A science-based heuristic to guide sector-level SDG investment strategy

ABSTRACT Aligning investments with Sustainable Development Goals (SDGs) has been a longstanding ambition for many private investors. The assessment of corporate impact on the SDGs is not a trivial task, and most present-day attempts often overlook SDG interactions, and lack scientific anchoring and transparency. We present an evidence-based review approach for investors to assess sector-level impacts on individual SDGs, and score these using a traffic-light system. Our initial review documents impacts of 81 economic sectors on SDGs 1-16. Results show that environmental SDGs are impacted negatively by most economic sectors, and that primary sector activities negatively impact the highest number of SDGs. Using the agricultural sector as a case, we draw on Causal Loop methodology to illustrate spillovers resulting from SDG interactions. Our findings point to three key considerations of relevance for sustainable investment strategies; the necessity to capture ‘impact shadows’, spillovers across SDGs, and the hierarchical nature of the SDGs.


Introduction
Assessing how economic activities impact the multiple dimensions of sustainability is a central challenge of the twenty-first century.Investors and financial institutions are often identified as key levers in contributing to sustainable outcomes by directing capital towards companies and economic sectors that ensure alignment with the Sustainable Development Goals (SDGs) (Crona, Folke, and Galaz 2021).The global reach of the financial services sector, and its pivotal role for real economic development, means that changes in investment practices can have a significant influence on our collective ability to achieve global sustainability targets (Weber 2014).It is therefore encouraging to see a growing number of institutional investors and rating agencies stepping up to the challenge of 'transforming' the investment landscape, whether by tracking or assessing how their portfolios impact sustainability, changing norms or influencing policy to enable and accelerate sustainability transitions. 1 While these initiatives often attempt to align with the SDGs, many still suffer from two primary shortcomings.
First, many initiatives reflect the optimism that negative impacts on the natural environment will be temporary and reversable.This rests on the idea that humanmade capital, such as technological advancements, will succeed in decoupling economic activities from environmental impact.For example, some SDG data providers use tools which allow for positive impact on certain dimensions of sustainability to compensate for negative impact (Van Zanten et al. 2023).When aligning investments with individual SDGs, spillovers onto other dimensions of sustainability are to be anticipated.For example, optimizing for specific sustainability goals (e.g. economic goals or reducing GHG emissions) can cascade into various positive and negative impacts on health, biodiversity and other sustainability themes.Additional spillovers are to be anticipated when investing in one economic activity, since every economic activity is embedded in a chain of operations.Consequently, investments in specific companies or sectors result in direct and indirect impacts throughout their supply chains and final use of products or services.However, most current practices do not yet capture spillovers beyond investments in specific assets, and upstream and downstream impacts are therefore not taken into consideration (Popescu, Hitaj, and Benetto 2021).
Another major issue with many current attempts by investors and rating agencies to address sustainability is that their methods are plagued by poor transparency (Popescu, Hitaj, and Benetto 2021).In many instances, it is not clear what data is included in the impact assessments, how the data was subsequently combined to develop impact metrics, and most importantly, if the measures used actually capture environmental and social impacts of corporate activities (Crona and Sundström 2023;Crona, Folke, and Galaz 2021).This reflects a general confusion over the definition of sustainability.For example, when referring to sustainability impacts, ESG (Environmental, Social, and Governance) and SDGs are used interchangeably by several actors of the financial sector, despite the fact that ESGs are a risk management tool for companies, and thus poor measures to capture impacts on sustainability.These weaknesses are increasingly acknowledged by scholars and sustainable finance practitioners (Van Zanten and Huij 2022; Crona, Folke, and Galaz 2021;Popescu, Hitaj, and Benetto 2021;Crona and Sundström 2023;Agnew, Klasa, and Mundy 2022;Tricks 2022;Wassénius, Crona, and Quahe 2023).On a positive note, new transparent ways to capture, measure and standardize impact across multiple environmental domains are being developed (cf.Lade et al. 2021), as well as standards for reporting progress on science-based targets (e.g.Science-Based targets.org).Yet, the handful of actors currently dominating the sustainability data (often ESG) provision for investors (Escrig-Olmedo et al. 2019), remain largely opaque in their methods and rely heavily on company-reported data, which often do not capture the most relevant aspects for assessing impact across multiple SDGs (Crona and Sundström 2023). 2 Company-reported data are also often inconsistent due to a lack of standardized approach to measure company impact on the SDGs.
Readers might note that several efforts have recently been initiated to address the standardization of corporate reporting.Most notably the Corporate Sustainability Reporting Directive (CSRD) by the European Union (EU), The Taskforce on Nature-related Financial Disclosures (TNFD), and the International Sustainability Standards Board (ISSB), initiated by the International Financial Reporting Standards Foundation.While encouraging, these accounting standards will take time to develop and harmonize.At the same time, the world is rapidly approaching 2030the year by which the SDGs should be largely achieved.Interim tools to help investors rapidly steer their capital towards that which reduces harm and potentially facilitates a green transition are therefore paramount.
In this paper, we present evidence-based indications of sector-level impacts on SDGs, which can guide equity investors in identifying general impacts typically expected by companies who belong in the same sector.It is important to note that sector-level assessments should be used by investors merely as proxies, not comprehensive assessments of investment impacts on sustainability.As such they can be especially useful for active portfolio management as a broad heuristic for assessing general sector-level impacts on SDG, which can and should ideally be complemented with company-specific information to identify laggards and leaders.The sector-level approach to mapping SDG impact avoids the difficulties and time-consuming aspects of dealing with individual company reports and is in line with recent recommendations by UNDP (2021).The approach presented in this paper rests on an extensive data-driven assessment of the impacts of 81 economic sectors on SDGs 1-16, where a total of 1382 publications were reviewed to identify unique causal mechanisms of impact.This resulted in a total of 1296 sector-SDG combinations, each capturing how a specific economic sector impacts a specific SDG.The approach is novel in several ways.First, it focuses on SDG impact of economic sectors (not individual companies) and is geared towards giving investors and investment scholars a broad understanding of portfolio-level impacts.It thus departs from the growing body of literature examining the impact of the private sector on SDGs using company-reported data (Bose and Khan 2022;Calabrese et al. 2021Calabrese et al. , 2022;;Pizzi, Rosati, and Venturelli 2021;Rosati andFaria 2019a, 2019b;García Meca and Ferrero 2021).It also adds to, and complements, the scholarship examining SDG impacts generated by a single economic sector (Fuso Nerini et al. 2018;Oliveira Neto et al. 2019;Monteiro, da Silva, and Neto 2019;Swain and Karimu 2020;Hazarika and Jandl 2019).Third, it builds on the growing literature on SDG interactions (Nilsson et al. 2018;Pradhan et al. 2017;Kroll, Warchold, and Pradhan 2019;Collste, Pedercini, and Cornell 2017;Biglari, Beiglary, and Arthanari 2022;Le Blanc 2015), and integrates such thinking with analysis of impacts from a broad set of economic sectors and Causal Loop methodology to illustrate how the documented interconnections between SDGs can cause impacts from one sector to ripple through and cascade to multiple SDGs (see Biglari, Beiglary, and Arthanari 2022 for related methodology).Our work builds on Van Zanten andvan Tulder (2020, 2021), yet we follow more granular methodological steps, use a distinct analytical approach, and focus our discussion on sustainability impact assessments for the financial sector.
Our results highlight which sectors appear to have the most positive and most negative respective impacts across all SDGs.They also show that across all sectors, the most consistent and negative impacts are to the environmental SDGs.This is alarming since a prerequisite for achieving many social and economic development goals is well-functioning ecosystems (Dasgupta 2021;IPBES 2019).We therefore discuss how impact assessments aiming to capture 'real' impact of investments on the SDGs would benefit from integrating the following three considerations.First, the notion that the SDGs are not of 'equal' importance for sustainability.The environmental SDGs underpin several social and economic SDGs (Folke et al. 2016).This has implications when measuring how economic activities impact the SDGs and subsequently, how 'actually sustainable' sustainable investments are.Second, impact assessment methods would be upgraded from taking a more holistic approach to impact, i.e. by integrating the notion that trade-offs exist and should be anticipated.We argue that any capital allocation in economic sectors will directly result in positive and/or negative impacts on certain SDGs as well as spillovers to several aspects of sustainable development.We also discuss the importance of considering not only immediate impacts of economic activities but also impacts from the entire supply chain.Third, impact assessment methods that are transparent in their assumptions and data sources are needed.These can contribute to the urgency of aligning our economic activities with the SDGs.We present our methodological approach and share a database of economic sectors´impacts on SDGs 1-16 that supports our findings.This database outlines more than 4000 impact mechanisms which are derived, to the extent possible, from scientific literature.We hope that these three insights can contribute to upgrading investment impact assessment methods and promote sustainable transitions.

