How Well does a Low-Carbon Index Performs Compared to its Benchmark Stoxx 600 Index?

(1)

1 Faculty of Economics and Business

How well does a low-carbon index

perform compared to its

benchmark Stoxx 600?

Abstract: With growing evidence of Climate change and its effect on the globe, investors are turning away from a conventional investment strategy to sustainable investment methods to protect against the risk associated with climate change. This research, in particular, looks at one such strategy to hedge against climate\carbon risk for long term investors inspired by the research literate of Andersson et al. (2016). Using the hedging model of Andersson et al. (2016) we construct a carbon-efficient portfolio and test its performance against the benchmark index of STOXX 600 for 3 and 5 year period data. The results suggest that the low carbon portfolio significantly outperforms STOXX 600 while keeping the Tracking Error low for both period dataset.

Study Programme : BSc Economics And Business Economics

Specialization: Economics

Name: Shubham Shreshtha

Student Number: 11600829

(2)

2

Introduction:

In recent times the issue of climate change is one of the most debated topics in the 21st_{-century society. Climate change is a complex global hazard that serves as} a significant challenge to society worldwide, the extent of this adverse challenge is perceived as an immediate risk varies significantly in the global population (Kim & Wolinsky-Nahmias, 2014; Swim et al., n.d.). Thus, having a polarizing effect on the members of the society making the issue of climate change that threatens the entire human civilization a partisan issue. This effect can be seen in the September 2019 Climate strike which took place from 20-27 September. This strike is one of the largest organised climate protests with up to 6 million participants and took place in almost 150 countries worldwide (Barclay & Resnick, 2019; Taylor et al., 2019). In the realm of investment and finance, the climate activists propose a divestment from fossil fuel and carbon-intensive firms to mitigate climate change and its consequences by reducing the carbon emission through divesting away from carbon-intensive firms. The divestment movement has reached its peak last year as the value of investment funds committed to selling off fossil fuel assets has jumped to $5.2tn, doubling in just over a year (The Guardian, 2016). The call for divestment intently identifies with the scientific and political discussion about the requirement for worldwide action to mitigate hazardous anthropogenic environmental change (Arbuthnott & Dolter, 2013). But at the same time, climate change scepticism has also increased with the advent of populism in the realm of politics. This induces uncertainty for investors on whether to price the risk of climate change policies, but this dilemma is more likely experienced by long term passive investors.

The main aim of this thesis is to put forward an investment strategy that does not involve divestment. The divestment movement posits a moral reason towards investors to do their part by considering the ecological impacts of the activities they finance next to traditional risk-return measures and therefore withdraw investments in publicly listed coal, oil, and gas companies (J. Ritchie & Dowlatabadi, 2014). Conforming to these moralistic values can be costly and or problematic for investors (Justin Ritchie & Dowlatabadi, 2015). This motivation’s main subject is long term passive investors such as pension funds, insurance firms and sovereign wealth funds as divestment leads to sacrificing financial returns to minimise the carbon footprint of the portfolio. This has a risk of the fund underperforming its benchmark due to divestments from high carbon

(4)

4

stocks which can have a detrimental consequence for the investment fund (Andersson et al., 2016). Thus, the main motive of the thesis is to create a simple dynamic investment plan which allows long term passive investors to hedge against climate risk without losing on returns by decarbonising the portfolio. Portfolio decarbonisation means reducing a portfolio’s exposure to carbon-intensive firms to build a hedge against risks related to climate change (De Jong & Nguyen, 2016). This leads to the main research question of the thesis i.e. “How well does a low-carbon index perform compared to its benchmark Stoxx 600?”. This paper is focused on European Markets. To create a low carbon index the paper uses Tracking error optimization technique using data of returns for 590 firms that constitute the Stoxx 600 index and carbon emission data of the 590 firms.

The structure of the paper is divided further into four sections. The first two sections discuss the underlying literature that this thesis is inspired by, the theoretical & methodological aspects of the research. The 3rd_{section discusses} the results of the research and the last section is for the conclusion based on the results obtained earlier and future avenue of analysis required in this topic.

Literature Review:

This section provides in detail the theoretical framework using a literature review for the empirical analysis conducted further for the thesis. To build the framework for the empirical research this section will try to examine if the financial market prices the risk of climate change correctly or not and discuss in-depth on literature that provides models on how to hedge against such mispricing by the financial markets.

