($)***cost of undergraduate room & board***Acceptance Rate (0-100%)****First Year Retention Rate (0-100%) –***percentage of students that finish their first year and continue on with their studies at the college for at least one more semester*- Abbreviated “1stYearRetentionRate”

**Peer Rating (1-5) –***“Each year, top academics – presidents, provosts and deans of admissions – rate the academic quality of peer institutions with which they are familiar on a scale of 1 (marginal) to 5 (distinguished). We take a two-year weighted average of the ratings. The 2021 Best Colleges ranking factors in scores from both 2020 and 2019.”***Percent of Faculty that are Full-Time (0-100%)**- Abbreviated “FTFaculty”

**Percent of Full-Time Faculty that have a doctorate (0-100%)**- Abbreviated “FTFacultyw/Doctorate”

**Student-to-Faculty Ratio (#)**- Abbreviated “Student : Faculty”

**Alumni Giving Rate (0-100%) –***“This is the average percentage of living alumni with bachelor’s degrees who gave to their school during 2017-2018 and 2018-2019. Giving measures student satisfaction and post-graduate engagement.”***Percent of Need-Based Aid Fulfilled (0-100%)**- Abbreviated “NeedBasedAid”

##### How do SDA colleges compare to the sample outcome averages?

The three graphs below show how SDA institutions compare to both their SDA and non-SDA peers on the three outcome measure variables for the sample population. It’s important to remember that these sample averages take all the 365 institutions in the dataset and pool them together so that, unlike the *USNWR* rankings, non-profit, private institutions of different classifications (national universities, regional colleges, and regional universities) are all part of the sample population. So, by only examining the sample averages, *Harvard University* and *Union College* are essentially being compared to one another, which is inappropriate. The econometric regression analyses neutralize this impropriety since the models control for all relevant variables.

The orange line represents the average **Median Starting Salary** of all 365 colleges in the sample. The blue line represents the Median Starting Salary values for all Adventist colleges. *Andrews University*, *Walla Walla University*, *Washington Adventist University*, *Pacific Union College*, and *Loma Linda University* have higher Median Starting Salaries than the sample average.

It’s worth noting how strongly *Loma Linda University *performs against its peers. The graphic to the right lists the institutions in the dataset with the fifteen highest median starting salaries. *Loma Linda *is fifth on the list; it outperforms all Ivy League institutions (*Harvard*, *Princeton*, *Yale*, etc.). You’ll notice that the top institutions on the list place a greater emphasis on STEM and medical fields than humanities disciplines, which have much lower starting salaries.

All Adventist institutions have a higher average student debt than the sample average. *Loma Linda University *does not report its average student debt so it is not included in the graphic above.

All Adventist institutions except for *Loma Linda University* have lower **Average 6-year Graduation Rates** than the sample average. So, for example, while *Washington Adventist University *enjoys a higher **Median Starting Salary** than the sample average for its graduates, more than 60% of students that attend the university won’t end up graduating (and won’t get to enjoy the higher salary).

## Model 1 (Median Starting Salary) Results

Model 1 uses the independent covariates on the right side of the equation (Beta1 through Beta9 and constant Beta0) to predict **Median Starting Salary. ****Adventist** is a binary random variable that equals “1” if the college is Adventist. Preliminary analysis showed no multicollinearity concerns (see **A2**). As shown in **A3**, there is a curvilinear relationship between **Median Starting Salary** and **Percent of Faculty that are Full-Time (FTFaculty)**. Because of this dynamic, a quadratic transformation of **FTFaculty **is included in the model: **FTFaculty².**

There are three iterations of the model: (1), (2), and (3). Regression (1) only includes the independent covariates shown above. Regression (2) controls for whether a college is a **National University**, **Regional University**, or **Regional College**, and also controls for whether a university’s admissions are considered *Least Selective, Less Selective, Selective, More Selective, *or *Most Selective*. While **Acceptance Rate** is one of the control variables, this does not capture the caliber of university’s applicant pool (the range of applicant ACT/SAT scores and GPAs). Regression (3) includes all Regression (2) controls but also adds state fixed effects. Because the results from Regression (3) are the most comprehensive, I’ll only discuss those results.

The Regression (3) **Adventist **coefficient equals 3759.7 and is statistically significant at the 5% level, which means that, when all other variables are held constant (“other things equal”), if a college is Adventist, we’d expect the **Median Starting Salary** of a college’s alumni to be $3,759.70 greater than if a college is not Adventist.

Another notable findings include the important relationship between **Peer Rating** and the dependent variable. **Peer Rating **is statistically significant at the .1% level and the coefficient means that, other things equal, on a 1-5 point scale, for every one point increase in the **Peer Rating **score (say from 2-3, for example), we’d expect to see a $4,229.60 increase in a college’s **Median Starting Salary**.

As shown in footnote 2, **FT Faculty **and **FTFaculty² **are jointly significant at the 5% level. The negative coefficient on **FTFaculty **and the positive coefficient on **FTFaculty² **indicate that there are increasing marginal returns to improvements in **Median Starting Salary**. When a college has a very low percentage of full-time employees, small increases in **FTFaculty** will yield little to no improvements in the dependent variable. However, past a certain threshold, moderate improvements in **FTFaculty** will lead to large improvements in a college’s **Median Starting Salary**.

