Long-term scenario modeling for investors: a practitioners - - PowerPoint PPT Presentation

Long-term scenario modeling for investors: a practitioner’s perspective

Hens Steehouwer, teehouwer, Nove vember mber 3 2020

2

Topics

• 1. Models
• 2. Scenario approach
• 3. Horizons and frequencies
• 4. The frequency domain
• 5. Long-term mean assumptions (poll 1)
• 6. Climate risk scenarios (poll 2)
• 7. Summary and Q&A

3

INQUIRE EUROPE 30th anniversary

Some personal memories

2008 2008 Guus Boender Michael Damm Jens Langewand

Program Committee

4

Topics

• 1. Models
• 2. Scenario approach
• 3. Horizons and frequencies
• 4. The frequency domain
• 5. Long-term mean assumptions (poll 1)
• 6. Climate risk scenarios (poll 2)
• 7. Summary and Q&A

5

OBJECTIVES UNCERTAINTY DECISIONS

INQUInvestmentRE

Risk – return – horizon – inflation – cash flows – liquidity – liabilities – solvency – ESG – … Asset allocation – rebalancing – matching –

• verlays – options – factors – ESG – …

Financial – economic – monetary policy – Philips curve – demographics – r* / low rates – COVID-19 – climate – …

6

INQUantitativeIRE

• To answer complex questions, such as how to control

the spread of a virus, we use models of reality to perform analysis that support us in making the right decisions for the future.

Stevens, H. (2020), “Why outbreaks like coronavirus spread exponentially, and how to flatten the curve”, The Washington Post, March 14 2020. https://www.washingtonpost.com/graphics/2020/world/corona-simulator/ https://www.ortecfinance.com/en/insights/whitepaper- and-report/financial-models-and-the-corona-crisis

• Also making investment decisions to achieve objectives

in the future is complex and relies heavily on models.

• Models help us to quantify, analyze and compare the

potential consequences of various decisions under different sets of assumptions. In doing so we learn in a structured way, which decisions work well and which decisions might not be effective.

7

Dices and planets

• All economic and financial models are based on

assumptions and cannot match the level of accuracy

• f models from the natural sciences
• Essential to use models in a sensible way: validate

assumptions, include parameter uncertainty, avoid sampling noise, perform sensitivity analysis, combine strengths of different approaches etc.

8

Realistic assumptions

Diebold, F.X. and C. Li (2006), “Forecasting the Term Structure of Government Bond Yields”, Journal of Econometrics, 130, 337-364.

Realistic ≠ Prudent Realistic ≠ Consensus Realistic ≠ Intuitive Realistic ≠ Rational Realistic ≠ Simple

9

Topics

• 1. Models
• 2. Scenario approach
• 3. Horizons and frequencies
• 4. The frequency domain
• 5. Long-term mean assumptions (poll 1)
• 6. Climate risk scenarios (poll 2)
• 7. Summary and Q&A

10

Scenario approach

• Flexibility  realistic
• Understanding  decisions (!)

“A scenario is a possible evolution

• f the future consistent with a

clear set of assumptions”

Bunn and Salo (1993)

“A scenario is a description

• f a possible future state of an
• rganization’s environment

considering possible developments

• f relevant interdependent factors

in the environment”

Brauers and Weber (1988)

11

The academic version

“The dynamic behavior of a pension plan is clearly dominated by rules and methodology which are discontinuous and non-linear function of its financial condition. The task of developing a closed-form solution to evaluate the potential state of a pension plan following a series of stochastic investment and inflation experiences would be extremely difficult, if not impossible. To date, the only approach that has proven feasible is the application of Monte Carlo Simulation, wherein an investment and inflation scenario is generated by random draws based

• n the expected probability distribution of year to year investment and

inflation behavior. In order to develop an accurate assessment of the range of potential uncertainties, it is necessary to repeat this simulation process by generating dozens or hundreds possible scenarios, consistent with statistical expectations.”

