The devastating global economic impact of the coronavirus disease 2019 (COVID-19) pandemic, worrying accounts of corporate negligence and relentlessly accumulating evidence for climate change and its possible impact have all contributed to a growing awareness of individual and collective responsibility. Modern investors are increasingly calling corporations to account, punishing miscreants by reducing holdings in their stock or shunning them completely. To make informed decisions about these tactics, investors – the bulk of whom cannot possibly investigate and assess each potentially investable stock – rely on environmental, social and governance (ESG) scores provided by dozens of ratings providers – five of whom (MSCI, Refinitiv, Sustainalytics, RepRisk and Institutional Shareholder Services [ISS]) currently (2022) dominate the market (Sikochi & Serafeim 2021). Using self-disclosed corporate information (such as board diversity (social score), carbon emissions (environmental score) and safety policies (governance score) and firm’s own, often concealed, analysis and algorithms, these providers formulate, synthesise and consolidate results into single ESG scores.

Sustainable investment increasingly relies on these ESG scores, but the investment strategies typically assemble investment portfolios using a ranked list of ‘good’ (high ESG score) stocks while ignoring portfolio completeness. Although the field has begun to flourish, literature that investigates sustainable investment asset allocation remains limited and no consolidated best-practice approach has been proposed.

Developing economies are particularly impacted by climate change and perceptions of poor governance. The literature is replete with research, which has repeatedly demonstrated that developing countries are suffering, and will suffer, considerably more than their developed country neighbours from the effects of climate change (Huang & Tian 2021; Kompas, Ha Pham & Nhu Che 2018) and the ongoing toxic effects of poor governance (Kłosowicz 2019; Yahyaoui & Bouchoucha 2021). At the November 2021 COP26 summit in Glasgow, for example, South Africa (the world’s 15th largest emitter of carbon dioxide [CO₂]) announced that it had secured commitments for US$8.5 billion in financing from 2022 to 2026 from France, the United Kingdom, Germany, the United States of America and the European Union to help install more clean energy and accelerate the country’s transition away from coal power. The funds will also be used to mitigate the impact on workers affected by this transition. South Africa relies overwhelmingly on coal, which supplies 87% of the nation’s electricity. This news was welcomed by investors, among others, wishing to capitalise on improved ESG scores and thus a wider universe of appropriate securities from responsible corporations.

Identifying and implementing an optimal investment solution in a mean-variance framework while heeding additional constraints (i.e. over and above ‘traditional’ constraints of simultaneously maximising excess returns and minimising associated risk) imposed by the new milieu such as selecting ‘optimal’ ESG scores for investable securities, remains. This article fills a gap in the literature by augmenting conventional portfolio selection models to embrace ESG constraints specifically for emerging markets where ESG scores are typically low compared with those observed in most developed markets. We aver that most investors still wish to generate risk-return performance obtained from conventional portfolio selection from portfolios constrained by ESG requirements and we find that for emerging economies, portfolio selection for sustainable investment and conventional portfolio selection are characterised by substantially different portfolio weights while achieving comparable levels of risk-adjusted returns. While this result has been demonstrated for developed markets, the results obtained from this research are important for sustainable investments because portfolio weights are the foundation of portfolio selection and investments. Using a sample of component stocks selected from South Africa’s Johannesburg Stock Exchange Index from 2008 to 2020, the ESG-constrained efficient frontier and the traditional efficient frontier were found to produce similar levels of risk-adjusted returns, a fortuitous endorsement of sustainable investment.

The remainder of this article is structured as follows. The next section provides an overview of the relevant literature and discusses the limitations and conclusions reached by the relevant authors. A discussion of the data used, and the methodology adopted for this research, is presented in the ‘Research design’ section, and the ‘Results and discussion’ section sets out the results obtained and considers possible outcomes derived from these results. The ‘Limitations and recommendations’ section discusses the limitations of the current work and presents some suggestions for future work, while the ‘Conclusion’ section summarises the results and observations obtained and concludes the article.

