The relationship between energy consumption and gross domestic product in Hong Kong (1992–2015): Evidence from sectoral analysis and implications on future energy policy

Abstract

The causal relationship between energy consumption and gross domestic product in Hong Kong from 1992 to 2015 is investigated in this study. Different from the previous studies focusing on the causal relationship between total energy consumption and total gross domestic product per capita, this study further investigates the causal relationship from sectoral perspective, including residential, commercial, industrial and transportation sectors. For each sector, the time series data of sectoral energy consumption and sectoral per capita value added are collected. To conduct the Granger causality test, the unit root test is first applied to analyse the stationarity of time series. The cointegration test is then employed to examine whether causal relationship exists in long-term. Finally, based on the aforementioned tests, both vector error correction model and vector autoregression model can be selected to determine the Granger causality between time series. It is interesting to find that the sectoral energy consumption and corresponding sectoral per capita value-added exhibit quite different causal relationships. For both residential sector and commercial sectors, a unidirectional causal relationship is found running from the sectoral per capita value added to sectoral energy consumption. Oppositely, for industrial sector and transportation sector, a unidirectional causal relationship is found running from sectoral energy consumption to sectoral per capita value added. Regarding the Granger causality test results, the indicative suggestions on energy conservation policies, energy efficiency policies and greenhouse gas emission reduction policies are discussed based on the background of Hong Kong’s economic structure and fuel types.

Keywords

Causal relationship gross domestic product and energy consumption sectoral analysis energy policy energy implications

Introduction

The Granger causal relationship test is first conducted in 1978.¹ They chose the United States as the target country and year 1947–1974 as testing period and conducted the test to determine the causal relationship between total energy consumption and gross national product (GNP). Thereafter, substantial literatures can be found on analysing the causal relationship inner relationship between economic performance (i.e. the growth of gross domestic product (GDP)) and total energy consumption, with an aim to provide advisory recommendations to policy makers. There are four types of causal relationship between economic performance and energy consumption. The causal directions and features are explained below, and a summary of related studies is tabulated in Table 1. The results have implications to policy makers on the environmental policies such as energy conservation policies, energy efficiency policies and greenhouse gas (GHG) emission reduction policies. Scarce research could be found analysing Hong Kong’s situation with comparison to China, but the colonial history makes Hong Kong politically and economically different from China and need to be studied independently. The uniqueness of economic structure and energy mix in Hong Kong will be mentioned subsequently. In addition, the energy consumption and economic performance are changing with time, therefore, there is a need to update the data on the energy consumption–GDP relationship close to the present time. More importantly, no detailed study on the causal relationship in different sectors has been conducted for Hong Kong yet. In particular, sectoral-based studies on analysing the causal relationship in Hong Kong have not been conducted yet. Therefore, this research is purposely designed to fill the research gap to provide advice for policy makers explicitly on residential sector, commercial sector, industrial sector and transportation sector. The contribution of this paper can be highlighted as renewing the time series data of both total and sectoral energy consumption and GDP up to year 2015 to conduct the Granger causality test, and more concrete test results are obtained. Furthermore, policy implications are discussed in accordance to the Granger causality results, which are crucial to energy-related policy makers in Hong Kong.

Table 1.

Summary of Granger causality studies.

Reference	Examined period	Examined country	Methodology	(1) Unidirectional causality		Bidirectional causality	Mutually independent
Reference	Examined period	Examined country	Methodology	(1.1) Energy consumption Granger causes economic performance	(1.2) Economic performance Granger causes energy consumption	Bidirectional causality	Mutually independent
Ikegami and Wang²	1996–2015	Japan	Autoregressive-distributed lag (ARDL) estimation model	✓
Ikegami and Wang²	1996–2015	Germany			✓
Shiu and Lam³	1971–2000	China	Error correction model	✓
Wolde-Rufael⁴	1952–1999	Shanghai, China	Modified version of the Granger causality test	✓
Cheng and Lai⁵	1955–1993	Taiwan			✓
Chiou Wei et al.⁶	1954–2006	Malaysia	Standard Granger causality test			✓
		Indonesia				✓
		Taiwan		✓
		Hong Kong		✓
		Philippines			✓
		Thailand			✓
Asafu⁷	1973–1975	India	Engle–Granger testing procedure	✓
	1973–1975	Indonesia		✓
	1971–1975	Thailand				✓
	1971–1975	Philippines				✓
Paul and Bhattacharya⁸	1950–1996	India	Standard Granger causality test			✓
Lee and Chang⁹	1954–2003	The United States	Standard Granger causality test				✓
		South Korea					✓
		Taiwan		✓
Menegaki¹⁰	1990–2010	OECD countries	Panel data cointegration regression	✓
Chen et al.¹¹	1971–2001	Mixed 10 Asia Countries	Single datasets and panel date procedure			✓
Narayan and Smyth¹²	1990–2010	12 oil exporting countries	Panel data cointegration regression		✓
Apergis and Payne¹³	1997–2007	27 European Countries	Multivariate panel framework				✓
Soytas and Sari¹⁴	1950–1992	Italy	Standard Granger causality test		✓
		Turkey		✓
		France		✓
		Germany		✓
		Japan		✓
	1950–1990	Argentina				✓
	1960–1992	Indonesia
	1953–1991	Korea			✓
	1965–1994	Poland				✓
Burke and Csereklyei¹⁵	1960–2010	132 countries	Panel data cointegration regression				✓
Borozan¹⁶	1992–2010	Croatia	Bivariate vector auto-regression	✓
Yuan et al.¹⁷	1963–2005	China	Vector error correction model	✓
Zhang and Xu¹⁸	1995–2008	China	Standard Granger causality test		✓
Zhang and Cheng¹⁹	1960–2007	China	Vector auto-regression and error correction model		✓
Narayan and Popp²⁰	1980–2006	93 countries	Bivariate vector auto-regression	✓
Ho and Siu²¹	1966–2002	Hong Kong	Standard Granger causality test	✓

OECD: Organisation for Economic Co-operation and Development.

