Risk Management and FInancial Engineering_Final Report
-
Upload
moises-martinez -
Category
Documents
-
view
41 -
download
0
Transcript of Risk Management and FInancial Engineering_Final Report
Royal melbourne institute of technology
Risk Management and Financial Engineering
Empirical testing of Binomial, Black-Scholes and Trinomial Models
Henry Santosa (s3583121), Jaimie Comungal (s3454408), Moises Martinez (s3434618), Naeem Qudeer (s3524294) and Muhammad Usman Khan (s3511699)
Instructor:Prof. Malick Sy.
Course:Risk Management and Financial Engineering (BAFI2081)
The aim of this report is to test the conventional models for covered warrant call option prices and evaluate which model is statistically more reliable along with the suggested statistical inferential modifications in those models
Table of Contents
Executive Summary:..............................................................................................................................1
1 – Covered, Normal Equity Warrants and Options...............................................................................2
2 – Equity Warrants...............................................................................................................................4
3 – Covered Warrants:...........................................................................................................................7
3.1 – Types of Covered Warrants:.....................................................................................................8
4 – Option Pricing Models – A comparison of Binomial, Black-Scholes and Trinomial model:............10
4.1 – Black-Scholes Model:..............................................................................................................10
4.2 – Binomial Model:......................................................................................................................12
4.3 – Trinomial Model:....................................................................................................................13
5 – Comparison between Option Pricing Models & their calculation methodologies:........................15
6 – Theoretical modifications in Binomial, Black-Scholes and Trinomial:............................................16
7 – Statistical evaluation of the models:..............................................................................................17
7.1 – Data collection and Assumptions:...........................................................................................19
7.1.1 – Assumptions:...................................................................................................................20
7.2 – Statistical Tests and findings:..................................................................................................20
7.3 – Modifications of the Binomial, Black-Scholes and Trinomial models:.....................................24
8 - Conclusion......................................................................................................................................26
9 – References.....................................................................................................................................27
Appendices:.........................................................................................................................................29
List of Appendices Appendix 1 – Binomial Model for CapitaLand Warrant
a) Simple regression output b) Correlogram of simple regression outputc) Serial correlation test for simple regression outputd) Heteroskadacity (White) test for Binomial Model e) Heteroskadacity (ARCH) test for Binomial Model f) Modified Binomial Model for CapitaLand g) VIF test for Modified Binomial modelh) Correlogram for Modified Binomial model i) Static forecasts for Modified Binomial model j) Dynamic Forecasts for Modified Binomial model
Appendix 2 – Black-Scholes Model for CapitaLand
a) Simple regression output b) Correlogram of simple regression outputc) Serial correlation test for simple regression outputd) Heteroskadacity (White) test for Binomial Model e) Heteroskadacity (ARCH) test for Binomial Model f) Modified Binomial Model for CapitaLand g) VIF test for Modified Binomial modelh) Correlogram for Modified Binomial model i) Static forecasts for Modified Binomial model j) Dynamic Forecasts for Modified Binomial model
Appendix 3 – Trinomial Model for CapitaLand
a) Simple regression output b) Correlogram of simple regression outputc) Serial correlation test for simple regression outputd) Heteroskadacity (White) test for Binomial Model e) Heteroskadacity (ARCH) test for Binomial Model f) Modified Binomial Model for CapitaLand g) VIF test for Modified Binomial modelh) Correlogram for Modified Binomial model i) Static forecasts for Modified Binomial model j) Dynamic Forecasts for Modified Binomial model
I
Appendix 4 – Binomial Model for SembCorp Marine
a) Simple regression output b) Correlogram of simple regression outputc) Serial correlation test for simple regression outputd) Heteroskadacity (White) test for Binomial Model e) Heteroskadacity (ARCH) test for Binomial Model f) Modified Binomial Model for CapitaLand g) VIF test for Modified Binomial modelh) Correlogram for Modified Binomial model i) Static forecasts for Modified Binomial model j) Dynamic Forecasts for Modified Binomial model
Appendix 5 – Black-Scholes SembCorp Marine
a. Simple regression output b. Correlogram of simple regression outputc. Serial correlation test for simple regression outputd. Heteroskadacity (White) test for Binomial Model e. Heteroskadacity (ARCH) test for Binomial Model f. Modified Binomial Model for CapitaLand g. VIF test for Modified Binomial modelh. Correlogram for Modified Binomial model i. Static forecasts for Modified Binomial model j. Dynamic Forecasts for Modified Binomial model
Appendix 6 – Trinomial Model for SembCorp Marine
a. Simple regression output b. Correlogram of simple regression outputc. Serial correlation test for simple regression outputd. Heteroskadacity (White) test for Binomial Model e. Heteroskadacity (ARCH) test for Binomial Model f. Modified Binomial Model for CapitaLand g. VIF test for Modified Binomial modelh. Correlogram for Modified Binomial model i. Static forecasts for Modified Binomial model j. Dynamic Forecasts for Modified Binomial model
II
Appendix 7 – Binomial Model for Wilmar International Warrant
a. Simple regression output b. Correlogram of simple regression outputc. Serial correlation test for simple regression outputd. Heteroskadacity (White) test for Binomial Model e. Heteroskadacity (ARCH) test for Binomial Model f. Modified Binomial Model for CapitaLand g. VIF test for Modified Binomial modelh. Correlogram for Modified Binomial model i. Static forecasts for Modified Binomial model j. Dynamic Forecasts for Modified Binomial model
Appendix 8 – Black-Scholes Wilmar International Warrant
a. Simple regression output b. Correlogram of simple regression outputc. Serial correlation test for simple regression outputd. Heteroskadacity (White) test for Binomial Model e. Heteroskadacity (ARCH) test for Binomial Model f. Modified Binomial Model for CapitaLand g. VIF test for Modified Binomial modelh. Correlogram for Modified Binomial model i. Static forecasts for Modified Binomial model j. Dynamic Forecasts for Modified Binomial model
Appendix 9 – Trinomial Model for Wilmar International Warrant
a. Simple regression output b. Correlogram of simple regression outputc. Serial correlation test for simple regression outputd. Heteroskadacity (White) test for Binomial Model e. Heteroskadacity (ARCH) test for Binomial Model f. Modified Binomial Model for CapitaLand g. VIF test for Modified Binomial modelh. Correlogram for Modified Binomial model i. Static forecasts for Modified Binomial model j. Dynamic Forecasts for Modified Binomial model
III
Risk Management and Financial Engineering (BAFI2081)
Executive Summary: The most widely used models for options and warrants pricing are Binomial model, Black-Scholes
model and Trinomial model. This report defines and tests these models with respect to the covered
warrants (calls) of three companies listed at the Singapore Stock Exchange. The foremost aim is to
test the models to analyze which one of the aforementioned models gives the most reliable warrant
prices with least error.
The report will start with a brief comparison of simple options versus those of warrants and covered
warrants. The focus is both in terms of their functions and how these instruments are used in
present day. A mathematical comparison of all the aforementioned models is also included which
compares the mathematical differences in the models. Since the trinomial model includes an
additional condition of stationary stock price, which Binomial does not include, and Black-Scholes
have long been criticized as inaccurate, it is expected that the Trinomial model would statistically be
more reliable. The tests were performed through simple regressions, followed by tests for serial
correlation, heteroskedacity and multicollinearity.
Based on a journal article and the results provided by the statistical tests, these models were then
modified using an Autoregressive and Moving Averages (ARMA) structure. Each model was then
fixed as per the nature of the statistical inference that suited the model. Finally for comparison
purposes, the modified models were then used for static and dynamic forecasts to gauge their
reliability. The root mean squared errors (RMSE) of the modified models were compared with that of
the actual warrant price data and individual warrant price predicted by each of the original model.
1 | P a g e
Risk Management and Financial Engineering (BAFI2081)
1 – Covered, Normal Equity Warrants and OptionsEquity options are known as contract that provide holders the right – though not the obligation – to
buy (call option) or sell (put option) shares of the underlying security at a particular price either on or
before a determined date. Those who grant these rights are known as the seller or the writer of the
contract. Although the sale of options are generally considered unprofitable, some investors still sell
options in order to hedge other positions, and ultimately buy futures (Bollen, NPB & Whaley, RE
2004).
Two of the most commonly accepted forms of equity options are the options on stock indexes or
sub-indexes, and options on individual shares.
Options are used by a large number of investors as a form of risk reduction in portfolios, to lock in
their target rates of returns, provide crisis insurance, and possibly even enhance their returns from
equity portfolios. Parallel to futures contracts, options also provide low transactions costs for rapid
asset allocation, making them desirable to strong investors (Equity Options, 1992).
There are large demands for options in the market. Reasons include, firstly, because the risks built in
the stock markets cannot be hedged against by trading stocks alone. Furthermore, Carr and Wu
(2009) argue that the presence of stochastic volatility causes the market for options to become the
true market to trade volatility risk. In addition, investors may decide to gain exposure to various
shares due to the higher leverage provided by the options (Black, 1975). Informed traders may also
prefer trading options as it allows them to shield themselves through multiple option contracts that
are available for one security (Easley, O’Hara, and Srinivas 1998).
All in all, we see that options provide many benefits and are widely used within the investment
world for many purposes. Therefore, we look further into the structure and nature of options to
obtain a more sophisticated understanding of options and how they are priced.
Beginning with the structure, we can define equity options by the following elements: their type (call
or put option), their style (American or European), their underlying security, units of trade, exercise
price and their expiration date. Going through each in more detail, we begin distinguishing the
difference between call and put options.
Call options refer to the right of the holder to buy shares at the exercise price either before or on a
determined date. On the other hand, put options provide the holder the right to sell shares at the
exercise price either before or on a determined date.
2 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Underlying security of an option refers to the shares/assets, at which the option will derive their
value hence being categorized as a derivative security. Since equity options often have a ‘unit of
trade’ of 100 shares, we may state that we can buy or sell 100 shares of the underlying security for
one option contract.
Next, we consider the exercise price (also known as strike price). This price represents the price at
which the option holder can buy or sell the shares of a company if he decides to exercise his right
against the writer of the contract. This price is often set close to the market price of the share. An
option may be considered in-the-money, out-the-money, or at-the money depending on the
difference between strike and market price of its underlying security. A call option would be in-the-
money if its exercise price is less than the current market price because this allows the holder to buy
the share for less than what it is currently priced at in the market. Similarly, a put option would be
in-the-money if the exercise price were greater than the current market price, as this allows the
holder to sell the share at a price higher than that being currently offered in the market. Therefore,
an option will be out of the money when the strike price is greater than the current market price for
a call option, and when the strike price is less than the current market price for a put option. Lastly,
if the strike price is equivalent to the current market price, we would state that the option is at the
money.
The price of an option is known as the premium which holders/buyers of the options must pay for
the right to buy or sell the underlying security at a locked price in the future. This premium is paid to
the writer of the option in exchange for the seller’s obligation to deliver the underlying security at
the strike price, if the buyer decides to exercise his call option (of to receive delivery if buyer decides
to exercise his put option). The premium is made up of two values: the intrinsic value and the time
value of money. The difference between the option premium and the intrinsic value makes up the
time value, which is ultimately affected by factors such as the interest rates, volatility, time to expiry,
share price, and dividend.
In regards to the style of an option, generally exchange-traded equity options are American-style
options and involve providing the option holder the right to buy or sell his option at any time prior to
the expiration date.1 If the option has not been exercised by this date, the option becomes worthless
and ceases existing as a financial instrument. Once a transaction is closed, this revokes the investor’s
previous position as the holder or writer of the option.
Finally, in order to exercise an option, the holder must advise his broker to submit an exercise notice
to the write, who then has to fulfil their obligations as stated in the contract. 1 A European option allows an option to be exercised any time to the date of expiration.
3 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Equity options are favoured by investors for its flexibility in dealing with intricate risks – especially,
risks regarding foreign equity price volatility and the foreign exchanges. In regards to foreign equity
options, there are various types with various payoff functions. This may provide the investors with
more choices for risk management and investment. Another reason why equity options are favoured
is because of exchange trade, meaning that the financial product has more liquidity. On top of that,
regulations allow investors to minimise other risks, e.g. counterparty risk. (Fan, K., Shen, Y., Siu T.K.,
Wang, R., 2013)
2 – Equity Warrants Now that we understand what equity options are, we now turn our focus to equity warrants, which
are quite similar in the way that they also provide investors with the right to buy a share at a
particular price at a specific date. The difference is that equity warrants and the actual shares can be
issued by the company themselves. Doing this allows the company to raise money and is also
beneficial for the investor as the warrants are often offered at a price lower than equity options.
Investors then profit by trading the warrants before the date of expiration, which is when the
warrant issuer must fulfil the investor’s requirements (Fung, H., Zhang, G. and Zhao, L., 2009).
Through equity warrants, investors can also enjoy them as a hedging tool. Specifically, a put warrant
could operate as a type of insurance policy for an investor’s share portfolio safeguarding them in the
event that the share price falls.
Equity warrants differ further in that the level of supply and demand within the market, or even its
trading volumes, do not determine its price. Instead, issuers of warrants price them according to
specific pricing models such as the Black Scholes model (Kui, L., 2007).
Another distinction between equity options and equity warrants is that warrants can be traded
continuously and more frequently. They also usually have lower transaction costs, which may be
because the value of the warrants contract may be lower than the value of the underlying asset,
which in turn result in lower brokerage and transactions fees. Further, warrants can be traded
directly through the interest systems provided by various brokers (Hunt, B. and Terry, C., 2011). As a
result, warrants are often favoured by retail investors who seek less intricate processes.
To trade the warrants, they are often put on a nation’s traditional stock exchange, such as the
Australian Stock Exchange (hereinafter ASX) or the Singapore Stock Exchange (SGX), similar to how
shares are usually traded. For example, in Australia, investors would have to use ASX’s equity trading
system, ASX Trade, which allows transactions to be certified and cleared through the use of CHESS 2.
