Vous êtes sur la page 1sur 53

Dublin Institute of Technology

Studying indications of equity bubbles through


economic variables in the S&P 500 Composite Index.

Author: Ivan Madaras

Supervisor: Brian McGillion


Submitted as a partial requirement for the B.Sc. (Economics and Finance)
April 2014

Declaration

This is an original work. All references and assistance are acknowledged

Signed
(Candidate)

Date

Acknowledgements
Firstly, I would like to thank Mr. Brian McGillion for his supervision and input in this
research project. Also, a big thank you to PhD. Damien Cassells, and mainly PhD. Lucia
Morales, for their assistance and encouragement.
I would also like to thank all the lecturing staff at the DIT for all their help and hard work
throughout the academic years. Their hard work and dedication made the time at DIT most
enjoyable and rewarding.
Lastly, Id like to thank my fellow classmates whose support and humour made the
education a bit easier and above all, more enjoyable.

Table of Contents
LIST OF EQUATIONS ................................................................................................................................ IV
LIST OF TABLES ...................................................................................................................................... IV
TABLE OF FIGURES ................................................................................................................................. IV
LIST OF APPENDICES ................................................................................................................................ V
LIST OF ABBREVIATIONS .......................................................................................................................... VI
ABSTRACT ............................................................................................................................................. 1
1.

INTRODUCTION ....................................................................................................................... 2
1.1. Empirical evidence of Irrational Exuberance ........................................................................ 2
1.2. Research aims and objectives ................................................................................................. 3
1.3. Research question................................................................................................................... 3
1.4. Introduction summary ............................................................................................................ 4

2.

LITERATURE REVIEW ................................................................................................................ 5


2.1. SEPARATION OF SPOT PRICE AND FUNDAMENTAL VALUE: PRICE BUBBLE ...................................................... 5
2.2. FACTORS INFLUENCING PRICING OF SECURITIES....................................................................................... 6
2.2.1.

Internal Factors ............................................................................................................ 8

Dividends................................................................................................................................................... 9
Earnings ..................................................................................................................................................... 9

2.2.2.

External Factors ......................................................................................................... 10

Gross Domestic Product .......................................................................................................................... 10


Interest Rate............................................................................................................................................ 11
Short Term Interest Rate ......................................................................................................................... 12
Long Term Interest Rate .......................................................................................................................... 12
Commodities ........................................................................................................................................... 12
Oil ............................................................................................................................................................ 12
Gold ......................................................................................................................................................... 13
Inflation ................................................................................................................................................... 14
Exchange Rate ......................................................................................................................................... 14

2.3. FACTORS INFLUENCING MISPRICING OF SECURITIES................................................................................ 15


2.4. SUMMARY OF LITERATURE REVIEW.................................................................................................... 17
3.

DATA AND METHODOLOGY ................................................................................................... 18


3.1. DATA DESCRIPTION ........................................................................................................................ 18
3.2. EMPIRICAL MODEL ......................................................................................................................... 19
3.3. ORDINARY LEAST SQUARES .............................................................................................................. 19
3.3.1. Spurious Correlation and Stationarity ............................................................................... 20
3.3.2. Multicollinearity................................................................................................................. 21
3.3.3. Serial Correlation ............................................................................................................... 21

II

3.3.4. Heteroskedasticity ............................................................................................................. 22


3.3.5. Specification ...................................................................................................................... 22
3.4. HYPOTHESIS FORMULATION ............................................................................................................. 24
4.

EMPIRICAL RESULTS ............................................................................................................... 26


4.1. STATIONARITY TEST RESULTS ............................................................................................................ 26
4.2. MULTICOLLINEARITY TEST RESULTS ................................................................................................... 28
4.3. SERIAL CORRELATION TEST RESULTS .................................................................................................. 30
4.4. HETEROSKEDASTICITY TEST RESULTS .................................................................................................. 30
4.5. SPECIFICATION TEST RESULTS ........................................................................................................... 30
4.5.1. Ramseys Specification Error Test Results .......................................................................... 31
4.5.2. Irrelevant Variable Test Results ......................................................................................... 31
4.6. REGRESSION ANALYSIS RESULTS ........................................................................................................ 33
4.7. INTERPRETATION OF REGRESSION ANALYSIS ........................................................................................ 34

5.

CONCLUSION ........................................................................................................................ 37

6.

REFERENCES........................................................................................................................... 39

III

List of Equations
EQUATION 1: DIVIDEND DISCOUNT MODEL ..................................................................................................... 7
EQUATION 2: EARNINGS DISCOUNT MODEL ..................................................................................................... 8
EQUATION 3: REGRESSION EQUATION........................................................................................................... 19
EQUATION 4: REGRESSION EQUATION (DETAILED) .......................................................................................... 28
EQUATION 5: ESTIMATED REGRESSION EQUATION .......................................................................................... 33

List of Tables
TABLE 1: HYPOTHESIS FORMULATION ........................................................................................................... 24
TABLE 2: DICKEY FULLER TEST................................................................................................................... 27
TABLE 3: METHODS USED FOR CALCULATING VARIABLES ................................................................................... 28
TABLE 4: CORRELATION MATRIX .................................................................................................................. 29
TABLE 5: VARIANCE INFLATION FACTORS ....................................................................................................... 29
TABLE 6: DURBIN WATSON D TEST ............................................................................................................... 30
TABLE 7: BREUSCH-PAGAN TEST FOR HETEROSKEDASTICITY ............................................................................... 30
TABLE 8: RAMSEYS SPECIFICATION ERROR TEST ............................................................................................. 31
TABLE 9: REGRESSION RESULTS (1) .............................................................................................................. 31
TABLE 10: REGRESSION RESULTS (2) ............................................................................................................ 32
TABLE 11: REGRESSION RESULTS (3) ............................................................................................................ 33

Table of Figures
FIGURE 1: DEPENDANT AND INDEPENDENT VARIABLES ..................................................................................... 26
FIGURE 2: S&P 500 AND DIVIDENDS VOLATILITY ............................................................................................ 35

IV

List of Appendices
APPENDIX 1 .......................................................................................................................................I
SUMMARY OF VARIABLES .......................................................................................................................... I
CORRELATION MATRIX WITH SIGNIFICANCE LEVELS ........................................................................................ I
REGRESSION ANALYSIS [1]........................................................................................................................ II
Variance Inflation Factor ............................................................................................................... ii
RESET test ...................................................................................................................................... ii
Heteroskedasticity Test ................................................................................................................. ii
Durbin Watson Test ....................................................................................................................... ii
APPENDIX 2 .....................................................................................................................................III
REGRESSION ANALYSIS [2]........................................................................................................................III
Variance Inflation Factor .............................................................................................................. iii
RESET test ..................................................................................................................................... iii
Heteroskedasticity Test ................................................................................................................ iii
Durbin Watson Test ...................................................................................................................... iii
APPENDIX 3 .................................................................................................................................... IV
REGRESSION ANALYSIS [3]....................................................................................................................... IV
Variance Inflation Factor .............................................................................................................. iv
RESET test ..................................................................................................................................... iv
Heteroskedasticity Test ................................................................................................................ iv
Durbin Watson Test ...................................................................................................................... iv

List of Abbreviations
DPS

Dividends per Share

EPS

Earnings per Share

GDP

Gross Domestic Product

INF

Inflation

ln

Natural Logarithm

LTIR

Long Term Interest Rate

NASDAQ

National Association of Securities Dealers Automated Quotation

NPV

Net Present Value

OECD

Organisation for Economic Co-ordination and Development

OLS

Ordinary Least Squares

PVM

Present Value Model

RESET

Ramseys Specification Error Test

S & P 500

Standard & Poor's 500 Composite Index

STIR

Short Term Interest Rate

VIF

Variance Inflation Factor

VI

Abstract
The research project attempted to answer the question, of whether or not the analysis of
fundamental valuation within the share valuation framework has the potential to indicate an
asset price bubble in the equity markets. The review of literature identified micro and macroeconomic factors directly or indirectly influencing the share price: dividends and earnings
per share, Gross Domestic Product, Short Term and Long Term Interest Rate, price of oil
and gold, Inflation and Exchange Rate, and warranted a construction of an econometric
model.
The OLS regression method was implemented to determine a linear relationship between the
aforementioned independent variables and dependant variable, share price. The empirical
evidence indicates that the variable had only a marginal influence on price, with the
exception of GDP. Additionally, the irrelevance of dividends in the estimated model
suggested that the share valuation is disrupted, and further investigation of the association
between price and dividend volatility allowed for an assumption supported by West (1988),
Lansing (2007) and Shiller (2005), that market prices of equities may be in a bubble territory
when the volatility of spot price significantly exceeds the volatility of the dividends, and
investors should be cautious when assessing investment opportunities.

1. Introduction
The implications of asset price fluctuations are reaching beyond the scope of the stock
market. An increase in price of assets may cause a feel-good factor among employees,
investors and homeowners, who are now feeling wealthier, and are spending more, thus
driving the demand for goods and services, adding to corporate profits and therefore,
fuelling economic growth (Chakraborty, Goldstein and MacKinlay, 2013). The opposite
happens when the asset prices decrease, wealth is being destroyed, pessimism is spreading,
demand is declining, profits are shrinking and the economy may experience stagnation or
even recession. Prices of securities traded on stock exchanges are rising and falling every
day without direct consequences on the everyday life; however, when the rise or fall is
significant enough, the consequences may be enormous, even for people who never open the
business section of their favourite newspaper.
As suggested, the scalable fluctuations in prices may affect the economy and thus everyday
life. Unprecedented increases in the stock prices, unsupported by equivalent increases in the
stocks fundamentals are, quite accurately, compared to a bubble. Fuelled by nothing but thin
air and destined to a fast and violent end (or burst).
The following research project will focus on the potential causes of the asset price bubbles
formation, the principles of asset price valuation and how economic indicators affect it. It
also attempts to answer the question of whether or not the analysis of company fundamentals
can indicate the presence of an equity bubble.
1.1. Empirical evidence of Irrational Exuberance
In the first edition of the book Irrational Exuberance, Shiller (2000) cautions against
inflated market prices and obvious mispricing, particularly with regards to companies
related to the internet. These dot-com companies were highly overpriced at the time of
Shillers research, and to illustrate this, the author points out an example of eToys. The
stock of the company was valued at $8 billion in 1999, when the sales in the previous year
were merely $30 million and the company reported a loss of $28.6 million. In comparison,
market capitalization of Toys R Us was $6 billion with the 1998 sales figure of $11.2
billion, and the company reported a profit of $376 million. The markets valuation of each
company is inconsistent with the underlying fundamentals. Furthermore, in the second
edition of the book, Shiller (2005) employed a similar approach and warned that there were
indications of an asset price bubble present in the property market.
With the benefit of the hindsight, it is clear that Shiller was correct in his assumptions about
the over-valuation in the market, and in both occasions the market experienced a correction.
2