Unidirectional causality

1.1 Energy consumption Granger causes economic growth

This type of causal relationship implies that the economic growth is associated with the increase of energy consumption. A typical case of this causality type is Japan. Study indicates that the booming development of Japan’s economy relies on its domestic electricity consumption from 1996 to 2015,² and the sudden loss of nuclear power supply after Fukushima accident has resulted in the first trade deficit in Japan 2011 since 1980 oil crisis. As a compensation to nuclear power, the substantial and expensive imports of fossil fuels have significantly struck Japan’s economic performance.²² This implies that Japan has a revived intention to reboot nuclear power in the future since the shortage of electricity supply will trigger the Japanese economy to a large extent.²³ Shiu and Lam³ examined China’s situation during 1971–2000 and the result also indicated unidirectional causality running from energy consumption to economic growth. This implication is consistent with Wolde-Rufael’s⁴ research that only examined Shanghai’s case. Under this causality type, the implementation of energy conservation policies, energy efficiency policies and GHG emission reduction policies may impede the development of an economy and should be carefully conducted based on continuous evaluation of the impact on economic performance.

1.2 Economic growth Granger causes energy consumption

The second type is another form of unidirectional causality relationship running from economic performance to energy consumption. Taiwan and Germany are good examples.⁵ This causal direction implies that economic growth leads to more energy consumption. That is to say, the implementation of energy conservation policies, energy efficiency policies and GHG emission reduction policies would not significantly affect economic growth or the impact could be mitigated if proper countermeasures are undertaken.²⁴

2. Bidirectional causality

The third causality type is the bidirectional causality between economic performance and energy consumption. Southeast Asian countries such as Malaysia, Indonesia, Philippines and Thailand are representative countries.^6,7 In South Asia, this causality type is also found in India by Paul and Bhattacharya.⁸ The bidirectional causality implies that the economic development relies on the energy consumption, but simultaneously it might be affected by the implementation of energy-related policies.

3. Mutual independence

The last type is the mutually independent relationship between energy consumption and economic development, which means no statistically causal relationship between these two variables. United States and South Korea are representative examples.⁹ Under this causality type, the economic development is statistically unrelated to the energy consumption. The implementation of energy and environmental protection policies is therefore in sound conditions and will not impede economic growth.

In addition to analysing the causality relationship in a particular country, some studies have examined the panel data of a set of countries by regions or by organizations. For example, a study has analysed the relationship between energy consumption and economic development in European countries and concluded that the energy consumption Granger caused economic growth.¹⁰ Other studies have also conducted this test on 10 Asian countries and found that long-term bidirectional causal relationship exists. G-7 countries and Organisation for Economic Co-operation and Development countries have also been investigated.^11–13 Test on nine independent countries has been delivered and diverse types of causal relationship were found among these countries.¹⁴ Furthermore, some studies do not simply analyse these two variables, but further decompose the variables by sectors or by fuel types. For instance, a study has decomposed the energy consumption and GDP elasticity in 132 countries by sectors and fuel types. It was found that residential energy consumption was inelastic to per capita household expenditure added if biofuels were counted in total energy consumption tallies. Moreover, from the sectoral perspectives, residential electricity consumption was proved to be more closely connected to GDP growth than other sectors in those sample countries.¹⁵ Besides, the causal relationship between economic performance and energy consumption by fuel types in Croatia has also been examined.¹⁶ A unidirectional causal relationship is found running from GDP to primary energy resources such as liquid fuel, and natural gas, as well as secondary energies such as solar power and wind power. No causal relationship is identified between GDP and coal consumption. From a policy maker’s perspective, it indicates that the formulation of energy-related policies should comprehensively take fuel types into consideration.

The results of Granger causality test might be different and cause ambiguity sometimes even examining the same countries or regions due to different time span being selected, methodologies being applied, and model specifications being employed or the bias appeared in the estimates of parameters in a regression analysis.²⁵ In addition, different results can also be caused by various conceptual frameworks and various policies adopted by different countries.²⁶ Studies have compared different Granger causality results in the Chinese context based on different model and time span, or reported the cross-section dependence between countries which is usually overseen, and this leads to biased results.^17,27 Taking China as an example, a study has concluded that the increase of energy consumption is associated with economic growth,¹⁸ which is consistent with the result delivered by Zhang and Cheng,¹⁹ but is contradicted to the results by others.²⁰

Hong Kong has a very different economic structure and energy mix from mainland China as a Special Administrative Region. After China adopting open-door policy since late 1970s, Hong Kong is gradually shifting its economic structure as well as its energy mix. A large amount of manufacturing activities outflows to Mainland China, leading to the industrial sector GDP share dropped from 33.2 in 1984 to 7% in 2014. However, the GDP share of the commercial sector increased dramatically from 20% to more than 80%, trailed by smaller rises in the residential sector and slight drop in the transportation sector. The relationship between energy consumption and economic consumption could be affected by economic structure change because the energy mix changes from industrial- to commercial-led economic structure and shows obvious discrepancies.^21,28

Regarding the Hong Kong Electrical and Mechanical Services Department (EMSD), the energy data are categorized into town gas and liquefied petroleum gas (LPG), electricity, and oil and coal products. This paper extracted the energy mix data in year 1992, 2002 and 2014 from EMSD for comparison purpose (Table 2). ‘Electricity’ refers to the electricity generated from natural gas, coal, solar energy as well as wind energy. ‘Town Gas and Liquefied Petroleum Gas’ includes town gas, bio-gas and LPG. ‘Oil and Coal Products’ include charcoal, biodiesel, gasoline, diesel, kerosene, semi-coking coal coking coal, anthracite as well as aviation fuel. It is evident that the energy mix in Hong Kong has shifted significantly and the use of oil and coal products is gradually replaced by electricity. Besides, it is found that residential sector, commercial sector, industrial sector and transportation sector are four dominating energy consumers in Hong Kong. The fuel type consumed in these sectors has also shifted drastically in the past decades.

Table 2.

Energy mix in 1992, 2002 and 2014.

