| 摘要: | 本文分別以美國耐久財、能源、科技與製造業的上市公司波動作為預測目標，並以 Inter‑Quantile‑Range‑based Volatility （IQRBV）作為波動代理並進行預測，樣本期間為 1990 年 1 月至 2021 年 12 月的月頻率資料，變數種類包含公司基本面的財務、經營、技術、供應、企業社會責任類別，以及代表其他未來資訊的情緒和總體經濟等類別共 414 個變數，並藉由不同的降維演算法萃取類別因子，並輸入分量迴歸預測分位數(10%,50%,90%)並依此建構 IQRBV。首先在四個產業中科技業波動預測表現最佳能源業次之，再者，形成波動的可能原因隨同時間變化，因此不同時空背景與經濟結構下，選擇使用何種變數因子與模型預測波動至關重要。最重要的是，本研究的模型不僅能預測產業的上行和下行波動，還能將這些波動與經濟因素結合，深入探討波動背後的影響因素。這種方法有助於市場參與者在決策時提高依據性，從而增強其資產配置和風險管理決策之間的因果關係。;This article targets the volatility of publicly listed companies in the U.S. durable
goods, energy, technology, and manufacturing sectors, utilizing the Inter-Quantile Range-based Volatility (IQRBV) as a proxy for volatility forecasting. The data spans
from January 1990 to December 2021 with a monthly frequency, incorporating 414
variables across categories including financial fundamentals, operations, technology,
supply, corporate social responsibility, as well as sentiment and macroeconomic
indicators that represent additional future information. Dimensionality reduction
algorithms are employed to extract categorical factors, which are then input into
quantile regression to predict volatility at the 10%, 50%, and 90% quantiles for the
construction of IQRBV.
Among the four sectors, technology exhibits the best performance in volatility
forecasting, followed by energy. Moreover, the factors contributing to volatility evolve
over time, indicating the critical importance of selecting appropriate variables and
models for volatility forecasting under different temporal and economic contexts. Most
importantly, the models developed in this study are capable not only of forecasting
industry-specific upward and downward volatility but also of integrating these
volatilities with economic factors to delve deeper into the underlying drivers of
volatility. This approach aids market participants in enhancing the substantiation of
their decision-making processes, thereby improving the causal relationship between
asset allocation and risk management decisions. |