The burst of a dot-com bubble on March 10th 2000 sent the NASDAQ index crashing down
and by 2002 effectively wiping out approximately $ 5 trillion of investors wealth. The more
recent episode in the market had, arguably, worldwide consequences. As a result of the
property bubble burst in 2008, almost $ 6 trillion was lost in property value and an estimated
two million construction jobs were lost in the U.S. alone, not to mention the long-reaching
implications for economies around the world (Ballew, 2011).
1.2. Research aims and objectives
As stated, the tendency of spot prices to sway away from the underlying fundamental
valuations can have far-reaching consequences. The aim of the research project is to review
academic literature related with the potential causes of formation of asset price bubbles and
to develop an understanding on how economic variables influence the fundamental valuation
of companies listed on stock exchanges. Subsequently, this research aims to offer discussion
on how investors may use the valuation model to recognize the market is overvalued and
might experience a correction.
The intentions of the following project are based on the potentially negative effects of asset
price bubbles burst on economy and society alike. The objectives are as follows:

Review of academic literature associated with definition, formation and potential


causes of asset price bubbles

Identification of economic variables with potential influence on stock valuation in


line with previous research

Construction of econometric regression equation

Analysis of econometric issues related with the intended regression method

Discussion on empirical results and potential implications

Concluding remarks and suggestions for further research


1.3. Research question

Assuming the investors are rational, and using a present value model to assess companies
investment potential, the market price should reflect the underlying value. However, it
appears that often this is not the case. The following research project is attempting to answer
the question, whether the analysis of companies fundamental value, compared to the spot
price, has the potential to indicate an asset price bubble in the stock market, while suggesting
investors should exercise caution when considering investment opportunities.

1.4. Introduction summary


The irrational exuberance in the stock markets has potentially far-reaching negative effects
on everyday life. It is therefore desirable to analyse the relationship behind the stock
valuation and market price, and offer discussion on how this may shed light onto the
indication of an asset price bubble.
The following research project is organised into the following sections: the literature review
section defines an asset price bubble, offers discussion on the companies valuation model
together with the economic variables with the potential to influence the price, and identifies
factors supposedly causing the separation of spot price from the fundamental value. The data
and methodology part discusses the origins of the data sample and proposes a regression
model with the analysis of econometric issues, which have to be addressed before the
regression can be performed. The following empirical results section analyses the results
from the econometric testing, and finaly the last section offers concluding remarks together
with suggestions for further research.

2. Literature Review
The tendency of asset prices to bubble above fundamental value and subsequently crash has
been observed throughout the history, with examples of this including: Dutch Tulip Mania
(1634 - 1637), South Sea Bubble (1719 - 1722), Black Tuesday; October 29th 1929, Black
Monday; October 19th 1987, dot-com bubble at the end of the millennium and most recently,
the global financial crisis (2008 - present).
The purpose of the following review of literature is to offer potential causes of market
bubbles creation and present a stock pricing model that may be used to understand the fair
market value of companies. Moreover, the discussion permits the construction of an
econometric model to find a possible relationship between stock prices and exogenous
variables with the potential to affect the pricing model and development of assumptions on
how to determine the presence of an equity price bubble in the stock markets. The
interpretation of subsequent regression results may assist potential investors with the
application of an investment strategy.
It is organised into four sections as follows; the first part offers definition of an asset bubble,
second debates the stock valuation model, together with a discussion on the factors
potentially affecting the stock price via the model, third part presents arguments on what
may cause the mispricing of securities and the final part summarizes the literature review
and suggests further analysis.

2.1. Separation of spot price and fundamental value: Price bubble


The term bubble, popularised by the press, describes a situation where the price of an asset,
or assets, increases to such levels that the price becomes unsustainable and market
consequently experiences downturn.
Barlevy (2007) and Lansing (2007) offer a more comprehensive description; a situation
where an assets price exceeds the fundamental value of the asset. The fundamental value
mentioned being the present value of properly discounted expected future cash-flows that
the asset yields over the lifetime. The spot price then represents a claim to a stream of future
payments. This definition allows for a construction of a valuation model that is essential for
the analysis and is further developed later in this research project.
The difficulty with the present value of a companys future dividends, and thus its present
fundamental value arises when there is uncertainty about the expected growth of the
company (Smidt, 2005). Lansing (2007) argues that this fundamental principle appears to be
disrupted while observing stock prices. The author has stated that the present value of the
5

future dividends is less volatile and has a tendency to grow in line with the historical average
return. However, the actual price levels of the S&P 500 Index suggest that the market
participants expect greater volatility and a higher return on their investment.
White (2006) stated that explaining stock price movements proved frustratingly difficult.
Through his analysis of the Gordon Growth Model (Gordon and Shapiro, 1956), author
noticed that changes in the risk free rate and the pay-out rate have contributed faintly to
gains in the market. Rational prices should be forward looking, encompassing the expected
course of dividends discounted appropriately; yet the prices tend to swift far more than the
estimated movement in the dividends.
Both Lansing and White seem to have reached a similar conclusion through their respect
analysis. While Lansing used a simple, average historical growth rate and White used the
Gordon Growth Model, both authors remark that, even using the most basic of the analysis
of the stock returns, the models have the potential to approximate the real return of the stock
market. However, this does not explain the increased levels of stock market prices at times;
whats more, this suggests that at times the volatility in stock prices is greater than the
volatility in the respective underlying fundamentals.
Furthermore, prices of equities will fluctuate; however, the movement has the tendency to
move around a computable figure; the mean. Price may move away from the value, but has
the tendency to revert back to it (Molodovsky, 1953 and Shiller, 2005). The fluctuations
around the mean value and mainly the upward movement can be understood as a price
bubble, correspondingly, the reverting behavior from the increased prices can be considered
as a market correction.
The literature suggests that market prices have the tendency to rise above the discounted
value of their respective future cash-flows, creating a price bubble, and are then prone to
revert back to the computed value, thus bursting the bubble. It is therefore vital to focus
further analysis on the valuation model that would allow for estimation of fair market price
based on expectations of future yields, and to analyse micro and macro-economic factors
that may affect the valuation.

2.2. Factors influencing pricing of securities


Previously mentioned research by White (2006) uses the Gordon Growth Model to
determine the fair value of assets. The model is widely accepted by practitioners and
academics alike. It is designed around the theory that the present value of stock is a stream
of discounted future payments. The examination of the model and the factors with the

potential to influence the output allows for an analysis of the effects of fluctuations in the
explanatory variables on the price.
Chen, Roll and Ross (1986) and Humpe and Macmillan (2007) use a Present Value Model
(PVM) in their respective research. The model is an alteration of the aforementioned Gordon
Growth Model. The stream of future cash flows, using dividends as expected payments, to
calculate current price presents itself as follows:
Equation 1: Dividend Discount Model

Where:
= current stock price

D = the expected cash dividend


r = the required rate of return (discount rate)
n = the year in which the dividend payment is expected
The model assumes a constant growth rate for the company that is being analysed. Thus, it is
most suitable for companies which are stable and mature, or current dynamic companies
going through the mature phase of business cycle. Companies with irregular or more cyclical
earning patterns require a more complex dividend capitalization model framework; however
the simplest PVM provides a convenient means of analysing the determinants of stock value
(Farrell, 1985).
The difficulty computing the present value arises when a company does not pay dividends.
At that point, earnings as stream of future payments may be discounted to compute the
present value. For the convenience of intended research and consistency with the reporting
standards the Earning per Share (EPS) will used for further analysis. The Earnings per Share
are calculated by dividing earnings after interest, depreciation and tax by the total amount of
outstanding shares.
Earnings after interest, depreciation and tax are a property of the shareholders. It is
important to note, that dividends may be paid out of the earnings. However, regardless of the
pay-out of dividends, the earnings belong to the shareholders. Therefore, investors and stock
brokers will observe the earnings per share to accurately estimate the fair value of equity
share (Bhatt & Sumangala, 2012).

According to the formerly mentioned, the corporate Earnings per Share are to be considered
as an aid to compute fair stock price by means of discounting future cash flows to present
value, and thus the model can be expressed as follows:
Equation 2: Earnings Discount Model

Where:
= current stock price

EPS = the expected earnings per share


r = the required rate of return (discount rate)
n = the year in which the cash flow is expected
The model discounts expected Earnings per Share by required rate of return r. Assuming the
constant discount rate in the calculations, the model suggests a positive relationship between
Earnings per Share and market share price.
Using the Present Value Model provides convenient means for assessing the fair market
value. Variables affecting the outcome of the pricing model (trivially) are: Dividends per
Share or Earnings per Share as a stream of future payments, required rate of return and time
horizon for which the asset will be held. The fluctuations in the variables, either positive or
negative, have direct effect on the price.
In addition, there are a number of micro-economic (internal) and macro-economic (external)
factors that may indirectly affect the inputs into the discount model and thus influence the
share price. The following section presents internal and external factors with indirect effect
on share valuation.
2.2.1.

Internal Factors

Internal factors describe the variables on the micro-economic level assumed to influence the
variables used in the Present Value Model. Based on Chen, Roll and Ross (1986) and
Humpe and Macmillan (2007), the internal factors are Dividends and Earnings.
Dividends and earnings are at the centre of stock values. The market value of assets, based
on the Present Value Model, consists of the expectations of returns from the asset. It may
appear that at times there is a little association between the value of common stock and
dividend payments. Typically, investors are looking for both dividends and capital
8

appreciation. However, investors facing high-income tax brackets may prefer securities
without stream of payments and actively seek companies habitually retaining large portion
of earnings and thus create long-term gain through capital appreciation (Molodovsky, 1953).
Therefore, the following section will focus on both Dividends and Earnings as Internal
factors influencing pricing of securities.
Dividends
Dividends are a hard core of stock valuation. The significance of dividend payments in the
assessment of present value of assets is one of the pillars of economic and finance
(Molodovsky, 1953).
The Present Value Model suggests that the dividends have a positive effect on stock prices.
The value for D is the numerator, thus regardless the size of the denominator (assuming the
value r remains constant) if dividend increases, the discounted stock price increases as well
and vice versa. Therefore, based on the model, there is a belief that there is a significant
positive relationship between asset prices and dividends.
Earnings
Since many companies would not have a dividend policy integrated into their respective
corporate structure and retain all their earnings, it is unfeasible to use dividends to calculate
the present value to evaluate the stock price. The Present Value Model then uses earnings as
a stream of future payments to assess the stock.
A company producing goods and services and earning revenue covering its cost of
production adds to its financial reserves. Once a company builds up sufficient reserves it
might invest into future operations and in turn earn higher profit. This may attract investors
demand which will result in increase in the market value of equity (Bhatt & Sumangala,
2012).
In addition, the objective of reported earnings is to help investors to evaluate the
fundamental worth of a company. The issue with use of accounting earnings comes from the
complexity of legislation and the changes in reporting standards over time. However, the
misunderstanding of earning and their importance in the financial time-series analysis may
lead to omitting a significant variable from the econometric model (Campbell & Shiller,
1988).
Similarly to the effect of dividends on the stock price, when Earnings per Share are
discounted to present value, the consequence on valuation is positive. Furthermore, as noted

by Bhatt and Sumangala (2012) and as evident from the Present Value Model, the effect of
increase in earnings will have a positive influence on share valuation.
2.2.2.

