Adjusting the social footprint methodology based on findings of subjective wellbeing research

Following some years of practical application, some weaknesses have been identified in the original 2018 version of the ‘social footprint’ methodology, where wellbeing was seen as exclusively related to consumption activities and as inseparably linked to production through the budget constraint, implying that the value of wellbeing was limited to be a mirror of the value of production. Several improvements in both methodology and data are presented here. The theoretical improvements are inspired by the suggestion of Juster et al. (Rev Income Wealth 27: 1–31, 1981) that wellbeing can be seen as the sum of the value added generated from work and the intrinsic activity benefits, i.e. the positive affect from performing or taking part in specific work or leisure activities. This implies a relatively low preference for income relative to intrinsic activity benefits, which is confirmed by recent findings of subjective wellbeing research. Other findings of subjective wellbeing research provide a constraint on the conversion factor between Disability-Adjusted Life-Years (DALY) and Quality-Adjusted person-Life-Years (QALY), leading to a surprising 0.3 QALY/DALY, against the more intuitive 1 QALY/DALY. These theoretical improvements, combined with the availability of more recent country-specific data on impacts on wellbeing, allow to calculate a global potential level of wellbeing of 0.958 QALY/person-life-year, replacing the global potential productivity of the 2018 version of the ‘social footprint’ methodology. The new country-specific data allows the valuation of impacts on wellbeing to be assessed separately from the valuation of inequality, the latter now done with equity weights relative to country-specific average income baselines, rather than to the global baseline used in the 2018 version. The new data confirm the dominating role of impacts of missing governance, now quantified at 78% of all sustainability impacts, which was the original motivation and rationale behind the 2018 version of the ‘social footprint’ methodology.


Introduction: the social footprint methodology
The social footprint methodology is a streamlined approach to provide comprehensive aggregated 'life cycle sustainability assessment' (LCSA) results, combining input-output data on value added and work-hours with an impact assessment focusing on the macro-scale impacts of the non-production-specific impacts, i.e. impacts unrelated to enterprise-specific actions and choice of technology, and therefore quantifiable from national statistics without the need to access detailed technology-or enterprisespecific data . Quality-Adjusted person-Life-Year (QALY) is suggested as a unit for 'sustainable wellbeing' (Weidema 2006, in parallel to the Disability-Adjusted Life-Year (DALY) measure used in comparisons of health impacts (although QALY and DALY will have opposite signs, due to DALYs expressing a detriment, while QALYs express wellbeing benefits), while the latter would then also include the nonhealth aspects of wellbeing.
Published examples of the application of the social footprint method can be found for milk in Pakistan and Communicated by Marzia Traverso.
The social footprint methodology, as suggested in , applies equity weights (also known as utility-weights or distributional weights) to changes in both income and in wellbeing, leading to an emphasis on changes that affect the poor. Although this may be seen as desirable, it implies an unfortunate dependency between the valuation of wellbeing and the valuation of inequality, and calculations on extreme cases have been found to yield implausible results, where wellbeing after the change exceeds 1 QALY/person-year. This paper suggests addressing these shortcomings by adjusting the social footprint methodology on the basis of the findings of subjective wellbeing research and improved data sources for the impacts that cause loss of wellbeing at the country level.
After these adjustments, the social footprint maintains its focus on impacts of missing governance and its core feature of being quantifiable from national statistics without the need to access detailed technology-or enterprise-specific data and with the option of additional quantification of credits for potential positive actions.

Subjective wellbeing research
Subjective wellbeing (SWB) research has advanced significantly over the last decades. Subjective wellbeing can be measured both as more general cognitive self-evaluation of life satisfaction and as self-evaluations of positive and negative moment-to-moment affect in relation to current activities (also known as 'experienced' or 'hedonic' wellbeing). For recent reviews of the scientific field, see Diener et al. (2013) and OECD (2013. Measures of positive affect appear to be well-correlated to the more general life satisfaction, while negative affect does not appear to have as lasting an influence (Helliwell et al. 2020b).
