Economic growth as a solution for global warming

(1)

Economic growth as a solution for global warming

Msc. Thesis Economics

Track: Monetary Policy and Banking

Tim Horsmeier

10047808

(2)

2 Verklaring eigen werk

Hierbij verklaar ik, Tim Horsmeier, dat ik deze scriptie zelf geschreven heb en dat ik de volledige verantwoordelijkheid op me neem voor de inhoud ervan.

Ik bevestig dat de tekst dit in deze scriptie gepresenteerd wordt origineel is en dat ik geen gebruik heb gemaakt van andere bronnen dan die welke in de tekst en in de referenties worden genoemd. De faculteit Economie en Bedrijfskunde is alleen verantwoordelijk voor de begeleiding tot het inleveren van de scriptie, niet voor de inhoud.

(3)

3

Abstract

This thesis investigates whether economic growth could be a solution for the CO2 emissions problem. It does so by researching whether the environmental Kuznets Curve applies for the relationship between CO2 emissions and economic growth. Next to that, the effect of individual energy sources is investigated. To that end, a cross country panel dataset (consisting of 18 countries for the time period 1971 – 2011) is subjected to methodology based on the paper of Grossman and Krueger (1994). The results show that the relationship between CO2 emissions and GDP is predominantly a positive one which stabilizes when countries reach a GDP of around 40.000 US Dollars. The results for the individual energy sources show that the use of coal has the most impact on CO2 emissions and the use of renewables has significantly negative impact on CO2. Therefore, this thesis concludes that economic growth will not be the solution for emission problem. In order to secure a sustainable future, the focus must be on reducing coal extraction and increasing renewable energy sources.

(4)

4

1. Introduction

Over the last few decades the concern of global warming and the consequences it could have for our planet grew within society. This has led to an ongoing discussion about whether we should actively try to reduce greenhouse emissions. On the one hand there is the opinion that dramatic action to reduce greenhouse gas emissions should be taken immediately as the consequences could potentially be catastrophic. The evidence for this view lies in the fact that the composition of the atmosphere has changed dramatically since the start of the industrial revolution, thus the development of all countries contributed greatly to the flow of greenhouse gasses into the atmosphere (Stern, 2007).

On the other hand the opponents of this view claim that climate change is not a human caused phenomena. Although some influential people on the world stage make this claim (for example: upcoming United States president Donald Trump (Business Insider Nederland, 2016)), a general consensus has been reached on political and scientific level: global warming and other climate change are human caused phenomena and pose an eminent danger to our planet in the nearby future (Stern, 2007).

The main goal of this thesis is to investigate the reduced form relationship between CO2 emissions and economic growth. It does so based on a research from Grossman and Krueger in 1994. They came up with the so called “environmental Kuznets curve”: an inverted U-shaped relationship between environmental deterioration and economic growth. This means that a country in its early stages of economic development has an increasing level of environmental degradation until it reaches a certain point of development. After that, the environmental degradation gradually decreases due to cleaner technologies and a change in behavior of the population. According to this theory economic growth can be seen as the solution against global warming and climate change. Grossman and Krueger (1994) did not estimate the relationship between CO2 emissions and economic growth, due to a lack of data. In the years after this research the World bank collected CO2 emission levels of a large group of countries in their database. Using this data, this thesis will fill that gap by estimating the reduced form relationship between CO2 emissions and economic growth.

By estimating this relationship the thesis will answer the following research question: to what extent could economic growth be a part of the solution for the problem of CO2 emissions?

Besides the relationship between CO2 emissions and GDP this research also estimates the relationship between CO2 and the use of fossil and renewable energy sources. A major topic within the environmental discussion is the use of fossil fuels and their impact on the environment (see Hassler et al., 2016). Figure 1 shows that most of the emissions of the last 45 years were the result of the use of

(6)

6 fossil fuels. This thesis will estimate the relationship between CO2 emissions and the major fossil fuels: oil and coal. Next to that, the relationship will be estimated between CO2 emissions and fossil fuel’s main clean substitute: renewables.

To estimate the reduced form relationships, described above, a panel data set is constructed with data from the top 21 countries with the highest total GDP. The time period of this data ranges from 1971 and 2011. This time period is interesting because the IPCC (2014) calculated that in this period more than half of the total CO2 was emitted (see figure 1). The reduced form relationship is estimated with CO2 emissions per capita and GDP (including quadratic and cubic GDP), a 3 years average of GDP and electricity productions from oil, coal and renewable sources.

The following section explains the consequences of the emission problem and summarises possible solutions that are linked to the empirical analysis in this thesis. In section 3 the regression formula and the corresponding variables are explained and next to that explanation about the dataset is given. Section 4 explains the use of the methodology. In section 5 the results of the empirical model are presented, analysed and discussed. Section 7 summarizes and concludes on all the findings of this thesis. Finally in section 8 suggestions for further research are given.

Figure 1. Global anthropogenic CO2 Emissions.

(7)

7

2. The topic in the literature

Before looking at possible solutions it is important to understand the problems that arise from an increasing level of CO2 emissions and other greenhouse gasses. The first part of this section gives a summary of these problems. After that, potential solutions are discussed that are presented in the literature about this topic.

2.1 The consequences of rising CO2 emissions

According to the Stern Review (2007) the current trend of CO2 and other greenhouse emission will have irreversible and very damaging effects to the global environment. Average global temperatures will rise by 2 - 3 degrees Celsius within the next 50 years and will rise even further if greenhouse gas emissions will continue to grow. These will have all sorts of impacts on the world. Melting glaciers will cause the sea level to rise. This increases flood risk and as a result of that reduce fresh water supplies, threatening one-sixth of the world’s population. These rising sea levels will also result in hundreds of millions more people being flooded each year. According to some estimations 200 million people need to move due to rising sea levels, heavier floods and more droughts. Declining crop yields could create a food shortage, especially in Africa. In some countries crop yields may increase due to a temperature rise of 2 - 3 degrees Celsius but when the temperature rises further global food production will be seriously affected. At this temperature change, ecosystems will be affected causing around 15 to 40 percent of species to face extinction. A direct result from rising CO2 levels is the acidification of the oceans causing threats to marine ecosystems and fish stocks (Stern 2007).

