Norwegian School of Economics

Bergen, Fall 2021

Analysis of Autocallable Notes

Who reap the benefits? The issuer, facilitator, distributor, or the investor?

Andreas Garborg Sie and Jonas Blom Helmersen

Supervisor: Petter Bjerksund

Master thesis, Economics and Business Administration

Major: Financial Economics

NORWEGIAN SCHOOL OF ECONOMICS

This thesis was written as a part of the Master of Science in Economics and Business
Administration at NHH. Please note that neither the institution nor the examiners are
responsible – through the approval of this thesis – for the theories and methods used, or
results and conclusions drawn in this work.
 i

Acknowledgements

This master thesis is written to mark the end of our master’s degree at Norwegian School
of Economics (NHH). During our studies, we have enrolled in classes thought by Jørgen
Haug and Svein Arne Persson that have awoken our interest of the derivatives topic. This
spark, together with great conversations with our supervisor Petter Bjerksund eventually
led to this master thesis.

During the process of investigating the complex autocallables, we have faced and overcome
numerous challenges that have proven to be very educational. We encountered new
advanced derivatives concepts and met challenges regards to financial engineering. This
have led us to get a better foundation of the derivatives field, in addition to become better
at programming. We hope, and strongly believe, that we will benefit from the acquired
knowledge after our graduation.

We want to thank our supervisor Petter Bjerksund who has been very helpful offering
great discussions and good feedback throughout the process. We will also thank our
families and friends for the continuous support.

The thesis assumes basic knowledge of option pricing theory.

Norwegian School of Economics

Bergen, December 2021

Andreas Garborg Sie Jonas Blom Helmersen

ii

Abstract
Autocallable structured products are complex instruments incorporating features and
conditions that make them difficult to assess for potential investors. Despite this, they
have become one of the most popular structured products in Norway. A potential reason
is that many investors believe that these notes offer a high fixed coupon combined with
limited risk. Finance experts do not share this belief suggesting that investors of these
products are either ignorant or idiots.

In this thesis we analyse two autocallable notes offered in the Norwegian market. Our
analysis suggest that investors pay price premiums of ca. 50% relative to the present
value of the notes. The products have a high probability of negative and strong negative
returns. In addition, the sales documents of these products appear biased. Based on this,
it seems like the benefits of these notes are not reaped by the investor, but by the issuer,
facilitator, and distributor.

Keywords – Autocalls, Structured products, Geometric Brownian motion, NHH

Contents iii

Contents
1 Introduction 1

2 Structured Products 3
2.1 Structured products in Norway . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Autocallable structured products . . . . . . . . . . . . . . . . . . . . . . 5
2.2.1 Features of autocallable structured products . . . . . . . . . . . . 6
2.2.2 Autocall Norske Selskaper and Autocall Telekom . . . . . . . . . 8

3 Valuation model for autocallable notes 10

3.1 The Black & Scholes model . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 Monte Carlo method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Geometric Brownian motion . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3.1 Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3.2 GBM under the risk-neutral framework . . . . . . . . . . . . . . . 14
3.3.3 Implicit dividends in GBM . . . . . . . . . . . . . . . . . . . . . . 14
3.3.4 Correlated geometric Brownian motion . . . . . . . . . . . . . . . 15
3.4 Payout function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4 In-data parameters 21
4.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Risk-free rate and discount rate . . . . . . . . . . . . . . . . . . . . . . . 22
4.3 Expected rate of return . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.4 Implicit dividend rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.5 Volatility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.6 Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

5 Results from valuation model 27

5.1 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.2 Model reliability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.2.1 Numerical versus theoretical price . . . . . . . . . . . . . . . . . . 31
5.2.2 Simulated versus historical correlation . . . . . . . . . . . . . . . 32
5.3 Autocall Norske Selskaper . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.3.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.3.2 Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.4 Autocall Telekom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.4.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.4.2 Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.5 Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

6 Expected return and probability distribution 44

6.1 From risk-neutral to the true probability measure . . . . . . . . . . . . . 44
6.1.1 CAPM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
6.2 Autocall Norske selskaper . . . . . . . . . . . . . . . . . . . . . . . . . . 45
6.3 Autocall Telekom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.4 Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
iv Contents

7 Analysis of prospect 53
7.1 Market participants and their role . . . . . . . . . . . . . . . . . . . . . . 53
7.2 Final terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
7.3 Sales brochure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7.4 Our thoughts of the supporting documents . . . . . . . . . . . . . . . . . 58

8 Discussion 60
8.1 Assessment of research questions . . . . . . . . . . . . . . . . . . . . . . 60
8.1.1 What is the fair price of the products? . . . . . . . . . . . . . . . 60
8.1.2 What can investors expect in returns from the products? . . . . . 61
8.1.3 How are the products presented to potential investors? . . . . . . 62
8.1.4 Who reap the benefits? . . . . . . . . . . . . . . . . . . . . . . . . 63
8.2 Strengths and weaknesses . . . . . . . . . . . . . . . . . . . . . . . . . . 63

9 Conclusion 65

References 66
List of Figures v

List of Figures
2.1 Outstanding volume of structured products in Norway 2000-2009 . . . . . 4
2.2 Discrete and continuous call dates . . . . . . . . . . . . . . . . . . . . . . 7
3.1 One GBM stock path. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.1 Figure illustrating simulated vs. historical correlations . . . . . . . . . . 33
5.2 One path generation of Autocall Norske Selskaper ’s four underlying assets 35
5.3 Sensitivity analysis of volatility and correlation, Autocall Norske Selskaper 37
5.4 Sensitivity analysis of drift, Autocall Norske selskaper. . . . . . . . . . . 38
5.5 Sensitivity analysis of coupon rate, Autocall Norske selskaper. . . . . . . 38
5.6 One path generation of Autocall Telekom’s four underlying assets. . . . . 40
5.7 Sensitivity analysis of volatility and correlation, Autocall Telekom . . . . 42
5.8 Sensitivity analysis of drift, Autocall Telekom. . . . . . . . . . . . . . . . 42
5.9 Sensitivity analysis of coupon rate, Autocall Telekom. . . . . . . . . . . . 43
6.1 Distribution of simulated HPR, Autocall Norske Selskaper . . . . . . . . 46
6.2 Analysis of investment length, Autocall Norske Selskaper . . . . . . . . . 47
6.3 Distribution of simulated HPR, Autocall Telekom . . . . . . . . . . . . . 50
6.4 Analysis of investment length, Autocall Telekom . . . . . . . . . . . . . . 50
vi List of Tables

List of Tables
2.1 Main feature of Autocall Norske Selskaper and Autocall Telekom . . . . . 8
3.1 Table illustrating payoff of one scenario of a autocallable note . . . . . . 20
5.1 Covariance matrices of the underlying assets . . . . . . . . . . . . . . . . 29
5.2 Annual and continuous dividend yields . . . . . . . . . . . . . . . . . . . 30
5.3 Numerical results of a European call option . . . . . . . . . . . . . . . . 32
5.4 Historical and simulated correlation matrices of underlying assets in Autocall
Norske Selskaper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.5 Parameters used to model paths for underlying assets in Autocall Norske
selskaper. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.6 Payoff specific parameters used in the valuation of Autocall Norske Selskaper 35
5.7 Results from valuation model - Autocall Norske Selskaper . . . . . . . . . 36
5.8 Parameters used to model paths for underlying assetsin Autocall Telekom 39
5.9 Correlation matrix of underlying assets in Autocall Telekom . . . . . . . 39
5.10 Payoff specific parameters used in the valuation of Autocall Telekom. . . 40
5.11 Results from valuation model - Autocall Telekom . . . . . . . . . . . . . 41
6.1 Parameters and expected returns of underlying assets in Autocall Norske
selskaper. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.2 Distirbution of IRR - Autocall Norske Selskaper . . . . . . . . . . . . . . 48
6.3 Parameters and expected returns of underlying assets in Autocall Telekom 49
6.4 Distribtion of IRR - Autocall Telekom . . . . . . . . . . . . . . . . . . . . 51
 1

1 Introduction
Autocallable structured products are a type of structured products that have become
increasingly popular the past decade. Common for all structured products is that they are
a package consisting of many financial instruments. Structured products have evolved over
time resulting in a variety of products with differences in characteristics and complexity.
Autocallables however, are a modern type of such structured products, with potentially
more complex features, which make them difficult to understand. Despite this, they have
become popular among private investors.

Autocallables as an investment alternative have received much criticism in Norway. Most

of the criticism is based on the complexity of the products. Firstly, the complexity makes
it hard for investors to understand how the products work and what they can expect
in terms of returns. Secondly, the complexity makes it difficult for investors to know
the true value of the products. Various experts have warned investors of these products.
Despite this, there seems to be an attractive market for autocallable structured products
in Norway as new notes are frequently offered. Garantum who is a broker of structured
products claimed to have over 30,000 customers who in total had invested over 80 billion
Swedish kronor in 2018 (Bjørklund, 2018).

In this thesis we will see if the criticism of these products can be justified. Our main goal
is to find out who benefits from the autocallable notes. Is it the issuer who constructs
the product? The facilitator and distributor who act as brokers? The investor? Or all of
them? To give a good answer we will research two specific autocallable notes and answer
the following questions:

1. What are the fair prices of these products?

2. What can investors expect in returns from these products?

3. How are the products presented to potential investors?

2

Outline
Chapter 2 of the thesis will present traditional structured products and autocallable
structured products. In the discussion of autocallable notes we will present various
features to illustrate how these notes might differ. The goal is to give the reader a clear
intuition of how these products work and how they can generate return or losses for
investors. We will also present the specific notes that we will value in this thesis.

Chapter 3 will present the valuation model used to value the autocallable notes. This
chapter will cover the option pricing theory applied in the model and illustrate how the
valuation model works. Chapter 4 will present the data and parameters used to value the
notes.

Chapters 5, 6, and 7 will analyse the autocallable notes. In chapter five we will value the
notes and compare the fair price derived from our valuation model with the actual price
of the notes. Chapter six will identify expected returns and returns distributions of the
notes. Whereas chapter seven will look at how the notes are sold in the Norwegian market
with great focus on the marketing material used.
 3

2 Structured Products
Structured products are typically composed of three elements. An obligation, a derivatives
product, and lastly, one or more underlying assets. Issuers of structured products package
these three components to make a single structured product (Osphare-Druilhe, 2021).
The only limitations of these products are the fantasy of the issuer and current demands
from investors.

Since structured products were first introduced to private investors in 1996 they have grown
rapidly in popularity. One explanation for this is that structured products offer retail
investors easy access to derivatives where they can invest in products that offer customized
exposure to hard-to-reach asset classes. Another reason might be that structured products
often are principal protected, meaning that the investor will have a minimum return
equal to the initial investment. However, principal protection is not always a feature of
structured products.

There are many potential risks that the investors of structured products should be aware
of. Firstly, there is a lack of liquidity in the structured products market due to the
highly customized nature of the products. Thus, investors should generally invest in
structured products with a buy-and-hold attitude as the products are hard getting rid of
once invested. Secondly, structured products vary in complexity. Some products are so
complex that the risk and return profile becomes unclear for the investor. Lastly, there
are risks regarding the issuer’s credit quality. If the issuer of the product goes bankrupt,
the investor might lose the invested money.

This chapter will present the role structured products have had in Norway and how
traditional structure products have developed in to modern structured products such as
the autocallable note. The discussion will start with the traditional guaranteed structured
products before introducing autocallable structured products. To wrap up the chapter we
will present the two specific autocallable notes we will analyse in this thesis.
4 2.1 Structured products in Norway

2.1 Structured products in Norway

Structured products were first introduced in Norway in 1992. Originally, they were only
offered to institutional investors, but in 1996 DnB started to offer these products to
private investors as well (Bøe, 2007). The structured products offered are categorised
as guaranteed structured products, characterised by a certain bond element where the
interest paid by the bond is allocated to an uncertain derivatives element. These products
are designed such that investors are guaranteed a minimum amount, often equal to the
original invested amount, in addition to an uncertain potential return dependent on the
derivatives part. The products became popular among private investors, and in 2006
private investors made up 90% of all outstanding structured products (Bøe, 2007). Figure
2.1 displays the total outstanding volume of guaranteed structured products in the time
period from 2000 to 2009 using data from Bøe and SSB (2007; 2009). The growth was
high and stable between 2000 to 2005. In 2005 the total volume was 47 billion NOK, more
than nine times higher than in 2000.

Figure 2.1: Outstanding volume of structured products in Norway 2000-2009

What Figure 2.1 also illustrates is that the growth stagnated between 2005 and 2006. By
looking at the numbers we can see that the total volume of outstanding structured products
decreased significantly between 2006 and 2009. During this period the products received
much critique in media, which was specially directed at the gearing and advertisement of
these products. As a result of the critique, regulations of structured products tightened
between 2006 to 2008. The tightened regulations required investors to have more knowledge
2.2 Autocallable structured products 5

of the structured products, and it was also stated that private investors normally would
not have the necessary knowledge to understand the risks of structured products. This
resulted in a stop in sales of guaranteed structured products to non-professional investors
(Helgerud, 2012).

In the 2010s we have seen a rise of new variants of structured products being offered
to private investors in Norway. Modern structured products differ from the guaranteed
structured products because of the regulations and interest rate environment. The interest
rates in the western world, and Norway, decreased significantly during the 2000s and
especially after the financial crisis in 2008. Lower interest rates affected the structured
products. Traditionally, the guaranteed structured products packaged a bond which
used the interest paid to finance a long position in a derivatives component. Certain
modern structured products however, such as the autocallable note, package the bond
with both long and short positions of derivatives with the goal of increasing the fixed
interest payment. The next section will present autocallable notes in more detail.

2.2 Autocallable structured products

Autocallable structured products are a specific type of structured products. Since first
introduced by BNP Paribas in 2003, autocallable structured products have been an
influential force in flow derivatives. A flow derivative is an instrument that provides
maximum leverage to profit from small movements in the market value of the underlying.
Autocallable products typically consist of a fixed income part and two or more barrier
options that depend on one or more underlying assets.

