The Attribution Puzzle: What Are We Actually Measuring?
Before we can attribute anything, we need to define what “performance” means for an SRI fund. The naive approach—just comparing the fund’s return to a broad index—is deeply flawed. Imagine a fund that avoids all fossil fuels. In 2022, when energy prices spiked, that fund underperformed the S&P 500 by 300 basis points. Is that a failure of the fund manager? Or is it the *expected* cost of screening, fully justified by the investor’s values? Performance attribution for SRI is not just about returns; it’s about decomposing returns into intentional value-driven decisions versus unintended factor exposures.
The traditional Brinson-Fachler model splits returns into allocation effect, selection effect, and interaction effect. But this model assumes the benchmark is a neutral, accepted baseline. For an SRI fund, the benchmark itself is contested. Many SRI funds use a broad market index as their stated benchmark, but their investment universe is a subset of that index. This creates a permanent “benchmark gap.” If the excluded sectors (say, tobacco or arms) perform well, the SRI fund will show a negative allocation effect—not because of any bad decision, but simply because the investor’s constraints made the fund structurally different from the benchmark.
In our internal work at BRAIN, we’ve adopted a two-layer attribution framework. The first layer measures the impact of *screening* (which sectors and companies are excluded). The second layer measures the impact of *selection* (which allowed stocks are overweighted or underweighted). This separation is crucial. Screening effects are often deterministic—you know exactly what you’re excluding. Selection effects are where the manager’s skill (or luck) shows up. Mixing these two layers is like putting a marble and a golf ball in a box and trying to measure their combined weight while they keep colliding—you get a number, but it tells you nothing about either object.
One case that opened our eyes was a Japanese pension fund client. They asked us to analyze a “green” fund that claimed to beat its benchmark. Our attribution showed that 85% of the outperformance came from *not holding Toyota Motor* (which had a governance scandal in 2022) and *overweighting semiconductor equipment makers* (which had nothing to do with sustainability). The ESG label was a distraction. The fund was really a quality-growth fund with a green makeover. The pension fund’s board was shocked. Our job was not to judge, but to make the returns legible. This legibility is the core purpose of performance attribution in the SRI space.
---Factor Decomposition: The Hidden Hand
If you’ve been in this industry for any length of time, you know the drill: every fund’s returns can be “explained” by a factor model—market, size, value, momentum, quality, low volatility. For SRI funds, factor decomposition is both a blessing and a curse. The blessing is that it allows investors to see whether the fund’s observed performance is simply compensation for taking on more systematic risk. The curse is that factor models are notoriously unstable when applied to constrained portfolios like SRI funds.
The instability comes from the constraint itself. Suppose an SRI fund excludes all companies with high carbon emissions. This exclusion is not neutral—it systematically removes heavy industries, some utilities, and many mid-cap manufacturers. A standard Fama-French value factor will now behave differently for this fund than for the broader market, because the fund’s opportunity set is truncated. Our research at BRAIN has found that the value factor’s loading for a typical fossil-fuel-free fund is roughly 20% lower than for an unrestricted portfolio. That means when value stocks rally, SRI funds lag—not because the manager is bad, but because the screening process changed the factor landscape.
To handle this, we use a custom factor model calibrated to the SRI universe. This is not a trivial exercise. We create “shadow factors” that mimic the returns of the excluded stocks, allowing us to quantify the screening effect separately. For example, if the shadow value factor returns +12% in a given quarter, and the SRI fund underperforms by -1.5%, we can say that -1.2% of that underperformance was “screening-driven” (the value shadow factor’s exposure cost), and the remaining -0.3% came from stock selection. This decomposition is far more actionable than a blanket statement like “the fund underperformed due to stock picks.”
I recall a specific case from early 2023. A client in the Netherlands had a “Paris-aligned” fund with a temperature alignment target. The fund’s returns in 2022 were abysmal—down 22%. Standard attribution pointed to a huge negative allocation to energy. But our custom factor model revealed something else: after removing the screening effect of excluding fossil fuel producers, the manager’s selection ability was actually *positive* by 180 basis points. The manager was doing a great job picking clean energy stocks; the problem was purely structural. We cannot overstate the importance of this distinction. The client’s initial reaction was to fire the manager. After our analysis, they doubled their allocation—and in 2023, the fund recovered strongly, outperforming its benchmark by 400 basis points.