Conceptualizations of sustainable developmenta brief background
Below we briefly review key conceptual developments related to the SDGs and environmental and social sustainability, to make explicit the assumptions on which our analytical approach is based.We start with the SDGs.They are a set of goals agreed upon by 195 countries in 2015, as part of the United Nations (UN) 2030 Agenda (https://sdgs.un.org/ goals).They comprise 17 goals, 169 targets and 231 indicators, and grew out of the Millennium Development Goals, which guided the UN development work from 2000 to 2015.The SDGs can be considered a guide for transitioning towards sustainable futures for all, and they encompass social, environmental and economic considerations.The original communication material initially presented the 17 goals alongside each other.However, critique has been raised that this gives the impression that each goal should receive equal consideration.In line with this, a nested approach to the SDGs has emerged.Developed by the Stockholm Resilience Centre, the SDG 'wedding cake' was first presented by Rockström and Sukhdev in 2016 to illustrate the relationship between the SDGs and the three sustainability pillars (environmental, social and economic; Rockström and Sukhdev 2016), yet layering them to recognize that natural capital provides essential functions that not only support, but make possible, the social and economic pillars.This nested conceptualization of the SDGs highlights that social well-being depends on a healthy biosphere and that our economies are underpinned by social and natural assets (Folke et al. 2016).
Before the release of the SDGs, the concept of 'Planetary Boundaries' was developed.These boundaries represent a set of bio-geo-physical processes that are paramount for a stable earth system (Rockström et al. 2009;Steffen et al. 2015).Transgressing them might lead to major environmental changes, with societal and economic impacts.In contrast, staying within them while simultaneously pursing social well-being represents a 'safe and just operating space' for humanity (Raworth 2017).
Together, the conceptualizations delineated here contribute three important insights about the nature of the SDGs which inform our analytical approach to assessing sectoral impacts on SDGs to support improved investment decisions.First, the nested nature of the three pillars of sustainability means that societal and economic prosperity depends on a functioning natural environment.Second, the bio-geo-physical processes that support the relatively stable planetary conditions in which civilizations have been shown to thrive are bounded by limits.Transgressing them is associated with significant risk of undermining many (if not all) of the SDGs.Third, the SDGs are interdependent.They interact with one another, either by constraining or enabling one another (Nilsson et al. 2018;Kroll, Warchold, and Pradhan 2019;Pradhan et al. 2017).Progressing on one SDG might therefore have positive or negative impacts on several others.

