Asymmetry of stock price reactions on oil price movements
Oil price sensitivity of stocks and indices in the Low Countries
Pieter Lont (1350269)
University of Groningen
Master’s Thesis MBA Finance (Corporate Financial Management)
December 11
th2009
Jel codes: C22, E44, G12, Q40, and Q43
Key words: oil prices, asset pricing, asymmetry, consumer expenditures, macroeconomy
Abstract
Introduction
Media, policymakers and economists give special attention to oil prices. Some
explanations for this large interest are that oil prices are relatively volatile compared
to other goods and services, the demand for oil is relatively inelastic, oil price
fluctuations are largely determined by exogenous forces to the economy, and oil
prices are often utilized as an instrument to explain movements in the economy
(Awerbuch & Sauter, 2006; Kilian, 2008a). Kilian & Park (2009) report strong
presumption in the financial press that oil prices also drive stock markets. Up to
midway 2008 the oil price increased above €90,- a barrel (~+100% in a single year).
At the same time companies operating in the energy sector reported very high profits,
while in general the stock markets moved down. During the global credit crunch and
the subsequent economic recession the oil price dramatically decreased to about €35,-
a barrel midway April 2009 (~-60% in a single year). In the same timeframe stock
markets declined considerably as well, oil producing companies stocks being no
exception. Stock price reactions appear to depend on the direction of the oil price
movement, at least for certain industry sectors. There has been a continuing interest
by researchers in the role and impact that oil has on the general economy as well as
financial markets and how these relations evolve. Extensive literature is available on
the oil-GDP relationship (e.g. Hamilton, 1983; Tatom, 1987; Hamilton, 1988a; Mork,
1989; Mork et al., 1994, Lee et al., 1995; Hooker, 1996; Hunt et al., 2001; Hamilton,
2003; Barsky & Killian, 2004). These papers suggest that oil price increases and oil
price volatility are responsible for lower economic growth (recessions), reduced
productivity, higher inflation, higher wages and higher unemployment. By affecting
actual as well as expected economic activity, the cost of production, corporate
earnings, balance of wealth between oil-consuming countries and oil-producing
countries, inflation, and monetary policy, changing oil prices are also expected to
have implications for equity and bond valuations
1, currency exchange rates and
government financing (e.g. Yurtsever & Zahor, 2007). Stocks should be valued for the
net present value of its future cash flows discounted using their risk characteristics
(Schleifer, 2000). Rising oil prices increase production costs and lower cash flows and
with that stock prices (Sadorsky, 2008). Furthermore rising oil prices often lead to
higher interest rates, which will lead to higher discount rates. Higher interest (discount)
1
rates make stocks less attractive and therefore normally lead to a fall in stock prices
(Basher & Sadorsky, 2006). The International Monetary Fund (IMF, 2000, 2004,
2007) reports that the transfer of income from oil consumers to oil producers will lead
to a fall in aggregate demand, which also leads to revaluations of equity and debt. The
sensitiveness of oil-importing countries depends on the degree to which they are net
importers and the oil intensity of their economies. A rise in the cost of production puts
pressure on profit margins. The consequences for the price level and inflation will
depend on monetary policies, consumers that seek higher wage increases, producers
that seek to restore profit margins, and the creditworthiness of oil-importing countries.
Currencies would adjust to changes in trade balances
2. This would raise the costs of
external debt and the oil-import bill. A loss of business and consumer confidence,
inappropriate policy responses and higher prices of other commodities would amplify
these economic effects. According to Kilian (2008a) oil prices rises could affect the
input (supply) as well as the output (demand) as it raises the marginal cost of
production, and reduces the demand for output. The negative oil price sensitivity will
be greatest in industries with a relatively high proportion of their costs devoted to
oil-based inputs. Also companies for which the costs of transportation are high are
affected. Furthermore some industries derive considerable revenue from oil-related
products (Faff & Brailsford, 1999). However the impact of oil on the general
economy is far larger than can be explained by means of these direct effects, as the
share of oil in production is relatively small. However oil prices could also induce
indirect effects (Lee & Ni, 2002). Killian (e.g. 2008a) and Hamilton (e.g. 2009b)
imply that the oil price changes affect the economy primarily through their effect on
consumer expenditures and firm expenditures. Multiple papers on the relation
between oil and the macro-economy report on an asymmetric response of the
economy in case of up and down movements of the oil price. Important explanations
for asymmetry are adjustment costs or sectorial shifts, financial stress, monetary
policy responses, changing demand composition, and investment under uncertainty
(Sadorsky, 2008). Market participants are in need of a framework that shows the
transmission mechanisms along which oil price changes will affect stock market
returns (Malik & Ewing, 2009). In this paper I will investigate whether oil returns are
an important factor in explaining stock returns, and whether the asymmetric relation
2
identified for economic output also translates into stock prices. There are multiple
views possible on the relationship between oil returns and stock returns.
The first view: The response will be in line with the classic supply-side effect in
which rising oil prices are indicative of the reduced availability of a basic input to
production. Oil users have higher production costs, and oil producers will have higher
revenues. The direction of the relation will depend on whether the company is an oil
producer (oil explorer) or an oil consumer. This direct effect suggests that oil price
changes have generally very little impact on stock prices, as most business activities
are not very energy intensive.
The second view: Oil price changes lead to (are related to) economic upheavals and
economic downturns and with that cause changes in investment and consumption
behaviour for many products. The indirect effect of oil price fluctuations will
influence revenues, profits, investments and cash flows of many firms. Cyclical
companies could show large reactions, but the consumption of products of oil-users
and oil-based products is expected to be affected the most. Furthermore indirect
effects such as adjustment costs could also apply for the oil-related stocks.
