Price Discrimination in Consumer Finance: Required, Permitted, and Prohibited
Price Discrimination in Consumer Finance: Required, Permitted, and Prohibited
Charging different people different prices for the same loan is, depending on the basis, forced by competition, tolerated, or illegal. A lender that charges everyone 19% loses its low-risk borrowers to a competitor at 16% and ends up charging 23.5% to the rest. A lender that charges existing mortgage customers 7.1% while offering 5.2% to new ones collects $4,207 a year from each one who does not move. The fairness argument in lending is an argument about which of three rows a practice belongs in.
What you'll learn
The industry where charging different prices is required, permitted, and banned at once
Price discrimination means charging different customers different prices for the same thing. In most industries it is a tactic. In consumer finance it is three different things depending on the basis for the difference.
| Basis for the price difference | Status | Example |
|---|---|---|
| Expected loss | Required in practice | A 16% rate for a low-risk borrower and 23.5% for a higher-risk one |
| Willingness to search or switch | Permitted | A new customer paying 5.2% while the existing customer pays 7.1% |
| Protected characteristics, or proxies for them | Prohibited | Any difference traceable to race, sex, age, or the variables that stand in for them |
The three rows are not three policies. They are three answers to one question: what is the lender allowed to know about you, and what may it do with it? Most of the argument about fairness in lending is an argument about which row a given practice belongs in.
Why the first row is close to mandatory
Suppose a lender refuses to discriminate at all and charges every borrower the same rate. Its pool has two groups: 60% with an expected loss of 3% and 40% with an expected loss of 10.5%, for a pooled loss of 6%. With funding at 8% and operating cost at 5%, the uniform rate has to be about 19%.
- A lender that prices by risk offers the low-risk group 16% and the high-risk group 23.5%.
- The low-risk borrowers leave the uniform lender for the 16% offer.
- The uniform lender's pool is now entirely the 10.5% group, and its rate has to rise to 23.5% to break even.
This is the adverse selection argument, and it is why risk-based pricing is required in practice rather than by rule. Per our rationing analysis, the alternative to pricing risk is rationing it, meaning declining the high-risk group entirely, and per our selection analysis that is the choice lenders make when a rate cap stops them pricing. With a 36% cap, a group whose expected loss is 28% needs a rate of 41% and is simply not served.
The row is also where most people's intuition about fairness breaks. Charging the riskier borrower more looks like penalizing someone for being poor. Charging everyone the same looks fair and ends with the low-risk borrower subsidizing the high-risk one until she leaves, after which nobody is subsidized and the high-risk borrower pays full price anyway. Per our price decomposition analysis, a loan price is mostly expected loss and fixed cost, and those are properties of the loan, not opinions about the borrower.
The second row: pricing on what you will tolerate
The second kind of price discrimination has nothing to do with risk. It charges people according to how much effort they will spend to avoid being charged.
A mortgage lender offers 5.2% to new customers and leaves existing customers at 7.1%. On $300,000 over 25 years, that is $1,789 a month against $2,140: a difference of $351 a month and $4,207 a year, paid by the customer who did not switch. Per our exit analysis, this is the loyalty penalty, and per our search analysis it exists because searching costs time and the lender can tell who will not pay that cost.
| Practice | What it prices |
|---|---|
| New-customer rates below back-book rates | Inertia. Per our refinance analysis, the customer who does not run the arithmetic |
| Fee waivers on request | Assertiveness. Per our exception analysis, the fee is paid by those who do not ask |
| Promotional rates that reset | Attention. Per our deferred interest analysis, the price depends on remembering a date |
| Relationship discounts | Concentration. Per our bundling analysis, a lower price for giving up the ability to leave |
| Opaque fee schedules | Comparison effort. Per our reference point analysis, a price that cannot be compared is not competed |
None of this is illegal, and some of it is defensible. Acquiring a customer costs money, and a lower introductory price that recovers that cost later is an ordinary commercial structure. The question is whether the gap between the front book and the back book reflects acquisition cost or simply the measured unwillingness of existing customers to move, and the $4,207 a year in the mortgage example is well past any acquisition cost.
The second row borrows the first row's legitimacy
Risk-based pricing earned public acceptance because it is tied to something real. The loyalty penalty is often presented in the same vocabulary, as pricing to the individual, and the resemblance is useful to the lender because the public has learned that individual pricing is fair. The test that separates them is simple: would the price difference disappear if both customers had identical expected losses? For the first row, yes. For the second, no. Any practice that prices two identical risks differently is in the second row whatever it is called.
The third row and the proxy problem
The third row is the one the law addresses directly. A price difference based on a protected characteristic is prohibited whether or not it would be profitable, and the prohibition extends to variables that stand in for the characteristic.