Impact assessment of economic sectors
There is little research connecting an examination of economic sectors with the knowledge of SDG interactions and spillover effects.A notable exception is the work of Van Zanten and van Tulder (2021), who map economic sector interactions with the SDGs using network theory.Based on this, they identify groups of sectors with similar SDG impacts and delineate strategic sustainability imperatives for each of these groups.
This can be valuable insights for companies in specific sectors or their owners aiming to shape future economies through active ownership.Some practitioners have started developing assessment methods that combine such sector-level with firm-level impacts, see Robeco´s SDG Framework 3 and Van Zanten et al. (2023).Although active ownership can be a useful tool for promoting sustainable corporate practices, particularly for industries that face reputational risks (Dimson, Karakaş, and Li 2015;Bauer, Derwall, and Tissen 2023), engagement at the firm-level is time and recourse intensive, and can have varying effect (Dimson, Karakaş, and Li 2015;Bauer, Derwall, and Tissen 2023;Sjöström 2020).Therefore, tools that can help guide higher resolution semi-automated investment decisions can be useful at initial portfolio development or assessment phases (e.g. by supporting broad comparisons across sectors), and can be well complemented by other more granular analysis at various investment phases to guide informed decisions.The evidence-based review approach of sectoral documented impacts on the SDGs and the traffic-light system for economic sectors we provide is a step in this direction.
The impact assessment undertaken in this paper formed part of a project commissioned by the Swedish corporate and investment bank Skandinaviska Enskilda Banken AB (SEB).Their aim was to develop a tool to assess SDG impacts of investment portfolios, as well as to allow for a broad comparison among sectors to help equity investors identify opportunities with positive impact.
Similar to Van Zanten andvan Tulder (2020, 2021) the approach rests on reviews of scientific scholarship combined with qualitative expert assessment of the literature documenting impacts on SDGs for each individual economic sector.Company reports were not considered due to their generally biased nature.Sectors were classified using the FactSet Revere Business and Industry Classification System (RBICS).The RBICS taxonomic structure consists of six levels and 14 overarching Economy sectors.Levels 1-3 of classifications are based on a market-defined approach, grouping companies by performance characteristics and correlated movements in stocks.For levels 4-6, categorizations are based on the services and products offered by the companies.Like Van Zanten and van Tulder (2020), the taxonomic levels chosen for this research vary between level 2 to level 6, depending on the sector, and were determined based on how well each level could be deemed to represent activities it comprised.Some sector categories in RBICS are relatively homogenous and level 2 is sufficiently well-defined to allow a search for academic assessments of impact from that sector category (e.g.'Healthcare Equipment').Other categories contain groupings of companies that are very different in terms of their potential impact on SDGs (e.g.'Business services' or 'Agriculture').These required a higher taxonomic resolution to meaningfully assess impact (see Supplemental Material section 1.1 for additional information).Furthermore, some specific subsectors (level 5) were specifically selected for inclusion by SEB because of their perceived role for sustainability transitions (e.g.'Renewable energy'), or ethical concerns (e.g.'Defense manufacturing').A total of 81 sectors were selected for review (see Table S1), covering 93.3% of all sub-industries (level 6) of the entire FactSet RBIC dataset.
Our ambition to assess the scientific evidence of sectoral impacts on each SDG (excluding #17) resulted in a total of 1296 SDG-RBICS code combinations.Our review process is based on what Grant and Booth (2009) refer to as a 'systematic search and review'.This approach combines a comprehensive search process with the evaluative element of a critical review to deliver a synthesis of the best available evidence.A literature search was done for each 1296 SDG-RBICS code combinations, with the aim of providing a representative overview of the scientific evidence.Our selection criteria for article inclusion were: (1) publication between 2000 and 2020, (2) relevance of scientific content (see Supplemental Material section 1.4 for specific search templates and for additional information on article inclusion selection criteria).
Non-peer-reviewed (grey) literature was only included when (1) our searches did not yield any peer-reviewed publications that fit the above criteria, but could be found in high-quality conference papers, working/discussion academic papers, textbook chapters, licentiate dissertations, or (2) a well-established institution (e.g.UN, Wolrd Health Organization, World Bank, Academic institution reports, EU publications) had published a review report specific to the SDG-sector nexus.A similar inclusion of grey literature was used by Van Zanten and van Tulder (2020).When grey literature was used, it was noted in the database.
We used Google scholar as the search engine for finding articles mainly because of the perceived value of transparency and accessibility to anyone (not just scholars at academic institutions).Tests and details preceding this decision are outlined in section 1.3 (Supplemental Material).We also developed templates to standardize the search procedure for each sector-SDG combination (see Supplemental Material section 1.4 and Table S4 for an example of the search terms used).All data collection and review were done during a ten-month period (April 2020-January 2021).
All articles identified using the search template were read.Sectoral impacts on an SDG documented in the readings were captured in the database.Information on economic sector impacts was included if the article; (1) articulated and supported a causal mechanism between an economic sector and an SDG (either an explicit mention of the SDG framework or a focus on the goal´s theme); (2) presented impacts of economic activities on an SDG at the macro level of economies, societies, or the environment, instead of the micro level of individuals, households or organizations (similarly to Van Zanten and van Tulder 2020); (3) presented effects of economic activities on an SDG that can generally be attributable to the economic sector in focus.We note that in many instances, specific sub-segments (or individual companies) of a sector may be developing improved practices, but unless these become so widespread that they can be viewed as representing the majority of the current sector, they have not been accounted for.If no reliable or scientific data was found on the mechanisms and impacts of sectors on the SDGs, this was captured in the database.
For many sector-SDG combinations, we identified several impact mechanisms.These were examined in aggregate through a qualitative expert assessment conducted by the six researchers who have authored this paper to classify the overall sectoral impact on a specific SDG according to a traffic-light system; where red signals that all impacts were generally negative, green signals that all impacts were generally positive, and yellow indicates that both positive and negative impacts were identified, or where evidence was not conclusive.An established methodology is not available for imposing a scientifically standardized way of assessing the scale and importance of all impacts.Yellow, therefore, indicates that the overall impact on the SDG is ambiguous and investors need to examine the impacts from these sectors more closely, and make a deliberate choice about investments in these sectors.Sector-SDG combinations where no documented impact mechanisms could be found were assigned a grey colour, and the impact was noted as 'Not Applicable'.

