Economic uncertainty and stock market volatility prediction - - PowerPoint PPT Presentation

Economic uncertainty and stock market volatility prediction

Vasiliki Skintzi Department of Economics, University of Peloponnese

Introduction

 There has been a growing interest in the role played by uncertainty

shocks in driving fluctuations in asset markets.

 Asset returns are functions of the state variables of the real

economy: productivity and policy shocks → output, inflation, interest rates, investment, employment → mean & volatility of asset returns.

 Various measures for economic uncertainty have been proposed in

the literature and are frequently used by market participants, policy makers, and researchers.

 I use the GARCH-MIDAS framework of Engle et al. (2013) to

examine the relationship between economic uncertainty and stock market volatility in the US.

Background – Motivation (1)

 There is ample theoretical and empirical evidence that economic

uncertainty affects stock returns and volatility.

 Veronesi (1999), Bollerslev et al (2009) and Pastor and Veronesi

(2012) provide a theoretical framework for the link between economic uncertainty and stock market volatility.

 Moreover, Pastor and

Veronesi (2012) find that individual US stock returns are more volatile and that pairwise US stock returns correlations are higher when economic uncertainty is higher.

 Liu and Zhang (2015) show that incorporating economic uncertainty

into volatility prediction models significantly improves forecasting performance.

Background – Motivation (2)

 Asgharian et al (2018) use an economic uncertainty index to

forecast stock market correlations and conclude that incorporating economic uncertainty into forecasting US-UK stock market volatilities and correlations improves out-of-sample forecasting performance.

 Brogaard and Detzel (2015) find that economic policy uncertainty

predicts stock market returns and that shocks in economic policy uncertainty earn a negative risk premium.

 Kelly et al (2016) examine the role of economic policy uncertainty

in the pricing of stock options.

Background – Motivation (3)

 Based on previous research, we expect that economic uncertainty

measures are of importance for stock market volatility.

 We draw upon several measures of economic uncertainty, and

examine the impact of economic uncertainty shocks on US stock market volatility.

 We apply the class of GARCH-MIDAS models initially proposed in

Engle et al (2013) that has been proven useful for analyzing the impact of the macroeconomic environment on financial volatility (see Asgharian et al., 2013, Conrad and Loch, 2015).

 More recently, Conrad and Kleen (2018) incorporate financial and

macroeconomic variables to forecast the long run US stock market volatility using both the multiplicative component model and the hard-to-beat Heterogenous-Autoregressive (HAR) model and conclude that MIDAS-GACH models improve upon the HAR for longer forecast horizons.

Main research question

 We investigate the question of whether economic uncertainty

measures contain information about future stock market volatility beyond that contained in common predictors such as, macroeconomic fundamentals and sentiment indicators.

Main contributions

 We link various measures of economic uncertainty to US stock

market volatility based on a MIDAS-GARCH model and examine both the in-sample estimation fit and the out-of-sample forecasting performance.

 We combine economic uncertainty measures with traditional

predictors of financial market volatility in the MIDAS-GARCH models to examine whether economic uncertainty incorporates any additional information in forecasting future volatility.

 The forecasting performance of a combined economic uncertainty

measure is also examined.

 As a robustness check we study the predictability of financial

volatility by economic uncertainty measures using standard predictive regressions for the future realized volatility.

Economic uncertainty measures

 Newspaper based measures

 Economic policy uncertainty (EPU) by Baker et al. (2016)

constructed by counting the number of articles including terms related to economic and policy uncertainty

 Geopolitical risk (GPR) by Caldara & Iacoviello (2016) by

counting the occurrence of words related to geopolitical tensions in leading newspapers

 Monetary policy uncertainty (MPU) by Husted et al (2017)

constructed by counting the number of articles including terms related to monetary policy uncertainty

 News-implied Volatility index (NVIX) by Manela & Moreira (2017)

focuses on investors’ concerns based on the co-movement between the front-page coverage of the Wall Street Journal and VIX using machine-learning techniques.

Economic uncertainty measures

 Econometric (and survey) based measures

 Scotti (2016) constructs an ex post, realized measure of

uncertainty about the state of the economy using recent economic data releases and survey (Bloomberg) forecasts.

 Jurado et al (2015) construct macroeconomic uncertainty indexes

(CMU) by aggregating the uncertainty around objective statistical forecasts across hundred of economic series.

 Ozturk and Sheng (2016) aggregate the mean square professional

forecast errors across eight variables and forecasters based on Consensus Forecasts.

 Rossi et al (2018) create an overall measure of uncertainty based

• n forecast densities using Consensus Forecasts and also provide

measures of Knightian uncertainty and ex-ante uncertainty.

Economic uncertainty measures - Descriptives

Methodology

 A MIDAS-GJR-GARCH model is employed to investigate the

relationship between stock market volatility and economic uncertainty.

 Stock market returns and volatility are modeled as follows:

Daily uncertainty measures

200 400 600 800 1,000 1,200 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14 16 DEPU GPRD SCOTTI

11/9 Iraq invasion Global financial crisis

50 100 150 200 250 300 350 400 86 88 90 92 94 96 98 00 02 04 06 08 10 12 14 16 CMU MPU NVIX Ozturk Rossi Ex ante Rossi Knightian Rossi Uncertainty

Monthly uncertainty measures

11/9 Iraq invasion Global financial crisis Brexit

Uncertainty measures - correlations

 Daily measures  Monthly measures

In-sample estimation results

Controlling for other predictors

 In daily frequency

 T

erm Spread

 In monthly frequency

 State of the economy variables

 Industrial Production growth rate (IP)  Chicago Fed National Activity index (NAI)  Housing Starts growth rate (HS)

 Sentiment indicators

 University of Michigan Consumer Confidence index (CC)  ISM New Orders index (NO)

Controlling for other predictors - Spread

Controlling for other predictors – IP

Controlling for other predictors – NAI

Controlling for other predictors – Housing

Controlling for other predictors – Consumer confidence

Controlling for other predictors – New orders

Out-of-sample forecasting

Model confidence set (Hansen et al, 2011)

Predictive regressions

 ln(Volt+1) = φ0 + φ1 Volt+1 + θi xt + vi,t+1  To deal with model uncertainty I

follow Christiansen et al (2012) and use a Bayesian model averaging approach to estimate the above predictive regression.

Conclusions

 I study the information content and predictive ability of

various economic uncertainty measures in the context of volatility forecasting.

 Economic uncertainty is important for volatility forecasting.  Economic uncertainty measures such as EPU, NVIX and

Knightian uncertainty (as measured by Rossi et al, 2018) are significant predictors of future volatility in an in-sample setting.

 Even when information on macroeconomic and sentiment

indicators is controlled for, economic uncertainty measures provide additional information about future volatility.

 Economic uncertainty measures improve out-of-sample

forecasting performance.