Skip to main content

Revisiting the American Statistical Association's Ethical Guidelines for Statistical Practice

Howard Hogan is Chair of the ASA Committee on Professional Ethics and Chief Demographer, U. S. Census Bureau. This paper reports the work of the ASA Committee on Professional Ethics. Any views expressed are those of the Author and not necessarily those of the U.S. Census Bureau.

The American Statistical Association (ASA) was founded in 1839, but it was 1949 before an ASA committee recommended developing a code of ethical practice. This was followed by years of discussion. In 1981, the ASA Board of Directors approved a Code of Conduct on a “three-year trial basis.” A few years later, the Committee on Professional Ethics was designated as a permanent ASA Committee. Within its duties is to “maintain and promulgate, subject to Board review and approval, the set of ASA Ethical Guidelines that describes the general view of ethics in statistical practice….” The most recent Guidelines were approved by the ASA Board in 1999 (http://www.amstat.org/committees/ethics/index.cfm).

A little background is important to understand the Guidelines and the revision process. First is the role of the Guidelines in the Association. The Guidelines are just that: guidelines. The Committee’s charter specifically states: “The Committee does not have the authority to act on, rule on, or arbitrate ethical matters. Execution of a program of professional certification is not included in this charge.” Indeed the Guidelines are explicitly “for Statistical Practice” and not just “for ASA members.” They are there to help those practicing statistics in its wider definition, not to police them.

Second, the Committee is made up of nine members, chosen by the ASA President to represent the Association. Its membership is diverse with respect to gender and ethnicity. Equally important, the Committee includes those active in government, academic and industry, with the industry “representatives” including those working in biomedical fields. Several members have worked within a legal/litigation context. To a great extent, the Committee includes people who have worked professionally in the field of statistical ethics; for example, taught a course or served on a review board. As this was a multi-year process, the Committee was assisted by previous members who had officially rotated off.

A principal motivation for the current revision was simply the passage of time; the Committee felt that good practice required that the Guidelines be reviewed at least once a decade. Were there developments in statistics that might require changing the Guidelines? Many people mentioned “Big Data.”

A second motivation was to make the Guidelines more accessible to our members. The 1999 Guidelines ran 11 pages, with almost four thousand words. The document also tended to be repetitious with an executive summary, followed by a Preamble and then the Guidelines. Further, the actual “rules” were intermixed with discussion and background in a very confusing way. Could we make the Guidelines more readable on Tablets and Smart-Phones, for example? Could we use hyper-links and nesting to quickly get the reader to the information being sought?

Exactly when the revision process started is hard to say. However, an important early milestone was a summary and comparison of ethics code of the ASA with that of the International Statistical Institute (ISI) and the Royal Statistical Society (RSS) prepared by Ron Wasserstein, Executive Director, ASA, in August 2009. Did the ASA have the right approach to the Guidelines? Were we missing something? The codes of other associations and national statistical offices were reviewed. However, after several years of discussion, the Committee decided there was nothing fundamentally wrong with the 1999 Guidelines and that they would make a good starting point for the revision.

The current effort started in earnest in late 2013, when the Committee on Professional Ethics agreed on a written set of goals for the revision. This included an agreement within the Committee on the purpose of the Guidelines: "The main purpose of the guidelines is to document shared values that distinguish statisticians as professionals with responsibility for beneficent use of our special knowledge and the privileged insights it grants."

It included a statement of the goals to the revision, highlighting:

  • Formatting and accessibility:
    • The revision should not be undertaken as an effort to reduce the scope of the professional code. Revisions should therefore focus on improving relevance, clarity.
  • Content:
    • Any revision should look to see whether there are new issues that the guidelines should address, or changes in the manner in which old issues are addressed to recognize particular new forms in which these issues are emerging. One area involves “Big Data…”
  • Dissemination and outreach:
    • The existing ethical guidelines seem to exist in the background and be sought out by the people least in need of exposure to them. It would help frame our work by envisioning a new system of engagement with members on ethics.
  • Professionalism:
    • It’s also recognized that “The guidelines may be partially conflicting in specific cases.” Where the guidelines are less forthcoming, however, is in addressing whether and how moral values and/or shared values might contribute to a prioritization of ethical obligations in situations where guidelines may seem in conflict.

With these goals set, the Committee began the process of drafting, discussing, and re-drafting. I count over 30 drafts in my archives, but there were certainly more. The goals and process were discussed at a poster session during the 2014 Joint Statistical Meetings (JSM). By the time of the 2015 JSM, the Committee had a draft ready to distribute and to discuss, which it did at a lobby table and within appropriate sessions. Additionally, the draft was sent to the chairs of all ASA Sections, which represent the different subject-matter interests of the Association.
In late 2015, a preliminary draft was presented to the ASA Board for their comments and direction. Their most important input was to adjust the tone of the guidelines, essentially to replace “the ethical statistician shall…” in that draft with “the ethical statistician acts/does/considers…”

This revision was sent by the ASA to the chairs of all its committees, sections and chapters. Only one substantive change was suggested, which would be to include a recommendation for pro bono work. Although interesting, this change was not accepted by the Committee. Many statisticians, because of their employment or community, would not have a meaningful opportunity for such work. This is not to say that pro bono work should not be encouraged; it is just that the Guidelines are not the right venue. Additionally, the ASA and the Committee presented a free webinar, advertised and open to all ASA members. At this webinar, the background, goals and substance of the revision were presented. This allowed for a free discussion by interested members. However, no substantive recommendations or “fatal flaws” were uncovered.

In terms of content, the Guidelines now cover:

A. Professional Integrity and Accountability
B. Integrity of data and methods
C. Responsibilities to Science, Public and Client
D. Responsibilities to Research Subjects
E. Responsibilities to Research Team Colleagues
F. Responsibilities to Other Statisticians
G. Responsibilities Regarding Allegations of Misconduct
H. Responsibilities of Employers

They are guided by the following philosophy:

"Good statistical practice is fundamentally based on transparent assumptions, reproducible results, and valid interpretations. Above all, professionalism in statistical practice presumes the goal of advancing knowledge while avoiding harm; using statistics in pursuit of unethical ends is inherently unethical. The principles expressed here should guide both those whose primary occupation is statistics and those in all other disciplines who use statistical methods in their professional work."

The guidelines are now around half the previous length. They may differ from the earlier set in emphasis and tone, but are not a fundamentally different approach. “Big Data” is not explicitly addressed, but the Committee feels that the ethical questions are well covered by:

  • D 5: Considers whether appropriate research-subject approvals were obtained before participating in a study involving human beings or organizations, before analyzing data from such a study, and while reviewing manuscripts for publication or internal use. The statistician considers the treatment of research subjects … when evaluating the appropriateness of the data source(s).
  • D 6: In contemplating whether to participate in an analysis of data from a particular source, refuses to do so if participating in the analysis could reasonably be interpreted by individuals who provided information as sanctioning a violation of their rights.

In March, the Committee met and unanimously voted to forward the current draft to the ASA board for approval at its April meeting. If adopted by the Board, the Guidelines will be presented to the Association at the 2016 JSM. The current draft is available at http://www.amstat.org/misc/webinarfiles/ASA_Ethical_Guidelines_Revised_12-30-15.pdf.

View Full Report