The relation is expected to be mostly negative, as most companies are consumers of
oil and oil derivatives. But some firms might be able to minimize the negative impact
on profitability by incorporating the higher oil costs in consumer prices. And certain
companies, such as oil explorers and oil producers, are even expected to obtain larger
cash flows, as their final product (oil) is more highly valued. Certainly as the price
elasticity of oil demand is small. Both the downside and upside risks of oil price
changes need to be considered (Fan, et al., 2008). With regard to the shape of the
relation I will focus on whether positive oil returns have the same (symmetrical)
impact as negative oil returns on the returns of individual and groups of stocks
3.
Asymmetry in this model means that positive oil returns lead to negative
consequences for most companies, while negative oil returns do not have a positive
impact of the same magnitude. The asymmetric reaction could be the consequence of
direct as well as indirect effects on the input as well as the output side of firms. This
will affect expected cash flows.
3
The following two main hypotheses are proposed:
1. Oil returns have a negative relationship with (most) stock returns.
2. Oil returns have asymmetric effects on (most) stock returns.
Transmission mechanisms
of crude oil are far less important for understanding changes in stock prices. Changing
types of oil shocks explain why regressions of macroeconomic aggregates on oil
prices tend to be unstable (Kilian & Park, 2009). The recent oil shock (2007/2008)
was driven primarily by global aggregate demand instead of by supply disruptions,
and therefore did not directly cause a recession
4. Apergis & Miller (2009) indicate
that there are differences in the impact between oil demand and oil supply shocks, but
all contribute significantly in explaining stock market returns. Kilian & Park (2009)
report that further insights can be gained from considering the responses of
industry-specific stock returns to global demand and supply shocks. Their results suggest that
appropriate portfolio adjustments depend on the underlying cause of the oil price
change. Outside the energy sector, the strongest responses are found in sectors for
which oil price shocks affect the demand for goods and services (in line with
Hamilton, 1988a).
Increasingly more researchers pay attention to the relationship between oil prices and
stock prices, also considering possible asymmetry
5(e.g. Huang et al., 1996; Sadorsky,
1999; Nandha & Hammoudeh, 2007; Sadorsky, 2008; Kilian & Park, 2009). These
papers indicate that for certain companies and sectors oil returns are an important
factor. Furthermore there is some empirical evidence of asymmetry in the reaction of
stock returns on the sign and volatility of the oil returns. Most studies relating oil
prices to financial market activity only examine the impact of oil price shocks on
stock prices across the entire market, rather than concentrating on individual stocks or
groups of stocks (Faff & Brailsford, 1999). Analysis at the aggregate level may hide
stock and sector specific sensitivities (Sadorsky, 2001). This study tries to fill this gap.
I choose to directly analyse the relationship between oil returns and stock returns as
well as sector/index returns, but am aware of the underlying complexity of this
relation. Incorporating all kinds of macroeconomic variables and different types of
shocks would make the model complex, which is beyond the intension of this paper.
Furthermore as weekly and monthly are used, it is difficult to obtain reliable data on
these variables. The main variables are the oil return variables in both the upward and
the downward direction and the oil volatility variable. The oil price volatility could
cause erratic behaviour of investors. A market factor is included to partially capture
4
the general macroeconomic situation. Exchange rates, interest rates and interest
spreads are often considered as explanatory variables of stock market returns
(Grinblatt & Titman, 2004), and are therefore also included as control variables to
give a rich picture of what drives stock returns. The data analysis will be quantitative
as well as qualitative. Correlations between the variables are determined and summary
statistics of the firms and sectors are provided. For every individual stock and sector a
separate regression is performed. There will be made use of time series data to study
the size and shape of the relation between company stock returns and oil price
movements. I want to obtain long term coefficients, but I will perform parameter
stability tests to check whether the relation dramatically changes over time. I will look
for differences between individual companies and industry groups in the reaction on
oil prices. Furthermore I will consider the impact of firm characteristics like size,
sector, and P/E ratio effects in a separate portfolio analysis. On the basis of the results
I can come to a conclusion on which transmission mechanisms are dominant. The
Dutch and Belgian stock markets are selected as the Netherlands and Belgium have
very open economies and are sensitive to other markets, especially to the US market.
Furthermore they are largely dependent on oil imports
6and are relatively energy
intensive. Therefore it is expected that the Belgium as well as the Dutch stock market
and with that most individual stocks will show a clear negative relation with the oil
price. An important question is whether oil price risk is a global factor risk or actually
a firm-specific risk factor that can easily be diversified away. I expect a negative
relation for most firms/sectors, except a few like oil producers and oil explorers, as oil
prices rises have a negative impact on the economy, and as oil is a direct or indirect
input for many industries. Asymmetry is widely expected as the uncertainty on the
consequences for the total market and the individual companies will lead to panic
reactions. Especially for the stocks with operations closely related to oil a significant
as well as asymmetric relation is expected, as both direct and indirect effects could
reinforce each other. The results will further the understanding of the interaction
between oil returns and equity returns and with that the scale of oil price risk, which
could be useful for investors, hedgers, managers, and policymakers (Basher &
Sadorsky, 2006). The empirical findings are useful to investors who need to
5
By symmetry, it is meant that irrespective of an estimated coefficient’s sign, the magnitudes of the
estimated coefficients are the same (Sadorsky, 2008).
