Core Indicators, Reports, Negotiated Performance Targets, and Perkins Funds Regional Workshops
Rule-based management of cash flows in Income Plus Funds€¦ · 06 A withdrawal example. 01...
Transcript of Rule-based management of cash flows in Income Plus Funds€¦ · 06 A withdrawal example. 01...
PRESENTED BY:
Rule-based management of cash flows in Income Plus Funds
Keith Forbes
Fixed Income Quantitative Analyst
4 June 2020
Disclaimer
Fairtree Asset Management (Pty) Ltd is registered as a financial services provider with the Financial Services
Board of South Africa, with registration number 2004/033269/07 and FSP number 25917.
A schedule of fees, charges and maximum commissions, as well as a detailed description on how performance fees are
calculated and applied, is available on request from the manager of each fund (“the Manager”), being either Sanne
Management Company (RF) (Pty) Ltd, Realfin Collective Investment Schemes (RF) (Pty) Ltd, Prescient Management
Company (RF) (Pty) Ltd or Boutique Collective Investments (RF) (Pty) Ltd, all being registered and approved managers of
Collective Investment Schemes. The name of the fund shall reflect the name of the approved manager of the fund.
Additional information, including key investor information documents, minimum disclosure documents as well as other
information relating to the portfolio is available, free of charge, on request from the Manager
The Manager retains full legal responsibility for any co-named portfolio and is responsible for the appointment of a trustee
in accordance with the provisions of the Collective Investment Schemes Control Act, 45 of 2002.
We believe the information displayed is accurate and reliable, but no warranty of accuracy or reliability is given and no
responsibility arising in any way for errors and omissions (including by way of negligence) is accepted by Fairtree Asset
Management (Pty) Ltd.
This information is not intended to provide advice to, or take into account individual investors’ objectives or
circumstances. This material should not be construed to represent a solicitation to invest in the portfolio and is disclosed
for reporting purposes only.
Collective Investment Schemes are generally medium to long-term investments. Please note that past performance is no
guarantee of future performance and that the value of participatory interests may go down as well as up. Collective
investments are traded at ruling prices and can engage in scrip lending and borrowing.
A Collective Investment Scheme may be closed to new investors in order for it to be managed more efficiently in
accordance with its mandate. The Manager does not provide any guarantee with respect to the capital or the return of the
portfolio. Excessive withdrawals from the portfolio may place the portfolio under liquidity pressure and in such
circumstances, a process of ring-fencing of withdrawal instructions and managed pay-outs over time may be followed.
Commission and incentives may be paid, and if so, are included in the overall costs. Investors should note that the value
of an investment is dependent on numerous factors which may include, but not limited to, share price fluctuations, interest
and exchange rates and other economic factors. Performance is further affected by uncertainties such as changes in
government policy, taxation and other legal or regulatory developments.
The performance of the portfolio is dependent on the making of correct assessments of the price movements of individual
securities and other investments. Financial markets have historically exhibited high levels of volatility and negative
movements that have affected the price of all assets within a specific class. The portfolio’s investments will thus be
subject to market risk. Through financial gearing via the long/short process, the portfolio may be leveraged. This will
mean enhanced positive gains but conversely can mean magnified losses. No taxation has been deducted in the
computation of returns. The taxation treatment of returns is the investor’s responsibility.
All returns are disclosed net of performance fees.
The Fairtree Global Flexible Income Plus fund is a sub-fund of Prescient Global Funds PLC and is managed by Prescient
Fund Services (Ireland) Limited. Prescient Fund Services (Ireland) Limited is authorised in Ireland and regulated by the
Central Bank of Ireland. The Fairtree Global Flexible Income Plus fund is approved as a foreign collective investment
scheme under Section 65 of the Collective Investment Schemes Control Act 45 of 2002 by the Financial Sector
Conduct Authority of South Africa. Fairtree Asset Management (Pty) Ltd (FSP: 25917) is the investment manager of the
Fairtree Global Flexible Income Plus fund.
