Dependence between oil price changes and sectoral stock returns in Pakistan: Evidence from a quantile regression approach

Abstract

This study explores the dependence between changes in world crude oil prices and the performance of the Pakistan Stock Exchange, at the aggregate as well as sectoral levels for the period from July 1997 to December 2016. Quantile regression approach is employed for a detailed examination of the structure and degree of dependence for three sub-periods corresponding to normal, rising, and falling oil price periods. We found that the dependence between changes in crude oil price and the sectoral stock returns is heterogeneous across industries and it exists in both bullish and bearish market trends. The dependence at the upper and lower quantiles is found to be a common feature across industries. Moreover, the dependence and direction of the relationship change at times of structural breaks. The findings highlight an external channel through which fluctuations in stock returns may impede the liquidity of the stock market of an oil-importing country such as Pakistan, thereby affecting the domestic economy.

Keywords

Crude oil stock returns quantile regression approach structural breaks Pakistan

Highlights

This study examines the dependence between changes in world crude oil prices and the performance of the Pakistan Stock Exchange.

We analyze structural breaks by using the Bai and Perron¹ model.

The degree of dependence is captured through Koenker and Bassett’s² quantile regression approach.

The degree of dependence is found heterogeneous across industries in both bearish and bullish market trends.

As a robustness test, the Vector Auto-Regressive (VAR) model is used to address the endogeneity issue in oil prices and stock returns nexus.

Introduction

On January 25, 2016, the Wall Street Journal reported:³

“The correlation between the price of Brent oil and the S&P 500-stock index is at levels not seen in the past 26 years—the correlation is 0.97, higher than any calendar month since 1990.”

This interesting observation raises a number of questions: Is there any dependence between the prices of oil and stocks? Is the association mainly a developed-economy phenomenon or could it be found in emerging and frontier stock markets as well? Do stocks of certain industries respond more to changes in oil prices than others?

Existing studies show a strong link between stock market returns and oil price shocks. On the one hand, studies documented a negative association.^4–8 On the other hand, researchers pointed out a positive relationship^9–12 or even evidence for no significant relationship.^13–15

The debate on oil prices and stock market returns nexus is started after the seminal paper of Jones and Kaul¹⁶ and Sadorsky.¹⁷ Subsequent researchers, for instance, Driesprong et al.¹⁸ conducted a study on a sample of developed and emerging countries and reported that a rise in oil prices drastically lowers future stock returns. In contrast, Ono¹⁹ examined the impact of oil prices on real stock returns for BRICs countries over the period from 1999 to 2009, and found a positive correlation between real stock returns and oil price indicators for BRICs countries except for Brazil. Similarly, Narayan and Narayan²⁰ explored the relationship among oil prices, stock prices, and nominal exchange rates for Vietnam during the period from 2000 to 2008 and reported that oil prices have a positive and significant impact on the Vietnamese stock prices. More recently, Wen et al.¹² explored oil price and stock market returns nexus by using dependence-switching copulas and time-varying single copulas in a broad range of emerging economies and reported positive crude oil-emerging stock markets linkages. Likewise, Tchatoka et al.²¹ investigated the oil price shocks - stock market returns relationship for 15 oil-importing and oil-exporting countries. They find a significant effect of oil price shocks on stock market returns.

Another strand of the literature studies the linkages between oil prices and sectoral stock returns. For instance, Zhu et al.²² explored the relationship between Chinese real sector stock returns and crude oil price changes for the period from March 1994 to December 2013. They found that the sensitivity of the relationship varies across time and industries. In the same vein, Kilian and Park²³ reported that the response of equity returns to oil price shocks varies significantly across industries depending on requisite rates of returns. It has also been shown that the response of sectoral or market level stock returns to oil price shocks may be negative or positive depending on whether the country is a net consumer or net producer of oil resources.^13,24 Kilian²⁵ reported that oil price shocks impact the real economy differently depending on whether oil shocks are associated with global demand shocks or global supply shocks. Bouri et al.²⁶ tested the link between oil prices and stock returns at sectoral level for Jordan during the Arab Uprising instigated in 2010 and noted an asymmetric effect across the sectors. Gokmenoglu and Fazlollahi²⁷ examined the association between commodity and financial markets and found a short-run impact of oil price changes on the stock market, particularly energy sector companies. Recently, Yun and Yoon²⁸ investigated the impact of oil price changes on returns of the airline industry. They found a strong effect of volatility spillover between the prices of airline stocks and crude oil as compared to the return spillover effect.

The aforementioned studies suggest a lack of consensus, thereby highlighting the need for country-specific examination of the nexus between oil price changes and stock returns.

In this study, we examined the relationship between crude oil price changes and the performance of the Pakistan stock market. Pakistan is a developing economy. The country’s Stock Exchange (PSX) has been one of the best performing stock markets over the past 15 years. It comprises of 663 listed companies with a total market capitalization of about US$75 billion. The impact of changes in world crude oil prices on the performance of PSX is examined both at the aggregate as well as the sectoral level. The significance of the current research lies in the fact that Pakistan meets over 85% of its crude oil demand through imports. A substantial change in world oil prices could therefore significantly affect the country’s stock market. This could happen chiefly in two ways. First, existing studies suggest that oil price shocks can lead to inflation, especially for oil-importing countries see for example.²⁹ Second, currently, general sales tax and petroleum levy are the two major categories of taxes that the government collects on oil-related products. The government charges 17% general sales tax on petroleum products compared to the 31% tax charged on high speed diesel mostly used in transport and agriculture sectors. Thus, the changes in oil prices may have a heterogeneous influence across industries in an oil-importing country like Pakistan.

