Built on research developed in The Wharton School of the University of Pennsylvania Insurance Department, the Human Capital Scoreâ„¢ combines credit-risk tools and metrics to give insight into each borrower’s future potential. Using merit data such as GPA, standardized test scores, college and major it provides a true, unbiased measure of the economic value of an education that empowers students to make better educational decisions and offers multiple advantages for both borrowers and lenders.
When you apply for a private student loan, lenders want to know what risk they would take by loaning money to you. To be able to do these assessments there needs to be a set of standardized (i.e. common and comparable) and verifiable (i.e. trustworthy) attributes.
The most common form of credit score is the “FICOÂ® score”. (Learn more about FICOÂ® scores). FICOÂ® scores are based upon such attributes as the number of credit cards you have, outstanding balances, payment history, bankruptcy. Based on this form of methodology, students (or young adults in general) will fare poorly, as they do not have a long (or even medium) positive history of payments. As a result, they will generally receive low FICOÂ® scores and, thus, will look like very risky propositions. This is where the Human Capital Scoreâ„¢ comes in.
The starting premise for the Human Capital Scoreâ„¢ is that there is a way to assess the relative riskiness of students by looking at a different set of standardized and verifiable attributes. These attributes help predict their future income, and hence their ability to pay back loans. These relevant attributes are items such as school, major, GPA, and standardized test score. The Human Capital Scoreâ„¢ uses these as inputs to create a score.
Why hasnâ€™t this been done before?
FICOÂ® scores are based on attributes that are easy to quantify and rank. More delinquencies are bad; a longer credit history is good. Ordering student attributes is not so simple. How do we know which schools or majors are better and worse? We must collect, clean, and integrate additional data about schools, majors and such. This requires expertise, time, and money.
What insights does the Human Capital Scoreâ„¢ provide?
Income predictions for the 10 years post graduation The Human Capital Scoreâ„¢ aims to predict the income paths of college students in the 10 years after graduation. This allows lenders to separately identify the ability of college students to repay loans of long- and short maturities. The ability to repay short-term loans can be assessed by looking at predicted income shortly after graduation; the ability to repay longer-term loans can be assessed by looking at predicted income over a longer period. Currently, we provide a Human Capital Scoreâ„¢ for students for the period 2 years and 8 years post-graduation.
Uncertainty of the income predictions Clearly the future is uncertain and any average income prediction is only part of the story. Not only is it important to know the predicted income, but also, which students are more (or less) likely than others to follow this path. Shortly, the Human Capital Scoreâ„¢ will include measures of uncertainty, to quantify the range of possible income paths. This makes it possible to predict the probability that income will fall below a certain threshold in a given year, or that average income will fall below a certain threshold in the 10 years following graduation.
Scores tailored to specific loans Not all loans are created equal. There is a world of difference between a medical student to have to repay a short-term loan within a couple of years of graduating from undergraduate school, and a loan that has deferred payments not coming due until 10 years later. We are working to develop Human Capital Scoreâ„¢ rankings that are tailored to specific loans, so that students would receive one ranking for a 3 year loan without principal deferment options and another for a 10 year loan with principal deferment options.
Who can use the Human Capital Scoreâ„¢?
The Human Capital Scoreâ„¢ can be used by borrowers, such as college students and prospective college students. It can also be used by lenders, whether individual lenders or financial institutions. Borrowers
College students, or prospective college students, can use their Human Capital Scoreâ„¢ to communicate to lenders about their ability to repay loans. While the model uses inputs such as college, major, GPA and standardized test scores, it recognizes that not all students will have information on all these attributes, e.g., prospective or new students will not have clear ideas about a major and, depending on your choice of college, you may not have needed to take any standardized tests. Even without all this information, the system can still generate a Human Capital Scoreâ„¢. (Note: currently the system only accepts the SAT as an option for a standardized test. We will be adding more standardized tests, such as the ACT and GRE shortly. Similarly we will be adding additional fields to add such items as secondary major, major GPA, etc. as we continue to develop and improve the system.) Lenders
Individual lenders can see the Human Capital Scoreâ„¢ for each borrower on the People Capital peer-to-peer lending site (coming soon). Or, they can use the Human Capital Scoreâ„¢ calculator to create a Human Capital Scoreâ„¢ for a potential borrower that does not already have one.