Causal loop diagrams
Causal loop diagrams (CLDs) are qualitative tools, used to apply a systems perspective, often in the context of social-ecological research (Meadows 2008;Biggs et al. 2021 and references therein).Although CLDs are simplified versions of how a system is understood, they are beneficial for understanding a system's characteristics, highlighting key interlinkages between multiple elements of the system, and capturing systemic impacts.A CLD is characterized as a set of variables, connected to each other by arrows (Meadows 2008).Each arrow represents an interaction between different elements of the system, such as positive (+) or negative (-) effects, i.e. an increase of the origin variable leads to an increase or decrease of the destination variable, respectively.When several nodes are mapped together, feedback loops emerge, and these can be reinforcing or balancing.In the CLD presented in the results section, each interaction is supported by scientific evidence (see Figure S1 and Table S5), following Downing et al. (2022) methodological approach.

Methodological limitations
Three limitations of our approach are worth highlighting.First, the impetus behind the creation of both industry classification codes and the SDGs introduce certain biases, as neither framework was designed for impact assessments.The dominant classification systems used by investment professionals (see also GICS and ICB) categorize sectors partly based on stock price movements, which is not ideal for classifying sectors based on their environmental and social impact, yet it provides a natural way to link impact assessments to the models used for portfolio allocation decisions.That said, it can be challenging to find a balance between a classification level that is homogenous enough to have a similar impact on the SDGs while at the same time maintaining a number of economic sectors that can realistically be assessed given practical limitations such as time constraints and availability of scientific studies on these sectors.The SDGs were designed to capture the dimensions on which humanity is not doing well and where progress is needed.As such, it is to be anticipated that additional positive impacts that corporate activities generate are not always well-captured.A further limitation of the SDGs as a framework against which to map impact is that high-resolution SDG indicators are sometimes conflicting, making the decision on the degree of impact of a sector on an SDG difficult to assess.This tension is supported by Kroll, Warchold, and Pradhan (2019) and Pradhan et al. (2017), who found negative correlations between internal indicators of SDGs 7-9, 13 and 15.We therefore refrain from structuring our assessment of impact on SDG indicators, and have instead searched for studies examining more general impact on, and interactions with, the SDGs.
A second limitation is that potential biases exist in academic literature.One is that the topics receiving research attention are highly dependent on what is interesting for academia and funding bodies for the advancement of science.Another bias is that some SDG impacts are well-established and 'obvious' and thus potentially less likely to be discussed in academic literature because no further research is needed.Third, certain sectors tend to be written about in a specific lighteither positive or negative.Temporality also plays a rolenewer sectors have had less time to accumulate research and thus the relevant literature may not yet exist.For these reasons, we were unable to yield results for some sectors and SDGs.In cases where an important consideration was missing in the literature, we included these impacts and mechanisms denoted with 'EJ' (expert judgement).The inclusion of such impact mechanisms resulted in 128 mechanisms (<3% of the entire database).
A third limitation is that when relying on assessment of impact across entire sectors using generic sector information (such as the scientific studies), instead of asset-level data, variation between individual companies of the same sector cannot be captured.A distinction between the best-and worst-in-class in a particular sector is therefore difficult.The kind of sectoral impact assessment approaches developed here will not allow investors to identify companies in need of significant changes or reward those leading at the front without additional asset-level analysis.Asset-level data can respond to this limitation by providing nuances, and consequently supporting the identification of firm behaviour that deviates from what is expected in the sector.Yet currently lack of standardized reporting and auditing of non-financial information, and consequent risk of biases in self-reported data present challenges to the use of company-level assessments as well.Despite the limitations mentioned above, a 'sectoral impact' approach captures broad impacts across sectors and provides a 'macro' view on impact (e.g.tobacco or soft drink production and retail will have minor variations on health impacts, even if some companies are selling a slightly healthier product than others), and can serve as an initial and important heuristic to guide screening and portfolio management.

Results
A total of 1382 publications were reviewed (89% peer-reviewed, 11% grey literature), and 4476 unique impact mechanisms were identified.For some sector-SDG combinations, multiple impact mechanisms were identified (e.g.'Metal Ore Mining' and SDG 15 on 'Life on Land').In contrast, for other interactions no impact mechanisms were identified in the assessed literature (e.g.'Home Improvement Retail' and SDG 10 on 'Reduced Inequalities').Using the traffic-light system classification, we conducted a set of comparative analyses.