Year	Energy consumption and percentage share in total and in sectors		Town gas and liquefied petroleum gas (TJ)		Oil and coal products (TJ)		Electricity (TJ)
1992	Total	Residential	25,871 (10%)	N/A	126,539 (51%)	N/A	94,152 (38%)	25,663 (27%)
		Commercial		N/A		N/A		33,062 (35%)
		Industrial		N/A		37,537 (30%)		15,103 (16%)
		Transportation		N/A		84,329 (67%)		N/A
2002	Total	Residential	39,194 (14%)	18,552 (42%)	107,615 (38%)	292 (0%)	137,112 (48%)	33,394 (24%)
		Commercial		13,721 (31%)		4200 (4%)		83,549 (61%)
		Industrial		1010 (2%)		4969 (5%)		16,112 (12%)
		Transportation		10,774 (24%)		88,797 (90%)		4057 (3%)
2014	Total	Residential	50,844 (18%)	17,838 (36%)	80,169 (28%)	8 (0%)	158,147 (55%)	43,415 (27%)
		Commercial		14,928 (29%)		5156 (6%)		102,627 (65%)
		Industrial		1244 (2%)		4244 (5%)		9230 (6%)
		Transportation		16,185 (32%)		70,939 (88%)		2875 (2%)

The changes of structure have flourished the local economy and shaped Hong Kong as an important financial centre in the world. Import–export, finance and insurance and business services have dominated Hong Kong’s economic structure. Under a rapidly developing economy and possessing a huge energy market, Mainland China has always drawn plenty of attentions from the global market and been studied by researchers. However, those studies on China are always not including Hong Kong since Hong Kong is a politically and economically autonomous region. So the relevant results and conclusions could not be generalized and applicable to Hong Kong. The integration with China has brought about tremendous challenges to Hong Kong, especially the localism procedure under political and economic dynamics.²⁹ Nevertheless, the feature of ‘post-colonial’ Hong Kong has exaggerated the contrasting from extended infrastructural projects to retain a local culture.³⁰ In addition, the extremely dense population and the sub-tropical climate makes the energy mix in Hong Kong very unique.³¹ In other words, Hong Kong needs to be studied independently. Nevertheless, insufficient studies have selected Hong Kong as a target city. Therefore, this paper considered the special background of Hong Kong and examined the causal relationship between energy consumption and economic performance of Hong Kong over the past decade on the basis of four sectors: commercial sector, residential sector, transportation sector and industrial sector. Up to date, only two papers could be found focusing on Hong Kong, one has applied both linear and non-linear Granger causality test to examine the causal relationship between total energy consumption and GDP of Hong Kong during period 1954–2006. The results show a unidirectional causality running from total energy consumption to GDP.⁶ Another research focuses only on the electricity consumption rather than energy consumption of Hong Kong from 1966 to 2002, which has indicated a unidirectional causality running from electricity consumption to GDP.²¹

To achieve the research purpose, the structure of this paper is organized as follows: the next section introduces the methodology in this paper. ‘Empirical research’ section consists of the data collection and empirical results. ‘Discussions and policy implications’ section discusses the policy implications with accordance to the test results in section three. The final section concludes and summarizes this paper.

Methodology

The methodology consists of three steps to achieve the aim of this paper, which includes unit root test, cointegration test and Granger causality test. The initial step is the unit root test, which is applied to check the stationarity of each time series, middle step is the cointegration test for checking the presence or absence of spurious regression and is able to identify whether stable long-term relationship of the variables exists or not. The last step is the Granger causality test for determining the direction of causal relationship.^32,33

Unit root test

Two preferred techniques in the unit root test are the Phillips–Perron (PP) test and the augmented Dickey Fuller (ADF) test.^34–36 The Kwiatkowski–Phillips–Schmidt–Shin (KPSS) test could be referred if results of PP test and ADF test are controversial. The time series variables are tested in level, in the first difference (and the second difference if necessary until the null hypothesis is rejected to all in the same order). MacKinnon p-value is used to determine the statistical significance.³⁷ The basic vector autoregression (VAR) model is presented as

Y_{it} = α_{i} Y_{i, t - 1} + β_{i} X_{it} + ε_{it}

(1)

where i denotes four sectors being observed over period t. X denotes exogenous variables, and ε is a stationary error process.

Cointegration test

To check if spurious regression exists or not, two testing techniques are suggested by Johansen and Juselius.³⁸ One is the trace statistics and the other is maximum eigenvalue statistics. Equations (2) and (3) explain trace statistics test and maximum eigenvalue statistics test, respectively. r denotes the number of relationships being tested. λ_i is the ith largest canonical correlation and T denotes the sample size³⁹

J_{trace} = - T \sum_{i = r + 1}^{n} ln (1 - λ_{i})

(2)

J_{\max} = - Tln (1 - λ_{r + 1})

(3)

For general cases, trace statistics and maximum eigenvalue statistics generate similar cointegration results when dealing with small test samples. If the sample is too large, it is found that the trace statistic test is more powerful than the maximum eigenvalue statistic test.⁴⁰ The cointegration test starts with the VAR of order p given by

Y_{t} = μ + A_{1} Y_{t - 1} + \dots + A_{p} Y_{t - p} + ε_{t}

(4)

where Y_t is an n×1 vector of variables. The function is commonly written as I(1) because the variables are integrated of order 1.

ε_{t}

is an n×1 vector of innovations.⁴¹

Granger causality test

The unit root test and cointegration test are cooperated to determine the presence or absence of the long-term cointegration relationship. If it is present, the vector error correction model (VECM) is employed to test the direction of causal relationship. If it is absent, the VAR is employed to test the Granger causality.⁴² The bivariate VECM regression form is presented as

Y_{t} = α_{0} + α_{1} Y_{t - 1} + \dots + α_{l} Y_{t - l} + β_{1} X_{t - 1} + \dots + β_{l} X_{t - l} + ε_{t}

(5)

X_{t} = α_{0} + α_{1} X_{t - 1} + \dots + α_{l} X_{t - l} + β_{1} Y_{t - 1} + \dots + β_{l} Y_{t - l} + μ_{t}

(6)

where l denotes the lag length. The null hypothesis is that X does not Granger-cause Y in formula (5) and at the same time Y does not Granger-cause X in formula (6).