External Factors

The prices of assets are commonly believed to react to various unanticipated economic
events and it may be the case that some news have greater effect on stock prices than others.
The modern financial theory focuses on widespread or systemic variables as a likely source
of increased volatility in the stock markets. Any systemic variables that affect the economys
pricing component or influence the stream of future payments would also affect the stock
market returns. Such a variable would have no direct effect on current cash flow; however, it
would impose changes in the investment opportunities and consequently stock prices (Chen,
Roll & Ross, 1986).
It follows that any economic variables that may influence stock returns would do so by
affecting the stream of future payments or the discount rate used in the Present Value Model.
The following section will offer the selection of systemic variables, based on reviewed
literature, that may influence the aforementioned asset pricing model, and thus affect stock
prices.
Gross Domestic Product
The Gross Domestic Product (GDP) is regarded as one of the key economic indicators. It
represents the amount of goods and services that are produced in an economy in a specific
period of time. Therefore, it is natural to consider the GDP to be a variable affecting stock
prices.
The Gross Domestic Product is a measure of all currently manufactured goods and services
in the economy at a specified time period valued at market price. Thus it then represents an
aggregate value of all industries in an economy (Liow, Ibrahim & Huang, 2006). Since a
vast number of the companies operating in various industries are in public ownership, and
are therefore listed on the stock exchanges, the economic growth should reflect market
conditions. It follows then, that the GDP could have forecasting power over stock markets.
In times of economic expansion, the confidence in the markets could drive additional
demand for goods and services which will cause increase in production, revenues and
potentially earnings. A number of firms seeking expansion may enlarge their operations or
invest additional funds in positive Net Present Value (NPV) projects. The expectations from
the expansion will be priced into the present value of future cash flows and thus increase the
stock price. On the other hand, in times of economic contraction accompanied by high

10

economic volatility, investors forward-looking confidence may be dulled. This may cause
decrease in demand, contraction in production and lower the expectation of return from
investment assets (Liow, Ibrahim & Huang, 2006).
According to the former, there may be econometric relationship between the Gross Domestic
Product and stock prices. The effect of positive economic outlook would channel through
expectations of increased future earnings to the discounted stock price. Therefore, there is a
belief that GDP has positive impact on stock prices.
Interest Rate
The interest rate has direct effect on the Present Value Model via the discount rate, and also
through the level of expected payments and accordingly the share valuation.
The interest rate possibly affects the stock prices through a number of channels. An increase
in interest rates may reduce companies spending and investment into future projects due to
increase borrowing costs. This may lower their earnings, dividends and thus the stock price.
In addition, investors who use loaned funds to purchase stocks may find themselves facing
higher transaction costs, which would result in increased required rate of returns and
therefore a reduction in demand leading to price depreciation (Maysami, Howe & Hamzah,
2004 and Kim, 2003). Also, an elevated interest rate may pressure the investors to adjust
their portfolios by shedding stocks and buying bonds with now more attractive and less risky
returns (Kim, 2003).
Furthermore, interest rate fluctuations have an indirect effect on stock prices through the
macroeconomic forces. High interest rates may slow down aggregate investment and also
GDP, therefore negatively affecting companies earnings and dividend growth, leading to a
deterioration of stock price (Udegbunam & Eriki, 2001).
The influence of interest rate can be observed on the internal, external and investor levels
alike. To allow for a better understanding of the interest rate effect, the time horizon should
be considered.
Interest rates can be broadly divided into two categories, the Short Term Interest Rate
(STIR) and Long Term Interest Rate (LTIR). The former is mainly caused by the business
cycle and monetary policy. On the other hand, the fluctuations in the latter are challenging to
anticipate or influence by fiscal policy. Therefore, it is assumed that the Long Term Interest
Rate might more accurately describe the view of the economy in the long term (Humpe &
Macmillan, 2007).

11

Short Term Interest Rate


It is important to observe the effect in the short term as a number of investors would
consider positions in stock for speculative purposes only. The increase in interest rate would
increase transactions costs and thus decrease demand for stock. Also, the short term
borrowing costs of corporations may affect the future earnings prospects in an inverse
manner. For the purpose of the empirical examination, the yields from three months of U.S.
Treasury Bills are considered as Short Term Interest Rate. The effect is therefore presumed
to be negative.
Long Term Interest Rate
As an appropriate measure of the Long Term Interest Rate, the yield on the 10 year U.S.
government bond is considered. An increase would make the investment in equities less
attractive for risk-adverse investors, as government bonds are presumed to be less risky than
equities. The subsequent move towards bonds would firstly lower the spot price as an
increased number of sellers would out-sell the buyers on the market and secondly the
subsequent decrease in demand would lower the price further.
Furthermore, the increased borrowing costs of companies relying on long-term funding
would have the potential to reduce future investment into growth; and this would also reduce
the companies earnings and, through the PVM, the stock price. Therefore, the effect of
Long Term Interest Rate on stock valuation is assumed to be negative.
Commodities
Commodities, such as gold and oil are used as raw inputs in production of final goods and
play a vital role in economic activity of individuals and corporations. The fluctuations of
prices may then have the potential to affect the pricing model. The following section
presents the literature on the role of commodities and their effect on stock valuation:
Oil
The prices of oil appear to have significant effect on stock markets. Oil is one of the
fundamental drivers of modern economic activity. In addition, there is generally the belief
that stock markets react to the shocks in oil prices (Nandha & Fuff, 2007).
The price of oil may have effect on the stock markets through a number of direct and
indirect channels. The direct channels are the transfer of wealth from oil consumer to oil
producer countries, the use of oil as a raw input in industries and the use of final oil products
by general public. The influence on consumer confidence and the overall economic activity
may also be considered as an indirect channel (Nandha & Fuff, 2007).
12

The oil consuming countries would suffer from higher oil price as the costs of production of
goods and services would increase. The cost of production would also increase in the oil
producing countries, however, the profits of oil producers would be higher, as well as the
higher Gross Domestic Product of countries with nationalized oil producers. In addition, the
general public suffers when the petrol prices are high. Increased fuel cost results in decrease
in spending on other goods and services. Similar trends apply to businesses: high fuel prices
increase operational costs and consequently reduce earnings. Corporations may opt to pass
the rising production costs onto consumers; however, this might reduce demand further as
consumers must also face higher fuel costs.
The effect of higher oil prices is reflected in the increase of day to day expenses of general
public and rise in costs of production, which will in turn reduce demand of goods and shrink
corporate earnings. Therefore, based on the previous section, the effect of oil prices on stock
valuation is believed to be inverse.
Gold
Gold has long been an important metal for many countries. It plays a vital role as a store of
value during the times of economic uncertainties. In comparison to other precious metal
traded on the markets, gold displays an advantage and outstanding position (Zhang & Wei,
2010). Furthermore, gold can be considered a hybrid asset as it is used as a raw material in a
number of industries (Sujit & Kumar, 2011).
The movements in the price of gold are important for the economics and finance perspective.
In general, gold is accepted as a store of value during challenging economic times and as a
hedge against rising inflation. Also, in more recent years, investors in gold recognized
significant profit making abilities (Zhang & Wei, 2010).
Gold has experienced significant increase in value after the 2008 slump in the stock market.
Also, gold is largely uncorrelated with other asset types, mainly in the interconnected
financial markets, and perhaps due to the description by financial media as a safe haven
(Baur & Lucey, 2009).
The appreciation of gold prices may be caused the increased demand from investors fleeing
from plummeting stock markets. However, the influence of gold prices on stocks arises from
the aforementioned hybrid properties. Gold is continually used as a raw material in the
production of a number of goods and is one of the main merchandises in the jewellery
industry. The rise in price of raw inputs of production will naturally cause the cost of
production to increase, which in turn might decrease earnings. Also, the jewellery industry
might experience decrease in profits, which will affect the earnings and through the Present
13

Value Model the stock price. Therefore, the influence of gold on stock prices is believed to
be negative.
Inflation
Inflation is another factor that may influence the pricing of assets through expected and
unexpected changes in the price level. The uncertainty among investors about the level of
inflation may also influence the discount rate, and therefore the present value of future cash
flows (Humpe & Macmillan, 2007).
Inflation is considered an important factor in financial asset pricing models. It is generally
expressed as a percentage change in Consumer Price Index (CPI), which measures the
current prices of basket of several thousand goods and services routinely consumed by
households (Liow, Ibrahim & Huang, 2006).
An increase in inflation may elevate the nominal risk-free rate and therefore the discount
factor leading to falling stock price (Naik, 2013). In addition, the rising inflation has a
negative effect on corporate revenues due to the immediate increasing cost and slower
adjustment pace of output prices, thus reducing profits and in turn, the share price (Humpe &
Macmillan, 2007).
According to the previous paragraphs it may be assumed that inflation has positive as well as
negative influence on stock price. The increase in price level would raise the nominal values
of profits for companies; however, a similar increase in cost of production may deteriorate
potential profits and within the scope of the discount model, the numerator may not increase.
In addition, the potential of inflation to raise discount rate may have superior outcome and
thus decrease stock prices. Therefore, there is a confidence in the inverse relationship
between inflation and stock prices.
Exchange Rate
As the dependant variable used for the analysis is the price levels of Standard & Poors 500
Composite Index quoted in US dollars, it is natural to assume that fluctuations in the dollar
exchange rate will significantly affect the corporate earnings and thus the market price.
The importance of research of the linkages between the stock price and exchange rate
increased following the generalized floating in the major currencies in the early 1973, and
especially in recent years as the dollar demonstrated unparalleled volatility. Also, the
increase in the world trade and capital movements has made the currency value one of the
main determining factors affecting corporate profitability and equity valuation (Kim, 2003).