Comparison by sector
The RBICS sectors with the most positive impact across all SDGs were 'Educational Services', 'Renewable Energy', and 'Telecommunications'.The ones with the most negative impacts across all SDGs were 'Tobacco Production', 'Tobacco Retail', and 'Metal Ore Mining'.The economic sectors with the most ambiguous impacts on the SDGs were 'Finance', 'Financially Operative Institutions' and 'Household Services'.For financial services, this ambiguity stems partly from the fact that financial sector impacts can be both positive and negative depending on the sectors and specific corporate activities capital is allocated to.For other sectors (such as 'Household Services' and most retail and serviceoriented sectors), insufficient evidence or scientific observations on impact were available to make an assessment.This may be because of the complexity of the sectors or due to a general lack of academic attention to tracing retail impact.The sectors 'Other Professional Services', 'Miscellaneous Retail' and 'General Merchandise Retail', generally lacked evidence of impact, most likely due to the broad definition of these sectors and a similar lack of attention from an impact tracing perspective.Figure 1 summarizes how each assessed economic sector impacts the 16 SDGs of the 2030 Agenda.

Comparison of impacts across SDGs
For all identified mechanisms, SDGs 1 (No Poverty), 9 (Industry, Innovation and Infrastructure) and 8 (Decent Work and Economic Growth) are those that are the most positively impacted by the assessed sectors.The SDGs with the most documented negative impacts are 13 (Climate Action), 15 (Life on Land) and 12 (Responsible Consumption and Production).SDG 8 (Decent Work and Economic Growth), 3 (Good Health and Well-being) and 5 (Gender Equality) had the most varied and ambiguous impacts (i.e. were impacted both positively and negatively), as illustrated in Figure 2. To exemplify how a sector results in ambiguous impact on an SDG, we present the results from the impact of the 'Telecommunications' sector on SDG 5 (Gender Equality).We found that this sector has generally positive impact by providing access via mobile phones for reproductive healthcare and women empowerment (Rotondi et al. 2020), as well as negative, by being a platform of online violence typically directed against women (Kerras et al. 2020).

A nested sustainability approach reveals uneven SDG impacts
Drawing on the 'wedding cake' metaphor introduced above (Folke et al. 2016), we ordered the 16 SDGs into clusters representing the 'environmental', 'social' and 'economic' sustainable development pillars, and assessed the aggregated impact from sectors on each pillar (Figure 3(A and B)).Although all three pillars are both positively and negatively impacted, the sectors assessed have a relatively stronger negative impact on the four SDGs comprising the environmental pillar.Over sixty percent of the impacts on the foundation of the 'wedding cake' are negative (Figure 3(B) bottom layer).Furthermore, if we disaggregate analysis by the three main economic sectors (primary, secondary, and tertiary), we see that primary economic activities result in both the highest negative and the smallest positive impact across all SDGs, when compared to secondary Figure 1.Summary of sectoral impacts on the 16 SDGs.All positive (green), negative (red), and ambiguous (yellow) impacts are aggregated per subsector, and the x-axis represents 100% of impacts on the 16 SDGs.As an example, 'Tobacco Production' has overarchingly negative impacts on 12 SDGs, positive impacts on 0 SDGs, and ambiguous impact on 1 SDG, according to reviewed literature.No documented impacts (grey) by tobacco production were found for 3 SDGs.and tertiary economic activities (Figure 3(C)).Across primary, secondary, and tertiary activities the environmental pillar is the most negatively impacted, and moving across Figure 3(C), from primary to tertiary sectors, the proportion of 'ambiguous' impact and lack of information generally increases.

Systemic impacts of corporate activities
A growing amount of research has shown that SDGs are interdependent (Nilsson et al. 2018;Kroll, Warchold, and Pradhan 2019;Pradhan et al. 2017;Le Blanc 2015;Van Zanten and van Tulder 2021;Biglari, Beiglary, and Arthanari 2022).Therefore, attempts Figure 2. Aggregated impact of 81 economic sectors on each SDG.All positive (green), negative (red), and ambiguous (yellow) impacts are aggregated per SGD, and the x-axis represents 100% of the 81 economic sectors.As an example, according to the assessed literature, SDG 13 is impacted negatively by 73% sectors, positively by 7% sectors, and ambiguously by 9%.For 11% of sectors, there is no evidence to assess how they impact SDG 13.
at understanding the impact of companies and their operations on the SDGs need to be cognizant that individual sectors are likely to have a suite of direct, but also indirect, effects on multiple SDGs (cf.Reyers and Selig 2020).We use the well-studied sector of agricultural crop production as a case to exemplify how progress in a specific SDG often triggers cascading effects and either reinforces or hinders progress on other SDGs.

