ESG indices – unintentionally active

• Supposedly passive ESG indices often show significant deviations from the benchmark.
• Portfolio construction using tracking error optimisation is susceptible to changes in correlation ratios, resulting in relative risks.
• Optimising the active share does not require an estimation process and ensures a significantly more stable portfolio with low relative risks.
   

ESG integration approaches

The integration of ESG and sustainability criteria into the asset allocation of institutional investors presents new challenges for the portfolio construction of passive investments. Essentially, there are two main approaches to incorporating ESG criteria into index-related investments.

Most ESG indices use a combination of filtering, sorting, and weighting for portfolio construction. For example, certain stocks are initially filtered out (negative list) or the investment universe is restricted to certain stocks (positive list). These selections may lead to sorting based on specific criteria, with the final weighting of securities determined by a combination of ESG criteria and market capitalisation. Naturally, this approach results in relative risks compared to the benchmark, as the portfolio construction focuses solely on the emphasised criteria, without considering the associated deviations from the benchmark.

A key issue for institutional investors is to integrate sustainability criteria into their investments while avoiding unwanted devia­tions from the benchmark. To achieve this, portfolio optimisation is required that replicates the benchmark as closely as possible while simultaneously fulfilling the desired sustainability criteria.

Tracking error optimisation

n practice, this type of optimisation is usually implemented by minimising the tracking error (the standard deviation of the difference in returns between the benchmark and the portfolio). This means that, given an improvement in the sustainability profile, such as a 30 per cent reduction in the carbon footprint relative to the benchmark, a portfolio is constructed to have the minimum ex-ante tracking error. However, this approach has a significant drawback: estimation errors can lead to considerable discrepancies between the expected ex-ante tracking error and the realised ex-post tracking error.

To calculate and optimise the expected tracking error, the vari­ances and correlations of all securities in the underlying universe are required. Estimating these correlations accurately is particu­larly challenging, as they tend to be unstable over time. Figure 1 illustrates the rolling 3-month correlation over the past five years between Apple and Microsoft, the two largest positions in the MSCI World.

Figure 1: Rolling 3-month correlation Apple & Microsoft
(31.12.19 – 29.02.24)

As it turns out, the correlation between the two index heavy­weights fluctuates significantly: even apart from the turbulent market phase surrounding the COVID-19 pandemic in March 2020, values ranged from around 0.5 to 0.8 within a few months in 2021.

The problem with tracking error optimisation is that it relies to a certain extent on the estimated correlation remaining stable over time. However, if the correlation changes – regardless of the direction – the realised tracking error inevitably increases, often significantly beyond the ex-ante expected level. This results in larger deviations from the benchmark, which can significantly negatively impact the tracking difference – the cumulative return difference between the portfolio and the benchmark.

Active share optimisation

An alternative measure of the relative risk to the benchmark is the active share. This measures the absolute weighting deviation of all securities in the portfolio compared to those in the bench­mark. This makes it possible to determine at any point in time to what extent the portfolio corresponds to the benchmark.

A perfect replication of the benchmark therefore has an active share of 0 per cent. A portfolio of securities that are not included in the benchmark has an active share of 100 per cent. This intuit­tive measure can be used in portfolio construction by directly minimizing the weight deviations of individual securities. The result is a portfolio with the lowest possible deviations from the benchmark.

Assenagon Equity Framework

To implement the optimisation described above, we use the squared individual stock deviations within the Assenagon Equity Framework. This results in a portfolio that has the lowest possible active share, which is also distributed across a maximum number of positions. This means, for example, that ten shares with a deviation of one basis point each are favoured over one share with a deviation of ten basis points. The desired profile of the portfolio – in this case the integration of ESG criteria – therefore has a broad base and is not dominated by just a few positions.

In addition to the individual stock deviations, a holistic view in portfolio construction is essential in order to preserve all relevant value drivers of the benchmark. Major deviations in the country/ sector allocation and biases in the factor profile are avoided through appropriate constraints in the optimisation. As a result, the tracking difference is minimised.

Figure 2: Relative performance vs global equity market
(14.09.21 – 29.12.23)

For the investor

The direct minimisation of individual stock deviations has major advantages, as these can be determined precisely at any time without the need to use statistical models to estimate the corre­lation and variances (calculation of ex ante tracking error). Figure 2 uses the relative performance of two portfolios to show how the two methods (tracking error and active share optimisation) can differ in their results. Based on the global equity market, a similar ESG strategy is implemented in both cases, which includes an ex­clusion list as well as an improvement in the ESG score (approx. 10 per cent) and a reduction in CO2 intensity (approx. 30 per cent).

Due to the problems of tracking error optimisation described above, the ESG Focus Index exhibits disproportionately high risks, which have a negative impact on the tracking difference. In addi­tion, the ex ante expected tracking error of 0.50 per cent is signifi­cantly exceeded at 0.72 per cent.

By design, the ESG core investment is as close as possible to the benchmark in all return and risk dimensions and thus demon­strates a high degree of robustness against rotations/changes in the market structure by systematically limiting individual security risks to the minimum. The low realised tracking error of 0.46 per cent confirms this.

P S: Stay tuned for the upcoming issue, where we delve into the risks and side effects of popular ESG approaches.