Optimal portfolio alignment to net-zero targets

The dual goals of portfolio implementation

The objective of our TargetNetZero strategy in equities is to align portfolios to a set of climate-related targets¹. The key elements of the strategy are the analytics provided by our sustainability research team that measure companies’ decarbonisation ambitions in the form of an implied temperature rise. While this methodology has been extensively discussed in a series of articles, its portfolio implementation has not been given the same level of coverage so far.      

In our view, portfolio implementation has a dual goal. The main goal is to ensure that the portfolio is fully aligned with its decarbonisation targets. As we highlighted previously, simply adding portfolio constraints is not enough to ensure true alignment. Uncertainty about the future decarbonisation pathways of companies tends to result in a consistent disparity or bias between the measured and the ‘true’ alignment of optimised portfolios. We argue that these biases can be mitigated by incorporating model uncertainty into the portfolio construction process.  

Read also: Capturing market upside while benefitting from net-zero tilts

The second goal of portfolio implementation is to ensure that targets are achieved optimally. Optimality is a rather ambiguous concept that can mean different things depending on the context, therefore, we feel the need to clarify our understanding of it. In this article, we will outline different approaches to portfolio implementation with a view that the optimisation process should ideally incorporate both risk and return dimensions.

Minimum impact solutions

The simplest approach to building an aligned portfolio is to transform the original benchmark index in a minimal way to align to given targets. The impact of such transformation can be measured, for example, by portfolio turnover².  

Minimising portfolio turnover is a natural objective as it lowers transaction costs. However, it is not widely used for portfolio construction as it does not always yield a unique optimal portfolio. A more traditional measure of impact that yields a unique optimal solution is the Euclidean distance³ between portfolio and benchmark weights. Minimising this distance is equivalent to maximising the holding-based correlation between the aligned portfolio and its benchmark⁴.

Both turnover and distance have the advantage of being model-free measures. On the flip side, these measures implicitly assume that stocks are independent; therefore, an underweight position in one stock cannot be ‘hedged’ by an overweight position in another. To allow for the natural complementarity of stocks, we must introduce a risk model with realistic correlations.

Tracking error minimisation

A risk model enables us to measure the ‘distance’ between portfolios in terms of the volatility of excess returns, or the tracking error (TE). Minimising the TE of an aligned portfolio is effectively equivalent to maximising its return correlation with its benchmark⁵.

Minimising TE is, however, a more challenging task as it is not model-free, therefore, its success will depend on the quality of the risk modelling. In practice, portfolio managers never fully rely on their risk models incorporating explicit constraints on systematic exposures such as sectors, regions and styles as well as single stock deviations.

In our view, TE minimisation is a better match for investors’ focus on active returns than holding-based measures. It is clearly easier to express risk tolerance in terms of the risk budget as opposed to portfolio distance or turnover. For example, portfolios with a TE below 1% are commonly considered to be low-risk vehicles.    

related content.

equitiesPlanetary Transition

equitiesPlanetary TransitionTNZ equities

Sustainability in the new Trump era

24 January 2025

equities: adapting to market shifts, optimising for net zero 

learn more

‹›

The ‘elephant’ in the room

We feel neither risk-based nor holding-based measures address the ‘elephant’ in the room – the expected returns. Optimising portfolios solely based on risk effectively means that either we have no convictions on individual names or investors are infinitely averse to risk. In our view, the latter is very unlikely since otherwise there would be no allocation to equities in the first place.

Minimising TE involves adjusting numerous positions: some adjustments are driven by the portfolio’s alignment to its targets while others are for pure risk-hedging purposes. While a minimal TE solution may be unique, there is still a broad variety of alternative portfolios with almost the same TE but with materially different compositions and potential returns. If a portfolio manager does have convictions on individual names, we believe it is more beneficial to use them as an additional selection criteria as long as it does not impact TE materially. The end result will be a portfolio with enhanced return potential without additional risk. Effectively, this offers investors the opportunity of a so-called ‘free lunch’.    

Taking advantage of a ‘free lunch’

Leveraging our years of experience in systematic investing, we have developed a dynamic alpha strategy that goes beyond traditional factor investing. Dynamic alpha is based on the timing of a broad range of equity styles across various sectors and regions driven by momentum, valuation, macro and machine learning signals.

Figure 1 provides a real-life illustration of the risk-return trade-off generated by enhancing a minimum TE portfolio⁶ by adding an alpha overlay⁷. The curve represents the efficient mean-variance frontier formed by portfolios that are fully aligned to net-zero targets while having different risk and return profiles.

“ Leveraging our years of experience in systematic investing, we have developed a dynamic alpha strategy that goes beyond traditional factor investing”.

The choice of a single optimal portfolio boils down to investor preference. The efficient frontier starts from the lowest TE portfolio, which is the optimal choice for the most conservative investor. At that point, the curve becomes almost vertical meaning that the portfolio return can be enhanced without a meaningful impact on TE. This is our ‘free lunch’ zone.

To assess where the optimal portfolio might be located, we assume that the investor uses the following mean-variance utility function:

utility function = excess return – risk aversion×TE²

The utility function is naturally increasing in the (expected) excess return and decreasing with the tracking error, while the risk aversion parameter determines the investor’s attitude toward risk. The optimal portfolio is the unique one on the efficient frontier with the highest utility.

FIG 1.  Efficient frontier of the TargetNetZero strategy⁸

To determine possible magnitudes of risk aversion, consider an investor who has the opportunity to also engage in a long-short strategy with a Sharpe ratio of as high as 3. Then suppose that the investor is extremely conservative and chooses to allocate only 1% risk to this strategy. This scenario allows us to recover the implicit risk-aversion parameter⁹.  Figure 1 highlights the location of the optimal portfolio for such a conservative investor. The optimal choice seems to be quite intuitive. By sacrificing a marginal increase in tracking error from 0.65% to 0.73%, the investor can achieve a material enhancement in the expected return of 0.40% – not quite but very close to getting a ‘free lunch’.

Read also: Decarbonisation: are companies on the right track?

How can we ensure the optimality of the TargetNetZero portfolios?

Our responsibility as portfolio managers is to ensure that the portfolios achieve their targets in the most optimal way. While we do not know the risk tolerance of each investor, even the most conservative estimates suggest that an optimal portfolio should incorporate elements of return enhancement. In our TargetNetZero portfolios, we have applied a marginal overlay – using our in-house dynamic alpha – to enhance potential returns, which, in our view, should be beneficial for investors.