The following trends motivate our research. First, the association between the dynamics of oil market and Pakistani stock market has so far not been examined from a sectoral perspective. Second, extant research work has mostly used simple regression model,³⁰ cointegration analysis,^29,31–33 and GARCH models³⁴ to value the dependence between oil price changes and stock markets in emerging markets such as Pakistan. In contrast, this study employs the quantile regression approach which offers a comprehensive view of the oil price changes and stock market returns that vary across different quantiles.^35,36 Using quantile regression, we can infer information on the co-movement between oil price changes and stock market returns in specific market conditions such as the time when the market is bullish (upper quantile), or bearish (lower quantile), or normal (intermediate quantile) where the market is neither bearish nor bullish.^37,38 Third, in the pre- and post-crisis periods, stock markets may react differently at the sectoral level due to heterogeneity among sectors. Thus, we obtained distinctive structural breaks at the sectoral level to achieve a complete picture of changes in the dependence caused by extreme and irregular events. Addressing structural changes is imperative because oil prices encounter wild swings due to extreme and irregular events such as financial turmoils (for example, the 1998 Russian crisis, the Global financial crisis 2008), or the 9/11 terrorist attacks. Moreover, during our sample period, two major oil price wars occurred. The first war took place between Saudi Arabia and Venezuela from November 1997 to March 1999. The second war started with an OPEC meeting in Vienna when Saudi Arabia implemented a policy of pump-at-will despite weak demand in response to the US shale revolution and production cuts of non-OPEC countries. According to Bloomberg calculations, oil prices collapsed during these price wars from $20 per barrel to below $10 and about $100 per barrel to $27 respectively. From an econometrics point of view, ignoring the structural breaks could lead to misleading conclusions about the integration order if it present in time series.³⁹

To the best of our knowledge, this is the first study on the impact of crude oil price changes on sectoral and aggregate stock returns in the Pakistani market. The analysis is conducted on the aggregate stock index as well as on 12 major industrial sectors using monthly oil price and stocks data for a 20-year period (July 1997–December 2016). We obtained our estimation results by employing the quantile regression approach regarding the dependence between oil price changes and stock returns. We analyzed corresponding structural breaks by using the Bai and Perron¹ model. The results of this study are compared with those obtained for three sub-periods: normal, high, and falling crude oil prices.

Pakistan’s stock market performance and oil imports

Figure 1 below shows the monthly price behavior for the Brent oil and Pakistan stock market index for the 1997–2016 period. The index rose sharply in the early 2000s. This phase ended with the onset of the global financial crisis and abrupt movement of oil prices in 2008. The Securities and Exchange Commission of Pakistan (SECP) had to freeze the index at the 9,144 points level. The index fell substantially once the floor was removed. The index has risen since that period and reached its record level of over 45,000 during the year 2016.

Figure 1.

Monthly prices behavior for Brent oil and PSX index (1997–2016).

Interestingly, both markets (the Brent oil and Pakistan stock market index) exhibit opposite trends during 2014–2015. A possible explanation for this observation could be the sharp fall in oil prices due to the second oil-price war which lasted for 22 months from November 2014 to August 2016. It started in November 2014 with an OPEC meeting in Vienna. Tired of non-OPEC countries freeloading on the cartel’s production cuts, and worried about the impact of the United States (U.S.) shale revolution, Saudi Arabia adopted a policy of pump-at-will in spite of weak demand. Crude oil price collapsed from about $100 per barrel to $27 during the period. Given Pakistan’s low domestic production and subsequent dependence on oil imports, the majority non-oil and gas stocks listed on the Pakistani stock market may reflect an opposite reaction.

In 2014, Pakistan’s import of oil and petroleum products stood at US$47.4 billion and accounted for 37% of the country’s electricity production (World Bank, 2016). The country’s oil imports were mostly stable during the 1990s as economic growth was low and local gas reserves satisfied the bulk of the country’s energy needs (Figure 2). Oil imports rose during the high-growth period of the mid-2000s. The rise in world oil prices further inflated the import bill. Oil imports slowed down during the global financial crisis but have since picked up again.

Figure 2.

Pakistan’s crude oil imports (1990–2013).

Data and methodology

Data description

The dataset for this study consists of monthly observations for crude oil prices and aggregate and industry-wise stock prices for the period from July 1997 to December 2016. Data on stock prices are based on 218 ordinary stocks about twelve non-financial industrial sectors listed on the Pakistan Stock Exchange. The industries covered include automobile and parts, chemicals, construction and materials, electricity, industrial engineering, financial, food producers, household goods, industrial mining, oil and gas, personal goods (textile), and pharmaceuticals.

Data on the prices of Brent oil are collected from the Thomson Reuters File (U.S. Energy Information Administration). Data for the stock market are obtained from the web site of Yahoo Finance while industry-level stock prices are taken from the web sites of Pakistan Stock Exchange (PSX) and Business Recorder.

The resulting data set comprises of 234 monthly observations for Brent crude oil, sectoral stock returns, and stock market returns. Summary statistics for the dataset are given in Table 1. Skew, Kurt and Obs. indicate the skewness, kurtosis and number of observation, respectively. Skewness captures the asymmetry in the distribution of the sample data while kurtosis represents the peakness of the data. A kurtosis value of 3 corresponds to a normal distribution, and the pattern is considered mesokurtic. If the value is greater than 3, then the pattern is leptokurtic, which is associated with a peaked, fat-tailed distribution. A kurtosis value of less than 3 is referred to as platykurtic and is associated with a less peaked distribution and thin tail. Several variables show a leptokurtic behavior and have negative skewness. Moreover, a fat tail and sharp peakedness are found in the case of kurtosis. Hence, non-normally distributed data regression estimation may produce incorrect results. Consequently, we follow quantile regression estimation in this study.

Table 1.

Summary statistics.