Empirical research

Data collection

The energy consumption datasets are collected from the Hong Kong EMSD, including the total and sectoral energy consumption from 1992 to 2015 (data available only to year 2015), with the original unit of ‘terajoule’ (Figure 1). The GDP and the value-added datasets in industrial sector, commercial sector and transportation sector, and the added household expenditure are extracted from the GDP (yearly) published by the Hong Kong Census and Statistics Department, all data with the unit of ‘Millions of Hong Kong Dollar’. To convert the GDP unit to USD per capita (Figure 2), the original data (with Hong Kong Dollar) are divided by the total population of Hong Kong which is obtained from the Hong Kong Census and Statistics Department. The population is selected by stages as according to the Census results in mid-1996, mid-2001, mid-2006, mid-2011 and mid-2016, respectively. To convert per capita Hong Kong Dollar to per capita USD, an average exchange rate of 0.128 (1Hong Kong Dollar = 0.128USD) is used for calculation (data available at the World Bank), since the linked exchange rate system mechanism stabilizes the exchange rate between Hong Kong Dollar and USD. Ten time series of data are collected and summarized in Appendix 1. The descriptive statistics and correlation matrix of 10 time series are represented in Tables 3 and 4, respectively. The Jarque–Bera test measures whether the series is normally distributed or not. The results shown in Table 3 indicate that the series of per capita industry value added should be rejected from the null hypothesis of normal distribution, whereas other series are statistically proved normal distribution. However, normality is not a prerequisite condition when testing the stationarity of a series.

Figure 1.

Total energy consumption (terajoule) and the share of four sectors.

Figure 2.

Total real GDP per capita (USD) and the share of four sectors. Note: The added household expenditure from 1992 to 1999 are lacking in the original publication of the GDP (yearly) as published by the Hong Kong Census and Statistics Department. GDP: gross domestic product.

Table 3.

Descriptive statistics.

	ECtot	ECres	ECcom	ECind	ECtra	GDPtot	GDPres	GDPcom	GDPind	GDPtra
Mean	27,4580.0	51,284.46	10,0176.8	28,717.58	94,401.25	21,7083.8	7259.507	16,5206.9	23,090.40	21,7083.8
Median	28,0601.0	53,453.00	10,4829.5	20,537.50	93,619.00	20,1980.9	6903.748	14,8198.6	21,058.61	20,1980.9
Maximum	28,9338.0	61,910.00	12,3674.0	59,812.00	10,6984.0	31,7685.4	10,754.78	26,3161.2	43,024.18	31,7685.4
Minimum	24,6562.0	35,677.00	63,416.00	14,583.00	87,657.00	12,5917.3	3863.703	79,014.35	16,089.82	12,5917.3
Std. Dev.	13,197.27	7557.146	18,821.24	14,488.89	5248.590	49,699.80	2323.510	49,697.13	7372.070	49,699.80
Skewness	−0.976150	−0.546797	−0.488005	0.661294	0.673646	0.428878	0.249404	0.364417	1.499380	0.428878
Kurtosis	2.467433	2.219730	1.962544	2.101016	2.491520	2.533790	1.690140	2.277155	4.607856	2.533790
Jarque–Bera	4.095102	1.804768	2.028911	2.557410	2.073746	0.953096	1.309696	1.053703	11.57776	0.953096
Probability	0.129051	0.405601	0.362600	0.278398	0.354562	0.620923	0.519521	0.590461	0.003061^a	0.620923
Sum	6,589,919	1,230,827	2,404,244	689,222	2,265,630	5,210,011	116,152	3,964,966	554,169	5,210,011
Sum Sq. Dev.	4.01E+09	1.31E+09	8.15E+09	4.83E+09	6.34E+08	5.68E+10	80,980,494	5.68E+10	1.25E+09	5.68E+10
Observations	24	24	24	24	24	24	16	24	24	24

^aDenotes rejection of the null hypothesis.

Table 4.

Correlation matrix.

	ECtot	ECres	ECcom	ECind	ECtra	GDPtot	GDPres	GDPcom	GDPind	GDPtra
ECtot	1.000000	0.521919	0.497169	−0.221157	−0.393065	0.729249	0.714930	0.708202	0.437474	0.729249
ECres	0.521919	1.000000	0.960745	−0.894097	−0.933886	0.893915	0.873237	0.911716	−0.327254	0.893915
ECcom	0.497169	0.960745	1.000000	−0.927597	−0.945729	0.904296	0.887722	0.927088	−0.382744	0.904296
ECind	−0.221157	−0.894097	−0.927597	1.000000	0.837458	−0.710015	−0.690995	−0.748118	0.579000	−0.710015
ECtra	−0.393065	−0.933886	−0.945729	0.837458	1.000000	−0.880172	−0.868869	−0.895568	0.395203	−0.880172
GDPtot	0.729249	0.893915	0.904296	−0.710015	−0.880172	1.000000	0.990633	0.997343	0.009926	1.000000
GDPres	0.714930	0.873237	0.887722	−0.690995	−0.868869	0.990633	1.000000	0.985857	0.033635	0.990633
GDPcom	0.708202	0.911716	0.927088	−0.748118	−0.895568	0.997343	0.985857	1.000000	−0.043020	0.997343
GDPind	0.437474	−0.327254	−0.382744	0.579000	0.395203	0.009926	0.033635	−0.043020	1.000000	0.009926
GDPtra	0.729249	0.893915	0.904296	−0.710015	−0.880172	1.000000	0.990633	0.997343	0.009926	1.000000

Empirical results

In this section, unit root test, cointegration test and Granger causality test are conducted and the results are shown from Tables 5 to 8.

Table 5.

Unit root test.

Variables	At level		At first difference
	ADF	PP	ADF	PP
ECtot	−1.71	−1.68	−5.70***	−5.67***
ECres	−3.11	−2.40	−9.90***	−9.94***
ECcom	−2.24	−3.94	−4.68***	−4.88***
ECind	−1.15	−1.15	−5.39***	−5.46***
ECtra	−1.78	−1.67	−5.19***	−5.36***
GDPtot	−3.73	−2.95	−2.76***	−2.74***
GDPres	−1.39	−1.39	−2.73**	−2.70**
GDPcom	−3.53	−3.62	−2.77***	−2.87***
GDPind	−2.40	−2.57	−7.00***	−11.49***
GDPtra	−2.99	−2.95	−4.30**	−4.56***

ADF: augmented Dickey Fuller; PP: Phillips–Perron.

*Signiﬁcant at conﬁdence level 10%.