Climate change is a real issue that threatens human wellbeing which in turn will cause a large number of economic losses. This sentiment is best echoed by the recent developments in policy efforts to curb CO2 emissions which raises the question of whether carbon emissions represent a material risk today for investors. 21st Conference of the Parties 2015 (COP21), which resulted in one of the most far-reaching policy initiatives is “Paris Agreement”, signed by 195 nations, to limit global warming to below 2°C (United Nations, 2015). This policy signals that the world has a consensus over the reality of climate change risk and is planning to transition to a low carbon economy. But the advent of populism

(5)

5

and economic crisis has a distortionary effect on this transition which induces uncertainty for investor regarding such risks. The risk from climate change is unique as in both the size and unpredictability of the climate change, the issue is also unprecedented as far as to the scale of the issue (i.e., worldwide) just as the timeline of the issue (i.e., extending over centuries) (Breakwell, 2010). Further, climate change is also a slow, cumulative and largely invisible process, it cannot be perceived and or experienced directly thus it is significantly different from the way that our ancestors have traditionally perceived threats in their local environment (Helgeson & Van Der Linden, 2012; Weber, 2010). This makes climate change risk a “novel” risk evolutionarily (Griskevicius et al., 2012). Thus, it is important to describe the climate or carbon risk in some detail. Carbon risk/ climate change risk is a new kind of risk includes all positive and negative impacts on firm values that arise from uncertainty in the transition process from a brown to a green economy, suggesting that all firms including brown firms and green firms are exposed to carbon risk (Görgen et al., 2019). Carbon risks include:

➢ Physical risks due to climate change such as an increase in water level which reduces the global landmass volume and can cause large scale flooding and other natural disasters which may affect the prices of physical assets such as land etc.

➢ Transitional risk due to the transition from brown economy (i.e. fossil fuel-intensive) to the green economy (i.e. clean energy-based or carbon-free) such as a change in consumer preferences.

➢ Carbon Pricing which may lead to taxes, cap & trade system to manage emission.

➢ Risk due to regulatory changes which may affect plants and manufacturing firm’s ability to emit carbon gas thus limiting carbon-intensive firms’ efficiency.

➢ Litigation risk against high-carbon emitters and investors.

➢ The “Carbon Bubble”: Potentially overvalued portfolio holdings due to stranded assets and mispricing or no pricing of carbon risks by investors. ➢ Technological risk due to innovation disruption by green energy usage and

needs.

Thus, the risk from an investors point of view is twofold, one from the climate change itself and the other from climate change mitigation policy. Thus, making the risk quantification for long term investors an ardent task. Institutional investors are hence more focused on environmental, social, and governance

(6)

6

(ESG) aspects of firm conduct, and are increasingly tracking the greenhouse gas emissions of listed companies. By some recent estimates, the total global assets under management of funds with some ESG tilt represented $30.7 trillion in 2018 (Bolton & Kacperczyk, 2019). This tilt may be influenced by the growing research field of climate finance. The concept of “climate finance” refers to the study of financial flows expected to contribute to the reduction of emissions and the adaptation to current climate variability as well as future climate change, encompassing private & public funds, domestic and international flows and covers three distinct objectives: mitigation, adaptation, coverage of losses and damage. (Boissinot et al., 2016). This research is mainly focused on the objective of covering loss and damage. There is growing evidence in recent asset pricing models highlighting the importance of climate risks as a long-run risk factor and the importance of carbon risks and environmental pollution in the cross-section of stock returns (Bansal et al., 2016; Bolton & Kacperczyk, 2019). There is evidence of carbon risk priced in the options markets and especially risk due to political uncertainty, as the cost of option protection against downside tail is larger for more carbon-intense firms and cost of downward option protection is magnified when public attention to climate change spikes (Sautner, 2019). Another main issue is the mispricing of the carbon risk by the market due to the unique nature of carbon risk. As Hong, Li, and Xu (2019) concludes that the exposure of food stocks to drought risks are incorrectly valued by markets and Kumar, Xin, and Zhang (2019) provides evidence that firms’ exposures to climate risks predict returns, which implies that stock markets misprice climate risks. There is also mispricing of carbon risk due to the investors habit to look at future cash flow projections through local thinking as described by Gennaioli and Shleifer (2010), ignoring unrepresentative information about global warming and its attendant risks. Cash-flow models widely used by financial analysts preclude any clear reference to carbon emissions and the potential of repricing. The risk of mispricing also increases due to the asymmetric nature of these risks on countries and industries. Thus due to the complex nature of carbon risk and its mispricing by the market, investor should find a compelling evidence to shift their strategy towards long term capital preservation by accounting for carbon risk considerations in their investment process that can create a win-win situation in which invested capital is protected and incentives are created for industries to reduce carbon emissions.