A well-proven measure of cognitive self-evaluation of life satisfaction on a scale from 0 to 10 is provided by the answers to the Cantril (1965) ladder question: 'Please imagine a ladder with steps numbered from 0 at the bottom to 10 at the top. The top of the ladder represents the best possible life for you and the bottom of the ladder represents the worst possible life for you. On which step of the ladder would you say you personally feel you stand at this time?' Cantril points seen as averages over a year can thus be converted to QALY by dividing by 10, and vice versa. However, subjective wellbeing only considers the current level of wellbeing. To capture also impacts that affect life expectancy, QALYs derived from Cantril scores must be adjusted for lost life expectancy.

Findings of SWB with relevance for the social footprint methodology
The theoretical improvements to the social footprint methodology presented in this article are inspired by the suggestion of Juster et al. (1981) that SWB can be seen as the sum of the instrumental value of income (equal to the value added generated from work) and the intrinsic activity benefits, i.e. the positive affect from performing or taking part in specific activities (named 'process benefits' by Juster et al.). This perspective emphasises the importance of intrinsic activity benefits relative to the instrumental value of income that played a central role in the Weidema (2018) formulation of the social footprint. On the data side, an important basis for extending and validating the social footprint results at country level is provided by the annual 'World Happiness Reports' (Helliwell et al. 2020a), which are based on the annual survey results from the application of the Cantril ladder question by the Gallup World Poll. In this context, it is particularly noteworthy that Helliwell et al. (2020a) also seek to explain the variation of the Cantril scores across countries by six explanatory variables: income (the natural logarithm of gross national income), healthy life expectancy at birth (years), social support (boolean), freedom to make choices (boolean), generosity (donation to charity during past month; boolean), and perceptions of corruption (0 to 1). The lowest national average score for each variable provides a benchmark (called 'Dystopia' by Helliwell et al. 2020a) at a Cantril ladder score of 1.97, against which the contributions of the six explanatory variables can be measured. For the difference in life satisfaction scores between the top 10 countries and the bottom 10 countries, Helliwell et al. (2020a, Statistical Annex, Table 21) find that the six variables explain 71%. Applied to the full range from Finland on top with a Cantril score of 7.809 to Afghanistan at the bottom with a Cantril score of 2.567, income explains 23% (1.19 Cantril point), social support 19% (1.00 point), health 15% (0.78 point), freedom of choice 7% (0.34 point), perceptions of corruption 6% (0.32 point), and generosity 2% (0.11 point) of the difference in life satisfaction scores, leaving 1.5 Cantril points (29%) unexplained (values may not add up to totals due to rounding).

On the value of QALYs and DALYs
The finding of Helliwell et al. (2020a) that differences in health only explain 15% of the difference in wellbeing could be applied to the 0.52 QALY/person-year difference between current and potential wellbeing from Table 1, implying that health impacts should only make up around 1 3 15%*0.52 = 0.078 QALY/person-year on average, or 0.6 billion QALY when applied to the global population of 7.7 billion. The global burden of disease according to IHME (2021) is 2.5 billion DALY, which would imply a conversion factor of 0.6/2.5 = 0.24 QALY/DALY. Considering that only a part of the global burden of disease is avoidable (Weidema and Fantke 2018) moves this factor slightly upwards to around 0.3 QALY/DALY. It is nevertheless a surprising finding that there is such large a difference between the two concepts that may intuitively be thought to be congruent. By combining the country-level data on avoidable health from IHME (2021) and the subjective wellbeing data from Helliwell et al. (2020a), we confirmed that for many countries, an assumption of 1 QALY/DALY would result in implausible counterfactual levels of wellbeing, exceeding 1 QALY per person-life-year.
The finding of Helliwell et al. (2020a) that log income explains 23% of the difference in average wellbeing could be used a basis for converting the contribution of income in monetary units to units of wellbeing. Juster et al. (1981) suggest that SWB can be expressed as the sum of the instrumental value of income (equal to the value added generated from work) and the intrinsic activity benefit, i.e. the positive affect from performing or taking part in specific activities. In the illustration in Fig. 1, the four fully drawn rectangles illustrate four types of contributions to the value of a QALY. Additionally, an unavoidable loss of wellbeing is added at the top, illustrating that it can never be possible to achieve full wellbeing for all people all the time, due to unavoidable life events, such as natural disasters, deaths of close relatives and friends, and unavoidable diseases.