Industry	Number of firms	Mean	Median	Std. Dev.	Skew	Kurt	Obs.	ADF	Cor. with Brent Oil
Chemical	18	0.049	0.030	0.070	3.420	9.834	234	−19.434***	−0.570
Const. & Material	24	0.027	0.027	0.001	2.230	25.869	234	−12.678***	−0.723
Oil & Gas	10	0.029	0.015	0.002	3.520	16.238	234	−15.231***	0.231
Engineering	7	0.028	0.028	0.003	−1.834	15.696	234	−15.255***	0.800
Automobiles & Parts	12	0.034	0.034	0.001	−2.151	22.526	234	−14.500***	0.539
Food Producers	19	0.034	0.033	0.001	1.048	11.746	234	−15.862***	−0.671
Households Goods	5	0.032	0.032	0.001	−2.004	21.721	234	−14.628***	0.295
Electricity	9	0.034	0.034	0.001	−3.528	29.780	234	−17.202***	0.278
Pharmaceuticals	5	0.035	0.035	0.001	4.102	42.089	234	−17.392***	0.669
Textiles	81	0.039	0.039	0.001	−3.636	31.536	234	−15.343***	0.562
Industrial Mining	7	0.041	0.041	0.001	2.156	13.659	234	−16.142***	0.150
Financial	21	0.042	0.041	0.001	−3.803	35.970	234	−15.976***	−0.841
Total	218
Overall		0.033	0.019	0.003	2.231	5.614	234	−14.021***	0.702
Brent Oil		0.004	0.004	0.001	−0.366	2.105	234	−12.183***	1.000

Note: This table presents the summary statistics for twelve industries, the Pakistan stock market (PSX), and the crude oil market (Brent Oil) based on monthly data for July 1997 to December 2016. ADF test values are significant at first difference. Finally, the last column reports the results for the correlation of sectoral stock returns and overall market returns with oil price changes.

Moreover, we test unit-roots for each series on raw data using the Augmented Dickey-Fuller (ADF) statistic.⁴⁰ We start testing unit roots for each series at level, and find all variables non-stationary at level, implying that the mean of the data is not constant over time. In the next step, we check the unit roots at first difference based on logarithmic transformation of the series. The result for the unit root test shows that all variables are stationary at first difference. This confirms the use of quantile regression instead of the standard OLS regression, given that the latter assumes data to be stationary at level.^a

The last column of Table 1 reports the results for the correlation of sectoral stock returns and overall market returns with oil price changes. The results show that a strong correlation exists between crude oil and sectoral returns such as construction and materials, engineering, food producers, automobiles, financial, pharma, and textiles. Moreover, overall stock market returns are strongly correlated with crude oil price changes. This is consistent with the statement documented by the Wall Street Journal mentioned earlier. Figure 1 also exhibits the strong correlation of monthly price behavior for crude oil and PSX index.

Econometric methodology

The following empirical strategy is employed:

In the first step, we obtained changes in oil prices ( $p_{o t}$ ) by taking the first difference of the natural logarithm of the oil prices. The mathematical expression is as follows:

p_{o t} = \ln ({O P}_{t} - {O P}_{t - 1})

where,

{O P}_{t}

and

{O P}_{t - 1}

are current and one-period lagged prices, respectively.

In the next step, we analyzed corresponding structural breaks by using Bai and Perron model. Further, we use the quantile regression technique to analyze the dependence between oil price changes and stock returns. The quantile regression technique is considered more appropriate than the classical regression model when the distribution of the dependent variable is highly non-normal.^36,38 Besides, considering only means using simple regression may lead to the omission of important relationships.⁴¹ In our case, the data is non-normally distributed. Hence the use of quantile regression is better suited to capture the dependence between variables of interest. More importantly, the quantile regression model answers the essential question as to whether shocks to oil prices influence stock returns in a different way for the bearish market (when returns are low) as compared to the bullish market (when returns are high).

The quantile regression approach suggests that the value of $ε_{i}$ conditional on the regressors in $τ$ th quantile is zero. We postulate a model for the $τ$ th quantile of the sectoral and market returns ( $y_{i}$ ) as a function of oil price shocks ( $x_{i}$ ) as:

Q_{y_{i}} (τ| x_{i}) = α^{τ} + x_{i}^{'} β^{τ}

(1)

where,

Q_{y_{i}} (τ| x_{i})

is the conditional

τ^{t h}

quantile of

y_{i}

that is the variable of interest (sectoral and market returns),

α^{τ}

is the intercept, which is allowed to depend on

τ

β^{τ}

is the estimated parameter in the equation associated with

τ^{t h}

quantile, and

x^{'}

contain variables supposed to affect the dependent variable, i.e. oil price changes.

The coefficients of $τ^{t h}$ quantile of the conditional distribution is computed as a solution to the minimization problem of the weighted deviations from the conditional quantile (Koenker and Bassett)² as:

_{\hat{β} ϵ R^{P}}^{\arg} \min \sum_{i = 1}^{n} ρ_{τ} (y_{i} - α^{τ} - x_{i}^{'} {\hat{β}}^{τ})

(2)

Where, $ρ_{τ} (u) = u (τ - I (u < 0))$ are a check function and $u = y_{i} - α^{τ} - x_{i}^{'} {\hat{β}}^{τ}$ is an indicator function. The estimator is vigorous as this approach spits the residuals into positive and negative values and provides weights of $τ$ as well as $1 - τ$ .

While examining whether changes in oil prices impact stock returns differently at a sectoral and overall market level during financial turmoil, we employed the structural breaks test. We modify equation (1) as:

Q_{r} (τ| x) = α (τ) + β (τ) o i l + \sum_{j = 1}^{n} δ_{j} (τ) D_{j} o i l

(3)

Where $r$ shows the monthly sectoral and market returns, $oil$ corresponds to changes in oil prices and $D_{j}$ indicates the dummy variable that takes on the value “1” if the returns refer to the period next to the eruption of $j - t h$ break and zero if prior to that. The dummy variable modifies the effects of changes in oil prices on sectoral and stock market returns. This model tests the dependence between oil price changes and sectoral stock returns in Pakistan. Following extant literature,^35,37,42 we consider seven quantiles “0.05, 0.10, 0.25, 0.50, 0.75, 0.90, and 0.95” classified as low “0.05, 0.10, and 0.25”, moderate “0.50” and high “0.75, 0.90, and 0.95”. $β (τ)$ captures the marginal changes of the different conditional variables in the non-break sub sample period for each quantile. While $δ_{j} (τ)$ captures the changes in dependence occurred by the $j - t h$ break. The values of $β (τ)$ and $δ_{j} (τ)$ for all quantiles, estimates the structure of dependence in the normal period and in the crisis period, respectively. The dependence decreases (increases) when $δ_{j} (τ)$ is negative (positive) in the crisis period.

