CSRHub Rating Rules


CSRHub follows a well-defined set of rules that determine when we can rate any part of a company’s performance and also when we can offer an overall rating. 

CSRHub ESG Ratings

Here is a summary of the steps:

A.   In order to rate a subcategory, we require:

a.    A minimum number of sources (it ranges depending upon a variety of circumstances between two and six sources) for each subcategory.  So, to give a company a rating for “Energy & Climate Change” we might need data from both CDP and Climate Counts. 

b.    A minimum amount of data.  We measure this in terms of “data weight.”  Some sources tend to predict and follow the consensus of our other sources—others diverge often from consensus.  Our software gives the sources that are good predictors a higher weight than those who are not.  Some sources invest a lot of resources in their work and/or generate original data.  Our software gives these sources additional weight compared to those who merely summarize work done by others.  Some sources offer one rating that covers a wide range of sustainability issues while others have many detailed ratings elements.  Those with more elements get more weight.

c.    If there is not good agreement between the data sources or if the resulting score is extreme (e.g., 0 or 100), we may exclude the result.  (Whether or not we do depends on the quality of the sources, number of sources, etc.)

B.   To score a category, we must have a rating for at least one subcategory.  We may suppress a category rating if we do not have enough weight in the subcategories underneath it to produce a reliable score.

C.   To offer an overall rating, we must have:

a.    Ratings for all four categories.

b.    Ratings for at least five subcategories (so at least one category must have two subcategories in it).

c.    Enough total weight.

d.    Enough total sources.

e.    If the weight is light or the number of sources is low, a reasonable score (we trim outliers that do not have enough support to justify). 

The above process is mechanical—our software handles the details of both converting the data we receive into a 0 to 100 score, mapping it into our subcategories and special issues, normalizing the data across all of the companies we follow, and then processing the data to produce ratings.  We have data on approximately 100,000 companies.  We analyze data on 31,658 companies.  We issue ratings on 17,268 companies (about 67% of the companies we analyze data on).  We offer full ratings on only 8,419 of these (about 70% of the companies we rate).  Each month, we conduct a separate “human review” of our ratings to make sure that we have not missed an obvious problem or outlier.  At present, only 26 companies are receiving “manual” adjustments. 

Surprisingly few of our users ask questions about this process.  We hope that this indicates that our ratings are reasonable and fit generally with expert opinion about how the companies we follow are performing.  We are actively working to increase the scope of our coverage.  We currently rate about 800 private companies, government agencies, and NFPs.  We hope our non-public company coverage will soon exceed that of public companies.