Session 54, Update on SOA Research Project – Modern Deterministic Scenarios for Interest Rates Moderator: Donald M. Walker, ASA, MAAA Presenter: Mark E. Alberts, FSA, MAAA Lori L. Helge, FSA, MAAA Donald M. Walker, ASA, MAAA
Session 54, Update on SOA Research Project – Modern Deterministic Scenarios for Interest
Rates
Moderator: Donald M. Walker, ASA, MAAA
Presenter:
Mark E. Alberts, FSA, MAAA Lori L. Helge, FSA, MAAA
Donald M. Walker, ASA, MAAA
PRELIMINARY DRAFT2016 Valuation Actuary SymposiumSession 54
© 2016 Willis Towers Watson. All rights reserved.
Update on SOA Research Project: Modern Deterministic Scenarios for Interest Rates
August 30, 2016
Lori Helge, FSA, MAAASenior Consultant, Willis Towers Watson
2
Contents
The Role of the Appointed Actuary
The Concept of Moderately Adverse
SOA Research Papers
Interest Rate Scenarios Today
3
The Role of the Appointed Actuary
Our role has become increasingly complex and challenging over the past decade
Professional Standards
Competing Interests
The fuzzy limits on reasonableness of assumptionsWhat is beyond moderately adverse?
4
The Concept of Moderately Adverse
ASOP No. 22, 2.15:
“Moderately Adverse Conditions – Conditions that include one or more unfavorable, but not extreme, events that have a reasonable probability of occurring during the testing period.”
ASOP No. 22, 3.4.2:
“Adequacy of Reserves and Other Liabilities – When forming an opinion, the actuary should consider whether the reserves and other liabilities being tested are adequate under moderately adverse conditions, in light of the assets supporting such reserves and other liabilities. To hold reserves or other liabilities so great as to withstand any conceivable circumstances, no matter how adverse, would usually imply an excessive level of reserves or liabilities.”
5
SOA Research Papers
Transition to a High Interest Rate Environment Published July 2015 Authors: Max Rudolph, Randy Jorgensen, Karen Rudolph
https://www.soa.org/Research/Research-Projects/Life-Insurance/research-2015-rising-interest-rate.aspx
Sustained Low Interest Rate Environment: Can it Continue? Why it Matters. Published June 2014 Author: Max Rudolph https://www.soa.org/files/research/projects/2014-sustianed-low-interest-
environment.pdf
6
Interest Rate Scenarios: The NY7……er, 5
The NY7 was developed in a period of much higher interest rates The full 3% and 5% movements created seven unique scenarios The floor was nothing more than a hypothetical For YE2015, NY did not require companies to pass the pop-down and
decreasing scenarios Now The down, down/up and pop down scenarios are tripping over one
another in years 1-10 The rates after 10 years are very similar for five of the seven scenarios And the increasing and pop up scenarios bring us back to historical
average-ish levels If this is all that you test, what risks are overweighted, and what risks
are ignored? And most important, does that satisfy the moderately adverse requirement?
7
Interest Rate Scenarios: Deterministic
Deterministic scenarios allow us to tell a story around the impact of interest rates on our results, and that story has meaning to a wider range of audiences
Easier to interpretEasier to review interim resultsEasier to describe to managementBe clear about best estimate vs. best case
Of course, the results are only as good as the model and assumptions behind them, but that’s not what we’re covering here today
8
Interest Rate Scenarios: Stochastic
Mean reversion targetsUnderstand the strength of the reversionHow far do you get in 3 years? 5 years? 10 years?What is your distribution of rates over all scenarios?
Black boxNot effective for a wide range of audiencesOne-size-fits-all generators
9
Interest Rate Scenarios: Testing Tail Risk
Source: Transition to a High Interest Rate Environment, Chart 28
Session 54:Update on SOA Research Project-Modern Deterministic Scenariosfor Interest Rates
1
Research Sponsored by:
• Financial Reporting Section
• Smaller Insurance Company Section
• Modeling Section
• Committee on Life Insurance Research
2
Lead Researcher:
Mark Alberts- Alberts Actuarial Consulting LLC
Project Oversight Group:Pam Hutchins (Chair)
Steve Cheung Zhixin Wu
Duncan Cook Corwin Zass
Kip Headley Ronora Stryker
Eric Janecek Jan Schuh
Don Walker
3
This is an Interactive Forum!
Audience Participation is encouraged!
4
Our Presenters
• Don Walker ASA,MAAAFarm Bureau Life of Michigan (Retired)
• Lori Helge FSA,MAAAWillis Towers Watson
• Mark Alberts FSA,MAAAAlberts Actuarial Consulting LLC
5
Our Agenda
• (Don) History: 25+ Years of Asset Adequacy Analysis(How the NY7 worked so well for many years
and then . . . )• (Lori) Current Challenges for Appointed Actuaries
(How do you define “Moderately Adverse”??) (Stochastic vs. Deterministic??)