The main idea of autocallable structured products is to pay predetermined coupons

at predetermined dates, just like a normal bond. However, the coupon payments for
autocallables contain some uncertainty in that they depend on the value of one or more
underlying assets. The underlying asset(s) must be equal to or higher than a predetermined
reference value named the coupon barrier for the coupons to be paid out. Another aspect
of autocallables and the reason behind the name is that they might be terminated or
“autocalled” before maturity. When the product is autocalled, the investor is repaid the
notional value plus one coupon payment, and all future coupons are cancelled. The autocall
6 2.2 Autocallable structured products

occurs when the value of the underlying asset(s) is above or equal to a predetermined
reference value called the autocall barrier. Typically, autocallable structured products also
have something called a risk barrier. If a product is not autocalled before maturity, and
the reference value of the underlying asset(s) is below the risk barrier at maturity, then
the investor will not be repaid the notional value. Instead, the investor will receive the
reference value of the relevant underlying asset. In the worst case scenario, this reference
value is 0 implying that the investor might lose the entire investment. Therefore, we often
say that autocallable structured products are not principal protected. The risk barrier
can be viewed as a written put option on the bond. Thus, the coupons of the autocallable
should increase as the risk/volatility of the underlying asset(s) increases to reflect the
higher value of the sold put option.

A question to be asked is why investors invest in autocallable notes. Some investors may
view autocallables attractive due to the potential high promised returns offered in low
yield environments. If stock markets move sideways the autocallable note may promise
higher returns than stocks. The autocall feature may also be perceived as attractive as
it offers flexibility if the market rallies. However, a disadvantage of the autocall feature
is that high underwriting costs might offset the returns if the note is autocalled early.
The obvious disadvantage of the autocallable note is the written put option (risk barrier),
which expose the investor to the downside of the underlying stock(s).

2.2.1 Features of autocallable structured products

Features of autocallable structured products vary, resulting in many different variants. The
most common variations are tied to discrete/continuous call dates, number of underlying
assets, currency of underlying asset(s), and features of the coupon payment structure.

Autocallable notes can have discrete or continuous call dates. Discrete autocallables can
only be autocalled if the autocall barrier is crossed at an observation date. An observation
date is a date where the prices of the underlying assets are observed to determine if there
will be a coupon payment or if the note should be autocalled. Continuous autocallable
notes on the other hand are called immediately when the relevant underlying asset hits
the autocall barrier. This implies that continuous autocallable notes are more likely
2.2 Autocallable structured products 7

to be autocalled compared to the discrete autocallable notes (Deng et al., 2011). The
difference between discrete and continuous call dates, for autocallable structured products,
is illustrated in Figure 2.2.

Figure 2.2: Discrete and continuous call dates

Autocallables may be referred to as univariate or multivariate. Univariate autocallables

have only one underlying asset, whereas multivariate autocallables have two or more
underlying assets. The most popular autocallable today is the worst-of multivariate where
the relevant underlying at all times is the worst performing asset. For such products,
investors should be aware of the correlation between the underlying assets. If for example,
the correlation is low, then there is a higher chance that one of the assets will perform
badly.

Some autocallable notes have underlying assets denominated in another currency than the
settlement currency. The settlement currency refers to the main currency of the product,
i.e., the currency of the notional amount and coupon payments. Autocallable notes where
the underlying assets are denominated in another currency than the settlement currency
usually have a quantos structure. Quantos are known for facilitating foreign returns in
domestic currency. For example, if the return of a foreign underlying asset is 20%, then
the return of the option would be 20% as well, independent of the exchange rate. This
means that the size of the domestic payout is directly dependent of the foreign asset(s)’
returns, and thus uncertain (Bjerksund et al., 1999).
8 2.2 Autocallable structured products

Features in the payment structure of coupons vary in that they might be accumulated or
not. For coupons that are not accumulative, any missed coupon barriers will result in
those coupons being lost. However, if the same note had an accumulative feature, the
missed coupons would instead accumulate and been paid out if the coupon barrier once
again was crossed.

2.2.2 Autocall Norske Selskaper and Autocall Telekom

Now that we have presented what autocalls are and how they might differ, we want
to present two specific types of autocalls that we will focus on in our thesis. The
specific autocalls are named Autocall Norske Selskaper and Autocall Telekom (Garantum
Fondkommission, 2021a,b). They are both structured by Goldman Sachs and sold to
Nordic customers through the broker firm Garantum. Their main features are summarized
in Table 2.1. As the table suggests, they have many similarities, but there are differences
in the underlying assets and the coupons.

Name of Autocall note

Description Autocall Norske Selskaper Autocall Telekom
Starting day 29th October 2021
Last possible ending day 18th November 2026
First possible autocall date 31th October 2022 (4th observation date)
Observation dates 20
Autocall barrier 90%
Coupon barrier 80%
Risk barrier 60%
Multi/Univariate Multivariate
Relevant underlying Worst-off
Discrete/Cont. observations Discrete
Coupons Accumulative
Coupon paid Quarterly, 14 business days after observation
Underlying (Ticker) NHY, YAR, TEL, SALM NOKIA, ERIC, VOD, TEL
Quantos structure No Yes
Quarterly coupons1 4.00% 3.65%

Table 2.1: Main features of the autocall notes Autocall Norske

Selskaper and Autocall Telekom.

.
1
Confirmed coupons for the autocallable notes are found in Garantum Fondkommission and Garantum
Fondkommission (2021c; 2021d)
2.2 Autocallable structured products 9

Both notes started on 29th of October 2021 and will run until 18th of November 2026, if
they are not autocalled at an earlier point in time. They both have notional values of
NOK10,000 per certificate and incur significant costs. Firstly, there are underwriting fees
of 3%, which means the investor must pay NOK10,300 per certificate. Secondly there are
facilitating costs of 6% that are said to be included in the notional value of NOK10,000.
Lastly, there are also issuer costs indicated to be 3.05%, also included in the notional value
of NOK10,000. In total there are costs of 12.05% based on the notional value. The way
we interpret this is that the investor must pay NOK10,300 for something that the issuer
values to NOK9,0951 . However, this value is as we will see in Chapter 7 not communicated
directly in the prospect. In later chapters we will value these products to check whether or
not the price of NOK10,300 paid by investors is acceptable. We will also analyse the sales
process of these products to get an understanding of how these products are presented
and sold to investors.

1
9, 095 = 10, 300 − (10, 000 × 12.05%)
10

3 Valuation model for autocallable notes

The autocallable structures are complex in that they are path dependent, have accumulative
coupons, and a potential early maturity. Because of this complexity, we cannot value the
products with a closed form solution. Instead we must use a numerical method. A popular
method to price such complex products is with the Monte Carlo simulation approach.
The idea behind the Monte Carlo simulation approach is to simulate a random variable
numerous times and using the average value as the estimate for the given variable’s
expected value (Glasserman, 2003). In our case, to simulate the price of the autocallable
notes, we must simulate the underlying stock paths and tie them to the payoff function.
As the stock path model is a stochastic model, every unique iteration of the simulation
will yield a different price path generation, and thus a different price estimate. If we
simulate these prices enough times, then the law of large numbers will ensure that the
average price will converge to the correct value of the product.

We will now introduce all relevant theory needed to perform the numerical valuation of the
autocallable strucutres using the Monte Carlo approach. The chapter can be summarized
as follows; In Subsection 3.1, we present the Black & Scholes model. This model is
the underlying reason for why we can produce a fair price estimate of the autocallables.
In Subsection 3.2, we present the underlying theory of Monte Carlo simulations. In
Subsection 3.3, we present the stock price model we use to model the underlying stock
prices. Lastly, in Subsection 3.4, we present the payoff function. In the Chapter 4 we will
discuss how the model parameters are defined. In Chapter 5, we will perform the actual
valuation of the two autocallable notes.

3.1 The Black & Scholes model

When valuing structured products we operate in a risk-neutral framework called the Black
& Scholes model. The risk-neutrality of this model is what allows us to make a fair price
estimate of the products, and is the cornerstone of modern option pricing. Brandimarte
sums up the key results of the valuation framework in his book Handbook in Monte Carlo
simulation (2014). Namely, given a risk-free interest rate and a risk neutral probability
3.1 The Black & Scholes model 11

measure Q, one can price an option by taking the expectation of the discounted future
payoffs. The expectation must be taken under the aforementioned risk-neutral measure,
which is often called the equivalent martingale measure, and the discounting must be done
using the risk-free interest rate. These findings are great news for us, as it eliminates
the necessity of having to find the subjective expected rate of returns and the associated
discounting rates of the underlying assets, in order the price the structured products.

Going from the real probability measure to the equivalent martingale measure is trivial in
the Black & Scholes model. This is accomplished by replacing the expected rate of return
parameter in the stock price model, with the objective risk-free rate. We will explain this
further when we introduce the stock price model in Section 3.3. Utilizing the Monte Carlo
simulation method, taking the expectation of discounted future payoffs is also a trivial
exercise. Thus, the difficulty in the valuation of the autocallable structures boils down to
how we tie the simulated stock prices to the payoff function.

The Black & Scholes model makes a couple of assumptions which must be clarified. These
are listed below.

1. The market is complete and there are no arbitrage opportunities.

2. There are no transaction costs, and investors have no restrictions regarding short-
sales, nor borrowing.

3. The volatility, expected rate of return, dividend, and the risk-free rate are constant
parameters.

4. Trading occurs in continuous time.

5. Stocks follow a geometric Brownian motion.

The first assumption is necessary for the risk-neutral valuation to hold. A risk-neutral
investor is indifferent between a certain and an uncertain payout with equal expected
return (McDonald, 2014). The law of one price ensures that there will be no arbitrage
opportunities which is a crucial element in risk-neutral valuation. Assumption five will
be introduced in Section 3.3. As for assumption two, three and four, these will be
implemented through the geometric Brownian motion model.
12 3.2 Monte Carlo method

3.2 Monte Carlo method

The Monte Carlo method is a mathematical technique for simulating stochastic systems
numerically. As mentioned above, in a risk-neutral valuation framework we can find the
option value by taking the expectation of a discounted future payoff. In short, this is
precisely what the Monte Carlo technique do. Thus, in cases where there are no easy ways
of finding the option value, Monte Carlo simulation, in combination with the risk-neutral
valuation framework, is a very handy tool. The pricing of autocallable structures is such
a case, as there is no analytical solution to the problem.

Monte Carlo simulation utilizes the randomness of the stock price model to build a
probability distribution of possible outcomes (Brandimarte, 2014, p.3). By drawing
from the underlying model’s probability distribution, the Monte Carlo model generates
numerous scenarios. The draws must be individual and identically drawn (IID) to ensure
that each scenario is random. This will provide a range of possible outcomes together
with their respective probabilities. The key feature with the simulated estimate-range
is that the average estimate will, by the law of large numbers (LLN), converge to the
model’s expected value as the number of simulations increases. Thus, in the aspect of
finding the value of a structured product, simulating the underlying stock paths and the
associated payoff function thousands of times are essential to get a good estimate of the
product’s value. Eq. 3.1 shows how the Monte Carlo method utilizes the LLN, where N
is the number of iterations, g n represents the individual discounted payoff, and µg is the
true value of the payoff.
N
1 X n
lim g = µg (3.1)
N →∞ N
n=1

According to MIT, the advantage of using Monte Carlo simulation is that the value
estimate is unbiased (na). When using the Monte Carlo framework it is important to
consider the standard error of the sample mean. In Eq. 3.1, the sample mean converges
towards the true expected value when N increases. However, it is worth noting that the
standard error of the estimate decreases with the square root of the number of simulations1 .

1
Equation showing how the standard error of the Monte Carlo estimator σȳ decreases by the square
σ
root of the number of simulations N: σȳ = √y
N
3.3 Geometric Brownian motion 13

This implies that to decrease the variation of the sample mean by a factor of 10, one
must increase the number of simulations by a factor of 100. Thus, it is necessary to
consider the computational time of the Monte Carlo simulations. For instance, in complex
situations, the number of required simulations could be problematic as it becomes too
time consuming. In our case, the computational cost is manageable as we are easily able
to simulate enough iterations to reduce the standard error.

3.3 Geometric Brownian motion

3.3.1 Properties

Geometric Brownian motion (GBM) is the most popular model used to model stock prices.
As mentioned in Section 3.1, the GBM is the stock price model used in the Black & Scholes
model. The GBM models the change in a stock price as a combination of a deterministic
drift-term and a stochastic dispersion-term (Hull, 2018, p.331). The drift-term is driven
by a drift parameter over the time interval dt, while the dispersion term is driven by a
random shock scaled by a volatility parameter σ. The random shock is called a Wiener
process, and is nothing more than a normally distributed random variable with mean zero
and variance equal to the time interval, dWt ∼ N (0, dt). Eq. 3.2 illustrates the GBM
model as a stochastic differential equation where we use the expected rate of return µ as
the parameter driving the drift-term. The model needs a starting value, called the initial
value, which is given by S0 = s > 0. The equation models how the stock price changes
from time t over a very small (infinitesimal) time step dt.

dSt = µSt dt + σSt dWt (3.2)

Instead of showing the change in the stock price we can instead utilize two mathematical
techniques, called Itô’s Lemma and the Feynman-Kac solution, to compute the final date
solution to the process instead (Glasserman, 2003). We can also utilize that the Wiener
process is equal in distribution to a random variable z multiplied with the square root of
d √
the time interval. Thus, as W (t) = tz and z ∼ N (0, 1), we can write the solution to the
GBM as Eq. 3.3.
1 2
√
St = S0 e(µ− 2 σ )t+σ tz
(3.3)
14 3.3 Geometric Brownian motion

where St is the time t stock price and the S0 is the initial stock price.

The GBM makes a couple of assumptions. Firstly, the assumptions under the Black &
Scholes model are also relevant for the GBM. Secondly, the GBM is nothing more than
an exponentiated Brownian motion. Taking the logarithm of the stock price from the
GBM will return a Brownian motion (Glasserman, 2003, p.93). As the Brownian motion
is normally distributed, it then follows that the geometric Brownian motion is lognormally
distributed. This property is important as it stops the stock prices to become negative.
Due to the the limited liabilities assumption of the stock market, negative prices are
unreasonable (Bodie et al., 2018, p.41). Additionally, the GBM follows a Markov process
(Hull, 2018, p.324). This implies that the only information relevant for pricing future
stock prices lies in the current price of the stocks. This also imply that price changes
are independent of each other. Lastly, the GBM assumes that the expected dollar return
is proportional to the stock price level. In other words, if the expected rate of return is
12% when the stock price is $30, it will still be 12% when the stock price is $80, ceteris
paribus. The same assumption applies to the volatility parameter as well. Thus, both
the uncertainty and the expected rate of returns grows proportionally to the price level,
which is a reasonable assumption according to (Haug, 2021).