Institutional lenders (that might have hundreds, or thousands, of individuals to create scores for) should contact us at firstname.lastname@example.org to discuss use of our API or XML link thatâ€™s allows multiple scores to be calculated from uploaded files.
On what scale is the Human Capital Scoreâ„¢ provided?
While the algorithm behind the scenes produces very fine graduations of credit risk, we are currently providing the results on a simplified 1-9 scale, with “+” and “-” to denote scores that are at the ends of the spectrum. (Ultimately, this produces a 27-point scale.) Currently, we provide a Human Capital Scoreâ„¢ for students for the period 2 years and 8 years post-graduation.
To help explain the various scores, we have provided the following summary table:
9 Highest 6 Above Average 3 Weak 8 Very Good 5 Average 2 Very Weak 7 Good 4 Below Average 1 Lowest
So how does Human Capital Scoreâ„¢ work?
Built on research developed in The Wharton School of the University of Pennsylvania Insurance Department, the Human Capital Scoreâ„¢ combines credit-risk tools and metrics to give insight into each borrower’s future potential.
We have data (that arenâ€™t publicly available) about a large number of students. We know their majors, schools, grades, scores, and a host of other attributes. We know how much these students earned in the years after they graduated. We can use this information to create income predictions for students based on their attributes. If students in our data who study engineering and have good grades had high and growing incomes after graduation, the HCS will assign high and growing incomes to engineering students with good grades who ask to be given an HCS score.
To do this, we merge into these data additional information from other sources on how much students with various majors earn, the attributes of the various schools, etc. This allows us to make high quality predictions about the future incomes of students, even when we donâ€™t have data on many (or even any) students who went to that school or had that major.
Naturally, we canâ€™t get data on the incomes of students 10 years after graduation except from students who graduated at least 10 years ago. To make sure the Human Capital Scoreâ„¢ reflects the most recent patterns in graduatesâ€™ incomes, we merge in, and extrapolate, from the most recent trends in the overall income distribution of college graduates.
Because we have individual-level data on many students, we can predict not only average income but also the range of possible income paths. Students from a given major and school may all have relatively similar incomes; another major or school may have wide variation in graduatesâ€™ incomes. The Human Capital Scoreâ„¢ will be able to provide a variety of statistics relevant for repayment, not just expected income. We can estimate the probability that income will fall below a certain value, or the expected income in the worst case scenario (e.g., worst 10 percent). We can compute the probability that lifetime income will fall below a certain threshold.
What canâ€™t the Human Capital Scoreâ„¢ do (aka “the fine print”)?
While the Human Capital Scoreâ„¢ measures the income, and thus the ability to pay, it does not measure the willingness to pay. If someone with a high income is unwilling to make loan payments, or someone with no income still makes loan payments, this isnâ€™t captured by the model. HCS measures ability to pay, not propensity to pay.
While the Human Capital Scoreâ„¢ provides income projections for college students in the 10 years after graduation, we do not have a crystal ball. These projections are based on data about the incomes of people who have already graduated from college and are working now. If economic conditions shift, we wonâ€™t capture that. Most obviously, if the current recession reduces the incomes of college graduates in the coming years, our estimates of income will be systematically too high. That said, it will continue to show the relative ranking of college students. So it will continue to show which students are relatively better options than others.
The Human Capital Scoreâ„¢ can only predict income using standardized attributes. An engineering major from MIT with high scores and grades will have a high HCS because past engineers from MIT with high scores and grades have on average enjoyed high incomes after graduation. If this particular student plans to join the circus (no disrespect to this particular career path intended, just that it traditionally affords a lower income level) after graduation, the Human Capital Scoreâ„¢ cannot reflect this. We have no way of reliably verifying information about specific studentsâ€™ work plans or expectations. We can only rely on information that we can verify.