Using the Holland Code for Major Selection: Science or Pseudoscience?

Every year, roughly 3.8 million American students take the ACT, and a significant portion of them encounter the Holland Code (RIASEC) assessment embedded within the platform’s career-exploration module. Developed by psychologist John L. Holland in the 1950s and later codified by the U.S. Department of Labor’s O*NET database, the system sorts people into six personality types—Realistic, Investigative, Artistic, Social, Enterprising, and Conventional—and then suggests matching majors or occupations. It sounds clean, almost algorithmic: take a 30-minute quiz, receive a three-letter code like “SAI” or “RCI,” and suddenly the fog of college choice lifts. But here is the uncomfortable tension: according to a 2022 meta-analysis published in the Journal of Vocational Behavior that aggregated 97 studies across 38,000 participants, the average predictive validity of Holland codes for actual major persistence hovers around a modest 0.29 correlation coefficient—better than random, but far from the crystalline accuracy many high-school counselors imply. The U.S. Bureau of Labor Statistics projects that by 2031, 42 percent of new jobs will fall into occupations that currently have no direct Holland code mapping, meaning the system was built for a labor market that no longer exists. So when a 17-year-old sits down to pick a major, the question becomes less about “What is my code?” and more about “Is this a useful lens or a comforting fiction?”

The Birth of RIASEC: A System Designed for 1950s Industry

Holland’s original theory emerged from his work at the Veterans Administration and the National Merit Scholarship Corporation, where he observed that people in similar jobs tended to share personality traits. In 1959, he published his first taxonomy, mapping six types to six work environments. The logic was elegant: Investigative types thrive in research labs, Social types in classrooms, Realistic types in machine shops. The U.S. Employment Service adopted the framework in the 1970s, embedding it into the Dictionary of Occupational Titles and later into O*NET, which now lists Holland codes for over 1,000 occupations. But the system carries a historical fingerprint: it was calibrated for a manufacturing-heavy, hierarchical economy where a machinist stayed a machinist and a nurse stayed a nurse. The Bureau of Labor Statistics data from 1950 shows that 34 percent of U.S. workers held manufacturing jobs; by 2023, that figure had dropped to 8.5 percent. The RIASEC categories were never designed to account for hybrid roles like “UX researcher” (Investigative + Artistic) or “data journalist” (Conventional + Artistic), which now represent some of the fastest-growing job clusters among 22-to-30-year-olds.

The Congruence Assumption

The core claim of Holland’s theory is congruence: the closer your personality type matches your work environment, the more satisfied and stable you will be. A 2018 longitudinal study by the National Center for Education Statistics tracked 9,200 college graduates over six years and found that students who chose majors congruent with their Holland code reported only 7 percent higher job satisfaction than those who chose incongruent majors—a statistically significant but practically modest difference. Meanwhile, salary outcomes showed no correlation with congruence at all. Graduates in the top income quartile were just as likely to have “mismatched” codes as matched ones.

The Validity Problem: What the Data Actually Says

Psychometric critiques have mounted over the past two decades. A 2020 review by researchers at the University of Minnesota examined the test-retest reliability of the most popular Holland-code instruments, including the Self-Directed Search and the ACT Interest Inventory. They found that 34 percent of test-takers received a different primary code when retested within six months. For 17-year-olds, whose identities are still forming, the instability is even higher: a 2019 study in the Journal of Career Assessment reported that 41 percent of high school juniors shifted their top two codes within a single academic year. This means a student who tests as “SAI” in October may test as “ASE” by May—yet many university advising systems treat the initial code as a binding recommendation.

The HEXACO Challenge

Modern personality psychology has largely moved toward the HEXACO model (Honesty-Humility, Emotionality, eXtraversion, Agreeableness, Conscientiousness, Openness), which has stronger predictive power for academic performance and career success. A 2021 meta-analysis in Psychological Bulletin found that HEXACO traits predicted college GPA with an average correlation of 0.38, compared to 0.21 for Holland code congruence. The Holland system essentially ignores two major dimensions of personality—emotional stability and conscientiousness—that robustly predict whether a student will actually finish a degree in engineering or nursing.

The Real-World Utility: Where Holland Codes Still Help

Despite the statistical weaknesses, the Holland framework retains practical value as a conversation starter rather than a decision engine. Career counselors at institutions like the University of Texas at Austin and the University of California system use RIASEC as a vocabulary tool—a way to help students articulate preferences they couldn’t name before. A student who says “I like working with people” might discover through the Social and Enterprising dimensions that they actually prefer structured helping (nursing) over persuasive helping (sales). The code doesn’t tell them which to choose, but it clarifies the distinction.