• (Mark) Our Research- Goals, Approach, and Results(and YOUR chance to give immediate feedback)
• (Open Forum) YOUR Chance to Delve Deeper, Ask More Questions, Give Comments, and Add Your Input to the Discussion
6
26 Years of Quarterly US Treasuries
7
0.00%
1.00%
2.00%
3.00%
4.00%
5.00%
6.00%
7.00%
8.00%
9.00%
10.00%
90 Day
10 Year
What Interest Rate Changes Can Do to Profitability in an In-force Model
• In “Normal” interest rate environments, a moderate increase in interest rates will tend to hurt profitability more than a similar decrease
• Back in the 1990s, we worried much more about the results in the “UP” scenarios than the results in the “DOWN” scenarios
• But, as rates trended downward, we enjoyed having an asset portfolio that was “longer” than our liabilities, and NY7 results became routinely benign
8
However . . .
• As we entered the 2000s, results began to show deterioration, both bottom line and CFT (especially in those “DOWN” scenarios)
• Companies would cut non-guaranteed elements, and actuaries would learn how to incorporate those cuts into their models
• At the same time, our investment colleagues were encountering more exotic instruments that we had to learn to model
• Asset Adequacy work was becoming more stressful, but we had it under control, didn’t we?
9
And Then . . .The Crisis Hits (and is not going away)
• When rates get low enough, policy guarantees can become “in the money” and this reverses the impact of further interest rate movements
• Now the “UP” scenarios have become our friends, and at least 2 of the “DOWN” scenarios are so bad that almost everyone wants to forget about them
• This is “regime change”. We’d better get used to it and we’d better be applying our actuarial skills to figuring it out.
10
The Horizon is Murky
• In 2008-2009, the pundits were predicting a steep but short recession (a “V”-shape). They were wrong.
• What we seem to have gotten is a sharp drop, followed by a very long and gradual economic recovery (a “J” shape)
• A significant increase in interest rates is seemingly always a year(or three) away
• And then there are GREXIT, Debt-ceiling, China, and BREXIT crises that seem to drive our rates downward
11
Inspiration for This Project• This project was inspired by last year’s ValAct Session
(Interactive Forum) on “Low Interest Rates and Their Impact on Asset Adequacy Analysis”
• Based on feedback gathered at that session, the SMALLCO Council approved the idea of proposing and funding a research effort on deterministic scenarios and the concept of “moderately adverse”
• Support was quickly gathered from the Financial Reporting Section, the new Modeling section, and the SOA (via CLIR)
• The project kicked off in January, an RFP was sent out, and Mark’s firm was selected in March. He began working on it in April, with one goal being to have something to present at this meeting. His work is continuing and we appreciate the opportunity to get feedback from you.
12
How This Fits Into the Great Scheme of Things• It is the role of the Society of Actuaries to Research
and to Educate. That is what we’ve embarked on with this project, which we hope will complement work already begun by others
• It will be the role of the American Academy of Actuaries to take our work and turn it into a public policy proposal, or other vehicle, if appropriate
• It will ultimately be the role of the regulatory community (NAIC and the individual states) to decide if, when, and how to make use of our work
13
2016 Valuation ActuarySymposium
Don Walker ASA, MAAAUpdate on SOA Research Project-Modern Deterministic Scenariosfor Interest RatesDATE August 30, 2016
Session 54: Update on SOA Research Project-Modern Deterministic Scenariosfor Interest Rates
1
MDS Project Overview
THANKS TO SOA STAFF!Ronora StrykerJan SchuhTeri Slager
AND TO PROJECT OVERSIGHT GROUP!Pam Hutchins – chairSteve CheungDuncan CookKip HeadleyEric JanecekDon WalkerZhixin WuCorwin Zass
2
MDS Project Overview
Primary Objective: Develop a set of deterministic cash flow testing scenarios that may be considered moderately adverse in varying interest rate environments
Secondary Objectives: Provide considerations in modeling inflation, investment spreads and equity returns
Research Methodology:• Identify or construct historical data series to use as basis of analysis• Perform Conditional Tail Expectation (CTE) and other statistical
analysis of historical data• Construct scenario algorithms based on the historical data analysis
3
MDS ProjectNY7 vs. Moderately AdverseThe following observations are based on analysis of the MDS Short and MDS Long rate data series
12/31/2015 Observations: Current long rates are toward the low end of the moderate historical range; current short rates are below the low end of the moderate range. Therefore:
• At 12/31/2015, level scenario IS NOT beyond moderately adverse for LONG RATES• At 12/31/2015, level scenario IS beyond moderately adverse for SHORT RATES (but not by a lot)
NY4 and NY7 Observations: A 3% pop-up or pop-down is a more than moderate change in current environment and in most environments:
• For long rates – a 3% Pop-up or 3% Pop-down is almost unprecedented in any environment• For short rates – a 3% pop-up is unprecedented when the initial rate is below 6%; a 3% pop-down is unprecedented when the
initial rate is below 9%• Only in very high rate environments would a 3% pop fall in the moderate range• Pops analyzed by comparing initial rate vs. the average rate over the subsequent 12 month period
NY2 and NY5 Observations: A 5% increase or decrease over 10 years is a more than moderate change in the current environment and in most environments:
• A 5% change over 10 years is generally a 5% tail event, with greater likelihood in higher rate environments• Clear asymmetries in historical data, with strong mean reversion tendencies in high or low rate environments
Parallel Shift Observations: Short rates exhibit significantly more volatility than long rates, so a rate change which is moderate at the short-end of the curve is more than moderate at the long end.