3.3.2 GBM under the risk-neutral framework

To value the autocallable notes we must use the risk-neutral valuation framework. As
mentioned in Section 3.1, this is a trivial implementation as the only difference from
Eq. 3.3 is the replacement of the subjective expected rate of return parameter µ, with
the objective risk-free rate rf . Thus, in a valuation framework, one can use Eq. 3.4 to
simulate the time t value of the underlying stock price.

1 2
√
St = S0 e(rf − 2 σ )t+σ tz
(3.4)

3.3.3 Implicit dividends in GBM

It is important to distinguish between total returns and capital gains of an asset when
analyzing the autocallable structures. As it is the stock price that determines the payoff
3.3 Geometric Brownian motion 15

and not the total return, we must adjust for this when simulating the stock prices. In the
Black & Scholes model it is easy to implement such a implicit dividend rate, as the only
implication is that we must subtract the rate in the drift-term. In the Black & Scholes
model it is usual to use delta δ as the notation for the continuous dividend yield. However,
as the quantos structure of the Autocall Telekom note will introduce some implications
later on in Section 4.4, we must distinguish between the implicit dividend rate and the
continuous dividend yield. In this thesis, we will thus use δ as the implicit dividend rate,
and λ as the continuous dividend yield. Eq. 3.5 shows the GBM in the risk-neutral
valuation framework, where implicit dividend rate is included.

1 2
√
St = S0 e(rf −δ− 2 σ )t+σ tz
(3.5)

3.3.4 Correlated geometric Brownian motion

The autocallable notes we are investigating have four underlying assets each determining
the payoff. Thus, we must account for the correlation between the assets when we model
the stock prices. According to Hansson, neglecting this adjustment will bias the value
estimate of the autocallable structures (2012). Therefore, to have a robust and credible
model, correlation must be taken into account.

Instead of simulating four individual GBMs, we must adjust each individual stock price
model such that all four of them have the correct correlation with the three others. To
implement this, we must adjust the dispersion-term. More specifically, we will adjust
the Wiener process W (t) in each GBM, such that the four Wiener processes have the
same correlation matrix pij as the underlying assets (Glasserman, 2003). According to
Glasserman, this can be accomplished by multiplying a vector of IID Wiener processes
(W (t)1 , ..., W (t)d ) with any matrix A, where AAT = Σ. The subscript d is the number of
underlying assets, and the matrix Σ is nothing more than the underlying assets’ d × d
covariance matrix. The matrix AT is transposed of matrix A, and matrix A and AT are a
lower and a upper triangular matrix, respectively.

To find matrix A we must utilize a mathematical technique called the Cholesky

decomposition (Haugh, 2004). The Cholesky decomposition factorizes any positive semi-
16 3.4 Payout function

definite matrix into a lower and a upper triangular matrix, where one is transposed of
the other. In most programming languages, including R which we use, the Cholesky
decomposition is implemented through a built-in function. Thus, finding matrix A in our
situation is accomplished by feeding the Cholesky function with the underlying assets’
covariance matrix. If the reader is interested in reading more about the theory behind the
Cholesky decomposition, then Haugh’s lecture slides on the topic is recommended (2004,
p. 5).

As in Section 3.3.1, we will utilize that the Wiener process can be written as a standard
d √
normal random variable multiplied by the square root of time, W (t) = tz, and
z ∼ N (0, 1). Eq. 3.6 shows the correlated GBM for asset i.

1
√
Sti = S0 e(rf −δi − 2 σi )t+
2 tai z
, i = 1, .., d, (3.6)

where ai is the ith row of the matrix A, and z a standard normal random variable
z ∼ N (0, 1). For each stock price generation, the z’s are IID. The parameter δi is the
implicit dividend rate for asset i.

3.4 Payout function

In the Monte Carlo simulation method, we compute the value estimate of the structured
products by taking the average of N simulated price estimates. In this section we present
how the price in each of those different scenarios are derived. As we want to price the
autocallables, we operate within the risk-neutral framework, thus using the risk-free rate
rf as the main driver of the drift-term in the stock price model.

The price solely depends on the payout tied to the underlying stock paths. As the
autocallable structures to be priced have complex features, it is difficult to make an easy
understandable function for the cash flow payouts. Instead we will simplify the illustration
by simulating the path of one stock, and show how we tie the payout to the path using
numerical vectors. It is worth noting that the only difference between one and four stock
paths regarding the payout function is that when having multiple underlying assets, we
evaluate the numerical vectors based on the lowest stock price of the four at each valuation
date.
3.4 Payout function 17

Figure 3.1 illustrate a simulated stock path for a stock with initial value of 100, volatility
of 20%, and drift equal the risk-free rate. The risk-free rate in this example is set to an
arbitrary number of 2%. The stock path in the figure is simulated for each consecutive day
for five years. This is done for an illustrative purpose. When doing the valuation of the
autocallable notes in Chapter 5, we will simply simulate the stock prices for a quarter at
a time, and then evaluate the payouts. The difference in length of simulation steps has no
importance on the estimated value of the notes. However, having longer simulation steps
(i.e., a quarter at a time) will decrease the computational workload, which is desirable.

The horizontal lines illustrate the different barriers, and the vertical lines illustrate the
discrete observations dates. The autocall barrier is set at 90% of initial value, the coupon
barrier at 80%, and the risk barrier at 60%. The first possible autocall date is the 4th
observation date illustrated by the highlighted vertical line. The autocallable to be valued
here has a notional value of NOK10.000 and a coupon of 4.00% (NOK400).

Figure 3.1: One GBM stock path.

The precise stock price at each valuation date is illustrataed in the column Valuation
prices in Table 3.1 at the end of this subsection. From Figure 3.1 we can see that the
autocallable note will be autocalled on the 15th observation date (3rd year, 3 qtr.). This
is also verified in Table 3.1 from the aforementioned Valuation prices column, where the
15th valuation price is the first price above the autocall barrier of 90 (disregarding the first
18 3.4 Payout function

three valuation dates). In the column Autocall the autocall condition is met, illustrated
by a vector turning 1 at the 15th valuation date.

To capture the coupon payouts, we use a similar approach as above by constructing

a vector that turns 1 when the condition for coupon is met, and 0 otherwise. This is
represented in the Coupons column in the same table. For coupons we must also adjust
for accumulation. The effect of accumulation is captured by counting observation dates
where the stock price is below the coupon barrier of 80. If the stock price moves above the
coupon barrier at a later valuation date, the counted streak will be added to that date.
From the Coupons column, we observe that the stock price is below the coupon barrier
between date 7 and date 13, resulting in a counting streak of 7. At observation date 14,
the stock price is once again above the coupon barrier. The counted streak of 7 is then
added to the coupon of date 14, resulting in a coupon payout of 8 × NOK400 = NOK3, 200.
The coupon vector must also reflect that there will be no coupons after the note has been
autocalled. This effect is captured by connecting the coupon vector to the autocall vector.
The coupon vector turns 0 for all remaining values if the autocall column turns 1.

There is also a vector for the risk barrier, represented by the Risk-barrier column. This
vector can only turn 1 at observation date 20. The conditions for the risk barrier vector
to turn 1 is that the sum of the autocall vector must be equal to 0 and that the stock
price at the last observation date is below the risk barrier of 60. In the example presented,
all the values in the risk barrier are 0, as the note is autocalled at valuation date 15.

With the Autocall, Coupons, and Risk-barrier columns in Table 3.1 identified, it is possible
to find the cash flows for the different observation dates, which are represented in the
CF column. For valuation date 1 to 19, the cash flows are identified by multiplying the
notional value of NOK10,000 with the Autocall vector and the coupon amount of NOK400
with the value in the Coupons vector.

For valuation date 20, there are three different scenarios. Firstly, the cash flow is zero
if the note has been autocalled during the first 19 valuation dates. Scenario two and
three however, depend on whether the stock price is above or below the risk barrier of 60.
This is represented by the value of the Risk-barrier column. If the Risk-barrier value is 1
3.4 Payout function 19

the cash flow is computed as the final date stock price divided by the initial stock price,
multiplied by the notional amount, St /S0 × NOK10, 000. If the Risk-barrier value is 0,
the cash flow will be similar to the computations of valuation date 1 to 19. However, the
notional amount of NOK10,000 will be paid out regardless of the value of the Autocall
vector. Thus, the cash flow will be the notional amount of NOK10,000, plus the the
coupon amount of NOK400 multiplied with the value in the Coupons vector.

The last column of Table 3.1, the Discounted column, reflects the discounted cash flows.
In this column we discount the cash flows from column CF by a continuous discount rate
of 2%. All observation dates are discounted according to correct time, represented by the
Observations column. Thus, the cash flow of observation date 15 is discounted 15 periods
back. To find the product’s value, we simply sum up all the discounted cash flows, which
in this scenario equals a value of NOK14,990.6.
20 3.4 Payout function

Observations Valuation prices Autocall Coupons Risk-barrier CF Discounted

1 90.60 0 1 0 400.0 398.0
2 88.30 0 1 0 400.0 396.0
3 80.50 0 1 0 400.0 394.0
4 84.80 0 1 0 400.0 392.1
5 82.60 0 1 0 400.0 390.1
6 82.10 0 1 0 400.0 388.2
7 71.30 0 0 0 0.0 0.0
8 74.20 0 0 0 0.0 0.0
9 62.00 0 0 0 0.0 0.0
10 70.20 0 0 0 0.0 0.0
11 68.70 0 0 0 0.0 0.0
12 77.10 0 0 0 0.0 0.0
13 74.90 0 0 0 0.0 0.0
14 86.90 0 8 0 3200.0 2983.7
15 111.50 1 1 0 10400.0 9648.5
16 102.30 0 0 0 0.0 0.0
17 114.90 0 0 0 0.0 0.0
18 112.30 0 0 0 0.0 0.0
19 113.40 0 0 0 0.0 0.0
20 113.50 0 0 0 0.0 0.0
Value 14,990.6

Table 3.1: Table illustrating the vectors used finding the discounted
cash flows from one simulated stock path. Column Observations
represents the different valuation dates. Column Valuation prices
represents the accompanied stock prices. Columns Autocall,
Coupons, and Risk-barrier represent the different conditions for
autocall, coupons and risk-barrier, respectively. Column CF and
column Discounted represent the exact cash flows and discounted
cash flows at each individual observation date. The value of this
particular autocallable note is 14,990.6, which can be seen at the
last row. The highlighted row, row 15, illustrate the valuation date
which the note is autocalled.
 21

4 In-data parameters
The value of the autocallable notes are determined by a set of different variables. In this
chapter we will present each variable, which are the risk-free rate rf , the discount rate,
the expected rate of return µ, the implicit dividend δ, the volatility σ, and the correlation
parameter ρ. We will also present the historical data which is used to compute the two
latter variables.

4.1 Data
For the volatility parameter σ and the correlation parameter ρ we will use historical data
for estimating the values. These parameters can vary substantially depending on the data
we estimate them from. Thus, it is necessary to discuss both the frequency and the length
of the historical data. We must also asses what type of data we will use in the models. Do
we want data that are adjusted for stock splits and dividends, or do we want raw data?
We must also consider whether we want to use logarithmic or arithmetic returns in our
estimation of the parameter values.

The length of the historical data depends on whether the range is regarded to be relevant
for the analysis or not. This is a subjective matter which needs to be evaluated in context
of the length of the products simulated. Usually, when simulating stock price behaviour
for two years, two years of historical data is considered to be relevant. In our case, this
translates to five years of historical data as the structured products have a five-year horizon.
There should also be an assessment regarding extraordinary events as well. Extraordinary
events create abnormal stock price movements, which can heavily affect the parameter
estimates. The last two years have been affected by the COVID-19 pandemic. However,
as the COVID-19 Omicron variant is on the rise (as of December 2021), we do not believe
excluding the COVID-19 data will be representative for the future. Thus, we choose not
to exclude any data from the historical data-set.

Regarding the frequency of the data, we believe a monthly frequency is an appropriate

frequency to be used. As the autocallable notes are evaluated once a quarter, a daily
22 4.2 Risk-free rate and discount rate

frequency can create too much noise, as auto-correlation can be an issue. A quarterly
frequency will match the data even better. However, this will lead to a data-set consisting
of twelve observations per underlying asset. When using historical data to estimate
parameters, it is preferable to use log normally distributed returns. However, it is difficult
to say anything about the distribution when having so few observations. A monthly
frequency will yield 60 observations per asset which we believe are more appropriate.

As we do not want the correlation and volatility estimates to be biased by events like stock
splits and dividends, we must use data that are adjusted for this. There are many sources
providing such adjusted data. However, we choose to download adjusted closing prices
from Yahoo finance. These prices have been adjusted according to the CRSP1 standards
(Yahoo!, na). As mentioned in the last paragraph, we will manipulate the historical data
to construct the logarithmic returns for each underlying asset. This will be done in the
programming language R.

To summarize, we use monthly adjusted stock prices from November 2016 to November
2021, downloaded from yahoo finance. In the programming language R, we remove any
missing values, and compute associated logarithmic returns for each underlying asset. The
logarithmic returns are used to estimate volatility and correlation.

4.2 Risk-free rate and discount rate

In a risk-neutral valuation framework, the risk-free rate rf is used as the parameter
responsible for the drift in the stock price model. The risk-free rate is also one of two
building blocks in the discounting rate, which is used to discount the simulated value
estimates in the Monte Carlo simulation. The second building block in the discounting
rate is the credit spread of the issuing bank, which is simply added to the risk-free rate
(Hull, 2018).

For the risk-free rate we find the yield of the Norwegian government bond with maturity
of five years to be a good proxy. Firstly, governments in stable countries such as Norway

1
https://www.crsp.org/products/documentation/crsp-calculations
4.3 Expected rate of return 23

have extremely low, if no, probability of default. Secondly, the settlement currency is
NOK. And thirdly, the maturity of the notes is five years.

