How Credit Scores Are Built

Part One · The price of money

How do credit scores work? Begin with a puzzle. Children under 18 generally have no credit report, the Consumer Financial Protection Bureau says. The CFPB's list of exceptions includes a file created in error, a loan taken out by an identity thief, and a child who is an authorized user or joint holder on an adult's account. Yet when that child later applies for an apartment, an insurance policy or a car loan, a stranger may sum the person up as a number between 300 and 850. A credit score is a prediction of credit behavior, such as how likely someone is to repay a loan on time. It is built from information in credit reports. To see why anyone wants that prediction, start with what a lender actually sells.

Suppose that young person, a few years on, asks a bank for a loan. A bank takes in deposits and uses them, with other borrowed funds, to make loans or buy securities, as the Federal Reserve Bank of San Francisco puts it. To earn a profit, the St. Louis Fed adds, a bank will generally set its loan rate above the deposit rate it pays. Two dated numbers show the gap. The FDIC's national average rate on savings accounts was 0.38 percent in October 2026. The Federal Reserve reported a preliminary average of 22.36 percent in August 2026 for card accounts that were charged interest. Both numbers are averages. For one borrower, the CFPB says, companies use scores to decide whether to lend, and to set the interest rate and the credit limit.

The bank's next question is what a loan costs over time. Interest is the price of using money, and it compounds. The SEC's investor education site gives the standard case. At 5 percent a year, $100 becomes $105 after one year and $110.25 after two, because the second year's interest is figured on $105. By our arithmetic, ten years gives $162.89. The Rule of 72 estimates doubling time: divide 72 by the rate. At 9 percent that is 8 years, and the exact answer is 8.04. At 5 percent the rule says 14.4 years, against an exact 14.2. A debt that is left unpaid grows the same way, which is one reason a lender wants a prediction of who will repay it.

Suppose, years later, the young person does get a card. A credit card's interest rate is the price of borrowing, stated as a yearly rate called the annual percentage rate, or APR. The CFPB explains that many card companies calculate interest daily from the average daily balance. On most cards, a purchase balance paid in full by the due date carries no interest. By our arithmetic, a 22 percent APR works out to about 0.06 percent a day. Suppose, as a teaching case, a $1,000 balance sat for a year with no payments and no new purchases. At simple interest it would cost $220. If each day's interest were added to the balance and charged again, it would cost about $246. That second figure is conditional, because card terms differ.

Part Two · The file

Now look at the file itself. Its raw material is the credit report, which the CFPB defines as a statement of credit activity, such as loan-paying history and account status. Three nationwide companies collect it from creditors: Equifax, Experian and TransUnion. Creditors are not required to report to all three, so the three files can differ. People can view each one free once a week at AnnualCreditReport.com. A scoring model, which the CFPB calls a mathematical formula, then turns a report into a number. So there is no single score. Each depends on the data, the model and even the day of calculation. For one person in four, the CFPB found, an educational score differs meaningfully from a lender's.

Two companies publish what their models weigh, and they disagree. FICO groups its data into five categories. Payment history counts 35 percent, amounts owed 30, length of credit history 15, new credit 10 and credit mix 10. VantageScore's 4.0 model, according to its page of June 2025, gives payment history 41 percent, depth of credit 20, credit utilization 20, recent credit 11, balances 6 and available credit 2. The categories overlap but do not line up one to one. FICO adds that their importance may vary from person to person, and that the exact effect of a single factor cannot be measured without the entire report. The percentages are a general guide, not a recipe.

What a score leaves out is as informative as what it holds. FICO says its standard scores are calculated only from the information in the credit report. Personal details such as income, age and where you live do not factor in, and neither do salary, occupation or employer. Lenders may weigh those things separately. The CFPB adds that most utility companies do not report on-time payments to the big three, although an unpaid bill sent to collections can appear. A score is therefore not a measure of income, effort or character. It is a statistical prediction of one behavior, repaying credit, made from one kind of record. Someone with a high income and no borrowing can still look blank to a scoring model.

Because lenders read it, the number reaches beyond loans. The CFPB lists mortgages, credit cards and auto loans, and also tenant screening and insurance. With a person's agreement, an employer may look at a credit report. Usually a higher score makes it easier to qualify, and it may bring a better rate. Most scores run from 300 to 850, a scale VantageScore also uses. Which model a lender picks may depend on the loan. A 2015 CFPB study made the same point from the other side. Lenders are much less likely to extend credit to someone with no record, or with too little history to assess.

Part Three · Who may borrow

One rule sits where scores begin. Under the Truth in Lending Act, as changed by the CARD Act of 2009, a card company generally cannot issue a credit card to anyone under 21. The exceptions are a person who shows an independent ability to meet payments, or an adult over 21 who co-signs, the CFPB says. Age cannot generally be used in other lending decisions. Richard Cordray, then director of the CFPB, defended the limit in October 2013. He said the Act was designed to protect young consumers from cards they cannot afford, because students had too easily run up debt they could not repay. He called the halving of under-21 cardholders an intended consequence.

Researchers have studied what the rule did. In a 2014 paper, a team from the Richmond Fed and Arizona State University found that it reduced young borrowers' access to credit, which was its aim. They also found that people who got a card before 21 defaulted less and were more likely to get a mortgage early. In 2020, researchers from the Boston Fed and the University of Delaware found more young adults with no credit history, or too little to be scored, than before the Great Recession. They called this likely an unintended outcome of the Act. So the same fall in young cardholders is the goal in one account and the cost in the other. Each paper says its views are the authors', not the Fed's.

So the number is a forecast built from a past, and for a teenager the past is often a blank page. A file appears when an adult adds a child to an account, when an error creates one, or when a thief uses the child's name, and otherwise it waits. When it does exist, it is made from what creditors report: paid late or on time, how much is owed, how long accounts have lasted, how many are new. None of that says whether a person is wise, kind or hardworking. Somewhere a bureau holds a file with your name on it, or does not. The score that comes out of it is a prediction about the next loan.