The lower rectangles in Fig. 1 express the current level of wellbeing, which is estimated by Helliwell et al. (2021)  The internal vertical lines in Fig. 1 separate both current and potential wellbeing into value added (left rectangles of Fig. 1), i.e. the value that work (productive activities) adds to products, and the activity benefits (right rectangles of Fig. 1), i.e. the value of the positive emotions that people obtain from performing or taking part in specific activities. The dotted line separates the activity benefits experienced during work (i.e. beyond the value of the work outputs) from those experienced from leisure activities. The size of these different parts of wellbeing is estimated to be approximately 25% for the value added, 25% for the activity benefits from work, and 50% for the activity benefits from leisure. This distribution is estimated on the one hand from the trade-off between leisure and work that implies that the benefits of a marginal hour of work equals the benefits of a marginal hour of pure leisure and on the other hand from the finding of Helliwell et al. (2020a) that income only explains approximately 25% of the difference in life satisfaction scores. The implication is that total subjective wellbeing (Q) has a value approximately four times the total value added (V A ) of production: thus providing a basis for expressing marginal subjective wellbeing (QALYs) in monetary units, and vice versa.
The approximate relationships in Fig. 1 can also be confirmed from the episodic (moment-to-moment) affect data collected by Krueger (2007) and Gershuny (2013). A useful insight from these data is that positive affect experienced from pure leisure exceeds the positive affect from work activities with a value around 0.25 on a 0-1 scale (after conversion from Krueger's 0-6 scale and the 0-10 scale used by Gershuny).
The above findings from SWB research provide an important correction to Weidema (2009), where wellbeing was seen as exclusively related to consumption activities and as inseparably linked to production through the budget constraint, implying that the value of wellbeing was limited to be a mirror of the value of production, actually only describing the consumption and production aspects of instrumental benefits (value added), respectively. Furthermore, the focus on the budget constraint led to the counter-intuitive result that the same percentage-wise change in instrumental productivity benefits and intrinsic activity benefits would provide the same change in utility (wellbeing). This would mean that e.g. a 10% reduction in wellbeing could potentially be compensated by a 10% increase in income. In , it was suggested to solve this problem by applying a different basis for calculation of equity weights for changes in wellbeing versus changes in productivity and consumption only (i.e. without any concurrent change in wellbeing), resulting in a ratio of a QALY to value added of However, this solution is blemished by the fact that the ratio is dependent on the number of income steps used in its determination (the value 2.36 reflects a normative choice of three income steps between zero and the current income level) and also maintains the fixed relationship-although no longer 1:1-between intrinsic activity benefits and instrumental productivity benefits; i.e. it implies an assumption that a change in intrinsic activity benefits will always be accompanied by a change in productivity (Weidema 2009). More fundamentally, it implies an unfortunate dependency between the valuation of intrinsic activity benefits and the valuation of inequality. All of these points of critique do not apply to the more comprehensive expression for the value of a QALY, which uses the finding of Helliwell et al. (2020a) that income only explains approximately 25% of the difference in life satisfaction scores. Note that the value of a QALY is quite sensitive to changes in the explanatory variable of preference for income (ratio of value added to total wellbeing). A change from 0.25 to 0.23 would increase the conversion factor from 1/0.25 = 4.00 to 1/0.23 = 4.35 times the value added of production. Previous studies have provided both even lower values for the explanatory variable, giving implausibly high values for non-income effects (Fujiwara and Campbell 2011), and much higher values (e.g. Fritjers et al. 2004, Sacks et al. 2010, which are similarly implausible because they would imply a larger difference in activity benefit between leisure and work activities than what is empirically observed from positive affect studies (Krueger 2007;Gershuny 2013). So, while the preferences for income, work benefits, and pure leisure benefits may change independently over time and between persons, the interdependence of the relationships expressed in Fig. 1 provides reason to expect that, at the population level, average changes in relative preferences will be moderate.
On the basis of data from the European Social Survey and the Gallup World Poll, as analysed by Helliwell et al. (2020b), inequality of wellbeing is found to explain a difference of 0.  in wellbeing between countries and to provide a more consistent explanation for the average levels of SWB than inequality in income. This is explained by changes in social connections and the quality of social institutions being of larger importance as buffers against wellbeing impacts for people with the lowest levels of wellbeing, so that happiness increases more for those, thereby reducing inequality. Table 1 shows how the total value of the current impacts on SWB is constrained by the theoretical limit of 1 QALY per person-life-year.