6
understand the exact effect of oil price changes on certain stocks across industries, to
determine an optimal savings and investment strategy. And also for managers of
certain firms in order to evaluate the efficiency of oil price risk hedging policies
(Devlin & Titman, 2004). Maybe even more interesting to investors is the possible
appearance of asymmetry. It could be argued whether asymmetry in the long run
actually is sustainable, as this suggests that price changes that are fully reversed will
leave net effects that are not neutral (Tatom, 1993). But the consequences for the cash
flows of firms could well be non-linear. With respect to asymmetry it also important
whether oil price changes are temporarily or permanent (Devlin & Titman, 2004).
Furthermore stock prices often are temporarily mispriced cause of noise trading or
erratic behaviour (uncertainty effect), which could provide interesting arbitrage
opportunities for rational investors. However a large variety of factors simultaneously
affect stock prices, so it is difficult to determine whether a stock is mispriced. It is
important for investors to know which stocks have oil price risk and also show
asymmetric behaviour, so they can cover that risk by taking measures and make
optimal portfolio allocation decisions. When oil price risk is a market wide factor
diversification is not that well possible. But a portfolio of oil-producing stocks and
oil-consuming stocks could capture some or even most of this risk, and be a good
hedging tool against possible asymmetry.
This paper is organised as follows. First there is reported on the general importance of
oil, including oil price risk hedging. In the next two sections theoretical and empirical
literature is examined to investigate the importance of oil for the macroeconomy as
well as in financial markets. A differentiation is made in the strength and shape of the
relationship. There is elaborated extensively on the exact routes to asymmetry. The
next section explains the methodology, model and data that will be applied. The
actual results are discussed in the next section. In a separate section also the
robustness of the results is checked. And the last section concludes the results and
gives some recommendations for further research. In the Appendix more extensive
graphs and tables on macroeconomic variables and tables with results are presented,
including a table on the companies selected
7. Every new section starts on a new page
to get a clear structure in the paper.
7
Importance of oil
Oil and oil-derivatives are used by almost all companies and consumers. Oil is needed
as an energy resource for industrial production, electric power generation, and
transportation. Oil price fluctuations are likely to influence all industrial sectors in
modern developed countries (Sadorsky, 1999). The price elasticity of demand is low
(e.g. Atkeson & Kehoe, 1999). Energy expenditures account for about 4% of GDP in
most industrial nations (Faff & Brailsford, 1999). According to the IMF (2004), a 10$
oil price increase will decrease world GDP with about 0.5% and inflation will rise
with 0.5%, and also unemployment is expected to increase
8. Oil is heavily traded
daily on spot and futures markets. Oil prices fluctuate from month to month because
of changes in expectations on supply and demand. Important factors are changes in
global economic conditions, political tensions in oil-producing countries, wars,
terrorist attacks, natural disasters, embargo’s, OPEC price agreements, the amount of
oil inventories, possible cutbacks in production, monetary policies that have an
inflationary effect, value change of US dollar, technological changes, decrease in oil
dependency, new oil discoveries, uncertainty on supplies from mature fields,
availability of substitute fuels, usage of alternative energy resources, increases in
spare capacities, stronger incentives for conservation, and oil consumption growth in
developing countries (e.g. IMF, 2004; Marimoutou, et al., 2009;)
9. When price
changes are considered permanent, the present value of future revenues is strongly
affected by changes in spot prices. The best strategy to deal with oil price volatility is
the use of market-based risk instruments (Devlin & Titman, 2004). Figure 1 shows the
oil price development of some oil spot and future series
10. The price of WTI is
normally higher than Brent oil as it is sweeter and lighter. The oil spot and futures
return series reveal significant volatility clustering. These different series show
similar behaviour, but the volatility is somewhat higher for the spot and short-term
futures. During the last decade the oil price increased from €10,- to €40,- a barrel,
with a peak of about €90,- a barrel. This oil price rise has been caused by a high
dependency on oil, world oil depletion and steady increase of world oil consumption.
The recent economic recession has sent the oil price back down, but the general
tendency is that oil demands and with that oil prices keep rising.
8
In Figure A2 the development of some macroeconomic variables like GDP, inflation and
unemployment are displayed for the Netherlands, Belgium and the Eurozone.
9
Figure 1: Multiple oil spot and oil future futures series (WTI & Brent euro prices)
0 10 20 30 40 50 60 70 80 90 100 99 00 01 02 03 04 05 06 07 08 09 Date Pr ic e ( eu ro) OILWTX OILWTXI OILWTX2 OILBREN OILBRNI OILBRNTFigure 2 shows the oil returns and oil price volatility
11. The sudden fluctuations in oil
price during the last year and around 9/11/2001 are clearly visible.
Figure 2: Oil returns and oil volatility (monthly data)
-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 99 00 01 02 03 04 05 06 07 08 09 Date R et u rn /V ol at il ity OILRETEU OILVOLMONTH
Figure 3 shows the stock returns for AF-KLM and RD-Shell. AF-KLM is a large oil
consuming (aviation) company, and is expected to have a clear negative relation with
oil returns. RD-Shell is an oil producer, and is expected to show a clear positive
relation with oil returns. It seems that the returns of AF-KLM and RD-Shell have an
opposite direction. The oil returns and stock returns show similar patterns, which
suggests that oil returns influence stock returns. However other variables could also
be simultaneously at work.
10
I have converted the US dollar prices into euro, in order to clear out the effect of currency differences.