(7)

7

Figure 1: Growth in Sustainable and Ethical Investing strategies Worldwide from 2016-18 (Source: Global Sustainable Investing Alliance)

As discussed in the above paragraphs about the detail of carbon (climate change) risk and resulting in different methods in attempts to hedge against such risk. The focus of this thesis is one such way to hedge against carbon risk. The primary model is inspired by the paper of Andersson et. al (2016) which describes a simple dynamic investment strategy that allows long‐term passive investors to hedge climate risk without sacrificing financial returns or buying expensive options to protect against the downside risk. According to the authors, this investment strategy is attractive as it negates the timing issue related to carbon risk, especially to long term investors. This strategy attempts to create a green index fund. Green indexes are of two types:

➢ Pure play index focuses on renewable energy stocks, clean technology, and environmental service stocks. Essentially a bet on clean energy and technology.

➢ “Decarbonized” index takes a standard benchmark index (Stoxx 600) and removes the component stocks with a high carbon footprint.

According to Andersson et al. (2016), Pure play indexes have underperformed the benchmark since the onset of the financial crisis of 2008. But also, pure-play indexes are not a hedge against carbon risk but a bet on clean energy and green technology. Thus, leaving the decarbonized index a relatively cheap and efficient hedge against carbon risk. To hedge against the carbon risk

(8)

8

the authors suggest few steps. The first is to divest from high carbon-intensive stocks from the underlying portfolio i.e. create a low carbon-intensive index. The second step is to optimize the composition of the low carbon index by weighing the constituent stocks in the index to minimize tracking error. It is a strategy best suited for long term investor who prefers to buy and hold in simple terms. Tracking error minimization is a commonly used technique by fund managers as a part of a long-short strategy. Tracking error measures, the difference from a pre-selected benchmark long portfolio or short portfolio and tracking error minimization yield the hedging positions(Coleman et al., 2006). To construct a low tracking error (TE) green index Andersson et. al. describes its main idea in the following steps:

➢ Constructing a portfolio with fewer composite stocks relative to the benchmark but with equivalent aggregate risk to all priced risk factor. ➢ The only difference in aggregate risk between the two indexes is only w.r.t

carbon risk factor.

➢ Thus, if and when the carbon risk starts to get priced appropriately the green index will start to outperform the benchmark significantly as the current consensus of analyst forecast assumes by default that there is no carbon or extremely low probability of carbon risk.

The optimization of the constructed portfolio using the steps above will be solved using two methods. The first method, as termed by Andersson et. al. is the “transparent” rule optimization where they eliminate high carbon footprint stocks from the benchmark intending to meet with the target carbon footprint reduction and then reweight the remaining stocks to minimize TE with the benchmark. The second method is called “Pure” optimization where the author imposes another constraint on maximum allowable carbon footprint relative to Benchmark and subject to this constraint, reweight stocks in the benchmark index to maximize the carbon footprint reduction and minimize the tracking error of the new green index. The pure optimization method dominates the transparent rule method with the same amount of carbon footprint due to fewer constraints used when optimizing. Andersson et al. also suggest that using these strategies can generate a positive externality by signalling clearly about which constituent stocks are in the green index, not only rewards the companies included for their effort and investment in reducing the firm’s carbon footprint but also discipline the firm’s that are excluded. It also inspires debate on whether CHG emission is properly

(9)

9

measured and lead to improvements in the methodology for determining the firm-specific carbon footprint. In this signalling aspect, the “transparent” rule method performs better as “pure” optimization has a significant drawback in term of the opacity of methodology and lack of clean signal by which companies are excluded from the index. The practice of such strategies in the financial market also seems to reduce the carbon footprint of the portfolio holding and match or outperform the benchmark (see fig.2 & 3) The authors cite the example of AP4 the fourth Swedish National Pension Fund which was the pioneer funds to deploy low TE green indexes to hedge against carbon risk. They achieved a 50% reduction in their carbon footprint with .50 % tracking error. They also find more evidence in favour of the transparent rule method as it leads to less TE due to concentration of most TE are in material and energy sector contributing in a higher TE for “pure” method.

Figure 2: Performance of the Low Carbon 100 Europe index vs Stoxx Europe 600 for 5 year period. (Source: THEAM, performance over five years at end-May 2015)

(10)

10

Figure 3: Performance MSCI Low Carbon Leaders vs MSCI Europe (Source: MSCI, Amundi)

Research Methodology:

The purpose of this section is to develop a framework to implement the research question and understand the data used and challenges faced to create the dataset used in the analysis of the research question.

The first step towards building a decarbonized portfolio is to find a benchmark index and calculate the optimal weights to set benchmark weights for the constituent underlying stocks. The benchmark used in this thesis is focused on European markets, hence Stoxx 600 is used for this purpose. To calculate the initial weights of the benchmark index underlying constituent stocks we use the minimization of tracking error optimization method. The optimization problem is solved using MATLAB R2019b software. We call this optimization problem the “benchmarking”. The “benchmarking” problem is designed in a similar way as the other optimization problem discussed in the earlier section. The formal representation of the “benchmarking” optimization is as follows:

(11)

11

Subject to: ∑ Wi = 1.00

1 ≥ Wi ≥ 0

Where Rg _{= price return matrix of the individual stocks of the Stoxx 600} constituents for a period. The row dimension represents the time (weekly periodicity) and the column dimension represents the return of individual firms in the portfolio.