These effects of climate change are not evenly distributed. The developing world will be hit the hardest by the consequences of climate change. Developing countries are on average already warmer than developed countries and as a result further temperature raises will create more costs and lower benefits from climate change. Next to that developing countries are mostly dependent on agriculture, the most sensitive of all economic sectors to climate change. These vulnerabilities of climate change will further decelerate the reduce of poverty and increase illness and the number of deaths in developing countries. Global warming could potentially result in an overall loss of 5 to 10 percent of GDP with developing countries suffering losses over 10 percent (Stern, 2007).

The IPCC (2014) has made projections for the future level of greenhouse gas emissions and their effects on the environment. These projections show that a temperature rise below 2 degrees Celsius is already very unlikely. This could only be possible if the greenhouse gas emissions would be reduced by more than 50 percent over the next 40 years and even then it is more unlikely than likely. According to the Stern review (2007) a stabilization at a concentration of 550 parts per million (CO2-eq) would be the

(8)

8 best achievable goal. This would mean that the reduction of greenhouse gas emissions would at least have to be between 19 percent and 49 percent in 2050 and between 59 percent and 89 percent in 2100 (IPCC, 2014).

2.2 The environmental Kuznets curve

A potential solution for the problem of CO2 emissions follows from the research of Grossman and Krueger (1994). This research forms the cornerstone for the empirical analysis done in this thesis. Grossmann and Krueger examined the relationship of economic growth and the environment. Using data from the Global Environmental Monitoring System (GEMS) they studied the reduced form relationship between four environmental indicators and GDP per capita. The indicators representing the environment are concentrations of urban air pollution, measures of the state oxygen regime in river basins, concentration of fecal contaminants in river basins and concentrations of heavy metal in river basins. They find that in the beginning of economic growth the environment deteriorates followed by a stage of improvement after further economic growth. The shifting point for most pollutants occurs when a country’s GDP per capita reaches $8.000.

As mentioned above the data on the environment is collected by the Global Environmental Monitoring System (GEMS). Their data for three types of indicators related to water quality and one related to air quality form the basis for the research of Grossman and Krueger (1994). The air quality indicator is commonly used to measure air pollution in cities but omits air pollution that affects global conditions, for example CO2 emissions.

Grossman and Krueger (1994) estimated the relationship between environmental degradation and per capita income using a reduced from approach. They used various indicators of environmental degradation as depended variables and GDP per capita as an independent variable. Besides GDP per capita of a specific year, the average GDP per capita of the 3 years prior to a specific year was also included. The authors added this variable because they reasoned that past levels of income are an important factor for the current level of GDP.

The indicators of air quality, sulphur dioxide and smoke display show an inverted U-shaped relationship: pollution rises with GDP at low GPD but after it reaches a peak, pollution falls when GDP increases further. Sulphur dioxide and smoke peak at a relatively early stage in national development. For water pollution in rivers they find an inverted U-shaped relationship between income and three estimates of pollution and a U-shaped relationship for one of the estimates. Three of the pollution estimates peak at an income of at least $7.500 and fall after when income rises beyond that. For fecal contamination of rivers the results are quite similar. Fecal coliform shows constant levels with an rising

(9)

9 level of per capita income until the income reaches a level of about $8.000 and after that fecal contamination falls sharply with income. Only total coliform shows a surprising relationship: concentrations of total coliform rise with per capita income at first, then fall and then rise again. Grossman and Krueger have no explanation for these findings. Finally, the authors discuss heavy metals: the results show that for lead the relationship is downward sloping, for cadmium the relationship is flat and for arsenic the inverted U-shaped relationship.

The authors conclude that the results show no evidence that economic growth does irreversible damage to the environment. On the contrary, the authors find that increases in income at low per capita income is related to the deterioration of the air and water quality but when a certain point of income is reached further GDP growth benefits the environment. The relationship between income and environmental deterioration follows an inverted U-shape: the so called environmental Kuznets curve (EKC). The turning point varies for different pollutants, but in most of the cases the turning point occurs at a per capita GDP of $8.000 (1985 dollars.). When a country reaches a GDP per capita level of $10.000, there is a 95 percent probability that environmental deterioration decreases.

Grossman and Krueger (1994) finally focussed on three issues regarding their findings. First, they stated that improving environmental quality with growth is not an automatic process. Other research states that the most important reason for an improved environment is a policy change. When countries grow richer, their citizens care more about their living conditions. Richer countries therefore have more stringent environmental laws and standards and thus have relatively better air and water quality than poorer countries. Secondly, it is possible that the inverted U-shaped relationship is caused by trading patterns. Richer countries shift pollution-intensive production to poorer countries. This process cannot continue forever: developing countries will not always be able to find poorer countries to shift their pollution-intensive production to. However, the available evidence does not support this hypothesis: the volume of this kind of trade is too small to explain the reduced pollution that accompanies economic growth.

The final issue discussed is that developing countries today can benefit from the work done in the past. Improving technologies and political and economic knowledge make it possible for developing countries to learn from the past and focus on the environment at earlier stages of their development. 2.3 Critique on the environmental Kuznets curve

Although the environmental Kuznets curve serves as a benchmark in many studies, there are also researches disproving the model. The main issue of many researchers, for example Arrow et al. (1995), is that in the environmental Kuznets curve model income was assumed to be exogenous, so there is

(10)

10 no reaction from climate change to economic production. This means that climate change and environmental damage cannot affect economic production enough to stop economic growth and any irreversible deterioration to the environment has no effect on future income. But attempting to grow fast in early stages of development at the cost of the environment could be counterproductive. Economic activity always disrupts the environment because production of goods requires use of energy and material. Efforts to reduce climate change while keeping economic growth could stimulate other problems (Stern, 2004).

The effects of trade on environmental damage is also an important issue. Both Arrow et al. (1995) and stern et al. (1996) suggested that if the environmental Kuznets curve is valid, it is largely due to the effects of trade in polluting production between developed and developing countries. The authors state that under free trade, developing countries would turn to production of goods that need more labour and natural recourses and thus put a bigger strain on the environment. Developed countries would specialize in human capital and capital intensive production and this causes polluting production to shift from developed countries to developing countries. Environmental laws in developed countries make this shift even bigger. This process cannot continue endlessly. At some point as developing countries become wealthy, they will be unable to find poorer countries to switch their polluting production to. Thus, when more stringent environmental regulation is applied in these countries they cannot outsource their production but are forced to abate it and this has a negative effect on economic growth (Stern, 2004).

