Sunday, September 16, 2007

Risk Based Pricing for UberNerds

by Tanta on 9/16/2007 10:41:00 AM

A new paper by Federal Reserve Board analysts I highlighted the other day contains some discussion of the issues of loan pricing and fairness to consumers. I encourage anyone who is interested in this issue to read the entire paper. However, I promised to write something UberNerdy about loan pricing, and this bit from the Fed paper provides me with a place to start:

As price flexibility has emerged in the mortgage market, so have concerns about the fairness of pricing outcomes. Such concerns generally fall into four broad categories. First are concerns about possible discrimination based on the race or ethnicity of the borrower. Such concerns are heightened because loan prices are not always determined strictly on the basis of credit risk or cost factors but can involve elements of discretion, in which loan officers or loan brokers may seek prices that differ from those on rate sheets or other techniques used by lenders to establish baseline prices.

Second are concerns about whether borrowers in the higher-priced segment of the loan market are sufficiently informed and whether they are willing or able to shop effectively for the loan terms most appropriate to their circumstances. For example, it may be difficult for borrowers to determine where they fit along the credit-risk spectrum.
For the moment I am going to largely ignore the first issue, of discrimination based on race or ethnicity. This is not because I don’t think it’s important; I do. But I don’t think we can really get there, in terms of understanding how pricing of a loan can be manipulated at the primary market level (in discriminatory or just equal-opportunity predatory ways) without getting a grip on how it works. What I will propose is that, indeed, it is very “difficult for borrowers to determine where they fit along the credit-risk spectrum.” Not only do borrowers, on the whole, lack the ability to size up their own risk, they don’t know how lenders price that risk. As long as information about “the market price” for certain risk factors is non-public, the public will not know whether the price it gets is the best on offer or not.

It is supposed to be the role of a mortgage broker to locate the best price for a consumer, since the idea is that the broker receives rate sheets from many wholesale lenders, and can choose the best-priced one on any given day or for any given kind of loan. This presumes that the broker is mostly motivated to offer a “competitive” rate/price to the borrower, not to maximize its own compensation by offering a worse-than-market rate/price to the consumer in exchange for higher fees from the wholesaler. The standard riposte of the brokers is that this kind of gouging can’t really happen, because the customers will “shop around” and know whether or not they’re getting the best deal. Or, at least, they should shop around, and if they don’t, it’s their own fault.

As I have argued before, this creates an odd conception of the broker’s role: the broker with access to all those wholesalers is supposed to be “shopping around” for you. If you go to another broker to get a “comparison” quote, you’re going to someone else who is “shopping” the same universe of wholesalers your original broker was, as a rule. So, in practical terms, what you would be “shopping” for here is differences in broker “markup” practices, not “best market rate.” Why you would pay a broker to “shop” for you and then do your own “shopping” is one of those things that beats me.

That’s my whole argument about “fiduciary” responsibilities in a nutshell: if the broker isn’t obligated to give you the best rate out there, what’s the point of using a broker? If you’re going to do your own shopping, why not shop a couple of different retail loan officers? Your loan is going to end up with the wholesaler anyway; you might as well skip dealing with someone who may not be around if problems ensue, or may not have regulators breathing down its neck about origination practices as a depository lender will (relatively speaking).

But in reality the whole “shop around” business is nearly meaningless when we look at this “risk based pricing” thing. You might know, from reading blogs or something, that, say, the Freddie Mac conforming fixed rate national average loan last week involved a rate of around 6.375% and 1.00% in points. Do you have any idea if you qualify for that? Are you “average”? Is the kind of loan you want “average”? If you were quoted a higher rate than this, would that mean that you are riskier than “average”? Says who?