4
MDS Interest Rate Source Data
Challenge:• Need a data series long enough to capture range of historical variation, but recent
enough to be applicable• Robust yield data on US Treasuries only available back to early 1950s. Is this adequate
to provide long-term expectation• Earlier U.S. Government bond yield data generally only available aggregated by term• Earlier U.S. Government bond yields represent poor measure of market interest rates,
for various reasons• The U.S. did not replace the UK as the world’s dominant economic power until after
World War I, and was essentially a developing nation through the latter part of the 19th
century
5
MDS Interest Rate Source Data
6
-
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Interest Rates since 1990
20/30CMT Avg
Average 1990- Source:
20/30CMT 5.39 FRB, H.15 Selected Interest Rates
MDS Interest Rate Source Data
7
-
2.00
4.00
6.00
8.00
10.00
12.00
14.00
16.00
Interest Rates since 1953
20/30CMT Avg
Average 1990- 1953- Source:
20/30CMT 5.39 6.21 FRB, H.15 Selected Interest Rates
MDS Interest Rate Source Data
8
0
2
4
6
8
10
12
14
16
1920
1922
1924
1926
1928
1930
1932
1934
1936
1938
1940
1942
1944
1946
1948
1950
1952
1954
1956
1958
1960
1962
1964
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
Interest Rates since 1920
20/30CMT Avg US Long - Contemp
Average 1990- 1953- 1920- Sources :
20/30CMT 5.39 6.21 NA FRB, H.15 Selected Interest Rates
US Long 6.25 6.87 5.81 Measuringworth.com
MDS Interest Rate Source Data
9
0
2
4
6
8
10
12
14
16
1798
1803
1808
1813
1818
1823
1828
1833
1838
1843
1848
1853
1858
1863
1868
1873
1878
1883
1888
1893
1898
1903
1908
1913
1918
1923
1928
1933
1938
1943
1948
1953
1958
1963
1968
1973
1978
1983
1988
1993
1998
2003
2008
2013
Interest Rates since 1798
20/30CMT Avg US Long - Contemp
Average 1990- 1953- 1920- 1798+ Sources :
20/30CMT 5.39 6.21 NA NA FRB, H.15 Selected Interest Rates
US Long 6.25 6.87 5.81 5.23 Measuringworth.com
MDS Interest Rate Source Data
10
0
2
4
6
8
10
12
14
16
1798
1803
1808
1813
1818
1823
1828
1833
1838
1843
1848
1853
1858
1863
1868
1873
1878
1883
1888
1893
1898
1903
1908
1913
1918
1923
1928
1933
1938
1943
1948
1953
1958
1963
1968
1973
1978
1983
1988
1993
1998
2003
2008
2013
Interest Rates since 1798
20/30CMT Avg US Long - Contemp UK Long Contemporary
Average 1990- 1953- 1920- 1798+ Sources :
20/30CMT 5.39 6.21 NA NA FRB, H.15 Selected Interest Rates
US Long 6.25 6.87 5.81 5.23 Measuringworth.com
UK Long 5.49 6.64 5.58 4.66 Measuringworth.com
MDS Interest Rate Source Data
11
0
2
4
6
8
10
12
14
16
1729
1735
1741
1747
1753
1759
1765
1771
1777
1783
1789
1795
1801
1807
1813
1819
1825
1831
1837
1843
1849
1855
1861
1867
1873
1879
1885
1891
1897
1903
1909
1915
1921
1927
1933
1939
1945
1951
1957
1963
1969
1975
1981
1987
1993
1999
2005
2011
Interest Rates since 1729
20/30CMT Avg US Long - Contemp UK Long Contemporary
Average 1990- 1953- 1920- 1798+ 1729+ Sources :
20/30CMT 5.39 6.21 NA NA NA FRB, H.15 Selected Interest Rates
US Long 6.25 6.87 5.81 5.23 5.23 Measuringworth.com
UK Long 5.49 6.64 5.58 4.66 4.42 Measuringworth.com
MDS Interest Rate Source Data
Challenge:• Need a data series long enough to capture range of historical variation, but recent enough to
be applicable• Robust yield data on US Treasuries only available back to early 1950s. Is this adequate to
provide long-term expectation?• Earlier U.S. Government bond yield data generally only available aggregated by term• Earlier U.S. Government bond yields represent poor measure of market interest rates, for
various reasons• The U.S. did not replace the UK as the world’s dominant economic power until after World
War I , and was essentially a developing nation through the latter part of the 19th century
Solution:• Constructed two interest rate series – MDS Long Rate series (1729-2015) and MDS Short Rate