4.3 Expected rate of return

In Chapter 6, we will investigate what returns the investor can expect when buying the
autocallable notes. We will also examine the probabilities of when the notes are autocalled.
In both instances we find ourselves outside of the risk-neutral framework. Thus, we must
use the subjective expected rate of return as the main driver of the drift when simulating
the stock prices.

According to Coval and Shumway, as we assume the underlying stocks follow a GBM
we can apply the Capital Asset Pricing Model (CAPM) to estimate the expected rate
of return (2001). The CAPM describes the relationship between risk and the expected
return. The general idea of the model is to compensate the investor in two ways. Through
the time value of money, denoted as rf , and through risk. For taking on risk the investors
are compensated by a risk premium rp , which in the CAPM is represented as the market’s
expected rate of return minus the risk-free rate, rp = rm − rf . Eq. 4.1 below illustrates
how the expected rate of return µ is estimated using the CAPM.

µ = rf + βrp (4.1)

where β is the asset’s measure of systematic risk, beta. We will use the CAPM to compute
the expected rate of return of all the individual underlying assets. As each company reacts
differently to the market movement, we must use the individual betas in the computation.
For the assets denoted in foreign currencies, we must also use the foreign risk-free rate
and the foreign market risk-premium, when calculating the expected rate of returns.

4.4 Implicit dividend rate

As mentioned in Section 2.2.1 there are some implications regarding the quantos structure
of the Autocall Telekom note and the implicit dividend rate δi . For notes with quantos
structure the underlying assets are denominated in another currency than the settlement
24 4.4 Implicit dividend rate

currency. The return of the autocallable is thus idependent of the development of the
exchange rates. As the size of the domestic payout are directly dependent on the foreign
assets’ return, we must adjust for this when modeling the underlying assets. This
adjustment will be implemented through the implicit dividend parameter.

In the article Pose og Sekk, the authors Bjerksund et al. illustrate how this adjustment
is implemented (1999). In the article, the implicit dividend rate is adjusted to include
not only the dividend yield, but also a component for the rate differences between the
domestic and foreign risk-free rate. This rate-difference is what distinguish the implicit
dividend rate δi with the continuous dividend yield λi . Eq. 4.2 shows the relationship
explicitly.
δi = λi + (rf − ri ) (4.2)

where the term (rf − ri ) corrects for the foreign assets being denominated in another
currency than the settlement currency. rf is as usually the domestic risk-free rate, and
ri is the i = 1, .., d individual foreign risk-free rates. In the Autocall Telekom note, one
of the underlying assets is a Norwegian company. In such instances, the correction-term
will become zero as the "foreign" risk-free rate will be equal to the domestic risk-free rate.
The article of Bjerksund et al. also include a component adjusting for the covariance
between the logarithmic return of the foreign assets, and the change in exchange rates.
This term is excluded from our equation as we find it to be so small that it has no real
affect on the implicit dividend rates.

Thus, for autocallable notes where all the underlying assets are denoted in the domestic
currency, the implicit dividend rate δi is equal to the continuous dividend yield of the
underlying asset λi as there is no rate difference. For Autocall Telekom which have the
quantos structure however, the implicit dividends will differ from the individual dividend
yields. The correction-term shows that if the domestic risk-free rate is higher than the
foreign risk-free rate, then it is expected that the foreign exchange rate will appreciate.
Such positive rate-difference will give a higher implicit dividend rate. In Table 5.8 we will
see that for the underlying assets in the Autocall Telekom note, this adjustment will make
a big difference.
4.5 Volatility 25

4.5 Volatility
Volatility measures the uncertainty of an asset. Plain vanilla call and put options increase
in value with increased volatility as increased volatility makes it more probable for such
options to be in the money, and deeper in the money. In the geometric Brownian motion
model used to model stock paths, the volatility is constant. However, in the real world
volatility varies over time. In some instances, like during the financial crisis in 2007-2008
or during the COVID-19 pandemic the two past years, the volatility was higher than
normal. In other times it is more subdued. This has led to criticism of the GBM, resulting
in solutions such as the Heston model. However, stochastic volatility models, like the
Heston model, does not guarantee any better results. An implementation of such models
would also incur even higher implementation difficulties, as one would have to calibrate the
model as well. Even though constant volatility does not reflect the real world correctly at
all times, it works as an okay proxy on average. Thus, we would use a constant volatility
in our valuation in the Results from valuation model, Chapter 5. To find the constant
volatility used in the GBM we have two approaches. Either we can look at historical
data or we could look at the implied volatility of the underlying assets in the options
market. As the options markets for the underlying assets are nonexistent or illiquid,
we find it unfeasible to use implied volatility. Instead, we will use historical volatility.
There are different methods to estimate historical volatility. The simplest method is the
historical volatility method. Other more sophisticated methods are the EWMA- and
GARCH-model, but these are more difficult to implement. Thus, we consider the simple
historical volatility method in this thesis illustrated in Eq. 4.3.
v v !2
n n n
u u
1 X u 1 X 2 1 X
(ui − ū)2 = t (4.3)
u
σ=t ui − ui
n−1 i=1
n − 1 i=1 n(n − 1) i=1

where σ is the volatility, ui the logarithmic daily returns in the interval i = 1, ..., n and ū
is the average logarithmic return.
26 4.6 Correlation

4.6 Correlation
Correlation measures how the underlying assets move together, and is illustrated by the
ρ parameter. The autocallable notes to be valued in this thesis have four underlying
assets, where the worst performing asset is the relevant for defining the payoff. In such a
structure, the correlation coefficients are very important. The lower the correlation, the
more likely it is for one of the assets to perform badly, assuming all other parameters
are held constant. In our model, the correlation is implemented through the Cholesky
factorization of the covariance matrix, as introduced in Section 3.3.4. Eq. 4.4 shows
how the covariance between two assets is estimated. xi and yi represent the i = 1, ..., n
individual logarithmic returns for the two assets, and x̄ and ȳ represent the mean returns.
Eq. 4.5 illustrates the relationship between the covariance and correlation. Here σx and
σy represent the respective standard deviations of the returns.
Pn
i=1 (xi − x̄)(yi − ȳ)
covx,y = (4.4)
n−1

covx,y
px,y = (4.5)
σx σy
 27

5 Results from valuation model

In this chapter we are going to conduct the valuation of the autocallable notes using the
methods introduced in Chapter 3. As mentioned in Section 3.1, when valuing a derivatives
product we operate in a risk-neutral framework. Thus, we will use the risk-free rate as
the parameter responsible for the drift in the stock price model. The framework will also
imply the use of the discount rate to discount the value estimates from the Monte Carlo
simulation in order to get a present value estimate of the true value.

The chapter will be divided into three parts. In part one, the estimated parameters
used in the valuation framework will be declared. In part two, the actual valuation of
the autocallable notes Autocall Norske Selskaper and Autocall Telekom will be executed.
Lastly, in part three we will summarise the numerical valuation.

5.1 Parameters
In Chapter 4 we examined how we would define the model parameters. In this section we
switch the focus to getting a real number. We distinguish between stock-dependent and
stock-independent parameters. For the independent parameters it is only necessary to
define the value once, while for the stock-dependent parameters we are forced to declare a
value for each underlying asset. We start by defining the stock-independent parameters,
which are the initial stock price S0 , the domestic risk-free rate rf , and the discounting rate.
Moving on, we will compute the stock-dependent parameters; volatility σi , correlation ρij ,
and the implicit dividend yield δi . For the assets denoted in foreign currencies, we must
also declare the foreign risk-free rates ri .

Initial stock price

We simplify the initial price of the underlying assets by setting them all to S0 = 100. In
Section 3.3.1 we saw that the GBM assumes expected dollar returns to be proportional to
the stock price level. The relative change in the stock prices will therefore be the same
independent of the initial value. To clarify, the initial stock price will only be identical at
date zero, as from that moment and on wards the stock prices will develop randomly.
28 5.1 Parameters

Domestic risk-free rate

As mentioned in Section 4.2, we will use the rate of the five-year Norwegian government
bond as the proxy for the domestic risk-free rate. This rate is given by Bloomberg to be
1.43% (as of 01.12.2021), equalling a continuous risk-free rate of 1.42%1 (Bloomberg L.P.,
2021).

Discount rate
As discussed in Section 4.2, the discount rate is composed of the domestic risk-free rate
and the credit spread of the issuing bank. We find the appropriate credit spread by
looking at the credit rating of the issuer, which in our case is Goldman Sachs. In the
sales brochure, Garantum states Goldman Sachs’ ratings from S&P and Moody’s, which
are A+ and A1 respectively. These ratings equals an appropriate credit spread of 0.980%
(Damodaran, 2020). If we adjust this credit default swap rate to be continuous, we get
a rate of 0.975%. This result in an appropriate continuous discount rate of 1.420% +
0.975% = 2.395%.

Foreign risk-free rates

To adjust the implicit dividend rate we must also find the risk-free rates for the underlying
assets denoted in foreign currencies. These assets are Nokia, Vodafone, and Ericcson,
which are listed in Finland, UK, and Sweden, respectively. The risk-free rates of these
countries are identified by the five-year treasury yield found using the Bloomberg terminal
(Bloomberg L.P., 2021). The foreign risk-free rates in Finland, UK9, and Sweden are
-0.030%, 0.707%, -0.407%, respectively. What we see is that all of these rates are lower
than the Norwegian rate. This implies a hidden dividend rate which is not favorable for
investors. The reason for this is that when we have a quantos structure, as in Autocall
Telekom, the return of the note is not affected by changes in foreign exchange rates. Taking
the return of the foreign underlying asset without including the exchange rate implies a
swap of returns, where the domestic interest rate is swapped with the foreign interest
rate. A positive difference between the domestic and foreign interest rate corresponds to
a situation where the foreign currency is expected to appreciate relative to the domestic
currency.

1
Where we use the equation cont.rate = ln(1 + annual rate) for finding the continuous rates.
5.1 Parameters 29

Volatility
The volatility parameters are estimated in R using a function that implement Eq. 4.3.
As mentioned in section 4.1, the estimates are based on five years of monthly returns
from 2016 to 2021. This yields a monthly volatility estimate. To properly implement
the volatility in the GBM model we must scale the volatility to be annual. This is easily
done by multiplying the monthly volatility with the square root of 12. For the underlying
companies in Autocall Norske Selskaper, Norsk Hydro, Salmar, Telenor and Yara, we get
annual volatility estimates of 0.319, 0.325, 0.167, and 0.195 respectively. Equivalently, for
Ericsson, Vodafone and Nokia, the underlying asset of Autocall Telekom, we get annual
volatility estimates of 0.248, 0.263, and 0.354, respectively. The Norwegian company
Telenor is included as an underlying company in both notes.

Covariance
As mentioned in section 4.6, the covariance matrix is fundamental for making correlated
stock paths in the GBM model. The covariance matrices for Autocall Norske selskaper
and Autocall Telekom are found in table 5.1. It may be difficult to relate to the values in
the covariance matrices which is why we also will present the correlation matrices later.
However, as the covariance matrices are crucial inputs in the model it is reasonable to
present them. The values in the matrices are found by using a function in the programming
language R that estimates the values according to Eq. 4.4.

Autocall Norske Selskaper

Norsk Hydro Salmar Telenor Yara
Norsk Hydro 0.12 0.02 0.01 0.04
Salmar 0.02 0.10 0.01 0.01
Telenor 0.01 0.01 0.04 0.01
Yara 0.04 0.01 0.01 0.06
Autocall Telekom
Ericsson Vodafone Telenor Nokia
Ericsson 0.11 0.01 0.01 0.05
Vodafone 0.01 0.05 0.02 0.01
Telenor 0.01 0.02 0.04 0.01
Nokia 0.05 0.01 0.01 0.11

Table 5.1: Covariance matrices of the underlying assets

30 5.2 Model reliability

Dividend yield
Finding appropriate estimates for dividend yield is difficult because it is uncertain how
companies will allocate their capital in the future. As mentioned in section 4.4, we
make a best guess based on the companies’ historical dividends. The estimated future
dividend yields are illustrated in Table 5.2, with both annual and continuous compounding.
According to our estimates, all the companies in Autocall Norske selskaper and Autocall
Telekom have positive dividend yields. Isolated, dividends represents a disadvantage to the
investors of the notes because they will decrease the capital gains of the underlying assets.
In other words, higher dividends decrease the chance for early autocall and increase the
chance of ending below the risk-barrier.

Annual yield Continuous yield

Norsk Hydro 3.00% 2.96%
Salmar 4.00% 3.92%
Yara 5.00% 4.88%
Telenor 5.00% 4.88%
Nokia 2.50% 2.47%
Vodafone 7.50% 7.23%
Ericsson 2.00% 1.98%

Table 5.2: Annual and continuous dividend yields

5.2 Model reliability

In this subsection we will test how our stock price model cope when pricing a derivatives
product with a well known solution. The intuition of doing this test is to ensure that
the model behaves as expected. In such, it can be considered as a robustness test. If the
simulated value-estimate converges to the theoretical price given the same parameters,
we can be assured that our stock price model is legitimate. We will also simulate 5,000
thousand stock price generations for one of the autocallable notes to assess whether the
historical correlation is reflected in the simulated prices or not.
5.2 Model reliability 31

5.2.1 Numerical versus theoretical price

As our model is based on the Black & Scholes framework, we consider the European call
option as a good product to use for the robustness test. As this derivative has a closed
form solution through the Black & Scholes formula, we can compute it’s theoretical price.
The Black & Scholes formula for a European call option with dividends, is given by Eq. 5.1.

C = S0 e−δT N (d1 ) − Ke−rf T N (d2 ) (5.1)

where
log( SK0 ) + (rf − δ + σ 2 /2)T
d1 = √
σ T
√
d2 = d1 − σ T

and N (d1 ) and N (d2 ) denotes the cumulative distribution function of the standard normal
distribution.

For both the Black & Scholes formula and for the numerical model, the arbitrary parameters
listed below will be used to calculate the value of the option.

