Quantification of corruption in preventative cost-based S-LCA: a contribution to the Oiconomy project

The UNEP guidelines for Social Life Cycle Assessment (S-LCA) of products mention corruption as a subcategory (UNEP/SETAC 2009, p. 49). But to date, the literature on addressing corruption in LCA is very limited. Some authors, usually citing the guidelines, mention corruption as an aspect to be studied, e.g. (Grießhammer et al. (2006), Benoît et al. (2010). Dreyer et al. (2010), describing a method for labor conditions, mentions its applicability for corruption. Ekener-Petersen and Finnveden (2013) identify corruption as a potential hotspot in the life cycle of a laptop computer. No attempt at further quantification of the specific aspect of corruption or research into potential indicators for that purpose could be found in the literature. We may conclude that there is a research gap in the quantification of corruption in S-LCA.

Life Cycle Attribute Assessment (LCAA) was developed by Norris (2006) and further elaborated by Andrews et al. (2009). The technique determines the share of an attribute, such as compliance to a certificate or “child labor-free,” in the supply chain of a product. It is a technique in two steps: 1. Determination of yes or no presence of the attribute and 2: Weighting of the presence by an activity factor, for which Norris takes workhours, as does Hunkeler (2006), but also other factors can be taken (Andrews et al. 2009). The results of LCAA are conclusions such as “the percentage of labor hours that is child labor-free.” However, where Hunkeler only uses country level background data, LCAA is based on both foreground, e.g. by measuring compliance percentages to certificates, and background data, e.g. from Input-Output databases. LCAA uses the simplification of a dichotomous variable per step in a product’s life cycle, inherently normalizes by measuring data as a percentage, and equally weights aspects and steps in the life cycle. Therefore, LCAA gains certainty, especially if combined with certification, but loses accuracy because of lack of gradation. A disadvantage is that the activity factor limits the applicability to aspects that are closely related to the activity factor and therefore also limits the aggregability over different categories.

Chhipi-Shrestha et al. (2015) recognize two types of quantitative Type-II models: 1. The “empirical model” and 2. the “ELCI-database” model. The empirical model constructs an empirical cause-effect pathway based on correlations between established (usually country-level) indicators. The ELCI-database model tries to follow the standard environmental-LCA (ELCA) procedure from life cycle inventory to characterization and sometimes even uses the LCI databases developed for ELCA.

Considering the empirical model, Norris (2006, p. 99) described a pathway from per capita income to health in the mathematical formula that describes the “Preston curve” relation between income and life expectancy, using World Bank country-level data, but saw too many shortcomings. Feschet et al. (2013) however, argue that, using the Preston curve, and assuming that the specific country location on the curve fully applies to that country, the health increment related to the added value of an activity in a country can be calculated and used as an empirical pathway. They demonstrate that the method works, although only for low income countries, because the Preston curve is almost flat for high income countries. In a similar way, Hutchins and Sutherland (2008) demonstrate a pathway from per capita income to child mortality, enabling a comparison of the effect of added value in different countries to child mortality, and Bocoum et al. (2015) used the relation between the country level GINI (inequality) coefficient and child mortality to develop a pathway from added value to health.

Empirical models are based on a correlation between a single available, but objective, country indicator, such as the GDP or the GINI coefficient, and a S-LCA midpoint category. Most of the authors themselves point out that reality is far more complex, involving a multi-criteria relationship instead. In addition, it is very questionable if the used correlations are causal. For instance, considering the Preston curve, showing the relation between the GDP per capita and life expectancy, there are several other indicators showing a similar relation to life expectancy, for example, relevant for this study, corruption. Figure 1 shows the typical Preston type of relation between the corruption index and the GDP per capita for the year 2011. But Fig. 2, showing a much weaker correlation between the same indicators for the year 2016/17, strengthens our doubt about a causal relationship between corruption and GDP. On the other side, Wu et al. (2015) show, based on multi-criteria statistical techniques, that although there is no significant direct causal relation between income and health, there is a significant indirect pathway from income, via health expenditures to health. However, they did not study governance in general and in our intuitive opinion, it is very likely that governmental governance in general is the leading indicator, leading to Preston-curve like relations with life expectancy for several of its components. Jørgensen, Lai, & Hauschild argue: “If there is no valid impact pathway, there is no way of telling whether and to what extent the indicators that we apply in S-LCA actually represent damage or benefits to the Area of Protection”. The authors studied the validity of the pathway from child labor to wellbeing (Jørgense, et al. 2010b) and from employment to wellbeing (Jørgensen et al. 2010a) and conclude that too many variables and unknowns affect the pathways to develop a valid pathway.

