CPSC Civil Penalties: A Statistical Analysis


CPSC Civil Penalties: A Statistical Analysis

Alert |


Jonathan Judge

Each February, consumer product safety professionals flock to the International Consumer Product Health and Safety Organization (ICPHSO) annual meeting to network, learn of new developments and hear from regulators, most prominently employees of the Consumer Product Safety Commission (CPSC).

Every year, there is also a keynote speaker or speakers, and it is common for the keynote to be handled by one or more speakers from the CPSC. In February 2015, that speaker was the Honorable Elliot F. Kaye, who had been confirmed as CPSC chairman that previous summer.

Kaye discussed many issues, but one statement grabbed the post-seminar headlines: that Kaye had directed CPSC staff to seek significantly higher penalties from companies found to be in violation of the Consumer Product Safety Act (CPSA) and other statutes administered by CPSC, and in particular, that he had directed staff to seek “double-digit” penalties when possible.

Since 2008, the maximum available penalty under the CPSA had been around $15 million; however, the highest penalty to date had comprised about a third of that amount. Virtually all of these penalties are imposed for failure to report product safety hazards to the agency in a timely manner.

It is now February of 2017, and as we approach another ICPHSO keynote, Kaye appears to have accomplished his stated goal, at least at first glance. Since Feb. 19, 2015, the average CPSC penalty settlement has been just over $2.9 million.

This is approaching double the average amount of the previous two years’ worth of penalties (just over $1.6 million). Above all, the CPSC has handed out its first $15+ million penalty, thus exercising discretion at the very top of its statutory authority.

This increase in penalties has not gone unchallenged. The most vociferous criticism has come from Commissioner Joseph Mohorovic, who has expressed repeated concern about the penalty-generation process and the criteria used to decide what penalty is appropriate.

Commissioner Mohorovic has asserted that the penalty assessment process “lack[s] any rigorous analysis,” fails to provide fair notice to affected companies, and is in a “generally disordered state.”

Shortly after Kaye’s 2015 speech, the BNA Product Safety and Liability Reporter published my statistical analysis of past CPSC fines. Analyzing 10 years of previous CPSC fines, my article used nonparametric statistical analysis to derive the factors that seemed to be driving the size of CPSC fines, and predicted the average fine to be expected for various hypothetical transgressions.

The article also explained why there are serious, constitutional concerns at work as well. Without repeating the full analysis, companies are entitled to federal constitutional protection, and while no company has the right to exact same penalty as everyone else, companies do have a right to be treated in a reasonably similar manner and for their punishments to reflect a reasonable proportionality to their alleged transgressions.

So, while serious conduct may of course be punished more seriously, it is probably inappropriate to summarily increase penalties without substantive distinctions between previous and later companies who have violated the same rules.

It seems like the right time to take stock of what has resulted from Kaye’s directive, for several reasons. First, although the average penalty has increased markedly under Chairman Kaye, mere averages tell us nothing about the circumstances underlying those penalties. We need to look deeper and confirm, statistically speaking, whether the penalties are in fact a departure from past practice, given the underlying conduct.

Second, with the election of a new administration, the chairmanship of CPSC will shift parties, and one of the first questions facing the new chairman may be whether the recent trend of increased penalty size is appropriate going forward.

Statistical Analysis: CPSC Penalties, 2015-2017
Over the last two years, there have been 13 CPSC penalties for company misconduct, ranging from $1.75 million to the practical maximum, $15.45 million. How do we analyze whether those penalties are, in fact, fair and proportional to both the conduct involved and to what other companies have received for comparable conduct?

The answer, once again, is statistical analysis. By preparing a model that analyzes the penalties that preceded Kaye’s speech, we can predict what subsequent penalties should have been and look for disparities. The clues as to what those penalties should have been can be drawn from the settlement agreement or consent decree that accompanies each penalty, laying out the factors the agency believes merit the fine.