Kingsland, L. (1982), “Combining financial and actuarial risk: Simulation analysis. Projecting the financial condition of a pension plan using simulation analysis”, Journal of Finance, vol. 37

12

From scenarios to decision making

Example: Insurance company under Solvency II

• 1. More efficient strategies
• 2. More effective strategies
• 3. Strategy decision

0% 5% 10% 5 7.5 2.5 10 12.5 20

• 3. Asset Allocation × Duration 10
• 1. Asset Allocation only

Probability available capital < required capital Expected market value funding ratio

• 2. Duration only

13

Approaches for constructing scenarios

• Historical averages: Use historical averages as expectations for the future
• 1. Future extremely unlikely to unfold in this way
• 2. Suggests level of certainty that is not there in reality
• Historical simulation: Simple way of dealing with uncertain future but also drawbacks:

1. History only one very specific realization of all that might have happened 2. Ignores economic and financial market conditions that prevail today

• By hand: Create scenarios by formulating what might happen in the future based on specific

narratives of possible events.

• Pros: simple, out-of-the-box, narratives, risk awareness, stress testing
• Cons: incomplete picture of uncertainties, risk-return tradeoff

14 30/09/2020

Interval Expected Realized 95 - 100% 5% 4% 75 - 95% 20% 15% 25 - 75% 50% 62% 5 - 25% 20% 15% 0 - 5% 5% 3%

Observations: 6440

Model performance

Observations

0% 10% 20% 30% 40% 50% 60% 70% 5% 20% 50% 20% 5% Expected Realized

Jun 2016 – Sep 2020

Out-of-sample forecasting performance of official monthly versions of Ortec Finance scenarios since launch per end of June 2016 until September 2020 based

• n 1, 3 and 12 month forecasts of financial market variables such as 3M and 10Y

government bond yields, 3M and 10Y IG and HY spreads, equity total return indices, commodities and exchange rates

Jun 2016 – Sep 2020

1% probability 4% 4% 1% 20% 20% 50%

Monthly asset only portfolio value and risk according to a typical investment strategy of a European pension fund

Model based approach

ASSUME  MODEL  CALIBRATE  GENERATE and VALIDATE

Likelihoods  Risk-return tradeoffs

15

Topics

• 1. Models
• 2. Scenario approach
• 3. Horizons and frequencies
• 4. The frequency domain
• 5. Long-term mean assumptions (poll 1)
• 6. Climate risk scenarios (poll 2)
• 7. Summary and Q&A

16

Horizon and frequencies

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.00 0.10 0.20 0.30 0.40 0.50 EQ US - EUR - monthly EQ US - EUR - annual EQ US - HY US - monthly EQ US - GSCI - monthly

1 Normal implied correlations based on Ortec Finance Economic Scenario Generator (ESG): correlation of a

bivariate Normal distribution which corresponds to the measured Tail Dependence Coefficient (TDC) per quantile (threshold) in the left part of the distributions, where the TDC corresponds to the probability that one margin exceeds a threshold under the condition that the other margin exceeds a threshold. See e.g. Frahm et al. (2005).

1: Tail correlations in the left part of return distributions1

Correlation Left tail of the distribution 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 5 10 15 20 25 30 CPI - Houses CPI - Equities Equities - Houses Equities - Interest rates

1 Based on Ortec Finance Economic Scenario Generator (ESG): correlations between cumulative (annualized) US

CPI, geometric US equity returns, geometric US house price returns and the 10-year US Government yield, calculated on investment horizons of 1 to 30 years.

2: Correlations of US cumulative returns across the investment horizon (in years)1

Correlation Horizon in years

17

How trend – cycle decompositions can help

–

0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 Trend

• 3%
• 2%
• 1%

0% 1% 2% 3% 1970 1980 1990 2000 2010 2020 Cycle

• 3%
• 2%
• 1%

0% 1% 2% 3% 1990 2000 2010 2020 Irregular 0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 Long-term UK government bond yield Sources: Bank of England, Bloomberg and Ortec Finance

1: Trend – cycle decomposition of long-term UK government bond yield time-series

Secular stagnation Savings glut Debt super-cycle Climate ... Business cycle Monetary policy Fiscal policy Covid-19 … Momentum Return reversal Alfa …

18

Returns of different frequencies

–

• 1.5
• 1.0
• 0.5

0.0 0.5 1.0 1.5 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 10Y return (log) Trend

• 0.8
• 0.6
• 0.4
• 0.2

0.0 0.2 0.4 0.6 0.8 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 1Y return (log) Cycle