The review of Dasgupta et al. (2002) is the most important, according to Stern (2004). This research discusses multiple viewpoints regarding the environmental Kuznets curve. A viewpoint on the type of pollutants states that some traditional pollutants have an inverted U-shape relation while new ones, such as carcinogenic chemicals, do not. When the traditional pollutants are dealt with, new ones come up so total pollution is not reduced. Another viewpoint does not reject the EKC model but suggests that it shifts down due to technological improvement. But the main contribution of the research of Dasgupta et al. (2002), is providing evidence that environmental improvements are also possible in developing countries. According to this evidence the greatest increase of pollution regulation happens in developing countries. Further, liberalization of developing countries over the last twenty years has resulted in better use of national resources and less subsidy for environmentally damaging economic activity. Many multinationals experience pressure from investors and consumers to enforce better environmental standards in the countries where they invest and produce. Next to that, better information and more experience in regulating pollution has resulted in better methods to regulate pollution. Regulatory capacity of developing countries seems to be underestimated in the environmental Kuznets curve (Stern, 2004).

(11)

11 Stern (2004) concludes that the true form of the relationship between emission and income is a mix of the viewpoints described by Dasgupta et al. (2002). New pollutants may increase environmental degradation with income but over time this curve shifts down due to improved technology and better capacity to regulate emission, even in developing countries (Stern, 2004).

2.4 The environmental Kuznets curve for CO2 emissions

As discussed earlier, the research from Grossman and Krueger (1994) does not include pollutants that affect global conditions, such as CO2. The question remains whether the relationship between CO2 emissions and economic growth also follows the environmental Kuznets curve.

Since the introduction of the environmental Kuznets curve in the early nineties, several studies on this topic have been conducted. Selden and Song (1994) used a cross-country dataset to examine the relationship between four air pollutants and economic development. They find substantial support for an inverted-U relationship. This is in line with the environmental Kuznets curve. However, the authors did find that the turning point, where the pollutants start decreasing, was relatively high. Heil and Selden (2001) expanded on this research by making future trajectories based of the relationship between CO2 and economic growth. Using cross-country data from 1880 until 1990 they find a similar relationship.

On the other hand Holtz-Eaking and Selden (1995) found evidence suggesting a diminishing marginal propensity to emit CO2 when a country develops economically. The results of Bruyn et al. (1998) also were in contradiction with the environmental Kuznets curve. However, they claimed that CO2 emissions are positively correlated with economic growth.

If you analyse the results of these papers an important limitation comes forward. The datasets used in these researches are outdated. As figure 1 shows, CO2 emissions have made a substantial development over the last 10 to 20 years. This development could have altered the relationship between CO2 emissions and economic growth.

Recent studies shifted the focus from a cross country analysis to a country specific analysis. The results from these researches also fail to give a definitive conclusion. Fodha and Zaghdoud (2010) find a monotonically increasing relationship between CO2 and GDP for the case of Tunesia. Al-Mulali et al. (2015) used data for Vietnam and find the same results. However, for the case of China, Jalil and Mahmud (2009) did find evidence for a relationship in line with the environmental Kuznets curve. From the results in the literature it is impossible to give an unambiguous conclusion about the relationship between CO2 emissions and economic growth. The empirical analysis of this thesis will add to the current literature by doing a cross country analysis with the most recent the data available.

(12)

12 This way the thesis will try to provide a general conclusion about the relationship between CO2 and economic growth and its ability to be part of a solution for the emission problem.

2.5 Energy use as a solution

The previous part has shown that the estimations of relationship between CO2 emissions and economic growth are ambiguous. That is why it is important to investigate other solutions to the CO2 emission problem.

Stern (2007) comes up with several measures that will help decline CO2 emissions. First, it is important to reduce non-energy emissions. Preventing deforestation is a good example of such a measure. But the biggest improvement can be made in the energy sector. Two-third of total greenhouse gas emissions are related to the energy sector. A big improvement can be made in energy efficiency. Although in the last century a lot of progress has been made in energy efficiency there is still a lot of potential for efficiency improvements. Studies by the International Energy Agency indicate that in the next 50 years energy efficiency could be the biggest source of emission savings in the energy sector. This could save money and emissions. next to that, the demand for carbon intensive goods and services need to be reduced and the use of low carbon energy sources need to be increased. The IPCC (2014) calculated that low carbon energy supply has to be increased by 185 percent in 2050, towards more than 40 percent of total energy supply. And in 2100 total energy supply must consist of 80 percent low carbon energy.

Hassler et al. (2016) come up with an alternative approach. They state that not only reducing fossil fuels and increasing low-carbon energy supply could be an effective measure but the composition of the use of fossil fuels could also be part of the solution.

The authors state that the major threat to the environment comes from the use of coal reserves and not the use of oil or gas reserves. The current estimate of oil and gas reserves indicates a stock of 300 gigatons of CO2 (GtC). If the total supply of oil and gas reserves will be fully used up this will lead to a temperature increase of 0,7 degrees Celsius in the long run (Hassler et al., 2016). According to the estimations of the IPCC (2014), such a temperature rise will not cause irreversible damage to the environment. Coal reserves are considerably larger than oil and gas reserves. The official global coal reserves are 640 GtC. However, this could be an underestimation. A research from Rogner (1997), estimates that coal reserves could amount up to 3.500 GtC. If all these reserves would be used it will cause irreversible damage to the environment (Hassler et al., 2016).

Fortunately, coal extraction is very sensitive to price distortions. This is due to the fact that profits on coal are very low. The extraction costs are almost equal to market prices. For example a carbon tax

(13)

13 could have a large impact on coal extraction. For oil and gas, the opposite is the case. Extraction costs for oil and gas are much lower than the market price. A potential carbon tax would have a small effect on oil and gas extraction because the extraction is too profitable (Hassler et al., 2016).

Hassler et al. (2016) conclude that coal use should be discouraged and the best way to do this is to institute a carbon tax. This tax would like have little effect on the use of oil and gas but this is not a problem. The main goal of this tax is that it will limit the use of coal.

The conclusions in the paper of Hassler et al. (2016) are based on a theoretical and mathematical framework. This thesis will try to add to that research by providing an empirical analysis. It will do so by estimating the relationship between CO2 emissions and the use of oil and coal. In addition to that the relationship between CO2 emissions and the use of renewables will also be estimated.

3. Data description

In section 3 the variables of the regression used in this thesis are explained in detail and the sources and periodicity of the data is provided. Furthermore the choice of the countries included in the dataset is given.