14

The effect of fluctuations in currency exchange rate can be observed via three price
channels. First is the domestic currency price channel which affects the domestic sales and
export revenues. Second is the foreign currency price channel, also influencing the domestic
sales and export revenues. The appreciation of the dollar causes the costs of imported goods
to decrease so the domestic price level tends to decrease. Consecutively, the U.S. exporters
are pressured to lower the dollar price of their goods to offset the rise of foreign currency
price. Thus, both channels are causing inverse relationship between earnings and dollar
movements. Third is the currency value channel. The variations in dollar exchange rate
directly affect the corporate costs of imported goods and intermediate production inputs. In
case of a strong dollar, the cost of imported goods decline and profits are likely to rise and
vice versa. Therefore, the relationship is positive. The three channels act simultaneously
fuelled by the fluctuations in the exchange rate; however, the positive effect of the currency
value channel is offset by the inverse effect caused by domestic and foreign currency
channels. In addition to the three trade channels, the portfolio adjustment effect will have a
positive effect on market prices. Investors may sell their foreign assets and purchase U.S.
assets when they assume the U.S. economy is stronger than other economies and thus create
demand for currency and securities which will cause the dollar to appreciate and the stock
price to move upwards. However, it is unlikely that the portfolio adjustment effect will have
major impact and also, it will be offset by the price channel (Kim, 2003).
This section provides evidence of the importance of the fluctuations in the dollar exchange
rate on the pricing of equities due to the magnitude of international trade, the flow of capital
and through the number of trade channels. Therefore, the assumption is that the dollar
exchange rate has a negative effect on stock prices.
Thus far, the discussion and research was concentrated on the evidence of the market price
of equities separating from the fundamental value and determining the appropriate model for
the price assessment together with micro and macro-economic variable affecting the model.
However, there are factors that may cause the separation of market price from fair stock
value and thus create a price bubble.

2.3. Factors influencing mispricing of securities


As debated, the literature offers arguments on the existence of price bubbles and the means
of computing a fair price of securities under consideration and in addition, the effect of
exogenous variables on the pricing model. In succession, the direction of the research should
encompass the factors with the potential to cause the separation of the market price from the
fundamental value. The discussion is presented in the following section.

15

The times of bubbles appear to overlap with the entry of inexperienced investors to the
market (Barlevy, 2007). At the time of Dutch Tulip Mania (1634 - 1637), a number of
middle class and moneyed working class investors started to speculate on the market for
tulips, which was previously a domain for investment specialists. Needless to say, the new
entrants were enthusiastic about their participation and thus optimistic and incautious.
During the late 1920s, the shift from short-term commercial bank notes to bonds and equity
financing, coupled with the general optimism from the previous decade of prosperity meant
that instead of institutions with considerable experience with valuating firms, the general
public became the main creditor to corporations. Yet again, as during the Tulip mania, new
participants were enthusiastic about the entry to the equity markets and prospects of
potential financial gains. The increase in a number of passionate investors drove demand and
prices upwards, coupled with the limited experience and knowledge; it is not surprising that
the price was pushed away from fundamentals. In the more recent episode from the 1980s,
the financial innovation and new investment opportunities drew inexperienced investors to
the markets from domestic and foreign regions alike (White, 2006).
Both Barlevy and White agree that throughout history, the inflow of inexperienced investors
to the markets may cause the increase in price above the fundamental value. More recently,
the globalization and fall of international boarders allowed for a virtually unlimited access to
the markets for foreign investors.

Shiller (2005) encompassed the behaviour at the

formation of a bubble - the success of investors is likely to attract public attention, which
spreads enthusiasm for the market. New, and often less sophisticated investors, enter the
market and freshly created demand bids up prices to touch new heights. The market surges
fuelled by forward looking, sophisticated theory with complete disregard for traditional
valuation methods. But eventually, prices may start to fall and pessimism might spread,
causing some to exit the market and subsequently for the market reach rock bottom.
Support of the arguments also comes from previous research by Smith, Suchanek and
Williams (1988). Authors conducted an experiment in which previously inexperienced
participants with limited knowledge about the functioning of the stock markets were fully
informed about the future dividends and thus had complete understanding of the
fundamentals so there was no reason for a bubble to arise. Nevertheless, during the course of
the experiment, the assets traded above the fundamental value and subsequently crashed
down as the asset reached end of life. This proposes that the true value of the asset was
calculated incorrectly, or the subjects speculated with the belief that other traders have
mispriced the value and can profit at the expense of their miscalculation. However, the
authors also found that once participants had some experience, the bubbles do not tend to

16

emerge, suggesting less experienced investors have higher tendency to behave irrationally
upon entering the market.
Authors suggest the creation of a bubble is linked with the inflow of enthusiastic,
inexperienced investors with limited knowledge about the functioning of the markets
,potentially using un-tested theory to price equities. Additionally, investors may be
expecting returns and growth unjustified by the underlying fundamental value and thus
behaving irrationally and potentially inflating an asset price bubble.

2.4. Summary of Literature Review


As Lansing (2007) stated, the valuation of future cash flows is less volatile than market
participants are anticipating and tends to grow in line with the average historical returns.
However, the analysis of market prices suggests that investors are pricing the equities with a
significant premium. The spread between the discounted future cash flows and market price
may be considered as an equity bubble.
The financial theory focuses on internal and widespread or systemic variables as a likely
source of increased volatility in the stock markets. Based on the research of Chen, Roll and
Ross (1986) and Humpe and Macmillan (2007), an econometric model can be constructed
including exogenous variables with direct or indirect effect on the endogenous variable - the
stock price. Further review of literature suggests the inclusion of following variables in the
regression model: Dividends per Share (Molodovsky, 1953 and Chen, Roll & Ross, 1986),
Earnings per Share (Campbell & Shiller, 1988 and Bhatt & Sumangala, 2012), Gross
Domestic Product (Liow, Ibrahim & Huang, 2006), Short Term Interest Rate (Maysami,
Howe & Hamzah, 2004 and Kim, 2003), Long Term Interest Rate (Udegbunam & Eriki,
2001 and Kim, 2003), price of Oil (Nandha & Fuff, 2007), price of Gold (Sujit & Kumar,
2011 and Zhang & Wei, 2010), Inflation (Humpe & Macmillan, 2007 and Naik, 2013) and
an Exchange Rate (Kim, 2003).
The discussion of the effects of Internal and External Factors on the asset pricing model
warrants a construction of an econometric model that would analyse the proposed
relationship between the variables and stock price. The results of the regression analysis then
may offer an explanation on how the individual factors are incorporated into the share
valuation framework and their respective effect on stock price. The findings and
interpretations are presented in the Empirical Results section of this paper

17

3. Data and Methodology


The following section will discuss the data collection and description, hypothesis
formulation, the method used to calculate the proposed econometric model, and will address
issues related with the use of Ordinary Least Squares.

3.1. Data Description


The analysis of the data was carried out with the use of secondary source material. The
prices of Standards & Poors 500 Composite Index, Dividends per Share and Earnings per
Share were obtained from file compiled by Shiller1. Data for Short Term Interest Rate2,
Long Term Interest Rate3, Consumer Price Index4, from which Inflation is calculated, and
Gross Domestic Product5 are from Organisation for Economic Co-ordination and
Development (OECD) iLibrary. The data for the remaining variables were obtained from
Thompson Reuters DataStream service.
The sample used for the proposed analysis contains 170 monthly observations from the time
period 1998.11 to 2012.12. The time period was chosen to capture the two asset price
bubbles bursts as described by Shiller; and also significant market shock imposed by
terrorist attack on World Trade Centre (11.9.2001).
The Dividends per Share, Earnings per Share and Gross Domestic Product are reported and
available on the quarterly basis, but for the purpose of the study were linearly interpolated
for monthly data as suggested by Shiller6. Remaining variables were obtained from their
respective sources as monthly observations.
The data is compiled in time series. This means that the observations of the values of
individual variables are collected at different times. The time series data are then denoted by
subscript t.

Available online at: http://www.econ.yale.edu/~shiller/data.htm


Available online at: http://stats.oecd.org/index.aspx?queryid=86
3
Available online at: http://stats.oecd.org/index.aspx?queryid=86
4
Available online at: http://stats.oecd.org/index.aspx?querytype=view&queryname=221
5
Available online at: http://stats.oecd.org/index.aspx?querytype=view&queryname=221
6
Available online at: http://www.econ.yale.edu/~shiller/data.htm
2

18

3.2. Empirical Model


The Literature Review section suggests inclusion of Internal factors: Dividends and
Earnings, and External factors: Gross Domestic Product, Short Term and Long Term Interest
Rate, price of Oil and price of Gold, Inflation and Exchange Rate. Therefore, the proposed
empirical model suggests the share price is a function of the following variables:
Equation 3: Regression Equation

Where:
SP:

S&P 500 Composite Index

DPS:

Dividends per Share

EPS:

Earnings per Share

GDP: Gross Domestic Product


STIR: Short Term Interest Rate
LTIR: Long Term Interest Rate
Gold: price of Gold
WTI:

price of West Texas Intermediate Light Sweet Crude Oil (Oil)

INF:

Consumer Price Index expressed by Inflation

FX:

Exchange Rate (USD/EUR)

3.3. Ordinary Least Squares


The Ordinary Least Squares (OLS) regression analysis is employed to estimate a functional
relationship between the dependant variable and the independent variables in the United
States financial market. The goal is to find the amount of variation in the dependent variable
explained by the variations in the independent variables and to determine their significance.
For the OLS estimators to be considered the best, a number of classical assumptions must be
met.
The Classical assumptions:
1. The regression model is linear, it is correctly specified, and has an additive error
term
2. The population mean of the error term is zero
3. There is no correlation between the error term and explanatory variables
4. There is no serial correlation between the error term observations
19