Rb _{= price return matrix of the benchmark Stoxx 600 index for the same period as} above. The row dimension represents the time (weekly periodicity) and the column dimension represents the return of individual firms in the portfolio. i = firm index number

Wi = a row vector of the weights of each firm stocks for the green index construction.

The first constraint describes the weights should sum up to 1 as we assume that the portfolio utilizes all the budget or can be understood as a budget constraint. The second constraint implies that the stocks cannot be shorted and cannot overinvest in one stock more than the budget. Implementing this optimization setup yields the benchmark weight given to the constituent stocks that will be used to create a decarbonized index further.

The second optimization problem is designed to decarbonize the portfolio using the “pure” optimization method that we discussed in the earlier section. This optimization also follows the minimization of tracking error heuristic model. The “Pure” optimization problem is stated formally as follows:

Minimize (TE) = standard deviation (Rg * Wi – Rb)

Wi

≥

0 for all i

∑ Wi * Ei ≤ CO2 Budget * x

Where Rg = price return matrix of the individual stocks of the Stoxx 600 constituents for a period. The row dimension represents the time (weekly periodicity) and the column dimension represents the return of individual firms in the portfolio.

(12)

12

Rb _{= price return matrix of the benchmark Stoxx 600 index for the same period as} above. The row dimension represents the time (weekly periodicity) and the column dimension represents the return of individual firms in the portfolio. i = firm index number

Wi= a row vector of optimal weights for constituent stocks given the constraint. Ei = a row matrix of total carbon emission per stocks in the portfolio.

CO2 Budget is calculated using the initial weights that we found in the benchmarking problem and multiply each weight with its total emission.

x represents the desired per cent reduction in the CO2 budget from the benchmark value.

The first constraint describes the weights should sum up to 1 as we assume that the portfolio utilizes all the budget or can be understood as a budget constraint. The second constraint implies that the stocks cannot be removed or divested from the green portfolio that we desire. The last constraint can we understood as the aggregate required carbon intensity by the investor.

The last optimization problem we design is the “transparent” rule method where we allow for divestment to take place for high emitting stocks from the index to attain the desired level of decarbonization of the green portfolio we desire. First, we need to sort the stocks in decreasing order of their total emission and give them rank, where index 1 is for the highest total emitting firm and index N, is for the lowest emitter. The “transparent” rule method is formally described as follow:

Minimize (TE) = standard deviation (Rg_{* Wi – R}b₎

Wj = 0 for all j = 1, 2, …., k

Wi ≥ 0 for i = k+1, k+2, ……., N

Where Rg = price return matrix of the individual stocks of the Stoxx 600 constituents for a period. The row dimension represents the time (weekly periodicity) and the column dimension represents the return of individual firms in the portfolio.

Rb = price return matrix of the benchmark Stoxx 600 index for the same period as above. The row dimension represents the time (weekly periodicity) and the column dimension represents the return of individual firms in the portfolio.

(13)

13

i = firm index number

Wi= a row vector of optimal weights for constituent stocks given the constraint. The first constraint describes the weights should sum up to 1 as we assume that the portfolio utilizes all the budget or can be understood as a budget constraint. The second constraint implies that the weights of kth most polluting stock should be 0 or is divested from. The third constraint implies that the remaining stocks in the portfolio should be optimized and cannot be shorted.

All the solution uses MATLAB and its “fmincon” optimization option to solves all three problems detailed above. Further code for all MATLAB computation used in this research can be found in the Appendix.

After understanding the research problem design and implementation it is necessary to discuss in detail the dataset used for generating the results of this research paper. The main dataset that we will use is the price returns for individual stocks in the Stoxx 600 index and the index itself for the same period obtained from FactSet Database. The Stoxx 600 index used in this research comprises of 590 individual stocks as 10 firms needed to be omitted as it had no data in the Asset4 ESG Database used for carbon emission data, this index will provide the basis of the construction of the decarbonized index. To make the Rg_{matrix for} the optimization problems detailed above we store the price return of individual stocks of the Stoxx 600 for a specific period. The Rb_{matrix is constructed using} the price returns of the Stoxx 600 index for the same period as the Rg matrix. The challenge faced in constructing these matrices was the dynamic nature of constituent stocks in the index. The index comprises of certain firms that may have been introduced in the index at different timelines thus causing missing cells in the matrix of Rb. Thus, there is a trade-off to be made in how to select the period of the observations. The longer the period the larger the number of missing returns from the matrix, thus higher number of individual stocks needs to be omitted to run the optimization that satisfies the budget constraint. Thus, the research uses 2 different period datasets of Stoxx 600 constituent stocks and the index itself to test the effect of the trade-off on the results. The 2 different datasets are of 3 year period from 31st_{December 2015 to 27}th_{December 2019 with a} weekly price return frequency and finally 5 years from 2nd_{January 2015 to 27}th December 2019 with a weekly price return frequency. Then we remove 34 firms from the 3-year dataset and 48 firms from the 5-year dataset to omit all the columns containing missing data. All the return data are based on Euro currency and it excludes dividend given in that period.