2 Clearing House Electronic Subregister System (CHESS)
4 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Further, using CHESS trading and its settling arrangements provides investors with more familiarity
than trading the options on the ASX, which are usually associated with margin trading.
When we consider their trading volume, we learn that there is an unmistakable pattern during the
trading day. Specifically, at the beginning and at the end of the day, we see that the volume of equity
warrants traded is much higher than any other time of the day. This may be because hedge traders
are more inclined to control positions that they were unable to during the night, as well as,
safeguard their open stock positions which are to be executed during the day. Similarly, speculative
traders may engage in trading during the time in order to exploit the fact that information declared
during the night has not been included in asset prices. We may see this occurring even in Australian
markets through the table and graphs given below. Evidently, we may conclude that during these
times of the trading day, demand for equity warrants may be much more inelastic (Segara, L &
Segara R, 2007).
5 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Table 1: Intraday patterns in the equity warrants
6 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Figure 1: Equity warrants trading volume
3 – Covered Warrants: A Covered Warrant is a listed security that is distributed by a financial organisation, of which
thereafter, is formed to be available for trade on a stock exchange. Covered warrants provide the
holder the right, but not a definite commitment in selling or buying the asset, at a certain amount
whether it was before or during the specific prearranged point in time of the underlying asset’s
expiry date (Chan et al. 2012).
For the masses, the similarity between options and covered warrants are significant and can be seen
that both are alike, however, a covered warrant ordinarily comprises of a faster maturity in
comparison to options, and is issued with a larger variety of assets (Aitken & Segara 2005). In
addition, the terms presented in covered warrants vary highly compared to options and to a greater
extent, are more flexible with its structure in meeting what the market demands.
On average, covered warrants usually have an ordinary lifespan between six to twelve months while
others may contain greater and/or unrestricted arrangements. The privilege of not being obligated
7 | P a g e
Risk Management and Financial Engineering (BAFI2081)
in exercising any right whether to buy or sell an underlying asset means any loss is limited to the
initial investment.
3.1 – Types of Covered Warrants:There are various types of covered warrants available and in circulation globally, the most prominent
being on stock warrants. Other covered warrants are on:
Baskets; a group of stocks that allows investors to gain exposure in the simplest form just by
purchasing a security in an arrangement of a covered warrant.
Indices; incorporating the overall market of covered warrants globally, indices are the most
heavily traded kind.
Bonds; bonds with option rights
Commodity; gives investors the capability in taking positions of commodities using covered
warrants.
Currency; accessible on a collection of exchange rates.
Barrier; additional terms where if a particular price is hit, the covered warrant will be void.
Terms can also be changed if the barrier price is met dependent upon the initial conditions.
Trigger; triggering a specific matter within the terms that leads to a fixed payout.
The addition of covered warrants as another financial instrument paved way for investors to have a
wider range of options to utilise (Horst & Veld 2008). Similar to options, the existence of covered
warrants can be found in two basic forms; a call and a put. These allow investors of covered
warrants to prevail as a result of either a rise or fall in the market. The maturity of covered warrants
is well established in advance, and dependent on its structure, it usually is the last day for which the
warrant could be exercised.
In obtaining the right for the issuance of warrants, it is relatively complex and requires the potential
issuer in meeting a stringent criterion that are set out in order for them to be approved. Over each
underlying security, warrant issuers offer a wide range of strike prices for investors, allowing them to
choose on the basis of how they perceive the market. The exercise style of covered warrants
however is also similar to those of options; an American or European style. In practice, in spite of the
differences between the exercise styles, these have small-scaled influences over the pricing of
covered warrants on the basis that selling the instrument is usually more profitable than exercising
the warrant earlier during its lifetime.
8 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Prices of covered warrants have two elements similarly to those of options; intrinsic and time value.
The first relates to the value of the instrument of being exercised instantly while the latter reflects
the time to its maturity date. The time value of a covered warrant decreases rapidly as it approaches
the expiry date, indicating warrants nearing maturity carry a higher risk. The gearing and leverage
offered by a covered warrant are one of the underlying reasons that attract investors. In other
words, covered warrants more than what you initially invested from price movements (leverage) and
exposure (gearing).
There are a number of risks associated with covered warrants ranging from counterparty risk, the
performance of the underlying asset to currency risks. The heavy stringent criterion that exist in
allowing an institution to issue warrants minimises the credit risk involved in negotiating with the
issuer. However, the risk is still present. A covered warrant is seen as a contract between two parties
(issuer and the holder) where the holder bares the risk of the warrant issuer not performing the
contractual obligation in the instrument.
The fundamental factor in measuring the outcome of success solely relies on the performance of the
underlying asset in question. In cases where a call warrant is present, its success relies on the
underlying assets value to rise and vice versa with put warrants. In a case where the value of the
underlying asset remains steady over time, the losses are limited to the costs incurred in obtaining
the warrant.
Where an antagonistic movement in value of the underlying asset against the desired expectation of
the investor occur, this is known as market risk. There is also a risk in the ability to offload the
warrant in the market for a suitable price. This could be due to a deficiency of liquidity in the actual
underlying asset alone therefore decreasing such demand for the particular warrant on offer.
9 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Table 2.0 – Covered warrants in circulation (2005-2007)
Covered warrants on overseas indices or foreign currencies expose the investor to currency risks
when an unfavourable movement in the currency occurs with regards to the relevant exchange rate
against a particular currency. Dependent on the terms within the warrant, the issuer may reserve
the right to terminate, withdraw or cancel the warrant on certain triggers like an extraordinary event
with regards to the underlying asset.
Geng, Qi Ding and Zhang (2013) discuss that a particular warrant’s size on a particular stock and its
level of liquidity will have a substantial impact on the price of the warrant. Their research looks into
the market of warrants in China and discuss that small & individual investors made up of most of the
warrants. Furthermore, it was also stated that the lack of understanding or lack of financial
knowledge in simply assuming and predicting the market were going to go up or down amongst
investors also played a major role in the price of warrants in the market.
4 – Option Pricing Models – A comparison of Binomial, Black-Scholes and Trinomial model: There is a high level of complexity in valuing an option as an option contract. It is dependent upon
the number of different variables which may affect the price of underlying assets. Over time there
have been many different types of models introduced to deal with the complexity of valuing options.
The most widely known models to date are Black-Sholes, Binomial and the Trinomial model. It is
necessary to have an idea about the difference between European style options and American style
options, in order to discuss the applicability of these models.
European style options: European style option is an option which cannot be exercised
before the expiration date.
American style options: American style option is an option that can be exercised at any time
during the life of that option.
4.1 – Black-Scholes Model: Black-Scholes model was developed by Fischer Black and Myron Scholes in 1973 (Black and Scholes
1973). Black-Scholes model is considered one of the best theoretical models for pricing a European
option and in due time, became one of the most foundation concepts within the realm of modern
financial theory. Its basic principles are used in the formulas found today for the evaluation of
almost all options. Black-Scholes pricing formulas for call options and put options are given below:
10 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Where:
C = call value
P = put value
S = stock price
K = exercise price of an option
T-t = time between expiration date and the valuation date of an option
r = risk free interest rate
N(d) = standard normal cumulative distribution in point d
∂ = volatility of underlying index
It can be observed from the above formula that an option with a higher volatility will be more
valuable in comparison to a one that has a lower volatility. Furthermore, the higher the ratio of stock
price to the exercise price is, the higher price it will be for that option.
4.1.1 – Assumptions:
The basic underlying assumptions from the original Black-Sholes model are:
The option is a European style option, which means that it cannot be exercised before the
expiration date
It assumes that the volatility of the underlying stock is constant.
There is no arbitrage due to the efficient markets.
Like volatility it assumes risk free interest rate remains constant over the time.
Return on the underlying stock follows a normal distribution
Markets are always open, giving an opportunity to buy or sell any option at any given time.
No transaction costs involved in buying or selling an option.
Zero dividends are paid during the option life.
11 | P a g e
Risk Management and Financial Engineering (BAFI2081)
4.2 – Binomial Model: Binomial option pricing model was invented by Cox-Rubinstein in 1979. He invented it as a tool for
the explanation of Black Scholes model to his students but soon it was found to be a more accurate
model in pricing American style options (Cox and Rubinstein 1979)). American style options are the
options which can be exercised before the date of expiration. Binomial model divides time to
expiration into a large number of short time intervals and produces a tree of prices working forward
from present to expiration step by step. It assumes that the value of current stock will either go up
or down in a certain time period. One step binomial model for option pricing is given below
One Step Binomial Model
Where So is the initial price of stock, p is the probability that value of stock will go up by factor u and
1-p is a probability that stock price will fall down by factor d.
A risk neutral world is assumed over a small period of time, given that the effective return of
binomial mode is equal to risk-free rate.
.
And also the variance of risk-free asset is equal to the variance of an asset in a risk-neutral world
given by the following equation.
The relation between upside factor and a downside factor is given by:
12 | P a g e
Risk Management and Financial Engineering (BAFI2081)
From above equations, values for p, u and d have been obtained as follows:
The values of p, u and d given by the Cox, Ross and Rubinstein (CRR) model means that the
underlying initial stock price is symmetric for a multi-step binomial model.
Two Steps Binomial Model
4.3 – Trinomial Model: Trinomial option pricing was proposed by Boyle in 1986 as an extension of the binomial model
(Boyle 1986). Trinomial model is considered to be more of an advanced form of a binomial model as
it gives three possible values that an underlying asset in a certain time period can be greater than,
less than or same as the current value of stock. The Trinomial model contains a third possible value
which assumes a zero change in the value of the stock makes this model more appropriate to deal
with the real life situations. Trinomial tree can be defined as
13 | P a g e
Risk Management and Financial Engineering (BAFI2081)
S(t)u with probability pu
S(t + ∆t) =
S(t) with probability 1 − pu − pd
S(t)d with probability pd
According to no arbitrage condition we have
E[S(ti+1)|S(ti)] = er∆tS(ti) …………………(a)
Var[S(ti+1)|S(ti)] = ∆tS(ti)2σ2 + O(∆t)…..(b)
Assuming that volatility of the underlying stock is constant during time interval t , r is the risk free
rate that the average return from the stock should be equal to the risk-free rate which can be
written as:
1 − pu − pd + puu + pdd = er∆t.
In order to look at upward and downward jump requires an extra constraint that size of upward
jump is a reciprocal of a downward jump i-e ud = 1…………………(c)
The value of the underlying stock can be find out by using the given knowledge of upward and
downward jump sizes u and d with transition probabilities pu and pd. If Nu, Nd and Nm are the
numbers of upward, downward and middle jumps respectively then the value of underlying stock at
node j and for time interval i is given as
Si,j = uNu dNd S(t0), where Nu + Nd + Nm = n
Three constraints (a), (b) and (c) have been imposed on u, d, pu and pm which results as a family of
trinomial tree models , jump sizes of the popular representative of that family are
14 | P a g e
Risk Management and Financial Engineering (BAFI2081)
And its changing probabilities are:
Stock index So will move up by Su or down by Sd or will remain same as So. Pu is the probability of
upward moment and pd is the probability that stock price will move upward, hence the probability
that the stock remains same will be given by (1-pu-pd).
5 – Comparison between Option Pricing Models & their calculation methodologies: All of above models have an edge on one another according to different circumstances. All of these
models use same inputs; stock price, strike price, time to maturity, risk-free rate and volatility. Black-
Scholes is a continuous time or closed form model while there are discrete steps in binomial and
trinomial models. In binomial model we compute future value of an option at time t by taking into
account the time value of money and then discounted it back to get the present value of the option.
Black-Scholes model is elegant and analytical. It includes the minimum value of the stock price So
minus the discounted strike price ‘ ’ and has added probability functions called standard
normal cumulative distribution functions. Binomial and trinomial models build a map of the future
stock prices with a number of steps and that number can go up to infinity. This means we are
converging to a Black-Scholes model, demonstrating that Black-Scholes model is a special case of
binomial and trinomial models where number of steps can be infinite.
Black-Scholes model is used in a wide range for option pricing, especially for European options.
Binomial model can price an American style options more accurately as it also considers the
possibility of early exercise of an option as it provides an insight of decision at different time
intervals before the expiration date that either an option should be exercised or should it be held for
a longer period. On the other hand, Black-Scholes model only considers the possibility of exercising
an option at the expiration date. However it can be implemented on American style options by
considering shorter times for expiration. It is easy to implement a binomial method in a spread
sheet for pricing options giving prices at every step. Black-Scholes model is much more convenient
15 | P a g e
Risk Management and Financial Engineering (BAFI2081)
for calculating a large number of option prices very quickly. Binomial model can be much more
complex than the Black-Scholes model in doing so. Trinomial model presents a more realistic view of
the behaviour of financial instruments. Trinomial models can sometimes become inconvenient and
inefficient but it crucially considers the third option of possibility of stock value - remaining
stationary at each step.
6 – Theoretical modifications in Binomial, Black-Scholes and Trinomial: After calculating theoretical warrant prices using Black-Scholes (BS) Model, Binomial Model and
Trinomial Model, it is apparent that none of the model is particularly close to the actual warrant
prices. Improved versions for all the models are necessary. This section will comment on current
problems of the 3 models that may affect their accuracies in this report, and propose ways to
improve them. Unless otherwise referenced, contents in this section are mainly derived from
Lauterbach and Schultz’s (1990) journal article titled “Pricing Warrants: An Empirical Study of the
Black-Scholes Model and Its Alternatives”.
Firstly, due to the normally long life of a warrant, the variance rate of return on stock and the risk-
free interest rate may be expected to change significantly during its life. To improve BS, Binomial and
Trinomial models, stochastic interest rates and stochastic variance rate of return should be used.