series (1825-2015)• UK government rates used through WWI• Treasury yields used for as long as reliably available – 90 day for the Short series, beginning
1931; average of 20 and 30 year for the Long series, beginning 1953• Short rates 1918-1930 – prime bankers acceptances, 90 day• Long rates 1923-1942 – Moody’s AAA less spread; 1943-1952 – US Government taxable bond
yields, 15 years+ to call or maturity
12
MDS Interest Rate Series
13
0
2
4
6
8
10
12
14
16
1729
1735
1741
1747
1753
1759
1765
1771
1777
1783
1789
1795
1801
1807
1813
1819
1825
1831
1837
1843
1849
1855
1861
1867
1873
1879
1885
1891
1897
1903
1909
1915
1921
1927
1933
1939
1945
1951
1957
1963
1969
1975
1981
1987
1993
1999
2005
2011
MDS Interest Rate Series since 1729
MDS Short Rate MDS Long Rate
Key Observations: • Long Rates historically floored above 2%. Current rates low, but not outside historical norms• Only precedent for this level of Short Rates is Great Depression/WWII – below 0.6% for 15
years and below 2% for 25 years; long rates also below 3% for 20 years• Short Rate volatility significantly greater than Long Rate volatility• Central banks began using short rate for monetary policy after WWI. Before that, inversions
were normal; after that, rare• Cycles are long and variable, with long periods near lows and short periods near highs
MDS Interest Rate Source DataIt’s Your TurnWhat do you think about the relevance of a historical data period extending back to the 1700s?
What do you think about the relevance of UK data prior to WWI?
What do you think about the proper weight of the 1970s-1980s period?
Other observations on the historical trends?
Other questions/comments?
14
MDS ProjectDefining Moderate AdversityChallenges/Questions:
• How to define “moderately adverse” and analyze historical data for moderate adversity?
• Are New York 7 moderately adverse now? Were they ever?• Is a moderately adverse rate change the same at the short and long end of the yield
curve?• Does “moderately adverse” vary with the initial interest rate environment?
15
MDS ProjectDefining Moderate AdversityConditional Tail Expectation (CTE) Framework
• CTE represents expected value of a variable, given that the value is in the defined tail. E.g. CTE70 for a stochastic CFT model with 1000 scenarios the average of the 300 worst scenarios
• CTE70 is a generally accepted measure of moderately adverse for actuarial reserving, codified in AG43 and VM20
• MDS analysis key assumptions• CTE70(interest rate input) CTE70(model output)• Empirical CTE70 analysis is predictive of future CTE70 ranges. IF the historical
data are chosen carefully• VM20 follows a similar empirical CTE70 approach in development of default rate
assumptions• Challenge in applying CTE70 to interest rates:
• 300 “worst” CFT results may reflect a mix of increasing and decreasing rate scenarios
• MDS scenarios address this challenge by calibrating to CTE15 to capture the left tail of interest rates and CTE85 to capture the right tail
16
MDS ProjectDefining Moderate AdversityData analysis frameworks – Four Analytic Bases for Scenarios
• Reversion Scenarios – based on empirical CTE15 and CTE85 measures of long-term average interest rates
• Rate Change Scenarios – based on empirical CTE15 and CTE85 measures of T-year changes in interest rates from the initial rate level
• Cyclical Scenarios – based on interest rate cycles• Academy Interest Rate Generator (AIRG) rate change targets – based on projected
CTE15 and CTE85 measures of interest rates projected using the AIRG
17
MDS ProjectModerately Adverse Reversion Targets
18
CTE05 CTE10 CTE15 CTE20 CTE30 Mean CTE70 CTE80 CTE85 CTE90 CTE95
MDS Short 0.13 0.26 0.52 0.80 1.26 3.34 5.43 5.89 6.35 7.22 8.68
MDS Long 2.35 2.48 2.59 2.67 2.80 4.11 6.15 6.89 7.56 8.52 9.94
-
2.00
4.00
6.00
8.00
10.00
12.00
MDS Interest Rate Series - CTE Analysis of Average Rates
MDS Short MDS Long