5. There is constant variance in the error term


6. No explanatory variable is a perfect linear function of any other explanatory variable
7. The error term is normally distributed
If any of the Classical assumptions or a combination of assumptions is violated, alternative
regression method might be better to estimate the proposed relationship (Studenmund, 2010,
p. 93-4). Therefore, the following sections will test whether the Classical assumption will
hold for the equation proposed.
3.3.1. Spurious Correlation and Stationarity
The data for the regression analysis are organized into time series. Such data can exhibit a
problem when the underlying trend in explanatory variable is the same a trend in dependant
variable. This would cause the explanatory variable to appear more significant than it
actually is (Studenmund, 2010, p. 417-18).
Similarity in trend between dependant and independent variables may out-weight any causal
relationship and cause variables to appear correlated, even if they arent. This problem is an
example of spurious correlation. A spurious correlation between dependant variable and one
or more explanatory variables, with a strong correlation caused by similar trend rather than
by real causal relationship, would cause spurious regression results, which are considered
overstated and untrustworthy (Studenmund, 2010, p. 417-18).
Spurious correlation may be caused by non-stationary time series. Such series would exhibit
basic property changes, for example changes in mean or variance in the sample over time.
Studenmund (2010, p. 421) suggest visual examination of the data to recognize if the mean
of the sample is changing dramatically over time indicating non-stationarity. The data for
individual variables are graphed in Figure 1.
Also, the Dickey-Fuller Test can be employed to determine the presence unit root,
suggesting a non-stationarity in the dataset. The test examines the hypothesis that the
variable has a unit root, and the regression would benefit from using the first differences,
and then substituting them into the proposed equation (Studenmund, 2010, p. 424). The
presence of unit root (

) would indicate a non-stationarity, and thus the hypothesis

presents itself as follows:


Null Hypothesis

Alternative Hypothesis

= 0 and significant bellow 95 per cent


:

< 0 and significant above 95 per cent

The results and interpretations are organized in Table 2 in Empirical Results Chapter.
20

3.3.2. Multicollinearity
Due to the use of multiple variables, the theoretical model must be tested for
multicollinearity. The presence of perfect multicollinearity in the model is a violation of 6 th
Classical assumption of the OLS regression. In contrast, presence of severe imperfect
multicollinearity does not violate the 6th assumption; however, it might cause difficulties in
the regression (Studenmund, 2010, p. 247)
If two variables are significantly related, it might be difficult to distinguish the effects of one
variable from the effects of another. Also, accurately estimating the coefficients can become
problematic. Nevertheless, if variables are only vaguely related, the estimation of
coefficients is accurate for most purposes (Studenmund, 2010, p. 248).
In order to detect the presence of multicollinearity, correlation coefficients between variables
must be assessed. The coefficients are presented in Table 4 in Empirical Results Chapter.
Severe multicollinearity can be indicated by a correlation coefficient greater than 0.8
(Cassells, 2013). No correlation coefficient suggests severe multicollinearity. The more
formal test is the Variance Inflation Factor (VIF). Any VIF score greater than 5 would
indicate a presence of multicollinearity (Cassells, 2013). The results from VIF test are
presented in Table 5 in Empirical Results Chapter.
3.3.3. Serial Correlation
Serial correlation is a violation of 4th classical assumption that there is no correlation
between error terms. The presence of autocorrelation in OLS model would not cause bias in
estimated coefficients; however, it would mean the linear unbiased estimators have no
longer minimum variance, and also, the estimators of standard error to be biased, resulting in
unreliable hypothesis outcomes (Studenmund, 2010, p. 304-14).
Serial correlation can be detected using the Durbin-Watson d Test. The test is used to
conclude whether or not there is a first-order serial correlation in the error term, and thus if
the model suffers from autocorrelation (Studenmund, 2010, p. 315-16). The test specifies a d
value on which the hypothesis decision rule is based on, and therefore:
Null hypothesis

: no serial correlation

Alternative Hypothesis

: serial correlation

The decision rule is:


If d <

: reject

21

If d >
If

: do not reject

: inconclusive

The critical values for

and

could be found in the Durbin Watson Test Statistical Tables.

The findings are presented in Table 6 in the Empirical Results Chapter.


3.3.4. Heteroskedasticity
Heteroskedasticity is a violation of 5th Classical assumption in OLS, for the assumption to
hold, the error terms have to have a constant variance. Even though the heteroskedasticity is
mostly common in cross-sectional data, the time-series data may exhibit the signs also. This
is especially important for the studies of financial markets using time-series (Studenmund,
2010, p. 339-61).
Heteroskedasticity does not cause the linear estimators to be biased; however, it causes bias
in the standard error estimator in OLS, resulting in unreliable hypothesis testing. Using the
STATA software for regression analysis offers a Breusch Pagan Test for heteroskedasticity.
The Breusch Pagan Test is as follows:
Null hypothesis

: Homoskedastic error term

Alternative Hypothesis

: Heteroskedastic error term

The decision rule is:


If Chi^2

: reject

If Chi^2 <

: do not reject

The findings from Breusch-Pagan Test for heteroskedasticity are organized in Table 7 of the
Empirical Results Chapter.
3.3.5. Specification
In order for equation to be estimated, it must be correctly specified; in particular, the
selection of explanatory variables must be advocated by the appropriate literature. The
review of literature suggests the use of aforementioned variables, and therefore there is a
belief that the model is correctly specified.
However, the proposed model should be tested for omitted variables. Omitting an important
variable from a model may cause the coefficients to be biased or no longer minimum
variance, or both. One of the popularly used tests is the Ramseys Specification Error Test
(RESET). The test measures whether or not the inclusion of generated variables is
22

improving the fit of the overall model. The results from the RESET test are organised in
Table 8 in Empirical Results Chapter.
Furthermore, the model should also be tested for the presence of irrelevant variables. Such
variable does not cause bias, but it does increase the variances in other estimated
coefficients. An irrelevant variable would present itself as insignificant and would decrease
the adjusted coefficient of determination. To verify the variable to be relevant, the following
conditions must hold;
1. The theory proposing the variable under consideration is unambiguous.
2. The variables coefficient is significant and in expected direction.
3. The adjusted coefficient of determination increases after a variable is added to the
equation.
4. The other variables coefficients and significance improve after adding a variable
into equation.
If all conditions hold, the variable belongs to the equation and thus cannot be removed.
However, if none of the conditions hold, the variable can safely be excluded from the
equation (Studenmund, 2010, p. 176-8). The results are summarized in Irrelevant Variable
Test Results part (p.31) in the Empirical Results section.

23

3.4. Hypothesis Formulation


Based on the literature reviewed and the model proposed, the hypotheses are formulated and
will be tested at 95 per cent significance level as follows:
Table 1: Hypothesis Formulation

Dividends per Share


Significant and positive relationship between the Stock Prices and Dividends per Share
Null Hypothesis

Alternative Hypothesis

0
0

0
>0

Tested at 95 per cent significance level


Earnings per Share
Significant and positive relationship between the Stock Prices and Earnings per Share
Null Hypothesis

Alternative Hypothesis

1
1

0
>0

Tested at 95 per cent significance level


Gross Domestic Product
Significant and positive relationship between the Stock Prices and Gross Domestic Product
Null Hypothesis

Alternative Hypothesis

2
2

0
>0

Tested at 95 per cent significance level


Short Term Interest Rate
Significant and negative relationship between the Stock Prices and Short Term Interest Rate
Null Hypothesis

Alternative Hypothesis

3
3

0
<0

Tested at 95 per cent significance level


Long Term Interest Rate
Significant and negative relationship between the Stock Prices and Long Term Interest Rate
Null Hypothesis

Alternative Hypothesis

4
4

0
<0

Tested at 95 per cent significance level

24

Gold
Significant and negative relationship between the Stock Prices and Gold
Null Hypothesis

Alternative Hypothesis

5
5

0
<0

Tested at 95 per cent significance level


West Texas Intermediate [Oil]
Significant and negative relationship between the Stock Prices and Oil
Null Hypothesis

Alternative Hypothesis

6
6

0
<0

Tested at 95 per cent significance level


Inflation
Significant and negative relationship between the Stock Prices and Inflation
Null Hypothesis

Alternative Hypothesis

7
7

0
<0

Tested at 95 per cent significance level


Exchange Rate
Significant and negative relationship between the Stock Prices and Exchange Rate
Null Hypothesis

Alternative Hypothesis

8
8

0
<0

Tested at 95 per cent significance level

25

4. Empirical Results
The following chapter will provide an analysis of the proposed regression model and will
address the econometric testing issues described in the Data and Methodology section.
Furthermore, the last part will offer an interpretation of empirical findings.

4.1. Stationarity Test Results


As Studenmund (2010, p. 417-18) suggests, initially a simple visual observation of the
graphed datasets is required. All variables suggested by the review of literature are graphed
in Figure 1.
Figure 1: Dependant and Independent variables
S_P_500
1600
1500
1400
1300
1200
1100
1000
900
800
700

Dividends
32
30
28
26
24
22
20
18
16
14

1999
2002
2005
2008
2011

19992002200520082011

STIR
7
6
5
4
3
2
1
0

5
4
3
2
1
19992002200520082011

INF
1.5
1
0.5
0
-0.5
-1
-1.5
-2

19992002200520082011

19992002200520082011

Gold
2000
1800
1600
1400
1200
1000
800
600
400
200

GDP_Growth
2
1.5
1
0.5
0
-0.5
-1
-1.5
-2
-2.5

LTIR
7

19992002200520082011

Earnings
90
80
70
60
50
40
30
20
10
0

WTI
140
120
100
80
60
40
20
0

1999
2002
2005
2008
2011

19992002200520082011

FX
1.2
1.1
1
0.9
0.8
0.7
0.6

19992002200520082011

19992002200520082011

A trend is noticeable from all of the data with the exception of Inflation (INF). To confirm
the visual observation assumption, a Dickey Fuller test was performed to determine the
presence of unit root in data set, which indicates a stationarity issue. The results are
organized in Table 2.