It’s not easy to find lots of wholesale rate sheets on the web to do comparisons, because most wholesalers put them behind registration walls to keep people like us out. There are, however, a few exceptions, and I found this one. Please understand that I am not “picking on” this lender because I have any particular beef with Chevy Chase, although I will say this “Cashflow Monthly ARM” you encounter on the first page of the rate sheet is perfectly nauseating. Other lenders have equally or probably even worse products, of course, but CCB is dumb enough to make its rate sheet publically available.

Whilst we are on this subject, notice the verbiage at the bottom of the page: “These rates are solely for the use of mortgage brokers, correspondent lenders, and other arrangers of credit and are not to be distributed to potential loan applicants.” All rate sheets have words to that effect on them. You are not given a copy of a rate sheet like this, if you are a loan applicant, and invited to price your own loan. The better reason for that is that you need a “professional” to assist you in this complex process (i.e., we acknowledge you do not understand “where you fit along the credit risk spectrum”). The less better reason is that if you can’t see the rate sheet, you won’t know if you got a higher rate than the best one you qualified for in order to increase the broker’s compensation. In any case, if CCB doesn’t want to see some blogger go to town with its wholesale rate sheet, CCB can invest in a better (more protected) web portal.

I am not, in fact, going to get anywhere near that “Cashflow Monthly ARM” today. We’re just going to look at the pricing for a plain old vanilla conforming fixed rate. If that makes your eyes cross, then do think about how lost in the weeds people are over these toxic ARMs. If you get the hang of the fixed rate pricing, you can play around with how to determine the price to the customer on the goofy ARM. So we’re going to look at page 5 of the rate sheet (page 6 of the "All Other Property States" pdf), which you might want to print if you’re following along at home, although you will need a magnifying glass to read it. Please note that if you’re following the link, the discussion below refers to the “all other states” rate sheet dated 9/14/2007 at 10:00 a.m. If you are now seeing different numbers, you’re looking at an updated rate sheet.

The first thing you see is a rate/price matrix. Eventually you will see that CCB doesn’t use the word “price.” It calls this “premium.” If you know anything about bond pricing, this will drive you crazy right off the bat. If you don’t know anything about bond pricing, it’ll just keep you very confused. “Everybody” knows (uh huh) about this thing called “discount points,” which are a fee, paid at closing, expressed as a percentage of the loan amount. You pay discount points in order to get a lower interest rate (from the lender’s perspective, the points bring the yield on the loan back to up to market).

“Premium points” would be, logically, money the lender pays you to take a higher rate. Now, lenders don’t hand out premium points in cash to anyone. If you the consumer get “paid” a premium, what you are getting is a credit on the final settlement statement against your closing costs (such as your credit report and appraisal fees, title fees, etc.). The “no cost” loan works on premium pricing; “no cost” just means “no cash outlay,” because the costs are there but paid for with premium. However, you do not always get that premium: it can be paid to the broker, not to you. We call this “back-end points” or “yield spread premium” (YSP).

So, anyway, CCB calls everything “premium” instead of “price," which in itself tells you something about the mindset here. I’m going to keep talking “price,” because I have a point to make. Another thing you bond-people will notice immediately is that the prices on this rate sheet are expressed in a “retail” format, not a dollar price (or “buy price”) format. This varies in the industry. Most correspondent rate sheets (remember, that’s a lender buying a closed loan from another lender) use dollar prices, and some wholesale (broker) rate sheets do. (Dollar prices are things like “par” or 100.00, 101.00, or 99.00, which numbers mean “percent of face value” of the bond or mortgage. The equivalent “retail” or consumer price would be 0.00, (1.00), or 1.00, respectively. Subtract the retail price from 100 to get the dollar price, and remember to change the sign on the adjustments. If that last sentence confused you, ignore it.)

That means that CCB’s rate sheet expresses a premium price as a negative number and a discount price as a positive number. You may also notice that CCB quotes 15-day locks in rounded ticks (1/32 increments) and 45-60 day locks in even eighths. This means that at any given rate, the borrower pays around 0.187 to go from 15 to 45 days, and .125 to go from 45 to 60 days.

