Macroeconomic news announcements, systemic risk, financial market volatility, and jumps

• Published date: 01 May 2018
• Author: Xin Huang
• Date: 01 May 2018
• DOI: http://doi.org/10.1002/fut.21898

Received: 17 August 2017

|

Accepted: 18 November 2017

DOI: 10.1002/fut.21898

RESEARCH ARTICLE

Macroeconomic news announcements, systemic risk, financial

market volatility, and jumps

Xin Huang

Risk Analysis Section, Federal Reserve

Board, Washington, District of Columbia

Correspondence

Xin Huang, Principal Economist, Risk

Analysis Section, Federal Reserve Board,

Mail Stop K1-91, Federal Reserve Board,

20th & C St., NW, Washington, DC 20551.

Email: xin.huang@frb.gov

I study the second-moment response to macroeconomic news announcements in

financial markets. Responses can be decomposed into contributions from continuous

volatility and discrete jumps. Disagreement and uncertainty are introduced to

measure the second moments of market forecasts. Two decades of high frequency

equity and bond futures data are examined including the global financial crisis. I

report evidence that uncertainty has a stronger effect on the second-moment response

than disagreement and the second-moment response is influenced by the level of

financial stress and monetary policy regime. The zero-lower-bound interest rate

policy constrains second-moment responses in the bond market.

KEYWORDS

disagreement and uncertainty, financial systemic risk, jumps, macroeconomic news announcements,

realized variance

JEL CLASSIFICATION

C5, G12, G14

1

|

INTRODUCTION

When macroeconomic news announcements bring new information to financial markets, how do markets respond? Do they

respond in a continuous or discrete pattern? Do their responses change over time with different factors, such as the level of

financial system stress or the monetary policy regime? These questions address the basic relationship between financial prices

and economic fundamentals.

The answer to the first question can support or contradict the efficient market hypothesis, as the new information

contained in macroeconomic news announcements differs from the market's original expectation. The answer to the

second question is not only of theoretical interest, but also has important practical implications. A continuous price path

allows for closed-form analytic solutions for both asset pricing and econometrics optimization problems, while

discreteness in the price path may render these solutions difficult. Moreover, differentiation between the continuous and

discrete components may provide further insight into the responsiveness of financial markets to news announcements.

Given the recent global financial crisis, both the level of financial systemic risk (which measures the stress level of the

financial system) and the related zero lower bound (ZLB) monetary policy could also be important in explaining the nature

of market responses to macroeconomic news.

A vast literature has addressed the first two questions, especially the first-moment responses in financial markets. Early

empirical evidence is mixed and relatively weak, especially for the equity market. Much of it is based on low frequency data and

Published 2018. This article is a U.S. Government work and is in the public domain in the USA.

J Futures Markets. 2018;38:513–534. wileyonlinelibrary.com/journal/fut

|

513

fails to consider market consensus expectations and forecast errors. Later literature reports more significant responses to news in

financial markets.

First, most market responses are short-lived, so they can be hard to detect based on daily or lower-frequency data. For

example, Jain (1988) finds that stock price responses to news are essentially complete one hour after the announcements.

Ederington and Lee (1993) show that interest foreign exchange rates respond within minutes of scheduled macroeconomic news

announcements and attendant volatility ensues.

Second, not all news announcements bring new information, and markets should respond only to the unexpected component.

Accordingly, Balduzzi, Elton, and Green (2001) propose a z-type standardized measure to convert the raw released value into a

news surprise variable.

Third, specific to equity prices, there are two competing factors: cash flows and discount rates. As argued by McQueen and

Roley (1993) and later confirmed by Andersen, Bollerslev, Diebold, and Vega (2003) and Andersen, Bollerslev, Diebold, and

Vega (2007), these two components are usually affected by a news announcement in the same direction, and the net effect

depends on their relative magnitudes. This offsetting effect thus makes it harder to detect equity market responses than bond

market responses. Meanwhile, the relative effects of news announcements on cash flow and discount rates can change over

business cycles, so separating expansion and contraction periods can help avoid opposite net effects cancelling each other out

over different business cycles.

Built upon the above observations, this paper extends the literature in three directions. First, market price responses are

separated into a continuous component and a discrete component in a statistically rigorous way, and they are studied in the same

framework. Specifically, the contributions of continuous volatility and discrete jumps to the second moment of returns are

examined. As jump detection and the measurement of volatility and jump contribution are nonparametric, I can study the second-

moment responses in financial markets independent of first-moment responses, which is helpful given the extensive literature on

first-moment responses. Some recent examples of studies of first-moment responses for various asset classes include: Goldberg

and Grisse (2013) on bond yields from four countries and exchange rates; Swanson and Williams (2014) on Treasury yields of

different maturities; and Chen, Liu, Lu, and Tang (2016) on Chinese stock index futures. In comparison, this paper focuses on the

second-moment responses to scheduled macroeconomic reports in financial markets.

The literature also includes studies on volatility or jump responses to news announcements separately. Ederington and Lee

(1993) were the first to use intraday 5-min returns to study the response of volatility to news releases. More recently, Wongswan

(2006) studies the response of volatility in the Korean and Thai equity markets to the U.S. and Japanese announcements. Some

recent papers have applied similar jump test statistics as in this paper but have focused on the first-moment jump responses to

news surprises. See, for example: Dungey, McKenzie, and Smith (2009) and Lee (2012) on jump arrivals; Evans (2011) on jump

size and risk premia; Lahaye, Laurent, and Neely (2011) on jump size and arrivals; and Jiang, Lo, and Verdelhan (2011) on jump

frequency, size, and trading activities. The joint analysis of volatility and jump contributions in this paper reveals the differential

nature of responses to news announcements in stock and bond markets with equity markets frequently exhibiting weaker

responses.

Second, consistent with the ma rket second-moment respon ses, I add the second moments of ne ws forecasts as

explanatory variables and differentiate between disa greement and uncertainty abo ut the news. Disagreement is def ined as

how much heterogeneous agent s disagree with others' forecasts on the upcoming news anno uncement and is measured by the

standard deviation of the sur vey forecasts. Uncertainty is de fined as an agent's uncertainty about his or her own forecast of

the upcoming news announceme nt. The aggregated uncertainty of the whole market is that of a representativ e agent and is

measured by the standard devia tion of the agent's forecast dist ribution, implied from the pr ices of options on news

announcements, known as econ omic derivatives (EDs). Disag reement and uncertainty may be re lated, as forecasts from

heterogeneous agents with hi gh disagreement may aggrega te into a widely distributed forec ast of the representative age nt,

resulting in high uncertain ty. But conceptually, disagr eement and uncertainty are di fferent. Empirically, Gürk aynak and

Wolfers (2006) find that the corr elation between disagreemen t and uncertainty is positive but not str ong. There are several

possible reasons. It is pos sible that the distribution of the representativ e agent's forecast is quite dispersed, but that the s urvey

responses are selected from po ints on the distribution curve th at are close to each other, resu lting in a small survey standard

deviation. In addition, sur vey forecasts usually come from 20 to 30 professional forecasters, wh ile option prices aggregate

information from all the inves tors who trade the security and often in clude the survey forecasters as well . Finally, people are

more careful when their own inve stment is at stake.

Just like the first moment of news surprise, the second moments of disagreement and uncertainty can also have intriguing

effects on volatility and jump contribution. For example, higher disagreement among agents may imply that they have more

different position holdings beforehand, which leaves them more room to trade with each other to restore the optimal positions

after the announcement, resulting in higher volatility. But higher disagreement may also imply that the agents have more

514

|

HUANG