Although recognizing these difficulties, Weidema (2006) is quite optimistic about finding valid pathways and argues that it is better to develop valid pathways for the less complex categories and use other techniques for the more complex categories. Following the common ELCA technique of using concrete biophysical pathways towards the endpoint category of health, expressed in disability-adjusted life years (DALYs), he proposes quality-adjusted life years (QALYs) as a characterization factor for S-LCA. By extending existing disability and disease classification systems (as a percentage of lost life years) with factors for the various social stressors, all established impact pathways leading to human well-being may be effectively interval quantified.

We will now assess the empirical Type-II models on our five criteria.

Quite a number of ELCI-database- type-II models have been described. Because of the scope of this article, we will only discuss three of the most quantitative models. Baumann et al. (2013) follow standard ELCA methodology transforming the inventory data of an airbag in automobiles, into DALYs. They calculate both the lost DALYs in the production stage of the airbag, and the saved DALYs during the use of them. Following the S-LCA guideline of assessing both the positive and the negative effects, they herewith express occupational health and safety aspects in the LCA endpoints of human health.

Hunkeler (2006) however, assessing the social impact of two different detergents and considering employment a key factor to social wellbeing, proposes breaking up the life cycle inventory into labor hours, divided over different countries. Thereafter the hours are comparatively weighted by the relative labor intensity factor in the different countries or regions, and thereafter characterized by the number of workhours a person needs, to locally access a combination of societal needs. This way an assessment is made of where a company’s activity would have the most positive effect on people’s ability to fulfill their needs.

Labuschagne et al. (2006) developed a model of four social Areas of Protection, leading via a quite comprehensive list of midpoint categories to concrete characterization factors. They distinguish both quantitative and qualitative categories. Results for most categories are quantitative and made aggregable by normalizing the indicators against regional conditions and weighting by comparison with stakeholders’ perceived- and target conditions. Their quantitative indicators are concrete (e.g. number of accidents and the number of employment opportunities) and based on data availability, but therefore not necessarily the most relevant to assess the aspect. In a test on three projects, the authors attempted to collect data for both company and country levels, but concluded that at the time of testing the model, insufficient data were available. Normalization is achieved by the determination of the fraction of the results on the specific studied unit relative to those in a larger reference system, such as the GDP, regional practices, or sector output. This type of normalization is especially effective to assess the relative importance of different aspects, because it gives a comparison of the relative contribution of different aspects in a greater system (e.g. 1% of a country’s CO2 emissions and 5% of the country’s child labor), which is lost in aggregation. Normalization however often comes with an undesired weighting effect. For example, considering corruption, it will make a huge difference, if in the international arena the results are compared with the standards of a corrupt production location or with those of a conscientious consumer or supplier in a country with low corruption, in which both are stakeholders. In addition, aggregation makes little sense if used for different aspects, and not all aspects are substantively related to the reference system.

We will now assess these ELCI-database methods with our five criteria.

Summarizing the above, quantitative impact-based S-LCA remains unsatisfactory for our goal of quantification of the aspect of corruption. In addition, the above described Tunisian incident demonstrates that if people become desperate, the consequences are totally unpredictable and that there are social tipping points, the exceeding of which may result in consequences never to be grasped in any statistical analysis. However, there are two developments towards company- and prevention-based assessment, the “Dreyer – company assessment” and the Oiconomy system.

Dreyer et al. (2006, p. 96) and Dreyer et al. (2010) argue that contrary to the assessment of environmental aspects, where the impact can be directly derived from the physical flows and processes in the life cycle of a product, social impacts are generally determined by the conduct of the companies which are engaged in the life cycle and the required specific data can only be obtained from the companies themselves. To accomplish this, a method is required both for the assessment of the companies and for the allocation of the companies’ scores to the product. Inspired by common certification standards, the authors argue that company performance can better be assessed by their efforts to prevent impact than by the impact itself, and propose an assessment method of the risk that a company presents, based on a combination of the company’s own governance quality and its contextual circumstances. For the company assessment the authors developed an extensive checklist and scoring system of “managerial measures” on three necessary “efforts”: planning, communication and active establishment and control. The three efforts are each scored one of three degrees of implementation for all managerial measures, added up and converted to a score between 0 and 1 by comparison with the maximum possible score and thereafter converted into one company risk factor by means of a complex system of division scores into risk classes and multiplication of the three effort scores, for which we refer to Dreyer et al. (2010).