Most of the “risk based pricing adjustments” on this rate sheet are also quoted in even eighths, although you’ll notice that the LMPI and Expanded Approval (EA) adjustments are not necessarily expressed in eighths. I will observe that: eighth increments are “traditional” in primary market pricing, and when you see pricing in other increments, you are likely to be seeing pricing that was derived from a much more exact model. The LPMI adjustments are based on the actual cost to the lender of mortgage insurance policies; the EA adjustments are based on the guarantee fee or loan-level pricing adjustments Fannie Mae comes up with in its Desktop Underwriter AUS. The rest of this stuff, my consumer friends, is ballpark. I know; I’ve been in the meetings. For years. We’ve been charging a quarter of a percent for escrow waivers since before Windows was copyrighted. If you tell me that number comes from some fancy cutting-edge servicing valuation model that looks at exact current float costs and some razor-sharp analysis of marginal credit risk differences, I will laugh in your adorable nerdly little face. I’m sure we had some data when we first made that one up, but we used a sextant, not LORAN and certainly not GPS.

In order to see how this works, let us imagine that Tanta wants a loan: a $140,000 cash-out refi. Tanta’s LTV is 90%, the property is her principal residence, a 1-unit home, and she is providing full documentation of income. However, since she has no idea where her tax returns are hiding, she needs a 45-day lock. Her FICO is 655. Because her brother-in-law’s boss’s Avon Lady’s financial advisor said it had “tax advantages,” she asks for an interest-only loan with “no MI” (that is, LPMI). On the other hand, Tanta read on some website that you should never pay points, so she wants the “no point” rate. You may reflect on how far Tanta and her loan request are or are not “average.”

You can see right now that you have to supply a lot of information to a broker or loan officer these days to get a simple rate quote. Actually, you can’t even get a real rate quote without someone running a credit report on you, because you do not know your own FICO (and even if you think you do, your lender will get that information directly from the credit bureaus anyway). You are already handing over your Social Security Number and incurring cost to someone who will want to recoup it by making a loan, even if all you thought you were doing was “comparison shopping.” And we’re just guessing on LTV at this point; that value isn’t conclusive until some sort of appraisal or AVM determines the value part. But you’ll be in this process fairly deep by the time that happens. Remember that this is a refi request: do you know, really, what your house is worth today? Does Tanta?

So how do we go about quoting a rate/price here? Well, what your broker is likely to do is first add up all the price adjustments you would be subject to. Tanta’s loan gets the following (COR = cash out refi):

Loan amount: 0.125
FICO: 0.750
COR: 0.750

Total: 1.625

Since Tanta doesn’t want to pay points, we must find a 45-day rate that shows premium of at least (1.625). Let’s take that 7.375 rate: it pays exactly (1.625) in premium. So Tanta’s base rate is 7.375: we add 1.625 to (1.625) to get zero points. However, Tanta has some rate adjustments in store:

Interest Only: 0.25
LPMI LTV: 0.30
LPMI COR: 0.10

Total: 0.75

Therefore, Tanta gets a rate of 8.125% (base 7.375 plus 0.75 adjustments) at zero points. Now, the trouble here is that the broker still has to make some money for going to all the trouble of taking Tanta’s loan application, so Tanta is highly likely to pay one or more “origination points.” Of the many things that makes Tanta a crazy person, indiscriminate use of the term “points” is one of them. Traditionally, the “origination fee” on a loan is the lender’s overhead, which includes commission to the loan officer or profit to the broker. Because it was traditionally expressed as a percentage of the loan amount, it is referred to as a “point,” but it is very important not to confuse it with a discount point: it does not “buy down” the interest rate.