The estimations are conducted both for the aggregate stock market index as well as for 12 categories of stocks. We categorize the full sample period examined into three sub-periods to take into account the spectacular fluctuations in oil prices that took place during the period. Oil prices rose from $50 in January 2007 to over $140 in July 2008 (Figure 1). Later on, with the onset of the global economic slowdown from 2008 to 2010, oil prices fell substantially. We examine the following three sub-periods:

The first period can be called the normal period (January 2000–December 2006), the second is the rise period (January 2007–July 2008), and the third is the crisis and the fall period (October 2008 onwards).

Baur⁴³ argues that the model without dummy variable $D_{j}$ means that all the dependence is captured by $β (τ)$ . In our study, the dependence may get altered by financial turmoil because of the difference in dependence for crisis and normal periods.

Our results are consistent with Moya-Martinez et al.⁴⁴ The F-statistic of Andrews et al.⁴⁵ is used to examine the null hypothesis with one break compared to the unknown timing. Using breakpoint $j$ our study then relates the Ordinary Least Squares residuals ${\hat{e}}_{j}$ from one regression for each subsample and the residuals $\hat{e}$ from the overall model as:

F_{j} = \frac{{\hat{e}}^{'} \hat{e} - {\hat{e}}_{j}^{'} {\hat{e}}_{j}}{{\hat{e}}_{j}^{'} {\hat{e}}_{j} / (n - 2 k)}

(4)

Where,

j = n_{i}, \dots, n - n_{i} (n_{i} \geq k)

The null hypothesis of the supF test assumes that there is no structural break and vice versa for the alternate hypotheses. This model is extended by Bai and Perron to identify the 0 versus L breaks and L versus L + 1 break. We also use the model presented by Bai and Perron for structural breaks identification.

The results of the test for structural breaks are shown in Table 2. Most sectors have two to three breaks. Significant heterogeneity in the number of structural breaks is found among different industrial sectors. Chemical, construction and material, oil and gas, engineering, food producers, household goods, pharma, textile, and industrial mining have two-breaks whereas the remaining sectors have three-breaks. Further, we divided the full period of each sector and overall market into sub-periods on the basis of the occurrence of these breaks.

Table 2.

Multiple structural breakpoint tests.

Sector	Breaks	Break-dates
		Pre-crisis	Post-crisis
		Suite 1	Suite 2	Suite 3
Chemical	2	2007/03	2011/01
Const. & Material	2	2007/03	2011/01
Oil & Gas	2	2007/07	2011/04
Engineering	2	2007/06	2011/01
Automobiles & Parts	3	2007/03	2011/01	2014/02
Food Producers	2	2007/09	2011/01
Households Goods	2	2007/10	2011/01
Electricity	3	2007/03	2011/01	2014/02
Pharma	2	2007/09	2011/01
Personal Goods (Textiles)	2	2007/03	2011/01
Industrial Mining	2	2007/11	2011/01
Financial	3	2007/03	2011/01	2014/02
Overall	3	2009/05	2011/05	2013/07

Note: This table reports the calculation of structural breaks for twelve industries, the overall market, and the oil market.

The first breakpoint pertains to the sub-period from 2007/03 to 2007/11. This breakpoint occurred during the pre-crisis period. Oil prices began their steep increase about that time, suggesting that the rise in oil prices may have influenced the industry stock market returns. The second breakpoint occurs from 2011/01 to 2011/05 while the third breakpoint occurs from 2013/07 to 2014/02.

Results and discussion

Baseline results

The results for the test of quantile regression are reported in Table 3. Following the quantile regression literature,^35,36,38 we show numerical results for seven quantiles from 0.05 to 0.95 while considering the global financial crisis. Our estimation results show that sectoral dependence on oil price changes is heterogeneous and the degree of dependence in some sectors is more severe as compared to others. Thus, these results validate the importance of exploring the dependence between stock market returns at the sectoral level and crude oil price changes. The marginal effects ignore the breaks (i.e., $β$ values) for each sectors and overall market.

Table 3.

Results for the quantile regression estimation.