As shown in footnote #1 below, we do not have evidence to reject the null hypothesis that the model does NOT suffer from omitted variable bias. The relationship (approximating a forty-five degree slope) between the fitted values (y-hat) and actual values (y) shown in **A5 **indicates that the model does a good job of predicting the dependent variable.

## Model 2 (Debt after graduation) Results

Model 2 uses the independent covariates on the right side of the equation (Beta1 through Beta9 and constant Beta0) to predict a logarithmic transformation of the variable **Average Debt After Graduation: log( AverageDebtAfterGraduation).** Preliminary analysis showed no multicollinearity concerns (see

**A5**). As shown in

**A6**, there is a curvilinear relationship between between

**Peer Rating**and

**Average Debt After Graduation**. Because of this dynamic, a quadratic transformation of

**Peer Rating**is included in the model:

**PeerRating².**In addition to the dependent variable, logarithmic transformations are conducted on several of the independent covariates leading to

**log(**,

*Tuition)***log(**,

*Room&Board)***log(**,

*MedianStartingSalary*)**log(**, and

*AcceptanceRate*)**log(**

**FTFaculty)**.In addition to allowing for a model that is a more accurate predictor of the dependent variable, the logarithmic transformations allow for an intuitive interpretation of the Beta-hat coefficients. When a logarithmic transformation is regressed on another logarithmic transformation, an elasticity is created. An elasticity is a measure of the percentage change in the dependent variable in response to a percentage change in the independent variable. This model has two types of logarithmic regressand (y) – regressor (x) relationships: Log-Level and Log-Log. The graphic below shows how the beta-hat coefficients should be interpreted.

Once again, I’ll only discuss the results from Regression (3). The fixed effects controls are identical to the setup in Model 1. The **log( Tuition) **coefficient is statistically significant at the 1% level and it indicates that, other things equal, a 1% increase in a college’s

**Tuition**cost will be associated with a .151% increase in average student debt after graduation. The

**log(**is statistically significant at the 1% level and its interpretation is that, other things equal, a 1% increase in a college’s

*AcceptanceRate*)**Acceptance Rate**will be accompanied by a .07% increase in the institution’s average level of student debt.

The regression (3) **Adventist **coefficient is significant at the 5% level. Its interpretation is that, other things equal, if a college is **Adventist**, we would expect the **Average Student Debt **to be 14.1% higher than if it were not Adventist.

As shown in footnote #1 below, unlike with Model 1, we **DO** have evidence to reject the null hypothesis that the model is not missing any variables. The relationship (again approximating a forty-five degree slope) between the fitted values (y-hat) and actual values (y) shown in A7 indicates that the model does a good job of predicting the dependent variable. However, the fit (R-Squared/Adj R-Squared) is not as strong as with Model 1 and the potential of omitted variable bias means the results should be interpreted with a degree of caution.

## Model 3 (Average 6-Year Graduation Rate) Results

Model 3 uses the independent covariates on the right side of the equation (Beta1 through Beta12 and constant Beta0) to predict **Average** **Six Year Graduation Rate****.** Preliminary analysis showed no multicollinearity concerns (see **A8**). As shown below, several of the independent variables have curvilinear relationships with **Average Six Year Graduation, **so four quadratic transformations were conducted – all of which are at least jointly statistically significant at the 5% level in Regression (3).

The regression (3) **Adventist **coefficient is significant at the 1% level. Its interpretation is that, other things equal, if a college is **Adventist**, we would expect the **Average Six Year Graduation Rate** to be about 7 percentage points LOWER than if the college were non-Adventist.

As shown in footnote #1 below, we do not have evidence (at the 5% level) to reject the null hypothesis that the model does NOT suffer from omitted variable bias. The relationship (approximating a forty-five degree slope) between the fitted values (y-hat) and actual values (y) shown in **A9 **indicates that the model accurately depicts the dependent variable.

## Conclusion

**In summary, we have significant evidence to conclude that, other things equal, Adventist schools have HIGHER ***Median Starting Salaries* than non-Adventist schools (of similar characteristics).

*Median Starting Salaries*than non-Adventist schools (of similar characteristics).

**We have evidence to conclude that, other things equal, Adventist schools have HIGHER ***Average Debt After Graduation* levels than non-Adventist schools (of similar characteristics).

*Average Debt After Graduation*levels than non-Adventist schools (of similar characteristics).

**We have significant evidence to conclude that, other things equal, Adventist schools have LOWER ***Average Six-Year Graduation Rates* than non-Adventist schools (of similar characteristics).

*Average Six-Year Graduation Rates*than non-Adventist schools (of similar characteristics).

# Appendices

∫ Bull, Malcolm., Lockhart, Keith. (2006). “Seeking a Sanctuary: Seventh-day Adventism and the American Dream.” Indiana University Press. 978-0-253-21868-1. p.113.

θ Adventist Colleges & Universities. (2020). https://adventistcolleges.org/

∝ U.S. News & World Report. (2020). “How U.S. News Calculated the 2021 Best Colleges Rankings.” https://www.usnews.com/education/best-colleges/articles/how-us-news-calculated-the-rankings

Γ U.S. News & World Report. (2020). “Loma Linda University.” https://www.usnews.com/education/best-global-universities/loma-linda-university-117636#summary