• Initial stock price, S0 = 100

• Strike price, K = 100

• Risk-free rate rf of 1.420%

• Volatility σ of 16.740%

• Implicit dividend rate δ of 4.879%

• Time to maturity T equalling 5 years

Given the arbitrary parameters, the theoretical price of the European Call option using
the Black & Scholes formula is 6.625. Table 5.3 shows the numerical results from the
stock price model.
32 5.2 Model reliability

N Price estimate error margin Upper c.i. Lower c.i.

1000 6.190 -0,435 7.177 5.204
10,000 6.723 0.098 7.060 6.386
100,000 6.616 -0.009 6.721 6.511

Table 5.3: Numerical results of a European call option

We observe that the simulated price estimates converges to the theoretical price when the
number of simulations increases. This argues that the model is reliable.

5.2.2 Simulated versus historical correlation

To check whether the correlation is reflected in the stock price model or not, we perform
a Monte Carlo simulation where we generate N number of stock path generations. For
each of these simulations we compute the logarithmic returns for each asset and compute
the simulated correlation. Then we can plot the N different correlation estimates in a
histogram to assess the distribution of the correlation. As mentioned in 3.2, as the number
of simulations N increases the average estimate will converge to the true expected value.
Thus, if we plot the historical correlations as a vertical line, we can quickly observe if the
simulated correlation distribution seems correct or not. Figure 5.1 shows the simulated
correlation distributions between every underlying asset in the Autocall Norske Selskaper
note, with the associated historical correlations plotted as a vertical red line. We have
used 5,000 simulations for the simulated correlations.

From the figure we observe that the average simulated correlations converge to the historical
correlation as expected. Thus, we are positive that we have managed to incorporate
correlation into our stock price model. Table 5.4 shows both the historical and simulated
correlation matrices for the underlying assets in Autocall Norske Selskaper. The simulated
correlation matrix is constructed using the average of the simulated correlations. We
observe that the simulated correlation matrix and the historical correlation is very similar,
almost identical. Thus we can argue that the stock price model manages to incorporate
the correct correlation when generating stock paths.
5.2 Model reliability 33

Figure 5.1: Figure illustrating the distribution of 5,000 simulated

correlations for the underlying assets in the Autocall Norske
Selskaper note. The red vertical lines display the historical
correlations between the assets.

Historical Norsk Hydro Salmar Telenor Yara

Norsk Hydro 1.000 0.039 0.006 0.404
Salmar 0.039 1.000 -0.116 -0.074
Telenor 0.006 -0.116 1.000 -0.001
Yara 0.404 -0.074 -0.001 1.000
Simulated Norsk Hydro Salmar Telenor Yara
Norsk Hydro 1.000 0.040 0.006 0.404
Salmar 0.040 1.000 -0.116 -0.074
Telenor 0.006 -0.116 1.000 -0.001
Yara 0.404 -0.074 -0.001 1.000

Table 5.4: Historical and simulated correlation matrices for

Autocall Norske Selskaper ’ underlying assets.
34 5.3 Autocall Norske Selskaper

5.3 Autocall Norske Selskaper

In this section we will value Autocall Norske selskaper using the quantified parameters
outlined above. The parameters are summarised in Table 5.5. The historical correlation
matrix identified in Table 5.4 can help the reader to get an intuition of how the underlying
assets in this product move together. From Table 5.4, we observe that Yara’s correlation
with Norsk Hydro is positive with a value of 0.404. While for both Salmar and Telenor,
Yara’s correlations are almost zero. Thus we can expect that Yara’s stock path will move
more with Norsk Hydro’s stock path, than with other two assets.

Norsk Hydro Salmar Telenor Yara

Initial stock price 100 100 100 100
Volatility 0.319 0.325 0.167 0.195
Risk-free rate 1.42% 1.42% 1.42% 1.42%
Dividend yield 2.96% 3.92% 4.88% 4.88%

Table 5.5: Parameters used to model paths for underlying assets

in Autocall Norske selskaper.

The underlying assets’ stock paths, together with the payout condition, determines the
value of the autocallable note. To get an intuition of the paths of the underlying assets
in Autocall Norske selskaper we have made figure 5.2. This figure illustrates the price
paths of the underlying assets according to the parameters and correlations presented
in tables 5.4 and 5.5. The vertical lines in this figure illustrates the observation dates,
and the dark vertical line signal the first possible autocall date. In this exact path the
note would be autocalled at observation date 5. Note that we for illustrative purposes
have used time-steps of one day in this figure. In the actual valuation model we will have
time-steps of one quarter for the purpose of increasing the efficiency of the model.

Table 5.6 presents the payoff specific parameters of Autocall Norske selskaper. These
parameters specify the conditions for payout. At each observation date the prices of the
underlying assets will be observed. The observed price of the worst performing asset
together with the payout conditions will determine what the cash flow will be. This
relationship highlights the fact that the value of the autocallable is determined by the
paths of the underlying assets and the payout conditions.
5.3 Autocall Norske Selskaper 35

Figure 5.2: One path generation of Autocall Norske Selskaper ’s

four underlying assets

Payoff specific parameters

Autocall barrier 90%
Coupon barrier 80%
Risk barrier 60%
Notional NOK10,000
Quarterly accumulative coupon N otional × 4.00%
Discount rate 2.40%

Table 5.6: Payoff specific parameters used in the valuation of

Autocall Norske Selskaper

5.3.1 Results

Table 5.7 displays the results from our valuation algorithm. Each iteration of the valuation
model derives different price estimates of the autocallable note. The price estimate column
in the table illustrates the average price estimate from the number of iterations presented
in the iterations column. What we see is that the price estimates becomes more precise
as the number of iterations increase, just as suggested in Section 3.2. The most precise
answer is found from a simulation with 1,000,000 iterations. With 1,000,000 iterations we
value Autocall Norske selskaper to NOK6,955.7, with a standard error (SE) of NOK4.0
giving us a 95% confidence interval between NOK6947.9 and NOK6963.6.

The results in table 5.7 suggest that Autocall Norske selskaper is extremely overpriced.
Investors must pay NOK10,300 for something that is only worth NOK6,955.7.
36 5.3 Autocall Norske Selskaper

Iterations Lower c.i. Price estimate Upper c.i. SE Time

10 4, 166.2 6, 018.1 7, 870.0 944.9 0.020 s.
100 5, 975.1 6, 773.2 7, 571.4 407.2 0.100 s.
1, 000 6, 752.0 6, 998.7 7, 245.4 125.9 0.960 s.
10, 000 6, 917.9 6, 996.6 7, 075.2 40.1 9.170 s.
100, 000 6, 922.9 6, 947.6 6, 972.4 12.6 1.71 min
1, 000, 000 6, 947.9 6, 955.7 6, 963.6 4.0 43.35 min

Table 5.7: Price estimates, confidence intervals, and standard

errors of Autocall Norske Selskaper from valuation model.

This implies a total price premium of 10,300/6,955.7 - 1 = 48.08%, which is far above the
costs indicated in the prospect and sales brochure. One reason why the price premium is so
much higher than the costs indicated in the prospect might be that the price of constructing
the autocallable note exceed the value of the note, due to the highly customized derivatives
components comprising the product. However, this does not mean that it is fair to allocate
all the costs incurred when constructing the product to the investors. For this reason
we believe that the issuer and brokers should clearly communicate what the value of
the product is in addition to all the costs related to construct the product. This would
ensure more transparency for the investor and give a better foundation for a well informed
investment decision. Another reason why our price estimate is so much lower than the
price paid by investors might be the uncertainty regarding the estimated parameters. To
give a more reliable conclusion regarding the price estimate, we will in the next section
perform a sensitivity analysis where we see how sensitive the price estimate is to changes
in the parameters.

5.3.2 Sensitivity analysis

As there are uncertainties regarding the estimated parameters used to value Autocall
Norske selskaper we find it necessary to do a sensitivity analysis. The sensitivity analysis
will see how the price estimate change when certain estimated parameters change, such
as the correlation and volatility. We will also see how sensitive the price estimate is
to parameters where we have had to make a best guess based on information, such as
the drift rate that is determined by risk-free rates and dividends. There will also be an
assessment of how sensitive the price estimate is to the coupon payments to see how high
5.3 Autocall Norske Selskaper 37

the coupons must be to justify the sales price of NOK10,000.

Figure 5.3 illustrates how the price would change if either the volatility, or the correlation
would change by relative +/− 30%. We observe that even if we have miss-estimated the
value of the parameters by +/− 30%, the value of Autocall Norske Selskaper would still
be much lower than the price paid by investors. This means that even if the estimation
error of the volatility or correlation was as high as +/− 30%, our conclusion stating that
Autocall Norske selskaper is extremely over priced would not change.

Figure 5.3: Effect of a +/− 30% relative change in volatility and

correlation on price estimate.

Figure 5.4 displays how the price estimate of Autocall Norske selskaper changes as the
drift-term changes. The drift-term in the valuation model is determined by the Norwegian
risk-free rate and the dividend yields of the underlying assets. From the figure we see
that even if the drift-term was 5% higher in absolute terms, the price estimate would
still be significantly below the price of NOK10,300 paid by investors. The price would
also still be lower than the price paid by investors minus the costs paid by investors. We
find it highly unlikely that the drift-term would deviate more than absolute 5% from our
estimate. Thus, we still believe that Autocall Norske selskaper is overpriced.
38 5.3 Autocall Norske Selskaper

Figure 5.4: Sensitivity analysis of drift, Autocall Norske selskaper.

In figure 5.5 we can also see what the coupons would have to be to justify a value of
NOK10,300. A coupon of 12% would justify a value of ca. NOK10,300. This represents a
coupon three times higher than the actual coupon of 4%.

Figure 5.5: Sensitivity analysis of coupon rate, Autocall Norske

selskaper.
5.4 Autocall Telekom 39

5.4 Autocall Telekom

In this section we will value Autocall Telekom. The procedure will be similar as we did
above for Autocall Norske selskaper. Table 5.8 displays all the parameters used to model
the stock paths, and Table 5.9 displays the correlation between the underlying assets. As
Autocall Telekom have a quantos structure there is an additional parameter in the table,
the foreign risk-free rate. As mentioned earlier, when the domestic risk-free rate is higher
than the foreign risk-free rate we get a higher implicit dividend yield. This is the case for
Autocall Telekom’ foregin underlying assets, as can be seen in Table 5.8 below.

Ericsson Vodafone Telenor Nokia

Initial stock price 100 100 100 100
Volatility 0.248 0.263 0.167 0.354
Domestic risk-free rate 1.42% 1.42% 1.42% 1.42 %
Foreign risk-free rate -0.03% 0.71% 1.42% -0.47%
Dividend yield 1.98% 7.23% 4.88% 2.47%
Implicit dividend yield 3.43% 7.94% 4.88% 4.36%

Table 5.8: Parameters used to model paths for underlying assetsin

Autocall Telekom

Ericsson Vodafone Telenor Nokia

Ericsson 1.00 0.10 -0.07 0.18
Vodafone 0.10 1.00 0.22 0.25
Telenor -0.07 0.22 1.00 0.06
Nokia 0.18 0.25 0.06 1.00

Table 5.9: Correlation matrix of underlying assets in

Autocall Telekom

To get an intuition of the paths of the underlying assets in Autocall Telekom we have
made Figure 5.6. This figure illustrates price paths of the underlying assets of Autocall
Telekom for one simulation by using the parameters and correlation from Table 5.8 and
Table 5.9. The vertical lines in this figure illustrate the observation dates, and the dark
vertical line signals the first possible autocall date. In the paths displayed in Figure 5.6
the note would be autocalled at observation date 4, which is the first possible autocall
date. Note that we also here have used time-steps of one day for illustrative purposes.
40 5.4 Autocall Telekom

Figure 5.6: One path generation of Autocall Telekom’s four

underlying assets.

Table 5.10 presents the payoff specific parameters. The payoff specific parameters for
Autocall Telekom are almost identical as the payoff specific parameters for for Autocall
Norske selskaper. The only difference is that Autocall Telekom pays a lower coupon.

Payoff specific parameters

Autocall barrier 90%
Coupon barrier 80%
Risk barrier 60%
Notional NOK10,000
Quarterly accumulative coupon N otional × 3.65%
Discount rate 2.40%

Table 5.10: Payoff specific parameters used in

the valuation of Autocall Telekom.

5.4.1 Results

Table 5.11 below displays the results of the valuation algorithm using the parameters of
Autocall Telekom presented above. The precision of the price estimates increases with
number of iterations. From 1,000,000 iterations we get a price estimate of NOK6633.1,
with a standard error (SE) of NOK4.0, which gives us a 95% confidence interval between
NOK6625.2 and NOK6640.9.
5.4 Autocall Telekom 41

Iterations Lower c.i. Price estimate Upper c.i. SE Time

10 5, 296.5 8, 056.8 10, 817.1 1, 408.3 0.020 s.
100 5, 593.6 6, 425.8 7, 258.0 424.6 0.110 s.
1, 000 6, 244.7 6, 487.9 6, 731.2 124.1 0.930 s.
10, 000 6, 537.4 6, 615.5 6, 693.5 39.8 9.270 s.
100, 000 6, 611.9 6, 636.6 6, 661.3 12.6 1.71 min
1, 000, 000 6, 625.2 6, 633.1 6, 640.9 4.0 38.77 min

Table 5.11: Price estimates, confidence intervals, and standard

errors from our valuation model

These results suggests that also Autocall Telekom is extremely overpriced with a price
premium of 10,300/6633.1 - 1 = 55.28%. The most likely reasons why the price premium
is so much higher than the costs indicated in the prospect, are the same as we discussed
for Autocall Norske selskaper. It might be that the costs of constructing the product
exceeds the value of the product, or it might be that the estimated parameters have some
errors. In the next section we will see how sensitive the price estimate of Autocall Telekom
is to the parameters.

5.4.2 Sensitivity analysis

This section will see how sensitive the price estimate of Autocall Telekom is. We will
do this in the same manner as we did for Autocall Norske selskaper by examining how
the price estimate of Autocall Telekom changes as the parameters; volatility, correlation,
drift-rate, and coupon change.

