Volatility forecasts for DaX, eUro StoXX 50 and euro-Bund futures

Authors dr. stefan Teis & Georg Gross

A report for

Good ViBRATions: tHe SeiSMoloGy oF Market Volatility

Published in November 2015 by Mondo Visione

Hamish McArthurBusiness Development DirectorMondo Visione Ltd.Central Point, 45 Beech StreetLondon EC2Y 8AD

mondovisione.comT +44 207 953 9800M +44 7906 147 527F +44 207 953 9798

Graham Taylor Graphic Design

cargocollective.com/mrgtmister.g.taylor@gmail.com

Contents

04 the value of forecasting volatility 06 accuracy of the forecast model 07 Capacity to predict volatility

signifi cantly improves

4 Market Volatility Mondo Visione

The value of forecasting volatilitySeismologists are still searching for ways to

successfully predict earthquakes. So far, they have only been able to introduce warning systems that provide alerts while an earthquake is already in progress. Shifts in fi nancial markets adhere to diff erent rules. Changes in volatility are often preceded by ‘rumblings’ or are caused by predictable events. Being able to ‘read’ the signs and predict market movements, also the smaller day-to-day movements, can create opportunity for both the buy- and sell-side.

Volatility is key to both alpha generation and trading performance measurement and is a crucial input for trading algorithms and risk methodologies. Th e capacity to forecast size and direction of volatility is therefore enormously valuable for traders of any strategy.

Th e newest Deutsche Börse analytic, the ‘Intraday Volatility Forecast’, is designed to provide that capacity. It aims to forecast the volatility for the DAX, EURO STOXX 50 and Euro-Bund futures; Eurex’ most important products. Th e Volatility Forecast creates a forward-looking insight into the peaks and troughs of market movements and provides the user with a major informational advantage over other traders.

How it worksTh e Intraday Volatility Forecast uses historical futures data to predict the direction (increasing or decreasing) and magnitude of volatility in a market. Th e volatility measure is designed to focus on the economic content of data rather than ‘noise’ created by order book micro structure, e.g. two algorithms competing with each other. Figure 1 gives an example of oscillation in price that refl ects short-term liquidity eff ects rather than volatility of the underlying price process. As can be seen in the example, volatility estimators that use time horizons of more than a few milliseconds average out this order book noise and focus on the economic price movements.

information provided by the intraday Volatility Forecast

Average realised volatility: a reference for the last second is sent every second

10-second forecast: published every fi ve seconds and covering the next three 10-second periods

1-minute forecast: published every 30 seconds and covering the next fi ve 1-minute periods

10-minute forecast: published every fi ve minutes, covering the three subsequent 10-minute periods.

Figure 1 — DAX future order book on 17 January 2015 at 11:00:32

Periods of price fi nding with intensive trading increase variance estimators

9760

9763

9766

9770

9774

9777

0

8

16

24

32

40

11:00:32 + 100 ms + 200 ms + 300 ms

Th e input going into the calculation of Deutsche Börse’s Intraday Volatility Forecast can be grouped into three categories: • Short-term memory of realised volatility • Historical intraday seasonality • Scheduled economic data releases

Th e short-term memory component of the volatility forecast model incorporates the fact that the future value of volatility is highly dependent on its prior value. Th is is realised via a statistical process that allows this auto-correlation to be captured. Based on a one second sampling frequency, the historically observed realised volatility is modelled.

