The big Freddie study is based on "tracts" not "appraiser defined neighborhoods." I work in a county that doesn't have census tracts to the 0.XX. Take Marin City, which isn't a 0.xx census tract,
Analyzing a decade of change
data.sj-r.com
It is described by Wiki:
Marin City is an unincorporated community and census-designated place in Marin County, California, United States. It is located 1.5 miles northwest of downtown Sausalito, 8 miles south-southeast of San Rafael, and about 5 miles north of San Francisco from the Golden Gate Bridge, at an elevation of 23 feet. Marin City was developed for housing starting in 1942, to accommodate wartime shipyard workers and other migrants to California. Among those were African Americans from the South in the Great Migration, which continued until 1970.
After the war, the area became predominantly African-American, as white residents were able to move freely to private housing elsewhere in Marin County. Since the 1980s, additional development has changed the makeup of the population while providing more local jobs. In 2018, Marin City's socioeconomic and racial makeup contrasts with the mostly wealthy and White population in Marin County overall. Wikipedia
Is Wiki a source of discrimination?
I doubt an appraiser today would be permitted to to include the narrative in bold. Based on Marin City's census tract it is considered a "diverse" neighborhood, with an index of 82, it has slightly more white than black:
* The USA TODAY diversity index shows on a scale of 0 to 100 how likely it is two people from an area would have a different race or ethnicity. A score of 0 would mean everyone had the same race and ethnicity; a score of 100 would mean everyone in an area had a distinctly different combination of race and ethnicity. Nearly everywhere is some place in the middle. The index was invented in 1991 by Phil Meyer of the University of North Carolina and Shawn McIntosh, who was then with USA TODAY. This score differs from the Census Bureau's version of the diversity index because of differences in how the bureau's formula counts race and Hispanic origin.
But its census tract is different than the other communities in Marin County:
Analyzing a decade of change
data.sj-r.com
The FHFA study is based on:
"What We Observed
From millions of appraisals submitted annually, a keyword search resulted in thousands of potential race-related flags. Individual review finds many instances of keywords to be false positives, but the following are examples of references when the appraiser has clearly included race or other protected class references in the appraisal."
I still can't see how the leap can be made that appraisers as a class have a racial based bias.