(14)

14

Another important dataset for this research is the firm-level carbon emission data for the European market. The dataset is obtained from Thomas Reuters Asset4 ESG database. Greenhouse gas (GHG) emissions are classified as per the Greenhouse Gas Protocol and are grouped in categories called Scope 1, Scope 2, and Scope 3. They are defined as follows:

➢ Scope 1 GHG emissions are those directly occurring "from sources that are owned or controlled by the institution, including on‐campus stationary combustion of fossil fuels; mobile combustion of fossil fuels by institution-owned/controlled vehicles; and "fugitive" emissions."

➢ Scope 2 emissions are "indirect emissions generated in the production of electricity consumed by the institution."

➢ Scope 3 emissions are all the other indirect emissions that are "a consequence of the activities of the institution, but occur from sources not owned or controlled by the institution" such as commuting; embodied emissions from extraction, production, and transportation of purchased goods; outsourced activities; contractor‐owned vehicles; and line loss from electricity transmission and distribution". In the data, Scope 3 emissions are conceptually divided into (a) upstream emissions, i.e. emissions stemming from a company’s supply chain and (b) downstream emissions, i.e. emissions from product “use phases” during their life cycle. (ESG —

YourSRI - Socially Responsible Investments, n.d.).

For this research, we use CO2 Equivalents Emission Total data for the firms in the STOXX 600 index from the year 2010 to 2020. This data is used to compile the emission per stock matrix and the total desired CO2 emission of the green portfolio index. The CO2 Equivalents Emission Total uses scope 1 and scope 2 emission data. The unit used is tonnes CO2 equivalent.

Results:

In this section, the results of the optimization problems as defined in the earlier section are described for all the different period datasets. The results will be discussed descriptively with the use of tables and graphs.

(15)

15

Parameters

3 year

5 year

Benchmark Return 0.0809 0.0951

Total Return Green Portfolio

0.1704 0.1808

Tracking Error 0.0955 0.1241

Information Ratio .94 .70

Firms excluded at the start

34 48

Number of Highest Emitting Firms Omitted

111 (Top 20 percentile) 108 (Top 20 percentile)

Initial Total CO2 emission of the portfolio

60019209.34 tonnes 65476551.11 tonnes

Reduced Total CO2 emission of the portfolio

2162397.44 tonnes 2238878.46 tonnes

Table 1: Financial and Carbon Efficiency Performance of “Transparent” method optimized Green Portfolio.

Table 1 presents the results from the “transparent” rule method optimization. The “transparent” rule applied is to remove the top 20 percentile highest emitter firms from the benchmark index to reduce the carbon footprint of the desired green portfolio. For the 3-year dataset, the green portfolio return is 17.04% but with a 9.55% excess volatility compared to the Stoxx 600 in that period. The information ratio for generated by this green portfolio is 0.94. For the 5-year dataset, the green portfolio return is 18.08% but with excess volatility of 12.41% compared to the Stoxx 600. The 5-year dataset green portfolio information ratio is .70. There is a positive relationship between the Tracking error of the green portfolio and the firms removed from the index (see fig. 4&5). The “transparent” rule outperforms the benchmark for both periods namely for 3 years and 5 years dataset. The carbon footprint is eliminated by 96.4% for the 3-year dataset and 96.6 % for the other dataset by eliminating 20 percentile highest emitting firms.

(16)

16

Figure 4: Graph plot showing the evolution of Tracking Error vs Number of High emitting firms removed of the green index constructed using the “Transparent” method for 3-year Data.

Figure 5: Graph plot showing the evolution of Tracking Error vs Number of High emitting firms removed of the green index constructed using the “Transparent” method for 5-year Data.

