From The New York Times:
The noxious penalty imposed on American Express and Discover consumers who exceeded their spending limit has finally died, quashed by legislation signed in May by President Obama to ease onerous fees for cardholders. In recent days, American Express customers began receiving notices of the fee elimination, which takes effect in October. Discover Card customers will soon get similar notifications.
There’s a section in this article that I want to unpack. Let me reproduce it in full:
The changes at Discover and American Express (which, by the way, included in its letter a notice that it would raise late fees and interest charges) come in the wake of the Credit Card Accountability, Responsibility and Disclosure Act, which bars card companies from penalizing consumers unless they specifically ask to breach their limit and agree to pay for the privilege.
Rather than levying the fee automatically, both firms will now use technology to decide whether and when to allow consumers to exceed their limit, based on the cardholders’ recent spending and creditworthiness.
But before you gush over the changes, consider this: Both firms probably arrived at their decision based on a calculus that showed it would be too costly to build a system that lets consumers opt to breach their credit limit, as the law required. Instead, it was cheaper to simply do away with the fee.
Desiree Fish, an American Express spokeswoman, said the company had determined that the cost to build such a mechanism would be more than it received from over-the-limit fees. Most other banks are likely to come to the same conclusion, particularly since they would have to tell customers why they are levying a fee when competitors are not doing so anymore.
A few things.
Notice what “building a system” entails here. On the back end, if a consumer has exceeded their limit with their purchase, they’d have to do a check to see if the consumer has signed off on paying a fee to go over their limit. This isn’t very hard, it’s a simple check to see if something in an account is turned on (it consists of something like “if Client.Overdraft == True, then Charge.Accept”). Instead, they are doing something far more complicated – they are doing some sort of risk modeling, probably based on some sort of statistical learning techniques, to analyze their recent “spending and creditworthiness” metrics. I do not know for sure, but I’m assuming that this is a check to see if the marginal expected loss is less than getting the juicy interchange fee from the transaction.
Since the back end now involves something more complicated, time-consuming (on the order of milliseconds, but still time-consuming), and energy using than a quick check to see if a client’s feature is turned on, this must not pass a cost-benefit analysis on the front end. Think about what that means – American Express thinks if they ask their customers if they want this feature they’ll overwhelmingly say no, so they aren’t going to bother. They’d rather keep the option of letting it go over to grab some interchange fees from time to time.
I see this as a giant gain to markets. We think of markets as giving people choices over what they want, but that also means giving people choices over not having what they don’t want. American Express looks at their massive set of consumer data, surveys, marketing as well as their own experience of the client and the product, and says there is no way people are going to say yes to this option, so let’s not even bother sending out a mailer asking them if they would want it. If the market had previously been working, everyone would naturally have signed up for what they previously had.
This is why I think Obama’s approach is very market friendly. It’s not about dictating what “rights” consumers have over their credit cards, where rights directly refers to specific options and features. It instead is focused on disclosure and choice. Allow credit card companies to decide whether or not they want to offer an option to sign into overcharge fees that customers must sign into. Customers have the choice where they didn’t previously. If customers don’t want to, the credit card companies will have to come up with other features that are worthwhile to consumers. For instance:
Robert Manning, the author of “Credit Card Nation,” said not everyone is so fortunate. Many people are living hand to mouth with just one card, and others have had their credit limits slashed through little or no fault of their own. He thinks some companies might come up with membership programs to serve this population, with new cards that charge annual fees and waive a handful of over-the-limit and late fees each year, no questions asked.
This is what Warren and others mean when they say a CFPA will allow actual innovation to take place. In order to convince consumers that a credit card overdraft charge is worth it, they’ll have to negotiate on other terms – by making these terms clear, and able to actually be chosen by consumers, there will be more variety and more competition among lenders. And that is what markets are all about.
Rortybomb 
I think your analysis is faulty. To approve a *transaction* based on whether the customer has agreed to fee-for-overdraft is very simple.