• 0.3
• 0.2
• 0.1

0.1 0.2 0.3 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 1M return (log) Irregular 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 Index (log)

Trend – cycle decomposition of US equities total return index

Sources: Goetzmann et al. (2001), Shiller, Bloomberg and Ortec Finance

19

How trend – cycle decompositions can help

Trend model Cycle model Irregular model Scenarios/Confidence bands/Forecasts Trend – cycle decomposition Historical time series  Consistent, efficient and realistic scenario modeling

20

Topics

• 1. Models
• 2. Scenario approach
• 3. Horizons and frequencies
• 4. The frequency domain
• 5. Long-term mean assumptions (poll 1)
• 6. Climate risk scenarios (poll 2)
• 7. Summary and Q&A

21

Time for the frequency domain

• Economies and financial markets move up and down all the time
• Natural way to analyze and model how they move up and down is in the frequency domain
• Fourier transform: any time-series {xt, t=0,…,T-1} can be written as a sum of cosine functions
• {Rj, j and j, j = 0,…,T-1} represent the amplitudes, frequencies and phases of T cosine functions that

together offer an alternative representation of the time-series xt

• Periodogram decomposes a time-series variance into frequencies
• Spectral density of stochastic process is the expected periodogram as sample length goes to infinity
• Applications (i) trend – cycle decompositions and (ii) spectral analysis



 

 

1

) cos(

T j j j j t

t R x  

 

   



1 1 2 2

1 T

t T j j t

R x T

Steehouwer, H. (2005), “Macroeconomic Scenarios and Reality. A Frequency Domain Approach for Analyzing Historical Time Series and Generating Scenarios for the Future”, https://research.vu.nl/en/publications/macroeconomic-scenarios-and-reality-a-frequency-domain-approach-f

22

Back to 2006

14

Deterministic simulation

 2 dimensional VAR(2) model

7 . 2 . 2 . 3 . 3 . 4 . 3 . 5 .

, 2 , 1 2 , 2 2 , 1 1 , 2 1 , 1 , 2 , 1

                                        

    t t t t t t t t

x x x x x x                             1 1 , ~

, 2 , 1

N

t t

 

15

Stochastic simulation

 x1: Slowly moving pattern together with fluctuations with a length of 4 to 5 years  x2: Mostly fluctuations with a length of 4 to 5 years  Short term fluctuations highly correlated  x1 leads x2 by 1 to 2 years for the short term fluctuations

23

Spectral densities of VAR model

29

Multivariate spectral density

Auto Spectrum x1 Auto Spectrum x2 Coherence Spectrum Phase Spectrum Fluctuations 4 to 5 years Long term fluctuations High phase corrected correlations x1 leads x2 by 1.5 years

24

Topics

• 1. Models
• 2. Scenario approach
• 3. Horizons and frequencies
• 4. The frequency domain
• 5. Long-term mean assumptions (poll 1)
• 6. Climate risk scenarios (poll 2)
• 7. Summary and Q&A

25

Zero frequency = long-term mean

–

0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 Trend 0% 2% 4% 6% 8% 10% 12% 14% 16% 18% 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 Long-term UK government bond yield

1: Trend – cycle decomposition of long-term UK government bond yield time-series

Sources: Bank of England, Bloomberg and Ortec Finance

Mean = 5.5% Mean = 5.5%

• Productivity slowdown?
• Demographic change?
• Lower inflation (expectations)?
• Safe haven asset demand?
• Government debt ratios?
• Monetary policy innovation?
• Persistency?
• Pandemics?
• Climate?
• …

26

Interest rate models

ken-nyholm.com Eser, Lemke, Nyholm, Radde and Vladu (2019) Bauer and Rudebusch (2020)

• Secular stagnation (e.g. Summers)
• Temporal savings glut (e.g. Bernanke)
• Debt supercycle (e.g. Rogoff)

27

Poll 1

What do you consider a realistic long-term assumption for the value of the nominal 10-year German government bond yield 40 years from now, so in 2060?