26

Table 2: Dickey Fuller Test

Dickey Fuller Test


Variable

p - value

Result

S&P 500 Index

-0.0304

0.8243

Cannot reject

Dividends

-0.0016

0.9922

Cannot reject

Earnings

-0.0085

0.9635

Cannot reject

Gross Domestic Product

-0.0908

0.168

Cannot reject

Short Term Interest Rate

-0.0128

0.9356

Cannot reject

Long Term Interest Rate

-0.0844

0.187

Cannot reject

Gold

-0.0475

0.5013

Cannot reject

Oil

-0.0715

0.3614

Cannot reject

Inflation

-0.5261

0.000

Can reject

Exchange Rate

-0.0381

0.5636

Cannot reject

All but Inflation data are indicating a presence of unit root and thus non-stationarity.
Studenmund (2010, p. 424).
To address the non-stationarity issue, Chen, Roll and Ross (1986) and Humpe and
Macmillan (2007) are using the differences in natural logarithms (ln) between
observations

This technique actually measures the

logarithmic rate of change between time periods in a variable. Also, it has the equivalent
effect as the first differences described previously.
Therefore, to avoid the potentially spurious regression results as a consequence of nonstationary time series data, the further research will use the differences in natural logarithms
between observations for the following variables: S&P 500, Dividends, Earnings, Gold,
Short Term Interest Rate, West Texas Intermediate and Exchange Rate. The Dickey-Fuller
Test indicates a non-stationarity issue in GDP variable; however, the data for GDP were
obtained from source as growth in GDP and thus already expressed as log return. Therefore,

27

the variable will be used in the model with no adjustments. Also, a very similar result to that
of GDP was observed in LTIR and thus no action will be taken with regards to this data. The
actions undertaken are summarized in Table 3
Table 3: Methods used for calculating variables

Variable

Notation

Method

S&P 500 Index


Dividends
Earnings
No change to dataset

Gross Domestic Product


Short Term Interest Rate

No change to dataset

Long Term Interest Rate


Gold
Oil
Inflation
Exchange Rate

After adjusting the data for the potential presence of non-stationarity, the econometric model
is expressed as follows:
Equation 4: Regression Equation (Detailed)

lnSPt = +

0lnDPSt

1lnEPSt

6lnWTIt

Where: is the constant term,

2lnGDPt

7INFt

3lnSTIRt

4LTIRt

5lnGoldt

+ 8lnFXt + t

are coefficients for their respective variables and t

indicates the stochastic error term.

4.2. Multicollinearity Test Results


Cassells (2013) suggests that correlation coefficient between any variables greater than 0.8
may indicate a multicollinearity issue. The correlation matrix is organized in Table 4.

28

Table 4: Correlation Matrix

Correlation Matrix
Variable

lnSP

lnDPS

lnEPS

lnGDP

lnSTIR

LTIR

lnGold

lnWTI

INF

lnDPS

0.01

1.00

lnEPS

0.32

-0.14

1.00

lnGDP

0.29

0.35

0.51

1.00

lnSTIR

-0.13

0.34

0.18

0.42

1.00

LTIR

-0.04

-0.16

0.03

0.29

0.15

1.00

lnGold

0.09

0.02

0.4

0.01

-0.18

-0.05

1.00

lnWTI

0.27

0.02

0.27

0.26

0.07

0.12

0.14

1.00

INF

0.11

0.05

0.18

0.17

0.06

0.13

0.09

0.58

1.00

lnFX

-0.19

-0.05

-0.06

0.01

0.26

0.06

-0.19

-0.21

-0.19

lnFX

1.00

For full Correlation Matrix see Appendix 1

The correlation coefficients between individual variables are all below the suggested value
of 0.8 and to confirm the assumption of no severe multicollinearity in the regression model,
a Variance Inflation Factor test was performed. The results are organized in Table 5.
Table 5: Variance Inflation Factors

Variance Inflation Factors


Variable

VIF

lnDPS

1.80

lnEPS

1.89

lnGDP

2.42

lnSTIR

1.52

LTIR

1.37

lnGold

1.08

lnWTI

1.65

INF

1.55

lnFX

1.20

For full VIF results see Appendix 1

The Variance Inflation Factor test confirms the results from the correlation matrix and
therefore suggests there is no severe multicollinearity issue in the proposed theoretical
model.

29

4.3. Serial Correlation Test Results


The Durbin Watson d Test was employed to determine the presence of autocorrelation. The
results are summarized in Table 6:
Table 6: Durbin Watson d Test

Durbin Watson d Test


d - statistic (9,169)

1.8497

Prob > chi^2

0.000

1.66302

1.83523

For full Durbin Watson d Test results see Appendix 1

The computed d value satisfies the test requirement 1.8497 = d >


the

= 1.83523 and therefore

cannot be rejected, suggesting there is no serial correlation in the proposed model.

4.4. Heteroskedasticity Test Results


The Breusch-Pagan test for heteroskedasticity was performed using STATA software. The
criteria for rejection of

are described in methodology section. Results are in Table 7.

Table 7: Breusch-Pagan test for heteroskedasticity

Breusch-Pagan Test
33.98
9

189.424

(9,159)

For full Breusch-Pagan test for heteroskedasticity see Appendix 1

The null hypothesis states that the error term is homoskedastic, the criteria for rejection are
not met; Chi^2 = 33.98 <

= 189.424, therefore, do not reject

However;

this does not prove the term to be homoskedastic.

4.5. Specification Test Results


The assumption based on the literature review is that the model is correctly specified;
however, the regression equation was tested for omitted variables using the Ramseys
Specification Error Test (RESET). The results are organized in Table 8. Furthermore, as
suggested by Studenmund (2010, p. 176-8), the model was also tested for irrelevant
variables. The outcomes and interpretations are presented in the following part of this
section.
7

Critical value for obtained from: http://www.stanford.edu/~clint/bench/dw05b.htm


Critical value for
obtained from: http://www.stanford.edu/~clint/bench/dw05b.htm
9
obtained from: http://www.medcalc.org/manual/chi-square-table.php
8

30

4.5.1. Ramseys Specification Error Test Results


Table 8: Ramseys Specification Error Test

Ramseys Specification Error Test


F(3,158)

4.51

p- value

0.046

For full RESET test results see Appendix 1

The Ramseys Specification Error Test results suggest that there is a specification error
present in the model. The test doesnt specify the omitted variables bias, it merely states that
there is an error.
4.5.2. Irrelevant Variable Test Results
As described in the Methodology section, an irrelevant variable would not hold for any of
the conditions defined. The results from regression analysis of the proposed model with all
suggested explanatory variables, are organized in Table 9:
Table 9: Regression Results (1)

R^2

0.2439

Adjusted R^2

0.2011

Regression Analysis
Variable

Coefficient

p-value

lnDPS

-0.0981858

0.815

lnEPS

0.0645356

0.103

lnGDP

2.153689

0.003

lnSTIR

-0.0827134

0.004

LTIR

-0.0042864

0.140

lnGold

-0.0015464

0.979

lnWTI

0.0826085

0.047

INF

-0.0055852

0.526

lnFX

-0.1396791

0.262

_constant

0.0065781

0.582

For full regression results see Appendix 1

Regression results in Table 9 indicate the coefficient for lnGold is in the expected direction.
However, it is minor and vastly insignificant and therefore second condition does not hold.
Consequently, gold, as an external factor affecting prices of equities must be confirmed to be
a relevant variable for the proposed model and as such, must hold for all remaining
conditions previously defined.

31

As noted in the Literature Review section, gold is a hybrid asset, used both as a raw material
and as a safe haven in the event of a market collapse (Sujit & Kumar, 2011). This suggests
the theory is not ambiguous and that the second condition does not hold. For the assessment
of outstanding variables, regression analysis excluding lnGold need to be performed. The
results are presented in Table 10. The empirical results, regarding the adjusted coefficient of
determination (increases after removal of variable) and remaining variables (all but one
significance level improve and none disprove) suggest that none of the remaining conditions
for relevant variable hold and therefore lnGold can be safely excluded from the model as
irrelevant.
Table 10: Regression Results (2)

R^2

0.2439

Adjusted R^2

0.2061

Regression Analysis
Variable

Coefficient

p-value

lnDPS

-0.0987925

0.813

lnEPS

0.0645087

0.102

lnGDP

2.153256

0.003

lnSTIR

-0.0825897

0.003

LTIR

-0.0042845

0.139

lnWTI

0.0825306

0.046

INF

-0.0055893

0.524

lnFX

-0.1393393

0.259

_constant

0.0065632

0.581

For full regression results see Appendix 2

The regression results from Table 9 and Table 10 also highlight lnDPS as highly
insignificant; the coefficient is in opposite direction as previously hypothesized and thus
second condition does not hold. Therefore the variable must be tested for relevance in the
model. The results from the regression analysis, excluding the variable lnDPS, are presented
in Table 11.

32

Table 11: Regression Results (3)

R^2

0.2436

Adjusted R^2

0.2108

Regression Analysis
Variable

Coefficient

p-value

lnEPS

.0690406

0.045

lnGDP

2.068081

0.001

lnSTIR

-.0845604

0.001

LTIR

-.0040015

0.128

lnWTI

.0827009

0.045

INF

-.0057282

0.512

lnFX

-.1344142

0.268

_constant

0.0065632

0.617

For full regression results see Appendix 3

Following the regression analysis excluding the lnDPS variable, all but one variables
coefficients improve as well as the adjusted coefficient of determination, therefore another
two conditions of relevance do not hold. Examining the unambiguity of the literature
presents a difficulty. As stated by Molodovsky (1953), dividends are a hard core of stock
valuation. The significance of dividend payments in the assessment of present value of
assets is one of the pillars of economic and finance. The empirical evidence, however,
suggests that the dividends are an irrelevant variable with insignificant effect on stock price.
The findings suggest a further interpretation of result, and thus the discussion is offered later
in this section.

4.6. Regression Analysis Results


The regression analysis was performed using the STATA 10 statistical software. Previously
described econometric issues were rectified and thus variables with the best fit for the
regression model based on the literature review are: Earnings per Share, Gross Domestic
Product, Short and Long Term Interest Rate, price of Oil, Inflation and Exchange Rate. The
result of the regression analysis is presented in Table 10 and specified in the following
equation:
Equation 5: Estimated Regression Equation

= 0.006 + 0.069 lnEPSt + 2.07 lnGDPt 0.08 lnSTIRt 0.004 LTIRt +


0.08 lnWTIt 0.005 INFt 0.134lnFXt
As all datasets representing variables included in the regression model measure the
logarithmic rate of change, expressed as percentage, between time periods of their respective
33

observations, the variables coefficients therefore characterize the percentage impact on


endogenous variable of one per cent change in the exogenous variable. All but one
regression coefficients (lnWTI) are in the expected direction. The variables lnEPS, lnGDP
and lnSTIR are significant at 95 percent confidence level and thus

can be rejected. The

lnWTI is significant at 95 per cent level; however, the coefficient is positive and therefore
cannot be rejected. The remaining variables LTIR, INF and lnFX are not significant at
the chosen significant level and therefore once again

cannot be rejected.