Both factors should not be assumed as constant and their inputs into the model have to be changed
on a daily basis. In particular, as the variance rate of return of a warrant’s underlying stock often
fluctuates and affects warrant prices greatly, the annualized variance rate of return has to be
modelled in a way that will more accurately reflect the market’s expected variance of the underlying
stock. Moreover, dividend should not be assumed as constant, as no/improper dividend adjustments
may lead to inaccurate theoretical model prices. This is because while constant dividends are used to
calculate model prices, it is logical to believe that markets would expect dividends to increase
(decrease) as underlying stock price increase (decrease). In other words, markets will expect the
dividend payout ratio to be the same. To address this problem, improved version of the models can
adjust dividend expectations in a daily basis. By simply adjusting daily expected dividends by a
percentage equivalent to the daily change in the underlying stock prices, the models’ theoretical
prices will provide a closer indication to actual warrant prices.
Secondly, high volatilities tend to result in extremely high theoretical warrant prices. This meant that
if an individual stock is having a fluctuating year (for example, SGX’s Wilmar from 1 June 2016 to 31
June 2016 as used in this report) theoretical prices tend to be overstated regardless of which model
was used. In order to address such issue, an improved model should utilize equity volatilities of a
16 | P a g e
Risk Management and Financial Engineering (BAFI2081)
basket of stocks that are similar to the underlying stocks (instead of only the warrant’s underlying
stock volatility) to calculate the annualized standard deviation for the models’ inputs.
Thirdly, it has been observed that warrant prices are less sensitive to underlying equity values than
the model predicts. As warrants usually trade less frequently than the underlying stock, implied
standard deviation (ISD) in a warrant tend to biased. ISD tend to be downward (upward) biased
when price increases (decreases), which means that actual warrant prices will not increase or
decrease by as much as it should be. In an improved model, sing lagged standard deviation/volatility
will minimize such bias. By taking yesterday’s stochastic annualized volatility to predict today’s
prices, any upward or downward pressure in ISD, and thus theoretical warrant prices, can be
neutralized. In addition, and more importantly, the improved model should allow an inverse relation
between equity volatility and equity value. This means that when equity value increases (decreases),
the model should decrease (increase) equity volatility, which will result in lower (higher) theoretical
prices when warrants are supposed to increase (decrease) in value. This mean movement in
theoretical prices will be less sensitive and more similar to movement in actual warrant prices. The
improved model can, for example, integrate Cox’s Constant Elasticity of Variance (CEV) or Square
Root CEV formula to calculate the standard deviation for the model.
7 – Statistical evaluation of the models:The empirical testing of the Binomial, Black-Scholes (BS) and Trinomial was done through running a
simple regression function for the models against the actual warrant prices of the covered warrants.
The structure of the regression was:
Binomial/Black-Scholes/Trinomial Model Price = Intercept + β actual warrant prices
Ideally the regression should produce the coefficient of the warrant prices (β) near to unit value and
the intercept close to zero. The basic framework included testing these regressions as to how
accurately they predict the warrant prices. The foremost statistical measures that were used to
evaluate the models included R-squared measure, Durbin Watson test along with Akaike Info,
Schwartz and Hannan-Quinn criterion. On the basis of the outputs provided by Eviews, further
statistical tests were performed to check for autocorrelation, heteroskedacity and mutlicollinearity.
To check for these effects, Correlogram, Serial correlation LM test, Heteroskedacity White test,
Heteroskedacity ARCH test were performed on the simple regression output of the warrant price
data and predicted values of warrants by the selected mathematical models.
Since coming up with a model that incorporates the complete dynamics of the covered warrants was
out of the scope of the current level of study and of this report, statistical inference approach was
17 | P a g e
Risk Management and Financial Engineering (BAFI2081)
selected. Once the results for these tests were obtained, an Autoregressive Moving Averages
(ARMA) structure was used to make the residuals more structured and make the model statistically
fit. Hence the modified Binomial, Black-Scholes and Trinomial models all contained some ARMA (p,q)
structure with improved aforementioned statistical measures. To further validate and compare the
performance of the modified models against that of the raw warrant prices data provided by the
actual models, Root Mean Squared Errors (RMSE) were compared. The RMSE values for raw data
were compared against the statistical forecasts and dynamic forecasts of the modified models.
Appendix 1-3 summarise the results for CapitaLand’s Binomial, Black-Scholes and Trinomial models,
Appendix 4-6 contain the same framework for SembCorp Marine and Appendix 7-9 illustrate the
results for Wilmar International. All Appendices summarise the statistical tests and outputs with the
following subpart classifications format:
a) Simple regression function output with normality tests for residuals
b) Correlogram of the simple regression output
c) Serial Correlation LM Test (Breusch Godfrey)
d) Heteroskedacity – White Test
e) Heteroskedacity – ARCH Test
f) Modified model with ARMA structure
g) Variance Inflation (VIF) test for mluticollinearity
h) Correlogram for the modified model
i) Statistical forecasts for modified models
j) Dynamic forecasts for the modified models
18 | P a g e
Risk Management and Financial Engineering (BAFI2081)
7.1 – Data collection and Assumptions: In this report, three existing covered warrants for Singaporean companies are considered. The
companies that were selected are:
CapitaLand (Industry: Real Estate)
SembCorp Marine (Industry: Offshore and Marine)
Wilmar International (Industry: Food Processing)
The selection of these companies was random and the main reason was based on the availability of
data. Data for actual warrant prices were collected from Macquarie Bank’s website. The warrants
are European in nature and have the following characteristics:
CapitaLand Covered Warrant – Main features:
Stricke Price 3.2Expiry date 12-Dec-16Risk free rate 1.86%Total days in years 261
Div. Yield 2.82%
SembCorp marine Covered Warrant – Main features:
Stricke Price 1.95Expiry date 12-Dec-16Risk free rate 1.86%Total days in years 261
Div. Yield 2.69%
Wilmar International Covered Warrant – Main features:
Stricke Price 3.3Expiry date 01-Dec-16Risk free rate 1.86%Total days in years 261
Div. Yield 2.48%
19 | P a g e
Risk Management and Financial Engineering (BAFI2081)
7.1.1 – Assumptions: The assumptions used in this study are the following:
The study is based on the historical data taken from the working days of 1-Jan-2016 to 30-
Jun-2016.
20 nodes are selected for the calculation of the warrant prices in Binomial and Trinomial
models
The risk free rate for the calculations was 1.86% (annualized), which was the yield on the 10
year Singaporean government bond at the time of checking
Time to maturity in years was calculated for each daily calculation. The official number of
working days in 2016 used is according Singapore’s official calendar.
Dividend yields for the warrant price calculation in the Trinomial model were taken from
Bloomberg and Yahoo Finance. These were checked against those listed on Thomson
Reuters and CNBC which were found to be identical
Volatility was the stock’s standard deviation in the selected time period. The volatility was
then annualized for the calculation of theoretical warrant prices
7.2 – Statistical Tests and findings:As shown in Appendices 1-9 (subpart a), the simple regression function output for the models for the
selected sample of companies show a very different result than what was ideally expected. The
values for the intercepts, as shown on Table 1 on the following page, and coefficients were larger
than unit value but remains statistically significant i.e. high f-stat values with less than 0.05 p-values.
The most notable difference was the coefficients for the warrant price variable which pointed out
that Trinomial model are better than the Binomial and Black-Scholes. The values for the actual
warrant price coefficients were as a whole smaller than those of the actual warrant price coefficients
for Binomial and Black-Scholes model. This depicts that less numerical adjustment is needed for the
values given by the Trinomial model to match with that of the actual warrant prices i.e. Trinomial
model prices are closer to the unit value coefficient.
The normality tests for the residuals of simple regression outputs for all three models for CapitaLand
and SembCorp Marine were non-normal as the Jarque-Bera values were high and p-value was less
than 0.05 (resulting in rejection of null of normality as shown in Table 1 on following page). The only
exception was the residuals of the simple regression outputs for the models for Wilmar
International’s warrants. They were all significantly normal following a very close normal distribution
and low JB stat with non-significant p value as shown by Appendix 7a, 8a and 9a. The regression
20 | P a g e
Risk Management and Financial Engineering (BAFI2081)
models also had a very low Durbin-Watson stat with the highest value of DW of 0.43 signalling a
severe serial correlation.
The correlogram tests for the simple regression functions for the models, as depicted in Appendices
1-9 (subpart b), showed that these regressions had a very strong AR 1 process with the highest AC
and PAC values at the first lag. On further testing through Breusch-Godfrey LM test with 5 lags, it
became evident that the regressions outputs have strong presence of serial correlation up to lag 2
on average, as the test results were significant and had F-stat. prob < 0.05 (Appendices 1-9 subparts
c). To test for Heteroskedacity, White test was run for all the models for simple regression which
showed a strong statistically significant presence of heteroskedacity for CapitaLand and SembCorp
Marine. In case of simple regression models for Wilmar International, although White test provided
statistically significant results, it was less severe than that of the models for other companies in the
sample (appendices 8 and 9 subparts c).
The strong presence of heteroskedacity in the residuals indicated a presence of Autoregressive
Conditional Heteroskedacity (ARCH) in the regressions. To confirm that, a heteroskedacity (ARCH)
test was run (appendices 1-9 subparts d). The tests proved significant presence of ARCH effects as all
the statistical tests resulted in the rejection of null of ARCH. Due to the limitation of the functionality
of Eviews, exact ARCH tests couldn’t be performed to further evaluate whether which level of ARCH
test was necessary to get all the ARCH effects removed or whether a GARCH would be a better fit. 3
In addition, an observation of the raw data (i.e. comparison of the actual warrant prices and prices
predicted by the selected models for the sample companies) showed a trend towards more accurate
prediction whenever the stock price moved closer to exercise price. The pattern was present in all
three models for all the companies, and was strongest for the trinomial model. This effect could be
the outcome of the in the money nature of the option prices which the models calculate.
3 Eviews at RMIT crashed every time whenever an ARCH/GARCH test was run even using the MyDesktop app.
21 | P a g e
Risk Management and Financial Engineering (BAFI2081)
CapitaLand SembCorp Marine Wilmar InternationalBinomial
ModelBlack-Scholes
ModelTrinomial
ModelBinomial
ModelBlack-Scholes
ModelTrinomial
ModelBinomial
ModelBlack-Scholes
ModelTrinomial
ModelIntercept -0.146 -0.146 -0.113 -0.043 -0.043 -0.034 0.333 0.334 0.308Coefficient (β) of Warrant Variable 3.625 3.612 2.790 1.230 1.226 0.951 0.505 0.498 0.479
R-squared 0.720 0.720 0.676 0.835 0.836 0.819 0.096 0.094 0.095Durbin Watson 0.190 0.194 0.189 0.430 0.424 0.438 0.234 0.236 0.234Akaike info. Criterion -4.603 -4.611 -4.919 -7.474 -7.493 -7.880 -1.696 -1.677 -1.757Schwarz Criterion -4.557 -4.566 -4.873 -7.429 -7.447 -7.835 -1.651 -1.630 -1.711Hannan-Quinn Criterion -4.584 -4.592 -4.900 -7.456 -7.474 -7.862 -1.678 -1.658 -1.738
F-stat of regression 316.557 316.824 257.133 620.338 627.319 556.348 12.999 12.387 12.421Prob. Of F-stat of regression 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.001
Jarque-Bera Stats 49.427 47.109 31.536 64.546 61.921 93.909 1.975 1.333 1.222Jarque-Bera Stats (prob) 0.000 0.000 0.000 0.000 0.000 0.000 0.373 0.514 0.543
Table 1 – Statistical Measures Table for the Simple Regression output results for Binomial, Black-Scholes and Trinomial Models vs. Actual Warrant Prices
22 | P a g e
Risk Management and Financial Engineering (BAFI2081)
CapitaLand SembCorp Marine Wilmar InternationalBinomial
ModelBlack-Scholes
ModelTrinomial
ModelBinomial
ModelBlack-Scholes
ModelTrinomial
ModelBinomial
ModelBlack-Scholes
ModelTrinomial
ModelIntercept -0.212 -0.212 -1.795 -0.053 -0.053 -0.042 0.380 0.337 0.311Coefficient (β) of Warrant Variable 4.507 4.508 3.706 1.412 1.423 1.119 0.022 0.022 0.023
R-squared 0.957 95.701 0.952 0.953 0.956 0.950 0.924 0.946 0.946Durbin Watson 1.938 1.919 1.922 1.781 1.902 1.920 2.025 1.836 1.844Akaike info. Criterion -6.831 -6.784 -7.096 -8.510 -8.599 -8.950 -4.405 -4.441 -4.542Schwarz Criterion -6.697 -6.649 -6.962 -8.349 -8.465 -8.815 -4.267 -4.348 -4.450Hannan-Quinn Criterion -6.777 -6.729 -7.042 -8.445 -8.545 -8.895 -4.349 -4.403 -4.505
F-stat of regression 526.303 500.851 449.603 361.552 486.811 428.006 261.545 683.728 699.554Prob. Of F-stat of regression 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Table 2 – Statistical Measures Table for the Modified (ARMA – structure) Binomial, Black-Scholes and Trinomial Models
23 | P a g e
Risk Management and Financial Engineering (BAFI2081)
7.3 – Modifications of the Binomial, Black-Scholes and Trinomial models: As discussed in the previous section, the regression outputs provide strong evidences of serial
correlation and heteroskedacity. These conditions lead to the modification of the models as per the
Autoregressive Moving Averages (ARMA) structure. A summary of the modifications that were
performed on the models is as following:
CompaniesModified Binomial
ModelModified Black-Scholes
ModelModified Trinomial
ModelAR Terms MA Terms AR Terms MA Terms AR Terms MA Terms
CapitaLand 1 1 1 1 1 1SembCorp Marine 1 2 (2,3) 1 1(2) 1 1
Wilmar Int. 1 1 1 0 1 0
Table 2 - ARMA Structure of the Modified Models for sample companies
After adjusting the models to the above ARMA structures, some notable differences were observed
in the models. First, the R-squared measure increased whereas the Durbin Watson measure became
much closer to 2. This suggests that although the modifications have taken care of serial correlation,
some effects still remain although is not as severe as before. Like previously, the coefficient for the
warrant price variable in the equation was the least for the Trinomial model here. This is consistent
with the original regression function. The results of the modified regression outputs are shown in
Appendices 1-9 (subsection f). The inverted AR Roots were also less than 1 (unit measure) and
Eviews did not give any warning related to the non-stationary relationship of the added ARMA
terms.