Industry		$Q_{0.05}$	$Q_{0.1}$	$Q_{0.25}$	$Q_{0.5}$	$Q_{0.75}$	$Q_{0.9}$	$Q_{0.95}$
Chemical	$β$	−0.370	−0.130	−0.124	0.512***	0.754***	0.946***	1.067***
	$δ_{1}$	0.148***	0.110***	0.122**	0.051	0.074***	0.083	0.138
	$δ_{2}$	−0.038**	0.048***	0.072**	0.040	0.130***	0.205***	0.217***
CM	$β$	−1.055***	−0.921***	−0.423***	−0.624*	−1.037***	−0.946***	−1.06***
	$δ_{1}$	0.078	−0.018	−0.190***	−0.178***	−0.067*	0.102***	0.089
	$δ_{2}$	0.129*	0.050	−0.008	0.018	0.074	0.169***	0.184***
Oil & Gas	$β$	−0.425**	−0.721***	−0.506***	1.568***	2.142***	2.610***	2.702***
	$δ_{1}$	−0.002*	−0.001*	0.004	0.009*	0.018*	0.032*	0.090
	$δ_{2}$	0.003*	0.006*	0.011*	0.025*	0.039*	0.063	0.125*
Engineering	$β$	0.210	0.271	0.558	0.907	1.791***	1.711***	1.397***
	$δ_{1}$	0.835***	0.779***	0.769***	0.613***	0.276***	0.242***	0.254***
	$δ_{2}$	0.343***	0.312***	0.234***	0.241***	0.693***	0.834***	0.863***
Auto Parts	$β$	1.670***	1.738***	1.801***	1.668*	1.499***	1.551***	1.576***
	$δ_{1}$	0.433***	.460***	0.447***	0.452***	0.237***	0.215	−0.267***
	$δ_{2}$	0.222***	0.190***	0.131*	0.117*	0.052	0.073***	0.147***
	$δ_{3}$	0.109*	0.162**	0.090**	0.061***	0.061***	0.087***	0.071***
Food Producers	$β$	−0.643	−0.497	−0.381	0.360***	0.282**	0.115*	0.318**
	$δ_{1}$	0.071	0.008	0.007***	0.006***	−0.257***	−0.470***	−0.568***
	$δ_{2}$	0.0009*	0.022*	0.028**	0.025*	0.025*	0.042*	0.046*
Household Goods	$β$	−1.633**	−1.434***	−1.235***	−1.227**	−1.706***	−2.366***	−2.338***
	$δ_{1}$	0.588***	0.595***	0.416***	0.342***	0.745***	0.816***	0.872***
	$δ_{2}$	0.557***	0.532***	0.501***	0.166***	0.142***	0.204***	0.220***
Electricity	$β$	−0.568***	−0.452**	−0.456**	0.234	0.335***	0.444***	0.482***
	$δ_{1}$	0.371***	0.384***	0.521***	0.410***	0.305***	0.347***	0.208***
	$δ_{2}$	−0.256***	−0.289***	−0.395***	−0.449***	−0.486***	−0.565***	−0.526***
	$δ_{3}$	0.254***	0.242***	0.209***	0.197***	0.184***	0.186***	0.174***
Pharma	$β$	−0.473	−0.413	−0.371	−0.263	−0.122	−0.027	−0.130
	$δ_{1}$	0.489***	0.448***	0.377***	0.330***	0.319***	0.307***	0.247***
	$δ_{2}$	0.163**	0.173***	0.226***	0.214***	0.195***	0.265***	0.298***
Textile	$β$	1.458***	1.095***	0.919***	1.085***	1.247***	1.432***	1.279***
	$δ_{1}$	0.392***	0.312***	0.277***	0.279***	0.287***	0.251***	0.202***
	$δ_{2}$	0.339***	0.349***	0.310***	0.267***	0.225***	0.223***	0.201***
Industrial Mining	$β$	−0.947***	−0.813***	−0.417***	0.220	0.473***	0.736***	0.765***
Industrial Mining	$δ_{1}$	−0.223***	−0.187***	−0.132***	0.122***	0.125	0.136**	0.172***
	$δ_{2}$	0.044	0.054**	0.091***	0.133***	0.179***	0.123**	0.071*
Financial	$β$	−0.934***	−0.902***	−0.835***	0.287	0.156	0.167	0.209
	$δ_{1}$	−0.109**	−0.107***	−0.121**	−0.289***	−0.359***	−0.380***	−0.380***
	$δ_{2}$	0.088**	0.095**	0.073	0.058	0.061***	0.054***	0.050**
	$δ_{3}$	0.050	0.076**	0.054	0.089*	0.091***	0.084***	0.062*
Overall	$β$	−0.464***	−0.555***	−0.198***	0.876**	2.606***	2.814***	2.941***
	$δ_{1}$	0.002**	0.005**	0.002***	0.001**	−0.025	−0.022**	−0.018
	$δ_{2}$	0.008*	0.003	−0.004*	−0.009*	−0.016*	−0.040**	−0.067*
	$δ_{3}$	3.512	−0.002*	−0.005*	0.014	0.022**	0.034*	1.457**

Note: This table exhibits the results of quantile regression estimation for twelve industries, Pakistan stock market (PSX), and crude oil market (Brent Oil) based on monthly data for July 1997 to December 2016. ***, **, * show that 1%, 5% and 10% levels, respectively.

First, we discuss key findings regarding sectoral dependence on oil price changes. The impact of changes in oil prices is positive and statistically significant for moderate and upper quantiles in chemical and food production sectors (bullish market trend) while significant and negative for lower quantiles in the financial sector (bearish market trend).

In the case of oil and gas sector, the effect of oil price changes is significant and positive for moderate and upper quantiles (bullish market trend) while negative for lower quantiles (bearish market trend). Similarly, the impact of changes in oil prices is positive and significant for upper quantiles (bullish market trend) and negative for lower quantiles (bearish market trend) in the industrial mining sector. Likewise, in the electricity sector, the impact of changes in oil prices is negative and statistically significant for lower quantiles (bearish market trend) and positive for upper quantiles (bullish market trend).

The impact of changes in oil prices is found positive and statistically significant for all seven quantiles in the automobile and textile sectors. While a significant and negative impact of changes in oil prices is observed for all seven quantiles in construction and materials as well as the household goods sectors. Furthermore, impact of changes in oil prices on returns of the engineering sector are found positive and significant for upper quantiles (bullish market trend) while no significant effect is found in case of the pharma sector.

Moreover, in the case of overall stock market returns, the effect of oil price changes is significant and negative for lower quantiles and positive for moderate and upper quantiles. During bullish market trends, the dependence is higher as the intensity of co-movement between the overall market and oil prices changes increases for upper quantiles. From this, we can infer that the dependence of stock returns on changes in oil prices for the Pakistani market varies across sectors from low to high quantiles.

Overall, the dependence exists in both bearish and bullish market trends. These results reveal that increase in oil prices during normal and bullish market trends increase stock returns in the Pakistani market while in the case of bearish market trends increasing oil prices lower the stock returns. One possible justification for these findings is that during bullish market trends, the economy also grows fast, thereby offsetting the negative effects of increasing oil prices. Our results for bullish market trends are consistent with the findings of Mensi et al.³⁸ who reported dependence between oil price changes and equity returns among BRICS economies.

In the case of bearish market trends, in particular, the results are consistent with the findings of Hu⁴⁶ who documented that dependence across financial markets is left-tailed and asymmetric. The reason for this asymmetry could be the irrational and herd behavior of investors in the Pakistani market as investors may react less rapidly to good news than they do to bad news. Javaira and Hassan⁴⁷ reported that a herd behavior was observed in the Pakistani stock market at the onset of liquidity crisis in March 2005 due to information asymmetry among investors, speculation and a local mechanism of financing (Badla financing).