(17)

17

“Pure” optimization method also reduces the carbon footprint of the green portfolio but produces a lower tracking error compared to the “transparent” rule method (See Table 2). The desired reduction level in the carbon footprint is the same level as the “transparent” method results discussed in the earlier paragraph (i.e. top 20 percentile firm omission level reduction) . The decarbonized portfolio using the 3-year dataset returns 16.15% with excess volatility of 5.7% compared to the Stoxx 600 in the same period. The information ratio generated by this decarbonized portfolio is 1.56. For the 5-year dataset, the decarbonized portfolio gives a return of 16.80% but with excess volatility of 7.62% compared to the benchmark in the same period. For this portfolio, the information ratio is .96. There is also a positive (direct) relationship between tracking error and the per cent reduction in CO2, as we increase the reduction percentage parameter the tracking error also increases (see fig.6&7).

Parameters

3 year

5 year

Benchmark Return 0.0809 0.0951

Total Return Green Portfolio

0.1615 0.1680

Tracking Error 0.0517 0.0762

Information Ratio 1.56 .96

Firms excluded at the start 34 48

Per cent reduction in CO2

emission

96.4% 96.6%

Table 2: Financial and Carbon Efficiency Performance of “Pure” optimization method Green Portfolio.

(18)

18

Figure 6: Graph plot illustrating Carbon Frontier of “Pure” method Green portfolio for 3- year Data.

Figure 7: Graph plot illustrating Carbon Frontier of “Pure” method Green portfolio for 5- year Data.

(19)

19

Conclusion:

In this final section, we will interpret the results using the theoretical framework developed in the earlier section and answer the main research question based on the interpretations. Lastly, we discuss the limitations of the results and future research opportunities related to this research thesis.

The “decarbonized” green index created using the tracking error minimization methods indeed performs similarly or above the benchmark index with a low amount of tracking error. This outperformance is evident from the positive information ratio generated by the green portfolio in both the dataset. In both period dataset, the green portfolio constructed using both the “pure” and “transparent rule” methods outperform the benchmark by approximately 8% to 9%. The “pure” optimization method portfolio gives significantly lower tracking error than the “transparent” method optimization for the same level of CO2 reduction as for both dataset the tracking error is almost 4 to 5 percentage point higher for the “transparent” method, but the returns are slightly larger than the “pure” method by at least 1% for both dataset. The significantly high returns for the constructed green portfolio is due to the carbon emission screen that may affect the portfolio diversification negatively because it limited the investment universe compared to the Stoxx 600 thus increasing the volatility of the green portfolio (Renneboog et al., n.d.). But this higher return conundrum is not in the spirit of Andersson et al. (2016), as they assume that the carbon is mispriced in the market for their hedging strategy to outperform the market due to the hedge acting as a “free” option on carbon risk. If we assume this mispricing then once the market prices the carbon risk efficiently the risk of the benchmark should increase to justify the higher return of the green portfolio. To elaborate more on this assumption we will use the risk model of Hoepner et al. (2010) literature. They put forwards three main drivers of portfolio diversification namely i) the number of stocks, ii) weighted average correlation of stocks & (iii weighted average specific risk of stocks. The carbon screen may have negatively affected the portfolio diversification of the green index through the first two drivers. But as we discussed in an earlier section about the carbon risk and its resulting cause of market inefficiencies can be attributed as a specific risk of the stocks in the index. Thus, portfolio diversification is improved as decarbonization reduces the specific risk of the green portfolio whereas this specific risk will negatively affect the benchmark index when such risk is realised. Hence, there is enough evidence

(20)

20

to support that the decarbonised green portfolio created using the tracking error minimization methods performs similarly or better than the Stoxx 600 index. Similarly to Andersson et al. (2016) finding, we find the “pure” optimization method a better strategy in terms of performance as it generates significantly higher information ratio, as for the same amount of carbon emission reduction “Pure” method gives a lower tracking error whereas the “transparent” rule may be a superior method for policy purposes due to its positive externalities generated by making firms compete to improve their carbon footprint.

The one major limitation of the results is that the returns are significantly higher as the diversification of the portfolio is not optimal due to the negative effect of carbon screen as the emission are reduced or removed without the sector-specific consideration thus creating a biased portfolio composition. This effect is more pernicious in a “transparent” rule method where we may have removed a large number of energy and mineral related firms and large capitalized firms which are generally carbon-intense thus causing unbalanced diversification. Another, limitation of the results is due to the omission of firms that have missing data for the entire period which may also unbalance the diversification of the green portfolio. Thus, for future research opportunities, it would be interesting the check if the sector-specific filtering of firms reduces the tracking error of such green portfolios. It would also be interesting to apply the framework developed in this paper to focus on different markets other than Europe and a cumulative global market.