Figure 5.7 displays how sensitive the price estimate of Autocall Telekom is to relative
changes in the volatility and correlation parameters. The figure illustrates that higher
volatility gives lower price estimates, and that lower correlation gives lower price estimates.
We can also observe the same findings as we did for Autocall Norske selskaper. Even
with relative changes of +/− 30% in the volatility and correlation parameters, the price
estimate would still be far below the price of NOK10,300 paid by investors. Based on
this, we still find our conclusion stating that Autocall Telekom is extremely overpriced to
be reasonable.
42 5.4 Autocall Telekom

Figure 5.7: Value-effect of a +/- 30% relative change in volatility

and correlation.

Figure 5.8 below displays how sensitive the price estimate of Autocall Telekom is to absolute
changes in the drift-rates of the underlying assets. The drift-rates in the valuation model
is determined by the Norwegian risk-free rate, and the implicit dividend yields. From the
figure we see that with an absolute increase of 5% in the drift-rates, the price-estimate is
NOK8,500. We find it unlikely that the estimated drift-rate parameter will have an error
greater than this. Thus, we still find it reasonable to conclude that Autocall Telekom is
overpriced.

Figure 5.8: Sensitivity analysis of drift, Autocall Telekom.

5.5 Summary of results 43

In Figure 5.9 we can see the sensitivity to the coupon rate. As for Autocall Norske
Selskaper, we can see that quarterly coupons would have to be 12% for the price of
Autocall Telekom to be ca. NOK10,000. This represents a coupon that is more than three
times higher than the actual coupon of 3.65%.

Figure 5.9: Sensitivity analysis of coupon rate, Autocall Telekom.

5.5 Summary of results

The numerical results of the valuation model suggests that both Autocall Norske selskaper
and Autocall Telekom are extremely overpriced. Deviations this large could be explained
by bad estimates of the input parameters. However, in the sensitivity analysis we saw
that even by adjusting the parameters to give a higher value, the price-estimates was still
far off the price of NOK10,300 paid by investors. Thus, based on the valuation model,
Autocall Norske selskaper and Autocall Telekom appears extremely overpriced.
44

6 Expected return and probability distribution

This chapter will examine what investors can expect from buying the autocallable notes.
We will investigate the holding period returns (HPR), the internal rate of return (IRR),
and the expected lifetime of both the Autocall Norske Selskaper and the Autocall Telekom
notes. Simulating 100,000 future values of the notes will provide a distribution for each
measure, which will give us an overview of the development of the notes. The HPR will
give insights of what investors can expect of total returns when holding the products.
The IRR will provide information of the notes’ dollar-weighted returns and the expected
annual returns. While the investigation of the maturity will give investors a greater
understanding of how long they can expect their funds to be locked in for. Contradictory
to Chapter 5 we will not use the risk-neutral valuation framework, thus we must make
some adjustments to the pricing model.

Subsection 6.1 below, addresses the model adjustments. In Subsection 6.2 and Subsection
6.3 we investigate each note respectively. Finally, Subsection 6.4 will summarize the
findings of the two autocall-specific subsections.

6.1 From risk-neutral to the true probability measure

In the previous chapter we used the risk-neutral valuation framework to simulate the fair
price estimates of the autocallable notes. In this chapter we are not interested in pricing
the products anymore. Instead, we are interested in how the notes are expected to develop
when using the real probability measure. The key difference between the two is that we
now must estimate the investors expected rate of return µi , and use this variable as the
main driver in the stock price model. Thus, we go from using the risk-free rate rf and the
equivalent martingale measure Q in Chapter 5, to the expected rate of return µi and the
real probability measure P in this chapter. As we are not interested in finding a present
value, we do not have to discount the Monte Carlo estimates anymore neither.
6.2 Autocall Norske selskaper 45

6.1.1 CAPM

As mentioned in Section 4.3, we will use the CAPM to estimate the underlying assets’
expected rate of returns. The CAPM requires certain parameters as can be seen in Eq.
4.1. These are the risk-free rate, the beta, and the market risk-premium. The risk-free
rate is already identified, whereas the betas are found using the Bloomberg terminal. The
risk-premium is identified through various sources, and will be discussed in the specific
sections below.

6.2 Autocall Norske selskaper

Most of the parameters needed to compute the expected rate of returns are already
outlined in Chapter 5, such as the risk-free rate, the implicit dividend, and the volatility.
In addition to these parameters we need the beta of each underlying asset and their
respective risk premiums. As mentioned, we get the individual betas from the Bloomberg
terminal, where the betas are calculated by a regression analysis based on five years of
monthly historical returns (Bloomberg L.P., 2021). For the Autocall Norske Selskaper note,
it is sufficient with the risk-premium of the Norwegian market as all the underlying assets
are listed in Norway. From a study by Ødegaard we find the appropriate risk-premium
in the Norwegian market to be 5% (2021). This is also confirmed in PWC’s annual
survey on the matter (PricwaterhouseCoopers, 2020). All parameters used to estimate the
expected returns of Autocall Norske selskaper are illustrated in Table 6.1. The table also
illustrate what the calculated expected rate of returns of each assets are, both annually
and continuously. The computations are done using the equation in Subsection 4.3, namely
Eq. 4.1.

To compute the holding period returns of the Autocall Norske Selskaper note, we run the
pricing model from Chapter 3 with the minor adjustment described in Section 6.1. As
we use the expected rate of returns, and skip the discounting of the Monte Carlo price
estimates, we get N simulated future cash flows. By dividing these future cash flows by
the notional value, we get N different HPR estimates.
46 6.2 Autocall Norske selskaper

Norsk Hydro Salmar Telenor Yara

Domestic risk-free rate 1.43% 1.43% 1.43% 1.43%
Beta 1.39 0.81 0.43 0.77
Risk-premium 5.00% 5.00% 5.00% 5.00%
Expected return 8.38% 5.47% 3.58% 5.27%
Continuous expected return, µi 8.05% 5.33% 3.51% 5.13%

Table 6.1: Parameters and expected returns of underlying assets in Autocall

Norske selskaper.

Figure 6.1 shows a distribution of 100,000 simulated expected holding period returns
(N = 100, 000). Looking at the figure, one can see a big spread in the distribution as the
simulated HPRs varies between -90% and 80%. The most frequent HPR is in the interval
of 10% - 20% however. The spike at this interval is caused by many notes being autocalled
at the first possible autocall-date. These notes will have a future payoff of 11,600, equalling
a HPR of 16%. There are also numerous outcomes with negative holding period returns.
For our simulation, there are nearly more negative holding period returns than positive!
For a note to receive a negative holding period return it must run to maturity and drop
below the risk-barrier. Thus, nearly half of the simulated notes dropped more than 40%.
It is worth noting that a note that goes below the risk-barrier can still have a positive
HPR if the coupon amount received during its lifetime is larger than the final date loss.
We also observe a high frequency of negative HPRs, falling between -30% and -70%.

Figure 6.1: Histogram of 100,000 simulated holding period returns

of Autocall Norske Selskaper note.
6.2 Autocall Norske selskaper 47

The findings from Figure 6.1 can also be illustrated in Figure 6.2. This figure illustrates
when, or if, the autocallable notes are autocalled. This statistic can be found by counting
the number of times the simulated notes are autocalled within each interval. The figure
indicates that 18.3% of the simulated notes were autocalled at the first possible autocall-
date, and 12.2% during the second year. The figure gives the intuition that if the notes
do not get autocalled during the first two years, they are very likely to run all the way
to maturity. 61.7% of the simulated notes did not get autocalled at all, and ran all the
way to the final date at year five. Only 13.4% of those ended above the risk-barrier and
as much as 48.3% ended below. Thus, 48.3% of the simulated 100,000 notes did not get
the notional amount of 10,000 back, which is confirmed by the high frequency of HPRs
between -30% and -70%. This illustrate a high probability of loosing a lot of money when
investing in the autocallable note.

Figure 6.2: Figure illustrating when or if 100,000 simulations of Autocall Norkse

selskaper are autocalled. Where 1=1st year, 2=2nd year, 3=3th year, 4=4th year,
5=5th year, 6=Not Autocalled, above riskb., 7=Not Autocalled, below riskb.

Knowing what to expect of annual returns when investing in the autocallable notes are
clearly in the best interest of the investors. As the lifetime and timing of the cash flows
of the autocallable notes are stochastic, we find the internal rate of return1 to be a
good alternative to estimate the annual returns. Thus, we have computed the annual
internal rate of returns of 100,000 simulations of Autocall Norske Selskaper, illustrated in
48 6.2 Autocall Norske selskaper

Table 6.2. The table show the distribution of those simulated IRR’s, where the different
IRR outcomes have been grouped into small blocks. From the row Sum: IRR > 0 we
observe that the investors can expect a probability of positive returns of about 52%,
and equivalently from the row Sum: IRR ≤ 0 a probability of 48% for negative returns.
The probabilities do not differ much when including the underwriting fees of 3% in the
simulation, as can be seen when comparing the two columns in the table. We can also see
that the positive returns are in the range of 12% to 18%, whereas the negative returns
are mostly in the range of -5% to -30%. The average IRR is however slightly negative.
This coincides with the above discussion where we saw that the note is most likely to be
autocalled either within the two first years, earning returns similar to the coupon, or run
to maturity and end below the risk-barrier earning large negative returns.

Internal rate of return Without fees With fees

≤ -30 % 3.66 % 3.97 %
-30 – -15 % 20.54 % 22.01 %
-15 – -5 % 19.80 % 18.42 %
-5 – 0 % 3.32 % 3.12 %
Sum: IRR ≤ 0 47.32 % 47.52 %
0–3% 4.70 % 5.45 %
3–6% 2.70 % 2.05 %
6–9% 1.60 % 1.5 %
9 – 12 % 2.09 % 2.93 %
12 – 15 % 9.38 % 40.55 %
15 – 18 % 32.20 % 0%
Sum: IRR > 0 52.68 % 52.48 %
Total: 100 % 100 %
Average IRR: - 0.95% - 2.28 %

Table 6.2: Table showing the distribution of Autocall Norske

Selskaper ’s IRR from 100,000 simulations. With and without
underwriting fees of 3%. The last row also displays the average
expected IRR.

1
The internal rate of returns were computed using a built in R-function called irr.
6.3 Autocall Telekom 49

6.3 Autocall Telekom

In Chapter 5 we identified the risk-free rate, the implicit dividend, and the volatility
of the underlying assets in Autocall Telekom. The beta of the underlying assets are
derived from the Bloomberg terminal just as for the underlying assets of Autocall norske
selskaper. Finding the correct risk-premiums requires more work for Autocall Telekom as
the underlying assets are based in different countries. The Norwegian risk-premium is
already identified as 5%. We use a survey by Pablo Fernandez to identify the risk-premium
in Sweden, Finland, and UK (2020). The risk-premiums are summarized in Table 6.3.
It is worth mentioning that the risk-premium from the survey of Fernandez suggests a
risk-premium in the Norwegian market to be 5.8% which differs from the 5% found in
Ødegaard’s and PWC’s surveys.

Ericsson Vodafone Telenor Nokia

Foreign risk-free rate -0.03% 0.71% 1.43% -0.47%
Beta 0.61 1.12 0.43 1.09
Risk-premium 6.10% 5.80% 5.00% 6.50%
Expected return 3.69% 7.22% 3.58% 6.62%
Continuous expected return µi 3.62% 6.97% 3.51% 6.41%

Table 6.3: Parameters and expected returns of underlying assets in Autocall

Telekom

Figure 6.3 shows the distribution of expected holding period returns from 100,000
simulations. The figure illustrates that Autocall Telekom have a similar return distribution
as the Autocall Norske selskaper note. We observe a large spread with HPRs between
-90% and 80%. Similarly to Autocall Norske Selskaper, we observe a large spike around
the interval of 10% - 20%, and frequent HPRs between -35% and -75%. However, by
comparing the figure with that of the Autocall Norske Selskaper, we observe that the
highest HPRs are a bit lower for Autocall Telekom. This is due to Autocall Telekom having
a lower coupon. The same findings can be seen in Figure 6.4 indicating that the note will
either autocall within the two first years, or go all the way to maturity. The early autocalls
explain the spike in the 10% - 20% HPR interval, as these instances return a HPR of
14.6% or higher. The frequent large negative HPRs are explained by the last column
in the figure, indicating that 51.3% of the notes end up going below the risk-barrier at
50 6.3 Autocall Telekom

maturity.

Figure 6.3: Histogram of 100,000 simulated holding period returns

for the Autocall Telekom note. Disregarding underwriting fees.

Figure 6.4: Figure illustrating when 100,000 simulated outcomes of

Autocall Telekom are autocalled, if autocalled at all. Where 1=1st
year, 2=2nd year, 3=3th year, 4=4th year, 5=5th year, 6=Not
Autocalled, above riskb., 7=Not Autocalled, below riskb.
6.4 Summary of results 51

Table 6.4 illustrates the distribution of IRR from 100,000 simulations of Autocall Telekom.
The results are again very similar to what we saw for Autocall Norske Selskaper. The
main difference is that Autocall Telekom have no outcomes in the IRR interval of 15%
to 18%. This is explained by the lower coupon in Autocall Telekom. We can also notice
that Autocall Telekom have a higher probability of negative IRR than Autocall Norske
selskaper. The table indicates that there is a higher probability of negative returns than
positive returns for Autocall Telekom.

Internal rate of return Without fees With fees

≤ -30 % 4.88 % 5.48 %
-30 – -15 % 24.80 % 26.16 %
-15 – -5 % 18.34 % 16.94 %
-5 – 0 % 2.87 % 2.80 %
Sum: IRR ≤ 0 50.89 % 51.38 %
0–3% 3.84 % 4.52 %
3–6% 2.24 % 1.68 %
6–9% 1.13 % 1.20 %
9 – 12 % 2.69 % 30.19 %
12 – 15 % 39.21 % 11.03 %
15 – 18 % 0% 0%
Sum: IRR > 0 49.11 % 48.62 %
Total: 100 % 100 %
Average IRR: - 3.04% - 4.47%

Table 6.4: Table showing the distribution of Autocall Telekom’s

IRR from 100,000 simulations, with and without underwriting fees.