The ARMA terms were statistically significant i.e. having large t-stat values with p-values < 0.05. The
modification also had a collective impact over the Akaike Info, Schwarz and Hannan-Quinn criterion
which became lesser than the ones for simple regression output as shown in Table 1 and Table 2 on
the previous page. To check for the presence of multicollinearity of the added variables and the
coefficients, Variance Inflation Factor (VIF) test was run for all the ARMA structure models which
showed that mutlicollinearity is not a problem with the modified models as shown in Appendices 1-9
(subparts g). Since all the coefficients of the modified models had Cantered VIF score of less than 2,
the modified models were deemed to be better than the original regressions.
The correlograms of the modified models, as shown in the Appendices 1-9 (subparts h), also showed
that the serial correlation problem is not as severe anymore. The AC and PAC values were within the
limits and any significant spike was not present. The concerned spikes in the AC and PAC values were
over 15 AR processes and thus, those were left unchanged to keep the parsimony principle in view.
24 | P a g e
Risk Management and Financial Engineering (BAFI2081)
The final step to evaluate the modified models was to run a static forecast through the end portion
of the sample data points. This was accompanied by more robust dynamic forecasts of the modified
models, then having their Root Mean Squared Errors (RMSE) compared with RMSE of the raw data
that was obtained from the original Binomial, Black-Scholes and Trinomial models. As shown in the
table below the modified models contained the lowest RMSE scores for the models. In addition,
even in the modified models, Trinomial model clearly has the least RMSE scores as compared to
Binomial and Black-Scholes model. Table 3 provides a summary of the RMSEs.
CompaniesRoot Mean Squared Errors (RMSE) of Raw Warrant Price Data with Normal
ModelsBinomial Model Black-Scholes Model Trinomial Model
CapitaLand 0.044 0.044 0.028SembCorp Marine 0.033 0.033 0.036
Wilmar Int. 0.318 0.317 0.291
CompaniesRoot Mean Squared Errors (RMSE) of Static Forecasts by Modified ModelsModified Binomial
ModelModified Black-Scholes
ModelModified Trinomial
ModelCapitaLand 0.044 0.044 0.028SembCorp Marine 0.033 0.033 0.036
Wilmar Int. 0.318 0.317 0.291
CompaniesRoot Mean Squared Errors (RMSE) of Dynamic Forecasts by Modified Models
Modified Binomial Model
Modified Black-Scholes Model
Modified Trinomial Model
CapitaLand 0.015 0.015 0.013SembCorp Marine 0.007 0.007 0.006
Wilmar Int. 0.060 0.058 0.056
Table 3 - Summary of Root Mean Square Errors
25 | P a g e
Risk Management and Financial Engineering (BAFI2081)
8 - ConclusionIn conclusion, the report has looked into the most widely used options and warrants pricing models,
namely the Binomial, Black-Scholes and Trinomial model. These models were empirically tested
using covered warrants for three selected Singaporean companies. The main objective was to
analyze the predictability and accuracy of the models in estimating call warrant prices. Although all
the models proved to be somewhat weak predictors of the covered warrant prices, Trinomial model
seems to be the least erroneous as compared to Binomial and Black-Scholes model. Nevertheless, all
3 models could be further improved.
These models were tested through simple regression models which were further tested for
statistical nature and error patterns. Based on the statistical tests, these models were then modified
to reign in the nature of the errors of which these models have in predicting the warrant values. It
was concluded that even in the modified models Trinomial models stands out as the more reliable
measure on the basis of its least root mean squared error (RMSE) terms.
26 | P a g e
Risk Management and Financial Engineering (BAFI2081)
9 – References
Aitken, M & Segara, R 2005, ‘Impact of warrant introductions on the behaviour of underlying stocks: Australian evidence’, Accounting and Finance, vol. 45, no. 1, pp.127-144
Black, F & Scholes, M 1973, 'The Pricing of Options and Corporate Liabilities', The Journal of
Political Economy, vol. 81, no. 3, pp. 637-654
Black, F., 1975, “Fact and Fantasy in the Use of Options,” Financial Analysts Journal, vol. 31, pp. 36–72.
Bollen, NPB & Whaley, RE 2004, 'Does Net Buying Pressure Affect the Shape of Implied Volatility Functions?', The Journal of Finance, vol. 59, no. 2, pp. 711-53.
Boyle, PP 1986, 'Option Valuation Using Three-Jump Process', International Options Journal, vol.
3, pp. 7-12
Carr, P., and L. Wu, 2009, “Variance Risk Premiums,” Review of Financial Studies, vol. 22, no. 3, pp. 1311– 1341.
Chan, CY, Peretti, CD, Qiao, Z & Wong, WK 2012, ‘Empirical test of the efficiency of the UK covered warrants market: Stochastic dominance and likelihood ratio test approach’, Journal of Empirical Finance, vol. 19, no. 1. pp. 162-174
Cox, J, Ross, S & Rubinstein, M 1979, 'Option Pricing: A Simplified Approach', Journal of Financial
Economics, vol. 7 pp. 229-263
Easley, D., M. O’Hara, and P. S. Srinivas, 1998, “Option Volume and Stock Prices: Evidence on Where Informed Traders Trade,” Journal of Finance, vol. 53, pp. 431–465.
Fan, K., Shen, Y., Siu T.K., Wang, R., 2013, “Pricing foreign equity options with regime-switching, Economic Modelling, vol. 37, pp. 296-297
Fung, H., Zhang, G. and Zhao, L., 2009, “China's Equity Warrants Market: An Overview and Analysis”,The Chinese Economy, vol. 42, no. 1, pp.86-97.
Horst, J & Veld, C 2008, ‘An Empirical Analysis of the Pricing of Bank Issued Options versus Options Exchange Options’, European Financial Management, vol. 14, no. 2, pp.288-314
Hunt, B. and Terry, C., 2011, “Australian Equity Warrants: Are Retail Investors Getting A Fair Go?”,
The Finsia Journal of Applied Finance, vol. 4, pp.48-64.
Kui, L., 2007, “Do Warrants Lead The Underlying Stocks and Index Futures?”, Singapore: Singapore Management University, p.3.
27 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Lauterbach, B & Schultz, P 1990, Pricing Warrants: An Empirical Study of the Black-Scholes Model
and Its Alternatives, The Journal of Finance, vol. 45, no. 4, pp. 1181-1209.
Segara L & Segara, R 2007, ‘Intraday trading patterns in equity warrants and equity options markets: Australian evidence’, The Australasian Accounting Business & Finance Journal, vol. 1, no. 2, p. 52
28 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendices:Appendix 1 (a) – Simple regression output for Binomial model for Capital Land
Dependent Variable: BINOMIAL_MODEL_PRICEMethod: Least SquaresDate: 10/17/16 Time: 22:28Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C -0.146451 0.013494 -10.85270 0.0000WARRANT_PRICE 3.625156 0.203751 17.79205 0.0000
R-squared 0.720173 Mean dependent var 0.090576Adjusted R-squared 0.717898 S.D. dependent var 0.045256S.E. of regression 0.024037 Akaike info criterion -4.602565Sum squared resid 0.071067 Schwarz criterion -4.557312Log likelihood 289.6603 Hannan-Quinn criter. -4.584181F-statistic 316.5572 Durbin-Watson stat 0.190229Prob(F-statistic) 0.000000
29 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 1 (b) – Correlogram for simple regression output for Capital Land
30 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 1 (c) – Serial correlation LM Test for simple regression output for Capital Land
Breusch-Godfrey Serial Correlation LM Test:
F-statistic 63.26895 Prob. F(5,118) 0.0000Obs*R-squared 91.04080 Prob. Chi-Square(5) 0.0000
Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 10/17/16 Time: 23:14Sample: 1/04/2016 6/24/2016Included observations: 125Presample missing value lagged residuals set to zero.
Variable Coefficient Std. Error t-Statistic Prob.
C -0.001515 0.007210 -0.210145 0.8339WARRANT_PRICE 0.025677 0.108916 0.235754 0.8140
RESID(-1) 0.851940 0.092193 9.240853 0.0000RESID(-2) -0.026605 0.119736 -0.222193 0.8245RESID(-3) -0.135686 0.119166 -1.138628 0.2572RESID(-4) 0.192246 0.119924 1.603066 0.1116RESID(-5) 5.78E-06 0.092624 6.23E-05 1.0000
R-squared 0.728326 Mean dependent var 5.73E-17Adjusted R-squared 0.714512 S.D. dependent var 0.023940S.E. of regression 0.012791 Akaike info criterion -5.825719Sum squared resid 0.019307 Schwarz criterion -5.667333Log likelihood 371.1074 Hannan-Quinn criter. -5.761375F-statistic 52.72412 Durbin-Watson stat 1.573603Prob(F-statistic) 0.000000
31 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 1 (d) – Heteroskadicity (White) test for simple regression output for Capital Land
Heteroskedasticity Test: White
F-statistic 14.86129 Prob. F(2,122) 0.0000Obs*R-squared 24.48760 Prob. Chi-Square(2) 0.0000Scaled explained SS 54.92773 Prob. Chi-Square(2) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/17/16 Time: 23:15Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C 0.008402 0.003878 2.166399 0.0322WARRANT_PRICE^2 2.597966 0.928224 2.798856 0.0060
WARRANT_PRICE -0.294102 0.121185 -2.426887 0.0167
R-squared 0.195901 Mean dependent var 0.000569Adjusted R-squared 0.182719 S.D. dependent var 0.001229S.E. of regression 0.001111 Akaike info criterion -10.74378Sum squared resid 0.000151 Schwarz criterion -10.67590Log likelihood 674.4864 Hannan-Quinn criter. -10.71621F-statistic 14.86129 Durbin-Watson stat 0.398478Prob(F-statistic) 0.000002
32 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 1 (e) – Heteroskadacity (ARCH) test for simple regression output for Capital Land
Heteroskedasticity Test: ARCH
F-statistic 17.49309 Prob. F(10,104) 0.0000Obs*R-squared 72.12199 Prob. Chi-Square(10) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/17/16 Time: 23:16Sample (adjusted): 1/18/2016 6/24/2016Included observations: 115 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 5.83E-05 4.74E-05 1.230806 0.2212RESID^2(-1) 0.592820 0.095840 6.185539 0.0000RESID^2(-2) 0.100839 0.110040 0.916384 0.3616RESID^2(-3) 0.242010 0.110015 2.199790 0.0300RESID^2(-4) -0.109843 0.112100 -0.979865 0.3294RESID^2(-5) 0.009521 0.111404 0.085462 0.9321RESID^2(-6) -0.021516 0.110696 -0.194370 0.8463RESID^2(-7) -0.046134 0.100315 -0.459891 0.6466RESID^2(-8) 0.128140 0.078521 1.631915 0.1057RESID^2(-9) -0.169137 0.076309 -2.216474 0.0288
RESID^2(-10) 0.110309 0.051196 2.154653 0.0335
R-squared 0.627148 Mean dependent var 0.000405Adjusted R-squared 0.591297 S.D. dependent var 0.000583S.E. of regression 0.000373 Akaike info criterion -12.85958Sum squared resid 1.45E-05 Schwarz criterion -12.59702Log likelihood 750.4259 Hannan-Quinn criter. -12.75301F-statistic 17.49309 Durbin-Watson stat 2.085835Prob(F-statistic) 0.000000
33 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 1 (f) – Modified model for Capital Land Binomial Model
Dependent Variable: BINOMIAL_MODEL_PRICEMethod: ARMA Maximum Likelihood (OPG - BHHH)Date: 10/18/16 Time: 18:41Sample: 2/01/2016 6/10/2016Included observations: 95Convergence achieved after 31 iterationsCoefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
C -0.212193 0.026576 -7.984247 0.0000WARRANT_PRICE 4.507190 0.259791 17.34927 0.0000
AR(1) 0.984027 0.019044 51.67061 0.0000MA(1) -0.373227 0.086969 -4.291509 0.0000
SIGMASQ 5.53E-05 4.98E-06 11.11552 0.0000
R-squared 0.959024 Mean dependent var 0.086453Adjusted R-squared 0.957203 S.D. dependent var 0.036939S.E. of regression 0.007642 Akaike info criterion -6.830930Sum squared resid 0.005256 Schwarz criterion -6.696516Log likelihood 329.4692 Hannan-Quinn criter. -6.776617F-statistic 526.6033 Durbin-Watson stat 1.938125Prob(F-statistic) 0.000000
Inverted AR Roots .98Inverted MA Roots .37
Appendix 1 (g) – VIF Test for Modified Binomial model for Capital Land
Variance Inflation FactorsDate: 10/18/16 Time: 18:41Sample: 2/01/2016 6/10/2016Included observations: 95
Coefficient Uncentered CenteredVariable Variance VIF VIF
C 0.000706 2.703830 NAWARRANT_PRICE 0.067491 2.540167 1.337280
AR(1) 0.000363 2.166351 2.162275MA(1) 0.007564 1.460435 1.427793
SIGMASQ 2.48E-11 2.372646 2.210507
34 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 1 (h) – Correlogram of Modified Binomial model for Capital Land
35 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 1 (I) – Static forecasts of modified binomial model for Capital Land
Appendix 1 (J) – Dynamic forecasts of modified binomial model for Capital Land
36 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 2 (a) – Simple regression output for Black-Scholes (BS) model for Capital Land
Dependent Variable: BS_MODEL_PRICEMethod: Least SquaresDate: 10/17/16 Time: 22:28Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C -0.146008 0.013439 -10.86453 0.0000WARRANT_PRICE 3.611772 0.202914 17.79955 0.0000
R-squared 0.720343 Mean dependent var 0.090144Adjusted R-squared 0.718069 S.D. dependent var 0.045084S.E. of regression 0.023938 Akaike info criterion -4.610804Sum squared resid 0.070484 Schwarz criterion -4.565551Log likelihood 290.1753 Hannan-Quinn criter. -4.592421F-statistic 316.8239 Durbin-Watson stat 0.193648Prob(F-statistic) 0.000000
37 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 2 (b) – Correlogram of Simple regression output for Black-Scholes (BS) model for Capital Land
38 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 2 (c) – Serial correlation test of simple regression output of Black-Scholes (BS) model for Capital Land
Breusch-Godfrey Serial Correlation LM Test:
F-statistic 61.78381 Prob. F(5,118) 0.0000Obs*R-squared 90.45012 Prob. Chi-Square(5) 0.0000
Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 10/18/16 Time: 18:49Sample: 1/04/2016 6/24/2016Included observations: 125Presample missing value lagged residuals set to zero.