4.7. Interpretation of Regression Analysis


The Ramseys Specification Error Test results suggest that there is a specification error in
the equation. This outcome was somewhat expected, as the fluctuations in a market
including approximately 500 publicly traded companies are extremely difficult to anticipate.
Previous academic research included the following additional economic variables in the
respective regression models: Consumption (Chen, Roll & Ross, 1986), Industrial
Production (Chen, Roll & Ross, 1986, Humpe & Macmillan, 2007 and Maysami, Howe &
Hamzah, 2004) and Money Supply (Liow, Ibrahim & Huang, 2006, Humpe & Macmillan,
2007 and Maysami, Howe & Hamzah, 2004). Inclusion of additional variables in the
regression model may improve the specification of the estimated equation. Also, additional
explanatory variables may increase the adjusted coefficient of determination and thus
improve the fit of the estimated linear model.
The significance of lnWTI variable in the equation is acknowledged; however, the direction
of the effect is in the opposite direction as hypothesised. The result is colluding with the
research by Nandha and Fuff (2007), where authors found a significant and negative
relationship between stock and oil prices. The positive regression coefficient may be caused
by the adjustment of the dataset to the logarithmic rate of change attributable to the nonstationarity of the sample, or to the time period and frequency of observations chosen for the
examination.
The exclusion of lnGold from the estimated model due to irrelevance may possibly be
attributed to the unambiguity of the literature. Gold is a hybrid asset (Sujit & Kumar, 2011)
and based on the empirical results, the effect on stock prices is indicated to be marginal.
Also, the golds safe haven properties (Baur & Lucey, 2009) may attract investors in case of
a market plummet, suggesting that fluctuation in the gold price is a result of market
fluctuation rather than a cause.
The irrelevance of the lnDPS variable in the estimated equation is a significant finding. The
theory suggests that the dividends are an essential component of share valuation framework.
34

However, empirical evidence indicates that the variable is irrelevant and has no effect on
share price, conflicting with the assumptions of White (2006) and Molodovsky (1953), but
supporting the statements of Lansing (2007), who argues that fundamental principles of
share valuation through discounting of streams of future payments are disrupted, and
investors are expecting greater volatility in the future dividends than warranted by the
historical fluctuations.
Furthermore, to support the finding, a standard deviation representing the volatility in the
variables over the examined time period can be calculated to offer further insight into the
discussion. The standard deviations of S&P 500 and Dividend prices are graphed in Figure
2.
Figure 2: S&P 500 and Dividends Volatility
S&P 500 price and Dividends Volatility
0.25

0.25

S&P Volatility (right)


Dividends Volatility (left)

Volatility

0.2

0.2

0.15

0.15

0.1

0.1

0.05

0.05

0
2000

2002

Note: Standard Deviation calculated as:

2004

2006

2008

2010

2012

, where

As can be observed in Figure 2, the volatility of S&P 500 is considerably greater than the
volatility in the Dividends realized from the index agreeing with the argument by Lansing
(2007). Furthermore, the finding also concurs with Shiller (2005), who assumed that the
equity bubbles may be formed when traditional share valuation is abandoned and investors
may be embracing unprecedented pricing theory, and thus possibly behaving irrationally.
Additionally, the effect of variables significant at 95 per cent level is only marginal. A one
hundred per cent increase in Earnings per Share would only account for 0.069 per cent
increase in the price according to the estimated model. Similarly, a one per cent increase in
Short Term Interest Rate would only explain 0.0002 per cent decrease in price. The variable
35

with the highest coefficient was Gross Domestic Product: the scale of percentage movement
in the variable would be more than doubled (2.07) in the movement of price. While
acknowledging the relatively low adjusted coefficient of determination (0.2108), the
empirical evidence suggests that the effect of exogenous variables on stock price, with the
exception of GDP, can only be described as minor.
The empirical evidence supported by the arguments of Lansing (2007) and Shiller (2005)
suggests an assumption that the market may be in a bubble territory when the market price
volatility exceeds the volatility in the dividends realized from investment in the stock.
Support of the assumption comes from West (1988), who concluded his research in
Dividends Innovations and Stock Price Volatility by the following statement:
The test indicates that stock prices are too volatile to be the expected present
discounted value of dividends, with a constant rate. Possible explanations for the
test results include that expectations are not rational, that discount rates vary and
that there are speculative bubbles. The econometric diagnostics and the informal
analysis were notably more consistent with the bubble explanation than with the
other two. (p.29)
It is important to note that the irrelevance of dividends in the regression equation may be
attributed to the time period or to the frequency of observations chosen for the econometric
analysis. However, the empirical results, together with supportive statements from academic
researchers (West, 1988 and Lansing, 2007) and recent Nobel Prize laureate (Shiller, 2005),
reinforce the assumption, thus the prices of equities traded on stock exchanges may be
mispriced when the volatility of the market price significantly exceeds the volatility of the
dividends realized from an investment into the underlying security and in such event,
investors are advised to be cautious when considering investment opportunities.

36

5. Conclusion
The prices of equities traded on stock exchanges fluctuate every day without direct
implications on day to day life; however, if the size of an increase is large and prolonged
enough, the historically observed tendency to revert back to the fundamental value can have
widespread consequences. Two such episodes, the dot com and property bubble bursts, had
the combined strength to wipe out approximately $ 11 trillion of investors wealth.
The research project attempted to answer the question of whether or not the analysis of
fundamental valuation within the share valuation framework, based on the research by Chen,
Roll and Ross (1986) and Humpe and Macmillan (2007), has the potential to indicate an
asset price bubble in the S&P 500 Composite Index. Further research of academic literature
identified micro-economic and macro-economic exogenous variables with potential to
directly or indirectly influence the share price, and warranted a construction of an
econometric model to analyse the relationship.
Additionally, academic researchers (Shiller, 2005 and Lansing, 2007) presented arguments
that the traditional share valuation is disrupted, and that at least some investors might be
speculating, or behaving irrationally and thus inflating a price bubble.
The time period 1998.12 to 2012.12 was chosen for the econometric analysis to capture the
two most significant market corrections - the dot com bubble burst and the property bubble
burst - together with an outside market shock caused by a terrorist attack on the World Trade
Centre. The Ordinary Least Squares regression method was adopted to analyse the proposed
linear relationship between the exogenous variables and endogenous variable, the stock
price.
The empirical evidence from regression analysis suggests the influence of selected
exogenous variables on stock price is only marginal (with the exception of Gross Domestic
Product), and the presence of gold and mostly dividends in the estimated regression equation
is irrelevant.
The exclusion of dividends from the regression equation implied that, in the time period
examined, at least some investors were not respecting the fundamental value of the
underlying securities and either speculated or behaved irrationally. Further investigation into
the association between dividends and share prices yielded an assumption supported by
leading academic researchers West (1988), Lansing (2007) and Shiller (2005) that the prices
of equities traded on the stock exchanges may be in a bubble territory when the volatility of
spot price significantly exceeds the volatility of the dividends realized from a position in the
security and investors should be cautious when assessing investment opportunities.
37

The author recognizes limitations of the research and recommends the following suggestions
for further investigation. The RESET test indicated that the model would benefit from an
inclusion of additional variables such as Consumption, Industrial Production or Money
Supply; additionally, adopting more advanced regression model to analyse the suggested
linear relationship, such as GARCH 1.1 (which does not assume the volatility over time to
be unconditional), would offer ability to extend the time period and perhaps improve the
results. Furthermore, future research may further examine the volatility assumption and
possibly allow for a recognition of numerical value representing an asset price bubble.

38

6. References
Ballew, P. D. (2011, August 4). Housing Bubble Burst: More than Just Foreclosures. CNBC.
Retrieved from http://www.cnbc.com/id/44022473
Barlevy, G. (2007). Economic theory and asset bubbles. Economic Perspectives, (Q III), 4459.
Baur, D. G., & Lucey, B. M. (2010). Is gold a hedge or a safe haven? An analysis of stocks,
bonds and gold. Financial Review, 45(2), 217-229.
Campbell, J. Y., & Shiller, R. J. (1988). Stock prices, earnings, and expected dividends. The
Journal of Finance, 43(3), 661-676.
Campbell, J. Y., & Shiller, R. J. (2001). Valuation ratios and the long-run stock market
outlook: An update (No. w8221). National bureau of economic research.
Cassells, D. (2013). Financial Econometrics. Lecture notes in the topic Multicollinearity
(DT399/3) Dublin, Dublin Institute of Technology.
Chakraborty, I., Goldstein, I., & MacKinlay, A. (2013). Do Asset Price Bubbles have
Negative Real Effects? Retrieved March 8, 2014 from
http://www.wharton.upenn.edu/jacobslevycenter/files/01.13.Goldstein.pdf
Chen, N. F., Roll, R., & Ross, S. A. (1986). Economic forces and the stock market. Journal
of business, 59(3), 383.
Farrell Jr, J. L. (1985). The dividend discount model: A primer. Financial Analysts Journal,
16-25.
Gordon, M. J., & Shapiro, E. (1956). Capital equipment analysis: the required rate of
profit. Management Science, 3(1), 102-110.
Humpe, A., & Macmillan, P. (2007). Can macroeconomic variables explain long term stock
market movements. A comparison of the US and Japan. University-School of
Economics and Management CDMA Working paper, (07/20).
Kim, K. H. (2003). Dollar exchange rate and stock price: evidence from multivariate
cointegration and error correction model. Review of Financial Economics, 12(3),
301-313.
Lansing, K. J. (2007). Asset price bubbles. FRBSF Economic Letter, 32, 25-29.