11
Figure 3: Stock returns of AF-KLM and RD-Shell (monthly data)
-0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 99 00 01 02 03 04 05 06 07 08 09 Date Re tu rn AF-KLM RD-SHELLTable 1 shows data on the oil consumption over the period 1998-2008 for certain
regions of the world. The oil usage in the Low Countries shows large increases and is
relatively high. The increase in energy efficiency has occurred because of reduced
energy intensity and more reliance on a diversified range of energy sources.
Table 1: Oil consumption (thousands of barrels per day)
Cons.
1998-2008
Cons.
Barrels/million
1998-2008
2008
% change
p/p 2008 USD GDP 2008
% change
World
84455
14.7
5
510
-43.1
US
19419
2.7
24
506
-34.0
Europe & Eurasia
20158
1.7
9
294
-59.4
European Union
14765
-0.6
11
291
-51.7
Belgium & Luxemb.
836
27.6
28
509
-36.2
Netherlands
982
15.0
22
398
-49.0
France
1930
-4.3
12
243
-50.4
Germany
2505
-14.1
11
247
-51.3
UK
1704
-2.1
10
230
-49.3
Asia Pacific
25339
29.6
2
617
-30.8
China
7999
89.2
2
664
-52.7
India
2882
46.8
1
877
-51.0
Japan
4845
-12.0
14
361
-29.9
Region/Country
Source: BP Statistical Review of World Energy, June 2009 (www.BP.com). GDP is according to
the International Monetary Fund.
Impact of higher oil prices on global economy
Figure 4: World oil production/demand 2001-2009 (million barrels per day)
70 75 80 85 90 01 02 03 04 05 06 07 08 09 Date M Il li on ba rr el s a day SUPPLY DEMANDSource: Energy Information Administration (www.eia.doe.gov/emeu/ipsr). Notes: For the oil
supply monthly figures for the world are used. For the oil supply of 2008 an eight month average
is used. For the oil demand the OECD demand is scaled up to a monthly world demand, by
means of the average annual world demand. For the oil demand of 2009 there is not enough data.
Importance of oil for the Netherlands and Belgium
Belgium and the Netherland are large net importers of fossil fuels
12. Furthermore they
are not running ahead in the usage of sustainable energy resources. The risk from oil
price changes in these countries is thus likely to play a large role in the development
of these economies and their financial markets. Oil is mainly used by the Industry and
Transport sectors. The energy intensity of the Dutch economy gradually decreased the
last decades, mainly for energy intensive sectors. The Dutch and Belgium economies
both are prosperous (high GDP/citizen), open, and highly dependent on foreign trade.
Furthermore they have large scale service and energy companies. The natural gas
discoveries in the Netherlands in the 1960s have lead to this sectorial dependence
13. In
the Netherlands services account for over half of the national income and are
primarily in transportation, distribution, and logistics, and in financial areas, such as
banking and insurance. Industrial activity generates about 20% of the national product
and is dominated by the metalworking, oil refining, chemical, and food-processing
industries. The same more or less holds for Belgium
14.
12
Even while the Netherlands is a large producer of natural gas.
13
Gas benefits of the Netherlands are ~
€
10 billion annually, ~7% of total revenues of the Government.
14
Oil price risk hedging
Theoretical literature review
firms (Sadorsky, 2003). This will lead to changes in investment as well as
consumption behaviour (e.g. Kilian, 2009).
volatile than the underlying stocks as they more heavily react on price changes cause
of changed expectations and options will expire without execution or are far out of the
money. While for a put option always holds that at least a fee is obtained, without
actually always having to deliver the stock. So options prices show asymmetric
behaviour as a reaction on underlying asset price changes. Furthermore companies
often have long-term oil contracts with suppliers. These prices are much less volatile,
and give protection against temporary higher oil prices.
it makes products more costly to manufacture (supply). Somewhat problematic is that
the greatest effects of an oil shock do not appear until about a year after the shock,
and strong recovery is observed within a couple of years also when oil prices remain
high (e.g. Hamilton & Herrera, 2004). According to Aguiar-Conraria & Wen (2007)
these models are not able to explain this accelerator effect. They propose a model to
explain the multiplier accelerator effect by including investment demand, which
depends on the production level of other firms. Hickman et al. (1987) state that the
aggregate price level responses are proportional to the magnitude of the oil price
shock. This implies that energy price changes have a symmetric effect on the
economy. When oil prices rise, energy-using capital is rendered obsolete, unless (1)
product prices adjust sufficiently, (2) product demand is unaffected, and (3) other
low-cost methods of production are unavailable (Tatom, 1987). If this is not the case,
alterations in the optimal employment of capital resources occur. Oil price shocks
reduce productivity by effectively destroying capital resources. Output and
employment can be altered only after sufficient time has passed. Such inelastic factor
proportions increase the short-run output loss associated with a rise in energy prices.
Relevant is to what extend capital and labour are considered interchangeable, as there
is a period of adjustment to a lowered desired capital/labour ratio. The dispersion
hypothesis posits that a considerable amount of unemployment can be accounted for
by sectorial shifts in demand, which require time for reallocation of labor and capital.
Sectorial shifts and short-run technological or nominal rigidities are likely to play an
important role in accounting for the correlations between oil prices and the economy
and underlying stock prices. Reallocation costs either across or within sectors could
result in a negative response. Hamilton (1988b) and Balke et al. (2002) state that
asymmetry could be the result of adjustment costs, as these lower economic activity in
both directions. Such adjustment costs could arise from sectorial imbalances
(Loungani, 1986), coordination failures between firms (Huntington, 1998) or because
the energy-to-output ratio is set in the capital stock (Atkeson & Kehoe, 1995, 1999).