Variable Coefficient Std. Error t-Statistic Prob.
C -0.001483 0.007242 -0.204719 0.8381WARRANT_PRICE 0.025296 0.109393 0.231238 0.8175
RESID(-1) 0.846586 0.092251 9.176994 0.0000RESID(-2) -0.031336 0.119674 -0.261843 0.7939RESID(-3) -0.109651 0.119283 -0.919255 0.3598RESID(-4) 0.176098 0.119785 1.470118 0.1442RESID(-5) -0.002281 0.092624 -0.024621 0.9804
R-squared 0.723601 Mean dependent var 2.86E-17Adjusted R-squared 0.709547 S.D. dependent var 0.023842S.E. of regression 0.012849 Akaike info criterion -5.816714Sum squared resid 0.019482 Schwarz criterion -5.658329Log likelihood 370.5446 Hannan-Quinn criter. -5.752371F-statistic 51.48651 Durbin-Watson stat 1.579401Prob(F-statistic) 0.000000
39 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 2 (d) – Heteroskadacity (White) test of Simple regression output for Black-Scholes (BS) model for Capital Land
Heteroskedasticity Test: White
F-statistic 15.04114 Prob. F(2,122) 0.0000Obs*R-squared 24.72533 Prob. Chi-Square(2) 0.0000Scaled explained SS 54.39302 Prob. Chi-Square(2) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 18:50Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C 0.008208 0.003805 2.157272 0.0329WARRANT_PRICE^2 2.544777 0.910622 2.794547 0.0060
WARRANT_PRICE -0.287637 0.118887 -2.419417 0.0170
R-squared 0.197803 Mean dependent var 0.000564Adjusted R-squared 0.184652 S.D. dependent var 0.001207S.E. of regression 0.001090 Akaike info criterion -10.78207Sum squared resid 0.000145 Schwarz criterion -10.71419Log likelihood 676.8796 Hannan-Quinn criter. -10.75450F-statistic 15.04114 Durbin-Watson stat 0.393388Prob(F-statistic) 0.000001
40 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 2 (e) – Heteroskadacity (ARCH) test of Simple regression output for Black-Scholes (BS) model for Capital Land
Heteroskedasticity Test: ARCH
F-statistic 17.13413 Prob. F(10,104) 0.0000Obs*R-squared 71.56300 Prob. Chi-Square(10) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 18:51Sample (adjusted): 1/18/2016 6/24/2016Included observations: 115 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 6.08E-05 4.77E-05 1.272659 0.2060RESID^2(-1) 0.618070 0.096025 6.436547 0.0000RESID^2(-2) 0.068679 0.111514 0.615879 0.5393RESID^2(-3) 0.254685 0.111404 2.286148 0.0243RESID^2(-4) -0.139011 0.113850 -1.221001 0.2248RESID^2(-5) 0.031538 0.113222 0.278555 0.7811RESID^2(-6) -0.046684 0.112998 -0.413135 0.6804RESID^2(-7) -0.006233 0.101233 -0.061571 0.9510RESID^2(-8) 0.107083 0.080283 1.333813 0.1852RESID^2(-9) -0.163184 0.078214 -2.086381 0.0394
RESID^2(-10) 0.107421 0.052533 2.044844 0.0434
R-squared 0.622287 Mean dependent var 0.000403Adjusted R-squared 0.585968 S.D. dependent var 0.000582S.E. of regression 0.000375 Akaike info criterion -12.85022Sum squared resid 1.46E-05 Schwarz criterion -12.58766Log likelihood 749.8878 Hannan-Quinn criter. -12.74365F-statistic 17.13413 Durbin-Watson stat 2.076010Prob(F-statistic) 0.000000
41 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 2 (f) – Modified model for Black-Scholes (BS) model for Capital Land Modified BS Model
Dependent Variable: BS_MODEL_PRICEMethod: ARMA Maximum Likelihood (OPG - BHHH)Date: 10/18/16 Time: 18:55Sample: 2/01/2016 6/10/2016Included observations: 95Convergence achieved after 35 iterationsCoefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
C -0.212298 0.027382 -7.753102 0.0000WARRANT_PRICE 4.507508 0.279429 16.13112 0.0000
AR(1) 0.982128 0.019722 49.79945 0.0000MA(1) -0.364819 0.082626 -4.415319 0.0000
SIGMASQ 5.81E-05 5.21E-06 11.14643 0.0000
R-squared 0.957008 Mean dependent var 0.086105Adjusted R-squared 0.955097 S.D. dependent var 0.036945S.E. of regression 0.007829 Akaike info criterion -6.783510Sum squared resid 0.005516 Schwarz criterion -6.649096Log likelihood 327.2167 Hannan-Quinn criter. -6.729197F-statistic 500.8514 Durbin-Watson stat 1.918621Prob(F-statistic) 0.000000
Inverted AR Roots .98Inverted MA Roots .36
Appendix 2 (g) – VIF test of modified Black-Scholes (BS) model for Capital Land
Variance Inflation FactorsDate: 10/18/16 Time: 18:55Sample: 2/01/2016 6/10/2016Included observations: 95
Coefficient Uncentered CenteredVariable Variance VIF VIF
C 0.000750 3.250658 NAWARRANT_PRICE 0.078081 2.992788 1.372972
AR(1) 0.000389 2.219858 2.217503MA(1) 0.006827 1.422085 1.393867
SIGMASQ 2.71E-11 2.475892 2.294333
42 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 2 (h) – Correlogram of modified Black-Scholes (BS) model for Capital Land
43 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 2 (I) – Static forecasts of modified Black-Scholes (BS) model for Capital Land
Appendix 2 (J) – Dynamic forecasts of modified Black-Scholes (BS) model for Capital Land
44 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 3 (a) – Simple regression output for Trinomial model for Capital Land
Dependent Variable: TRINOMIAL_MODEL_PRICEMethod: Least SquaresDate: 10/17/16 Time: 22:29Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C -0.113260 0.011522 -9.830179 0.0000WARRANT_PRICE 2.789592 0.173965 16.03538 0.0000
R-squared 0.676429 Mean dependent var 0.069134Adjusted R-squared 0.673799 S.D. dependent var 0.035934S.E. of regression 0.020523 Akaike info criterion -4.918660Sum squared resid 0.051807 Schwarz criterion -4.873407Log likelihood 309.4162 Hannan-Quinn criter. -4.900276F-statistic 257.1333 Durbin-Watson stat 0.189091Prob(F-statistic) 0.000000
Appendix 3 (b) – Correlogram of Trinomial model for Capital Land
45 | P a g e
Risk Management and Financial Engineering (BAFI2081)
46 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 3 (c) – Serial correlation test for Trinomial model for Capital Land
Breusch-Godfrey Serial Correlation LM Test:
F-statistic 65.35230 Prob. F(5,118) 0.0000Obs*R-squared 91.83616 Prob. Chi-Square(5) 0.0000
Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 10/18/16 Time: 19:03Sample: 1/04/2016 6/24/2016Included observations: 125Presample missing value lagged residuals set to zero.
Variable Coefficient Std. Error t-Statistic Prob.
C -0.001558 0.006090 -0.255813 0.7985WARRANT_PRICE 0.026147 0.092010 0.284172 0.7768
RESID(-1) 0.866331 0.092331 9.382921 0.0000RESID(-2) -0.056403 0.121080 -0.465836 0.6422RESID(-3) -0.088670 0.120933 -0.733218 0.4649RESID(-4) 0.163466 0.121217 1.348543 0.1801RESID(-5) -0.000839 0.092735 -0.009050 0.9928
R-squared 0.734689 Mean dependent var 4.30E-17Adjusted R-squared 0.721199 S.D. dependent var 0.020440S.E. of regression 0.010793 Akaike info criterion -6.165513Sum squared resid 0.013745 Schwarz criterion -6.007128Log likelihood 392.3446 Hannan-Quinn criter. -6.101170F-statistic 54.46025 Durbin-Watson stat 1.595759Prob(F-statistic) 0.000000
47 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 3 (d) – Heteroskadacity (White) test of Trinomial model for Capital Land
Heteroskedasticity Test: White
F-statistic 14.61061 Prob. F(2,122) 0.0000Obs*R-squared 24.15436 Prob. Chi-Square(2) 0.0000Scaled explained SS 47.37663 Prob. Chi-Square(2) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 19:03Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C 0.004847 0.002648 1.830309 0.0696WARRANT_PRICE^2 1.565007 0.633804 2.469229 0.0149
WARRANT_PRICE -0.172787 0.082747 -2.088140 0.0389
R-squared 0.193235 Mean dependent var 0.000414Adjusted R-squared 0.180009 S.D. dependent var 0.000838S.E. of regression 0.000758 Akaike info criterion -11.50685Sum squared resid 7.02E-05 Schwarz criterion -11.43897Log likelihood 722.1781 Hannan-Quinn criter. -11.47927F-statistic 14.61061 Durbin-Watson stat 0.406609Prob(F-statistic) 0.000002
48 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 3 (e) – Heteroskadacity (ARCH) test of Trinomial model for Capital Land
Heteroskedasticity Test: ARCH
F-statistic 16.81761 Prob. F(10,104) 0.0000Obs*R-squared 71.05786 Prob. Chi-Square(10) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 19:04Sample (adjusted): 1/18/2016 6/24/2016Included observations: 115 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 5.02E-05 3.71E-05 1.355171 0.1783RESID^2(-1) 0.622102 0.096796 6.426907 0.0000RESID^2(-2) 0.073477 0.112808 0.651339 0.5163RESID^2(-3) 0.249660 0.112679 2.215679 0.0289RESID^2(-4) -0.144028 0.115045 -1.251929 0.2134RESID^2(-5) 0.043006 0.114808 0.374590 0.7087RESID^2(-6) -0.058190 0.114678 -0.507418 0.6129RESID^2(-7) -0.012578 0.103354 -0.121699 0.9034RESID^2(-8) 0.099355 0.085893 1.156724 0.2500RESID^2(-9) -0.146181 0.083482 -1.751042 0.0829
RESID^2(-10) 0.096549 0.056882 1.697374 0.0926
R-squared 0.617894 Mean dependent var 0.000310Adjusted R-squared 0.581153 S.D. dependent var 0.000446S.E. of regression 0.000289 Akaike info criterion -13.37034Sum squared resid 8.68E-06 Schwarz criterion -13.10778Log likelihood 779.7943 Hannan-Quinn criter. -13.26376F-statistic 16.81761 Durbin-Watson stat 2.061901Prob(F-statistic) 0.000000
49 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 3 (f) – Modified Trinomial model for Capital Land
Dependent Variable: TRINOMIAL_MODEL_PRICEMethod: ARMA Maximum Likelihood (OPG - BHHH)Date: 10/18/16 Time: 19:13Sample: 2/01/2016 6/10/2016Included observations: 95Convergence achieved after 30 iterationsCoefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
C -0.179483 0.023913 -7.505584 0.0000WARRANT_PRICE 3.706089 0.229780 16.12888 0.0000
AR(1) 0.983461 0.020076 48.98651 0.0000MA(1) -0.345738 0.090491 -3.820688 0.0002
SIGMASQ 4.24E-05 3.78E-06 11.23754 0.0000
R-squared 0.952341 Mean dependent var 0.066327Adjusted R-squared 0.950223 S.D. dependent var 0.029996S.E. of regression 0.006692 Akaike info criterion -7.095962Sum squared resid 0.004031 Schwarz criterion -6.961547Log likelihood 342.0582 Hannan-Quinn criter. -7.041648F-statistic 449.6032 Durbin-Watson stat 1.922426Prob(F-statistic) 0.000000
Inverted AR Roots .98Inverted MA Roots .35
Appendix 3 (g) – VIF test for modified Trinomial model for Capital Land
Variance Inflation FactorsDate: 10/18/16 Time: 19:14Sample: 2/01/2016 6/10/2016Included observations: 95
Coefficient Uncentered CenteredVariable Variance VIF VIF
C 0.000572 2.856546 NAWARRANT_PRICE 0.052799 2.582642 1.336992
AR(1) 0.000403 2.262807 2.262636MA(1) 0.008189 1.371007 1.347798
SIGMASQ 1.43E-11 2.460336 2.288745
50 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 3 (H) – Correlogram of modified Trinomial model for Capital Land
51 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 3 (I) – Static forecasts for modified Trinomial model for Capital Land
Appendix 3 (J) – Simple regression output for Trinomial model for Capital Land
52 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 4 (a) – Simple regression output of Binomial model for Sembcorp Marine
Dependent Variable: BINOMIAL_MODEL_PRICEMethod: Least SquaresDate: 10/17/16 Time: 22:36Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C -0.043047 0.002244 -19.18147 0.0000WARRANT_PRICE 1.230189 0.049392 24.90658 0.0000
R-squared 0.834530 Mean dependent var 0.011377Adjusted R-squared 0.833185 S.D. dependent var 0.014001S.E. of regression 0.005719 Akaike info criterion -7.474323Sum squared resid 0.004022 Schwarz criterion -7.429070Log likelihood 469.1452 Hannan-Quinn criter. -7.455939F-statistic 620.3375 Durbin-Watson stat 0.430419Prob(F-statistic) 0.000000
53 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 4 (b) – Correlogram for simple regression output of Binomial model for Sembcorp Marine
54 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 4 (c) – Serial Correlation test for Binomial model for Sembcorp Marine
Breusch-Godfrey Serial Correlation LM Test:
F-statistic 41.14236 Prob. F(5,118) 0.0000Obs*R-squared 79.43478 Prob. Chi-Square(5) 0.0000
Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 10/18/16 Time: 19:53Sample: 1/04/2016 6/24/2016Included observations: 125Presample missing value lagged residuals set to zero.
