Determinants of Starting Salaries for Filipino Graduates: An Explainable Machine Learning Approach
This study employs an explainable machine learning approach on crowd-sourced survey data to demonstrate that job role and industry are the primary determinants of starting salaries for Filipino graduates, significantly outweighing institutional prestige and suggesting a need to prioritize sector-specific skills in career guidance and policy.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the Philippines, a significant number of college graduates find themselves without work, creating a sharp disconnect between the years spent in the classroom and the reality of the job market. For those who do find employment, the starting salary they receive acts as a clear signal of how much the economy values their specific background. Traditionally, researchers have tracked these outcomes by simply counting how many graduates are employed and what their average pay is, much like taking a headcount at a school reunion. However, these counts tell us what happened, not why it happened or what will happen to the next group of students. To understand the true drivers of pay, one must look beyond simple averages and ask which specific factors actually push a salary up or pull it down.
A team of researchers from the University of the Philippines sought to answer this question by applying modern computer learning techniques to a unique set of data. They analyzed responses from a crowd-sourced survey called "First Pay," which gathered information from nearly three thousand Filipino graduates about their first jobs. Because the data came from an open online form, it was messy and unstructured; job titles were written in many different ways, and some important details like company size or location were missing. The researchers treated this noisy information as a puzzle, using machine learning to predict which salary bracket a graduate would fall into based on their education, gender, and job details. Their goal was not just to make a prediction, but to use a method called explainable artificial intelligence to understand exactly which pieces of information the computer relied on to make its decision.
The study revealed a clear and somewhat surprising truth about the local labor market: the specific job role and the industry a graduate enters are far more important than the prestige of the university they attended. When the researchers let the computer analyze the data, it consistently ignored the reputation of the school and focused almost entirely on what the person actually does and where they work. For instance, the computer learned that a graduate working in a specific sector like mining or management would likely earn significantly more than someone in construction or clerical work, regardless of whether they graduated from a private or a state university. This finding was so strong that it appeared across three different ways of testing the data, suggesting that the market rewards functional skills and sector alignment much more than it rewards the brand name of a degree.
To ensure this result was not a fluke, the researchers tested their approach in several different ways. First, they used a technique that breaks down a prediction to see which input contributed the most, confirming that job titles and industry names were the primary drivers. Second, they tried to improve the computer's ability to read the messy text of job descriptions by using advanced language tools, only to find that the computer still relied most heavily on the specific words describing the job and industry. Finally, they tried to reframe the problem so the computer could reason about the economic meaning of a salary range rather than just matching keywords. Even with this more sophisticated reasoning approach, the computer still leaned heavily on the raw text of the job role and industry to make its best guesses. This consistency across different methods suggests that the dominance of job role and industry is a real feature of the data, not an artifact of a specific computer model.
The researchers also discovered that the difficulty in predicting exact salaries was not due to a lack of computing power, but rather the limitations of the data itself. Because the survey relied on self-reported information and lacked details like company size or years of experience, there was a natural ceiling to how accurately the computer could predict outcomes. The data was particularly fuzzy for graduates earning middle-range salaries, where many different types of jobs overlap in pay. While the researchers found that grouping salaries into broader, economically meaningful categories improved the computer's accuracy, the fundamental lesson remained the same: the specific path a graduate takes matters more than the school they came from.
Ultimately, this work suggests that career guidance and education policy in the Philippines should shift their focus. Instead of emphasizing the prestige of a university, the evidence points toward helping students align their skills with specific high-value sectors and roles. The study demonstrates that while the data was imperfect, the signal was clear enough to show that the labor market values what a graduate does far more than where they studied. By moving beyond simple descriptions of employment rates to a deeper, data-driven understanding of what drives pay, this research offers a clearer map for graduates, educators, and policymakers navigating the complex landscape of the Philippine job market.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.