Portfolio construction

Subsection 5.1.3 briefly introduced how the portfolios examined in this study are con-structed as the reader needed the information to understand the descriptive statistics presented in the subsection. However, portfolio construction is discussed more pro-foundly in this subsection.

The empirical analysis of this study focuses on four different portfolios, although those portfolios’ performance is also examined in different time periods. As briefly discussed in subsection 5.1.3, the ESG momentum tries to generate alpha by overweighting the previous year’s best ESG improvers and underweighting the worst ESG improvers. One of the study’s restrictions was that in order for a company to be selected in a portfolio for a particular year, for example 2005, ESG data needed to be available for the company at the end of 2003 and 2004 to calculate the ESG growth rate. Also, the financial data needed to be available at the end of 2004 and at the end of 2005 to calculate the com-pany’s financial performance for the given fiscal year. This restricted the investment uni-verse slightly as the data was not always available, yet it was necessary to complete the empirical analysis.

Furthermore, the whole S&P500 index was divided into two separate subsamples, “Sub-sample 1” and “Sub“Sub-sample 2”. The S&P500 index was divided into half based on the size of the companies, as this study tries to examine whether the size factor matters when implementing the ESG momentum strategy. Focusing on examining whether the size fac-tor matters is based on three scientifically proven arguments: firstly, O’Rourke (2003) argues that usually, larger companies have more resources to use in CSR activities. Sec-ondly, Lins et al. (2017) state, that companies with higher ESG scores performed better

during the financial crisis than companies with low ESG scores. Thirdly, small-cap com-panies are usually more hard-hit in economic turmoil (CME Group, 2020).

Portfolio median market capitalization was used as a dividing point when constructing the two subsamples. Ideally, the examination of whether the size factors matter would have been conducted comparing “large-cap” and “small-cap” companies, but for exam-ple, the ESG data for the Russell 2000 index only available from 2017 onwards. This com-parison would have narrowed the sample period significantly and made the results un-reliable. Although the comparison is conducted with four portfolios of S&P500 compa-nies, which are generally considered “large-cap” compacompa-nies, the difference between the portfolio average market capitalizations is significant, as shown in table 2. The mean for the average portfolio market capitalization in the sample period for the “Subsample 1, Top 10%” is 41260.44 million US dollars and 45256.80 million US dollars for the “Sub-sample 1, Top 25%”. In contrast, the market capitalizations for the “Sub“Sub-sample 2, Top 10%” and “Subsample 2, Top 25%” are 7533.89 and 7831.34 million US dollars, thus mak-ing the difference in market capitalizations significant.

Table 2. Descriptive statistics regarding market capitalizations for the four portfolios over the sample period 2005-2019.

Figure 7 below gives a visual presentation of the four portfolios’ average market capital-ization development in the sample period 2005-2019. Even though figure 7 might not indicate it, the “Subsample 2” market capitalizations have risen more percentage-wise than the “Subsample 1”-portfolios. However, as we can see, the actual dollar difference

between the portfolio average market capitalizations has risen drastically in the sample period. This is in line with the current market situation, as the larger companies are get-ting even larger daily, and thus the market share of the smaller companies is continu-ously diminishing.

Figure 7. Visual representation of the portfolios’ average market capitalization (in million USD) development over the sample period 2005-2019. The portfolio market capitalizations are based on the previous year’s ending market capitalization.

After dividing the investment universe into half by market capitalization and ranking the companies with the previous year’s ESG score growth rate, the four different portfolios were constructed by choosing the top 10% and top 25% ESG improvers of the subsam-ples. These portfolios are equally weighted. This methodology is similar to Bergskaug (2019); however, the methodology is slightly modified. Bergskaug constructed Top 10%, Bottom 10% and Long-Short -portfolios for each of his investment universes. This study focused only on the positive ESG momentum effect, as the short-portfolios’ preliminary results were extremely unreliable. The reason to include Top 25%-portfolios in addition to Top 10%-portfolios is to do a robustness check for the results and thus increase the reliability of the results.