39

Liow, K. H., Ibrahim, M. F., & Huang, Q. (2006). Macroeconomic risk influences on the
property stock market. Journal of Property Investment & Finance, 24(4), 295-323.
Maysami, R. C., Lee, C. H., & Hamzah, M. A. (2005). Relationship between
macroeconomic variables and stock market indices: cointegration evidence from
stock exchange of Singapores all-S sector indices. Jurnal Pengurusan, 24, 47-77.
Molodovsky, N. (1953). A theory of price-earnings ratios. The Analysts Journal, 65-80.
Naik, P. K. (2013). Does Stock Market Respond to Economic Fundamentals? Time-series
Analysis from Indian Data. Journal of Applied Economics & Business
Research, 3(1).
Nandha, M., & Faff, R. (2008). Does oil move equity prices? A global view. Energy
Economics, 30(3), 986-997.
Pushpa Bhatt, P., & Sumangala, J. K. (2012). Impact of Earnings per share on Market Value
of an equity share: An Empirical study in Indian Capital Market. Journal of
Finance, Accounting & Management, 3(2).
Smidt, E. T. (2005). Asset Price Bubbles. Denmark: University of Copenhagen. Retrieved
from: http://www.econ.ku.dk/sos/Old_files/Asset%20Price%20Bubbles.pdf
Shiller, R.J. (2000). Irrational Exuberance. New Jersey: Princeton University Press.
Shiller, R. J. (2005). Irrational Exuberance. New York: Broadway Books.
Smith, V. L., Suchanek, G. L., & Williams, A. W. (1988). Bubbles, crashes, and endogenous
expectations in experimental spot asset markets. Econometrica: Journal of the
Econometric Society, 1119-1151.
Studenmund, A. H. (2010). Using Econometrics: A Practical Guide (6th ed.). Boston:
Pearson
Sujit, K. S., & Kumar, B. R. (2011). &Study on Dynamic Relationship Among Gold Price,
Oil Price, Exchange Rate and Stock Market Returns'. International Journal of
Applied Business and Economic Research, 9(2), 145-165.
Udegbunam, R. I., & Eriki, P. O. (2001). Inflation and stock price behavior: Evidence from
Nigerian stock market. Journal of Financial Management & Analysis, 20(14), 1.
West, K. D. (1988). Dividend Innovations and Stock Price Volatility (No. 1833). National
Bureau of Economic Research, Inc.
40

White, E. N. (2006). Bubbles and Busts: The 1990s in the Mirror of the 1920s (No.
w12138). National Bureau of Economic Research.
Zhang, Y. J., & Wei, Y. M. (2010). The crude oil market and the gold market: Evidence for
cointegration, causality and price discovery. Resources Policy,35(3), 168-177

41

Appendix 1
Summary of Variables
. summarize

lnsp lndps lneps lngdp lnstir ltir lngold lnwti inf lnfx;

Variable |
Obs
Mean
Std. Dev.
Min
Max
-------------+-------------------------------------------------------lnsp |
169
.0012862
.0412615 -.2280448
.1135225
lndps |
169
.0038937
.0091245 -.0232969
.0185273
lneps |
169
.0049439
.0993652
-.52828
.7038552
lngdp |
169
.0051776
.0062651 -.0215171
.0188836
lnstir |
169
-.0181221
.1252972 -.6045938
.4054651
-------------+-------------------------------------------------------ltir |
169
.0414728
.0115435
.0153
.0666
lngold |
169
.0102196
.0504911 -.1909511
.1599647
lnwti |
169
.0114014
.0884894 -.3366812
.2031284
inf |
169
.1990986
.4032481 -1.933869
1.21459
lnfx |
169
-.0006052
.0251578 -.0645733
.0753924

Correlation Matrix with Significance Levels

. pwcorr

lnsp lndps lneps lngdp lnstir ltir lngold lnwti inf lnfx;

|
lnsp
lndps
lneps
lngdp
lnstir
ltir
lngold
-------------+--------------------------------------------------------------lnsp |
1.0000
lndps |
0.0092
1.0000
lneps |
0.3209 -0.1419
1.0000
lngdp |
0.2948
0.3484
0.5129
1.0000
lnstir | -0.1356
0.3435
0.1836
0.4184
1.0000
ltir | -0.0425 -0.1663
0.0344
0.2946
0.1488
1.0000
lngold |
0.0960
0.0238
0.0406
0.0112 -0.1845 -0.0512
1.0000
lnwti |
0.2754
0.0212
0.2794
0.2613
0.0073
0.1232
0.1391
inf |
0.1152
0.0562
0.1868
0.1711
0.0658
0.1340
0.0925
lnfx | -0.1920 -0.0550 -0.0673
0.0105
0.2652
0.0600 -0.1926
|
lnwti
inf
lnfx
-------------+--------------------------lnwti |
1.0000
inf |
0.5786
1.0000
lnfx | -0.2148 -0.1916
1.0000

Regression Analysis [1]


. regress lnsp lndps lneps lngdp lnstir ltir lngold lnwti inf lnfx;
Source |
SS
df
MS
-------------+-----------------------------Model | .069764763
9
.00775164
Residual | .216257001
159 .001360107
-------------+-----------------------------Total | .286021764
168
.00170251

Number of obs
F( 9,
159)
Prob > F
R-squared
Adj R-squared
Root MSE

=
=
=
=
=
=

169
5.70
0.0000
0.2439
0.2011
.03688

-----------------------------------------------------------------------------lnsp |
Coef.
Std. Err.
t
P>|t|
[95% Conf. Interval]
-------------+---------------------------------------------------------------lndps | -.0981858
.418169
-0.23
0.815
-.924068
.7276964
lneps |
.0645356
.0393251
1.64
0.103
-.0131314
.1422026
lngdp |
2.153689
.7057715
3.05
0.003
.7597931
3.547585
lnstir | -.0827134
.0279575
-2.96
0.004
-.1379293
-.0274975
ltir |
-.428643
.2888976
-1.48
0.140
-.9992147
.1419288
lngold | -.0015463
.0586508
-0.03
0.979
-.1173815
.1142888
lnwti |
.0826085
.0413248
2.00
0.047
.0009922
.1642247
inf | -.0055852
.0087825
-0.64
0.526
-.0229306
.0117601
lnfx | -.1396791
.1241221
-1.13
0.262
-.3848197
.1054615
_cons |
.0065781
.011914
0.55
0.582
-.0169521
.0301082
------------------------------------------------------------------------------

Variance Inflation Factor


Variable |
VIF
1/VIF
-------------+---------------------lngdp |
2.42
0.414069
lneps |
1.89
0.530218
lndps |
1.80
0.556081
lnwti |
1.65
0.605425
inf |
1.55
0.645482
lnstir |
1.52
0.659759
ltir |
1.37
0.727944
lnfx |
1.20
0.830272
lngold |
1.08
0.923178
-------------+---------------------Mean VIF |
1.61

RESET test
Ramsey RESET test using powers of the fitted values of lnsp
Ho: model has no omitted variables
F(3, 156) =
4.74
Prob > F =
0.0034

Heteroskedasticity Test
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: fitted values of lnsp
chi2(1)
Prob > chi2

=
=

33.98
0.0000

Durbin Watson Test


Durbin-Watson d-statistic( 10,

169) =

1.850172

ii

Appendix 2
Regression Analysis [2]
. regress lnsp lndps lneps lngdp lnstir ltir lnwti inf lnfx;
Source |
SS
df
MS
-------------+-----------------------------Model | .069763817
8 .008720477
Residual | .216257946
160 .001351612
-------------+-----------------------------Total | .286021764
168
.00170251

Number of obs
F( 8,
160)
Prob > F
R-squared
Adj R-squared
Root MSE

=
=
=
=
=
=

169
6.45
0.0000
0.2439
0.2061
.03676

-----------------------------------------------------------------------------lnsp |
Coef.
Std. Err.
t
P>|t|
[95% Conf. Interval]
-------------+---------------------------------------------------------------lndps | -.0987925
.4162295
-0.24
0.813
-.9208048
.7232198
lneps |
.0645087
.0391889
1.65
0.102
-.0128856
.1419029
lngdp |
2.153256
.7033735
3.06
0.003
.7641627
3.54235
lnstir | -.0825897
.0274752
-3.01
0.003
-.1368505
-.028329
ltir | -.4284511
.2879027
-1.49
0.139
-.9970305
.1401283
lnwti |
.0825306
.0410901
2.01
0.046
.0013817
.1636794
inf | -.0055893
.0087537
-0.64
0.524
-.0228769
.0116984
lnfx | -.1393393
.1230652
-1.13
0.259
-.382381
.1037023
_cons |
.0065632
.0118634
0.55
0.581
-.0168659
.0299923

------------------------------------------------------------------------------Variance Inflation Factor


Variable |
VIF
1/VIF
-------------+---------------------lngdp |
2.41
0.414294
lneps |
1.88
0.530576
lndps |
1.79
0.557769
lnwti |
1.64
0.608535
inf |
1.55
0.645680
lnstir |
1.47
0.678859
ltir |
1.37
0.728406
lnfx |
1.19
0.839318
-------------+---------------------Mean VIF |
1.67

RESET test
Ramsey RESET test using powers of the fitted values of lnsp
Ho: model has no omitted variables
F(3, 157) =
4.70
Prob > F =
0.0036

Heteroskedasticity Test
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: fitted values of lnsp
chi2(1)
Prob > chi2

=
=

33.97
0.0000

Durbin Watson Test


Durbin-Watson d-statistic(

9,

169) =

1.849776

iii

Appendix 3
Regression Analysis [3]
. regress lnsp lneps lngdp lnstir ltir lnwti inf lnfx;
Source |
SS
df
MS
-------------+-----------------------------Model | .069687674
7 .009955382
Residual |
.21633409
161
.00134369
-------------+-----------------------------Total | .286021764
168
.00170251

Number of obs
F( 7,
161)
Prob > F
R-squared
Adj R-squared
Root MSE

=
=
=
=
=
=

169
7.41
0.0000
0.2436
0.2108
.03666

-----------------------------------------------------------------------------lnsp |
Coef.
Std. Err.
t
P>|t|
[95% Conf. Interval]
-------------+---------------------------------------------------------------lneps |
.0690406
.0341223
2.02
0.045
.0016556
.1364256
lngdp |
2.068081
.6031678
3.43
0.001
.8769406
3.259222
lnstir | -.0845604
.0261137
-3.24
0.001
-.13613
-.0329908
ltir | -.4001543
.2612895
-1.53
0.128
-.9161509
.1158423
lnwti |
.0827009
.0409632
2.02
0.045
.0018064
.1635954
inf | -.0057282
.0087084
-0.66
0.512
-.0229257
.0114693
lnfx | -.1344142
.1209472
-1.11
0.268
-.3732616
.1044332
_cons |
.0054165
.0108034
0.50
0.617
-.0159182
.0267513
------------------------------------------------------------------------------

Variance Inflation Factor


Variable |
VIF
1/VIF
-------------+---------------------lngdp |
1.79
0.560082
lnwti |
1.64
0.608721
inf |
1.54
0.648580
lneps |
1.44
0.695736
lnstir |
1.34
0.747083
lnfx |
1.16
0.863879
ltir |
1.14
0.879160
-------------+---------------------Mean VIF |
1.43

RESET test
Ramsey RESET test using powers of the fitted values of lnsp
Ho: model has no omitted variables
F(3, 158) =
4.51
Prob > F =
0.0046

Heteroskedasticity Test
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: fitted values of lnsp
chi2(1)
Prob > chi2

=
=

34.41
0.0000

Durbin Watson Test


Durbin-Watson d-statistic(

8,

169) =

1.849169

iv

Vous aimerez peut-être aussi