But factor decomposition has its limits. Most factor models assume linearity and normality, which break down during stress events. In March 2020, the COVID crash produced non-linear correlations across nearly all risk factors. SRI funds, paradoxically, showed *less* idiosyncratic volatility but *more* systematic sensitivity during that month. This is because the selloff was most violent in cyclical, high-beta stocks—many of which are excluded from SRI portfolios. So the screening actually increased the fund’s beta to the downside. A good attribution system must be tested across market regimes, not just in calmer periods.
---Unintended Consequences of Exclusion-Based Strategies
Let me give you a concrete example that still makes me wince. A few years ago, we worked with a Nordic asset manager who had a strict exclusion list: no companies with coal revenue above 5%, no tobacco, no gambling, no weapons. They wanted us to attribute their fund’s performance relative to the MSCI World. The results were counterintuitive. The fund was underperforming by 200 basis points annually, but our screening-layer analysis showed that *exclusions alone* were responsible for 180 basis points of that drag. The manager wasn’t bad—the policy was costly.
There’s a term for this in the industry: the “exclusion burden.” Exclusion-based SRI strategies inherently sacrifice diversification, and the cost of that sacrifice is real and measurable. But here’s the twist: a growing body of research suggests that the exclusion burden varies enormously by sector and region. Excluding tobacco, for example, has historically cost investors very little—tobacco stocks are few, and their performance has been mediocre relative to the broader market. Excluding energy, by contrast, has been brutally expensive in 2021-2022.
However, the unintended consequences go deeper than just raw return drag. Consider *ownership-based* effects. When a large SRI fund excludes a company, that company’s cost of capital rises. This is the “virtuous circle” that SRI proponents tout. But the opposite is also true: excluded stocks sometimes outperform *because* they’re excluded. The stigma of being cut from an SRI index can attract short-sellers, create information asymmetry, and sometimes lead to contrarian buying. We’ve seen situations where the exclusions create a “value trap premium” for the excluded stocks—they become so underpriced that they offer outsized returns to non-SRI investors. This is a perverse feedback loop that makes attribution even more complex.
At BRAIN, we’ve developed a metric called “exclusion opportunity cost” (EOC). This measures the return differential between a company’s actual returns and its expected returns based on its fundamental and risk-profile characteristics, given that it’s excluded from an SRI universe. We found that EOC is positive in emerging markets (excluded stocks tend to outperform their fundamentals) and negative in developed markets (excluded stocks are often overvalued, and exclusion is a blessing). What this means is that the same SRI policy has opposite effects in different geographies—performance attribution that ignores this is dangerously misleading.
---Timing and Portfolio Rebalancing Effects
You can have the best stock-picking model in the world, but if your rebalancing policy is slow, your performance attribution will show phantom losses. This is especially true for SRI funds because their investment universe is dynamic—companies get added or removed from ESG indices based on annual reviews, controversies, and ratings changes.
Imagine a company that receives a sudden ESG downgrade (e.g., an oil spill). An SRI fund with a monthly rebalancing policy might still hold that stock for up to 30 days, while an index-based SRI product would have sold it immediately. During that window, the stock drops 10%. Whose “fault” is that drop? The fund manager’s? The rating agency’s? Or simply the cost of a less agile portfolio construction process? Attribution models that don’t explicitly capture the rebalancing lag will misattribute this loss to stock selection, creating a downward bias in the manager’s skill estimate.
There’s also the issue of what I call “events-based reconstitution.” Many SRI indices reconstitute their constituents once a year, in March. Between March and the next reconstitution, the index can become stale. Our research at BRAIN shows that SRI indices typically beat their own fundamental benchmark in the months immediately after reconstitution (because they’ve just refreshed their screens), but lag in the months before the next reconstitution (because they’re holding older, potentially less “green” names). This seasonal pattern is purely mechanical, yet it can dominate the short-term attribution results.
For our internal attribution engine, we separate rebalancing effects from selection effects. We define a “shadow portfolio” that assumes instantaneous execution of all screening decisions on the first trading day after a rating change. The difference between this shadow portfolio’s return and the actual fund’s return is purely operational—it captures execution slippage, liquidity costs, and timing delays. In practice, for many SRI funds, this operational drag is between 40 and 150 basis points per year. That’s a huge number that most investors never see because their attribution reports lump it into “residual” or “other.”