3.1 Description of the variables

The regression formula used in thesis is as follows:

(1) 𝑪𝑶𝟐𝒊𝒕= 𝜷𝟎+ 𝜷𝟏 𝑮𝑫𝑷𝒊𝒕+ 𝜷𝟐 𝑮𝑫𝑷𝒊𝒕𝟐 + 𝜷𝟑 𝑮𝑫𝑷𝒊𝒕𝟑 + 𝜷𝟒 𝑮𝑫𝑷 𝑨𝑽̅̅̅̅̅̅̅̅̅̅̅𝒊𝒕 +

𝜷𝟓 𝑮𝑫𝑷 𝑨𝑽̅̅̅̅̅̅̅̅̅̅̅𝒊𝒕 𝟐 + 𝜷𝟔 𝑮𝑫𝑷 𝑨𝑽̅̅̅̅̅̅̅̅̅̅̅𝒊𝒕𝟑 + 𝜷𝟕 𝑶𝑰𝑳𝒊𝒕 + 𝜷𝟖 𝑪𝑶𝑨𝑳𝒊𝒕+ 𝜷𝟗 𝑹𝑬𝑵𝑬𝑾𝑨𝑩𝑳𝑬𝒊𝒕 + 𝜺𝒊𝒕

CO2 stands for CO2 emissions per capita in metric tons. GDP is measured per capita in 2010 constant US dollars. GDP AV is the average GDP per capita of the previous three years. Oil is the product of electricity production from oil sources (% of total) and total electric power consumption in kilowatt hours (kWh). Coal is the product of electricity production from coal sources (% of total) and total electric power consumption in kWh. Renewable is the product of electricity production of renewable sources, excluding hydroelectric (% of total) and total electric power consumption in kWh.

All data is taken from the World Bank Databank. The periodicity of the data is annual and the time period used in this thesis is from 1971 until 2011. More specific details about the data are explained below.

The variable GDP (Gross Domestic Product) is measured per capita in constant 2010 US Dollars. The sources for this data are the World Bank National Accounts data and OECD Accounts data files. The

(14)

14 World Bank’s definition for this data is: ‘‘GDP per capita is gross domestic product divided by midyear population. GDP is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources.’’ (World Bank, A, 2017) As stated above the base year for this data is 2010, the periodicity is annual and the aggregation method used is a weighted average.

The variable CO2 emissions is measured per capita in metric tons. The source for this data is The Carbon Dioxide Information Analysis Center, Environmental Sciences Division, Oak Ridge National Laboratory. The Worlds Bank’s definition is: ‘’Carbon dioxide emissions are those stemming from the burning of fossil fuels and the manufacture of cement. They include carbon dioxide produced during consumption of solid, liquid, and gas fuels and gas flaring.’’ (World Bank, B, 2017) The aggregation method is a weighted average.

The variable Oil is the product of two datasets. the first dataset is electricity production from oil sources (% of total). The source of this data is the International Energy Agency Statistics (OECD/IEA 2014). The definition of the world Bank is as follows: “Sources of electricity refer to the inputs used to generate electricity. Oil refers to crude and petroleum products. Electricity production from oil sources (% of total) is the share of electricity produced by oil and petroleum products in total electricity production which is the total number of KWh generated by power plants separated into electricity plants and CHP plants.’’ (World Bank, C, 2017) The periodicity of this data is annual. The second dataset is Electric power consumption. The source of this data is the International Energy Agency Statistics (OECD/IEA 2014). The definition of the world Bank is as follows: “Electric power consumption measures the Electricity production of power plants and combined heat and power plants less transmission, distribution, and transformation losses and own use by heat and power plants.” (World Bank, D, 2017) The periodicity is annual and the aggregation method is a weighted average. As stated above the variable Oil is the product of these two datasets.

The variable coal is the product of two datasets. The first dataset is electricity production from coal sources (% of total). The source of this data is the International Energy Agency Statistics (OECD/IEA 2014). The definition of the world Bank is as follows: Sources of electricity refer to the inputs used to generate electricity. “Coal refers to all coal and brown coal, both primary (including hard coal and lignite-brown coal) and derived fuels (including patent fuel, coke oven coke, gas coke, coke oven gas, and blast furnace gas). Electricity production from coal sources is the share of electricity produced by coal related products in total electricity production which is total number of kWh generated by power plants separated into electricity plants and CHP plants Peat is also included in this category.” (World

(15)

15 Bank, E, 2017) The periodicity of this data is annual. The second dataset is Electric power consumption. The source of this data is the International Energy Agency Statistics (OECD/IEA 2014). The definition of the world Bank is as follows: “Electric power consumption measures the Electricity production of power plants and combined heat and power plants less transmission, distribution, and transformation losses and own use by heat and power plants.’’ (World Bank, D, 2017) The periodicity is annual and the aggregation method is a weighted average. As stated above the variable Coal is the product of these two datasets.

The variable renewable is the product of two datasets. The first dataset is electricity production from renewable sources (% of total). The source of this data is the International Energy Agency Statistics (OECD/IEA 2014). The definition of the world Bank is as follows: “Electricity production from renewable sources (% of total) is the share of electricity produced by geothermal, solar photovoltaic, solar thermal, tide, wind, industrial waste, municipal waste, primary solid biofuels, biogases, bio gasoline, bio diesels, other liquid bio fuels, non-specified primary biofuels and waste, and charcoal in total electricity production which is the total number of GWh generated by power plants separated into electricity plants and CHP plants. Hydropower is excluded.” (World Bank, F, 2017) The second dataset is Electric power consumption. The source of this data is the International Energy Agency Statistics (OECD/IEA 2014). The definition of the world Bank is as follows: “Electric power consumption measures the Electricity production of power plants and combined heat and power plants less transmission, distribution, and transformation losses and own use by heat and power plants.’’ (World Bank, D, 2017) The periodicity is annual and the aggregation method is a weighted average. As stated above the variable Coal is the product of these two datasets. The time period for this data used in this thesis is 1971 until 2011.