Variable Coefficient Std. Error t-Statistic Prob.
C -3.51E-05 0.001390 -0.025270 0.9799WARRANT_PRICE 0.001419 0.030580 0.046411 0.9631
RESID(-1) 1.019348 0.092035 11.07569 0.0000RESID(-2) -0.391474 0.131412 -2.978972 0.0035RESID(-3) 0.059469 0.136399 0.435989 0.6636RESID(-4) 0.061421 0.132055 0.465116 0.6427RESID(-5) 0.024109 0.092820 0.259742 0.7955
R-squared 0.635478 Mean dependent var -2.24E-17Adjusted R-squared 0.616943 S.D. dependent var 0.005695S.E. of regression 0.003525 Akaike info criterion -8.403492Sum squared resid 0.001466 Schwarz criterion -8.245106Log likelihood 532.2182 Hannan-Quinn criter. -8.339148F-statistic 34.28530 Durbin-Watson stat 1.911429Prob(F-statistic) 0.000000
55 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 4 (d) – Heteroskadacity (White) test for Binomial model for Sembcorp Marine
Heteroskedasticity Test: White
F-statistic 58.73830 Prob. F(2,122) 0.0000Obs*R-squared 61.31945 Prob. Chi-Square(2) 0.0000Scaled explained SS 126.4006 Prob. Chi-Square(2) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 19:54Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C 0.000685 7.63E-05 8.979479 0.0000WARRANT_PRICE^2 0.346644 0.034959 9.915622 0.0000
WARRANT_PRICE -0.030938 0.003329 -9.293724 0.0000
R-squared 0.490556 Mean dependent var 3.22E-05Adjusted R-squared 0.482204 S.D. dependent var 6.67E-05S.E. of regression 4.80E-05 Akaike info criterion -17.02819Sum squared resid 2.81E-07 Schwarz criterion -16.96031Log likelihood 1067.262 Hannan-Quinn criter. -17.00062F-statistic 58.73830 Durbin-Watson stat 1.146321Prob(F-statistic) 0.000000
56 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 4 (e) – Heteroskadacity (ARCH) test for Binomial model for Sembcorp Marine
Heteroskedasticity Test: ARCH
F-statistic 8.586962 Prob. F(10,104) 0.0000Obs*R-squared 52.00941 Prob. Chi-Square(10) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 19:54Sample (adjusted): 1/18/2016 6/24/2016Included observations: 115 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 1.13E-05 6.40E-06 1.768791 0.0799RESID^2(-1) 0.816257 0.098012 8.328171 0.0000RESID^2(-2) -0.383675 0.125771 -3.050577 0.0029RESID^2(-3) 0.124272 0.131298 0.946491 0.3461RESID^2(-4) 0.039004 0.131976 0.295541 0.7682RESID^2(-5) 0.066848 0.132098 0.506048 0.6139RESID^2(-6) -0.018846 0.132143 -0.142619 0.8869RESID^2(-7) -0.054702 0.132065 -0.414206 0.6796RESID^2(-8) 0.022830 0.131570 0.173523 0.8626RESID^2(-9) -0.002918 0.125421 -0.023263 0.9815
RESID^2(-10) -0.014661 0.095331 -0.153787 0.8781
R-squared 0.452256 Mean dependent var 2.88E-05Adjusted R-squared 0.399588 S.D. dependent var 6.66E-05S.E. of regression 5.16E-05 Akaike info criterion -16.81563Sum squared resid 2.77E-07 Schwarz criterion -16.55307Log likelihood 977.8987 Hannan-Quinn criter. -16.70906F-statistic 8.586962 Durbin-Watson stat 1.840431Prob(F-statistic) 0.000000
57 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 4 (f) – Modified Binomial model for Sembcorp Marine
Dependent Variable: BINOMIAL_MODEL_PRICEMethod: ARMA Maximum Likelihood (OPG - BHHH)Date: 10/18/16 Time: 20:00Sample: 2/01/2016 6/10/2016Included observations: 95Convergence achieved after 43 iterationsCoefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
C -0.052906 0.002917 -18.13432 0.0000WARRANT_PRICE 1.412092 0.052747 26.77128 0.0000
AR(1) 0.913174 0.079114 11.54250 0.0000MA(2) -0.391104 0.137418 -2.846098 0.0055MA(3) -0.352013 0.099698 -3.530798 0.0007
SIGMASQ 1.03E-05 1.78E-06 5.793034 0.0000
R-squared 0.953078 Mean dependent var 0.013051Adjusted R-squared 0.950442 S.D. dependent var 0.014890S.E. of regression 0.003315 Akaike info criterion -8.510135Sum squared resid 0.000978 Schwarz criterion -8.348838Log likelihood 410.2314 Hannan-Quinn criter. -8.444959F-statistic 361.5521 Durbin-Watson stat 1.781168Prob(F-statistic) 0.000000
Inverted AR Roots .91Inverted MA Roots .89 -.44+.45i -.44-.45i
Appendix 4 (g) – VIF Test for modified Binomial model for Sembcorp Marine
Variance Inflation FactorsDate: 10/18/16 Time: 20:01Sample: 2/01/2016 6/10/2016Included observations: 95
Coefficient Uncentered CenteredVariable Variance VIF VIF
C 8.51E-06 9.343338 NAWARRANT_PRICE 0.002782 10.35827 1.381015
AR(1) 0.006259 5.119722 4.705390MA(2) 0.018884 3.551382 3.422117MA(3) 0.009940 1.822758 1.779146
SIGMASQ 3.16E-12 1.529709 1.463460
58 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 4 (h) – Correlogram for modified Binomial model for Sembcorp Marine
59 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 4 (i) – Static forecasts for modified model for Sembcorp Marine
Appendix 4 (j) – Dynamic forecasts for modified Binomial model for Sembcorp Marine
60 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 5 (a) – Simple regression output of Black-Scholes model for Sembcorp Marine
Dependent Variable: BS_MODEL_PRICEMethod: Least SquaresDate: 10/17/16 Time: 22:37Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C -0.042696 0.002224 -19.19906 0.0000WARRANT_PRICE 1.225895 0.048945 25.04633 0.0000
R-squared 0.836070 Mean dependent var 0.011538Adjusted R-squared 0.834737 S.D. dependent var 0.013940S.E. of regression 0.005667 Akaike info criterion -7.492506Sum squared resid 0.003950 Schwarz criterion -7.447253Log likelihood 470.2817 Hannan-Quinn criter. -7.474123F-statistic 627.3188 Durbin-Watson stat 0.423673Prob(F-statistic) 0.000000
61 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 5 (b) – Correlogram of simple regression output of Black-Scholes model for Sembcorp Marine
62 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 5 (c) – Serial corerrlation LM test of Black-Scholes model for Sembcorp Marine
Breusch-Godfrey Serial Correlation LM Test:
F-statistic 42.28204 Prob. F(5,118) 0.0000Obs*R-squared 80.22300 Prob. Chi-Square(5) 0.0000
Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 10/18/16 Time: 20:09Sample: 1/04/2016 6/24/2016Included observations: 125Presample missing value lagged residuals set to zero.
Variable Coefficient Std. Error t-Statistic Prob.
C 2.02E-06 0.001365 0.001476 0.9988WARRANT_PRICE 0.000579 0.030041 0.019289 0.9846
RESID(-1) 1.034105 0.091894 11.25323 0.0000RESID(-2) -0.426402 0.132407 -3.220394 0.0017RESID(-3) 0.121896 0.137870 0.884135 0.3784RESID(-4) -0.011467 0.133039 -0.086195 0.9315RESID(-5) 0.060449 0.092669 0.652310 0.5155
R-squared 0.641784 Mean dependent var -2.51E-17Adjusted R-squared 0.623570 S.D. dependent var 0.005644S.E. of regression 0.003463 Akaike info criterion -8.439126Sum squared resid 0.001415 Schwarz criterion -8.280740Log likelihood 534.4454 Hannan-Quinn criter. -8.374782F-statistic 35.23504 Durbin-Watson stat 1.903320Prob(F-statistic) 0.000000
63 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 5 (d) – Heteroskadacity (White) test for Black-Scholes model for Sembcorp Marine
Heteroskedasticity Test: White
F-statistic 58.09654 Prob. F(2,122) 0.0000Obs*R-squared 60.97631 Prob. Chi-Square(2) 0.0000Scaled explained SS 123.4637 Prob. Chi-Square(2) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 20:10Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C 0.000666 7.45E-05 8.937991 0.0000WARRANT_PRICE^2 0.336481 0.034115 9.863117 0.0000
WARRANT_PRICE -0.030033 0.003248 -9.245123 0.0000
R-squared 0.487810 Mean dependent var 3.16E-05Adjusted R-squared 0.479414 S.D. dependent var 6.49E-05S.E. of regression 4.68E-05 Akaike info criterion -17.07708Sum squared resid 2.67E-07 Schwarz criterion -17.00920Log likelihood 1070.318 Hannan-Quinn criter. -17.04951F-statistic 58.09654 Durbin-Watson stat 1.087966Prob(F-statistic) 0.000000
64 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 5 (e) – Heteroskadacity (ARCH) test of Black-Scholes model for Sembcorp Marine
Heteroskedasticity Test: ARCH
F-statistic 9.692826 Prob. F(10,104) 0.0000Obs*R-squared 55.47627 Prob. Chi-Square(10) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 20:10Sample (adjusted): 1/18/2016 6/24/2016Included observations: 115 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 1.08E-05 6.03E-06 1.796662 0.0753RESID^2(-1) 0.849764 0.097959 8.674655 0.0000RESID^2(-2) -0.369647 0.127723 -2.894136 0.0046RESID^2(-3) 0.053860 0.132760 0.405692 0.6858RESID^2(-4) 0.113790 0.132815 0.856754 0.3936RESID^2(-5) 0.000110 0.133325 0.000822 0.9993RESID^2(-6) 0.035107 0.133296 0.263374 0.7928RESID^2(-7) -0.091452 0.132903 -0.688115 0.4929RESID^2(-8) 0.031280 0.132927 0.235318 0.8144RESID^2(-9) 0.022800 0.127023 0.179495 0.8579
RESID^2(-10) -0.040199 0.095062 -0.422874 0.6733
R-squared 0.482402 Mean dependent var 2.83E-05Adjusted R-squared 0.432633 S.D. dependent var 6.47E-05S.E. of regression 4.87E-05 Akaike info criterion -16.93051Sum squared resid 2.47E-07 Schwarz criterion -16.66795Log likelihood 984.5042 Hannan-Quinn criter. -16.82394F-statistic 9.692826 Durbin-Watson stat 1.822560Prob(F-statistic) 0.000000
65 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 5 (f) – Modified Black-Scholes model for Sembcorp Marine
Dependent Variable: BS_MODEL_PRICEMethod: ARMA Maximum Likelihood (OPG - BHHH)Date: 10/18/16 Time: 20:17Sample: 2/01/2016 6/10/2016Included observations: 95Convergence achieved after 20 iterationsCoefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
C -0.053299 0.002625 -20.30663 0.0000WARRANT_PRICE 1.423312 0.048616 29.27654 0.0000
AR(1) 0.985305 0.076036 12.95832 0.0000AR(2) -0.354304 0.071882 -4.928993 0.0000
SIGMASQ 9.61E-06 1.33E-06 7.243185 0.0000
R-squared 0.955823 Mean dependent var 0.013206Adjusted R-squared 0.953859 S.D. dependent var 0.014825S.E. of regression 0.003184 Akaike info criterion -8.599133Sum squared resid 0.000913 Schwarz criterion -8.464719Log likelihood 413.4588 Hannan-Quinn criter. -8.544820F-statistic 486.8107 Durbin-Watson stat 1.901681Prob(F-statistic) 0.000000
Inverted AR Roots .49-.33i .49+.33i
Appendix 5 (g) – VIF test for Black-Scholes model for Sembcorp Marine
Variance Inflation FactorsDate: 10/18/16 Time: 20:17Sample: 2/01/2016 6/10/2016Included observations: 95
Coefficient Uncentered CenteredVariable Variance VIF VIF
C 6.89E-06 9.316912 NAWARRANT_PRICE 0.002364 9.892869 1.329405
AR(1) 0.005782 1.594611 1.424966AR(2) 0.005167 1.490718 1.455466
SIGMASQ 1.76E-12 1.354383 1.327363
66 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 5 (h) – Correlogram of modified Black-Scholes model for Sembcorp Marine
67 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 5 (i) – Static forecasts for modified Black-Scholes model for Sembcorp Marine
Appendix 5 (j) – Dynamic forecasts of modified Black-Scholes model for Sembcorp Marine
68 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 6 (a) – Simple regression output of Trinomial model for Sembcorp Marine
Dependent Variable: TRINOMIAL_MODEL_PRICEMethod: Least SquaresDate: 10/17/16 Time: 22:37Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C -0.033557 0.001832 -18.31805 0.0000WARRANT_PRICE 0.950987 0.040318 23.58704 0.0000
R-squared 0.818944 Mean dependent var 0.008515Adjusted R-squared 0.817472 S.D. dependent var 0.010926S.E. of regression 0.004668 Akaike info criterion -7.880300Sum squared resid 0.002680 Schwarz criterion -7.835047Log likelihood 494.5187 Hannan-Quinn criter. -7.861916F-statistic 556.3484 Durbin-Watson stat 0.437609Prob(F-statistic) 0.000000
Appendix 6 (b) – Correlogram of simple regression Trinomial model for Sembcorp Marine
69 | P a g e
Risk Management and Financial Engineering (BAFI2081)
70 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 6 (c) – Serial correlation LM test for Trinomial model for Sembcorp Marine
Breusch-Godfrey Serial Correlation LM Test:
F-statistic 42.68507 Prob. F(5,118) 0.0000Obs*R-squared 80.49526 Prob. Chi-Square(5) 0.0000
Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 10/18/16 Time: 20:26Sample: 1/04/2016 6/24/2016Included observations: 125Presample missing value lagged residuals set to zero.
