Another Modernization Thread

[RCA]

Market conditions comments from the report:

Market conditions surveyed for the past twelve months reveal that general property values have been stable.


365 days ago, the SP/SF was $190.00; as of the effective date, it is $208.00.

View attachment 52991

Actually, if you were doing non-linear regression, the plot would curve upwards starting around about 220 +/- days since sale.

 

[Mejappz]

Actually, if you were doing non-linear regression, the plot would curve upwards starting around about 220 +/- days since sale.

Yes, but it would not improve the model. I already plotted the residual against the Days Since Sale, and there was no curve-linear relationship.

 

[TerryRohrer]

So


So - let's recap:

1. Borrower got ripped off;
2. Appraiser got ripped off (and tried to make up for it by not being thorough);
3. AMC got exactly what it paid for.

On another note I am a little concerned that the "powers that be" will try to time the introduction of the "newfangled mobile-friendly, dumbed down appraisal forms" in a few years to coincide with the influx of these "PAREA"-trained homunculi who might just be getting trained on this "newfangled software" while all of us are out doing real work. And that they might be looking to "ease us out of the business" via this route.

I am probably being a bit too paranoid (easy to become that way hanging around here), but - anyone else have any thoughts on this?

From the information provided, I would not conclude the appraiser got ripped off. Appears to have gotten a premium based on their qualifications/efforts. But your spot on regarding the PAREA trained folk and bifurcated appraisal process. Couple that with the unwillingness in Congress and the government at large to allow a firewall between mortgages and the Treasury, this will catch on like wildfire!

 

[Meandering]

Just wait on Cuomo to sue the GSEs again.

 

[RCA]

Yes, but it would not improve the model. I already plotted the residual against the Days Since Sale, and there was no curve-linear relationship.

View attachment 53046

Without overfitting, a more complex spline could be used that woud, from right to left, go up because of that one high grouping, then probably go down a ways then go back up again. That would give you a higher R2. You would want to use non-parameterized statistics. If you want to send me the data as a CSV, I can run it through R-earth, --- or you can.