But the process of signing up the customer is more complicated and expensive. You have to send them a letter providing all the information the law says you must provide. You have to enclose a business reply envelope (unless you want to get only customers willing to buy their own stamp to sign up). So you end up paying for postage both ways. Then you have to hire people to open the returned forms and activate the flag in the database for the correct customer. *Then* you have to keep the form on file in case you are later sued for charging an overdraft to a customer who claims they didn’t sign up. Since your personnel will sometimes make mistakes, you sometimes *will* charge overdrafts to a customer who didn’t actually approve them, so you need a dispute investigation and resolution procedure and a way of revoking erroneous overdraft charges (hopefully before the customer gets pissed off enough to sue you over it). And if any of them sue you anyway, you’ll need to defend the suit (or settle it, particularly if it was your own people who made mistakes).
All of these are solvable problems. But they’ll take a lot more than one line of computer code and the solutions will cost you money. And the cost is per customer – unlike a computer change that you can make once and apply to every account at a marginal cost of zero.
And as a result of that, not all customers will overdraft at all; you only collect fees from the ones who do (and then only from the ones whom you can *collect* the fees from; it does you no good to charge a fee to someone’s account only to be forced to write it off later). Presumably the model helps them catch that factor in parentheses, but because they don’t have to interact with the consumer to approve the overdraft (at no extra charge beyond normal interchange fees and interest), it’s cheaper.
Actually you just have to call them up on the phone. My credit card company is constantly calling me and offering me add ons and services (for ‘free trials’ that require opt out to avoid paying once the trial is over). I’m sure the adoption rate of those services figured heavily into the Am Ex and Disc decisions.
That’s likely not your credit card company, but “partners” to whom they’ve sold your personal information for a profit.
True, but the infrastructure exists and probably has a high number of agreements. If the CC did it themselves they could easily reach a significant audience.
Chris, I’m sorry should have clarified. All of the very relevant stuff you bring up is what I meant by the “front-end” charges. The cost of asking people to adopt a new technology does not outweigh the benefit of it. But that’s fine, as far as it goes – it is what markets are supposed to do.
The costs of processing and storing legal documents may be high on the margins, but that’s how it goes with debt contracts – it should require documentation well stored! – but my suspicion is that they suspect, rationally, that many people will actively not want it, if only to act as a mental accounting mechanism. And I also suspect they want trustworthy consumers to be able to go over so the cc company can collect interchange fees, and consumers may actually want a strict limit enforced, so they’d rather have the option to collect the fee.
I think the revenue opportunity is not so much the interchange on the transaction as the interest income on the actual debt, isn’t it?
I think it’s a matter of the cost of information. Giving overdrafts-for-a-fee to everyone (if “giving” is the right word here) was cheap and profitable, but finding out which customers want them and which don’t is expensive and makes the whole effort unprofitable. (The profitability, of course, depends on the number of customers who *will* want them. But it also depends on the costs to obtain the information to distinguish one from the other, particularly in a way that complies with the law.)
I also agree with dsquared that the interchange fees are probably not the main source of profit: these are *maxed out* customers, i.e. ones who are already carrying a very large balance month-to-month. Interchange fees are practically insignificant compared to the interest – and that is what you get to still receive if your secret formula grants them an automatic credit limit increase. (Subject to the risk of being forced to write off, which is what the secret formula attempts to estimate.)
The overdraft fee was probably *also* insignificant compared to the interest – they just tacked it on because they could – which is why they’re trying to preserve the idea of issuing more credit to selected cardholders, while working within the new restrictions.
Re: interchange being more valuable than interest. I don’t deal with financial algorithms for credit card companies, but if I did here is what I’d consider (and this is my logic for what I wrote above):
You go over your limit by $25. The algorithm has to figure out whether or not it should go through.
Your APR is 20%, so is about 1.52% monthly. I’m assuming the expected interest paid on that $25’s lifetime in the account is half that; it is very likely a majority of consumers will pay it back down immediately (ie within that month) to get under budget, so no interest is paid. Certainly consumer not in distress will likely do that; even those in distress may go pay that $25 off just to get under balance, so the card is useable, to maintain internal mental accounting, etc. I’d be surprised if a sizable percent of consumers had that $25 balance over account limit the entire year.
