Measuring and Testing the Impact of News on Volatility
Source: Engle, R. F. & Ng, V. K. (1993) · Journal of Finance 48(5), 1749–1778 · doi:10.1111/j.1540-6261.1993.tb05127.x
TL;DR
Introduces the News Impact Curve (NIC) — the mapping from a return shock to next-period conditional volatility, holding the past constant — as a common yardstick for comparing ARCH-family volatility models, plus a set of diagnostic tests for asymmetry. Estimated on daily Japanese TOPIX returns (Jan 1980–Dec 1988), the curve is asymmetric: negative shocks raise volatility more than positive shocks of equal size (the leverage effect). Symmetric GARCH fails the asymmetry tests; the GJR (Glosten-Jagannathan-Runkle) and Nelson EGARCH models fit best.
What it models
The conditional variance response to news. Different volatility specifications imply different responses to good vs bad news; the NIC and the bias tests let models be compared on how they translate the sign and size of shocks into next-period volatility.
Specification
News Impact Curve: plot the one-step-ahead conditional variance h_t as a function of the current shock ε_{t-1}, holding all earlier information at its unconditional level. For symmetric GARCH the curve is a centered parabola; for EGARCH/GJR it is asymmetric and (for EGARCH) recentered/steeper for bad news.
Partially Non-Parametric (PNP) model: a piecewise-linear ARCH specification that lets the data estimate the NIC shape flexibly, nesting many parametric forms.
Estimation
Daily TOPIX index returns, full sample 1 Jan 1980 – 31 Dec 1988; robustness on a pre-crash subsample 1 Jan 1980 – 30 Sep 1987.
The mean is pre-filtered (day-of-week dummies plus an AR adjustment, à la Pagan-Schwert) so the analysis focuses on the unpredictable return component.
Parametric models (GARCH, EGARCH, GJR, AGARCH, VGARCH, NGARCH) and the PNP model are estimated; NICs are compared against the non-parametric benchmark.
What it captures
Volatility asymmetry / leverage effect: bad news increases volatility more than equally sized good news.
Diagnostic tests run on standardized residuals: the sign-bias, negative-size-bias, and positive-size-bias tests (t-ratios on indicator/size regressors) detect misspecified responses to the sign and magnitude of shocks. Symmetric GARCH fails these; GJR and EGARCH pass best (results weaker on the shorter pre-crash sample).
Use & extensions
The NIC and bias tests became standard tools for building and validating volatility models, formalizing the leverage effect that GJR and EGARCH were designed to capture.
Limitations
A daily, parametric framework; predates high-frequency realized-volatility methods.
Estimated curves are sensitive to conditioning information and sample; evidence here is from a single (Japanese) market.
Key references
Engle, R. & Ng, V. (1993) — Measuring and Testing the Impact of News on Volatility — Journal of Finance
Glosten, L., Jagannathan, R. & Runkle, D. (1993) — On the Relation between the Expected Value and the Volatility of the Nominal Excess Return on Stocks — Journal of Finance
Nelson, D. (1991) — Conditional Heteroskedasticity in Asset Returns: A New Approach (EGARCH) — Econometrica
Provenance: verified/generated from the paper's full text.