Brokers (or lenders) don’t have to charge origination points; you can and people do just throw in a bunch of flat fees for this and that which end up being profit to someone. I bring this up in part to highlight a real problem in “predatory pricing” land, which is the concept of the “bona fide discount point.” Things have gotten so bad that we actually have to use that term, because you see loans where a borrower got charged one origination point and one “discount” point (they show up separately on the disclosures and settlement statement), but the rate didn’t get discounted appropriately. A whole lot of brokers seem to think that you can charge “discount points” without reducing a premium rate. A whole lot of consumers can never know whether this is happening or not, since, of course, the consumer doesn’t see this rate sheet.

You will notice that, in our example, we picked 7.375 as the “base rate” because we are an honest broker who will get compensation for this loan on the “front end,” by charging some kind of origination fee to Tanta. However, we could have chosen 7.50 as our base rate, which paid (1.750) in premium. If we had done that, after our price adjustments, there would be 0.125% of the loan amount to end up in someone’s pocket. If the broker applies it as a closing credit, Tanta’s closing costs will be reduced by that amount (she would, say, pay a 0.875 origination fee instead of a 1.00 fee). If the broker doesn’t give it to Tanta, it becomes YSP or compensation to the broker. If Tanta doesn’t know that 0.125 is on the table someplace, Tanta doesn’t know whether she should be paying 0.875 or 1.00 in origination fee (or the equivalent in flat fees). She also probably doesn’t know that it doesn’t have to be on the table, because she could have gotten that 7.375 base rate.

Does Tanta have any idea whether that 0.75 add-on for LPMI compares favorably to paying MI herself? Not unless somebody runs some “scenarios” and gives her the figures, carefully explaining the advantages and drawbacks of LPMI. Does she know whether CCB’s add-ons for LPMI are going to be the same at any other lender or not?

Does she understand that the biggest impacts on the price she was offered were her FICO and the fact that she’s taking cash out at a high LTV for cash-outs? If the amount of cash she’s taking is rather modest, does she know that it might not be worth it, considering that she might be able to save up that modest amount of cash in a fairly short time by doing a rate/term refi at a lower rate and therefore lower payment? On the other hand, does she realize that if she reduced her loan amount significantly, her loan amount adjustment would increase?

Does she know she’d get a better rate/price with 5 more FICO points? Is there a way for her to manipulate her FICO in the short term to squeeze 5 more points in? Is there a “credit counseling” company who would be willing to extract some fee from her for assistance in this matter? Will that fee be worth the rate/price break on a 660+ FICO loan?

If this whole rate quote blows up when the appraisal comes back, showing that unfortunately our LTV is over 90%, what will happen? Will the broker go back to Tanta and tell her she has to borrow less, which reduces any of the broker’s compensation that takes the form of a percentage of loan amount, or will the broker lean on the appraiser until a “better” number comes back (or fool around with the inputs on an AVM until it complies)? In the latter case, will Tanta even know that that is going on?

An important thing to bear in mind is that we just looked at one rate sheet by one wholesaler. For some if not most of you, this will be the only exposure to this sort of thing you’ve ever had. If you are now thinking something along the lines of, “OK, so 75 bps is a normal market price adjustment for a FICO in the 620-659 range,” you are making a logical error. It would be a profound logical error if you then assumed that a different wholesaler’s rate sheet with a higher or lower FICO adjustment were “overcharging” or “undercharging.”

The fact is, it depends on how you calculate that base price up in the rate/lock days matrix. A competitor of CCB’s could easily calculate a base price that is better, relative to CCB’s, by 25 bps. That competitor’s FICO adjustments could, then, be worsened by 25 bps, so the FICO adjustment for 620-659 might be 100 bps, not 75 bps, although the end price to the consumer is the same. Even those things that are fairly consistent throughout the industry—like the classic 25 bps adjustment for escrow waivers—can confuse the unwary: some lenders calculate the base price assuming that all loans have escrows, and then worsen the price by 25 bps if escrows are waived. Some lenders calculate the base price assuming all escrows are waived, and improve it by 25 bps if escrows are established. You might see the same value but different signs on different rate sheets.