A strong aspect, in our opinion, is the division in efforts and multiplicative effect of these three efforts, because non-compliance with any of the three efforts, means that governance is not effective.

However, comparing their managerial measures with the criteria in risk-based certification standards, the authors only use part of these, and choose to mix these general governance types of criteria with concrete subject-related managerial measures. In contrast, the standards leave the subject-related measures to the company, including the obligation to define these in one of the criteria, but apply, instead of three efforts, a system of more than 60 general governance criteria. It must be noted that Dreyer et al. describe a system for labor rights, where most subject-related requirements are well defined by ILO conventions. For other aspects, like safety and corruption, managerial measures may be far more company-specific, therefore less easily pre-definable and therefore better left to the company. Because of their choices Dreyer et al., in our opinion, miss the opportunity to measure the degree of company performance by means of checking and ticking the complete set of criteria and therefore apply a complex system of scoring which turns a potentially interval quantitative score into a nominal quantitative score.

After the determination of a company score, according to Dreyer et al. (2010), a share factor is required to determine the relevance of the company score for the product and, if more than one product is produced, to distribute the company impact over the different products. In principle, there are various ways to determine a share factor, but, without a standardized and verified system it will be very difficult to consistently execute such calculation over all supply chain actors. Where Hunkeler (2006), Norris (2006) and, Andrews et al. (2009) usually use workhours as the activity factor to determine the contribution or distribution of an impact in the supply chain, Dreyer et al. (2006, p.90) and Hauschild et al. (2008, p.23) argue that there is no single objective choice of an activity factor, and mention physical weight, cost contribution, value contribution and workhours as options for the distribution of a company score to a product. The disadvantage of any choice is however the same, which is the limited applicability to aspects and share- or activity factors that are relevant to each other (Hauschild et al. 2008). For example, one cannot use workhour distribution for the allocation of scores of the bribery type of corruption, because bribery negatively affects the society and competitors more than the workers.

We assess this method on our five criteria:

Contrary to the conventional impact-based LCA, in preventative cost-based LCA, the impact is not the (un)sustainability measure, but the costs of preventing the impact. Although knowledge of the impact pathway is required to develop a preventative measure, the costs of the measure are independent of impact pathways and the perception of the seriousness. The EcoCost system (Vogtländer and Bijma 2000) and the envisioned Oiconomy system (Croes and Vermeulen 2015) weight aspects by the preventative cost-based distance to a target, or performance reference point (PRP). The EcoCost system is a conventional LCA system, intended for comparative assessment and limited to environmental aspects. It uses predetermined marginal preventative costs, which are the costs of the last (most expensive) employed available preventative measure necessary to reach the target (assuming that the cheapest measures are employed first). The Oiconomy system follows the supply chain, copying the financial gradual price build-up and bookkeeping in the supply chain for the hidden preventative costs of complete products. Central in the Oiconomy system is a comprehensive measuring standard (Croes 2013), leading the supply chain actor through all sustainability aspects and resulting in “Eco Social Cost Units (ESCUs)” for every aspect, added together to a total ESCU score which is transferred to the next actor in the supply chain. Every ESCU is, just as for standard financial costs, the product of a quantitative factor, provided by the supply chain actor, and a price factor. The system challenges companies to self-calculate their specific product preventative costs and only to use the predetermined marginal costs as default values where companies cannot demonstrate more accurate data. Because the Oiconomy system, by means of certification, makes all background data into foreground data, it overcomes the difficulties of the collection of product and company-specific data in S-LCA and also overcomes the need for activity variables like workhours (discussed in the section on ELCI-database- type-II models) and extends the EcoCosts ELCA-system with social aspects.

The Oiconomy system is normative and, added to the market price, aims to provide the real price of a damage-free alternative of the product. The ESCUs may be used both for comparative analyses and as a score for the sustainability of an individual product.

In order to include our new proposals for quantification of corruption in the five-criteria assessment of the Oiconomy system, we will now first present these proposals.

The impact category is product-related corruption, characterized by the costs required to prevent corruption.