3.2 Content of the dataset

The dataset used in this thesis consists of the top 21 countries with the highest total GDP. The main reason to use the countries with the highest GDP is that these countries have made a big transition in as well CO2 Emissions as in GDP per capita. The ten most polluting countries in the world are also part of the top 21 countries with the highest GDP. A second reason to choose these countries is the availability of data. Especially data on CO2 emissions is not widely available for all countries within the time period 1971 to 2011. This time period is an interesting one. The IPCC (2014) calculated that that more than half of total emitted CO2 since 1750, was emitted between 1970 and 2010 (see figure 1). Germany and Russia are excluded from the dataset due to a lack of data within the time period 1971 to 2011. Data for Saudi Arabia are treated as outliers and it is decided to exclude them from the dataset. Saudi Arabia’s economy is heavily depended on oil. It has the world largest proven oil reserves

(16)

16 and is one of the world largest producers and consumers of electricity. The reason for the high consumption partly lies in the domestic pricing policies that keep energy prices far below international prices (Alyousef & Stevens, 2011). This results in electricity production from oil sources that is far bigger than other countries relative to the size of the economy and CO2 emissions. Next to that almost all electricity production comes from oil sources and nothing from coal sources. If you include data for Saudi Arabia, this will potentially result in an significant overestimation of the effect of the use of oil on CO2 emissions. The results of the regression will then be biased. Therefore it is chosen to omit the data of Saudi Arabia.

The 18 countries included in the dataset are the United States, China, Japan, the United Kingdom, France, India, Italy, Brazil, Canada, Republic of Korea (South Korea), Australia, Spain, Mexico, Indonesia, the Netherlands, Turkey, Sweden and Argentina. Per country the dataset consists of 41 observations so the total number of observations are 738.

4. Methodology

The methodology section of the thesis explains how the analysis of the reduced form relationship between the depended and independent variables is measured and what steps are taken to come to this set-up.

The analysis in this thesis has two goals. The main objective is to estimate the reduced form relationship between CO2 emissions and economic growth. The other objective is to estimate the effect of the use coal, oil and renewable sources in the electricity production of a country. As seen in the previous section, the formula used in this thesis is based on the formula used in the paper of Grossman and Krueger (1994). With the addition of the variables for electricity production from oil, coal and renewables sources. The use of a random effects model is also in accordance with Grossman and Krueger (1994).

(1) 𝑪𝑶𝟐𝒊𝒕= 𝜷𝟎+ 𝜷𝟏 𝑮𝑫𝑷𝒊𝒕+ 𝜷𝟐 𝑮𝑫𝑷𝒊𝒕𝟐 + 𝜷𝟑 𝑮𝑫𝑷𝒊𝒕𝟑 + 𝜷𝟒 𝑮𝑫𝑷 𝑨𝑽̅̅̅̅̅̅̅̅̅̅̅𝒊𝒕 + 𝜷𝟓 𝑮𝑫𝑷 𝑨𝑽̅̅̅̅̅̅̅̅̅̅̅𝒊𝒕 𝟐 +

𝜷𝟔 𝑮𝑫𝑷 𝑨𝑽̅̅̅̅̅̅̅̅̅̅̅𝒊𝒕𝟑 + 𝜷𝟕 𝑶𝑰𝑳𝒊𝒕 + 𝜷𝟖 𝑪𝑶𝑨𝑳𝒊𝒕+ 𝜷𝟗 𝑹𝑬𝑵𝑬𝑾𝑨𝑩𝑳𝑬𝒊𝒕 + 𝜺𝒊𝒕

To study the relationship between pollution through CO2 emissions and economic growth this thesis estimates a reduced form equation that relates the level of CO2, in country i and year t, to current and lagged GDP per capita (GDP AV) of a country. This lagged GDP consists of the average of GDP per capita in the three previous years. According to Grossman and Krueger (1994) the use of lagged and cubic GDP should be included to account for the effect of permanent income and because it is likely that

(17)

17 past income is a relevant factor in the current environmental state. The use of a reduced form relationship has two major advantages. Firstly the reduced form approach straight away estimates the net effect of a nation’s GDP on CO2 emissions. Secondly with the reduced form approach it is not needed to collect data on, for example, pollution regulation and the state of technology, which are not always easily available and could be of doubtful validity. A downside of the reduced form approach is that it does not give an explanation why the relationship between CO2 emission and GDP is as it is estimated (Grossman and Krueger, 1994).

All the variables are taken per capita. This way the results are controlled for population growth. This makes it easier to compare the countries and the years with each other. Besides population, GDP is also controlled for inflation. This is done by measuring all data for GDP in constant 2010 US dollars. This way different years in time can be compared to each other.

For the relationship between CO2 emissions and the electricity production from oil, coal and renewable sources, the oil, coal and renewable electricity production as a percentage of the total electricity production is used. If the variable is used in this way, then the results could be biased. The reason for this is that a variable with electricity production from oil, coal or renewable sources as a percentage of total production is a non-trending variable while all the other variables (including the dependent variable) are trending variables. This could lead to potentially biased results. To account for this, the percentages are scaled using data for electric power consumption per capita measured in kWh. As a result of this scaling, the variables for oil, coal and renewable sources are trending and therefore fit better into the regression.

5. Results

In this section the results of the regression are displayed and explained. From there the findings are linked to the relevant literature and used to give an answer to research question.

5.1 Analysis of the countries separately

Before equation 1 is estimated and displayed, the next part first shows the data for a selection of countries out of the data set. Data for a total of six countries is analyzed in a scatter plot with a trend line. A separate estimation of equation 1 for each of the countries will not be useful due to the fact that the number of observations (41) is too small to give clear and significant results. The six countries that will be analyzed are the United States, China, India, Brazil, Republic of Korea and the Netherlands. These countries fall into three levels of economic development measured by GDP per capita. China

(18)

18 and India are still in the early stages of development. Their GDP per capita has not exceeded 5.000 US dollars (constant 2010 US Dollars). Brazil and Korea developed from being a country in an early stage of development to a more advanced state of the economy (GDP per capita above 10.000 US dollars in 2011). The United States and the Netherlands are two well developed countries with higher GDP per capita (above 40.000 US dollars in 2011).

Figure 2 and 3 show the data for the relationship between CO2 emissions per capita and GDP per capita for China and India respectively. Each blue dot represents the country in a specific year in time. Both countries experience increasing levels CO2 emissions per capita when their GDP per capita grows. China went from a emission level of around 1 kiloton in 1971 to a level 6,5 kilotons in 2011 while GDP per capita grew from 234 US dollars to 4.900 US dollars in 2011. India’s emissions went up from 0,3 kiloton to 1,7 in 2011 and GDP per capita grew from 380 US dollar in 1971 to 1.500 US dollars in 2011. The difference between the two countries is that China seems to have more of a linear trend in the data. This means that the level of increase of CO2 emissions per capita stays the same when GDP per capita rises. On the other hand in India’s CO2 emissions per capita follow a more concave trend. The level of CO2 emissions per capita increase reduces when GDP per capita goes up.