**Signiﬁcant at conﬁdence level 5%.

***Signiﬁcant at conﬁdence level 1%.

Table 6.

Lag length selection.

Energy variable	Economic variable	LR		AIC		SC		HQ		Lag selection
Energy variable	Economic variable	Lag 1	Lag 2	Lag 1	Lag 2	Lag 1	Lag 2	Lag 1	Lag 2	Lag selection
ECtot	GDPtot	87.43^a	9.13	−8.80	−8.97^a	−8.50^a	−8.47	−8.72	−8.85^a	2^b
ECres	GDPres	34.48	11.03^a	−6.42	−7.08^a	−6.15	−6.62^a	−6.45	−7.12^a	2
ECcom	GDPcom	132.42^a	3.57	−8.99^a	−8.83	−8.69^a	−8.34	−8.92^a	−8.72	1
ECind	GDPind	72.95^a	2.58	−2.99^a	−2.78	−2.70^a	−2.29	−2.92^a	−2.66	1
ECtra	GDPtra	25.01^a	1.64	−5.75^a	−5.49	−5.46^a	−4.99	−5.68^a	−5.37	1

AIC: Akaike information criterion; HQ: Hannan–Quinn information criterion; LR: sequential modified LR test statistic (each test at 5% level); SC: Schwarz information criterion.

^aIndicates lag order selected by the criterion.

^bLR and SC test indicate lag 1 should be selected, however AIC and HQ suggest lag 2, so the final prediction error (FPE) test is applied to make the final decision. The FPE test result suggests that lag 2 should be selected.

Table 7.

Cointegration test.

Variable	Trace statistics		Maximum eigenvalue statistics		No. of cointegration equation(s)
	None	At most 1	None	At most 1
ECtot GDPtot	21.37^* (0.0058)	0.60 (0.4398)	20.77^* (0.0041)	0.60 (0.4398)	1
ECres GDPres	21.87^* (0.0048)	0.89 (0.3449)	20.98^* (0.0038)	0.89 (0.3449)	1
ECcom GDPcom	19.10^* (0.0137)	2.24 (0.1347)	16.86^* (0.0190)	2.24 (0.1347)	1
ECind GDPind	14.84 (0.0626)	4.10^* (0.0429)	10.74 (0.1679)	4.10^* (0.0429)	no
ECtra GDPtra	17.23^* (0.0271)	1.84 (0.1753)	15.39^* (0.0330)	1.84 (0.1753)	1

*Denotes rejection of the hypothesis at the 0.05 level. Values in brackets represent the Mackinnon–Haug–Michelis p-values.

Table 8.

Granger causality test.

Energy variable	Economic variable	Null hypothesis and Chi-sq		Causality
Energy variable	Economic variable	EC_i ≠ GDP_i	GDP_i ≠ EC_i	Causality
ECtot	GDPtot	16.61*** (0.0053)	11.33*** (0.0035)	ECtot <-> GDPtot
ECres	GDPres	3.45 (0.1785)	5.12* (0.0772)	GDPres → ECres
ECcom	GDPcom	4.36 (0.4991)	33.61*** (0.0000)	GDPcom→ ECcom
ECind	GDPind	4.93** (0.0264)	2.51 (0.1134)	ECind → GDPind
ECtra	GDPtra	5.02** (0.0250)	0.08 (0.7770)	ECtra → GDPtra

Note: VEC Granger causality for cointegrated pairs and VAR Granger causality for non-cointegrated pair.

*Significant at confidence level 10%.

**Significant at confidence level 5%.

***Significant at confidence level 1%.

The unit root test results are presented in Table 5. The results suggest accepting the null hypothesis (the existence of unit root) at level but rejecting the null hypothesis at first difference, which implies that all time series are non-stationary at level but are stationary at the first difference. To improve the test accuracy, this paper also adopts KPSS test to prove the rationality of the results,⁴³ which is presented in Appendix 2.

The selection of the appropriate lag length for time series variables is extremely important, since a non-optimal lag length may negatively affect the test results due to mistaking dynamic status of time series or distorting the power of the variables.⁴⁴ The lag order selection is commonly determined by LR, AIC, SC and HQ criterion and the lag length selection result is shown in Table 6. The results indicate that the optimal lag length of two models (total energy consumption and real GDP per capita, residential sector) is two, and optimal lag length of other three models (commercial sector, industrial sector and transportation sector) is one. The FPE test is used for the total energy consumption and GDP model for final decision.

The results of cointegration test are displayed in Table 7. No cointegration relationship is found between the time series of industrial energy consumption and per capita industry value added. The cointegration relationship between energy consumption and per capita of added value for other sectors is observed in the long-term. Therefore, for the Granger causality test, we add an ECM term only to the models which are cointegrated.

The results of Granger causality test regarding the energy variables and economic variables are presented in Table 8, which shows great differences among different sectors. Bidirectional causal relationship is found between the total energy consumption and per capita GDP. For the decomposed sectors, unidirectional causal relationships are found running from the added per capita household expenditure to energy consumption in residential sector and from per capita commerce value added to energy consumption in commercial sector. This finding implies that for the residential and commercial sectors, the increase of energy consumption is associated with the improvement of sectoral economic performance. Therefore, energy conservation policies, energy efficiency policies and GHG emission reduction policies for these two sectors would not significantly affect economic performance, or the negative impacts of implementing such kind of policies on economic performance could be mitigated. On the other hand, unidirectional causal relationship is found running from energy consumption in industrial sector to per capita industry value added and from energy consumption in transportation sector to per capita transport value added. This reveals that the boost of economic performance in industrial sector and transportation sector relies on the energy consumption. As a result, implementing energy conservation policies, energy efficiency policies and GHG emission reduction policies might impede economic performance of these two sectors.

Discussions and policy implications

Evidence from ‘Empirical results’ section indicates that the economic performance of industrial sector and transportation sector is more sensitive to energy-related policies. Energy policies for residential sector and commercial sector should be carefully formulated and evaluated to avoid its negative effect on the per capita GDP. A testing period is suggested to measure the performance of the electricity conservation policy on specific buildings or firms. Some voluntary policies seem to be ineffective and thus mandatory policies should be implemented. Since energy policies for industrial sector and transportation sector may affect the per capita value added, energy policies for industrial sector and transportation sector should be carefully examined for the pros and cons, especially regarding the pollutant emission reduction policies.

