The CSRHub Ratings Methodology
Our objective is to provide consistent ratings of Corporate Social Responsibility (CSR) performance for as broad a range of companies as possible. Given this objective, we face several methodological challenges:
- Our sources track different topics in different ways. For instance, one source might measure how a company treats its community by measuring how much money it contributes to local charities. Another might ask if a company has programs that allow its employees to take time off for charitable work. A third source might count the number of charity board memberships held by the company’s board members. All are valid estimates of a single aspect of corporate social performance—and each might give a different reading for any given company.
- Our sources each have their own rating and measurement methodology. Some sources given companies a numerical score (e.g., between 0.0 and 1.0). Some use “+” or “-” signs. Many sources offer only a relative ranking (e.g., “Top 50” or “Best Performing”).
- Each source tracks a different universe of companies. Some sources cover only specific industries. Many sources focus on one region or a single country. None of our sources offer data on more than about 60% of the companies we cover.
- Company performance changes over time. Many of our sources update their information only once per year. If a controversy arises regarding a particular company, it may take as much as two years for its effect to be reflected among all of our sources.
- Some sources rate company subsidiaries or individual products. Our ratings are given at the parent level of a company. It is difficult to fit together sometimes conflicting ratings on a company’s subsidiaries or on its products.
Our rating system attempts to remove most of the above sources of bias and inconsistency, by using this approach:
- Map to a central schema. We have divided Corporate Social Responsibility performance into twelve subcategories. These subcategories roll up into four categories. We have established an open-ended number of special issue topics to hold CSR issues that do not fit our twelve subcategory schema. We map each element of data we receive from a data source into one or more subcategory and/or one or more Special Issue. For instance, if a data source reports that a company is involved in Burma, we include this information in our Leadership Ethics subcategory and in our “Involved in Burma” special issue. We have mapped over [data_elements] data elements.
- Convert to a numeric scale. We take each data item from our sources and convert it into a rating on a 0 to 100 scale (100 = positive rating).
- Normalize. We compare the scores from different data sources for the same company. By analyzing the variations between our sources, we can determine their biases. We then adjust all of the scores from a source to remove bias and create a more consistent rating.
- Aggregate. We weight each source based on our estimate of its credibility and value. We then combine all of the available data on a company and generate base ratings at the subcategory level. We then aggregate these ratings further to the category level.
- Trim. We drop ratings when we do not have enough information. We currently do not rate about 100,000 companies for whom we do not have enough information.
- We research each rated company and attempt to determine which industries it participates in. We gather contact information, a description of the company’s business, and the location of its Web site. This information allows us to create industry and country averages. We have set up our own industry category system, based loosely on the NAICS code structure. This page gives the conversion from our industry codes to the NAICS codes.