(19)

19 Figure 4 and 5 display a scatterplot for Korea and Brazil. According to the data both countries have experienced an increase of CO2 emissions per capita when the GDP per capita developed, just as India and China. Korea experienced a high growth of GDP in comparison to the other countries. GDP per capita increased from 2.100 in 1971 to 23.000 US dollars in 2011. Brazil had a smaller growth level of GDP: from 5.000 US dollars in 1971 to 11.500 in 2011. This higher GDP growth for Korea also seem to have caused a much larger increase in CO2 emissions. While Brazil had an increase in CO2 emissions from 1 kilotons per capita to 2,1 kilotons in 2011, Korea’s CO2 emissions per capita increased from 1,8 kilotons in 1971 to 11,8 kilotons in 2011. However there is a difference in the trend that the relationship follows for both countries. The trend for Brazil is linear which means that the level of CO2 emissions increase does not diminishes when GDP per capita rises. On the other hand Korea has a concave shaped trend so the level of CO2 emissions increase does diminishes when GDP per capita rises.

(20)

20 The scatterplots for the United States and the Netherlands show completely different trends (see figure 6 and 7). Both countries showed decreasing CO2 emissions per capita when GDP per capita increases which stabilizes with higher levels of GDP per capita. CO2 emissions per capita decreased from 21 kilotons in 1971 to 17,1 in 2011 for the United States and from 10,6 kilotons in 1971 to 10

Figure 4. Korea: scatterplot with CO2 emissions p.c on y-axis and GDP p.c on x-axis.

(21)

21 kilotons in 2011 for the Netherlands, while GDP per capita increased from 24.000 US dollars to 48.000 in 2011 for the United states and from 25.000 towards 51.000 US dollars for the Netherlands. However both countries experienced a decreasing GDP per capita in the period between 2008 and 2011. The trend for the relationship between CO2 emissions and GDP per capita seem to have stabilized.

Figure 7. Netherlands: scatterplot with CO2 emissions p.c on y-axis and GDP p.c on x-axis. Figure 6. U.S.: scatterplot with CO2 emissions p.c on y-axis and GDP p.c on x-axis.

(22)

22 The data of these six countries partly supports the environmental Kuznets curve of Grossman and Krueger (1994). In the early stages of development, CO2 emissions are rising when GDP increases and when a country is more developed CO2 emissions stabilize or even decrease slightly. However the concave trend that the environmental Kuznets curve follows does not come forward for all the countries. Only the data for India and Korea follow a concave trend line.

5.2 Regression results for the whole dataset

Each variable of the equation (1) described in section 4 is estimated in table 1. In this table you see the estimated effect of the independent variables GDP per capita, average GDP per capita and scaled electricity production from oil, coal and renewable sources on the dependent variable CO2 emissions per capita. Besides the estimated effect of the independent variables the table also displays the result of the t-test and the p-values correspondent to each of the variables. The results in table 1 show that the coefficients for the current and lagged GDP per capita are significant at a 5 percent level, with the coefficient for GDP per capita being highly significant (t-value = 7,65, p-value = 0,00). This indicates that national income per capita appears to be an important factor in the level of CO2 emissions per capita. The R-squared of 0,8562 shows the variance of the model is relatively well explained. The model fits the data well. The result of the F-test (F(9, 728) = 481.65 ) rejects the null-hypothesis that all of the variables are equal to zero with a significance level of 1 percent.

Variable Coefficient t-value (p-value)

GDP per capita 0,0001081 t = 7,65

(p = 0,000)

GDP per capita^2 1,28e-08 t = 1,99

(p = 0,047)

GDP per capita^3 -2,15e-13 t = -2,17

(p = 0,030)

GDP per capita average 0,0001647 t = 3,25

(p = 0,001)

GDP per capita average^2 - 1,58e-08 t = - 2,40

(p = 0,016)

GDP per capita average^3 2,21e-13 t = 2,16

(p = 0,031)

Oil 0,0012599 t = 7,22

(p = 0,000)

(23)

23 Coal 0,0017021 t = 32,34 (p = 0,000) Renewable - 0,0018665 T = -2,80 (p = 0,005) Constant 0,5441464 t = 2,63 (p = 0,009)

R-squared = 0,8562 Number of observations = 738 F(9, 728) = 481.65 (0,000) The results show that the coefficients of current and lagged GDP per capita are relatively similar and show a positive relationship between CO2 emissions per capita and GDP per capita. But the coefficients for quadratic and cubic GDP differ. The quadratic form of current GDP per capita is positive and that of cubic GDP per capita negative. This contradicts with the findings for the coefficients of lagged GDP per capita which are the other way around. A negative quadratic function gives evidence for a concave shaped trend curve as described by Grossman and Krueger (1994) with their environmental Kuznets curve. This negative quadratic function is found with the relationship between lagged GDP per capita and CO2 emission per capita but is missing in the relationship with current GDP per capita. The results for the coefficients for the quadratic and cubic variables of current GDP per capita do not indicate that the relationship between CO2 emissions and GDP is described by a concave shaped curve.

(24)

24 In figure 8 a scatter plot is displayed with CO2 emissions in metric tons per capita on the y-axis and GDP per capita in constant 2010 US dollars on the x-axis. Each blue dot represents a country in a specific year in time. In the figure it is shown that most countries that have a lower GDP per capita in a specific year in time also had a lower emission level per capita of CO2. When the countries reach a higher level of GDP per capita, CO2 emissions per capita generally also are at a higher level. From a certain point CO2 emissions per capita seem to stabilize as the countries have a higher level of GDP per capita. After that point of stabilization, the figure does not show a significant decline in CO2 emissions per capita when countries reach a higher level of GDP per capita.

The results of table 1 and figure 8 show that the reduced form relationship between CO2 emissions per capita and GDP per capita is predominantly a positive one which seems to stabilize when a country reaches a certain point of development (around 40.000 US dollars). Significant evidence for a negative relationship between CO2 emissions and GDP per capita after that point of stabilization is not found. Therefore economic development alone is unlikely to solve the problem of CO2 emissions.