Findings for structural breaks

The value of $δ_{j}$ shown in Table 3 captures the changes in the dependence that occurred by the first, second, or third break. We also illustrate graphical results for all the quantiles and conditioning variables for the Pakistani stock market and sectoral returns in Figure A1 in Supplemental material. Figure A1 depicts the dynamic trace of quantile regression coefficients for all structural breaks. If we look at the chemical sector, the first break has positive and statistically significant effects from lower-level quantiles to upper-level quantiles with steady behavior while the second break has the same significant effects but shows a volatile behavior. Likewise, in the case of the oil and gas sector, we found a significant and negative effect for lower-level quantiles, while positive and significant effects at moderate and upper-level quantiles. In the second structural break, the effect is significant and positive for all quantiles. The engineering sector shows significant and positive effects with a decreasing trend for the first break and same positive effect in the upward direction. In the case of the automobile, electricity and financial sectors, three breaks are identified. In the automobile sector, the significant and positive effect is found for lower and moderate quantiles while the significant negative effect is found for one upper quantile in the first break. Besides, break 2 and break 3 exhibit a significant and positive effect. Finally, the electricity and financial sectors show significant and positive effects for all quantiles in three breaks.

In the case of the overall market, the three breaks show different effects for lower, moderate, and upper quantiles. In the first and second breaks, the effect is significant and positive for lower and moderate quantiles while negative for upper quantiles. In the case of the third break, the effect is significant and positive for one lower quantile and negative for the next three lower quantiles. Moreover, the effect is significant and positive for moderate and upper quantiles.

All sectors have two to three breaks with significant heterogeneity in the number of structural breaks present among different industrial sectors and different effects of dependence. Our findings are in-line with previous work on the said association, see for instance,^48,49 among others. These studies argued that the heterogeneous effects for industries could be due to the different industrial attributes such as competition, cost structures, regulations, and dependence on oil. Hence, our findings conclude that portfolio managers, investors, and policy makers should be aware of the dependence between the Pakistani stock market and oil price changes in both bearish and bullish market trends.

Robustness test: endogeneity issue

The results reported in the preceding section provide a significant relationship for changes in oil prices and returns of stocks. However, one might argue that this effect might be the result of endogeneity. The extant literature has found evidence of the endogeneity problem caused by reverse causality between oil prices and stock returns.^50,51 Thus, to address the endogeneity issue, we follow Masih et al.⁵² and used a Vector Auto-Regressive (VAR) model. VAR allows us to examine the endogeneity caused by reverse causality between oil prices and stock returns.

Following recent studies, for instantce, Wang and Zhang,⁵³ we performed unit root and co-integration tests for time series properties before moving to the VAR model. All variables are fund to be integrated at first difference implying that variables have a unit root. Cointegration test proposed by Johansen and Juselius⁵⁴ is used with both the maximum Eigenvalue and the trace tests. Cointegration test results suggest no association between changes in oil prices and returns of stock in the long-run however the results are not documented for brevity purpose. Consequently, to analyze the dependence between changes in world crude oil prices and the performance of the Pakistan Stock Exchange both at the aggregate as well as sectoral level, we obtain “Generalized Impulse Response Functions” GIRF through using the VAR model. Furthermore, the Granger-causality test is employed to obtain the unidirectional causality between the returns of stocks and changes in oil prices. Existing studies, for example, Diaz and Perez de Gracia¹⁰ and Cunado and Perez de Gracia,⁸ among others use the VAR model for estimating the influence of changes in oil prices and stock returns.

Figure A2 in Supplemental material exhibits the generalized impulse-response functions of aggregate as well as sectoral level stock returns to oil price shocks for 1997 to 2016. The graph shows a significant dependence of stock returns on oil price changes and it varies across industries. The results of the VAR model are qualitatively similar to the findings of quantile regression for the dependence of stock returns and oil prices.

Table 4 presents the results of the Granger-causality test. Our estimation results show the rejection of the null hypothesis for no causality for several industries with oil price changes.

Table 4.

Granger causality test for the full sample.

Chemical→Brent	1.3348	Electricity→Brent	0.5559
Brent→Chemical	5.7128***	Brent→Electricity	11.4788***
Const. & Material→Brent	0.0648	Pharmaceuticals→Brent	1.0985
Brent→Const. & Material	2.6196*	Brent→Pharmaceuticals	14.7991***
Oil & Gas→Brent	1.4947	Textiles→Brent	0.9128
Brent→Oil & Gas	0.1811	Brent→Textiles	1.8471***
Engineering→Brent	0.3344	Industrial Mining→Brent	1.1797
Brent→Engineering	5.9017***	Brent→Industrial Mining	4.3770**
Automobiles & Parts→Brent	1.5348	Financial→Brent	0.0087
Brent→Automobiles & Parts	0.2533	Brent→Financial	12.4930***
Food Producers→Brent	4.5034	Overall→Brent	0.2614
Brent→Food Producers	3.0937**	Brent→Overall	12.6091***
Households Goods→Brent	0.9857
Brent→Households Goods	1.3136

Note: Rejection of null hypothesis at 1%, 5% and 10% is presented by ***, ** and *.

Hence, the causality exists between stock returns and changes in oil prices in uni-direction for companies listed on the Pakistan stock exchange. This is in line with the results of Diaz and Perez de Gracia.¹⁰

Conclusion and policy recommendations

In this study, we examined the effects of crude oil price changes on the emerging stock market of Pakistan at both sectoral and market-level using monthly data for 20 years (July 1997 to December 2016). We obtained changes in oil prices by natural-logarithmic subtraction of lagged period prices from the current period prices and then dependence between returns of stocks and changes in oil prices was analyzed using the quantile regression approach. The period of estimation was divided into three sub-periods: normal oil prices period (January 2000–December 2006), rising oil prices period (January 2007–July 2008), and falling oil prices period (October 2008 onwards). Since Bai and Perron model has been proven effective to analyze corresponding structural breaks, we applied this model because we believe that structural breaks may have been caused by a large event of the 2008–09 global financial crisis included in our sample period. We observed some interesting results; three of them are more salient:

We found that overall the dependence between changes in crude oil price and the sectoral stock returns is heterogeneous across industries. All sectors have two to three breaks.We found significant heterogeneity in the number of structural breaks among different industrial sectors. Two breaks were found for the chemical, construction and material, oil and gas, engineering, food producers, household goods, pharma, textile, and industrial mining sectors while three breaks were found for the other sectors.