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 Subsample 1, top 10% Subsample 1, top 25%

Subsample 2, top 10% Subsample 2, top 25%

Previous main studies concerning ESG momentum strategy, such as Nagy et al. (2013 &

2016) and Verheyden et al. (2016), use different methods, as those studies include the whole investment universe in the portfolio. After that, the best ESG improvers are over-weighted, and the worst ESG improvers are underweighted in the portfolio. However, these weighting methods are not further specified in the studies. The studies most prob-ably use their own financial institutions’ models, so a similar method could not be used in this study. After this, the financial performances of the portfolios are calculated, and those performances are next further discussed.

Table 3 below presents the annual returns for the “Subsample 1, Top 10%”-portfolio. The portfolio gained a cumulative return of 197.30% over the sample period 2005-2019. The table also presents the risk-free rate of returns over the sample period and the annual excess returns over the risk-free rates of return.

Table 3. “Subsample 1, Top 10%”-portfolio annual returns over the sample period 2005-2019.

Same portfolio performance statistics for the “Subsample 1, Top 25%”-portfolio are pre-sented in table 4 below. “Subsample 1, Top 25%” portfolio had a cumulative return of 192.01% in the sample period. The cumulative return is very close the cumulative return of the “Subsample 1, Top 10%”-portfolio, but it slightly underperformed compared to it.

Table 4. “Subsample 1, Top 25%” portfolio annual returns over the sample period 2005-2019.

The similar portfolio performance statistics for the “Subsample 2, Top 10%”-portfolio are presented in table 5 below. Over the sample period, “Subsample 2, Top 10%”-portfolio gained a cumulative return of 302.23%. This portfolio outperformed significantly both of the “Subsample 1”-portfolios, which was expected, as companies with smaller market capitalizations have historically outperformed the companies with larger market capital-izations.

Table 5. “Subsample 2, Top 10%” portfolio annual returns over the sample period 2005-2019.

For the last portfolio, “Subsample 2, Top 25%”, the financial performance is illustrated in table 6 below. It earned a cumulative return of 237.64% over the sample period and expectedly outperformed the “Subsample portfolios. As with the “Subsample 1”-portfolios, the Top 10%-portfolio outperformed the Top 25%-portfolio between the

“Subsample 2”-portfolios.

Table 6. “Subsample 2, Top 25%” portfolio annual returns over the sample period 2005-2019.

Lastly, table 7 below summarizes the portfolio performances over the sample period and presents the portfolios’ descriptive statistics. As discussed earlier, “Subsample 2”-port-folios that include companies within the S&P500 index with lower market capitalizations outperformed the “Subsample 1”-portfolios with the larger market capitalization panies. However, the returns for the “Subsample 2”-portfolios were more volatile com-pared to “Subsample 1”-portfolios, standard deviations being 15.41%, 16.95%, 17.22%

and 18.42% for the “Subsample 1, Top 10%” -, “Subsample 1, Top 25%” -, “Subsample 2, Top 10%”- and “Subsample 2, Top 25%”-portfolios, respectively. When analyzing the an-nual returns even further, the difference between variability can be mostly explained by

the time periods around negative return years. Companies with smaller market capitali-zations in the US are generally more affected by economic turmoil than companies with larger market capitalizations in the US. However, the companies with smaller market capitalizations outperform the companies with larger market capitalizations after the economic turmoil. Thus, smaller companies tend to be associated with higher risk. (CME Group, 2020)

Table 7. Descriptive statistics of the portfolio financial performances.

Finally, table 8 below illustrates the portfolio cumulative returns in different time periods.

The sample period 2005-2019 is here further divided into three sub-periods; Pre-crisis period 2005-2006, Crisis period 2007-2009 and After-crisis period 2010-2019. Here for simplicity purposes, the Top 10%-portfolios are compared together, and the Top 25%

portfolios are compared together. In every time period, the cumulative returns are higher for both the Top 10% and the Top 25%-portfolios constructed from Subsample 2.

This is again in line with the previous literature that companies with smaller market cap-italizations tend to outperform companies with higher market capcap-italizations.

Table 8. Cumulative returns of the portfolios in the different time periods.