5.3 The relationship between CO2 emissions per capita and energy production

The second part of the regression (see table 1) estimates the relationship between CO2 emissions per capita and a country’s electricity production from oil, coal and renewable sources scaled with electricity consumption per capita (measured in kWh). The coefficients for this variables are highly significant [the t-values (and p-values) are -2,80 (0,005), 7,22 (0,000) and 35,34 (0,000) for renewable, oil and coal respectively]. Especially the t-value for coal is very high. This means that electricity production from renewable sources, oil sources and especially coal sources are relevant determinants for the level of CO2 emissions per capita of a country.

Looking at the coefficients of the variables, the results show that oil and coal have a positive effect on CO2 emissions while electricity production from renewable sources has a negative effect. So, in line with what you expect, when electricity production from oil and coal sources increases, CO2 emissions per capita also increase. And when electricity production from renewable sources increase, CO2 emissions per capita decrease. The value of the coefficient of electricity production from oil is smaller than that of coal (respectively 0,00126 and 0,0017). This result indicates that each kilowatt hour of electricity per capita produced with coal leads to more CO2 emissions than a kilowatt hour produced with oil. This result provides basic empirical evidence for the findings of Hassler et al. (2016) that coal is the major threat to the environment. Not only because the reserves of coal are far greater than that of oil but also because the use of coal simply creates more CO2 emissions than the use of oil. The value of the coefficient for electricity from renewable sources is - 0,0018. This significant negative

(25)

25 relationship supports the suggestion of the IPCC (2014) that increasing low-carbon energy use will diminish CO2 emissions.

The results provide basic empirical evidence for the conclusions of Hassler et al. (2016) and the IPCC (2014). If the focus shifts towards decreasing coal extraction and besides this, the use of renewable energy use increase, global CO2 emissions will go down. These measures could be a solution for the emissions problems society faces.

6. Summary and conclusions

In the last 25 years, the discussion about global warming intensified. A general consensus has been reached that global warming and the rise of CO2 emissions are human-caused phenomena and action has to be taken in order to limit the damage of these problems in the future (Stern, 2007).

If you look at the future trends, the current path of greenhouse gas emissions will cause major problems globally in the nearby future: from declining living spaces to food shortages. Stern (2006) calculated that this will lead to a GDP loss of at least 5 to 10 percent.

A potential solution for this problem could be economic development. This solution is based on a paper of Grossman and Krueger (1994). They researched the relationship between pollution and GDP and came to the conclusion that economic development could be the solution for the environmental problems. They find no evidence that the environment deteriorates with economic growth. Instead they find that in early stages of economic development the environmental indicators deteriorated. This was followed by an improvement of the environment when the economy grows further. They presented their findings in the so called “environmental Kuznets curve”. The turning point for most of the economic indicators was at a per capita income of 8.000 US dollars.

Over time some studies disproved the environmental Kuznets curve of Grossman and Krueger. The main critique came from Dasgupta et al. Who claimed that developing countries already have the means to make environmental improvements causing these countries to perform better than the results of Grossman and Krueger indicate.

The research of Grossman and Krueger did not include CO2 emissions. But in the last 25 years several other researches did test the environmental Kuznets curve for CO2. The results from those researches are ambiguous. On the one hand there are studies that find evidence for an environmental Kuznets curve (Selden and Song, 1994; Heil and Selden, 1994; and Jalil and Mahmud, 2009). On the other hand papers from Bruyn et al. (1998), Fodha and Zaghdoud (2010) and Al-Mulali et al. (2015) find a positive

(26)

26 relationship. Holtz-Eaking and Selden (1995) even find evidence suggesting a diminishing marginal propensity to emit CO2 when a country develops economically.

Because there is not yet a clear answer on the relationship between CO2 and economic growth, it is important to investigate other possible solutions. Stern (2007) suggests measures such as increasing energy efficiency, stopping deforestation and increasing low-carbon energy supplies. However, Hassler et al. (2016) have an alternative approach. They suggest that the composition in which fossil fuels are used could provide a solution for the emission problem. The problem of fossil fuels mainly lies with the use of coal and not oil. The main reason for this is that oil resources are very limited while there are coal resources in abundance. Hassler et al. (2016) conclude that the focus must be on reducing coal and that a small carbon tax would be sufficient to reduce coal extraction greatly.

This thesis estimates the reduced form relationship between GDP per capita and the environmental indicator that Grossman and Krueger did not include (due to a lack of data): CO2 emissions (per capita). In addition to this, this thesis tries to estimate the effect of different energy sources on CO2 emissions. It does so by using a random effects model, based on the model of Grossman and Krueger. The data used is a cross country panel dataset for the time period 1971 – 2011. The countries in this dataset are the top 21 countries when it comes to GDP, excluding Russia, Germany and Saudi Arabia.

The results of the regression show that the relationship between CO2 emissions per capita and GDP per capita is mainly a positive one. In early stages of economic development CO2 emissions increase when GDP per capita rises and at a certain point the emission level stabilizes with more GDP growth. The average turning point lies at approximately 40.000 dollars (constant 2010 US dollars). Significant evidence for decreasing CO2 emissions after this turning point does not come forward out of the results. These findings contradict with the environmental Kuznets curve model. First of all the turning point of stabilization in this model lies at 8.000 US dollars. This is significantly lower than the outcome in this thesis, even taking into account the 1985 US dollar level. Secondly the declining pollution after the turning point in the environmental Kuznets curve is not found for the relationship between CO2 emissions and GDP per capita. Based on these results it can be concluded that economic development will not solve the emission problems the world faces. CO2 emissions will not decline but stabilize and major polluting countries such as China, India and Brazil are still far away from stabilizing their CO2 emissions.

The results of the second part of the regression show that the use of coal has a significantly greater positive effect on CO2 emissions than the use of oil. This is in line with the conclusion of Hassler et al. (2016) that coal is the biggest problem. The results on the relationship between CO2 and the composition of energy sources provide basic empirical evidence that the focus should be on reducing

(27)

27 coal use. However the results also confirm that the use of renewable sources has a negative impact on CO2 emissions. Therefore, in order to reduce emissions, the focus should not only be on reducing coal use but increasing the use of renewables should also remain an important goal.