Forecasting volatility

volume (2nd axis)price per 10 ms bintraded price

PR

ICE

VO

LU

ME

TIME

11:00:32

9760 0

16

32

8

24

40

9766

9774

9763

9770

9777

+ 100 ms + 200 ms + 300 ms

Mondo Visione Market Volatility 5

Forecasting volatility

Figure 2 — Average realized volatility per 1 minute bin (in %)

Figure 3 — Effect of news on realised volatility Realised volatility on the 5th of June 2014 per 2 minute bin

Figure 4 — Average volatility impact of economic releases

Region Scheduled news* Impact

US Nonfarm Payrolls 737%

EU ECB Rate Direction 541%

US FOMC 396%

US Gross Domestic Product 315%

US FOMC Minutes 278%

DE ZEW 250%

US Retail Sales 249%

US Housing Starts 233%

US ADP Unemployment 215%

US Philadelphia Fed 202%

US Constructing Spending 199%

US Durable Goods Orders 163%

EU Euro Area Inflation 163%

US New Home Sales 160%

US Producer Price Index 157%

0900 2200210020001900180017001600150014001300120011001000

0.000

0.020

0.040

0.060

0.080

0.100

0.120

0900 2200210020001900180017001600150014001300120011001000

0.000

0.020

0.040

0.060

0.080

0.100

0.120

0.000

0.005

0.010

0.015

0.020

0.025

0.030

0800 0900 2200210020001900180017001600150014001300120011001000

DAX future

EURO STOXX 50 future

Euro-Bund future

0.05%

0.00%

0.10%

0.15%

0.25%

0.30%

0.35%

0.40%

0800 0900 2200210020001900180017001600150014001300120011001000

pattern observedintraday

Sour

ce: D

euts

che

Bör

se A

G

Th e model captures the historical intraday seasonal pattern derived from historical data for the respective products. In addition, the model accounts for relevant holidays and calendar eff ects as well the extra time diff erence between Europe and the US due to the diff erence in the daylight savings conventions, which contributes to the intraday volatility pattern, as do other regular events such as the opening and closing auctions of the diff erent markets. Figure 2 depicts the intraday seasonality pattern for the three relevant futures.

Finally, important economic releases that are known to have signifi cant market impact are also integrated into the Intraday Volatility Forecast in a way that captures the average eff ects of the respective releases. For example, the model accounts for the average eff ects of interest rate announcements or US job market fi gures.

Figure 3 provides an example showing the impact of major economic releases on realised volatility on a typical trading day.

TIME

TIME

TIME

Figure 4 provides a list of the most relevant news events ranked by their impact on realised volatility.

Visible impact of scheduled releases • 13:45 ECB interest rate decision • 14:50 Building permits and initial jobless claims

Spikes at 09:40, 12:20 and 15:55 CET cannot be explained by economic releases

*M

ay –

Jun

e 20

14

PER

CEN

T

0.120

0.120

0.030

0.030

0.040

0.080

0.080

0.020

0.020

0.020

0.020

0.005

0.005

0.100

0.100

0.025

0.025

0.035

0.040

0.040

0.010

0.010

0.060

0.060

0.015

0.015

0.000

0.000

0.000

0.0000900

0900

0800

0800

1300

1300

1300

1300

1700

1700

1700

1700

1100

1100

1100

1100

1500

1500

1500

1500

1900

1900

1900

1900

2100

2100

2100

2100

1000

1000

0900

0900

1000

1000

1400

1400

1400

1400

1800

1800

1800

1800

1200

1200

1200

1200

1600

1600

1600

1600

2000

2000

2000

2000

2200

2200

2200

2200

6 Market Volatility Mondo Visione

The forecast model

Deutsche Börse’s volatility forecast model is well able to predict the structural intraday volatility pattern. Overall, the model forecasts the direction of a volatility change with approximately 70% accuracy depending on the trading day and sampling frequency. Figure 5 provides an example for one-minute bins.

Accuracy of the forecast model

While prices remain difficult to forecast, the Intraday Volatility Forecast captures much of the variation in prices in the near term. One cannot expect ‘perfect’ agreement between forecast and de facto realised volatility as the model does not include a) unscheduled news and trading events and b) only contains the ‘average’ effects of scheduled news events and cannot predict the actual outcome of the events and their specific effects on trading behaviour.