Variable Coefficient Std. Error t-Statistic Prob.
C 4.33E-05 0.001122 0.038597 0.9693WARRANT_PRICE -0.000469 0.024689 -0.018993 0.9849
RESID(-1) 1.046455 0.091928 11.38345 0.0000RESID(-2) -0.426625 0.133238 -3.201975 0.0018RESID(-3) 0.084685 0.138904 0.609668 0.5433RESID(-4) 0.012923 0.133839 0.096554 0.9232RESID(-5) 0.054628 0.092721 0.589167 0.5569
R-squared 0.643962 Mean dependent var -1.52E-17Adjusted R-squared 0.625858 S.D. dependent var 0.004649S.E. of regression 0.002844 Akaike info criterion -8.833018Sum squared resid 0.000954 Schwarz criterion -8.674632Log likelihood 559.0636 Hannan-Quinn criter. -8.768674F-statistic 35.57089 Durbin-Watson stat 1.931104Prob(F-statistic) 0.000000
71 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 6 (d) – Heteroskadacity (White) test for Trinomial model for Sembcorp Marine
Heteroskedasticity Test: White
F-statistic 61.90164 Prob. F(2,122) 0.0000Obs*R-squared 62.95852 Prob. Chi-Square(2) 0.0000Scaled explained SS 153.6181 Prob. Chi-Square(2) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 20:27Sample: 1/04/2016 6/24/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C 0.000483 5.46E-05 8.837945 0.0000WARRANT_PRICE^2 0.248886 0.025015 9.949340 0.0000
WARRANT_PRICE -0.022038 0.002382 -9.251941 0.0000
R-squared 0.503668 Mean dependent var 2.14E-05Adjusted R-squared 0.495532 S.D. dependent var 4.83E-05S.E. of regression 3.43E-05 Akaike info criterion -17.69759Sum squared resid 1.44E-07 Schwarz criterion -17.62971Log likelihood 1109.099 Hannan-Quinn criter. -17.67001F-statistic 61.90164 Durbin-Watson stat 1.087646Prob(F-statistic) 0.000000
72 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 6 (e) – Heteroskadacity (ARCH) test for Trinomial model for Sembcorp Marine
Heteroskedasticity Test: ARCH
F-statistic 12.22804 Prob. F(10,104) 0.0000Obs*R-squared 62.14522 Prob. Chi-Square(10) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 20:27Sample (adjusted): 1/18/2016 6/24/2016Included observations: 115 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 7.58E-06 4.20E-06 1.806566 0.0737RESID^2(-1) 0.945942 0.097967 9.655747 0.0000RESID^2(-2) -0.463757 0.134140 -3.457248 0.0008RESID^2(-3) 0.059376 0.141575 0.419395 0.6758RESID^2(-4) 0.163452 0.141606 1.154273 0.2510RESID^2(-5) -0.057433 0.142444 -0.403200 0.6876RESID^2(-6) 0.066209 0.142345 0.465129 0.6428RESID^2(-7) -0.100208 0.141654 -0.707419 0.4809RESID^2(-8) 0.041297 0.141792 0.291247 0.7714RESID^2(-9) 0.006904 0.133835 0.051583 0.9590
RESID^2(-10) -0.046210 0.096272 -0.479993 0.6322
R-squared 0.540393 Mean dependent var 2.01E-05Adjusted R-squared 0.496200 S.D. dependent var 4.91E-05S.E. of regression 3.49E-05 Akaike info criterion -17.59917Sum squared resid 1.26E-07 Schwarz criterion -17.33661Log likelihood 1022.952 Hannan-Quinn criter. -17.49260F-statistic 12.22804 Durbin-Watson stat 1.842243Prob(F-statistic) 0.000000
73 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 6 (f) – Modified Trinomial model for Sembcorp Marine
Dependent Variable: TRINOMIAL_MODEL_PRICEMethod: ARMA Maximum Likelihood (OPG - BHHH)Date: 10/18/16 Time: 20:29Sample: 2/01/2016 6/10/2016Included observations: 95Convergence achieved after 15 iterationsCoefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
C -0.042378 0.002354 -18.00585 0.0000WARRANT_PRICE 1.118704 0.039248 28.50377 0.0000
AR(1) 0.569424 0.102628 5.548407 0.0000MA(1) 0.462442 0.114595 4.035434 0.0001
SIGMASQ 6.76E-06 1.00E-06 6.750518 0.0000
R-squared 0.950056 Mean dependent var 0.009858Adjusted R-squared 0.947836 S.D. dependent var 0.011696S.E. of regression 0.002671 Akaike info criterion -8.949753Sum squared resid 0.000642 Schwarz criterion -8.815339Log likelihood 430.1133 Hannan-Quinn criter. -8.895440F-statistic 428.0056 Durbin-Watson stat 1.919620Prob(F-statistic) 0.000000
Inverted AR Roots .57Inverted MA Roots -.46
Appendix 6 (g) – VIF test for modified Trinomial model for Sembcorp Marine
Variance Inflation FactorsDate: 10/18/16 Time: 20:30Sample: 2/01/2016 6/10/2016Included observations: 95
Coefficient Uncentered CenteredVariable Variance VIF VIF
C 5.54E-06 6.817023 NAWARRANT_PRICE 0.001540 6.692471 1.213586
AR(1) 0.010533 2.269428 1.939369MA(1) 0.013132 2.438627 2.188084
SIGMASQ 1.00E-12 1.539229 1.536801
74 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 6 (h) – Correlogram for Modified Trinomial model for Sembcorp Marine
75 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 6 (i) – Static forecasts for Modified Trinomial model for Sembcorp Marine
Appendix 6 (j) – Dynamic forecasts for Modified Trinomial model for Sembcorp Marine
76 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 7 (a) – Simple regression output for Binomial model for Wilmar International
Dependent Variable: BINOMIAL_MODEL_PRICEMethod: Least SquaresDate: 10/17/16 Time: 22:43Sample: 1/04/2016 6/30/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C 0.332709 0.013055 25.48497 0.0000WARRANT_PRICE 0.504644 0.140011 3.604311 0.0005
R-squared 0.095529 Mean dependent var 0.366112Adjusted R-squared 0.088175 S.D. dependent var 0.107658S.E. of regression 0.102802 Akaike info criterion -1.696147Sum squared resid 1.299903 Schwarz criterion -1.650894Log likelihood 108.0092 Hannan-Quinn criter. -1.677763F-statistic 12.99106 Durbin-Watson stat 0.234457Prob(F-statistic) 0.000453
77 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 7 (b) – Correlaogram for Simple regression output for Binomial model for Wilmar International
78 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 7 (c) – Serial correlation test for Binomial model for Wilmar International
Breusch-Godfrey Serial Correlation LM Test:
F-statistic 148.8570 Prob. F(5,118) 0.0000Obs*R-squared 107.8943 Prob. Chi-Square(5) 0.0000
Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 10/18/16 Time: 20:47Sample: 1/04/2016 6/30/2016Included observations: 125Presample missing value lagged residuals set to zero.
Variable Coefficient Std. Error t-Statistic Prob.
C 0.024902 0.005019 4.961478 0.0000WARRANT_PRICE -0.387212 0.054788 -7.067498 0.0000
RESID(-1) 0.590192 0.077539 7.611593 0.0000RESID(-2) 0.237470 0.087616 2.710346 0.0077RESID(-3) 0.187075 0.088029 2.125150 0.0357RESID(-4) 0.001597 0.087673 0.018220 0.9855RESID(-5) -0.015496 0.077349 -0.200342 0.8416
R-squared 0.863154 Mean dependent var -6.94E-17Adjusted R-squared 0.856196 S.D. dependent var 0.102387S.E. of regression 0.038827 Akaike info criterion -3.605048Sum squared resid 0.177886 Schwarz criterion -3.446663Log likelihood 232.3155 Hannan-Quinn criter. -3.540705F-statistic 124.0475 Durbin-Watson stat 1.145562Prob(F-statistic) 0.000000
79 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 7 (d) – Heteroskadacity (White) test for Binomial model for Wilmar International
Heteroskedasticity Test: White
F-statistic 4.944652 Prob. F(2,122) 0.0086Obs*R-squared 9.372731 Prob. Chi-Square(2) 0.0092Scaled explained SS 8.533927 Prob. Chi-Square(2) 0.0140
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 20:48Sample: 1/04/2016 6/30/2016Included observations: 125
Variable Coefficient Std. Error t-Statistic Prob.
C 0.008783 0.005738 1.530569 0.1285WARRANT_PRICE^2 0.051178 0.122329 0.418362 0.6764
WARRANT_PRICE 0.017700 0.100361 0.176366 0.8603
R-squared 0.074982 Mean dependent var 0.010399Adjusted R-squared 0.059818 S.D. dependent var 0.014319S.E. of regression 0.013884 Akaike info criterion -5.692462Sum squared resid 0.023517 Schwarz criterion -5.624583Log likelihood 358.7789 Hannan-Quinn criter. -5.664886F-statistic 4.944652 Durbin-Watson stat 0.188648Prob(F-statistic) 0.008613
80 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 7 (e) – Heteroskadacity (ARCH) test for Binomial model for Wilmar International
Heteroskedasticity Test: ARCH
F-statistic 31.75125 Prob. F(10,104) 0.0000Obs*R-squared 86.62599 Prob. Chi-Square(10) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 20:48Sample (adjusted): 1/18/2016 6/30/2016Included observations: 115 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.001109 0.000971 1.142060 0.2561RESID^2(-1) 0.719418 0.097227 7.399392 0.0000RESID^2(-2) 0.072816 0.120129 0.606147 0.5457RESID^2(-3) 0.019719 0.119336 0.165238 0.8691RESID^2(-4) 0.224577 0.118949 1.888016 0.0618RESID^2(-5) -0.079318 0.120749 -0.656883 0.5127RESID^2(-6) 0.050907 0.119755 0.425089 0.6717RESID^2(-7) -0.098749 0.117312 -0.841761 0.4019RESID^2(-8) -0.150543 0.117672 -1.279343 0.2036RESID^2(-9) -0.005391 0.118190 -0.045613 0.9637
RESID^2(-10) 0.114540 0.095823 1.195330 0.2347
R-squared 0.753270 Mean dependent var 0.009913Adjusted R-squared 0.729545 S.D. dependent var 0.014674S.E. of regression 0.007631 Akaike info criterion -6.822311Sum squared resid 0.006057 Schwarz criterion -6.559752Log likelihood 403.2829 Hannan-Quinn criter. -6.715739F-statistic 31.75125 Durbin-Watson stat 1.968904Prob(F-statistic) 0.000000
81 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 7 (f) – Modified Binomial model for Wilmar International
Dependent Variable: BINOMIAL_MODEL_PRICEMethod: ARMA Maximum Likelihood (OPG - BHHH)Date: 10/18/16 Time: 20:51Sample: 2/01/2016 6/10/2016Included observations: 91Convergence achieved after 28 iterationsCoefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
C 0.380037 0.047176 8.055651 0.0000WARRANT_PRICE 0.022098 0.022871 0.966204 0.3367
AR(1) 0.946363 0.025812 36.66330 0.0000MA(1) 0.209361 0.084102 2.489375 0.0147
SIGMASQ 0.000622 0.000111 5.606801 0.0000
R-squared 0.924040 Mean dependent var 0.406945Adjusted R-squared 0.920507 S.D. dependent var 0.091015S.E. of regression 0.025661 Akaike info criterion -4.405001Sum squared resid 0.056631 Schwarz criterion -4.267042Log likelihood 205.4276 Hannan-Quinn criter. -4.349343F-statistic 261.5452 Durbin-Watson stat 2.024783Prob(F-statistic) 0.000000
Inverted AR Roots .95Inverted MA Roots -.21
Appendix 7 (g) – VIF test for modified Binomial model for Wilmar International
Variance Inflation FactorsDate: 10/18/16 Time: 20:51Sample: 2/01/2016 6/10/2016Included observations: 91
Coefficient Uncentered CenteredVariable Variance VIF VIF
C 0.002226 1.087213 NAWARRANT_PRICE 0.000523 1.006811 1.006564
AR(1) 0.000666 1.175969 1.121351MA(1) 0.007073 1.033015 1.029597
SIGMASQ 1.23E-08 1.194666 1.125863
82 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 7 (h) – Correlogram for modified Binomial model for Wilmar International
83 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 7 (i) – Static forecasts for modified Binomial model for Wilmar International
Appendix 7 (j) – dynamic forecasts for modified Binomial model for Wilmar International
84 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 8 (a) – Simple regression output for Black-Scholes model for Wilmar International
Dependent Variable: BS_MODEL_PRICEMethod: Least SquaresDate: 10/18/16 Time: 20:57Sample: 1/04/2016 6/24/2016Included observations: 121
Variable Coefficient Std. Error t-Statistic Prob.