Results: the consequences for the social footprint methodology
The findings from SWB research described in the previous section implies that: • The social footprint should be calculated from the countryspecific impacts on wellbeing instead of solely from the country-specific and global productivity impacts. • The use of equity weighting should be limited to function as a distribution key for overall SWB impacts within the constraints of the findings from SWB research, such as the relatively low preference for income relative to intrinsic activity benefits indicated in Fig. 1 and the relatively low preference for health improvements relative to overall wellbeing improvements.
The original social footprint methodology ) is a streamlined methodology, in that it allows quantification of impacts based exclusively on: • Value added per skill-level, industry, and country • Workhours per skill-level, industry, and country • A country-specific correction factor for purchasing power (PP) applied to the value added • An estimated global potential productivity (value added per workhour) • An estimated global elasticity of marginal utility of income In the adjusted methodology, a number of additional data inputs are applied: • Using the value of 0.958 QALY/person-life-year from Table 1 as the global potential level of wellbeing • Using the country-specific Cantril scores from the World Happiness Report (divided by 10 to convert to the 0-1 QALY scale, and modified by lost years of life expectancy) as the country-specific levels of current wellbeing • Using the country-specific proportion between nonproduction-specific wellbeing impacts and wellbeing impacts that are attributable to specific activities, based on Weidema (2022a) (see Annex 1) • Using a conversion factor of 1 QALY per 388′000 USD 2019 based on the updated global potential productiv-ity of 97′000 USD 2019 per person-life-year from Weidema (2022b) and the Q = 4 * V A from Eq. 1.
The calculation procedures to derive the social footprint of a specific activity are shown in Table 2, indicating the differences between the original and the adjusted methodology. The main difference is that the starting point (see step 1 in Table 2) is now the full non-production-specific wellbeing impacts (rather than only productivity impacts) by country, as shown in Table 4 in Annex 1. In step 2, the key for distribution of the impact over industries has been changed from value added to direct income from industries (compensation of employees + operating surplus), thus excluding taxes and subsidies on products and production. This change is done to avoid skewed results for industries that receive substantial subsidies, given the current lack of adequate data for redistribution of taxes and subsidies between income groups. In steps 3 to 5, the main change is the use of country-specific rather than global equity weights and that these are now used only as distribution keys for the impacts, so that the total impacts measured in QALY are the same before and after equity weighting.
The adjusted method maintains the core idea from the original social footprint method ) that local enterprises have a co-responsibility for non-productionspecific governance impacts because they benefit from the current low internal costs of labour. These impacts should thus be distributed over the industries in proportion to their responsibility. A simple distribution relative to the value added of the industries, i.e. mimicking a value added tax, would punish industries that actually do pay a fair wage. The social footprint method therefore makes the distribution in a Table 2 Differences (in italics) between the original and the adjusted calculation procedures to derive the social footprint Original social footprint methodology  The adjusted social footprint methodology 1. Calculating the country-specific productivity impact as difference between the potential productivity and the country-specific productivity after subtraction of the global productivity impacts that are attributable to specific activities 1. Calculating the country-specific non-production-specific wellbeing impact as difference between the potential and the country-specific level of wellbeing after subtraction of the impacts that are attributable to specific activities 2. Calculating the industry-specific productivity impact as the share of the country-specific productivity impact proportional to the industry's share in the country-specific value added 2. Calculating the industry-specific share of the non-production-specific wellbeing impact as the share of the country-specific impact proportional to the industry's share in the country-specific direct income from industries (compensation of employees + operating surplus) 3. Calculating the group-specific productivity ratio of the population group affected by the change as the ratio of the global average productivity to the population-group-specific productivity 3. Calculating the group-specific wellbeing ratio of the population group affected by the change as the ratio the country-specific average productivity to the population-group specific productivity 4. Equity weighting the industry-specific productivity impact by multiplying the impact by the group-specific productivity ratios raised to the power of the global elasticity of marginal utility of income 4. Equity weighting the industry-specific share of the non-productionspecific wellbeing impact by multiplying the impact by the groupspecific wellbeing ratios raised to the power of the global elasticity of marginal utility of income and rescaling to maintain the countryspecific totals before equity weighting unaffected 5. Finally deriving the equity-weighted social footprint by subtracting the likewise equity-weighted value added (representing the impact of monetary re-distribution) 5. Finally deriving the equity-weighted social footprint by subtracting the likewise equity-weighted and rescaled QALY-equivalent of value added (representing the wellbeing impact of monetary re-distribution) somewhat more complex way, including the differences in wage levels between and within industries, with the intention to give more weight to those industries that have low-paid employees. From data on wages and numbers of work-hours, provided per industry and skill level in databases such as EXIOBASE (Stadler et al. 2018), the wage/work-hour can be calculated. The wage data make up only part of value added. However, in the absence of detailed data on tax redistribution, taxes and subsidies on products and production are ignored, and the assumption is made that the operating surplus will end up with the same population groups as the wages, i.e. proportionally to the wages. With this assumption, the direct sources of income (compensation of employees + operating surplus) per country are now divided over industries and divided over wage groups (income groups) for each country. In the resulting direct income matrix for each country, each cell DI i,g where subscript i indicates industry and subscript g indicates income group, is then weighted with an equity weight EW, where WH is work-hours and subscript c is country): The justification for the specific equity weighting is provided in . However, in Weidema (2018), the world average wage level is used for the equity weighting, while here, the country average wage level is used, as justified above.