So that’s 1.52%/2 = .76%. Now we need to adjust that downward for counterparty risk – that the guy with the credit card never pays. I’m sure there’s a lot of statistical learning backflips, but let’s say it’s estimated new expected outcome is like .7%
Interchange is 1.8%, with a third going to consumers in bonus features, so that’s 1.2%. Paid immediately, no discounting for waiting. No counterparty risk as it comes from the business, goes straight into the account.
That’s fantastic compared to an expectation of .7% over the course of some months and payments and hiring someone to threaten if payments aren’t made etc.
If the fee is $50, that changes a lot – suddenly $25 turns into $75 balance, so a giant bonus on principle – plus that gathers interest. So we have to multiple the .7% by 3, scale it back just a bit more because of more counterparty risks and more risks of prepayment, but we are approaching the 2% mark here with overdraft fees. Very, very profitable.
You seem to be assuming that people who are over max will pay back down to max almost immediately *and* that they wouldn’t have used the same funds to pay down their balance if the transaction had been refused.
That seems a bit fishy – if they had the money and inclination to pay down their balance at all, why are they sitting at max? If they’ll pay down an over-max balance, why not a max balance? If they were going to pay down a fixed amount anyway, then anything added to the balance offsets the paydown and you get to count the whole interest from purchase to the remote future when they pay off completely as profit from approving the transaction, because otherwise they would have had that much less balance all along.
Of course, I’m assuming that the interest the CCs make from having you carry a balance exceeds their counterparty risk + cost of capital. Given their business model and practices this seems like a safe bet. It follows that the more balance they can induce you to carry (by any means necessary), the more profit they make.
Consumers not in distress don’t enter into this calculation at all because they don’t have the prerequisite maxed out card.
(I’m assuming no fee – obviously the 200% profit from the fee itself would swamp both the interchange and interest profit opportunities, even with the interest calculated on the tripled amount.)
“You seem to be assuming that people who are over max will pay back down to max almost immediately *and* that they wouldn’t have used the same funds to pay down their balance if the transaction had been refused.”
Yeah, it may be fishy, but I throw in ‘even those in distress may go pay that $25 off just to get under balance, so the card is useable, to maintain internal mental accounting, etc.’
If you are at $4980 on a $5K card, you can still use it to buy $15 worth of gas. Very useful if you are desperate, and something you can’t count on at $5005 on the card – so there’s a certainty piece of mind issue at play, and something you don’t get more of at $4955. And the difference between $4980 and $4955 may be comparably minor if $25 towards the electric bill can make a major difference.
Though rereading what I wrote, you are correct I’m assuming way too high of a prepayment. Particularly ending it at the first month. Let’s assume X% of people pay off that $25 the 0 month (so no interest paid), and X% remaining pay it off after the first month, X% remaining of that after the next month, etc.
Add that up over a year, the percents come to (total no discounting, interest paid on new interest):
X = 75% payoff; the interest collected is .6%
X = 50% payoff; the interest collected is 3%
X = 25% payoff; the interest collected is 16% (!!!)
I can understand why they spend so much money doing research on psychological profiles of debtors.
Right. And people who usually pay off most of their balance don’t max out in the first place unless they’re unemployed or something (in which case they temporarily become people who don’t pay off their balance).
You’re right about the utility of not being maxed out, but it seems a suspiciously reasonable argument to be imputing to someone who carries around an almost maxed-out credit card. Unlike credit card companies, though, and possibly unlike you, I haven’t really studied the psychological profile of people who carry large credit card balances. But given the terms of credit cards, they almost have to be irrational about *something*.
> This isn’t very hard, it’s a simple check to see if something in an account is turned on (it consists of something like “if Client.Overdraft == True, then Charge.Accept”).
Uh, have you worked on critical software at large companies before?
I suppose not, or you wouldn’t have written this. You’re off by a factor of five or six orders of magnitude on what implementation would involve is my best guess.
> The costs of processing and storing legal documents may be high on the margins, but that’s how it goes with debt contracts – it should require documentation well stored!
There is a difference between processing and storing legal documents as text on one hand and putting the documents into databases that can be used by programs on the other hand. Not at all the same thing.
My opinion — this is a card feature that people won’t want until they want it, and then they will be screaming about not having it. (It’s similar to overdraft protection on a checking account.) Though I’m sure the fees are too high and that’s a problem.