Financial panel presentation for 2009 TRI conference

Sources of data

EPA has data of interest to the financial community: on releases of pollution, hazardous wastes, chemical accidents, contaminated sites (e.g. Superfund), and enforcement/penalties.

The first and most important task that has to be done for using any of these data for company-specific purposes is to match EPA's tracked entities (facilities, sites, accidents) to parent companies.

Some major EPA data collections

The Toxic Release Inventory is the most important and commonly used of these, plus it can be matched to RSEI risk screening information.

How good are TRI parent company identifiers?

Not very good.

Here, as an example of data quality, are TRI records for Honeywell International, a Fortune 500 company. They are from 2007 only, the latest TRI year. For that year Honeywell seems to have 72 facilities within TRI.

Breakout of Honeywell International facilities by Parent D&B number

D&B numbers are often blank (i.e. NA) or invalid, and sometimes apparently incorrect.

Parent company names are better, but not really reliable.

Breakout of Honeywell International facilities by Parent Company Name

There are some spelling and punctuation variations, sufficient to keep a computer program from grouping them together easily -- note especially the "Interanational". But more significantly, 3 have NA, and one lists UOP (a Honeywell company).

This is not good data quality. It's easy for a large company to send out a memo to all their TRI reporting employees telling them to fill in the same parent name. Some companies do it, and we have direct evidence that some don't. It's an open question among people who use these data whether this kind of variation is just inattention on the company's part or whether it's sometimes a deliberate attempt to make it more difficult to find out what the company is doing.

This is bad for reasons of environmental management as well as for reporting reasons. New efforts at coordinating reporting should also encourage company oversight.

What does PERI do to assign TRI parent companies?

The Political Economy Research Institute (PERI) is an independent unit of the University of Massachusetts. PERI must assign a parent company to every TRI facility as part of its Toxic 100 project, which reports on the top U.S. air polluters for chronic human health effects as scored by RSEI. Toxic 100 data are publicly accessible through the Web at both the company and individual facility levels.

The process can be described both technically, in terms of what is actually done, and philosophically, in terms of what overall decisions have been made to guide it. First, the technical steps, since they go along with the data tables presented earlier. There are currently about 53,000 TRI facilities to assign, if you include all TRI years.

Technical steps for parent company assignment

1. First, each facility is assigned a regularized parent name -- regularized by dropping all extensions like "Inc." or "Co.", dropping most punctuation, and putting abbreviations in common format e.g. "International" becomes "Intl". Then the facilities are grouped together by exact matches on this regularized name;
2. It is assumed that when parent D&B numbers exist, they may be wrong, but they are unlikely to match with those of other facilities if thee are wrong. Therefore, sets of facilities grouped together by name are again grouped together if they have the same parent D&B number;
3. The first two steps above are done by computer program, but this is never good enough. So at this stage the facilities and their group assignments as created by the program are put in a form that human beings can check over, and they are checked over;
4. The final, corrected groups of facilities can have parent company names assigned to them manually as part of the step above, but if they don't, the parent company name is assigned by a computer program to each group of facilities by taking the most common parent company name within the group.

Manual checking of parent companies

This is the most important, lengthy, and costly step. It's difficult to find people who can do it fairly reliably. In general, someone at grad school level seems to be needed, but it's very boring. It consists largely of looking over the list with an experienced eye and checking Web-accessible sources of company ownership information.

One of the most important of these is now the company's own Web site (if it is major enough so that you're not likely to confuse the company with another of the same name.) If a company is unclear on its own Web site about who owns it, it's noted down as a probable bad actor.

Sometimes the TRI public or technical contact for the company must be called. In a several-decade history of working with TRI, I've never had a call answered or returned by a public contact, so I tell people to go straight to the technical contact.

Companies sometimes attempt to obstruct callers. One grad student was told that she needed to supply her student ID so that the company could check on her before they would give her information about who owned them. A good deal of the time, lack of information about who owns a company seems to be purposeful on that company's part.

Issues in parent company assignment

Definition of parent company

Before you can assign parent companies, you have to decide what they are. Some EPA programs (not TRI) seem to treat "parent company" as "holding company", so that not every company has one. PERI treats every company as having a parent company, according to the following rules:

1. If an entity owns 51% or more of the facility, that entity is the parent company;
2. If two entities each own 50% of the facility (a joint venture), the joint venture is the parent company;
3. If an entity owns 50% of the facility, and no other single entity owns the other 50%, that entity is the parent company;
4. If no single entity owns 50% of the facility, the facility is its own parent company.

In that way, parent companies can be assigned for public, private, or government-owned entities. It's an artificial decision to break down ownership into a maximum of two entities rather than tracking e.g. 40%, 30%, 20% shares, but it limits what needs to be tracked.

In all cases, ownership is treated as ultimate ownership -- e.g. when a facility is owned by a U.S. subsidiary of a foreign company, it's treated as owned by the foreign company.

Time issues and ownership

Facilities change ownership over time, and companies go through acquisitions, mergers, splits, and bankruptcies. How should past pollution be handled?

For its purposes, PERI has decided that when a company buys a facility, they also take responsibility for all of that facility's past pollution. So ownership of facilities is not tracked year by year: the latest owner of the facility is treated as the owner for all years.

Because TRI data are reported with a two-year lag, e.g., the 2009 Public Data Release reports 2007 data, the listed TRI parent company may no longer be the parent company. PERI starts with the information from the latest public data release, but this information may no longer be current. Therefore, PERI attempts to set parent company ownership to what it is at the time when PERI does its analysis, which may use later parent company information than the latest within TRI.

What could be done to help this process?

1. EPA could issue reporting guidance to better define what a parent company is for TRI reporters;
2. EPA could change its TRI reporting package, TRI-ME, to emphasize parent company checking at the point where data is entered. It could also begin to reject TRI submissions for parent company problems;
3. Financial services industry users could share their TRI parent company assignments;
4. EPA could create new data fields with TRI parent company assignments;
5. EPA could replace D&B identifiers with a more publicly accessible ID system.