Forecast adjustment under shocks: similarity-based solutions to unprecedented events

Lundquist, David Patrick

Permalink

https://hdl.handle.net/2142/129291 Copy

Description

Title

Forecast adjustment under shocks: similarity-based solutions to unprecedented events

Author(s)

Lundquist, David Patrick

Issue Date

2025-05-02

Director of Research (if dissertation) or Advisor (if thesis)

Eck, Daniel J

Doctoral Committee Chair(s)

Eck, Daniel J

Committee Member(s)

• Liang, Feng
• Lee, Jihyung
• Dayanikli, Gokce

Department of Study

Statistics

Discipline

Statistics

Degree Granting Institution

University of Illinois Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• forecasting
• shock
• unprecedented events
• uncertainty

Abstract

This work examines various forecasting challenges under conditions that are not ideal. In the first chapter, we develop a procedure for forecasting the volatility of a time series immediately following a news shock by exploiting series that have experienced similar shocks. By positing news shocks to be affine functions of exogenous risk proxies, we are able to express particular news shocks as random realizations of a common data-generating process. In our model, random effects induce excess volatility. Those excess volatilities are then estimated as fixed effects and aggregated to adjust the GARCH volatility forecast of the time series under study by an additive term. The adjusted and unadjusted forecasts are evaluated using the unobservable but easily-estimated realized volatility (RV). Asymptotic results are established for our excess volatility estimators as well as our adjustment estimator. Numerical simulations are provided to illustrate conditions and hyperparameters under which our method thrives. A detailed real data application for forecasting volatility after the outcome of the 2016 United States presidential election demonstrates the applicability of the method. The second chapter dramatically abstracts from the comfort of a particular model and instead probes situations in which one’s default times series forecasting model must be adjusted, augmented, or abandoned completely, generating second-order questions about algorithmic design and the role of human judgment, as well as the information leveraged for forecast adjustment. This work further systematizes and unifies the rich landscape of model adjustment and model correction methods, with a special focus on forecast adjustment under the presence of news shocks, when unanticipated events may give an observer reason to doubt the credibility of the default forecasting function. We demonstrate the usefulness of similarity-based methods in forecasting and present a general framework dubbed Similarity-based Parameter Correction (SPC). We highlight several specific time series models that can benefit from SPC, along with formal results for some of those special cases.We close with a thorough discussion of the state of forecasting under non-ideal conditions and the directions in which it may find the most success.

Graduation Semester

2025-05

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/129291

Copyright and License Information

Copyright 2025 David Lundquist

Owning Collections