• > 4.0%
• [3.0% - 4.0%]
• [2.0% - 3.0%]
• [1.0% - 2.0%]
• [0.0% - 1.0%]
• < 0.0%
• Impossible to say

28

Long-term mean assumptions: building blocks

Short (3-month) real rate

• 0.50%

+ Long-term expected inflation 1.75% + Inflation risk premium 0.25% + Maturity spread 0.75% = Long (10-year) nominal rate 2.25%

Building blocks GER interest rates

Short (3-month) Nominal rate 1.25% Break-even inflation 2.00% Nominal maturity spread 1.0% Long (10-year) real rate 0.25%

29

Long-term mean assumptions: uncertainty

Trend 10 year GER government bond yield Irregular Cycle Long-term mean 2.25%

+ + =

30

Topics

• 1. Models
• 2. Scenario approach
• 3. Horizons and frequencies
• 4. The frequency domain
• 5. Long-term mean assumptions (poll 1)
• 6. Climate risk scenarios (poll 2)
• 7. Summary and Q&A

31

Economic impact of climate change

• Transition risks and opportunities related to the transition to a less carbon intensive world
• Changes in carbon prices, technological innovation, policy change and so on
• Global impact of the transition relatively small, but some countries / sectors will loose and some will win
• Physical risks related to the impact of increased temperatures and changing precipitation patterns
• Gradual impacts: temperature impact on agricultural, industrial and worker productivity
• Extreme weather: frequency and severity of wildfires, flooding and tropical storms
• On a global level, gradual physical risk expected to have the largest impact on the economy
• Impact on investment risk and return
• Economic growth, different across countries and sectors
• Inflation, interest rates, credit spreads, equities, real estate, …

32

Climate scenarios describe how temperature targets can be reached

CLIMATE SCENARIOS

(policy, technology, liability, natural system, etc.)

IEA PIK CICERO IIASA Cambridge Econometrics Etc

Scenario analysis of climate science

Scientific temperature pathways (IPCC)

Global surface temperature change (oC)

33

Integrated financial and climate scenario analysis

Scientific temperature pathways (IPCC)

Global surface temperature change (oC)

Traditional economic and financial scenarios

34

Climate informed investment decision making

Climate-related risks are systemic ic and therefore unhed edgeabl ble.

1 3

Climate te change will fundamentally impact how the econom

• my

y perform rms as a whole.

4

‘Stock-picking’ insufficien ficient t to manage systemic risk.

5

Take climate change into account as a risk driver ver in strateg egic ic investmen ent t decis ision ion-mak aking.

2

Investors face both transitio ition and physica ical risks related to climate change

35

Global warming pathways (scenarios)

In line with: Limiting avg. global warming to below 2°C by 2100 (IPCC RCP 2.6) In line with: Limiting avg. global warming to below 2°C by 2100 (IPCC RCP 2.6) In line with:

• Avg. global temperaturas rising to

above 4°C by 2100 (IPCC RCP 8.5) Orderly transition pathway Locked-in physical impacts Disorderly transition pathway Locked-in physical impacts

2oC

Business-As-Usual (only currently committed transition efforts) Dramatic physical impacts

4+oC

Paris Orderly Transition Paris Disorderly Transition Failed Transition

Narratives Dashboard https://climatemaps.app/hello/?redirect_to=https%3A%2F%2Fclimatemaps.app%2F “Scenario Analysis for Systemic Climate Risk”, in “Case studies of Environmental Risk Analysis Methodologies”, Network of Central Banks and Supervisors for Greening the Financial System (NGFS), September 2020 https://www.ngfs.net/sites/default/files/medias/documents/case_studies_of_environmental_risk_analysis_methodologies.pdf

36

Poll 2

Which global warming pathway (scenario) do you consider the most likely for the coming 40 years?