A case study of industrial crop production
For the purpose of this exercise, crop production is assumed to be an export-oriented, large-scale and conventional production system, characterized by intensive and mechanized farming methods, and use of agrochemicals.Although we acknowledge that variations exist in agricultural practices, crop species, and geography, and that regulations impact sustainability dimensions differently, we opted for the type of crop production practices conducted by most listed agrifood companies to illustrate plausible generic impacts on SDGs.We excluded impacts from the production of fertilizers and other agrochemicals since we did not conduct a full life-cycle analysis.Biofuel crops were included in the assessment because these crops are sometimes used both for food and fuel.
In their network analysis of corporate activities and SDG impact, Van Zanten and van Tulder (2021) found that crop production had the highest out-degree centrality of all 67 economic sectors examined, indicating that crop production is generally both positively and negatively impacting other SDGs.
Building on our assessment of the crop production sector, we summarize some key generic evidence-based sectoral impact mechanisms to the SDGs through a CLD (Figure 4). Figure 4 illustrates both direct impacts from crop production on the SDGs as well as indirect impacts resulting from the interconnections of the SDGs (see Supplemental Material Table S5 for additional information on the evidence that supports each interaction).Therefore, the CLD captures systemic effects that originate from crop production and shows whether the interaction enables or constrains progress towards the SDGs.More specifically, our review finds evidence that crop production positively impacts poverty reduction by generating agricultural opportunities in rural areas, especially in developing countries (Feliciano 2019;Blesh et al. 2019).It can reduce hunger by increasing food security (Godfray et al. 2010;West et al. 2014;Friedrich and Kassam 2016); promote clean energy since crops and crop residues can be used for bioenergy production (Immerzeel et al. 2014;Hill 2009;Wilhelm et al. 2004); promote industry innovation and infrastructure if rural infrastructure investments are designed to support SDG 9 (Industry, innovation and infrastructure) (Llanto 2012), and can similarly support sustainable communities, e.g. when farms help reduce heat island effects (Pearson, Pearson, and Pearson 2011;Ackerman et al. 2014;Lovell 2010).However, the same sector also has notable negative impacts on health as current production practices often have many negative health impacts on agricultural workers (Udeigwe et al. 2015).Crop production is one of the most significant freshwater users affecting local hydrology (Chai et al. 2015;Udeigwe et al. 2015), and conventional industrial practices negatively affect multiple environmental SDGs (13,14,15).The latter include deforestation, soil and water contamination, GHG emissions, land use change and concomitant biodiversity decline, as well as over-use of fertilizers, to name a few (Udeigwe et al. 2015;Scialabba and Mller-Lindenlauf 2010;Goh 2012;Hill 2009;Baudron et al. 2009;Lal 2008;Mora and Sale 2011;Perfecto and Vandermeer 2008).We found ambiguous evidence on how crop production affects SDGs 4, 5, 8, 12 and 16.For example, it can both positively affect economic growth (Wheeler and Kay 2010) and have adverse impacts on the target of 'decent work' (both part of SDG 8) due to the exposure of farmers to harmful substances (Shannon et al. 2015).
Figure 4 thus illustrates evidence-based interconnections between SDGs and systemic ripple effects.In summary, crop production has both direct positive and negative impact on multiple SDGs.It also influences SDGs indirectly, due to their inherent interconnections (dashed lines, Figure 4).This means that the overall impact of an economic sector on the SDGs is highly complex.Every sector is likely to impact a particular SDG even if these connections are not currently established by scientific literature.As such, approaches that aim to assess the impact of economic sectors on the SDGs must be designed to incorporate this complexity to some degree.A generic Causal Loop Diagram on conventional crop production, illustrating direct impacts from crop production to the SDGs and indirect impacts resulting from SDG interconnections.SDGs are coloured according to which sustainable development pillar they contribute to.'+' indicates a positive relationship between two variables (an increase in one, increases or enables another), and a '-' indicates a negative relationship between two variables (a decrease in one, decreases or constraints another).Solid lines represent impact that emanates from crop production activities, dashed lines indicate second-order effects between SDGs or effects from crop production.The red-and blue-coloured arrows indicate two causal pathways elaborated on in the main text.For supporting evidence of each interaction and additional details, see Supplemental Material Table S5.
By tracing two causal pathways originating from crop production, we exemplify how the SDGs interact and result in spillovers.The first causal pathway (red-coloured arrows in Figure 4) departs from the fact that conventional crop production systems generally rely heavily on agrochemicals (e.g.fertilizers), which leach from soils and pollute freshwater (Lester and Boulton 2008).Agricultural runoff might lead to eutrophication of coastal areas or lakes and negatively impact aquatic biodiversity (SDG 14) (Fountain and Wratten 2013).Such biodiversity loss often negatively impacts the livelihoods of those dependent on fisheries or aquaculture for income (SDG 1) and food (SDG 2) (Worm et al. 2006;Brown et al. 2008).Low income, in turn, affects food security (SDG 2) (Webb 2014), which affects health (SDG 3) (Ruel and Alderman 2013), education (SDG 4) (Webb 2014), and unless addressed may lead to poverty traps (SDG 1) (Lade et al. 2017).Poverty traps are characterized by reinforcing feedback loops, such as the one shown in Figure 4.
The second example (indicated in blue-coloured arrows, Figure 4) refers to the high water consumption of conventional crop practices (Chai et al. 2015).When groundwater is reduced, the aquifer might get contaminated if it is located close to a coastal area, known as the 'sea intrusion' phenomenon (Maneas et al. 2019;Mazi, Koussis, and Destouni 2013).This constrains the ability of local people to access safe freshwater (SDG 6).Consumption of unsafe water has negative health implications (SDG 3) (Blomstedt et al. 2018).Poor health, in turn, is a constraint to education (SDG 4), reduces income, and increases the likelihood of poverty (SDG 1) (Webb 2014), and can also reinforce poverty traps.