(21)

21

Appendix:

Benchmarking 3 year: clear;

benchmark = xlsread('3year','Sheet3'); portfolio = xlsread('3yearedit','Sheet4');

%tracking error minimization function

fun = @(w) nanstd(((portfolio * (w).') - benchmark));

%optimization constraints setup:

w0 = 0*ones(1,556)*(1/556); %initial point

A = []; b = [];

Aeq = ones(1,556); % A*w = beq i.e. sum of all weights = 1

beq = 1.00; nonlcon = []; options =

optimoptions('fmincon','Algorithm','sqp','MaxFunctionEvaluations',Inf,'MaxIterations',Inf);

%needed to increase my default iteration as the fmincon was stopping prematurely error was given till 60000 value

lb = (zeros(1,556)); ub = (ones(1,556));

[weights,risk]= fmincon(fun,w0,A,b,Aeq,beq,lb,ub,nonlcon,options);

Benchmarking 5 year: clear

portfolio = xlsread('5yr','Sheet1'); benchmark = xlsread('5yr','Sheet5');

w0 = ones(1,542)*(1/542); %initial point

A = []; b = [];

beq = 1.00; nonlcon = [];

options = optimoptions('fmincon', 'MaxFunctionEvaluations',650000); %needed to increase my default iteration as the fmincon was stopping prematurely error was given till 60000 value

lb = (zeros(1,542)); ub = (ones(1,542));

(22)

22

“Pure” optimization 3 year:

clear;

benchmark = xlsread('3year','Sheet3'); portfolio = xlsread('3yearedit','Sheet4'); perstockemit = xlsread('3yearedit','Sheet2'); initialbud = 60019209.34;

A = perstockemit; %(w) * emission per stock

b = (initialbud)* .50 ; %pure optimization constraint: w*emission<= budget

Aeq = ones(1,556);

beq = 1.00; % A*w = beq i.e. sum of all weights %=1

lb = (zeros(1,556)); ub = (ones(1,556)); nonlcon = [];

options = optimoptions('fmincon','Algorithm','sqp',

'MaxFunctionEvaluations',Inf,'MaxIterations',Inf); % needed to increase my default iteration %as the fmincon was stopping prematurely error was given till 60000 value

“Pure” optimization 5 year:

clear

portfolio = xlsread('5yr','Sheet1'); benchmark = xlsread('5yr','Sheet5'); perstockemit = xlsread('5yr','Sheet2'); initialbud = 65476551.11;

A = perstockemit; %(w) * emission per stock

b = (initialbud)* .50 ; %pure optimization constraint: w*emission<= budget

beq = 1.00; nonlcon = [];

options = optimoptions('fmincon', 'MaxFunctionEvaluations',650000); %needed to increase my default iteration as the fmincon was stopping prematurely error was given till 60000 value

(23)

23 ub = (ones(1,542));

“Transparent” rule 3 year:

clear;

benchmark = xlsread('3year','Sheet3'); portfolio = xlsread('3yearedit','Sheet6');

A = []; b = [];

Aeq = [ones(1,556);eye(111,556)]; % A*w = beq i.e. sum of all weights = 1 and w = %0 for highest 111 emitter

beq = [1;zeros(111,1)]; nonlcon = [];

options = optimoptions('fmincon','MaxFunctionEvaluations',650000); %needed to increase my default iteration as the fmincon was stopping prematurely error was given till 60000 value

lb = zeros(1,556) ; ub = (ones(1,556));

“Transparent” rule 5 year:

clear

portfolio = xlsread('5yr','Sheet4'); benchmark = xlsread('5yr','Sheet5');

A = []; b = [];

Aeq = [ones(1,542);eye(108,542)]; % A*w = beq i.e. sum %of all weights = 1 and w %= 0 for highest 108 emitter

beq = [1;zeros(108,1)]; nonlcon = [];

options = optimoptions('fmincon', 'MaxFunctionEvaluations',650000); %needed to increase %my default iteration as the fmincon was stopping prematurely error was given till 60000 %value

lb = (zeros(1,542)); ub = (ones(1,542));

(24)

24

Reference List:

Andersson, M., Bolton, P., & Samama, F. (2016). Hedging climate risk. Financial

Analysts Journal, 72(3), 13–32. https://doi.org/10.2469/faj.v72.n3.4

Arbuthnott, K. D., & Dolter, B. (2013). Escalation of commitment to fossil fuels.

In Ecological Economics (Vol. 89, pp. 7–13). Elsevier.

https://doi.org/10.1016/j.ecolecon.2013.02.004

Bansal, R., Ochoa, M., & Kiku, D. (2016). Climate Change and Growth Risks. Barclay, E., & Resnick, B. (2019). Youth climate strike crowd estimate: More

than 4 million people took part, activists say - Vox. Vox.

https://www.vox.com/energy-and- environment/2019/9/20/20876143/climate-strike-2019-september-20-crowd-estimate

Boissinot, J., Lame, G., & Huber, D. (2016). Finance and climate: The transition to a low-carbon and climate-resilient economy from a financial sector perspective. OECD Journal: Financial Market Trends, 2015(1), 7–23. https://doi.org/10.1787/fmt-2015-5jrrz76d5td5

Bolton, P., & Kacperczyk, M. T. (2019). Do Investors Care about Carbon Risk?