What can market participants do with the Volatility Forecast?As a feed, traders can use the forecast to support the assessment of the likelihood of price changes and therefore the risk involved in using certain automated strategies. Using an automated trading strategy set against a benchmark, such as a VWAP algorithm, is less risky if the likelihood

Figure 5 — Bund volatility forecast versus Bund future trading

Figure 6 — Forecasting the probability of price slippage for DAX® index

0

125

250

375

500

158.00

158.25

158.50

158.75

159.00

1400 1405 15001455145014451440143514301425142014151410

1400 1405 15001455145014451440143514301425142014151410

0.0000

0.0005

0.0010

0.0015

0.0020

contracts per trade (no 2nd axis)

observed (in%)

price in EUR

forecast (in %)

TIME

TIME

TR

AD

ES

RE

AL

ISE

D V

OL

AT

ILIT

Y,

1 m

inu

te b

in

Volatility had already spiked prior to publication with model expectation being very accurate in its peak

At 14:30 US increasing non-farm playrolls surpise markets and drive down BUND yields

of huge price changes in the underlying can be gauged. As such, the volatility forecast is a natural choice for a state variable in the generation of a trading algo.

On the other hand, traders who are looking for high volatility opportunities and trade respective options strategies can use the forecast to determine a good point in time for order submission, e.g. for strangle option strategies.

The Volatility Forecast is not only interesting for algo traders, but also for screen traders who can use the forecast to support their live trading.

Pre-trade transaction cost analysis (TCA) can also be enhanced considerably as the risk of price slippage can be assessed using the volatility forecast. Brokers, keen to differentiate their value-added services for buy-side clients can include this indicator into TCA offerings. Given the broad timeframes that are covered by the indicator – from 10 seconds to 30 minutes – brokers can include an estimate of possible price slippage in their TCA offering.

Finally, the volatility forecast can be used as input into intraday risk systems as volatility plays an essential role in any risk methodology. For example, as protection against large price jumps, risk limits are typically set as multiples of volatility which scales with the square root of time – sqrt(T) around the current price. Instead of using estimates for the volatility – and the approximate scaling with sqrt(T), risk managers could employ the Intraday Volatility Forecast, and thus practically include volatility changes due to “normal trading activities” in their systems.

158.00 0

125

250

375

500

0.0000

0.0010

0.0005

0.0015

0.0020

158.50

158.25

158.75

159.00

1400

1400

1425

1425

1405

1405

1430

1430

1415

1415

1440

1440

1455

1455

1410

1410

1435

1435

1450

1450

1420

1420

1445

1445

1500

1500

1400 1405 15001455145014451440143514301425142014151410

12200

12220

12210

12230

12240

12250

12280

12270

12280

simulated sample pathshistorical price evolution

TIME

With a probability of 95% the price will lie within this range for the indicated time period

10s 1m 10m

DA

X®

IND

EX

ill

ust

rati

ve

12240

12260

12250

12230

12270

12280

1400 14251405 14301415 14401410 1435 14501420 14451400 1405 15001455145014451440143514301425142014151410

12200

12220

12210

12230

12240

12250

12280

12270

12280

Mondo Visione Market Volatility 7

The forecast model

Deutsche Börse released the Intraday Volatility Forecast in November 2015. As a consequence, traders, investors and risk managers will see a marked improvement in their capacity to predict intraday volatility. The direct users and the clients of firms who build the forecast into their customer-facing systems, will be better supported in their assessments of short-term risk and slippage probabilities for trading across the DAX, EURO STOXX 50 and Euro-Bund markets.

As margins become compressed for sell-side firms and the buy-side is asked to be ever more transparent by both regulators and clients, the Intraday Volatility Forecast will deliver real insight to both enhance and appraise trading decisions.

For more information, please go to:

www.mds.deutsche-boerse.com/volaforecast

Conclusion Capacity to predict volatility significantly improves

Market Volatility Mondo Visione MMXV