Because corruption is defined as a crime by the 2004 UN Convention against Corruption and by most countries’ legislation, the legal target or PRP must be zero corruption. However, no country is without corruption and Fig. 2 demonstrates that an index score of 43 in the Corruption Index presents a quite sharp index score above which corruption does not seem (on average) to have a severe impact on life expectancy (as indicator for health). However, if we, equally justifiably, replace life expectancy with the happy life years’ index (Fig. 3), which multiplies life expectancy with an indicator on life satisfaction, this point becomes less sharp and shifts to higher values, demonstrating that we may not exclude a corruption-induced impact on wellbeing below an index score of 64, which is only accomplished in a very limited number of countries. We therefore propose a zero target for the company but a calculation and verification tolerance in countries with a score > 64, based on data from the year 2011.

Palmier (1985, p. 271) identified three important causes for corruption: Low salaries, opportunities and low risk of detection and punishment, factors which according to Palmier should be addressed all three together. Quah (2011, p. 17–20) points to the cause that in many countries, civil servants’ salaries are by far insufficient to sustain their families and often lower than in the private sector. Governmental bureaucracy and regulations on the access to resources and activities provide opportunities, and if police and other members of the juridical system are also underpaid, policing against corruption will be inadequate. Important factors in prevention are transparency, political will and the example of the leadership. With corrupt leaders and policing officers, the chances of being detected and punished are perceived to be small. Although governmental bureaucracy and public officers commonly provide the demand side of corruption and business see themselves as victims (Runde et al. 2014, p. 5), business usually does provide the supply side of corruption (Myint 2000, p. 42). In addition, corruption between private companies, with businesses on both the demand and supply side, is perceived to be just as common (Heinrich and Hardoon 2011, p.19).

Because corruption is an illicit activity, it is much easier to measure an organization’s preventative measures against corruption than its actual involvement in corruption incidents. There is quite a body of literature on combatting corruption e.g. (Asian Development Bank 2010; International Chamber of Commerce et al. 2008; OECD 2011; Runde et al. 2014; United Nations Global Compact 2010; World Bank 1997, 2008; Transparency International 2009; Tanzi 1998; Walcher et al. 2013; World Economic Forum 2005). All this literature demonstrates that good governance principles are key to combatting corruption. Well-developed governance systems and standards on various subjects are available, suitable for independent verification, e.g. (ISO 2004; OSHAS 2007). However, the costs of establishing and maintaining good governance are difficult to estimate, and are dependent on factors such as location and culture, availability of qualified people, size, sector, complexity, capital- and labor intensity, and experience with other operational standards.

By its business principles for countering bribery, Transparency International (2009) (TI) provides a comprehensive standard on organizational governance. The World Bank distinguishes three levels of combatting corruption for an organization. 1. Internal: Assessment of the internal risk, implementation of anti-corruption measures as part of an overall compliance program, and provision of guidance to the managers within the organization. 2. External: Sharing the internal policies with stakeholders. 3. Collective: Reaching out externally via independent facilitators to joint activities to fight corruption (World Bank 2008). Based on these principles, United Nations Global Compact (2010) provides comprehensive guidance in anti-corruption practices approaching the different stakeholders with corruption challenges. But Sedex and Verité (2013, p. 9) recommend to use the principles of common risk based management systems, such as ISO 14001, to address ethical aspects.

Because of dependence on the location, the size of the organization and specific activity, the industry sector, the stakeholders and internal organization, on collective cooperation, and the highly multi-aspect character of combatting corruption, the direct costs of measures against corruption will be very different to determine. The magnitude and persistence of bribes often depend on the ease with which the organization could relocate its activities to a less corrupt environment, which is one of the major reasons why location-dependent activities such as mining and forestry, are corruption-sensitive sectors (Bai et al. 2015; Svensson 2003, p.211; World Bank 1997, p.18). However, this also means that the ultimate measure against corruption is to refuse to pay bribes and accept the potential necessity to refrain from the activity altogether, or relocate activities away from the corrupt environment.

To obtain the marginal preventative costs, we need to sort the available preventative measures by their costs and deploy them starting from the cheapest until the target is reached. The last deployed measure presents the marginal preventative costs. From our literature study of currently proposed preventative measures against corruption, we recognize four cost levels.