For putty-clay energy intensive capital goods
15a change in oil prices could have
negative output consequences as firms adjust to new energy prices (Atkeson &
Kehoe, 1995, 1999). It is all about adjustment costs associated with shifts among
economic sectors in response to supply shocks (Loungani, 1986; Hamilton, 1988b).
15
decreases. For example because wages and other prices are downward sticky. Balke,
et al. (2002) report that the effect of oil prices on output is reflected primarily through
interest rates. The relatively fluid market rates move in anticipation of asymmetric
real effects in the future. Bernanke et al. (1997) report that contractionary monetary
policy accounts for the decline in aggregate economic activity following an oil price
increase. If prices are sticky downward, an oil price increase leads to important GDP
losses if monetary authorities do not maintain nominal GDP constant by means of
unexpected inflation. After a decline in oil prices, real wages must grow to obtain a
new market equilibrium. However also interest rates respond asymmetrically to oil
price shocks. Monetary policy can accommodate an oil price increase by raising
aggregate demand and lessen the negative effect on output, but at the cost of higher
prices. Central banks can reduce aggregate demand and lessen the price effect, but at
the cost of lower output. Furthermore also the short-term interest rate response
suggests substantial asymmetry as they incorporate the asymmetric response of the
Fed through the term structure, the expectation of financial markets on the real effect
of oil price changes, and financial stress cause of the oil price shock.
Killian (2008a) and Hamilton (e.g. 2009b) report there is general consensus that the
primary transmission mechanism involves a reduction in the demand for goods and
services. According to Kilian (2008a) higher oil prices cause both a reduction in
aggregate demand and a shift in expenditures. Kilian (2008a) and Edelstein & Kilian
(2009) report that oil price changes affect consumer expenditures, as they:
1. Could change discretionary income, as consumers have more or less money to
spend after paying their energy bills.
2. Could create uncertainty about the development of oil prices, causing
consumers to postpone purchases of consumer durables.
3. Could reduce consumption, as consumers increase their precautionary savings.
4. Could reduce consumption of energy-using durables even more than other
durables, as these durables require energy which is more costly.
Literature suggests that it depends on the sector whether oil price shocks will mainly
affect supply or demand. Real balance effects, monetary policy, and income transfers
have to do with the supply side, while consumer and firm expenditures have to do
with the demand side. The real balances, income transfer and potential output effects
are expected to have a symmetric relationship between oil price changes and output
growth. Monetary policy will cause non-linear output responses if central banks
tighten policy in response to oil price increases but do not pursue expansionary policy
in the face of oil price declines. Both the sectorial shocks and uncertainty effect can
explain the asymmetry if oil price shocks produce increased volatility and uncertainty.
The increased volatility in case of sharp rises of oil prices reinforces the other
negative effects, while volatility generated by falling oil prices offsets the other
positive effects.
Empirical literature review
Many researchers have studied the strength and shape of the relationship between oil
and the economy (e.g. Tatom, 1981; Mory; 1993; Mork et al., 1994; Lee et al., 1995;
Ferderer, 1996; Huntington, 1998; Davis & Haltiwanger, 2001; Hunt et al., 2001;
Balke et al., 2002; Hamilton, 2003) and mostly found a non-linear relationship.
Increasingly more researchers also study the impact of oil prices on stock markets
(e.g. Hamilton, 1988a; Huang et al., 1996; Jones & Kaul, 1996; Sadorsky, 1999;
Papapetrou, 2001, Jones et al., 2004; Bachmeier, 2008). New papers of Hamilton
(2009) and Kilian (2008/2009) study whether the response to the more recent oil price
shocks (and economic downturn) is similar. However few studies have attempted to
determine through what transmission mechanisms oil price shocks exactly operate to
produce an asymmetric response in aggregate economic activity and stock prices.
Historical oil prices rises have generally resulted in falling aggregate output (GDP), a
higher price level, and higher interest rates. As is shown in Table 2, the best
explanation is a classic supply-side effect in which rising oil prices are indicative of
the reduced availability of a basic input to production. However the basic supply
shock effects can only partially explain the large effect that oil price shocks have on
aggregate economic activity. Additional explanations for the intensity of the response
are proposed, such as restrictive monetary policy, adjustment costs, coordination
externalities, and financial stress. The different effects are consistent with observed
facts, and may be contemporaneously at work (Brown et al., 2004). The effects of
systematic monetary policy are not as important (anymore) as historically suggested
(e.g. Bachmeier, 2008; Herrera & Pesavento, 2009).
Table 2: Expected response to rising oil price
Economic theory
Real GDP
Price Level
Interest Rate
Historical Record
DOWN
UP
UP
Classis Supply Shock
DOWN
UP
UP
Aggregate Demand Shock
DOWN
DOWN
DOWN
Monetary Shock
DOWN
DOWN
UP
Real Balance Effect
DOWN
DOWN
UP
Evidence on oil-GDP relationship
Most studies initially focused on the relation of oil prices with economic activity.
Hamilton (1983) reports on statistically significant correlation between high crude oil
prices and historical recessions. The acceptance of a linear relationship has led to the
widespread usage of oil prices as a macroeconomic variable (Hooker, 1996). However
Mork (1989) found evidence that rising oil prices slow down economic activity more
than falling oil prices stimulate it. Also later studies (e.g. Hamilton, 2003; Lee et al.,
1995; Sadorsky, 1999; Jones et al., 2004) report on clear evidence of a non-linear
relation between oil price changes and GDP growth. Multiple researchers report clear
negative correlations between oil prices and aggregate measures of economic activity
such as recession, excessive inflation, low productivity and low economic growth.
