The simple version

Every Wednesday morning a number comes out counting how many people applied for a mortgage the week before. It moves markets a little and it produces headlines about the housing market.

It counts applications. Not loans approved, not houses bought, not money that changed hands. Somebody filled out a form, and that is the entire event being measured.

The numbers

• Mortgage applications decreased 1.5% from one week earlier on a seasonally adjusted basis, for the week ending September 18, 2026 (Mortgage Bankers Association (MBA), Weekly Applications Survey, released September 23, 2026)
• On an unadjusted basis the same index increased 9% compared with the previous week (MBA, Weekly Applications Survey)
• The Refinance Index decreased 3% from the previous week and was 62% lower than the same week one year ago (MBA, Weekly Applications Survey)
• The seasonally adjusted Purchase Index decreased 1% from one week earlier, and the unadjusted Purchase Index was 11% lower than the same week one year ago (MBA, Weekly Applications Survey)
• The refinance share of applications was 39.3%, down from 39.4% the previous week, and the average contract rate on a 30-year fixed mortgage with conforming balances of $832,750 or less rose to 7.12% from 6.97% (MBA, Weekly Applications Survey)
• The survey covers closed-end residential mortgage applications originated through retail and consumer direct channels, has been conducted weekly since 1990, and its respondents are mortgage bankers, commercial banks, thrifts, and credit unions (MBA, Weekly Applications Survey)
• Under the Home Mortgage Disclosure Act (HMDA), a lender reports one of eight outcomes for an application, including loan originated, application denied, application withdrawn by the applicant, and file closed for incompleteness (Consumer Financial Protection Bureau, HMDA loan-level data fields)

An application is an intention, not a transaction

Applying for a mortgage costs a person an afternoon and some paperwork. It commits nobody to anything. That is why the count reacts quickly when rates move, and why it is the earliest housing signal available each week.

It is also why not everything counted here becomes a loan. The federal reporting standard makes that explicit: when a lender reports what happened to an application, loan originated is one of eight possible answers.

The others include denied, approved but not accepted, withdrawn by the applicant, and file closed for incompleteness. How often each of those happens is a separate question this article does not answer, and the weekly index does not track outcomes at all.

Two different populations in one headline

The index blends two groups with little in common. People buying a home, and people refinancing a loan on a home they already own.

Refinancing responds to rates far more directly, because the decision is arithmetic rather than life circumstance. A rate move can swing refinance applications hard in a single week while purchase applications barely move.

So a headline about mortgage applications can describe a week in which the number of people trying to buy a house hardly changed. The purchase index and the refinance index are published separately for exactly this reason.

The same week, down 1.5% and up 9%

This week's release shows the trap plainly. The composite index fell 1.5% seasonally adjusted and rose 9% unadjusted, in the same week, from the same responses.

Seasonal adjustment strips out patterns that repeat at the same time every year, so the adjusted figure answers a different question from the raw count. The release notes that the prior week's results included an adjustment for the Labor Day holiday.

Neither number is wrong. They answer different questions, and a headline usually quotes one of them without saying which.

The Real Cost lens on a headline percentage

Here is how a large total move can hide a flat market. The split below is round numbers we chose to show the arithmetic.

• Start with a total index of 100, made up of 60 purchase applications and 40 refinance applications.
• Refinance applications rise 20%, taking that piece from 40 to 48.
• Purchase applications do not move at all, staying at 60.
• The new total is 108, so the headline reads that applications rose 8%, while purchase activity changed by 0%.

The 60 and 40 split is ours, chosen to make the arithmetic legible, and it is not this week's composition or any week's. The lesson holds at any split. When the total moves, the first question is which piece moved.

What this means

The application index is early, weekly, and jumpy, which is a reasonable trade if you know what you are getting. It tells you how people responded to last week's rates faster than any other housing number.

It does not tell you how many homes sold. Those counts come later and from different organizations, and how they decide when a sale happened is its own subject, covered separately.

What this is NOT

This article does not forecast mortgage rates, home sales, or application volumes, and it does not advise anyone on when to apply for a mortgage, when to refinance, or whether to lock a rate. It takes no position on whether the housing market is strong, weak, recovering, or cooling, and it recommends no lender, broker, or loan product. It makes no claim about how many applications become loans or how often any particular outcome occurs, because the weekly survey does not measure outcomes and this article cites no rate for them. The 60 and 40 split in the Real Cost section is an illustration we chose to show how a blended index behaves, not this week's composition. The survey described here is conducted by an industry association rather than a government agency, so it reflects its own respondents rather than a complete count of the market, and its index values and methodology are the association's own.

Sources

• Mortgage Bankers Association, Mortgage Applications Decrease in Latest MBA Weekly Survey, September 23, 2026: https://www.mba.org/news-and-research/newsroom/news/2026/09/23/mortgage-applications-decrease-in-latest-mba-weekly-survey
• Consumer Financial Protection Bureau, Home Mortgage Disclosure Act loan-level data fields, action taken: https://ffiec.cfpb.gov/documentation/publications/loan-level-datasets/lar-data-fields