The main question in this thesis is ‘to what extent can economic growth be part of a solution for the emission problem?’ The results of the empirical analysis in this thesis give a clear answer: economic growth does not lead to a fall in CO2 emissions and therefore is not a solution. CO2 emissions only seem to stabilize at a high level of GDP and many countries are still far away from this stabilizing point. Therefore, alternative measures should be taken to ensure the reduction of emissions. The solution from Hassler et al. (2016) to focus on reducing coal extraction is a plausible one. This thesis found empirical evidence that reducing coal use, together with increasing the use of renewables will lead to a reduction in CO2 emissions.

One thing is certain: immediate action is necessary to limit the irreversible damage caused by global warming. In order to do so, the emission of CO2 and other greenhouse gasses has to be dramatically reduced. Due to the positive relationship between CO2 and GDP, this will probably come at the cost of GDP. But these costs are negligible in comparison with the costs society faces if we do nothing.

7. Suggestions for further research

Society still faces a lot of challenges with regards to global warming and other environmental problems. Further research on this topic remains vital in securing a sustainable future.

For future research it could be important to estimate the factors that shape the relationship between CO2 and GDP. This thesis uses a reduced form approach for the relationship between CO2 emissions and GDP. While this approach gives clear results on the effect of GDP on CO2, it provides little information about the driving factors that shape this relationship. This information could potentially be very useful in solving the environmental problems while keeping costs for GDP limited.

8. References

Al-Mulali, U., Saboori, B., Ozturk, I., (2015). Investigating the environmental Kuznets curve hypothesis in Vietnam. Energy Policy, vol. 76, pp: 123 – 131, January 2015

Alyousef, Y., Stevens, P., (2011). The cost of domestic energy prices to Saudi Arabia. Energy Policy vol. 39, no. 11, pp 6900-6905, November 2011.

(28)

28 Arrow, K., Bolin, B., Costanza R., Dasgupta, P., Folke, K.G., Perrings, C.A., Pimentel, D., (1995). Economic growth, carrying capacity, and the environment. Science vol. 268, pp: 520 – 521, May 1995.

Bruyn, de, S.M., Bergh, van den, J.C.J.M., Opschoor, J.B., (1998). Economic growth and emissions: reconsidering the empirical basis of environmental Kuznets curves. Ecological Economics vol. 25, pp: 161 – 175, April 1998.

Business Insider Nederland (2016). President-elect Donald Trump doesn’t believe in climate change. Here’s his platform on the environment. November 2016. https://www.businessinsider.nl/donald-

trump-climate-change-global-warming-environment-policies-plans-platforms-2016-10/?international=true&r=US

Dasgupta, S., Laplante, B., Wang, H., Wheeler, D., (2002) Confronting the environmental Kuznets curve. Journal of Economic Perspectives. Vol. 16, pp: 147 – 168, spring 2002.

Fodha, M., Zaghdoud, O., (2010). Economic growth and pollutant emissions in Tunisia: An empirical analysis of the environmental Kuznets curve. Energy Policy vol. 38, pp: 1150 – 1156, December 2010. Grossman, G.M., Krueger, A.B., (1994). Economic Growth and the Environment. NBER Working Paper series nr. 4634. National Bureau of Economic Research, Februari 1994.

Hassler, J., Krussell, P., Nycander, J., (2016). Climate policy. Economic Policy vol. 31, pp: 503 – 558, October 2016.

IPCC, (2014). Climate Change 2014 Synthesis Report, Summary for Policymakers. 2014

Holtz-Eakin, D., Selden, T.M., (1995). Stoking the fires? CO2 emissions and economic growth. Journal of public economics vol. 57, pp: 85 – 101, June 1995.

Jalil, A., Mahmud, S., (2009) Environment Kuznets curve for CO2 emissions: A cointegration analysis for China. Energy Policy vol. 37, pp: 5167 – 5172, January 2009.

Rogner, H.H., (1997). An assessment of world hydrocarbon recourses. Annual Review of Energy and the Environment, vol. 22 pp: 217 – 262, November 1997.

Selden, T.M., Song, D., (1994). Environmental quality and development: Is there a Kuznets curve for air pollution emissions? Journal of Environmental Economics and Management, vol. 27, pp: 147 – 162, September 1994.

Stern, D.I., Common, M.S., Barbier, E.B., (1996). Economic growth and environmental degradation: the environmental Kuznets curve and sustainable development. World Development vol. 24, pp: 1151-1160, December 1996.

(29)

29 Stern, D.I., (2004). The Rise and Fall of the Environmental Kuznets Curve. World Development vol. 32, no. 8, pp 1419 – 1439, July 2004.

Stern, N.H., (2007). The economics of climate change: the Stern review. Cambridge university press, 2007.

World Bank, A, (2017). Data about CO2 emissions per capita. http://data.worldbank.org/indicator/EN.ATM.CO2E.PC

World Bank, B, (2017). Data about GDP per captia in constant 2010 US dollars. http://data.worldbank.org/indicator/NY.GDP.PCAP.KD

World Bank, C, (2017). Data about electricity production from oil (% of total) http://data.worldbank.org/indicator/EG.ELC.PETR.ZS

World Bank, D, (2017). Data about electric power consumption in KWH per capita. http://data.worldbank.org/indicator/EG.USE.ELEC.KH.PC

World Bank, E, (2017). Data about electricity production from coal (% of total) http://data.worldbank.org/indicator/EG.ELC.COAL.ZS

World Bank, F, (2017). Data about electricity production from renewable sources, excluding hydroelectric (% of total). http://data.worldbank.org/indicator/EG.ELC.RNWX.ZS

(30)

30

9. Appendix

Scatterplots of the other countries in the dataset.

Figure 9. Japan: scatterplot with CO2 emissions p.c on y-axis and GDP p.c on x-axis.

(31)

31

Figure 11. Italy: scatterplot with CO2 emissions p.c on y-axis and GDP p.c on x-axis.

(32)

32

Figure 13. Spain: scatterplot with CO2 emissions p.c on y-axis and GDP p.c on x-axis.

(33)

33

Figure 15. Indonesia: scatterplot with CO2 emissions p.c on y-axis and GDP p.c on x-axis.

(34)

34

Figure 17. Argentina: scatterplot with CO2 emissions p.c on y-axis and GDP p.c on x-axis.

(35)

35

Figure 19. France: scatterplot with CO2 emissions p.c on y-axis and GDP p.c on x-axis.

Economic growth as a solution for global warming