A casual comparison of price adjustments across even a large number of wholesaler rate sheets will not tell you what the “going market price” of a given risk factor is; you have to analyze the base prices together with the adjustments to get that. “You” in this case are a broker or a loan officer; “you” are unlikely to be a consumer.

You must also remember that risk adjustments are tailored to the guidelines of the loan program being priced. Why is there no adjustment for cash-outs over 90%? This loan program does not allow an LTV that high, so it does not price one. If you wandered over to an Alt-A or subprime rate sheet, you would see additional risk adjustments because crazier things are allowed; you might also find very different values in those adjustments, because the “base price” is set very differently.

I’m guessing by now that you all have spotted the trouble with the idea that risk based pricing is “individualized”: it is, but what an individual loan gets are adjustments based on average performance of loans of that type. How reliable those calculations are will be a matter, among other things, of how a pricing model considers variables singly or in conjunction. In other words, these adjustments you see may be “net of” a lot of factors that aren’t obvious to consumers.

Let me observe that I purposely picked an example loan with a lot of risk-based adjustments on it, so you could see how the process works. But if you hypothesize a loan with fewer adjustments, you can see that a broker could end up with a lot more than 0.125 in YSP off of this CCB rate sheet (and there are, or at least were until recently, wholesale rate sheets out there paying a lot more premium—quoting much higher rates—than CCB’s. We are looking at a “post-turmoil” rate sheet.). It may seem counterintuitive to you, but quite often it is the borrowers with the best credit history and the most conservative loan terms who are most at risk for getting a high interest rate—as long as they stay ignorant about what “YSP” is and why it is showing up as a charge on their settlement statement.

It is perfectly true that a similar mechanism works in retail loan origination; in that context this additional premium is generally referred to as “overage” rather than “yield spread,” and it can, depending on the lender’s practices, be additional profit to the lender or additional commission to the loan officer or (frequently) a split. However, it is even less visible in a retail environment, since "overage" isn't disclosed on the settlement statement the way YSP is. On the other hand, depository retail lenders (so far) face a great deal more regulatory pressure to keep "overages" under control than brokers do.

But to get back to the question raised by the Fed analysts, does every borrower with good credit and conservative loan terms think of him or herself that way? Does the rest of the world, or a vocal subset of it, think that way? How widespread is the belief that minority and low-income people are almost always “subprime” borrowers? If you can make a 110-pound adult believe she’s “fat,” or a college student who gets a C believe he’s “stupid,” can you convince someone who makes $30,000 a year fixing cars and happens to have a Hispanic surname that a 655 FICO makes you “subprime”?

I think you can. If I tried to link to every media article that defines “subprime” as “loans to low-income people,” I would blow the server. A loan to a low-income person that the person cannot, patently, afford is certainly subprime. That is not the same thing as saying that all low-income people are “subprime credits,” but how often do you hear that distinction being made?

How often are prime-credit borrowers given subprime loans? Good question, and I am not making or endorsing a particular claim about an empirical matter for which I don’t have satisfactory data (see the Fed's analysis for the difficulties in sorting that out). I am pointing out that we have the necessary if not always the sufficient conditions for predation when we have “risk based pricing” that is opaque to consumers, and consumers who are not educated about what constitutes “risk.” We are not exactly making it hard for discrimination or predation to occur here.

We also have an industry which hasn’t done a particularly fine job lately proving that it really knows how to price risk anyway. Do price adjustments (of any amount) for FICOs “make sense”? Are they enough? Too much? Beside the point? Are you sufficiently convinced of the predictive power of a FICO score to want to justify pricing a loan on that basis? Particularly when an entire sub-industry of various forms of more or less “respectable” FICO manipulation has grown up around this practice? In the midst of all of that, are you confident that putting the onus on consumers to “shop around” still makes the most sense as a “solution” to the problem of pricing distortions? I’m not.