Moderately adverse long-term reversion targets: • Reversion targets capture both right tail and left tail of the interest rate distributions• Based on CTE analysis of rates over the entire period 1729-2015• For long rates, left tail very flat, meaning there are a large number of periods near minimum
values• CTE15 (low target) is approximately 0.5% for short rates and 2.5% for long rates• CTE 85 (high target) is approximately 6.5% for short rates and 7.5% for long rates
MDS Project – Recent Long Rates vs.Moderately Adverse Reversion Targets
19
0
2
4
6
8
10
12
14
16
1/1/
1980
1/1/
1981
1/1/
1982
1/1/
1983
1/1/
1984
1/1/
1985
1/1/
1986
1/1/
1987
1/1/
1988
1/1/
1989
1/1/
1990
1/1/
1991
1/1/
1992
1/1/
1993
1/1/
1994
1/1/
1995
1/1/
1996
1/1/
1997
1/1/
1998
1/1/
1999
1/1/
2000
1/1/
2001
1/1/
2002
1/1/
2003
1/1/
2004
1/1/
2005
1/1/
2006
1/1/
2007
1/1/
2008
1/1/
2009
1/1/
2010
1/1/
2011
1/1/
2012
1/1/
2013
1/1/
2014
1/1/
2015
MDS Long Rate - CTE Analysis Since 1980
Long Rate CTE15 Mean CTE85
MDS Long rates in recent periods: • Entire period 1987-2007, rates above long-term mean; above CTE85 – beyond moderately
adverse – 1980-1990.• Rates since 2008, generally below long-term mean, but not below CTE15 – not beyond
moderately adverse.
MDS Project – Recent Short Rates vs.Moderately Adverse Reversion Targets
20
MDS Short rates in recent periods: • Short Rates driven by Fed monetary policy; much greater volatility than Long Rates• Rate drops follow recessions; while long rates move modestly in these periods• Since 2008, rates below CTE15 - beyond moderately adverse. Only precedent for such rates is
Great Depression/WWII.• 1980-1990 – rates above CTE85 – beyond moderately adverse.
0
2
4
6
8
10
12
14
16
1/1/
1980
1/1/
1981
1/1/
1982
1/1/
1983
1/1/
1984
1/1/
1985
1/1/
1986
1/1/
1987
1/1/
1988
1/1/
1989
1/1/
1990
1/1/
1991
1/1/
1992
1/1/
1993
1/1/
1994
1/1/
1995
1/1/
1996
1/1/
1997
1/1/
1998
1/1/
1999
1/1/
2000
1/1/
2001
1/1/
2002
1/1/
2003
1/1/
2004
1/1/
2005
1/1/
2006
1/1/
2007
1/1/
2008
1/1/
2009
1/1/
2010
1/1/
2011
1/1/
2012
1/1/
2013
1/1/
2014
1/1/
2015
MDS Short Rate - CTE Analysis Since 1980
Short Rate CTE15 Mean CTE85
MDS Project - Moderately Adverse Rate Change Stats (yearly avg)
21
Notes on Yearly Average Rate Change Targets: • Objective is to develop CTE statistics around changes from initial rate to rate T years from
starting date, and build scenarios from these statistics• Short rate exhibits larger changes than long rate, supporting non-parallel shifts• Generally 15CTE and 85CTE changes are smaller than those in the NY7 through 10 years, but
continue much longer
-8
-6
-4
-2
0
2
4
6
8
1yr Chg 5yr Chg 10yr Chg 15yr Chg 20yr Chg 25yr Chg 30yr Chg
Rate Change CTE Statistics - Total
Long Rate - CTE15 Long Rate - CTE85 Short Rate - CTE15 Short Rate - CTE85
MDS Project - Moderately Adverse Rate Change Stats by Int Rt Grp (yearly avg)
22
Notes on Yearly Average Rate Change Targets: • Hypothesis: Both direction and magnitude of rate changes vary by the initial rate level.
Hypothesis is borne out by the data.• Significant mean reversion effect when rates are significantly high or low• Absolute magnitude of changes in rates tend to be greater when starting rates are higher• Graph shows CTE15 changes for Long Rates - Patterns for 85CTE and for Short Rates
-12
-10
-8
-6
-4
-2
0
2
4
1yr Chg 5yr Chg 10yr Chg 15yr Chg 20yr Chg 25yr Chg 30yr Chg
Long Rate Changes - CTE15 by Initial Interest Rate Group
Total 2.00-2.75 2.75-3.75 3.75-6.00 6.00-10.00 10.00+
MDS Project - NY7 scenarios 4 and 7 vs. Moderately Adverse Rate Changes
23
NY4/NY7 Notes
• Pops of +/- 3% over 1 year are very extreme historically.
• When initial rate < 6%, Long rates have never moved more than 1.25% in 1st year and Short rates have never moved more than 2.5%.
• Pops at the short end of the curve are larger than pops at the long end.