Our estimation results for the quantile regression approach showed that sectoral dependence on oil price changes is heterogeneous and the degree of dependence in some sectors is more severe as compared to others. These results corroborate the findings of previous studies which report that the nature of oil shocks can be negative or positive for stock returns of specific industry depending on whether the industry is a net importer or net exporter of of oil resources.^13,23,24

We found that the dependence exists between changes in oil price changes and Pakistani sectoral returns at both the lower and upper quantiles as well as in both the bearish and bullish market trends.

Our findings can help different stakeholders in their decision making. Particularly, the results of this study have implications for investors, managers, and analysts involved in portfolio selection, risk management, or asset allocation. The study highlights an external channel in the form of world oil price changes through which fluctuations in stock returns may impede the liquidity of the stock market of an oil-importing country like Pakistan thereby hurting the domestic economy. This is a serious concern in Pakistan. The country reduce the import of oil and encourage the use of indigenous resources. The consumption of domestically-produced natural gas in Pakistan has grown by 5.1 percent annually during the past ten years in addition to the higher consumption of local coal (Pakistan Economic Survey 2016–17). This attempt at substituting the consumption of imported oil may have helped dampen the effect of world oil price changes on stock returns of the firms included in the Pakistan Stock Exchange.

This study also has important industry-specific policy recommendations. Since our study is based on a sectoral perspective, implications for portfolio investments involving different equity sectors can be deduced. Moreover, portfolio investors should carefully account for the differential effect of oil price changes on different sectors. Our findings show that sectoral dependence on oil price changes is heterogeneous and the degree of dependence in some sectors is more severe as compared to others. Because, in production process, oil is considered as an important input; therefore, shocks to oil prices may adversely impact the cost of production which in turn affect firms’ profits and stock price. Besides, shocks in oil prices have different effects on stock returns of firms in different sectors thereby causing the transfer of wealth from oil consumers to oil producers. This response of sectoral returns to oil price shocks provides information to agents about the sectors of the stock markets in which to invest in order to optimize risk.

Footnotes

Author contributions

All authors contributed equally. All authors read and approved the final manuscript.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Junaid Ahmed

Note

a. The ADF tests the null hypothesis that a time series is I (1) against the alternative hypothesis that it is I (0). The ADF estimating the following regression: $Δ Y_{t} = β_{1} + β_{2} t + δ Y_{t - 1} + α_{i} \sum_{i = 1}^{m} Δ Y_{t - i} + ε_{t}$

If the null hypothesis is rejected, it means that the variable is stationary, whereas accepting the null hypothesis means the variable is non-stationary at that level. To make it stationary the variable requires to be differenced.

Supplemental material

Supplemental material for this article is available online.

Mushtaq Hussain Khan is Lecturer in Banking and Finance, Department of Management Sciences, University of Azad Jammu & Kashmir, Muzaffarabad Pakistan. He received his PhD in Management Sciences, Finance, from Capital University of Science & Technology, Islamabad, Pakistan. His PhD dissertation covered three areas; corporate governance, corruption and dual banking systems. He has published his research work in reputed journals internationally such as Finance Research Letters and Soft Computing.

Junaid Ahmed works as Senior Research Economist at Pakistan Institute of Development Economics. He holds a PhD in Economics from Georg August University, Germany. His main research interest includes migration, remittances, and trade policy. He has published in numerous refereed journals.

Mazhar Mughal is Professor of Economics at Pau Business School, France and Associate Researcher at CATT, Pau University, France. He specializes in the empirical study of financial flows entering and exiting the emerging countries as well as issues pertaining to development, inequality and demographic change. He holds Habilitation in Empirical Economics (2016) and PhD in Development Economics (2012). He advises PhD and master students in Economics. His teaching experience includes various undergraduate and graduate-level courses in Economics and Statistics taught at French institutions.

References

Bai

Perron

Estimating and testing linear models with multiple structural changes. Econometrica 1998; 66: 47–78.

Koenker

Bassett

Regression quantiles. Econometrica 1978; 46: 33–50.

Stubbington

Kantchev

Oil, stocks at tightest correlation in 26 years. Wall Street J 2016; 25. https://www.wsj.com/articles/oil-stocks-dance-the-bear-market-tango-1453722783.

Westerlund

Sharma

SS.

Panel evidence on the ability of oil returns to predict stock returns in the G7 area. Energy Econ 2019; 77: 3–12.

Xiao

Zhou

Wen

, et al. Asymmetric impacts of oil price uncertainty on Chinese stock returns under different market conditions: evidence from oil volatility index. Energy Econ 2018; 74: 777–786.

Joo

Park

SY.

Oil prices and stock markets: Does the effect of uncertainty change over time?

Energy Econ 2017; 61: 42–51.

Diaz

Molero

de Gracia

FP.

Oil price volatility and stock returns in the G7 economies. Energy Econ 2016; 54: 417–430.

Cunado

de Gracia

FP.

Oil price shocks and stock market returns: evidence for some European countries. Energy Econ 2014; 42: 365–377.

Gupta

Oil price shocks, competition, and oil & gas stock returns—global evidence. Energy Econ 2016; 57: 140–153.

10.

Diaz

de Gracia

FP.

Oil price shocks and stock returns of oil and gas corporations. Finance Res Lett 2017; 20: 75–80.

11.

Luo

Qin

Oil price uncertainty and Chinese stock returns: new evidence from the oil volatility index. Finance Res Lett 2017; 20: 29–34.

12.

Wen

Bouri

Cheng

The crude oil–stock market dependence and its determinants: evidence from emerging economies. Emerg Markets Finance Trade 2019; 55: 2254–2274.

13.

Huang

Masulis

Stoll

HR.

Energy shocks and financial markets. J Fut Mark 1996; 16: 1–27.

14.

Chen

N-F

Roll

Ross

SA.

Economic forces and the stock market. J Bus 1986; 59: 383–403.

15.

Sukcharoen

Zohrabyan

Leatham

, et al. Interdependence of oil prices and stock market indices: a copula approach. Energy Econ 2014; 44: 331–339.