6.4 Summary of results

In this chapter we studied the expected returns and returns distributions of the autocallable
notes. Our findings are similar for Autocall Norske selskaper and Autocall Telekom. The
results indicate large variations in potential returns where we have seen holding period
returns varying from up to 80% and down to -90%. A reason for the strong negative returns
is the exposure to unsystematic risk these notes offer. When there are four underlying
assets and the relevant underlying is the worst performing, the exposure to negative
returns is high. Another aspect is the relationship between when the notes are autocalled
and the returns. The results suggest that the notes are likely to either be autocalled early
52 6.4 Summary of results

and earn positive return as all the coupons are paid, if not, run to maturity and end under
the risk barrier. Overall, we believe that such relationship between risk and return, where
there is limited upside potential and high probability of strong negative returns will not
be found attractive by few investors.
 53

7 Analysis of prospect
In the previous chapters, Chapter 5 and Chapter 6, our results indicate that the autocallable
notes are not very attractive. The results suggest that the products are sold with price
premiums of ca. 50%, and that there are ca. 50% chance for negative returns for both
notes. Despite this, there have been sold Autocalls for billions of NOK and there are still
new notes frequently being offered in Norway. This suggests that there is an attractive
market for these products in Norway, and that people are willing to invest in these
products. This chapter will look at the Norwegian market of autocallable notes, identify
the sellers and customers, and investigate how the products are presented to potential
investors.

7.1 Market participants and their role

Structured products are typically structured by an issuing bank, facilitated by a facilitator,
and sold to customers through a distributor. For the autocallables presented in this
thesis, the issuer is Goldman Sachs, the facilitator is Garantum Fondskomission and the
products are distributed to the Norwegian market through the distributor Garantum
Norge. The latter is an agent acting on behalf of Garantum Fondskomission. Therefore
we recognise the facilitator and the distributor as one participant, under the name broker.
It is important to distinguish between the issuer and broker as they serve different roles
and earn profits from different operations.

The role of the issuer is to structure, value, and create prospects of the products. In the
process of structuring the products, the issuer faces certain costs related to the structuring
itself and hedging. To cover the expenses and earn money, the issuer adds a margin to
the constructed product. As we will see, it is hard to determine what this margin really
is because the issuer do not give any information regarding the value of the constructed
product, nor the costs of constructing it. For investors, this mean they will not be able
to know the true value of what they buy. And for the issuer this represent a conflict of
interest. The higher the price they are able to sell the product for, the more money they
will earn. As an example, imagine that the issuer constructs the product for NOK7,000,
54 7.2 Final terms

and get NOK9,000 of the total sales price of NOK10,300. This would mean earnings of
NOK2,000 for the issuer paid by the investor. One could argue that the distributor would
not agree to such terms. However, it is the issuer who have the relevant information
regarding the value of the product, and the distributor might not be aware of such high
margins. Even if the facilitator and distributor is aware of these margins they might
find it acceptable because they do not buy the product, they just act as middlemen and
operate with profits as long as they are able to sell the products.

The role of the facilitator and the distributor is hard to distinguish. From our
understanding, their job is to sell the products to potential investors and act as brokers.
Profits are strongly connected to number of notes sold. Based on this it also appears as
the distributors have conflict of interest. The more attractive they portray the products
the more they will sell and eventually earn. Earlier happenings can back the belief of
such conflict. There have for example been instances where Garantum Norge have under
communicated the potential risks of loss (Bjørklund, 2018). Another example is Garantum
Norge who did not allow media to attend at their investor presentation in 2018 (Jordheim,
2019). Articles in the media also suggest that these products have been aggressively
marketed to investor through direct mails and phone calls to potential investors (Bjørklund,
2018).

The distributor’s goal is to sell these products to investors. In the prospects of the
notes, it is stated that the targeted customers are both professional and non-professional
investors. People that invest in these products must understand them and view them
as good investments. The main material to get an understanding of the product and an
intuition of the products as investments, are the prospects and sales documents. In the
following sections we will give a review of the final terms documents which is a finalised
form of the prospect, and the sales brochures.

7.2 Final terms

The final terms documents are created by the issuer with purpose of meeting legal
requirements in the EU Prospectus Regulation. A final terms document is a version of the
base prospect, but with more specified characteristics. In the EU Prospectus Regulation
7.3 Sales brochure 55

article 8 they state: “Whereas the base prospectus contains options with regard to the
information required by the relevant securities note, the final terms shall determine which
of the options is applicable to the individual issue by referring to the relevant sections of
the base prospectus or by replicating such information” (Regulation (EU) 2017/1129 of
the European Parliament and of the Council, 2017). Most of the information in the final
terms documents are general. Thus, we will do a common review of the documents for
Autocall Norske selskaper and Autocall Telekom.

As the final terms and base prospects are made for legal reasons the main focus seem to
be presenting all the characteristics and formalities of the products in a clear manner.
It does not present any suggestions of whether or not the product described is a good
investment. The reader will however, get a good intuition of how the products work and
important formalities. One thing that caught our attention is that the conflict of interest
faced by the issuer is declared. On page 30 in both of the final terms documents, it is
written that the issuer is subject to a number of conflict of interests between its own
interests, and those who holds the security (Goldman Sachs International, 2021a,b). One
thing that we did miss was a better presentation of the costs incurred, in addition to a
discussion regarding the true value of the product.

Overall the final terms are very informative, but there is a lot of information and
terminology which can be overwhelming for non-professional investors. We see the
importance of such a document and believe that potential investors should become
familiar with it. However this is a demanding task, as there is a lot of information to
process. Thus, the risk of missing vital information is high.

7.3 Sales brochure

The sales brochure is created by the distributor Garantum with the purpose of marketing
the product. It presents much of the same information as in the final terms document,
but the information is more condensed with a greater focus on what the notes can offer as
an investment. Also for the sales brochures we will do a common review of the content in
Autocall Norske selskaper and Autocall Telekom as much of the information is the same.
56 7.3 Sales brochure

The content of the sales brochures summarise the features, risks, and costs of the
autocallable notes. Much of the content is the same as in the final terms documents, but
the content is more condensed and presented more reader friendly. The brochures present
the features, scenarios, risks, and important formalities in a way that gives the reader
a better intuition of what to expect when investing in the autocallable notes. However,
after we have carefully studied the sales brochures, we have made some remarks regarding
the presentation of the products.

The main remark is regarding the costs of the product and how they are presented.
Information about the costs are first encountered at page 2 in both documents (Garantum
Fondkommission, 2021a,b). Here the costs are presented in terms of amount and to whom
they are paid. It is specified that there will be an underwriting fee of 3% accrued to
the distributor. In addition, there is specified a margin of 6% where maximum 3.5% is
accrued to the distributor and the rest to the Garantum Fondkomission AB. We find this
presentation confusing as they have not specified who the distributor is and neither have
properly specified who Garantum Fondkomission is. In fact, the distributor is Garantum
Norge who act as an agent on behalf of Garantum Fondkomission. So, in total Garantum
Norge and Garantum Fondkomission charge fees of 9%. The way we interpret this is
that Garantum buys a product from Goldman Sachs for 9,400 and add total costs of 9%
relative to the notional amount of NOK10,000. At first glance this seem like all the costs.
However, if one reads the paragraph carefully, they state that there also is a cost falling
to the issuer. This cost is not specified, but they refer to a later section called Viktige
opplysninger where they better explain this cost. In this section the cost is specified to
cover the expenses occurred by the issuer (Goldman Sachs) when constructing, hedging,
and distributing the product. The cost is indicated to be 3.05% and 3.03% for Autocall
Norske selskaper and Autocall Telekom, respectively (Garantum Fondkommission, 2021a,b,
p. 6). To get a better understanding of the costs of Autocall Norske Selskaper :

1. Goldman Sachs makes a product worth NOK9,095. To cover costs incurred when
constructing the product, they add a margin of 3.05% * NOK10,000. As a result,
they sell the product for NOK9,400 to Garantum.

2. Garantum add facilitating costs of 6% *NOK10,000 and underwriting fees of

3%*NOK10,000. As a result, they sell the product for NOK10,300.
7.3 Sales brochure 57

This suggests that the investor pays a total price premium of 10, 300/9, 095 − 1 = 13.25%.
Relative to the notioanl value, this imply a total costs of 12.05%. For Autocall Telekom,
the corresponding price premium and total costs are 13.22% and 12.03%, respectively.
There are two additional aspects that make these findings even more interesting. Firstly,
they have a section discussing the total costs, but here they exclude the issuer margin.
Instead of referring to the total costs as 12.05% and 12.03%, they only refer to the costs
accrued to Garantum of 9%. We believe that this can be misleading for the investor.
Secondly, there are no discussions regarding the margins charged by the issuer. The
margins are indicated as a cost of 3.05% and 3.03%. A margin of 3.05% of the nominal
value would imply that Goldman Sachs values the autocallable note to 9,095. However,
there is no additional information regarding this value. The higher the value the issuer
can justify for the product, the less this indicated cost will be. This illustrates a prime
example of the conflicts of interest the issuer face. As we saw in the valuation chapter, the
values of both Autocall Norske Selskaper and Autocall Telekom are far below NOK9,095
and NOK9,097. This suggests that Goldman Sachs charge costs way above the 3,05% and
3.03% indicated in the sales brochures. Overall we find the communication of the costs to
be very unclear. They are presented more complicated and with more possibilities for
misinterpretation than necessary. It would be more transparent if they clearly stated the
value of the products and the costs of constructing them. This would give the reader a
better intuition of how much they actually pay in costs. A possible explanation to why
such information is not included in the prospects, nor in the sales brochures, is that it
would make the products appear more unattractive.

Another remark we made was regarding a figure presented. On page 4 they present a
histogram of when the autocallable note will be autocalled based on simulations using
historical data (Garantum Fondkommission, 2021a,b). In the document of Autocall Norske
selskaper 60% of the simulations are autocalled at the first possible date whereas ca. 10%
of the simulations go below the risk barrier. In Autocall Telekom’s document 40% of
the simulations are autocalled at the first possible date and ca. 20% of the simulations
go below the risk-barrier. These results contradict our findings in Chapter 6. The sales
brochures states that the simulations are based on historical values and are not reliable. If
this is true, why are they included in the brochures and why are the data and assumptions
58 7.4 Our thoughts of the supporting documents

used not presented?

What they write about the underlying assets and their associated industries, seems biased.
For instance, in the sales brochure of the Autocall Telekom note, Garantum portray the
Telekom industry as a very promising industry. Suggesting that it will revolutionize the
way we communicate, leading to a new way of living life (Garantum Fondkommission,
2021b, p. 3). The description of the industry is not necessarily wrong, but it has an
immense focus on the positive aspects and do not mention the risks of the Telekom
industry at all. Such narrative will potentially lead to a unfortunate intuition of the
underlying assets.

The specified coupons in the sales brochures and also the final terms are only indicative.
For Autocall Norske selskaper the indicative coupon is 3.5% and for Autocall Telekom
the indicative coupon is 4%. Both these coupons ended up being adjusted. For Autocall
Norske selskaper the coupon was adjusted upwards to 4% and for Autocall Telekom it was
adjusted downwards to 3.65% (Garantum Fondkommission, 2021c,d). In neither the sales
brochures nor the prospects it is clearly specified why the coupons are indicative. There
are also no information regarding which factors determine a potential coupon adjustment.
There is most likely a good reason for the indicative coupons but the reason for this should
be presented to ensure transparency, especially in the prospects.

7.4 Our thoughts of the supporting documents

Overall the supporting documents of the autocallable notes give an overview of the notes
and what to be aware of. The majority of important information is presented clearly which
gives the reader a good intuition of the products, especially in the final terms documents.
For the sales brochures, the information are overall clear but there are more room for
misinterpretation, and some of the information presented appears more biased. In both the
final terms and sales brochures we find some deficiencies which can result in uninformed
investment decisions. First of all, the incurred costs are not communicated clear enough.
Secondly, the value of the notes and the associated costs of constructing them, are not
specified. We believe that such information would help the investors to make a more
informed investment decision. Thirdly, we believe there are important information not
7.4 Our thoughts of the supporting documents 59

emphasised enough, or not emphasised at all. Neither of the documents give a thorough
explanation of the effect of dividends, the quantos structure, nor the unsystematic risk
exposure of the notes. All of these aspects represent a clear disadvantage to investors, and
it would perhaps portray the autocallable notes as more unattractive, if the information
were included. Both the final terms and the sales brochures present the autocallable notes
in a way that it is possible for potential investors to understand them. However, the
biased information suggests that the documents are presented in a way that do not give
the investors sufficient knowledge to do a well informed investment decision.
60

8 Discussion
In this chapter we will discuss and answer the main research question of the thesis. We
will also discuss strengths and weaknesses regarding the thesis itself, and regarding the
models implemented throughout the thesis.

8.1 Assessment of research questions

This section addresses the main research question of the thesis. Namely: Who reap the
benefits? The issuer, facilitator, distributor, or the investor? To get a proper foundation to
answer this question we must first address the three sub-questions introduced in Chapter
1. The first of which focuses on the price of both the Autocall Norske Selskaper and
the Autocall Telekom notes, and asks what would a fair price of the autocallable notes
be. The second question draws the attention to what investors could expect of returns
when investing in the notes. The third and final sub-question addresses how the notes are
presented to investors. Ultimately, we will conclude on the main research question of the
thesis.

8.1.1 What is the fair price of the products?

In Chapter 5 we conducted a numerical valuation of both the Autocall Norske Selskaper

and the Autocall Telekom notes. For both notes our results from the valuation chapter
were significantly lower than what are presented in the sales brochures by Garantum. For
Autocall Norske Selskaper, we found a price of NOK6, 955.7, while for Autocall Telekom
we found a price of NOK6, 633.1. These prices are excluding any margins, thus a fair
price of the products are obviously higher than these estimates. As discussed in Section
7.1, the autocallable notes are created by Goldman Sachs, facilitated by Garantum
Fondkommission, and finally distributed and sold to Norwegian investors by Garantum
Norge. Neither of these market participants do things without getting paid, implying
there must incur some fees to the products. Based on our price estimates, the price
premium of the Autocall Norske Selskaper and Autocall Telekom are 48.08% and 55.28%,
respectively. As Goldman Sachs get NOK9,400 for the products, this corresponds to
8.1 Assessment of research questions 61

35.14% and 41.71% of the price premiums of Autocall Norske Selskaper and Autocall
Telekom respectively. Garantum Fondkommission and Garantum Norge receive in total
NOK900 per autocallable note that are sold to investors. These NOK900 comprise 12.94%
and 13.57% of the total price premium of Autocall Norske Selskaper and Autocall Telekom
respectively. The question to be asked is whether the price premius are too high. We
will argue that these premiums are enormous. Of course, Goldman Sachs incur extra
costs regarding hedging of their exposure, however, premiums of 35.14% and 41.71% seem
enormously high. Even Garantum’s part of the premium, which is 9% of the notional
value is huge in the financial environment today, where funds are considered expensive
when exceeding fees of 2-3% (Lorvik, 2021). Especially considering it is Goldman Sachs
that bear the market risk of the autocallable structures, and not Garantum.