Hong and Kacperczyk (2009) found that ‘sin’ stocks generate higher average returns because they are rejected by ethical investors, and instead embraced by mutual and hedge funds (i.e. considerably larger investors). Investors are also willing to accept lower returns from socially responsible investments, a direct reflection of their investment motives which involve social signalling and preferences, rather than superior performance (Barnett & Salomon 2006; Barracchini 2012; Riedl & Smeets 2017). Lee et al. (2020) and Barber, Morse and Yasuda (2021) found that high ESG score venture capital funds generate returns which are 4.7% lower than funds with comparable geography and vintage. Investors were again found to be willing to accept these lower returns as compensation for social impact. Bolton and Kacperczyk (2020) found that companies which produce high CO₂ generate higher-than-average returns possibly because a carbon risk premium is included which compensates investors for the risks associated with future carbon pricing regulations (see also Bilbao-Terol, Arenas-Parra & Cañal 2012).

Gompers, Ishii and Metrick (2003) found that portfolios which incorporate long stocks with high governance scores (e.g. those exhibiting strong shareholder rights) and those with low governance scores (e.g. those with poor employee satisfaction ratings) outperform the market by 8.5% annually. Core, Guay and Rusticus (2003) further explored Gompers et al.’s (2003) work and found that weak governance firms have worse operating performance and vice versa, which could explain the high market outperformance reported by Gomers et al. (2003, 2009) and Renneboog, Horst and Zhang (2008).

Using an extensive proprietary database of corporate social responsibility engagements with US public companies spanning from 1999 to 2009, Dimson, Karakas and Li (2015) found that those engagements which addressed ESG concerns generated positive abnormal returns. Engaged firms concerned about reputational issues and with an enhanced capacity to implement changes were more likely to be engaged and after engagement, accounting performance in the relevant companies improved and better governance led to increased institutional ownership.

Brav, Jiang, Partnoy, & Thomas (2008) found that, for activist funds at around the time of investment announcements, stock prices increased and exhibited abnormal returns of up to 7%. Alok, Kumar & Wermers (2020) found that money managers underweighted holdings in climate-change-affected geographies more than other managers. This was attributed to the psychological tendency to overestimate severe event probabilities in proportion to their proximity and emotional significance.

Friede, Busch and Bassen (2015) undertook a comprehensive review of primary research on the relationship, if any, between ESG criteria and corporate financial performance. They found that 90% of studies observed a positive ESG–corporate finance performance relation and that most studies reported positive findings. The positive ESG impact on corporate finance performance appeared to be stable over time.

Choi, Gao and Jiang (2020) found that investors reduced holdings of high emission companies after higher-than-average seasonal temperatures and that – after abnormally hot periods – high carbon intensity stocks underperform low carbon emission stocks.

Interest in ESG investing in emerging markets is also widespread and constantly gaining traction (Cole et al. 2020), although caution is frequently advised because of poor non-financial disclosure and a lack of fundamental research which typify these geographies (Chauhan & Kumar 2018; Mobius & Ali 2021).

Limited work has also been conducted on South African ESG investment. Johnson, Mans-Kemp and Erasmus (2019) investigated the relationship between ESG scores and corporate financial performance measures. McCallum and Viviers (2020) explored key barriers (as well as potential opportunities) to impact investment in an emerging market, and Johnson (2020) connected ESG disclosure to South African firms’ capital costs. While these works are relevant and fill a gap in emerging market ESG literature, they did not assess empirical South African ESG performance, and ESG portfolio optimisation in an emerging market milieu was outside their ambits.

Despite the often-inconclusive evidence that investment in ESG stocks generate superior returns, ESG investing remains immensely popular among institutional and private investors, with no signs of abating. It remains, however, an evolving field. No universally agreed process has been devised which optimally combines profit-maximising metrics with ESG ones.

Méndez-Rodríguez et al. (2013) proposed a flexible approach to measuring mutual funds’ social responsibility score considering different dimensions and strategies of SRIs. Efficient frontiers corresponding to different SRI strategies and dimensions were plotted and the results show that including SRI constraints in the optimisation framework shifts the efficient frontier to the southeast in mean–variance space. Thus, for the same return (as portfolios on the traditional efficient frontier) the risk (on the ESG-constrained efficient frontier) tends to be greater as the level of social responsibility increases.

Beiting, Ioannou and Serafeim (2014) found that firms with better corporate social responsibility performance faced considerably lower capital constraints. No attempt was made to minimise capital or maximise risk-adjusted returns. Chen and Mussalli (2020) introduced an ESG investment framework which optimally collates the dual objectives of sustainability performance and α (excess market returns). While this work is important, we wished to explore optimal ESG risk-adjusted returns, not only the maximisation of α.