Furthermore they find significant correlations between oil prices and microeconomic
data on output, employment, and real wages (e.g. Hunt et al., 2001; Papapetrou, 2001;
Barsky & Kilian, 2004). Huntington (1998) finds that the timing and the pattern of
product price movements as a reaction on oil price changes are different:
1. A significant part of the asymmetry is found in the energy sector.
2. Consumer prices respond asymmetrically to energy price changes.
3. Aggregate output responds asymmetrically to crude oil price changes.
utilization is sensitive to oil prices and supplies. Hamilton (2009b) therefore
concludes that the oil price has also contributed to the recent economic recession.
Multiple researchers have investigated the development of the shape and strength of
the relation over time:
1. Nonlinearity in the relation: Oil price increases have a bigger effect on the
economy than oil price decreases (e.g. Hamilton, 2003).
2. The causes (type) of the oil price shock: Higher global demand has a less
disruptive effect than lower global supply (e.g. Killian, 2009).
3. A changing relation over time: The modern economy knows better how to
cope with an oil price shock than previously (e.g. Blanchard & Gali, 2008).
Evidence on oil-stock relationship
Apergis & Miller (2009) show that different types of oil shocks all play a significant
but small role in explaining the adjustments in stock market returns. Nandha & Faff
(2008) argue that oil shocks can have adverse effects on output as well as profitability
of firms, especially for those firms where oil is used as an input. Driesprong et al.
(2004) report evidence that investors in stock markets underreact to oil price changes
in the short run. Increasing oil prices lower future stock market returns. This
predictability effect is less strong for oil-related sectors. Oberndorfer (2009) finds that
the relation over time between the oil returns and the oil and gas portfolios is not
restricted to a linear relationship. While the oil price change positively impacts both
oil and gas stock returns, oil volatility has a strong negative effect on stock returns.
This suggests that an increase in oil price volatility is only relevant to oil and gas
corporations. Yurtsever & Zahor (2007) specifically study the impact of oil on the
Dutch stock market, but restricted the study to some large capitalization companies
and sectors. They find that oil shocks have a negative impact on stock returns of some
industries and individual companies whereas they have a positive impact on oil and
gas companies. Furthermore their analysis shows that oil price increases and
decreases have an asymmetric effect on the equity markets. Huang, et al. (1996) find
evidence for significant causality from oil futures to stocks of individual oil
companies, but find no impact on the entire index (S&P500). However Nandha &
Hammoudeh (2007) find significant nonlinear causality from crude oil futures returns
to S&P500 index returns. Huang et al. (2005) find that the optimal threshold level
seems to vary with the dependency of the economy on imported oil and the attitude
towards adopting energy-saving technology. There are many possible scenarios as to
how oil returns and stock returns are linked, including feedback effects, lead-lag
relationships, and market price and volatility spillovers across markets. Huang et al.
(1996) find that stock index futures lead the underlying stock prices within the day,
but there is no feedback from stocks to oil futures. Also Sadorsky (1999) states that
the economic activity has little impact on oil price. This suggests that oil futures are a
good vehicle for diversifying stock portfolios.
Methodology and regression model
Multiple studies make use of multi-factor market models (e.g. Basher & Sadorsky,
2006; Sadorsky, 2008). In a multi-factor market model expected returns are linearly
related to risk factors and risk premiums (Basher & Sadorsky, 2006). A multi-factor
market model is in line with both the arbitrage pricing theory (APT) and a multi-beta
capital asset pricing model (CAPM). An important question for this research is
whether oil price risk should be seen as a common factor, or as a firm-specific factor.
In many previous studies two-factor models are used including market returns and oil
returns to explain stock returns. But these models are somewhat underspecified as
exchange rates and interest rates are not included (Sadorsky, 2001). Factors can also
be estimated by using portfolios formed on the basis of firm characteristics (Grinblatt
& Titman, 2004). An oil price factor should be well visible in the Dutch and Belgian
financial markets. Both countries are largely dependent on oil imports, are relatively
energy intensive, have relatively open economies, and depend on the global market
conditions. The relationship between oil prices and stock prices will be analyzed on a
company specific (microeconomical) and sectorial (macroeconomical) level. I will
use the following (general) multi-factor model to investigate how the different risk
factors influence stock/indices/portfolio returns:
it t t t t t t t it
c
Rm
D
Ro
D
Ro
Ov
Rir
Ris
r
R
=
+
β
1+
β
2 1+
β
3(
1
−
1)
+
β
4+
β
5+
β
6+
β
7Re
+
ε
(1)
In equation (1) the dependent/endogenous variable are the stock/indices/portfolio
returns denoted as R
it(i indicates the firm/index/portfolio and t indicates the time
period), and the independent/exogenous variables are the market returns Rm
t,the oil
returns Ro
t, the oil price volatility Ov
t,the interest rate returns Rir
t, the interest spread
returns Ris
t, and the exchange rate returns Rer
t. The betas indicate the different
coefficients that belong to the variables. The main variables are the variables on oil
returns and oil return volatility, and the other exogenous variables serve as control
variables. To analyse the asymmetrical behaviour the oil price changes are separated
into positive changes and negative changes. In equation (1) this is done by introducing
a Dummy. D
1is a dummy variable that takes a value of one (zero) if oil price
movement is positive (negative). Henceforth
D
1Ro
tis replaced with
Rou (return oil
tand interest spreads are included in the model to partially grasp the effects of
inflation. And by incorporating a market index there is partially reckoned with the
general economic situation. I have decided to directly test the influence of oil returns
on stock returns. The exact transmission mechanisms are not considered in the
regression model. Incorporating the type of oil price shock and macroeconomic
variables like GDP, unemployment and inflation is considered too complex, certainly
for weekly and monthly data
16. Furthermore Apergis & Miller (2009) reported that the
different types of oil shocks only play a small role in explaining the adjustments in
stock market returns. Interpreting weekly and even monthly changes in business
environment and type of oil supply shock is very arbitrary, and goes beyond what can
be achieved with long-term statistics. I focus on the longer term relationship and
thereby also consider other sampling frequencies. For every firm/index/portfolio I will
perform the regression for a 10 year period using monthly and weekly data
17. Daily
data contains too much noise, and shows some highly undesired behaviour like large
non-normality in the residuals. Separate portfolios are constructed to consider the
impact of firm characteristics like size, and sector more specifically for the monthly
data. The portfolio regressions are not performed for the weekly data, because of
limitations in Excel and Eviews. Furthermore the portfolios for the monthly data
already show very undesirable behaviour. Size is measured with market value and
sales as a proxy, as well as the stock market it is currently located (AEX, AMX,
ASCX, BEL20, BELMID, and BELSMALL). Also P/E ratios are included as this is
considered informational on stock returns. Sectors are determined on basis of the
classification and firm description of Euronext. Furthermore portfolios are constructed
on the oil usage and oil dependency, based on the firm descriptions of Euronext (see
Table A1
18).