Let me share a personal story. In late 2022, we had a client who switched from a passive SRI index fund to an actively managed SRI fund. The active fund underperformed in the first quarter. The client demanded an explanation. Our attribution showed that the active fund had made a deliberate decision to *delay* selling a controversial stock (a mining company) because they believed the controversy was temporary. The passive fund sold immediately. In quarter one, passive won; in quarter two, the mining stock rebounded, and active won. Timing is not just a risk; it’s a philosophical choice. And attribution must respect that.
---Thematic Overlay: Green Alpha and Social Beta
Now, let’s talk about something that’s become a buzzword at every ESG conference since 2020: “green alpha.” The idea is that companies with high ESG ratings or low carbon footprints will deliver superior risk-adjusted returns. Performance attribution of SRI funds needs to test this claim head-on.
From a mechanical perspective, we treat thematic overlay as a distinct “style factor.” For example, if an SRI fund has a formal commitment to invest at least 20% of assets in renewable energy projects, then the return contribution of the renewable energy basket is comparable to the allocation effect in Brinson attribution. We isolate this basket and measure its performance. Does it add value or not? As of mid-2024, the evidence is mixed. Renewables have been hugely volatile—2020-2021 was golden, 2022 was terrible, 2023 was decent. Any thematic overlay attribution that doesn’t adjust for the market timing of when the overlay was funded will be misleading.
There’s also “social beta”—the performance of stocks that are perceived as socially responsible but not necessarily high-ESG-rated. Think of education companies, healthcare providers, or community banks. Some SRI funds implicitly tilt toward social beta by favoring companies with employee-friendly policies or unionized workforces. Social beta is harder to define than green alpha, and attribution models often ignore it entirely. At BRAIN, we built a proxy social beta factor using companies that score highly on labor-management relations and human capital development. We’ve found that this factor has a low correlation with the market and with value/quality factors, meaning it provides genuine diversification.
But the most dangerous pitfall in thematic attribution is what I call “narrative contamination.” This happens when a fund’s performance is attributed to a thematic factor, but in reality, the factor is just a proxy for something else. For instance, a “clean water” fund might actually be a mid-cap industrial fund in disguise. Thematic attribution requires rigorous regression analysis to ensure the factor loads are economically meaningful, not just statistically significant. We need to be alert to the risk of “greenwashing by attribution.” A fund that claims its outperformance came from ESG integration, when in fact it came from a sector bet, is engaging in a subtle form of misleading marketing.
---Data Quality and Scope: Garbage In, Garbage Out
I’d be lying if I said the biggest challenge in SRI attribution is the math. It’s not. The biggest challenge is the data. ESG data is a mess—inconsistent, overlapping, sometimes contradictory. One rating agency gives Tesla a 95/100; another gives it a 40/100. The variance is so large that any attribution model relying on a single ratings source is fundamentally unreliable.
In our daily work at BRAIN, we don’t just pull ESG scores from one vendor. We aggregate data from five major providers—MSCI, Sustainalytics, Refinitiv, S&P Global, and Bloomberg—and we create a composite score. Even then, we’re careful to use this composite only for portfolio construction, not for attribution. Why? Because the attribution process should be agnostic to the ratings. Instead, we focus on *outcome-based* metrics: carbon emissions, water usage, waste generation, gender pay gaps. If a fund claims to be a low-carbon fund, the most robust attribution uses actual carbon data, not a rating.
The data problem is compounded by scope issues. When we attribute performance, we should be including cash flows. But many SRI funds, especially mutual funds, have daily cash inflows and outflows. The standard time-weighted return (TWR) correctly neutralizes the impact of external cash flows, but it also masks the impact of the manager’s decisions on when to hold cash. A fund that is fully invested in equities but holds 5% cash because of valuation concerns has made a tactical decision. TWR-based attribution will not capture this. We need money-weighted return (MWR) attribution for this purpose, but MWR is hard to calculate for funds with daily subscriptions. We must acknowledge the need for blended valuation approaches.
Another scope issue: corporate actions. Mergers, spin-offs, and index changes happen frequently in the SRI space because companies are constantly trying to improve (or game) their ESG profile. A company that spins off its coal assets into a separate entity—thereby cleaning its own ESG score—creates a shadow problem for attribution. The SRI fund might still hold the coal spin-off (if it was received as a dividend), but its ESG mandate might force it to sell. The return from the spin-off period must be attributed somewhere. In practice, we’ve seen funds with 10-15% of their total return in a single year coming from corporate actions related to ESG thresholds. Standard models just throw this into “residual.” That’s not good enough.