Transportation and industrial sectors

The changing structure has also contributed to the transportation sector. The declining importance of manufacturing industry in Hong Kong leads to a dropping necessity on the raw materials transportation into China, and consequently a decreasing necessity on transporting back again for exporting to the foreign markets. Hence the energy consumption of goods vehicles rapidly drops from 33,946 TJ in 1992 to 25,009 TJ in 2014.⁴⁵ However, the passenger transport demand is inexorable changing with the movement of inhabitant from Hong Kong Island (20.6% in 2004 and 17.2% in 2014) and Kowloon (31.0% in 2004 and 30.2% in 2014) to the New Territories (48.1% in 2004 and 52.6% in 2014). As more people decide to transport by cars and taxis, the energy consumption by cars and taxis is gradually increasing in the past decade. However, the energy consumption by rails and buses is decreasing, probably because of the ‘Area Approach Bus Route Rationalisation’ plan by the Transport and Housing Bureau in 2013 to alleviate traffic congestion and vehicle exhaust gas pollution problems.

Nevertheless, although energy consumption by goods vehicles drops rapidly over the past decades, the goods vehicles still consume 30% of the oil and coal products in 2014, followed by cars and motorcycles (27%) as well as bus and taxis (21%). Therefore, there is a large space to improve the energy transit from grey energy to green energy in the future. Replacing gasoline by LPG or electricity for small sized vehicles (taxis, private cars and motorcycles), and replacing diesel by biodiesel for generous sized vehicles (buses and goods vehicles) should be encouraged by the Government and the Transport Department. Changing the energy structure is the most effective way to eliminate pollutant emissions; however, it is usually done by energy legislation and regulation. In addition, fuel tax, road tunnel tolls and parking controls are also effective mandatory ways to reduce the number of private cars.

To stimulate environment sustainability, some voluntary policies are necessary. The government should encourage local people to choose public transportation for daily travelling. ‘Area Approach Bus Route Rationalisation’ plan by the Transport and Housing Bureau in 2013 presents a good example to attract passengers to choose bus for travelling. This plan has merged, changed or shortened some bus routes to meet passengers’ needs. However, this plan is implemented mainly in Kowloon and New Territories, so efforts should be made in the Hong Kong Island in the future. With more people choosing to use public vehicles and a higher usage rate of green energy, energy consumption could be reduced, and air pollution could be eliminated.

The share of energy consumption in industrial sector drops drastically by 19.1% and the per capita industry value added reduced by 11.3% in the past two decades. In accordance with the economic structure change in Hong Kong, the mandatory policies have been effectively implemented to address the pollutant emission problem by power plants over the past years although the total electricity demand has increased significantly. From year 1997, the coal-fired power plants have been prohibited, and gas-fired power has been encouraged afterwards. Currently, there exist some laws and regulations aiming to control the pollutant emissions from construction activities, power plants, manufacturing, etc. such as the Technical Memorandum for Allocation of Emission Allowances in Respect of Specified Licences issued by Hong Kong Environmental Protection Department.

Residential and commercial sectors

The residential and commercial sectors consume 92% of the total electricity, 65% of the total town gas and LPG in 2014. According to ‘Residential and commercial sectors’ section, space conditioning, cooking, hot water and lighting are the most energy-consuming items. The local people become wealthier and are more willing to pay for durable household appliances, which has stimulated the rising trend of per capita household expenditure added.²¹ However, it happens when per capita income rises, the energy consumption in residential sector reversely decreases. Another study pointed out that it might be true in some countries under specific development stages,²⁸ that the occupants have higher energy conservation consciousness and the household appliances are more energy efficient. Nevertheless, it does not seem to happen in Hong Kong so far.

Although the government has enacted some energy conservation policies, the reactions by the users are pessimistic because most of those policies are voluntary instead of mandatory. For instance, the BEAM Plus scheme, which only requires public housing to fulfil the rating requirements, nevertheless is voluntary for private housings. More importantly, the standards and specifications in BEAM Plus regarding thermal performances of building envelopes and air-conditioning units are inexplicit and the criteria are described ambiguously and the expected outcomes are not well-defined.⁴⁶ Another example is a voluntary agreement target on shopping malls to adjust the air-conditioning thermostat to 25.5°C for energy saving, since usually the shopping malls in Hong Kong set up very low temperatures.⁴⁷ Their research reveals that the enterprises with higher compliance costs would prefer not to comply the energy saving agreement, and they would not be punished due to the voluntary nature.

Therefore, the voluntary energy conservation and energy efficiency policies are ineffective and could not draw public attention. It is suggested that mandatory regulations should play a dominant role, whereas voluntary agreements work as supplementary role, which would be more efficient to draw public attention and raise their consciousness about sustainability and environment protection. However, it is of necessity to implement the mandatory policies with caution. Although the Granger Causality test indicates that sectoral per capita value added Granger causes sectoral energy consumption, it does not mean that the energy conservation or energy efficiency policies would not negatively affect economic performance. It means that there might be negative influences but are able to be addressed or eliminated by suitable countermeasures. Therefore, this paper suggests that the mandatory policies need to be implemented under mutual compliances between the government and the enterprises, with a testing period to identify possible drawbacks, monitor the performance, and gradually adjust the standards and specifications to reach consensus by both parties.

Renewable energy policy

The utilization of renewable energy is limited in Hong Kong. Although Hong Kong has abundant renewable energy resources, the feasibilities of adopting those technologies are restricted. A study points out that 239,833 out of 309,606 buildings in Hong Kong are suitable for installing solar panels. The electricity generation potential by solar photovoltaic (PV) on building rooftops is estimated as 5.944 TW h annually,⁴⁸ which accounts for about 16% of total electricity consumption.⁴⁹ However, the application of PV panel is still a dilemma issue for Hong Kong, owing to the low average potential PV per building, the high upfront cost of solar PV system and the relatively low energy efficiency under real outdoor conditions.⁵⁰ Alternatively, solar water heating systems for domestic hot water usage, as well as for floor/wall/ceiling heating are becoming more popular due to higher efficiency and much lower cost, thus the wide application of solar water heating for residential and commercial sectors should be highly encouraged in the future.