Research questions
I want to examine the strength and shape of the relationship between oil returns and
sector/stock return in the Low Countries. The relation is expected to depend on the
16
In Tables A2-A5 correlations are given between macroeconomic variables like GDP, inflation and
unemployment for the Netherlands, Belgium and the Euro-zone, with euro oil prices, the S&P1500
(market factor) and Euribor interest rates and interest spreads. It is difficult to filter out an exact
relation between these parameters and oil returns, certainly as the variables considerably differ between
countries. But it gives an impression of the suitability of the model.
17
I believe the sampling period sufficiently covers the multiple types of oil shocks and business cycles.
18
dependency on oil as an energy resource. The impact can differ, dependent on the
time scope of the sample and the related range and movement of the oil price. The
actual dependency and oil usage of the individual stocks will be an explorative
research. Conclusions will be drawn on the basis of the results. In the portfolio
analysis I will analyse whether there are differences between certain groups of
companies. Upfront I consider the oil price factor as a common factor which is widely
visible. The stocks and sectors directly related to oil are expected to show the largest
response. A differentiation is made in the direct impact of oil for oil dependent
companies, and the indirect route of oil affecting the investment and consumption
behavior (output) of many firms. The direct (allocative) effects mainly focus on the
supply/input side of firms, and the indirect (aggregate) effects concentrate on the
demand/output side of firms. Furthermore I expect clear asymmetric behaviour. For
the individual stocks it is a combination of the different theories that amplify each
other, such as sectorial shifts, adjustment costs, changing customer demands,
changing investment behaviour, financial stress, monetary policy responses, business
cycle asymmetry, and investment under uncertainty. On the supply side oil price rises
are not in the same way and as quickly integrated in product prices as oil price
decreases. Furthermore changing oil prices will lead to adjustment costs in the
operation mix. On the demand side oil price changes lead to changes in consumption
expenditures. Consumers continue to spend money on primary goods such as
transportation, but will safe on luxury goods. Customer demand will be lower for
many products, which will reduce cash flows. These indirect (aggregate) effects
suggest that asymmetry could be obtained throughout the whole market and not only
for specific sectors with operations closely related to oil. However product cost prices
and product consumption of companies and sectors that are very oil intensive are
expected to show the strongest reaction, as the consequences for their cash flows will
be the most dramatic. But oil price changes could lead to investor sentiment or noise
trading, and other non-rational investment strategies. It could be difficult for investors
to assess the impact of oil prices changes on cash flows of firms. The following two
main hypotheses are proposed:
Oil returns have a negative relationship with stock returns.
(1)
The following additional hypotheses are also evaluated:
• Hypothesis 1: Market returns and stock returns are positively related
• Hypothesis 2: Oil returns are positively related with oil-producer stock returns
• Hypothesis 3: Oil price volatility is negatively related with stock returns
• Hypothesis 4: Oil price risk is a significant factor for many stocks/indices
• Hypothesis 5: The impact of positive oil returns is different from negative oil
returns.