• Paris Orderly Transition (1.5oC)
• Paris Disorderly Transition (2.0oC)
• Failed Transition (4+oC)
• Another
• Impossible to say

37

Systemic climate risk scenarios

38

Macroeconomic model

https://www.e3me.com/

• Address national and global economy-

environment policy challenges

• Inputs: policy and technology assumptions

per global warming pathway

• Outputs (e.g.): climate adjusted GDP and CPI
• Strong empirical basis, not assume optimising

behaviour and full utilisation of resources

• Two-way linkages between the economy,

wider society and the environment

• Both short and long-term impacts, up to 2050
• 61 global regions, with a detailed sectoral

disaggregation in each one

• Includes transition and gradual physical risks,

extreme weather risk requires separate (catastrophe) modeling

39

3 ºC

4+ ºC 1.5 ºC

Climate-adjusted GDP growth (cumulative difference to climate-uninformed baseline)

Region B Region A

Systemic climate risk scenarios

40

Impact of on expected returns

Policy and technology assumptions per global waring pathway (scenario)

Climate adjusted GDP and CPI Climate informed asset risk and return scenarios

Environment – economic model Economic – financial market model

41

Topics

• 1. Models
• 2. Scenario approach
• 3. Horizons and frequencies
• 4. The frequency domain
• 5. Long-term mean assumptions (poll 1)
• 6. Climate risk scenarios (poll 2)
• 7. Summary and Q&A

42

Summary

• Models
• Support decision making in a complex and uncertain world
• Based on assumptions and to be used in a sensible way
• Scenario approach
• Flexibility and insights to deal with the complexity of the real world
• Stylized facts valuable basis for constructing realistic assumptions
• Frequency domain
• Powerful toolkit to help think about horizons, frequencies and dynamics
• Integrating impact of climate change into investment decision making and models key theme

43

Q&A

Contact

www.ortecfinance.com

Rotterdam

Ortec Finance bv Boompjes 40 3011 XB Rotterdam The Netherlands

• Tel. +31 10 700 50 00

Amsterdam

Ortec Finance bv Naritaweg 51 1043 BP Amsterdam The Netherlands

• Tel. +31 20 700 97 00

London

Ortec Finance Ltd Bridge House 181 Queen Victoria Street London, EC4V 4EG United Kingdom

• Tel. +44 20 3770 5780

Pfäffikon

Ortec Finance AG Poststrasse 4 8808 Pfäffikon SZ Switzerland

• Tel. +41 55 410 38 38

Toronto

Ortec Finance Canada Inc 250 University Avenue Toronto, ON M5H 3E5 Canada

• Tel. +1 416 736 4955

Hong Kong

Ortec Finance Asia Ltd Unit 211, 2/F, Building 12W, Phase 3 Hong Kong Science Park Shatin, New Territories Hong Kong

• Tel. +852 2477 92 88

Melbourne

Ortec Finance Pty Ltd WeWork Level 7 222 Exhibition Street Melbourne, VIC 3000 Australia

• Tel. +613 8899 6455

Hens Steehouwer, PhD

+31 6 12 12 40 20 Hens.steehouwer@ortec-finance.com

Head of Research

45

PCA factors per component

• 1. Decompose economic and financial market

time-series across asset-classes and regions

• 2. Standardize to mean 0 and stdev 1
• 3. Calculate PCA factors
• 4. Standardize PCA factors to mean 0 and stdev 1

1970 1980 1990 2000 2010 2020

Cycle PCA 1

1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Trend PCA 1

1990 2000 2010 2020

Irregular PCA 1

46

Multiple PCA factors needed

• 4
• 3
• 2
• 1

1 2 3 1970 1980 1990 2000 2010 2020 Cycle PCA1 Cycle PCA2 Cycle PCA3

• 3
• 2
• 1

1 2 3 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 Trend PCA1 Trend PCA2 Trend PCA3

• 8
• 6
• 4
• 2

2 4 6 1990 2000 2010 2020 Irregular PCA1 Irregular PCA2 Irregular PCA3

• Typically 5 to 10 PCA factors per frequency band

for a parsimonious representation of the correlation structure of hundreds of time-series

• Dynamics of PCA factors and actual financial

market time-series can be modeled with conventional Dynamic Factor Models

Disclaimer

Please note that this report has been prepared with care using the best available data. This report may contain information provided by third parties or derived from third party data and/or data that may have been categorized or otherwise reported based upon client

• direction. This report is not intended as an investment advice. Ortec Finance assumes no responsibility for the accuracy, timeliness
• r completeness of any such information. Ortec Finance accepts no liability for the consequences of investment decisions made in

relation on information in this report. All our services and activities are governed by our general terms and conditions which may be consulted on www.ortecfinance.com and shall be forwarded free of charge upon request.