16.

Jones

Kaul

Oil and the stock markets. J Finance 1996; 51: 463–491.

17.

Sadorsky

Oil price shocks and stock market activity. Energy Econ 1999; 21: 449–469.

18.

Driesprong

Jacobsen

Maat

Striking oil: another puzzle?

J Financ Econ 2008; 89: 307–327.

19.

Ono

Oil price shocks and stock markets in BRICs. Eur J Comparat Econ 2011; 8: 29–45.

20.

Narayan

Modelling the impact of oil prices on Vietnam’s stock prices. Appl Energy 2010; 87: 356–361.

21.

Tchatoka

Masson

Parry

Linkages between oil price shocks and stock returns revisited. Energy Econ 2019; 82: 42–61.

22.

Zhu

Guo

You

An empirical research of crude oil price changes and stock market in China: evidence from the structural breaks and quantile regression. Appl Econ 2015; 47: 6055–6074.

23.

Kilian

Park

The impact of oil price shocks on the US stock market. Int Econ Rev 2009; 50: 1267–1287.

24.

Nandha

Brooks

Oil prices and transport sector returns: an international analysis. Rev Quant Finan Acc 2009; 33: 393–409.

25.

Kilian

Exogenous oil supply shocks: how big are they and how much do they matter for the US economy?

Rev Econ Stat 2008; 90: 216–240.

26.

Bouri

Awartani

Maghyereh

Crude oil prices and sectoral stock returns in Jordan around the Arab uprisings of 2010. Energy Econ 2016; 56: 205–214.

27.

Gokmenoglu

Fazlollahi

The interactions among gold, oil, and stock market: evidence from S&P500. Procedia Econ Finance 2015; 25: 478–488.

28.

Yun

Yoon

S-M.

Impact of oil price change on airline’s stock price and volatility: evidence from China and South Korea. Energy Econ 2019; 78: 668–679.

29.

Fatima

Bashir

Oil price and stock market fluctuations: emerging markets (a comparative study of Pakistan and China). Int Rev Manag Bus Res 2014; 3: 1958.

30.

Siddiqui

Mashkoor

Muhammad

Oil price fluctuation and stock market performance-the case of Pakistan. Journal of International Business and Economics 2014; 2(1): 47–53.

31.

Wang

Toward to economic growth without emission growth: the role of urbanization and industrialization in China and India. J Clean Prod 2018; 205: 499–511.

32.

Wang

, et al. The effects of energy prices, urbanization and economic growth on energy consumption per capita in 186 countries. J Clean Prod 2019; 225: 1017–1032.

33.

Wang

Zhang

Does increasing investment in research and development promote economic growth decoupling from carbon emission growth? An empirical analysis of BRICS countries. J Clean Prod 2020; 252: 119853.

34.

Naurin

Qayyum

Impact of oil price and its volatility on cpi of pakistan: bivariate EGARCH model, 2016.

35.

Zhu

Guo

You

, et al. The heterogeneity dependence between crude oil price changes and industry stock market returns in China: evidence from a quantile regression approach. Energy Econ 2016; 55: 30–41.

36.

Nusair

Olson

The effects of oil price shocks on Asian exchange rates: evidence from quantile regression analysis. Energy Econ 2019; 78: 44–63.

37.

Naifar

Do global risk factors and macroeconomic conditions affect global Islamic index dynamics? A quantile regression approach. Q Rev Econ Finance 2016; 61: 29–39.

38.

Mensi

Hammoudeh

Reboredo

, et al. Do global factors impact BRICS stock markets? A quantile regression approach. Emerg Markets Rev 2014; 19: 1–17.

39.

Bai

Carrion-I-Silvestre

JL.

Structural changes, common stochastic trends, and unit roots in panel data. Rev Econ Stud 2009; 76: 471–501.

40.

Dickey

Fuller

WA.

Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica 1981; 49: 1057–1072.

41.

Binder

Coad

From average Joe’s happiness to miserable Jane and cheerful John: using quantile regressions to analyze the full subjective well-being distribution. J Econ Behav Organ 2011; 79: 275–290.

42.

Tiwari

Jena

Mitra

, et al. Impact of oil price risk on sectoral equity markets: implications on portfolio management. Energy Econ 2018; 72: 120–134.

43.

Baur

DG.

The structure and degree of dependence: a quantile regression approach. J Bank Finance 2013; 37: 786–798.

44.

Moya-Martínez

Ferrer-Lapeña

Escribano-Sotos

Oil price risk in the Spanish stock market: an industry perspective. Econ Modell 2014; 37: 280–290.

45.

Andrews

Lee

Ploberger

Optimal changepoint tests for normal linear regression. J Econometr 1996; 70: 9–38.

46.

Dependence patterns across financial markets: a mixed copula approach. Appl Financial Econ 2006; 16: 717–729.

47.

Javaira

Hassan

An examination of herding behavior in Pakistani stock market. Int J Emerg Markets 2015; 10: 474–490.

48.

Arouri

MEH

Nguyen

DK.

Oil prices, stock markets and portfolio investment: evidence from sector analysis in Europe over the last decade. Energy Policy 2010; 38: 4528–4539.

49.

Wang

Zhang

The impacts of global oil price shocks on China’s fundamental industries. Energy Policy 2014; 68: 394–402.

50.

Reboredo

Ugolini

Quantile causality between gold commodity and gold stock prices. Resour Policy 2017; 53: 56–63.

51.

Salisu

Isah

Raheem

ID.

Testing the predictability of commodity prices in stock returns of G7 countries: evidence from a new approach. Resour Policy 2019; 64: 101520.

52.

Masih

Peters

De Mello

Oil price volatility and stock price fluctuations in an emerging market: evidence from South Korea. Energy Econ 2011; 33: 975–986.

53.

Wang

Zhang

The effects of trade openness on decoupling carbon emissions from economic growth–evidence from 182 countries. J Clean Prod 2021; 279: 123838.

54.

Johansen

Juselius

Maximum likelihood estimation and inference on cointegration—with appucations to the demand for money. Oxford Bull Econ Stat 2009; 52: 169–210.