As for the wind energy development, since offshore wind resources are very abundant in Hong Kong, some researches have investigated the potential offshore wind energy to evaluate the viability for Hong Kong to develop offshore wind farms. The investigation scope of Gao’s research was within Hong Kong’s marine boundary of around 360 km², and the estimated wind power generation was 11 TW h,⁵¹ which accounts for more than 30% of the annual electricity consumption in year 2011. Therefore, it is suggested that the Hong Kong Government should consider constructing offshore wind farms and developing offshore wind power in the future.

In addition, biomass is also feasible in Hong Kong since 400 km² (accounts for 36% of Hong Kong’s total land area) of the land area is used for energy crop plantation, although the crop biomass only accounts for insignificant energy share (approximately 1%).⁵² Other renewable resources such as tidal power, geothermal energy and hydropower are negligible regarding the contribution to energy demand, so would not be discussed in this paper.

Conclusion

The causal relationship between energy consumption and GDP in Hong Kong during period 1992–2015 from a sectoral prospective is investigated in this paper; the methodology employed consists of unit root test, cointegration test and Granger causality test. The results of the Granger causality test have indicative evidence to policy makers for energy policy decision making related to energy conservation policies, energy efficiency policies as well as GHG emission reduction policies.

The results indicate great differences among different sectors. Unidirectional causal relationships are found running from per capita household expenditure added to energy consumption in residential sector and from per capita commerce value added to energy consumption in commercial sector. This finding implies that energy conservation and energy efficiency policies aiming to residential sector and commercial sector would not significantly affect economic performance, and effort should be made to eliminate the negative effect. Currently, Hong Kong has enacted some voluntary energy conservation and energy efficiency policies, but they are ineffective. Therefore, voluntary agreements are proposed to work together with mandatory regulations, in order to draw more public attention and power. However, it should be noted that the mandatory regulations need to be made with caution. Although the Granger causality test indicates the causal relationship running from per capita GDP to energy consumption, it does not mean that the energy conservation policies would not affect GDP, it is because that the Granger causality test reflects statistical significance rather than real significance. Thus, the mandatory policies need to be implemented under mutual cooperation by the government and the specific body, to monitor the improvement or discrepancy by a conservation policy in a testing period, and gradually adjust to reach consensus. Unidirectional causal relationship is found running from industrial sector energy consumption to per capita industry value added, as well as from transportation sector energy consumption to per capita transport value added. Therefore, the emission reduction policies and change in fuel mix may impact the economic performance contributed by these two sectors. It is noted that changing the energy structure is the most effective way to eliminate polluted emissions. Furthermore, the wide application of renewable energy in Hong Kong is strongly recommended to be encouraged by government, especially for solar water/space heating system and wind power energy.

There are two suggestions for the future study. First, a longer period could be investigated in a panel data regression model integrated with other countries or districts similar to Hong Kong’s context, for instance, Singapore and Taiwan. Thus, a comparative study on the Granger causality direction and policy implications could be completed. Second, it is suggested that future studies should focus on investigating the Granger causality between segmental energy consumption and per capita GDP under four sectors. For instance, residential sector could be further divided into public housing, private housing, Housing Authority subsidized sale flats and other housing. Commercial sector could be further divided into restaurant segment, retail segment, office segment, education segment, human health segment, accommodation segment and others. Industrial sector is then separated into textile and wearing apparel, food and beverage, metal and machinery, non-manufacturing and others. Finally, transportation sector could be considered as the combination of freight segment, passenger segment and others. Followed by obtaining a more sophisticated and well-directed causality direction in a specific segment, it will be more helpful for expertise to stipulate energy and environmental policies in practice.

Footnotes

Declaration of conflicting interests

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

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Ruixiaoxiao Zhang is a PhD candidate at Ng Wing Hong Laboratory for Sustainable City, Department of Building and Real Estate, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, China.

Geoffrey QP Shen is a Chair Professor at Ng Wing Hong Laboratory for Sustainable City, Department of Building and Real Estate, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, China.

Meng Ni is a Professor and Associate Head of Department Ng Wing Hong Laboratory for Sustainable City, Department of Building and Real Estate at The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, China.

Johnny Wong is an Associate Professor at School of the built environment, University of Technology Sydney, Australia.

Appendix 1

References

Kraft

On the relationship between energy and GNP. J Energy Dev 1978; 3: 401–403.

Ikegami

Wang

The long-run causal relationship between electricity consumption and real GDP: evidence from Japan and Germany. J Policy Model 2016; 38: 767–784.

Shiu

Lam

P-L.

Electricity consumption and economic growth in China. Energy Policy 2004; 32: 47–54.

Wolde-Rufael

Disaggregated industrial energy consumption and GDP: the case of Shanghai, 1952–1999. Energy Econ 2004; 26: 69–75.

Cheng

Lai

TW.

An investigation of co-integration and causality between energy consumption and economic activity in Taiwan. Energy Econ 1997; 19: 435–444.

Chiou Wei

Chen

Zhu

Economic growth and energy consumption revisited – evidence from linear and nonlinear Granger causality. Energy Econ 2008; 30: 3063–3076.

Asafu

AJ.

The relationship between energy consumption, energy prices and economic growth: time series evidence from Asian developing countries. Energy Econ 2000; 22: 615–625.

Paul

Bhattacharya

RN.

Causality between energy consumption and economic growth in India: a note on conflicting results. Energy Econ 2004; 26: 977–983.

Lee

C-C

Chang

C-P.

Structural breaks, energy consumption, and economic growth revisited: evidence from Taiwan. Energy Econ 2005; 27: 857–872.

10.

Menegaki

AN.

Growth and renewable energy in Europe: a random effect model with evidence for neutrality hypothesis. Energy Econ 2011; 33: 257–263.

11.

Chen

Kuo

Chen

CC.

The relationship between GDP and electricity consumption in 10 Asian countries. Energy Policy 2007; 35: 2611–2621.

12.

Narayan

Smyth

Energy consumption and real GDP in G7 countries: new evidence from panel cointegration with structural breaks. Energy Econ 2008; 30: 2331–2341.

13.

Apergis

Payne

JE.