• When rates are moderate, little mean reversion seen over 1 year; but we do see mean reversion in the lowest and highest interest rate groups, in comparing the min and max pops.
Historical CTE Level for:Rate Groupings
NY7 - Pop-down 3%
NY4 - Pop-up 3% Min Pop Max Pop
Short Rate Initial Level:Total CTE02 CTE98 (4.70) 6.46 0.00-1.00 < CTE01 > CTE99 (0.79) 1.88 1.00-3.00 < CTE01 > CTE99 (1.77) 1.66 3.00-6.00 < CTE01 > CTE99 (2.33) 2.06 6.00-10.00 < CTE01 CTE93 (3.35) 6.46 10.00+ CTE40 CTE93 (4.70) 4.35 Long Rate Initial Level:Total < CTE01 > CTE99 (3.04) 2.64 2.00-2.75 < CTE01 > CTE99 (0.04) 1.36 2.75-3.75 < CTE01 > CTE99 (0.84) 1.31 3.75-6.00 < CTE01 > CTE99 (1.23) 1.24 6.00-10.00 < CTE01 > CTE99 (2.03) 2.64 10.00+ < CTE01 > CTE99 (3.04) 2.30 Note: Based on analys is of long and short treasury rates , ini tia l
da i ly va lues compared to subsequent 1 year average.
MDS Project - NY7 scenarios 2 and 5 vs. Moderately Adverse Rate Changes
24
NY2/NY5 Notes
• Movements of +/- 5% over 10 years are fairly extreme historically.
• When initial rate < 6%, Long rates have never moved more than 1.25% in 1st year and Short rates have never moved more than 2.5%.
• 10 year changes at the short end of the curve are larger than pops at the long end.
• Considerable mean reversion seen at the extremes. When long rates >10%, the maximum 10 year increase is a decrease of 2.73%. When Short rates < 0.50%, the maximum 10 year decrease is an increase of 0.10%.
Historical CTE Level for:Rate Groupings
NY5 - Down 5% over 10 yrs
NY2 - Up 5% over 10 yrs
Min 10 yr Chg
Max 10 yr Chg
Short Rate Intial Level:Total CTE07 CTE96 (8.66) 9.71 0.00-0.50 < CTE01 > CTE99 0.10 2.24 0.50-2.50 < CTE01 > CTE99 (2.44) 3.56 2.50-4.00 < CTE01 > CTE99 (3.42) 4.29 4.00-6.00 CTE07 CTE91 (5.68) 9.71 6.00-15.00 CTE50 CTE99 (8.66) 5.00 Long Rate Initial Level:Total CTE01 CTE98 (5.44) 7.46 2.00-2.75 < CTE01 > CTE99 (0.74) 1.82 2.75-3.75 < CTE01 > CTE99 (1.81) 3.92 3.75-6.00 < CTE01 CTE92 (2.68) 7.46 6.00-10.00 < CTE01 > CTE99 (3.31) 4.38 10.00+ CTE70 > CTE99 (5.44) (2.73) Note: Based on analys is of MDS Long and MDS Short interest rate
series , yearly average for year T compared with yearly average
for year T+10.
MDS ProjectDefining Moderate AdversityChallenges/Questions:
• How to define “moderately adverse” and analyze historical data for moderate adversity?• Are New York 7 moderately adverse now? Were they ever?• Is a moderately adverse rate change the same at the short and long end of the yield curve?• Does “moderately adverse” vary with the initial interest rate environment?
Solutions/Answers:• Use generally accepted CTE70 definition of moderately adverse. Considering both tails, this
translates to CTE15 for the left tail and CT85 for the right tail• Reversion scenarios: Scenarios that revert to 15CTE and 85CTE-based reversion targets
capturing left tail and right tail• Rate change scenarios: Scenarios whose T-year change from the starting rate are based on
CTE analysis of the historical T-year changes in rates, varying by initial rate• Cyclical scenarios: Scenarios that project a cycle of interest rates• AIRG scenarios: Scenarios based on similar CTE analysis of stochastic scenarios generated by
the Academy Interest Rate Generator
25
MDS Project - Defining Moderate Adversity – It’s Your TurnDoes CTE70 applied to model inputs imply CTE70 on model outputs?
Is CTE70 an appropriate threshold for calibrating moderately adverse in a deterministic framework?
Is it reasonable to bifurcate the tail of interest rates into CTE15 and CTE85 to approximate CTE70 overall?
Other questions/comments?
26
MDS ProjectScenario ConstructionChallenges/Questions:
• How to turn a set of historical statistics into projection scenarios?• What is the right number of scenarios? Is there a right number of scenarios?• What dynamics should be captured in the scenarios?• With parameters that vary by interest rate grouping, how to avoid discontinuities at the
boundaries of the rate groupings?• What is the right balance between simplicity and reasonable reproduction of historical
patterns?