Considering the first level of costs, no direct data on costs of compliance to corruption criteria could be found in literature. However, the Ponemon Institute (2011, p. 13) studied compliance costs of USA industry to USA legislation, such as on privacy, data protection and regulations. The report finds a range of yearly per capita compliance costs of $ 134–$ 535, inversely related to the size of the companies. At an average yearly labor cost of about $ 60,000 per capita, that would represent about 0.22–0.89% of the wages. Considering the second cost level, more studies have been made on costs of compliance to taxation. Allers (1994) reports, based on a 1989 survey, an average compliance costs of 3% of the wage sum in the Netherlands (including costs of the tax office itself). The European Commission (2004, p.4) reports compliance costs of 0.02% of sales for large companies and of 2.6% for small and medium sized companies. These costs are very inaccurate, but this is irrelevant for our purpose, because they are lower than the costs of the marginal measure necessary to globally reduce corruption to the target level. The third level of costs for refusing to pay bribes and if necessary take the consequence of refraining from or relocating the activity, needs to be deployed. In fact, these consequences are already widely taken by all the companies that previously decided not to enter, or left corrupt markets. Fourth level costs will only be an option if these are lower than the profit margin. Therefore, we argue that, because in the long term companies will always end their activities when no profits are made, the margin made on a product represents the marginal preventative costs for the aspect of corruption. However, profit margins are volatile and may be unequally shared by companies in the supply chain, e.g. by the use of subcontractors, and multi-year product-specific averages are unusable in the case of fast moving products. Therefore, we propose to use the local five-year sector average net profit margin as a maximum default value but, without public availability of reliable local sector averages, to take the readily available S&P 500 data. An indication of the magnitude is presented in Table 2 showing the five-year sector average net profit margins in the S&P 500, from April 2016. If a company can demonstrate reliable data on all its products’ profit margins, product distinctiveness could be further improved by using the product’s share in companies’ profit as a share factor.

The question left now is how to derive specific company- and product ESCUs from this maximum value. Literature demonstrates that, before taking the ultimate measure of exiting or relocating, good governance is the way to combat corruption.

As described above, Dreyer et al. propose a system of scoring the degree of compliance to a set of managerial measures for the determination of a company risk factor. However, they do not use the complete set of general governance criteria commonly used in risk-based certification standards, and thereby, in our opinion and because they do not base their system on actual certification, miss the opportunity to interval quantify the effort scores. Weighting profitability against preventative measures is typically a top management responsibility. Therefore, it seems fair to set the actual preventative costs at a proportion of the measured quality of management preventative efforts. Therefore, for the Oiconomy system, we propose a modification of the Dreyer system as follows:

The scoring system, filled with example scores, is presented in Appendix 1 (Electronic Supplementary Material).

The first three effort classes are very similar to Dreyers’ efforts. We added the fourth effort class for three reasons. 1. In common standards, the management review probably is the most important criterion, because it demonstrates continuous management commitment, evaluation and correction and requires top management to be included in audits. 2. The multiplication with an extra effort score adds to fast reduction of the company risk score with non-conformances, which is in accordance with certification practices of rejecting a certificate already with very few non-conformances. 3. The Plan-Do-Check-Act (PDCA) principles are the core of most well recognized governance standards.

The formula gives equal weight to all 4 effort classes, because compliance of all four need is necessary for effective governance. However, within the classes the criteria carry more weight the lower the number of criteria in the effort class. This answers well to the above mentioned importance of the management review with only 4 criteria in the effort class.

An explanation is required about the Oiconomy ESCU allocation of indirect costs to companies’ different products in action 5. The Oiconomy system is a bookkeeping system for hidden preventative costs, as much as possible copying companies’ financial bookkeeping system. To enable companies to exactly use their standard bookkeeping and cost allocation methods, and to avoid special “greenwashing” allocations for S-LCA purposes, for the Oiconomy system we propose to require an exact copy of companies’ standard financial cost allocation methods. The mentioned advantages outweigh the disadvantage of a not fully harmonized allocation to products, because it only concerns indirect costs (e.g. overhead). All direct costs can be allocated directly to the product.

We do not need Dreyer’s generic contextual company risk factor, because the Oiconomy system is based on actual and verified company data. Instead, as mentioned before in step 2, in order to avoid unnecessary burdens to the supply chain actors we propose to allocate zero ESCUs in countries with a corruption index score above 64, which also means that in a risk-based certification system the aspect needs no verification in such countries.