Roseville, CA · NMLS 281056 · Licensed CA, TN, TX (916) 705-2557 adebusk@loandepot.com
← Research homeRestoring the American DreamDownload PDF ↓

Chapter 8 of 12

Jobs Reports and Mortgage Rates: What’s the Connection?

Separate what economic-data revisions show from what they cannot establish about borrower costs.

Jump to a page on the report 1-59 + 37A-37C

The online edition groups several PDF pages into each chapter. Select a page to open its place in the web edition. Pages 1 and 2 correspond to the cover and contents on the research home page. The revised PDF adds Chapter 7 pages 37A-37C while retaining the original page numbers.

Who Does What: Jobs Data, the Fed, and Mortgage Markets

The Bureau of Labor Statistics produces employment estimates. The Federal Reserve sets monetary policy. Mortgage lenders and financial markets price expectations about inflation, growth, policy, Treasury yields, mortgage-backed securities, and other risks. These institutions interact, but they do not perform the same function.

Preliminary payroll estimates are revised as more employer reports arrive and are later benchmarked to more comprehensive administrative records. A benchmark correction can change the picture of the economy that was available in real time. It does not by itself show that officials deliberately manipulated the data or that a particular borrower was charged an identifiable excess rate.

How the Jobs Numbers Changed After Review

Compare Jobs Revisions Using the Same Definitions

Data table · Original report p. 38
March yearPreliminary NSAFinal NSA comparisonFinal SA level
2021-166-7+374
2022+462+506+568
2023-306-187-266
2024-818-598-589
2025-911-861-898
2026-79Not finalNot final

Thousands of jobs. NSA means not seasonally adjusted; SA means seasonally adjusted. Final NSA comparison removes scope changes. March 2025 headline NSA revision is -862,000. March 2026 remains preliminary. [9, 10, 86]

The table uses March benchmark years rather than publication years. Its final unadjusted comparison removes scope changes, following BLS’s historical technical table. For March 2025, the headline unadjusted level revision was minus 862,000, while the historical comparison was minus 861,000 after a 1,000-job reconstruction adjustment. The seasonally adjusted level revision was minus 898,000. All three figures can be correct when their definitions are stated. [9, 10]

The record includes an upward revision in 2022 and downward revisions in 2023 through 2025. It does not support describing each of those three years as a 0.4% to 0.6% overstatement of annual job growth. A benchmark percentage concerns the employment level at a point in time, not a percentage of that year’s job growth. The March 2026 preliminary adjustment was minus 79,000, with a final benchmark scheduled for February 2027. [9, 10]

What the Research Says About Market Reactions

George Perry’s study of releases from 1994 through mid-2004 found that payroll surprises explained a substantial share of Treasury yield changes in a narrow announcement window. The surprise was the difference between the initial release and the forecast available beforehand. It was not the eventual benchmark revision. The same paper found no consistent effects for the two-and ten-year maturities over a full-day window, an important limit on claims about persistence.

[11]

More recent San Francisco Fed research, published in February 2025, examines rolling responses to payroll and inflation surprises from 2014 through January 2025. It finds that Treasury sensitivity to payroll news increased sharply in early 2024. That evidence supports the transmission mechanism and shows why a single constant response coefficient is inadequate. It does not identify a counterfactual mortgage-rate path. [12]

The strongest supported conclusion is that unexpected employment news can affect market yields. Determining whether inaccurate initial data caused a lasting mortgage-rate premium requires further identification. Markets also respond to inflation, wage data, Federal Reserve communications, Treasury supply, and changes in risk premiums. Lender hedging and mortgage spreads add another layer between a Treasury price move and a borrower’s lock.

Can We Tell What Data Revisions Cost a Borrower?

The sharp mortgage-rate rise in 2022 coincided with inflation and monetary tightening, while that year’s March employment benchmark was later revised upward. This timing weakens an explanation that attributes the entire rate increase to payroll overstatement. It does not rule out incremental effects on particular release days.

A defensible causal estimate would separately establish the unexpected information, the yield response, its persistence, the pass-through to mortgage quotes, and the number and duration of affected loans. Treating final revised data as if traders possessed them at the initial release would introduce look-ahead bias. Multiplying a temporary yield change by 360 payments would also exaggerate losses for loans refinanced or repaid earlier.

The payment table in Section 5 is consequently a sensitivity calculation. The paper does not assign its half-point scenario to BLS errors. The policy response supported by the evidence is stronger data collection, transparent vintage comparisons, and clear explanation of uncertainty while preserving statistical independence.

Prefer the original layout? Open this section in the PDF →