Overview
01 Motivation
02 Risk, return and correlation
03 Portfolio strength
04 Mathematical optimisation
05 Our implementation
06 A withdrawal example
MOTIVATION01
Our situation
▪ We manage 14 credit funds with different risk / return targets
▪ Fund flows alone will cause the funds to differ
▪ The funds are also governed by different legislation that imposes different constraints on
portfolio composition
▪ Positions tend to be illiquid bonds that are usually acquired at issuance and held to
maturity
Our situation
▪ Our funds are all different from one another, and the assets are such that they cannot
necessarily be bought / sold in the marketplace
▪ A well-diversified portfolio must be built up over time
▪ One cannot simply acquire all desired assets at a single point in time
Software is used to compute optimal trade sizes in three manners
Rule-based appetite, allocation and withdrawal
WITHDRAWAL
Remove the assets that are
seen as the least useful for the
fund and determine whether
they can benefit other funds
under management
(re-allocation)
APPETITE
Determines how much of an
asset to bid for based on the
funds’ appetite for the
security
ALLOCATION
Once an issuance is
acquired, allocate it across
the funds in a way that
provides maximum benefit to
all funds
Sizing positions
▪ Our approach is designed to treat customers fairly by algorithmically sizing portfolio positions
▪ This avoids manager bias, where one fund may inadvertently be favoured over another or the
manager’s risk preference becomes confused with the client’s
▪ It allows us to optimally move each fund towards its unique target keeping, ex ante Sharpe
ratios as high as possible and fulfilling all regulatory constraints
Sizing matters
▪ Simply thinking buy / sell / hold is not enough
▪ There is an optimal size for each position – too large or too small a position will lead to risk
and return characteristics that are not as good as they could be
▪ We need to be able to measure the effects of sizing. That is, we must compute the portfolio
distribution
▪ The portfolio distribution is the ex ante probabilistic distribution of return outcomes
Why compute portfolio distributions?
▪ An investor is ultimately concerned with portfolio performance, not the performance of
individual positions per se
▪ Heuristics and rules-of-thumb are sub-optimal
▪ Instruments interact in complex non-obvious manners. Intuition is our enemy – the maths
must be done
▪ Knowing the portfolio distributions allows us to optimise the risk targeting and
diversification within each fund in a common framework
Risk targeting example
Proportion of target risk achieved
Qualit
y o
f risk (
Div
ers
ity)
Goal: (1,1)
100%
Perfect diversification:
all unpriced risk removed
too much risktoo little risk
Fund C
Fund BFund A
100%
Risk targeting example - an instrument matures
Proportion of target risk achieved
Qualit
y o
f risk (
Div
ers
ity)
Goal: (1,1)
100%
Perfect diversification:
all unpriced risk removed
too much risktoo little risk
Fund C
Fund BFund A
100%
Risk targeting example – new issuance allocated
Proportion of target risk achieved
Qualit
y o
f risk (
Div
ers
ity)
Goal: (1,1)
100%
Perfect diversification:
all unpriced risk removed
too much risktoo little risk
Fund C
Fund BFund A
100%
RISK, RETURN &
CORRELATION02
Moments of portfolio return distributions
Mean of the return distribution:
Variance of the return distribution:
An even-handed model for risk and return
▪ Driven by current market levels rather than backward-looking
▪ No position is considered to be inherently more favourable than any other
▪ More specifically, we use a CAPM-like approach in which the undiversifiable risk in every
position is priced equivalently
▪ Risk and return come directly from the distributions implied by the spread
Multi-sector hierarchical model (illustrative)
▪ More closely related obligors, have higher pairwise correlation
▪ Model proposed by Gregory & Laurent (2004)
Moments of portfolio return distributions
Mean of the return distribution:
Variance of the return distribution:
Returns, risk, correlation – model summary
▪ Positions have expected returns – this is estimated from the par spread (the expected return
is the mean of the return distribution)
▪ Positions have risk – this is also estimated from the par spread (the risk is the standard
deviation of the return distribution)
▪ Position P & L is correlated with other positions – this is imposed by the hierarchical
correlation model
BCI Income Plus positions on 08 May 2020
NPV
Par
Spre
ad
BCI Income Plus on 08 May 2020 – Vanilla Bonds
Ex ante Sharpe
Expecte
d R
etu
rn
BCI Income Plus on 08 May 2020 – Vanilla Bonds & Sector Sub-Portfolios
Ex ante Sharpe
Expecte
d R
etu
rn
BCI Income Plus on 08 May 2020 – Vanilla Bonds & ABS’s
Ex ante Sharpe
Expecte
d R
etu
rn
BCI Income Plus on 08 May 2020
Ex ante Sharpe
Expecte
d R
etu
rn
BCI Income Plus on 08 May 2020
Ex ante Sharpe
Expecte
d R
etu
rn
PORTFOLIO
STRENGTH03
Twin goals
▪ An investment manager aims on having -
1. Ex-ante Sharpe ratios as high as possible, and
2. Actual risk as close as possible to target risk
The dilemma
28
Which is better ?
1. Obtaining a high Sharpe with tiny
risk
2. Or a mediocre Sharpe with on-
target risk?