C 0.334196 0.013352 25.03015 0.0000WARRANT_PRICE 0.498048 0.141511 3.519505 0.0006
R-squared 0.094278 Mean dependent var 0.367446Adjusted R-squared 0.086667 S.D. dependent var 0.108597S.E. of regression 0.103784 Akaike info criterion -1.676615Sum squared resid 1.281768 Schwarz criterion -1.630404Log likelihood 103.4352 Hannan-Quinn criter. -1.657847F-statistic 12.38691 Durbin-Watson stat 0.235534Prob(F-statistic) 0.000613
85 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 8 (b) – Correlogram of simple regression output for Black-Scholes model for Wilmar International
86 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 8 (c) – Serial correlation test for Black-Scholes model for Wilmar International
Breusch-Godfrey Serial Correlation LM Test:
F-statistic 144.2704 Prob. F(5,114) 0.0000Obs*R-squared 104.4872 Prob. Chi-Square(5) 0.0000
Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 10/18/16 Time: 20:59Sample: 1/04/2016 6/24/2016Included observations: 121Presample missing value lagged residuals set to zero.
Variable Coefficient Std. Error t-Statistic Prob.
C 0.025091 0.005129 4.892365 0.0000WARRANT_PRICE -0.386775 0.055362 -6.986257 0.0000
RESID(-1) 0.585623 0.078911 7.421303 0.0000RESID(-2) 0.239502 0.088943 2.692752 0.0082RESID(-3) 0.189896 0.089313 2.126199 0.0356RESID(-4) 0.003134 0.088946 0.035236 0.9720RESID(-5) -0.016506 0.078656 -0.209849 0.8342
R-squared 0.863531 Mean dependent var 5.96E-17Adjusted R-squared 0.856348 S.D. dependent var 0.103351S.E. of regression 0.039171 Akaike info criterion -3.585625Sum squared resid 0.174922 Schwarz criterion -3.423885Log likelihood 223.9303 Hannan-Quinn criter. -3.519937F-statistic 120.2254 Durbin-Watson stat 1.146114Prob(F-statistic) 0.000000
87 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 8 (d) – Heteroskadacity (White) test for Black-Scholes model for Wilmar International
Heteroskedasticity Test: White
F-statistic 4.596896 Prob. F(2,118) 0.0120Obs*R-squared 8.746094 Prob. Chi-Square(2) 0.0126Scaled explained SS 7.535418 Prob. Chi-Square(2) 0.0231
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 21:00Sample: 1/04/2016 6/24/2016Included observations: 121
Variable Coefficient Std. Error t-Statistic Prob.
C 0.012130 0.005793 2.093835 0.0384WARRANT_PRICE^2 0.114730 0.122673 0.935246 0.3516
WARRANT_PRICE -0.038323 0.100737 -0.380423 0.7043
R-squared 0.072282 Mean dependent var 0.010593Adjusted R-squared 0.056558 S.D. dependent var 0.014198S.E. of regression 0.013791 Akaike info criterion -5.705175Sum squared resid 0.022441 Schwarz criterion -5.635858Log likelihood 348.1631 Hannan-Quinn criter. -5.677022F-statistic 4.596896 Durbin-Watson stat 0.194622Prob(F-statistic) 0.011955
88 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 8 (e) – Heteroskadacity (ARCH) test for Black-Scholes model for Wilmar International
Heteroskedasticity Test: ARCH
F-statistic 30.36941 Prob. F(10,100) 0.0000Obs*R-squared 83.50393 Prob. Chi-Square(10) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 21:00Sample (adjusted): 1/18/2016 6/24/2016Included observations: 111 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.001127 0.000998 1.129201 0.2615RESID^2(-1) 0.728278 0.099139 7.346023 0.0000RESID^2(-2) 0.060456 0.123050 0.491312 0.6243RESID^2(-3) 0.012886 0.122088 0.105547 0.9162RESID^2(-4) 0.227287 0.121762 1.866651 0.0649RESID^2(-5) -0.072092 0.123639 -0.583085 0.5611RESID^2(-6) 0.039068 0.122247 0.319585 0.7499RESID^2(-7) -0.081427 0.119518 -0.681293 0.4973RESID^2(-8) -0.152865 0.119752 -1.276517 0.2047RESID^2(-9) -0.007846 0.120247 -0.065250 0.9481
RESID^2(-10) 0.113341 0.097191 1.166167 0.2463
R-squared 0.752288 Mean dependent var 0.010039Adjusted R-squared 0.727516 S.D. dependent var 0.014508S.E. of regression 0.007573 Akaike info criterion -6.834537Sum squared resid 0.005735 Schwarz criterion -6.566025Log likelihood 390.3168 Hannan-Quinn criter. -6.725609F-statistic 30.36941 Durbin-Watson stat 1.965859Prob(F-statistic) 0.000000
89 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 8 (f) – Modified Black-Scholes model for Wilmar International
Dependent Variable: BS_MODEL_PRICEMethod: ARMA Maximum Likelihood (OPG - BHHH)Date: 10/18/16 Time: 21:03Sample: 1/04/2016 6/24/2016Included observations: 121Convergence achieved after 16 iterationsCoefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
C 0.337010 0.074657 4.514121 0.0000WARRANT_PRICE 0.022268 0.025167 0.884820 0.3781
AR(1) 0.969535 0.017881 54.22204 0.0000SIGMASQ 0.000631 9.18E-05 6.873716 0.0000
R-squared 0.946038 Mean dependent var 0.367446Adjusted R-squared 0.944654 S.D. dependent var 0.108597S.E. of regression 0.025548 Akaike info criterion -4.440754Sum squared resid 0.076367 Schwarz criterion -4.348331Log likelihood 272.6656 Hannan-Quinn criter. -4.403218F-statistic 683.7279 Durbin-Watson stat 1.835533Prob(F-statistic) 0.000000
Inverted AR Roots .97
Appendix 8 (g) – VIF test for Modified Black-Scholes model for Wilmar International
Variance Inflation FactorsDate: 10/18/16 Time: 21:03Sample: 1/04/2016 6/24/2016Included observations: 121
Coefficient Uncentered CenteredVariable Variance VIF VIF
C 0.005574 1.194675 NAWARRANT_PRICE 0.000633 1.009553 1.007207
AR(1) 0.000320 1.373864 1.154906SIGMASQ 8.43E-09 1.206686 1.151733
90 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 8 (h) – Correlogram for Modified Black-Scholes model for Wilmar International
91 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 8 (i) – Static forecast for Modified Black-Scholes model for Wilmar International
Appendix 8 (j) – Dynamic forecasts Modified Black-Scholes model for Wilmar International
92 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 9 (a) – Simple regression output for Trinomial model for Wilmar International
Dependent Variable: TRINOMIAL_MODEL_PRICEMethod: Least SquaresDate: 10/18/16 Time: 21:10Sample: 1/04/2016 6/24/2016Included observations: 121
Variable Coefficient Std. Error t-Statistic Prob.
C 0.308355 0.012826 24.04175 0.0000WARRANT_PRICE 0.479086 0.135936 3.524339 0.0006
R-squared 0.094513 Mean dependent var 0.340339Adjusted R-squared 0.086904 S.D. dependent var 0.104332S.E. of regression 0.099696 Akaike info criterion -1.756992Sum squared resid 1.182776 Schwarz criterion -1.710780Log likelihood 108.2980 Hannan-Quinn criter. -1.738224F-statistic 12.42097 Durbin-Watson stat 0.233876Prob(F-statistic) 0.000603
Appendix 9 (b) – Simple regression output for Trinomial model for Wilmar International
93 | P a g e
Risk Management and Financial Engineering (BAFI2081)
94 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 9 (c) – Serial correlation test for Trinomial model for Wilmar International
Breusch-Godfrey Serial Correlation LM Test:
F-statistic 144.9007 Prob. F(5,114) 0.0000Obs*R-squared 104.5493 Prob. Chi-Square(5) 0.0000
Test Equation:Dependent Variable: RESIDMethod: Least SquaresDate: 10/18/16 Time: 21:12Sample: 1/04/2016 6/24/2016Included observations: 121Presample missing value lagged residuals set to zero.
Variable Coefficient Std. Error t-Statistic Prob.
C 0.024095 0.004917 4.900095 0.0000WARRANT_PRICE -0.370186 0.053070 -6.975389 0.0000
RESID(-1) 0.585134 0.078935 7.412882 0.0000RESID(-2) 0.244359 0.088910 2.748377 0.0070RESID(-3) 0.184901 0.089410 2.068020 0.0409RESID(-4) 0.000295 0.088904 0.003314 0.9974RESID(-5) -0.013737 0.078657 -0.174640 0.8617
R-squared 0.864044 Mean dependent var -1.19E-16Adjusted R-squared 0.856888 S.D. dependent var 0.099280S.E. of regression 0.037558 Akaike info criterion -3.669768Sum squared resid 0.160806 Schwarz criterion -3.508028Log likelihood 229.0209 Hannan-Quinn criter. -3.604079F-statistic 120.7506 Durbin-Watson stat 1.140804Prob(F-statistic) 0.000000
95 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 9 (d) – Heteroskadacity (White) test for Trinomial model for Wilmar InternationalHeteroskedasticity Test: White
F-statistic 4.583399 Prob. F(2,118) 0.0121Obs*R-squared 8.722266 Prob. Chi-Square(2) 0.0128Scaled explained SS 7.457200 Prob. Chi-Square(2) 0.0240
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 21:13Sample: 1/04/2016 6/24/2016Included observations: 121
Variable Coefficient Std. Error t-Statistic Prob.
C 0.011600 0.005326 2.177965 0.0314WARRANT_PRICE^2 0.113859 0.112776 1.009605 0.3148
WARRANT_PRICE -0.042513 0.092609 -0.459058 0.6470
R-squared 0.072085 Mean dependent var 0.009775Adjusted R-squared 0.056357 S.D. dependent var 0.013051S.E. of regression 0.012678 Akaike info criterion -5.873422Sum squared resid 0.018966 Schwarz criterion -5.804105Log likelihood 358.3420 Hannan-Quinn criter. -5.845270F-statistic 4.583399 Durbin-Watson stat 0.192656Prob(F-statistic) 0.012106
96 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 9 (e) – Simple regression output for Trinomial model for Wilmar International
Heteroskedasticity Test: ARCH
F-statistic 30.66501 Prob. F(10,100) 0.0000Obs*R-squared 83.70381 Prob. Chi-Square(10) 0.0000
Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 10/18/16 Time: 21:13Sample (adjusted): 1/18/2016 6/24/2016Included observations: 111 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C 0.001011 0.000915 1.104728 0.2719RESID^2(-1) 0.730726 0.098954 7.384536 0.0000RESID^2(-2) 0.062973 0.123018 0.511903 0.6098RESID^2(-3) 0.003506 0.122251 0.028679 0.9772RESID^2(-4) 0.227036 0.121922 1.862137 0.0655RESID^2(-5) -0.066249 0.123819 -0.535047 0.5938RESID^2(-6) 0.037191 0.122524 0.303542 0.7621RESID^2(-7) -0.081821 0.119856 -0.682657 0.4964RESID^2(-8) -0.139310 0.120118 -1.159775 0.2489RESID^2(-9) -0.026123 0.120465 -0.216855 0.8288
RESID^2(-10) 0.120482 0.097187 1.239700 0.2180
R-squared 0.754088 Mean dependent var 0.009241Adjusted R-squared 0.729497 S.D. dependent var 0.013333S.E. of regression 0.006935 Akaike info criterion -7.010749Sum squared resid 0.004809 Schwarz criterion -6.742237Log likelihood 400.0966 Hannan-Quinn criter. -6.901821F-statistic 30.66501 Durbin-Watson stat 1.963487Prob(F-statistic) 0.000000
97 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 9 (f) – Modified Trinomial model for Wilmar International
Dependent Variable: TRINOMIAL_MODEL_PRICEMethod: ARMA Maximum Likelihood (OPG - BHHH)Date: 10/18/16 Time: 21:15Sample: 1/04/2016 6/24/2016Included observations: 121Convergence achieved after 16 iterationsCoefficient covariance computed using outer product of gradients
Variable Coefficient Std. Error t-Statistic Prob.
C 0.311134 0.071083 4.377026 0.0000WARRANT_PRICE 0.022516 0.023435 0.960746 0.3387
AR(1) 0.970092 0.017642 54.98841 0.0000SIGMASQ 0.000570 8.17E-05 6.976230 0.0000
R-squared 0.947194 Mean dependent var 0.340339Adjusted R-squared 0.945840 S.D. dependent var 0.104332S.E. of regression 0.024281 Akaike info criterion -4.542383Sum squared resid 0.068977 Schwarz criterion -4.449960Log likelihood 278.8142 Hannan-Quinn criter. -4.504846F-statistic 699.5537 Durbin-Watson stat 1.843858Prob(F-statistic) 0.000000
Inverted AR Roots .97
Appendix 9 (g) – Simple regression output for Trinomial model for Wilmar International
Variance Inflation FactorsDate: 10/18/16 Time: 21:16Sample: 1/04/2016 6/24/2016Included observations: 121
Coefficient Uncentered CenteredVariable Variance VIF VIF
C 0.005053 1.191672 NAWARRANT_PRICE 0.000549 1.009684 1.007340
AR(1) 0.000311 1.363339 1.147384SIGMASQ 6.68E-09 1.190825 1.143986
98 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 9 (h) – Correlogram for Modified Trinomial model for Wilmar International
99 | P a g e
Risk Management and Financial Engineering (BAFI2081)
Appendix 9 (i) – Static forecasts for Modified Trinomial model for Wilmar International
Appendix 9 (j) – Dynamic forecasts for Modified Trinomial model for Wilmar International
100 | P a g e