---Behavioral and Governance Factors in SRI Portfolios
Last but not least, let’s talk about humans. Because that’s what performance attribution always comes down to—human decisions, human biases, and human governance.
SRI fund managers face a peculiar psychological pressure. They are judged not only on returns but also on their adherence to values. This dual mandate creates a “crowding effect” into what we might call “woke high-quality stocks.” Apple, Microsoft, Salesforce, NVIDIA—these are overwhelmingly popular in SRI portfolios. Not because they’re the best investments, but because they’re safe from a moral and reputational perspective. Attribution analysis reveals this crowding effect empirically: the “residual” returns of many SRI funds cluster tightly around the tech sector, which means their idiosyncratic risk is actually quite low. The managers are doing themselves a favor, but they’re not doing their investors a favor.
Governance is another issue. Many SRI funds are structured with an advisory board or an ethics committee. These committees can override the fund manager’s decisions. Suppose the committee decides to divest from a pharmaceutical company because of pricing controversies. That decision has a real performance impact. But whose “selection effect” is it? The committee is not an investment manager, yet its actions directly affect returns. Our attribution framework includes a third layer: “governance effect.” This captures the performance impact of decisions made outside the formal investment process. It’s an uncomfortable revelation for many clients, but we feel it’s necessary for full transparency.
I remember a bank in Singapore who almost dropped us as a vendor because we showed them that their own responsible investment committee’s decisions, taken months apart, had cost the fund 150 basis points of alpha. The committee hadn’t done anything “wrong,” but the timing of their decisions was terrible—selling ethically dubious stocks right before they rebounded. After a tense phone call, they came around and now use our governance attribution as a standard reporting tool. The truth hurts, but it also helps.
--- ## Conclusion: The Future is Integrated, Not Isolated So where do we stand? After all this analysis, what have we learned? First, performance attribution for SRI is not a single number—it’s a multidimensional framework that must include screening effects, factor loadings, rebalancing impacts, thematic overlays, data quality adjustments, and governance effects. I predict the field will mature in the next 3-5 years, with standardized ESG attribution becoming as common as Brinson attribution is today. Second, we’ve learned that SRI performance is highly contextual—it depends on the market regime, the sector rotation, and the specific universe of excluded stocks. Third, and most importantly, we need to shift the conversation from “does SRI perform?” to “for whom, in what conditions, and at what opportunity cost?” That’s the only way SRI can move beyond its current niche status. My recommendations: (1) Asset owners should demand better attribution reports from their SRI fund managers, specifically asking for a separate reconciliation of screening versus selection effects. (2) Data vendors need to harmonize their ESG definitions to reduce model noise. (3) Researchers should focus on outcome-based and irreversible impact factors, not static ratings. And (4) the next frontier is *dynamic* attribution—using machine learning to detect non-linear relationships between ESG factors and returns that standard linear models miss. At BRAIN TECHNOLOGY LIMITED, we’ve built a proprietary attribution engine that handles all these layers. It’s not perfect, but neither is the data. What it does do is give investors a crystal-clear view of *why* their SRI portfolio is moving the way it does. And in a world where “greenwashing” is a four-letter word, that clarity is worth its weight in gold. --- ## BRAIN TECHNOLOGY LIMITED’s Perspective on Performance Attribution of SRI At BRAIN TECHNOLOGY LIMITED, we view performance attribution not as a backward-looking report, but as a forward-looking diagnostic tool. Our experience building AI-driven financial models has taught us that the biggest risk in SRI investing is not poor returns—it’s *misleading narratives*. Without rigorous attribution, investors can’t distinguish between a fund that’s truly generating alpha through superior ESG integration and a fund that’s accidentally loading onto a momentum factor while talking about saving the planet. We believe that the future of SRI depends on transparency, and transparency depends on state-of-the-art attribution. Our platform integrates real-time ESG data feeds, custom factor libraries, and automated rebalancing tracking to give our clients a granular view of every basis point. We are particularly excited about using natural language processing to incorporate news sentiment into attribution analysis—so that when a company has an environmental scandal, the immediate and longer-term performance effects are captured with high precision. Ultimately, we believe that the convergence of AI and granular ESG data will create a new gold standard for investment accountability, making SRI not just an ideological choice but a quantitatively validated one. We are proud to be part of that journey.