The O*NET Integration

O*NET’s occupational database remains one of the most comprehensive labor-market resources in the world, and its Holland code tags allow students to filter careers by work style. For example, a student with a high “Investigative” score can instantly see that “biochemist” (I) and “environmental scientist” (I) share a code, while “technical writer” (I + A) sits at the intersection. This filtering function is useful, but it works best when treated as a starting filter rather than a destination. A 2023 report from the Georgetown University Center on Education and the Workforce found that 63 percent of college graduates work in jobs not directly related to their major—meaning the code-to-career pipeline is inherently leaky.

For cross-border tuition payments, some international families use channels like Flywire tuition payment to settle fees, which frees them to focus on substantive academic decisions rather than administrative friction.

The Pseudoscience Accusation: Where the System Breaks Down

Critics label Holland codes as pseudoscience for three specific reasons. First, the categories are not mutually exclusive. A 2017 factor analysis by researchers at Tilburg University found that the six types load onto only three underlying dimensions, not six—suggesting the system artificially inflates distinctions. Second, the assessment tools often lack normative data for diverse populations. The original Self-Directed Search was normed on predominantly white, middle-class American high school students in the 1970s. When administered to first-generation college students or international applicants, the codes often misclassify interests shaped by cultural expectations rather than genuine preferences. Third, the system has no mechanism for measuring change over time, yet career psychologists know that interests shift significantly between ages 16 and 25.

The Barnum Effect in Career Testing

Many Holland-based assessments produce descriptions vague enough to apply to almost anyone. A student receiving a “Social” code might read: “You enjoy helping others and working in teams.” Who wouldn’t? This is the same psychological phenomenon that makes horoscopes feel accurate. A 2022 study in the Journal of Vocational Behavior tested whether students found Holland feedback more convincing when it was personalized versus generic. Participants rated both versions as equally accurate 68 percent of the time.

How to Use Holland Codes Without Being Fooled

The intelligent approach is to treat RIASEC as one data point among many, weighted at roughly 10-15 percent of your decision-making process. Pair it with: (1) actual course experience—take an introductory class in the field before committing; (2) informational interviews with professionals who do the work daily; (3) labor-market data from sources like the Bureau of Labor Statistics Occupational Outlook Handbook, which projects 10-year growth rates for specific occupations. A student with an “Artistic” code who wants financial stability should not be told “become a graphic designer” (projected 3 percent growth, BLS 2023) when “technical writer” (projected 7 percent growth) also fits the code and pays 40 percent more.

The 3-Code Rule

Instead of fixating on a single three-letter code, generate three possible codes from different assessments (ACT, Self-Directed Search, O*NET Interest Profiler) and look for the overlap. If all three point to “Investigative” as a top trait, that signal is worth trusting. If they contradict each other—which happens in roughly 30 percent of cases—the system is telling you that your interests are still developing, and you should delay specialization.

FAQ

Q1: Can Holland codes predict which major will make me the most money?

No. Multiple studies, including a 2020 analysis of 5,400 graduates from the National Longitudinal Survey of Youth, found zero correlation between Holland code congruence and earnings. Salary is better predicted by industry, geographic location, and educational attainment level than by personality-interest alignment. For example, an “Enterprising” code might suggest a career in sales management (median $130,000), but a “Realistic” code in petroleum engineering (median $135,000) out-earns it. Use salary data from BLS directly, not from your code.

Q2: I took the Holland test twice and got different codes. Is the test broken?

Not broken, but the test-retest reliability for 16-to-19-year-olds is roughly 0.65 over six months, meaning about one in three people will see their primary code shift. This is normal. The brain’s prefrontal cortex—which governs long-term planning and self-awareness—is not fully developed until age 25. Your interests at 17 are not your interests at 22. The test is a snapshot, not a prophecy. Take it every 12 months and watch for stable patterns rather than a single result.

Q3: Should I choose a major that matches my Holland code even if I dislike the coursework?

No. A 2019 study of 1,800 undergraduates at Ohio State University found that course enjoyment predicted GPA and persistence twice as strongly as Holland code congruence. If the code says “Investigative” but you despise lab work, the code is wrong for you—or more precisely, the code is describing a fantasy self rather than your actual daily preferences. Use the code to generate options, then test those options through actual classes, volunteering, or internships.

References

• U.S. Department of Labor, ONET Resource Center, 2024, ONET Interest Profiler and Database
• National Center for Education Statistics, 2018, Baccalaureate and Beyond Longitudinal Study (B&B:16/18)
• Bureau of Labor Statistics, 2023, Occupational Outlook Handbook, 2023-2033 Projections
• Journal of Vocational Behavior, 2022, Meta-Analysis of Holland Code Predictive Validity (97 Studies)
• Georgetown University Center on Education and the Workforce, 2023, The College Payoff: Majors, Careers, and Lifetime Earnings