The impact of oil returns on stock returns can be tested with the following hypotheses:
H0 = β
2= 0;
H0 = β
3= 0;
H1 = β
2< 0
H1 = β
3< 0
The symmetry hypothesis can be tested with the following hypotheses:
H0
= β
2= β
3;H1 = β
2≠ β
3Conversely a positive oil return sensitivity is expected in oil-related industries, in
which oil directly impacts revenues. Furthermore the impact of oil price changes on
equity prices will depend on the ability of firms to pass on the effect to customers. It
is also expected that oil price increases have a larger impact on stock returns than do
oil price decreases. I expect that oil price volatility will be negatively related to stock
returns, as higher volatility increases risk. Increases in the market returns are expected
to lead to increases in individual stock returns (Chen, 1991). Chen also signals the
interest rates should have a negative impact. Positive interest spread returns are
generally observed during economic upheavals and negative interest spread returns
are observed prior to periods of economic downturns. Positive exchange rate returns
are expected to lead to higher returns in the short run, but worsen the competitiveness
of firms. However the Low Countries have transactions in both directions with the US,
so the actual impact is difficult to assess. Summary statistics will be provided for the
microeconomical (individual stocks) and macroeconomical analysis (indices) as well
as the portfolios. Multiple correlation diagrams will be constructed for the correlations
between stock returns and the different factors. The regression model is estimated by
OLS. Preliminary test results suggest considerable heteroskedasticity. Therefore
White heteroskedasticity period robust standard errors are used in calculating the
F-statistics. Model adequacy is tested using various regression diagnostic tests to check
for non-normality, autocorrelation, heteroskedasticity and GARCH effects. The
returns in stead of the levels of the variables are selected to prevent possible problems
with unit root. For the interest spread variable the logarithmic returns are replaced
with simple returns, as the interest spread can also turn negative. Wald tests will be
used to test the symmetry of the oil price effects. For the sectorial and individual stock
regressions there will be made use of monthly data. Robustness is checked by also
considering weekly data. Furthermore portfolio regressions are executed to investigate
the relation between oil returns and firm characteristics like size and sector more
specifically using monthly data.
Regression diagnostic tests
Data
Data is needed on the different risk factors included in the multi-factor model, as well
as on the stock returns. The data cover the period January 1st, 1999, to May 31
st2009
for a total of 125 monthly and 543 weekly observations respectively. High frequency
data increases sample size and will give a more detailed picture of the oil price and oil
price volatility sensitivity. Daily data samples are not considered as there could be a
lag in the incorporation of oil returns (new information) in stock returns. Furthermore
daily data also contains more noise and has some undesired properties. The data is
expressed in the local currency (Euro). Data only available in dollars in Datastream
are converted into euros by means of the exchange rate. Stock returns of all traded
Dutch and Belgium funds (AEX, AMX, ASCX, BEL20, BELMED, and BELSMALL
funds
19), market returns of the S&P1500 index, oil futures prices, interest rates,
exchange rates, and firm sizes are obtained from Datastream. Euribor rates are used as
a proxy for the risk-free interest rate
20. Operating sector classifications are obtained
from Euronext
21. Continuously compounded stock/index/portfolio returns are the
dependent variable in each regression model. The data for the dependent variable
consists of monthly and weekly returns calculated out of the monthly and weekly
closing prices of the return index (prices including added back dividends) of mostly
Dutch and Belgium stocks and indices. Daily excess returns are calculated by
subtracting the yield of 3-month Euribor from the continuously compounded monthly
and weekly returns. For European investments the usage of Euribor rates for the
risk-free interest rate is common. For the risk-risk-free return, the average rate throughout the
period is taken, and transformed into weekly and monthly returns. Market returns are
measured in the same way as stock returns as the excess returns on the S&P1500
index
22. The market return is a proxy for changes in aggregate economic wealth that
affects risk premiums and expected returns. Oil futures returns are measured as the
log difference of the daily return on the West Texas Intermediate (WTI) crude oil
futures contract which trades on the NYMEX. Oil futures prices are selected as they
are less affected by short-run price fluctuations. Oil price volatility could be
calculated using this data. For the monthly analysis, oil price volatility is measured by
19
These are 174 stocks.
20
www.euribor.org.
21
www.euronext.com.
22
Results
In this section I will present the main results. I make a differentiation in the individual
stock analysis, indices analysis, and portfolio analysis, using monthly data. This to
identify differences between the microeconomical and macroeconomical level, and
more specifically consider firm and market characteristics that could make the
relation more clear. For every group summary statistics, correlations, and regression
results are presented. In Table 3 the summary statistics are presented for the six
independent variables. The average market return is negative and the average oil price
return is positive
23. Positive oil returns are by definition positive and negative oil
returns are by definition negative. Oil price volatility is always positive, as only the
magnitudes of the price changes are relevant and not the signs. It is defined as the sum
of a certain number of squared daily oil returns. The interest return on average is
negative. Interest rates have declined the last decade. Interest spreads have a positive
mean, which indicates that the interest spread between 12-mth and 3-mth Euribor has
increased in this period. The exchange rate has an average return which is just
positive which indicates that the Euro has somewhat increased in value against the US
Dollar.
Table 3 Summary statistics main (independent) variables (monthly data)
VARIABLE NO
MEAN
MEDIAN
SKEW
KURT
STDEV
T-VALUE
R
m125
-0.005
-0.002
-0.351
-0.265
0.051
-0.997
R
o125
0.012
0.033
-0.859
2.276
0.109
1.267
R
ou75
0.081
0.074
1.395
3.815
0.056
12.481
R
od50
-0.090
-0.076
-1.977
4.656
0.086
-7.415
R
ov125
0.111
0.100
3.001
13.382
0.046
26.648
R
ir125
-0.010
0.001
-2.264
7.068
0.073
-1.595
R
is125
0.049
0.005
-7.381
76.552
2.945
0.185
R
er125
0.001
-0.001
-0.049
1.159
0.030
0.558
Notes: The first column indicates the name of the variable. R
mstands for market return, R
ostands for oil price return, R
oustands for oil price return up, R
odstands for oil price return down,
R
ovstands for oil volatility return, R
irstands for interest rate return, R
isstands for interest
spread return and R
erstands for exchange rate return. The second column indicates the number
of data points (returns). The third to sixth row indicate the mean, median, skewness, and kurtosis
for the returns of the specific variable. The last two columns indicate the standard deviation in
returns and the T-statistic of the return distributions, which gives an idea of the significance.
23