To summarise the first sub-question. We are in no position to tell what the fair price of
the autocallable notes really should be. However, we strongly believe that the autocallable
notes distributed and sold by Garantum are hugely overpriced.

8.1.2 What can investors expect in returns from the products?

In Chapter 6 we investigated the development of 100,000 simulated cases of both Autocall

Norske Selskaper and Autocall Telekom. This gave us a wider understanding of what
investors can expect when investing in these types of structured products. The analysis
included an investigation of the total returns, the annual dollar-weighted returns, and the
length of the investment period. The latter analysis also examined whether notes ended
above or below the risk-barrier if not getting autocalled at all. This analysis illustrated
that almost half of the 100,000 simulated notes ended up below the risk-barrier at year
five, which in most cases also led to a negative return for the investor. These results were
observable for both of the notes. The analysis for both the holding period return and the
analysis of the dollar-weighted annual returns revealed that the notes that fell below the
risk-barrier often incurred large losses as well. For holding period returns, the simulation
implied negative returns between -30% and -70% to be very common. Similarly, for the
dollar-weighted returns, annual returns between -5% and -30% were found to be very
likely. There are however opportunities for decent gains for the investors as well. The
most frequent annual return for Autocall Norske Selskaper was 16%, which is associated
62 8.1 Assessment of research questions

with the note being autocalled at the first possible autocall-date about 18% of the times.
For Autocall Norske Selskaper, we saw that investors would retrieve dollar-weighted annual
returns between 12% and 18% around 41% of the time. Similar tendencies were found
with the Autocall Telekom note.

Thus to summarise the sub-question, we believe that the investors could expect a high
probability of huge losses. Although, there is also a high probability of receiving returns
of about 12-18%. However, the probability of getting negative returns are larger than for
positive. Also the expected losses are much larger than eventual positive returns, leading
to a negative expected annual return for both notes.

8.1.3 How are the products presented to potential investors?

To answer the question regarding how investors are presented the autocallable notes, we
investigated two of the supporting documents enclosed with the notes, the final terms
and the sales brochures. The first documents we found to be very informative, providing
detailed information of most of the key aspects of the notes. Although, the documents are
quite large. In addition, there are used a lot of terminology which can be overwhelming
for one of the intended customer groups, the non-professional investors. We would also
like the documents to be both more clear regarding the incurred costs, and to provide
information regarding the true value of the notes. We also found the sales brochures to be
informative. However, these documents left room for more misinterpretations, and were
a little biased. The documents had a larger focus on the upside than for the downside,
which can give investors the wrong expectations. The major problem with the supporting
documents however, were the lack of information regarding the effects of regular dividends,
hidden dividends imposed by the quantos structures, and of unsystematic risk exposure.
These are elements that represents disadvantages for investors, and that could make the
notes less attractive if had been included.

To summarise the discussion regarding this sub-question, we believe that the investors are
provided with enough information so that it is possible to understand the autocallable
notes. However, the sales brochures provide biased information, and key information is
either lacking or not included in all of the supporting documents. Thus, we believe that
8.2 Strengths and weaknesses 63

the products are presented in a way that does not give the investors sufficient knowledge
to do a well informed investment decision.

8.1.4 Who reap the benefits?

Regarding the main research question: Who reap the benefits? The issuer, facilitator,
distributor, or the investor? The two autocallable notes analysed in this thesis appear
overpriced with unfavorable conditions and biased marketing for the investor. In Chapter
5 we found a value of these products that suggested a price premium of ca. 50% paid by
the investor. Further in Chapter 6, we identified unfavorable returns distributions and
high probability of losses. Lastly, in Chapter 7 we saw that these products are prone to
biased information in their marketing material, leading investors to invest in something
different than what they might expect. The results suggest that the benefits of these
products are reaped by the issuer, facilitator, and distributor at the expense of the investor.
This does not mean that the investor will not be able to earn returns, but it means that
the products incur high costs putting the investor in an unfavourable position.

8.2 Strengths and weaknesses

The Monte Carlo method used to value the autocallable notes have been able to perform
many iterations, leading the estimated values to converge to the true value. Thus, we
are certain that the value estimates are precise. To check the validity of the valuation
framework we valued another derivatives product, where we had a closed form solution.
The value estimates from this test converged to the closed form solutions, suggesting that
the numerical valuation model is valid. We also ensured that the historical correlation
were incorporated into the model, by creating a correlation matrix based on simulated
correlations between the underlying assets. This simulated correlation matrix were almost
identical to the historical correlation matrix.

The major weakness of the analysis in Chapter 5 and Chapter 6 are the historical data
which were used in the models. The parameters going into the model are only estimates of
the correct values. Furthermore, the Black & Scholes assume that these parameters also
are constant. This is seldom the case. However, the parameters provided to the model are
64 8.2 Strengths and weaknesses

our best guess of the correct values. In addition, we performed sensitivity analysis of the
key parameters of both Autocall Norske Selskaper and Autocall Telekom. These sensitivity
analyses revealed that for us to change our conclusion regarding the autocallable notes,
the estimated parameters would have to be substantially wrong.

Lastly, We did only investigate two autocallable notes in this thesis. To provide a more
generalized answer of the the research questions, we would preferably have analysed
substantially more notes. Regarding the scope of the thesis, this were however not feasible.
 65

9 Conclusion
We have in this thesis aimed to identify who reap the benefits of autocallable structured
products. To examine this, we investigated two autocallable notes named Autocall Norske
Selskaper and Autocall Telekom. Both of these notes were issued by Goldman Sachs,
facilitated by Garantum Fondkommission, and distributed and sold by Garantum Norge.
After conducting a numerical valuation of the notes, we found that both were extremely
overpriced, suggesting a price premium of ca. 50%. We investigated further what investors
can expect when investing in the two autocallable notes. We found that the probabilities
of ending up with negative returns were larger then ending up with a surplus. This finding
were present for both the Autocall Norske Selskaper and the Autocall Telekom notes.
What is even more shocking is that if ending up with negative returns, these were most
likely in the range of -5% to -30% annually. Thirdly, we investigated how the products
are presented to potential investors. In this analysis we found a lack of vital information
and supporting documents that seemed biased. These findings indicate that investors are
not given sufficient information to be able to make well informed investment decisions.
Ultimately, we believe that it is the issuer, the facilitator, and the distributor that reaps
the benefits of the autocallable strucutred products. This on the behalf of the investors.
66 References

References
Bjerksund, P., Carlsen, F., and Stensland, G. (1999). Aksjeindekserte obligasjoner – både
i pose og sekk? Praktisk økonomi og finans, 2:74–82.
Bjørklund, I. (2018). Vil ikke ha media på presentasjon av komplisert spareprodukt.
Dagens Næringsliv.
https://www.dn.no/marked/garantum/sparing/spareprodukter/vil-ikke-ha-media-pa-
presentasjon-av-komplisert-spareprodukt/2-1-479889, Last accessed 14th December 2021.
Bloomberg L.P. (2021). World bond markets: Europe, middle east and africa.
https://www.bloomberg.com/company/, Last accessed 4th December 2021.
Bodie, Z., Kane, A., and Marcus, A. J. (2018). Investments. McGraw-Hill, New York,
11th edition.
Brandimarte, P. (2014). Handbook in Monte Carlo simulations. Wiley, New Jersey.
Bøe, G. M. (2007). Analyse av strukturerte spareprodukt. pages 10–12.
Coval, J. D. and Shumway, T. (2001). Expected option returns. The journal of finance,
56(3):983–1009.
http://www-personal.umich.edu/~shumway/papers.dir/optret.pdf, Last accessed 20th
November 2021.
Damodaran, A. (2020). Ratings, interest coverage ratios and default spread.
https://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/ratings.htm, Last
accessed 23rd November 2021.
Deng, G., Mccann, C., and Mallett, J. (2011). Modeling autocallable structured products.
Journal of Derivatives Hedge Funds, 17:326–340.
Fernandez, P. (2020). Survey: Market risk premium and risk-free rate used for 81 countries
in 2020.
Garantum Fondkommission (2021a). Autocall norske selskaper kvartalsvis pluss/minus nr
4265.
https://www.garantum.se/webapi/pdfservice/getpdfdocumentInfoHub?id=2364e528-
1fd2-435f-8918-9dc8bb15c5d2&filename=ProductBrochure_NO0011109530_NOK_sv.pdf&typeid=
41, Last accessed 15th December 2021.
Garantum Fondkommission (2021b). Autocall telekom selskaper kvartalsvis pluss/minus
nr 4266.
https://www.garantum.se/webapi/pdfservice/getpdfdocumentInfoHub?id=a0e96316-
8809-4233-983c-622380998ae3&filename=ProductBrochure_NO0011109555_NOK_sv.pdf&typeid=
41&fbclid=IwAR0-A6UwMaM0A0pmwzY7ahbMYiL89HwZH84y5AAIb8nimIDjQIIWZf4DEJ8,
Last accessed 15th December 2021.
Garantum Fondkommission (2021c). GS AC Norske Sel PM 4265 NOK.
https://www.garantum.no/Produktinformasjon/aktuelle-tilbud/produktside/?isin=
NO0011109530&currency=NOK&fbclid=IwAR3phLSBCdv4LcdeGpxOtWziRFgEs-
L0divDh6PhVNU6G_BUquw44bGKHgY, Last accessed 15th December 2021.
References 67

Garantum Fondkommission (2021d). GS AC Telekom Sel PM 4266 NOK .

https://www.garantum.no/Produktinformasjon/aktuelle-tilbud/produktside/?isin=
NO0011109555&currency=NOK&fbclid=IwAR2vYaeIzk6AaevvBPoSLS2nK-
VN95FkJEnHDpEofEGWAnhmclITXixhsHA, Last accessed 15h December 2021.
Glasserman, P. (2003). Monte Carlo Methods in Financial Engineering. Springer.
Goldman Sachs International (2021a). Final terms autocall norske selskaper.
https://www.garantum.se/webapi/pdfservice/getpdfdocumentInfoHub?id=60e819f1-90f7-
4adf-940d-4a4f994be5c7&filename=ProductFinalTerms_NO0011109530_NOK_sv.pdf&typeid=
42&fbclid=IwAR0nvaaA5RRlT6z51OEvrhkbD9Fe6Q5z1hD4KW-
fiWYSWI3e7g5cfVQf4wY, Last accessed 15th December 2021.
Goldman Sachs International (2021b). Final terms Autocall Telekom.
https://www.garantum.se/webapi/pdfservice/getpdfdocumentInfoHub?id=3331c26f-
9439-41c5-9ac2-f859be12ea60&filename=ProductFinalTerms_NO0011109555_NOK_sv.pdf&typeid=
42&fbclid=IwAR0R3AF9RzkASIKIAIMJ3QWGO5mquhywDXdBdfTRzTFCqSZMQgnC-
C8TAH8, Last accessed 15th December 2021.
Hansson, F. (2012). A pricing and performance study on auto-callable structured products.
pages 32–33.
Haug, J. (2021). Modeling prices as diffusions [pdf slides].
Canvas.https://nhh.instructure.com/courses/1336/files/folder/Lecture%
20notes?preview=196099, Last accessed 14th December 2021.
Haugh, M. (2004). The monte carlo framework, examples from finance and generating
correlated random variables.
Helgerud, A. (2012). En analyse av strukturerte spareprodukter før og etter innføringen
av mifid-direktivet. pages 4–9.
Hull, J. C. (2018). Options, Futures, and other Derivatives. Pearson.
Jordheim, H. M. (2019). Finanstilsynet: Meglerhus har gitt galt bilde av høyrisikable
produkter. E24.
https://e24.no/boers-og-finans/i/dlLGEq/finanstilsynet-meglerhus-har-gitt-galt-bilde-av-
hoeyrisikable-produkter, Last accessed 14th December 2021.
Lorvik, N. (2021). Spareøkonom advarer mot høye gebyrer.
https://www.nettavisen.no/okonomi/her-er-verstingfondene-mange-vet-ikke-hva-de-
betaler/s/5-95-322270, Last accessed 14th of december 2021.
McDonald, R. L. (2014). Derivatives Markets. Pearson.
MIT, O. (n.a.). Error estimates for the monte carlo method.
https://ocw.mit.edu/courses/aeronautics-and-astronautics/16-90-computational-
methods-in-aerospace-engineering-spring-2014/probabilistic-methods-and-optimization/
error-estimates-for-the-monte-carlo-method/, Last accessed 20th November 2021.
Osphare-Druilhe, N. (2021). BNP Paribas Wealth Management: Understanding Structured
Products. BNP Paribas Wealth Management.
https://wealthmanagement.bnpparibas/en/expert-voices/understanding-structured-
products.html, Last accessed 12th November 2021.
68 References

PricwaterhouseCoopers (2020). Risikopremien i det norske markedet.

https://www.pwc.no/no/publikasjoner/pwc-risikopremie-2020.pdf, Last accessed 30th
November 2021.
Regulation (EU) 2017/1129 of the European Parliament and of the Council (2017). Article
8, The base prospectus.
https://www.legislation.gov.uk/eur/2017/1129/article/8#, Last accessed 8th December
2021.
SSB, S. N. (2009). Finansforetak, balanser, mars 2009.
https://www.ssb.no/bank-og-finansmarked/statistikker/orbofbm/maaned/2009-05-
08#content, Last accessed 12th December 2021.
Yahoo! (n.a.). What is the adjusted close?
https://help.yahoo.com/kb/SLN28256.html, Last accessed 14th December 2021.
Ødegaard, B. A. (2021). Risikopremien i det norske markedet.