Degree

Doctor of Philosophy (PhD)

Department

Department of Finance

Document Type

Dissertation

Abstract

This dissertation contains three essays, two on option pricing and one on residential appraisal.

Chapter 2, with Don Chance (Journal of Derivatives 2025), fits S&P 500 call premiums with a Kernel Support Vector Machine and a Gaussian Process. Which model leads depends on the loss metric and the scoring protocol, and the ranking reverses across the moneyness and maturity surface.

Chapter 3 turns the reversal into a forecast combination condition. A simple average of two predictions with opposite sign errors beats the better of the two whenever the larger absolute error is less than three times the smaller, an algebraic threshold that holds under any loss strictly increasing in the absolute error. An analyst can check it from two realized validation errors before fitting any combination rule. Equal weighting captures 75.7 percent of the gain a Random Forest gate delivers over the better single model under cross validation and 88.2 percent out of time, holding the Chapter 2 specification fixed.

Chapter 4 takes up residential appraisal. The purchase to refinance comparison has the appearance of contract anchoring, yet a benchmark fit on one sample and scored on the other reproduces most of that pattern even when the two behave identically. A controlled random split of Federal Housing Administration appraisals drives the dispersion ratio to 1.82 in the thinnest census tracts. In the main sample that ratio is 1.039 and falls to 1.003 under a benchmark that scores both groups the same way. Measured against the contract price, the share of purchase appraisals reported exactly at the contract is 0.34, a concentration the design identifies while leaving its cause open.

Date

7-17-2026

Committee Chair

Pace, R. Kelley

LSU Acknowledgement

1

LSU Accessibility Acknowledgment

1

Download

Available for download on Sunday, July 31, 2033

Share

COinS