27
MDS ProjectScenario ConstructionData analysis frameworks revisited – Four Analytic Bases for Scenarios
• Reversion Scenarios – based on empirical CTE15 and CTE85 measures of long-term average interest rates
• Rate Change Scenarios – based on empirical CTE15 and CTE85 measures of T-year changes in interest rates from the initial rate level
• Cyclical Scenarios – based on interest rate cycles• Academy Interest Rate Generator (AIRG) rate change targets – based on projected
CTE15 and CTE85 measures of interest rates projected using the AIRG
28
MDS ProjectMDS Scenario Set – Reversion Scenarios
29
Reversion Scenario Construction Notes: • Eight reversion scenarios - Grade to either high or low reversion target over 15 years• To capture different risks, some scenarios employ a 5-year delay and/or an initial pop-up/down• Reversion targets
Short Rate Long RateLow CTE target 0.50% 2.60%High CTE target 6.25% 7.50%
Scenario Number Scenario Name Scenario DescriptionReversion Scenarios:MDS1 Reversion - High Grade linearly to an 85CTE (right tail) reversion target over a 15 year periodMDS2 Reversion - Low Grade linearly to a 15CTE (left tail) reversion target over a 15 year periodMDS3 Delayed Reversion - High Long and Short Rates level for 5 years, then grade linearly to 85CTE reversion
target over a 10 year periodMDS4 Delayed Reversion - Low Long and Short Rates level for 5 years, then grade linearly to 15CTE reversion
target over a 10 year periodMDS5 Pop-up with Reversion - High Initial pop-up, then Grade linearly to 85CTE reversion target by year 15MDS6 Pop-down with Reversion - Low Initial pop-down, then Grade linearly to 15CTE reversion target by year 15MDS7 Delayed Pop-up with Reversion - High Long and short rates level for 5 years followed by pop-up, then Grade
linearly to 85CTE reversion target by year 15MDS8 Delayed Pop-down with Reversion - High Long and short rates level for 5 years followed by pop-down, then Grade
linearly to 15CTE reversion target by year 15
MDS Project - MDS Scenario Set Reversion Scenarios 12/31/15
30
12/31/15 Reversion Scenario Notes: • Ultimate high rate comparable to NY2, higher than NY4• Ultimate low rate lower than level, but higher than NY5 and NY7• Changes much more gradual than NY7 change, and pops much smaller
- 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00
MDS Long Rate - Reversion Scenarios -High
MDS1 MDS3 MDS5 MDS7
-
0.50
1.00
1.50
2.00
2.50
3.00
ProjYear
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
MDS Long Rate - Reversion Scenarios -Low
MDS2 MDS4 MDS6 MDS8
MDS Project - MDS Scenario SetRate Change, Cyclical and AIRG Scenarios
31
Rate Change Scenarios:MDS9 85CTE Rate Changes Change from initial rate based on 85CTE (right tail) historical change statistics
for the applicable interest rate groupMDS10 15CTE Rate Changes Change from initial rate based on 15CTE (left tail) historical change statistics
for the applicable interest rate groupMDS11 85CTE Change w Transitional Pop-up Change from initial rate based on 85CTE (right tail) historical change statistics
for the applicable interest rate group, with initial pop-up based on 85CTE transitional change
MDS12 15CTE Change w Transitional Pop-down Change from initial rate based on 15CTE (left tail) historical change statistics for the applicable interest rate group, with initial pop-down based on 15CTE transitional change
MDS13 Cyclical, 20 year cycle Change from initial rate based on 15CTE (left tail) historical change statistics for the applicable interest rate group, with initial pop-down based on 15CTE transitional change
MDS14 Cyclical, 40 year cycle Change from initial rate based on 15CTE (left tail) historical change statistics for the applicable interest rate group, with initial pop-down based on 15CTE transitional change
AIRG Scenarios:MDS15 AIRG 85CTE Rates based on 1000 scenarios from Academy interest rate generator, 85CTE
of cumulative average rates, annualized.MDS16 AIRG 15CTE Rates based on 1000 scenarios from Academy interest rate generator, 15CTE
of cumulative average rates, annualized.
