PORTFOLIO METHODOLOGY

In the following section a thorough description of the portfolio methodology and its importance will be explained. The objective of portfolio analysis is to examine the cross-sectional relation between two variables and is one of the most commonly used statistical methodologies in asset pricing (Bali, Engle, & Murray, 2016). In this assignment, portfolio analysis will be employed to understand the cross-sectional relationship between ESG scores and stock returns.

5.7.1 Univariate portfolio analysis

Portfolio sorting has been an important part of modern empirical financial research and is widely used to test for, and establish, cross-sectional relationships between expected asset returns and asset class characteristics (Cattaneo, Crump, Farrell, & Schaumburg, 2016). Motivated by Blume (1970) the majority of empirical financial research in modern asset pricing today uses portfolios instead of individual stocks when analyzing the relationship between expected return and asset class characteristics.

Additionally, several studies have examined if the usage of portfolios compared to individual stocks enhances the analysis (Fama & Macbeth (1973), Fama & French (1992), Black, Jensen, &

Scholes (1972)). The studies found that by using portfolios, the standard errors of factor loadings are reduced significantly due to decreasing idiosyncratic risk. Intuitively, this makes sense and is in accordance with Markowitz’s (1959) Modern Portfolio Theory that explains that by investing in more than one stock an investor can achieve benefits from diversification and reduce the riskiness

Page 54 of 146

of the portfolio (McClure, 2020). There exist multiple ways of conducting portfolio sorting. In this thesis, we have chosen to use the univariate sorting methodology introduced by Banz (1981).

The objective of a univariate portfolio analysis is to assess the cross-sectional relation between X (Stock returns), the sort variable, and Y (ESG scores), the outcome variable. In this thesis, we will follow the four-step procedure put forth by Bali, Engle, & Murray (2016).

5.7.2 Number of portfolios

The first step is to group the stocks in the sample into portfolios based on values of the ESG score. We have pulled data from a combined 4,415 listed companies in Oceania, Asia, and Europe from 2007 to 2020. When we segment these stocks according to their ESG score, it results in 3,043 observations for Oceania, 12,841 observations for Asia and 7,233 observations for Europe with an ESG score for every year. In each year, the stocks have been divided into decile portfolios based on their ESG score for the respective year. The ten portfolios range from PF1 (lowest ESG score) to PF10 (highest ESG score), where PF1 consist of the 10% lowest ESG scores while PF10 consist of the 10% highest ESG scores. This sorting methodology is applied for all regions.

Choosing an appropriate number of portfolios is a trade-off, as fewer portfolios lead to less cross-sectional dispersion within the factor loadings. A decreased dispersion of the sort variable X across the portfolios can make it relatively more difficult to detect the cross-sectional relationship between ESG scores and stock returns, as the difference in ESG score across the portfolios are reduced (Bali, Engle, & Murray, 2016). On the other hand, it is assumed that a higher number of portfolios would result in more dispersed information in the cross-section. As the number of portfolios increases, the number of companies in each portfolio decreases which will result in increased noise when using the sample mean to calculate the true mean value for each portfolio.

Historically, well known empirical studies like Fama & Macbeth (1973) used 20 portfolios, Fama

& French (1992) used 25 portfolios and Black, Jensen, & Scholes (1972) used 10 portfolios. In our study within the field of ESG scores, it is assumed that ten portfolios sorted on a univariate variable (ESG score), is suitable and will provide results that are easy to interpret.

5.7.3 Portfolio Analysis

In this section we will describe different ways to weight each stock in a portfolio and which method we have applied on our ten decile ESG portfolios. The most common methods to weigh stocks in a portfolio or index is by equally weighting (EW) each stock or value weight (VW) each

Page 55 of 146

stock based on their market capitalization. Even though two portfolios/indices consist of the same companies, they can behave very differently and can affect investments substantially (Hayes, 2020).

In an EW portfolio/index, the investor places an equal bet on every company’s success, which is a passive decision. On the contrary, a VW portfolio/index based on market capitalization has a higher concentration of larger companies and assumes that yesterday’s winners will continue to win. Figure 9 below illustrates how the S&P 500 Equally Weighted Index (S&P 500 EWI) and S&P 500 Market Weight Index (S&P 500 MWI) has performed since May 2009. The table illustrates that the indices perform almost identically over an 11-year period from 2009-2020.

However, it is worth mentioning that the S&P 500 EWI is a bit more volatile than the S&P 500 MWI, which might be caused by the greater volatility among small-cap stocks and their larger weights compared to S&P 500 MWI.

Figure 9: Weekly return for S&P 500 equally weighted index and S&P 500 market-weighted-index Figure 9 illustrates the weekly return for S&P 500 EWI and S&P 500 MWI. The illustration shows that S&P 500

EWI is indeed more volatile, but experienced are higher average return than S&P 500 MWI.

The fact that small-cap stocks are more volatile is not necessarily negative, because over a long-time horizon small-cap stocks has performed a better risk-adjusted return than large-cap stocks, a view that is supported by Fama & French (1992).

Page 56 of 146

Based on the descriptions above, we have chosen to VW the companies based on their market capitalization for the previous year. This means that the stocks in portfolio one for 2007 is weighted based on their respective market capitalization in 2006. If a company does not have a market capitalization in the previous year the weight will be based on the company’s market capitalization for the same year. An analysis of whether this method results in the best risk adjusted ESG portfolio return or not, is outside the scope of this master thesis and will not be examined.