Discussion
Our results indicate that the assessed economic sectors have a disproportionately negative impact on the environmental pillar, when compared to the social and economic sustainable development pillars (Figure 3(B)).Moreover, our data illustrate that primary economic activities result in the highest negative impact across all SDGs, when compared to secondary and tertiary economic activities.In contrast, tertiary economic activities appear to have the lowest negative impact on the environmental SDGs, compared to primary and secondary economic activities (Figure 3(C)).
However, our findings are a consequence of how we, and most other impact assessment developers in the financial sector, measure the impact of different types of sectors (Popescu, Hitaj, and Benetto 2021).In assessing secondary and tertiary industries we accounted for downstream, but not upstream, impacts and, as a result, assessed the impact of each RBICS sector without considering the full range of inputs into this sector.This was an explicit choice made by SEB.For secondary and tertiary sectors, we assessed both impacts associated with operations of the specific value chain segment in focus and forward-looking impacts, i.e. end-use/consumption of the product or service.Consequently, the food retail sector appears to impact SDG 15 only through shop location, operation, and consumption of food sold.As a result, the embedded biodiversity footprint of sold food was not taken into account.Measuring impact without considering upstream impacts is not viable, because it appears as if secondary and tertiary sectors are not dependent on materials and energy, i.e. they appear as decoupled from the primary sectors of the economy.Capturing upstream supply chain impacts is therefore essential in future development of portfolio impact assessment tools (Popescu, Hitaj, and Benetto 2021 and references therein).
However, there are three main obstacles when attempting to account for upstream impacts.First, most secondary and tertiary sectors are characterized by high complexity (i.e. they depend on inputs from multiple primary sectors) and supply chains generally suffer from low traceability and transparency (Schäfer 2022;Gardner et al. 2019;Boström et al. 2015;Wognum et al. 2011).Second, currently there are barriers to accurately assessing supply chain impacts in the form of life-cycle assessments (LCAs) due to constraints in data availability (Wognum et al. 2011).Third, although LCA approaches are one of the best available tools to date, they too have drawbacks, since LCA output values are highly sensitive to model assumptions and input data, and therefore not necessarily applicable beyond the specific commodity analysed in each LCA study (Hellweg and Milà 2014;Reap et al. 2008;Guo and Murphy 2012;Crenna et al. 2020).

Addressing spillovers and capturing the impact shadow of non-primary sectors
Our analysis points to three considerations for better conceptualizing and supporting the generation of impact data for sustainable investment objectives.
First, considering the position of investee companies in the supply chain and taking a supply network approach is essential to capture a fuller, and more accurate, picture of impacts (Figure 5(A)).As an example, investors wanting to understand the societal impacts of allocating funds to companies in the food processing sector need to not just examine the immediate impacts of the processing segment, but also understand the impacts associated with the primary agrifood sectors supplying the raw materials.In this paper, we refer to this as understanding one's 'impact shadow'.It is similar to the notion of scope-3 impacts used in corporate GHG protocols but encompasses more categories of impacts, such as those outlined by the SDGs, and we therefore use it to differentiate between the narrowly GHG focused scope 3, and to avoid confusion.The notion of an 'impact shadow' falling from one (particularly primary) sector across downstream segments is thus merely a heuristic to make explicit the need to include impacts incurred by upstream industries in the impact evaluations of downstream industries.These impacts have the greatest detrimental effects on the four SDGs underpinning the SDG wedding cake and are not the burden simply of primary industries, but of all economic sectors that demand the raw materials.Hence, we let the light shine squarely on these sectors, in Figure 5(A).
The importance of tracing impact along supply chains is increasingly recognized for environmental inputs such as biodiversity impacts and water (Lähtinen et al. 2016;Wilting and van Oorschot 2017;Wang et al. 2020) but the added insights we provide is the need for investors to also apply a systemic approach to impact, as many of both environmental and social impacts are interconnected (Popescu, Hitaj, and Benetto 2021;Downing et al. 2022).
Second, it is essential that impact assessment methodologies for sustainable investing reflect that the environmental pillar supports all other sustainable development pillars, given that humans are embedded inand dependent onthe biosphere (Dasgupta Review;IPBES 2019;Folke et al. 2016).Our analysis shows that the most consistent and negative impacts resulting from economic activities are to the environmental SDGs (Figure 3).Failing to meet the environmental SDGs adds pressure on the biogeo-physical processes that support the relatively stable earth system (Rockström et al. 2009;Steffen et al. 2015) which is very alarming, since environmental deterioration undermines all other SDGs.We exemplify the challenge of addressing the pressures on the climate and living biosphere by presenting the impacts of an active SDGaligned portfolio (see Table 1A).
Third, it is necessary that investors embrace the complexity and high interconnectivity between the SDGs, which result in cascading, systemic impacts and trade-offs.Every investment will result in trade-offs, even when designed with sustainability as an objective (see Table 1B).Hence it is important to anticipate them and not to focus only on positive impacts.Despite the extensive academic literature on SDG interactions (Nilsson et al. 2018;Kroll, Warchold, and Pradhan 2019;Pradhan et al. 2017;Le Blanc 2015), when the financial industry operationalizes the SDGs for impact assessments, interactions are generally not discussed or captured by assessment tools.This indicates a gap between research and practice and signals an important area for future developments and science-industry collaborations.
We acknowledge that a major concern for all investors is the issue of (short-, or, as in the case of sustainable investing, long-term) returns.However, lack of consideration of portfolio-induced impacts is now increasingly recognized as a significant and increasing risk, especially by large institutional investors (Principles for Responsible Investment -PRI).By virtue of essentially owning a slice of the real economy, these investors, often termed Universal Owners (Hawley and Williams 2000;Lukomnik and Hawley 2021;Quigley 2019), will be directly affected by systemic risks that are already materializing and affecting entire regions and markets (Svartzman et al. 2021;Quigley 2019).However, improved impact accounting and understanding is not relevant only to large mutual and pensions funds.Credit portfolios will also feel the effects of a changing climate and turbulent social and economic dynamics resulting from resource scarcity (Quigley 2019;Wu et al. 2023), and for asset managers it will become increasingly hard to ignore the fact that environmentally induced systemic risks (i.e.'beta') are shown to increasingly outstrip any excess return ('alpha') (Lukomnik and Hawley 2021;Ibbotson 2010;Hensel, Ezra, and Ilkiw 1991).
Regulations such as the EU Taxonomy, CSRD and Sustainable Finance Disclosure Regulation (SFDR), all create incentives and pressures for investors to reflect on activities causing 'significant harm'.Such regulations have the power to affect norms and practices within the financial sector (Ahlström and Monciardini 2022), particularly in conjunction with various international voluntary initiatives that forge pre-competitive alliances and promote learning and leadership among a wide set of financial actors, such as the Climate Action 100 + on climate, the UN PRI, and Finance for Biodiversity, to name only a few.Marti et al. (2023) refers to these norm-shaping practices as promising 'field building' efforts, and Bauer, Derwall, and Tissen (2023) find that when likeminded investors exert pressure to investee companies, their engagements are more likely to be effective.It is our hope that this paper can provide inspiration and concrete examples to those financial sector actors who develop impact methodologies, manage equities and fixed income investments, or develop impact related policy.This paper emphasizes the value of using the best available science when attempting such impact assessments, and when designing impact-assessment methodologies (cf.Popescu, Hitaj, and Benetto 2021).
We recognize that the analysis in focus here is tilted towards the practices of equity investors.While impact assessment is a valid, and increasingly necessary, tool for most  .The information presented exhibits the limited positive impact concerning the environmental SDGs, even in SDGaligned portfolios.First, two (SDG 14 and SDG 15) of the four environmental SDGs are not featured, meaning that the investment did not have positive impacts on these SDGs.Second, the other two environmental SDGs (SDG 13 and SDG 6) were impacted, but by inadequately addressing the actual pressures on the biosphere.More precisely, the positive impact on SDG 13 was depicted as 'Providing 10 clients with natural catastrophe insurance (13.1)'.Natural catastrophe insurance mainly supports adaptation and does not aid mitigation efforts by reducing carbon.Third, SDG 6 is mentioned, but the investment resulted in improving water access and treating wastewater, not actively restoring freshwater ecosystems.This example demonstrates the enormous challenge of aligning with, and progressing on, the environmental SDGs when investors manage SDGfocused portfolios.
An example that illustrates trade-offs between SDGs due to investments is the analysis of biodiversity impacts of the largest European listed companies assessed by the Groupe Caisse des Dépôts' Asset Management Division.In a webinar in 2021, their analysis showed that a lowcarbon portfolio had a biodiversity footprint over 25% larger than the benchmark portfolio (CDC, Biodiversité 2021).This shows inherent trade-offs between environmental sustainability objectives, such as climate and biodiversity (as reflected in SDG 13 and 15 and illustrated in Figure 5(B)).
investors (Grewal and Serafeim 2020), a full review of the processes by which different types of investors (ranging from credit providers to venture capital equity) can incorporate the use of impact assessment is beyond the scope of this paper.However, scholars in the growing academic nexus between sustainability science and investment scholarship are embarking on precisely such mapping.