Of course the answer depends on the
actual numbers
In order to answer these questions and to
allocate positions to a fund or to a suite
of funds, it is necessary to specify a single
number that somehow combines
distance from the risk target with the
distance from best-case Sharpe ratio
Moments of portfolio return distributions
Portfolio strength:
Research
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3407419
Sensitivity to position size
Position
Port
folio
Str
ength
Sensitiv
ity to P
ositio
n S
ize
Sensitivity to position size
Position
Port
folio
Str
ength
Sensitiv
ity to P
ositio
n S
ize
Sensitivity to position size
Position
Port
folio
Str
ength
Sensitiv
ity to P
ositio
n S
ize
MATHEMATICAL
OPTIMISATION04
2-Dimensional optimisation problem (illustrative)
Maximisation in n dimensions
▪ BCI Income Plus has 167 positions – this means our withdrawal problem involves finding the
maximum of a 167-dimensional surface
▪ Although we cannot visualise high-dimensional functions, we know that if they are smooth,
then they are locally approximately quadratic (think of a parabola / paraboloid)
▪ If we know our surface’s gradient (first derivatives) and curvature (second derivatives) then
we know the quadratic’s formula and can step directly to its maximum
Maximisation in n dimensions
▪ This maximum will in general be a good approximation to our surface’s maximum, but we
can obtain an even better approximation by repeating the procedure iteratively until
improvements are negligible
▪ The modern optimisation algorithms that we use are extensions of this idea
1-Dimensional optimisation problem (illustrative)
Position Size
Port
folio
Str
ength
1-Dimensional optimisation problem (illustrative)
Position Size
Port
folio
Str
ength
1-Dimensional optimisation problem (illustrative)
Position Size
Port
folio
Str
ength
1-Dimensional optimisation problem (illustrative)
Position Size
Port
folio
Str
ength
1-Dimensional optimisation problem (illustrative)
Position Size
Port
folio
Str
ength
1-Dimensional optimisation problem (illustrative)
Position Size
Port
folio
Str
ength
1-Dimensional optimisation problem (illustrative)
Position Size
Port
folio
Str
ength
1-Dimensional optimisation problem (illustrative)
Position Size
Port
folio
Str
ength
OUR
IMPLEMENTATION05
COPS Software – Our implementation
▪ COPS stands for Constrained Optimisation of Position Sizes
▪ COPS determines the optimal position sizes for new allocations and withdrawals across our
suite of credit funds
▪ We have derived analytical solutions for the objective function (portfolio strength) and its
gradients including modelling special cases such as CLNs and ABSs
▪ Legal constraints are specified mathematically
COPS Software – Our implementation
▪ Function and gradients that are used as input to a non-linear constrained optimisation library
(namely MIT’s NLOPT library)
▪ Monte-Carlo sampling methods are not used. We use closed-form analytical solutions to
ensure speed and smoothness of the objective function
▪ The mathematics is coded in C++ with a Mathlab frontend. An object-orientated approach is
followed
Portfolio mathematics
Portfolio strength gradient (chain rule):
Sharpe from hazard rate and spread:
Accounting for ABS effects: Smoothed UCITS constraint:
ABS effects on gradients:UCITS gradient:
Implementation - ~ 25, 000 lines of C++ and Mathlab
EXAMPLE06
Withdrawal example
Position #
Am
ount
Withdrawal example
Position #
Am
ount
Withdrawal example
Position #
Am
ount
Withdrawal example
Withdrawal example
Reallocation of non-cash assets to other funds
portfolio strength
before after improvement
Fund 1 0.7508 0.9313 0.1805
Fund 2 1.3162 1.3706 0.0545
Fund 3 1.2883 1.3256 0.0374
Fund 4 1.2828 1.3063 0.0235
Fund 5 1.2355 1.2355 0.0000
Fund 6 1.0892 1.1582 0.0689
Fund 7 1.1817 1.2015 0.0198
ABKS1 AGL03 FRB24 FRC319 IV035 IV046 NGLT1B SBT103 SBT206
Fund 1 6,000,000 0 0 0 0 4,000,000 17,000,000 2,000,000 0
Fund 2 0 0 2,000,000 0 2,000,000 5,281,546 1,000,000 8,000,000 0
Fund 3 0 0 0 5,000,000 0 0 0 0 0
Fund 4 0 0 0 10,000,000 0 0 0 0 0
Fund 5 0 0 0 0 0 0 0 0 0
Fund 6 0 6,000,000 0 0 0 3,000,000 2,000,000 1,000,000 0
Fund 7 0 5,000,000 0 0 0 1,000,000 1,000,000 0 0
Summary
60
We use our in-house COPS
software to optimally size
instruments across all funds
during withdrawals and
allocations
THE CONCEPT OF PORTFOLIO STRENGTH ALLOWS
US TO MEASURE THE STATE OF OUR PORTFOLIOS
WITH RESPECT TO BOTH DIVERSIFICATION AND
RETURN TARGETING
Algorithmic allocation and withdrawal ensures that customers are treated fairly and that portfolios are
optimally positioned with respect to diversification and return targeting
The portfolios strength of our suite of funds has improved from withdrawals. This is because
withdrawals allow us to optimally dispose of and re-allocate assets
Thank You