MDS Project - MDS Scenario SetRate Change and AIRG Scenarios
32
Rate Change Scenario Construction Notes: • Scenarios MDS9-MDS12 are based on historical rate change CTE data• MDS11 and MDS12 include an initial pop-up/down; MDS9 and MDS10 don’t
Cyclical Scenario Construction Notes:• Scenarios MDS13 and MDS14 reflect a cyclical pattern of rates
• MDS13 is a 20 year cycle with less extreme rates• MDS14 is a 40 year cycle with more extreme rates
• Since 1729, only about 6 cycles, with no two similar to each other. As such, cyclical scenarios are more judgmental
AIRG Scenario Construction Notes:• AIRG scenarios apply the rate change CTE analysis to a set of 1000 stochastic scenarios
generated using the Academy Interest Rate Generator• MDS15 is the high AIRG scenario, using CTE85• MDS16 is the low AIRG scenario, using CTE15
MDS Project - MDS Scenario SetRate Chg and AIRG Scenarios 12/31/15
33
12/31/15 Rate Change and AIRG Scenario Notes: • Low scenarios are in a tight range due to low starting point• High scenarios have varied patterns, capturing a range of risks• Changes generally much more gradual than NY7 changes• Peaks in high rate scenarios comparable to the NY2 ultimate rate of 7.84%• Low MDS scenarios more moderate than NY5 and NY7 ultimate rate of 1.96%• AIRG high rate scenario is milder than the others. AIRG low rate scenario is comparable to
others
- 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00
ProjYear
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
MDS Long Rate - Rt Chg CTE Scenarios -High
MDS9 MDS11 MDS13 MDS14 MDS15
-
0.50
1.00
1.50
2.00
2.50
3.00
3.50
ProjYear
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
MDS Long Rate - Rt Chg CTE Scenarios -Low
MDS10 MDS12 MDS16
MDS Project – MDS Scenario SetComplete Scenarios 12/31/15
34
12/31/15 Complete Scenario Notes: • NY2, NY4, NY5, NY7 rates for comparison:
NY2 NY4 NY5 NY7Short 5.16% 3.16% 0.00% 0.00%Long 7.84% 5.84% 1.96% 1.96%
• Long Rates – both high and low MDS scenarios milder than NY Seven. Low rates never drop as low as NY5 and NY7; high rates reach NY2 levels, but more slowly
• Short Rates – Low MDS scenarios somewhat milder than NY5 and NY7; high MDS scenarios more extreme than NY2 and NY4
• AIRG scenarios milder than historically-based scenarios, except for Long Rate low scenario.
-
2.00
4.00
6.00
8.00
ProjYear
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
MDS Short Rate Scenarios
MDS1 MDS2 MDS3 MDS4
MDS5 MDS6 MDS7 MDS8
MDS9 MDS10 MDS11 MDS12
MDS13 MDS14 MDS15 MDS16
-
2.00
4.00
6.00
8.00
ProjYear
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
MDS Long Rate Scenarios
MDS1 MDS2 MDS3 MDS4
MDS5 MDS6 MDS7 MDS8
MDS9 MDS10 MDS11 MDS12
MDS13 MDS14 MDS15 MDS16
MDS Project - Key ObservationsMDS Scenarios and NY7 Scenarios• NY7 scenarios were: a) calibrated to the environment of the 1970s-1980s;
and b) developed before actuaries had developed a statistical framework to define moderately adverse
• Historically, speed of NY7 rate changes is beyond moderately adverse, even in the context of 1970s-1980s interest rates
• Historically, a +/- 5% range in interest rates is beyond moderate in most environments, at least for long rates
• Short rates exhibit much greater historical volatility than long rates, so parallel shifts tend to over-stress long rates or under-stress short rates
• Interest rate changes vary with the level of interest rates, so moderately adverse scenarios should consider the starting rate environment
• Impact of starting rate levels on the magnitude of projected interest rate changes• Impact of starting rate levels on the direction of projected interest rate changes – i.g.
impact of mean reversion
• MDS scenarios are constructed to address these issues
35
MDS Project - Key ObservationsMDS Scenario Construction• In developing projection scenarios from historical rate analysis, one must
consider: a) A long enough period to be statistically valid; and b) The applicability of the rates over the entire historical period
• MDS scenarios consider period back to early 1720s, the period when some authors consider major government debt to have been effectively risk free. This period also covers all of the industrial revolution and the period leading up to it
• MDS scenarios consider UK rates up through World War II, when UK was the world’ leading economic power and the US was a developing economy
• The period from the late 1970s to the late 1980s was a historically extraordinary period, and it is dangerous to give it too much (or too little weight in calibrating expectations for future interest rates
• Even 287 years of data provides a limited data set. Converting historical data analysis into scenarios requires a great deal of judgment, smoothing, and reasonableness-checking
36
MDS Project - Key Observations12/31/2015 MDS ScenariosAs of 12/31/2015:
• Low rate scenarios - The level scenario was at the low end of moderately adverse for long rates, but beyond moderately adverse for short rates
• High rate scenarios – Increases of 5% over 15-20 years are consistent with moderately adverse
• Beyond 15 years, AIRG scenarios appear to be mild relative to historical patterns
37
Now it’s really your turn!
The floor is open.
38
2016 Valuation ActuarySymposium
Mark E. Alberts, FSA, MAAAUpdate on SOA Research Project-Modern Deterministic Scenariosfor Interest RatesAugust 30, 2016