Renewable energy consumption and economic growth: evidence from a panel of OECD countries. Energy Policy 2010; 38: 656–660.

14.

Soytas

Sari

Energy consumption and GDP: causality relationship in G-7 countries and emerging markets. Energy Econ 2003; 25: 33–37.

15.

Burke

Csereklyei

Understanding the energy-GDP elasticity: a sectoral approach. Energy Econ 2016; 58: 199–210.

16.

Borozan

Exploring the relationship between energy consumption and GDP: evidence from Croatia. Energy Policy 2013; 59: 373–381.

17.

Yuan

J-H

Kang

J-G

Zhao

C-H

, et al. Energy consumption and economic growth: evidence from China at both aggregated and disaggregated levels. Energy Econ 2008; 30: 3077–3094.

18.

Zhang

Retesting the causality between energy consumption and GDP in China: evidence from sectoral and regional analyses using dynamic panel data. Energy Econ 2012; 34: 1782–1789.

19.

Zhang

X-P

Cheng

X-M.

Energy consumption, carbon emissions, and economic growth in China. Ecol Econ 2009; 68: 2706–2712.

20.

Narayan

Popp

The energy consumption-real GDP nexus revisited: empirical evidence from 93 countries. Econ Model 2012; 29: 303–308.

21.

C-Y

Siu

KW.

A dynamic equilibrium of electricity consumption and GDP in Hong Kong: an empirical investigation. Energy Policy 2007; 35: 2507–2513.

22.

Hayashi

Hughes

The policy responses to the Fukushima nuclear accident and their effect on Japanese energy security. Energy Policy 2013; 59: 86–101.

23.

Tanaka

Big Bang in Japan’s energy policy. Energy Strategy Rev 2013; 1: 243–246.

24.

Lee

Causal relationship between energy consumption and GDP revisited: the case of Korea 1970–1999. Energy Econ 2004; 26: 51–59.

25.

Belke

Dobnik

Dreger

Energy consumption and economic growth: new insights into the cointegration relationship. Energy Econ 2011; 33: 782–789.

26.

Hondroyiannis

Lolos

Papapetrou

Energy consumption and economic growth: assessing the evidence from Greece. Energy Econ 2002; 24: 319–336.

27.

Damette

Seghir

Energy as a driver of growth in oil exporting countries?.

Energy Econ 2013; 37: 193–199.

28.

Chow

CL.

A study of sectoral energy consumption in Hong Kong (1984–97) with special emphasis on the household sector. Energy Policy 2001; 29: 1099–1110.

29.

Hartley

Tortajada

Biswas

AK.

Political dynamics and water supply in Hong Kong. Environ Dev 2018; 27: 107–117.

30.

Charrieras

Darchen

Sigler

The shifting spaces of creativity in Hong Kong. Cities 2018; 74: 134–141.

31.

Tso

GKF

Yau

A study of domestic energy usage patterns in Hong Kong. Energy 2003; 28: 1671–1682.

32.

Granger

Investigating causal relations by econometric models and cross-spectral methods. Econometrica 1969; 37: 424.

33.

Granger

Testing for causality: a personal viewpoint. J Econ Dyn Control 1980; 2: 329–352.

34.

Dickey

Fuller

WA.

Distribution of the estimators for autoregressive time series with a unit root. J Am Stat Assoc 1979; 74: 427–431.

35.

Dickey

Fuller

WA.

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

36.

Phillips

PCB

Perron

Jun

Testing for a unit root in time series regression. Biometrika 2007; 75: 335–346.

37.

MacKinnon

JG.

Numerical distribution functions for unit root and cointegration tests. J Appl Econ 1996; 11: 601–618.

38.

Johansen

Juselius

Maximum likelihood estimation and inference on cointegration – with applications to the demand of money. Oxf Bull Econ Stat 1990; 52: 169–210.

39.

Johansen

Likelihood-based inference in cointegrated vector autoregressive models. New York: Oxford University Press, 1995.

40.

Lutkepohl

Sakkonnen

Trenkler

Maximum eigenvalue versus trace tests for the cointegrating rank of a VAR process. Econmetr J 2001; 4: 287–310.

41.

Hjalmarsson

Osterholm

Testing for cointegration using the Johansen methodology when variables are near-integrated. Washington, DC: International Monetary Fund, 2007.

42.

Hassapis

Pittis

Prodromidis

Unit roots and Granger causality in the EMS interest rates: the German dominance hypothesis revisited. J Int Money Fin 1999; 18: 47–73.

43.

Shi

Alexiadis

Testing the real interest parity hypothesis in six developed countries. Int Res J Fin Econ 2012; 86: 168–180.

44.

Hendry

Juselius

Explaining cointegration analysis: Part II. Energy J 2000; 22: 75–120.

45.

Chow

Pei

Fong

, et al. Modeling and application of direct-expansion solar assisted heat pump for water heating in subtropical Hong Kong. Appl Energy 2010; 87: 643–649.

46.

Jia

Lee

WL.

Drivers of moderate increase in cooling energy use in residential buildings in Hong Kong. Energy Build 2016; 125: 19–26.

47.

Kwok

Wong

PT.

Complying with voluntary energy conservation agreements (I): air conditioning in Hong Kong’s shopping malls. Resour Conserv Recycling 2017; 117: 213–224.

48.

Wong

Zhu

Liu

, et al. Estimation of Hong Kong’s solar energy potential using GIS and remote sensing technologies. Renew Energy 2016; 99: 325–335.

49.

Leung

MKH

Sumathy

, et al. Potential of renewable hydrogen production for energy supply in Hong Kong. Int J Hydrogen Energy 2006; 31: 1401–1412.

50.

Gaglia

Lykoudis

Argiriou

, et al. Energy efficiency of PV panels under real outdoor conditions – an experimental assessment in Athens, Greece. Renew Energy 2017; 101: 236–243.

51.

Gao

Yang

Study on offshore wind power potential and wind farm optimization in Hong Kong. Appl Energy 2014; 130: 519–531.

52.

EMSD. Study on the potential applications of renewable energy in Hong Kong, 2004.

53.

Salamaliki

Venetis

IA.

Energy consumption and real GDP in G-7: multi-horizon causality testing in the presence of capital stock. Energy Econ 2013; 39: 108–121.