Figure 3 .
Figure3.Impact of all assessed economic sectors on the three sustainable development pillars.Panel A: Each box represents an assessed SDG-RBICS code combination (1296 in total) and is coloured according to the assessed aggregate impact of the sector on the SDG (see Figure1).SDGs are clustered to correspond to the three pillars of sustainability (environment, social and economic).Clusters refer to the environmental SDGs (6, 13-15, bottom layer); the social SDGs (1-5, 7, 11, 16, middle layer); the economic SDGs (8-10, 12, top layer).Panel B: The data from Panel A represented as the 'wedding cake' to visualize the pattern of economic sector impacts across the nested sustainability pillars.Panel C: Representation of primary, secondary, and tertiary sectors separately.Primary sector refers to natural resource extraction sectors (e.g.forestry) and includes 10 economic activities (which cover 13% of the entire FactSet L6), Secondary sector refers to sectors that process natural resources (e.g.manufacturing) (33 sectors, 37% coverage of FactSet L6), Tertiary sector refers to all sectors that are neither primary nor secondary (e.g.service sectors) (38 sectors, 44% coverage of FactSet L6).

Figure 4 .
Figure 4.A generic Causal Loop Diagram on conventional crop production, illustrating direct impacts from crop production to the SDGs and indirect impacts resulting from SDG interconnections.SDGs are coloured according to which sustainable development pillar they contribute to.'+' indicates a positive relationship between two variables (an increase in one, increases or enables another), and a '-' indicates a negative relationship between two variables (a decrease in one, decreases or constraints another).Solid lines represent impact that emanates from crop production activities, dashed lines indicate second-order effects between SDGs or effects from crop production.The red-and blue-coloured arrows indicate two causal pathways elaborated on in the main text.For supporting evidence of each interaction and additional details, see Supplemental Material TableS5.

Figure 5 .
Figure 5.The hierarchical nature of SDGs necessitate a systemic conceptualization of supply chain impacts.Panel A: Impact shadow cast by a generic supply chain, where primary industries often incur a significant portion of the environmental impacts for production of raw materials used by downstream segments; Panel B: Spillovers resulting from SDG interactions illustrated in a hierarchical SDG format.Figure 5B is modified from Folke et al. 2016 (CC BY-NC 4.0) by inserting the arrow and '-' symbol and by reducing the figure's opacity.The original version of this figure was first presented by Rockström and Sukhdev 2016.