Missing HMDA Interest Rates: Avoid a Misleading Average

A missing historical interest rate should not become zero in your spreadsheet. Doing that can make a mortgage selection look much cheaper than the observed rates support. Keep numeric rates, missing values and source-specific nonnumeric labels separate before calculating any summary.

Show the count behind the rate summary

FinToku’s explorer preserves unavailable and exempt-style labels instead of turning every field into a number. An average calculated from only numeric values describes that subset, not all selected records. The result also mixes loan characteristics unless you control the action, purpose and other available filters. Historical note rates are not current APR quotes.

Worked example

Suppose five invented records show 5%, 6%, 7%, ‘Not reported’ and ‘Exempt.’ The mean of the three numeric rates is (5 + 6 + 7) ÷ 3 = 6%. Replacing the other two values with zero gives 18 ÷ 5 = 3.6%, an artificial reduction. Report the valid numeric count of three and the excluded count of two. Even the 6% subset average does not describe a currently available product.

This example is illustrative, not a finding about a named borrower, company or property.

A practical research workflow

Create a numeric-rate flag and retain the original displayed value beside it. Calculate a summary only after defining the eligible record selection. Report how many records contributed and how many did not. If comparing counties or institutions, use the same treatment on both sides and review whether different missing-value rates could affect the interpretation.

Checks before you use the result

• Keep the original rate text alongside converted numbers.
• Never replace missing or exempt labels with zero.
• Report the numeric observation count with every average.
• Keep note-rate research separate from current APR shopping.

What this evidence cannot establish

An arithmetic average is not automatically a typical borrower experience. A different product mix, term mix or reporting pattern can change it. Without the appropriate controls and data definitions, a precise percentage can conceal an imprecise comparison.

Questions readers ask

Can I drop every record without a rate?

You can define a numeric-rate subset for a specific calculation, but disclose the exclusion and keep counts of the removed categories. Do not describe the subset as the complete dataset.

Is the median safer than the average?

A median can reduce the effect of extreme numeric values, but it does not repair missingness or a mixed selection. Both statistics require a clearly defined population.

Continue your research

• Explore the relevant directory
• Browse the free loan resource library
• Find more loan research guides
• Why County Mortgage Volume Does Not Predict Your Loan Rate

Sources and scope

• Official public HMDA field definitions
• 2025 source release and snapshot context

Prepared November 7, 2026 using the linked public source context and directory methodology. Historical source periods are identified in each directory. Current offers, individual account facts and program availability require separate verification.

Explore this topic further

Check your assumptions with these tools and related guides.

Try a relevant calculator

• Mortgage & Home Loan — Principal, interest, taxes and insurance in one monthly number.
• Loan Comparator — Put two or more loan offers side by side and see which costs less.

Read related guides

• Credit Score Range vs Percentile: Avoid a Misleading Loan Chart
• Missing HMDA County Records: Check Coverage Before Comparing
• Car Payment Cleared the Bank but Is Missing From the Loan

Further reading

CFPB guide to reading a Loan Estimate