SSRN Electronic Journal. https://doi.org/10.2139/ssrn.3398441

Breakwell, G. M. (2010). Models of risk construction: some applications to climate change. Wiley Interdisciplinary Reviews: Climate Change, 1(6), 857–870. https://doi.org/10.1002/wcc.74

Coleman, T., Li, Y., & Henniger, J. (2006). Minimizing tracking error while restricting the number of assets. The Journal of Risk, 8(4), 33–55. https://doi.org/10.21314/jor.2006.134

De Jong, M., & Nguyen, A. (2016). Weathered for climate risk: A bond investment proposition. Financial Analysts Journal, 72(3), 34–39. https://doi.org/10.2469/faj.v72.n3.2

ESG — yourSRI - Socially Responsible Investments. (n.d.). Retrieved June 10,

2020, from https://yoursri.com/esg-and-carbon/introduction/carbon

Görgen, M., Jacob, A., Nerlinger, M., Riordan, R., & Rohleder, M. (2019).

(25)

25

arising from the transition process to a low-carbon. 1–65.

Griskevicius, V., Cantú, S. M., & van Vugt, M. (2012). The Evolutionary Bases for Sustainable Behavior: Implications for Marketing, Policy, and Social Entrepreneurship. Journal of Public Policy & Marketing, 31(1), 115–128. https://doi.org/10.1509/jppm.11.040

Helgeson, J. F., & Van Der Linden, S. (2012). The role of knowledge, learning

and mental models in public perceptions of climate change-related risks.

https://doi.org/10.3920/978-90-8686-757-8_21

Hoepner, A. G. F. (2010). Corporate Social Responsibility and Investment

Portfolio Diversification. Ssrn, May, 1–16.

https://doi.org/10.2139/ssrn.1599334

Kim, S. Y., & Wolinsky-Nahmias, Y. (2014). Cross-National Public Opinion on Climate Change: The Effects of Affluence and Vulnerability. Global

Environmental Politics, 14(1), 79–106.

https://doi.org/10.1162/GLEP_a_00215

Renneboog, L., Horst, J. Ter, & Zhang, C. (n.d.). Socially responsible

investments: Institutional aspects, performance, and investor behaviour.

https://doi.org/10.1016/j.jbankfin.2007.12.039

Ritchie, J., & Dowlatabadi, H. (2014). Understanding the shadow impacts of investment and divestment decisions: Adapting economic input-output models to calculate biophysical factors of financial returns. Ecological

Economics, 106, 132–140. https://doi.org/10.1016/j.ecolecon.2014.07.005

Ritchie, Justin, & Dowlatabadi, H. (2015). Divest from the Carbon Bubble?

Reviewing the Implications and Limitations of Fossil Fuel Divestment for Institutional Investors 1. http://blogs.ubc.ca/dowlatabadi/welcome/

Sautner, Z. (2019). Carbon Tail Risk. December.

Swim, J., Clayton, S., Doherty, T., Gifford, R., Howard, G., Reser, J., Stern, P., & Weber, E. (n.d.). Psychology and Global Climate Change: Addressing a

Multi-faceted Phenomenon and Set of Challenges A Report by the American Psychological Association’s Task Force on the Interface Between Psychology and Global Climate Change Members. Retrieved June 25, 2020,

from http://www.apa.org/science/about/publications/climate-change.aspx Taylor, M., Watts, J., & Bartlett, J. (2019). Climate crisis: 6 million people join

(26)

26

the latest wave of global protests | Environment | The Guardian. The

Guardian. https://www.theguardian.com/environment/2019/sep/27/climate-crisis-6-million-people-join-latest-wave-of-worldwide-protests

The Guardian. (2016). Fossil fuel divestment funds double to $5tn in a year |

Environment | The Guardian.

https://www.theguardian.com/environment/2016/dec/12/fossil-fuel-divestment-funds-double-5tn-in-a-year

United Nations (2015) Framework Convention on Climate Change

FCCC/CP/2015/L.9/ Rev.1 Adoption of the Paris Agreement.

https://unfccc.int/resource/docs/2015/cop21/ eng/l09r01.pdf.

Weber, E. U. (2010). What shapes the perceptions of climate change? In Wiley

Interdisciplinary Reviews: Climate Change (Vol. 1, Issue 3, pp. 332–342).

How Well does a Low-Carbon Index Performs Compared to its Benchmark Stoxx 600 Index?