Wholesale mortgage prices are those delivered by wholesale lenders to their clients, mortgage brokers and smaller ("correspondent") lenders. Wholesale lenders do not have any of the infrastructure necessary to deal with borrowers. The clients mark up these prices to deliver retail prices to borrowers.

Source: Twelve large wholesale mortgage lenders, who post prices daily to mortgage brokers and correspondent lenders. The data are provided by Amerisave, a correspondent lender. The price shown is always the lowest of those posted by the 12 lenders.

Timing: Daily, showing the state of the market on the date shown. Some of the tables using data are compiled only weekly, see below.

Coverage: Everything but sub-prime.

Instrument Coverage: 6 fixed-rate mortgages, and 8 adjustable rate mortgages.

Geographical Coverage: Each of 48 states and US average. Regions could be developed if needed.

Transaction Characteristics: For any of the 14 types of loans, prices can be obtained for any combination of the following variables: Loan amount, property value, type of property, state location, purpose of loan, use of property, borrower’s FICO score, type of documentation, and whether borrower escrows taxes and insurance. All prices are for a 30-day lock period on refinances, 60 days on purchases.

Price Quotes: The price tables and charts do not show rate and points. Rather, they show a rate at zero points. This rate is interpolated between the rate just above and the rate just below zero points. Wholesale points include the small amount of fixed-dollar fees that each individual wholesale lender charges.

Availability: Coverage begins on a daily base on September 6, 2007. However, we have data for May 4, June 11 and August 3 for some of the tables, and these can be accessed along with the most recent data.

I had three purposes in developing these data:

1. Provide an accurate measure of day-to-day changes in the market as a whole. For this purpose data are required daily for a wide variety of loan types.

2. Provide accurate measures of how, at any one time, mortgage rates vary with various features of the loan transaction. These features include loan size, FICO score of borrower, down payment, type of documentation provided, and loan purpose. For this purpose data for one day a week suffice, because these "structural" features of the market ordinarily do not change from day to day.

3. Provide the means by which an individual borrower can find the wholesale price of his loan. For this purpose, a pricing engine is required that adjusts the price to the individual features of the transaction.

The wholesaler price is a critical input to Equalizer, the professor’s mortgage shopping tool, which is in the works.

These three purposes will be discussed in turn.

Sensitivity: Wholesale price data are a much better gauge of market conditions than retail price data because they contain much less statistical "noise". The data are offer prices on the day indicated. Retail price series, such as the one published by Freddie Mac, are weekly averages and the timing used by reporting retail lenders is not clear. The lenders could be reporting offer prices, prices on locked loans, or prices on closed loans.

There is also much less price dispersion in the wholesale market because wholesale prices are net of retail markups which can vary widely from one transaction to another.

The ARM features that affect future rate adjustments are also well defined. The margin on all the ARMs is 2.25%. The index is one-year Libor on the fully-amortizing ARMs, 6-months Libor on the interest-only versions. Maximum rates are 5% or 6% above the initial rate. The initial rate adjustment cap is 5% except for the 3-year ARM on which it is 2%. Periodic rate adjustment caps after the initial rate period are all 2%.

Tables and Charts: The user can access these data in one table showing all 14 programs, or on three charts where the programs are divided into 3 groups. The data are available daily but also combined into weekly and monthly averages. The averages will become more useful after some time has elapsed.

Go to Tables and Charts and select those available daily.

Five tables available weekly show how rates on 30-year fixed-rate mortgages vary with FICO score, loan size, loan purpose and type of property, type of documentation, and down payment. The major purpose is educational, but these tables also measure the occasional changes that occur in how the market appraises risk.

Such a change, in fact, occurred while these tables were being developed. A trial run was done on May 4, 2007, which was before the full eruption of the sub-prime crisis. While the May 4 data covered California rather than the US, the differences between California and the US average are very small.

Comparing the data for May 4 with that for September 21, we see a marked increase in risk premiums. The difference in rate between a $417,000 loan which is saleable to Fannie Mae and Freddie Mac, and a $418,000 loan, which is not, rose from 0.278% to 0.745%. The difference between a full documentation and a no-documentation loan rose from .525% to 1.022%. And the difference between a 740 FICO score and a 620 score rose from 0.30% on May 4 to 1.37% on September 14. A week later, the 620 was not available.

A caveat is that the non-conforming data are not as reliable as the conforming data because the number of lenders quoting prices declines in the riskier niches.

Go to Tables and Charts and select from those available weekly.

Under Construction