Finally, DI i,g * EW i,g is used as distribution key for the non-production-specific QALYs for each country (from Table 4) to the cells in the matrix and thus to each industry. The resulting vector is a vector of 'non-production-specific impacts' per industry that does not need any further characterisation of weighting because it is already expressed in QALY and may be directly applied for the calculation of footprints, using the normal life cycle assessment calculation routines (Heijungs and Suh 2002).
An example of the above calculation for implementation in the year 2011 hybrid version of EXIOBASE can be found in Weidema (2022a).

Discussion and conclusion
Findings from SWB research have been applied here to the social footprint method, reducing the previous focus on productivity, removing the dependency between the valuation of wellbeing and the valuation of inequality, and increasing the emphasis on the intrinsic activity benefits. Besides these improvements in the justifications and theoretical basis for the social footprint method, the described changes to the implementation of equity weighting ensures that results cannot exceed the plausibility threshold for wellbeing of 1 QALY/person-year.
Besides these improvements, new calculations validate the findings from the original 2018 version of the social footprint methodology that a large share of the total impacts on sustainability is non-production-specific, i.e. unrelated to enterprise-specific actions and choice of technology. As shown in Table 4 in Annex 1, non-production-specific impacts make up 78% of the total impacts at the global level. This percentage varies from 87.2% of all impacts in Syria to 26.5% of all impacts in Bahrain. As shown in Table 3 in Annex 1, the largest shares of non-production-specific impacts are participation restrictions, discrimination, inequality, and insufficient development of skills (education and learning), but also issues such as impacts from armed conflicts and infectious diseases are included. The nonproduction-specific impacts on wellbeing can be further decomposed using country-specific data for the more specific impacts from Weidema (2022a), replacing the decomposition provided in Weidema (2017).
The co-responsibility for local governance impacts should not be seen as an argument for avoiding to locate enterprises in countries with missing governance . The calculations described above provide a baseline of coresponsibility, applicable to an average enterprise that takes no actions to improve the local conditions. However, the same calculation procedures can be applied to an improvement, e.g. in payment of wages or taxes, in working conditions, or in support to local communities. Because of the larger potential for relative improvements in a country with a low initial level of wellbeing and the amplifying effect of the equity weighting, the potential credits for specific positive actions provide a compelling argument for placing activities in countries with missing governance, provided that the enterprise is allowed to follow an active strategy to create shared value.

Appendix 1. Non-production-specific impacts vs. impacts attributable to specific activities
As a streamlined approach, the social footprint methodology initially focuses on the macro-scale impacts of non-production-specific impacts, i.e. impacts unrelated to enterprisespecific actions and choice of technology . Therefore, it is important to separate these non-productionspecific impacts from the impacts that are attributable to specific activities, especially to avoid double-counting when the latter are treated separately in the impact assessment. This annex describes the division of the overall wellbeing impacts into these two groups. Table 3 provides an overview of the size of the global annual impacts for year 2019, expressed in million QALY, following Section 5 in Weidema   Table 4 provides the shares of non-production specific impacts per country and the absolute numbers in million QALY for the 64 countries with largest contributions to the global total.
Obviously, the 64 countries in Table 4 include mainly developing countries and a few large developed economies, while the smaller and richer countries with smaller contributions are part of the 'Rest of World'