Qi and Li (2020) also extended the traditional portfolio selection model by imposing ESG constraints, computed the efficient frontier and found that the ESG-constrained frontier’s portfolio weights all lie on a ray (half line). The authors showed that portfolios on the ESG-constrained frontier and those which constitute the traditional efficient frontier comprise substantially different portfolio weights. Because portfolio weights are the foundation of asset selection and portfolio composition, this result is important for sustainable investments. Using component stocks of the Dow Jones Industrial Average Index from 2004 to 2013, Qi and Li (2020) found that the ESG-constrained and traditional efficient frontiers were similar (i.e. both occupied very similar loci in mean-variance space) which demonstrated that investors can still obtain the risk-return profile of (or very similar to) traditional efficient portfolios after imposing strong ESG requirements, despite the considerable differences in portfolio weights.

Utz et al. (2014) and Pedersen, Fitzgibbons and Pomorski (2020) attempted to address this issue by installing an equilibrium asset pricing model with ESG preferences and then deriving an ESG-efficient frontier for use by investors to invest in stocks based on risk-adjusted performance and ESG scores. The authors found that optimal portfolios could be generated by combining the minimum variance portfolio, the risk-free asset, the maximum Sharpe ratio (SR) (optimal risk-adjusted) portfolio, and a novel ‘ESG-tangency’ portfolio, that is, analogous to the maximum SR portfolio on the traditional efficient frontier. Gupta, Mehlawat and Saxena (2013) and Pedersen et al. (2020) found that ESG portfolios benefit investors as they are used to update expectations of returns and risk which in turn raises the efficient frontier. However, a cost arises from investor’s pursuit of ESG preferences: they leave the traditional efficient frontier and do not obtain optimal SR portfolio returns or risk.

The data comprised a sample of some 389 global emerging economy stocks grouped into seven business sectors (financials, industrials, basic materials, energy, consumer cyclicals, technology, and academic and educational services), sourced from Reuters Refinitiv database (Refinitiv 2021) and are summarised in Table 2.

While several ESG databases exist, Refinitiv was selected as it offers one of the most comprehensive ESG databases in the industry covering over 80% of global market capitalisation, providing details of more than 450 different ESG metrics over two decades, since 2002. This sample was used to ascertain broad insights into emerging economy ESG trends and patterns. A smaller sample comprising 12 South African stocks from three economic sectors, namely, financials (banks, insurance and reinsurance agencies, and asset managers), consumer cyclicals (retailers) and technology (software and IT services, Internet providers and mobile telecommunications) was used to focus on a single emerging economy’s ESG evolution. Like most other countries, South African corporate ESG data only date from 2007, so the timespan used was 2007–2020 with an annual sampling frequency (at this stage) much like the reporting of credit rating agency grades. Several institutions have advocated for years for a more frequent assessment and reporting of ESG data (CFA Institute 2019) and as a direct result, ESG information is now a regular feature of quarterly earnings reports (AFME 2021).

The ESG information provided include, on a 0 – 100 numeric scale, an ESG score, an ESG controversies overlay, a combined ESG score and the mappings of these scores to grades on an A – D scale (with additional ‘±’ notching identifiers). Environmental, social and governance scores provide a comprehensive evaluation of a company’s ESG performance based on publicly available, reported information covered in the ESG pillars (see Table 1) with an ESG controversies overlay. The function of the ESG controversy overlay is based on 23 ESG controversy topics and captured from global media reports, and its function is to discount ESG performance scores based on negative media reporting. Such commentary increases the impact of material ESG controversies in the overall ESG score. If scandals occur, the affected company is penalised, which affects their overall ESG score and grading. Negative event impacts may be dragged into subsequent years if new developments related to the negative event occur, such as lawsuits, rolling legislation disputes or fines. The overall ESG score is determined as the weighted average of the two component scores per fiscal period, with recent controversies reflected in the latest complete period. To account for this, only the overall ESG score was used in this work. We did not separate and independently investigate the individual components of the overall ESG score, mainly because controversy scores are relatively recent additions, so using only these would have constrained the somewhat limited data set even more. Note that different ESG scales, ranges and score methodologies are used by different ESG data providers.

Having selected the ESG-scored corporates, weekly share prices were assembled over the relevant period (as far back as 2007 in some cases). These share prices were then used to calculate share returns, volatility, portfolio variance-covariance matrices (where relevant) and SRs. These data were again gathered from Refinitiv (2021). Weekly risk-free rates (90-day rates of South African treasury securities) were sourced from the St Louis Federal Reserve Database (2021).