That extension is where the first and third rows collide. Per our proxy analysis, a variable can be genuinely predictive of loss and also closely correlated with a protected group, and a lender using it is doing first-row pricing with a third-row effect. Per our fair lending analysis, the legal question is then whether the variable is justified by business necessity and whether a less discriminatory alternative exists, which is a question about the variable rather than about intent.
- A variable that predicts loss only through its correlation with the group is a proxy and belongs in the third row.
- A variable that predicts loss within the group as well as across groups is doing first-row work and carries a third-row burden of proof.
- Per our insurance score analysis, the same file used to price car insurance raises the same question in another industry, and several states have answered it by prohibition.
The practical consequence is that the first row is not as clean as the adverse selection argument makes it sound. Risk-based pricing is required by competition and constrained by the third row, and the constraint binds exactly where risk is most correlated with group membership, which is also where the pricing has the most effect on who gets credit.
Who ends up paying more
Put the three rows together and ask who pays the highest price. The answer is not simply the riskiest borrower.
- The riskiest borrower pays the most under the first row, by construction, and is declined entirely where a cap prevents pricing.
- The least attentive borrower pays the most under the second row, and attention correlates with time, income and familiarity with the system. The households with the least time to shop pay the loyalty penalty most often.
- The two rows stack. A borrower who is both higher risk and less able to search pays the risk premium and the inertia premium, and per our poverty premium analysis that stacking is a large part of why having less money costs more.
The second row is where the stacking is avoidable. Expected loss is a property of the loan. Inertia pricing is a choice the lender makes about how to treat people who do not leave, and per our contract analysis it is embedded in terms nobody negotiates and nobody reads.
The case for the second row
Economists have a real argument that inertia pricing can increase access. If a lender can charge inattentive customers more later, it can charge attentive customers less now, and the low introductory price brings in people who would not have qualified or applied at the full price. The subsidy runs from the inattentive to the attentive, and the attentive include some who could not otherwise afford the product at all. The argument is correct as far as it goes. Its weakness is that the subsidy runs in the wrong direction by any measure of need, since the inattentive are disproportionately the time-poor, and that the lender captures a share of the gap rather than passing all of it forward. A regulator that bans the back-book gap should expect introductory prices to rise, and a regulator that permits it should expect the gap to be paid by the people least able to notice it.
Testable implications
- Markets where risk is easily observed should show narrow price dispersion within risk tiers and wide dispersion between them. Where risk is hard to observe, the reverse.
- The front-book to back-book gap should be largest in products with the highest switching cost and the longest term, and smallest in products that can be switched in an afternoon.
- Rate caps should produce declines concentrated in the risk bands just above the cap, not a general reduction in credit, with borrowers below the cap unaffected.
- Removing a predictive variable on fairness grounds should raise prices for the group it was meant to protect in the first round, as the lender reverts to pooled pricing, before any second-order effect on access.
The fourth implication is the uncomfortable one and it is worth stating plainly. Per our boundary analysis, a decision rule that cannot use a variable does not stop deciding; it decides with less information, and less information means more pooling, and pooling moves prices toward the average for everyone in the pool. Whether that trade is worth making is a value judgement, and per our adverse action analysis it is one that has to be made in the open, because the price of the rule is paid by identifiable people.
Frequently asked questions
Because uniform pricing collapses under adverse selection: the low-risk borrowers leave for 16% and the uniform lender ends up at 23.5% for the rest.
Charging existing customers more than new ones for the same risk. Here $4,207 a year on a $300,000 mortgage. Legal, and distinct from risk pricing.
At proxy variables that predict loss and also track protected groups. Business necessity and less discriminatory alternatives decide which row they belong in.
Borrowers who are both higher risk and less able to search, because the risk premium and the inertia premium stack. Only the second is avoidable.
Key takeaways
- Price differences based on expected loss are forced by competition; on inertia, tolerated; on protected traits, prohibited.
- A uniform 19% rate collapses to 23.5% once a competitor offers low-risk borrowers 16%, because they leave.
- The alternative to pricing risk is rationing it, which is what a rate cap produces for groups above the cap.
- A 5.2% new-customer rate against 7.1% for existing customers costs $4,207 a year on a $300,000 mortgage.
- The test separating the rows: would the gap vanish if both customers had identical expected losses?
- Proxy variables are where risk pricing and fair lending collide, and the constraint binds where it matters most.
- Risk premiums and inertia premiums stack on the same households, and only the inertia premium is avoidable.
- Removing a predictive variable moves prices toward the pool average for everyone in the pool, which is a trade made in the open or not at all.
This analysis is educational and does not constitute legal, financial, or compliance advice. All figures are stylized illustrations using assumed loss rates, funding and operating costs, group shares, and mortgage terms, and do not describe any particular lender, product, or market. Fair lending law, the treatment of proxy variables, disparate impact analysis, rate caps, and permissible pricing practices are governed by federal and state law and supervisory guidance that vary by product and jurisdiction and change over time; consult qualified counsel before relying on any framework described here.