By tracking those factors into a statistical model that compares the factors both to each other and also to the size of the penalty, we can put all penalty recipients on substantially the same playing field.

It turns out that what I will call the “previous” penalties (dating from about 2005 through mid-February of 2015) follow a discernable pattern. While I cannot speak to the official “formula,” or even guarantee that CPSC used the exact same numbers I did, I can report that the formula being used can be substantially reverse-engineered as follows:

  1. The penalties begin with a base amount;
  2. If the conduct arose or at least partially occurred after the enactment of the CPSIA in 2008, CPSC adds a layer on top of that amount;
  3. CPSC adds a multiplier times the number of reported incidents;
  4. CPSC adds a multiplier times the number of units sold; and
  5. There is some evidence that an additional layer is added when the OGC finds that the firm engaged in misconduct.

Despite a potential penalty range of $15 million, this model successfully predicts all “previous” penalties, on average, within about $120,000, which, in my biased opinion, is pretty darn good. Now, let’s put the model to work.

Statistical models typically predict an average penalty. Using our model, what would have been the fair average penalty for the penalties handed down in 2015 and 2016, had they been punished instead before that time? Here is a sample of some of the penalties from those subsequent two years, along with the relevant predictors:

Products Sold Incidents Misconduct Model Prediction Actual Penalty Margin of Error
127,444 32 Yes $2,966,467 $4,250,000 $5,377,299
233,000 11 Yes $2,904,393 $5,550,000 $5,335,289
445,000 302 No $3,205,868 $3,750,000 $5,739,596
2,500,000 119 Yes $5,327,678 $15,450,000 $9,372,693
3,025 44 No $1,293,988 $2,940,000 $2,232,077
160,000 100 No $430,978 $1,543,500 $655,886

Here’s how to read this table:

  • All of these cases involve post CPSIA conduct.
  • The relevant number of products sold and incidents are noted, as are whether CPSC alleged misconduct by the firm in connection with the investigation or underlying recall.
  • The prediction is what the model would expect based on “previous” penalties.
  • The actual penalty is what was eventually imposed and/or agreed to.
  • Finally, the margin of error provides CPSC with one standard deviation of discretion above the predicted amount. The point of the margin of error is to acknowledge that CPSC may perceive individual circumstances that in its judgment justify varying from the standard fine, while also providing a statistically reasonable limitation upon that discretion.

At least on its face, Kaye’s initiative indeed seems to have succeeded in punishing these “subsequent” companies more severely than other companies were being punished before. Controlling for the underlying conduct, 10 out of the 11 companies subsequently fined were punished more severely, on average, than a typical company would have been punished previously, according to the model.

Several companies paid two or three times more than other companies were asked to pay before. The majority of the subsequent penalties (7 out of 12) even exceeded the “margin of error” I provided for agency discretion.

The infamous $15.45 million penalty arguably exceeded the average expected penalty by over $10 million. A standard statistical test suggests that it is unlikely these across-the-board increases have occurred by chance; rather, these results are consistent with a change in the fine schedule pursuant to the described directive.

Although some might deem these results a success story, others will find them more troubling. Of course, no company should be knowingly violating CPSC laws or regulations for any reason. However, there is a five-year statute of limitations on CPSC penalties, and many investigations last a fair amount of additional time.

If companies whose investigations just so happened to drag into 2015 began receiving disparately severe punishments, that calls into question the fairness, and perhaps even the constitutionally of those proposed fines.

Is it appropriate for an agency to dramatically increase its average fines years after the underlying conduct has already occurred? And if punishments are suddenly being increased, what does that say about the reasonableness of the punishments being handed down before?

These are policy questions for others to answer. However, as the CPSC adjusts to the realities (and composition) of a new administration, this analysis suggests a fork in the road stemming from one question: is this recent inflation in penalties constitutionally and prudentially appropriate?

If not, then steps may need to be taken — either by the commission itself or by companies/litigants involved in a civil penalty process — to ensure that companies facing future fines are treated more like the companies that preceded them.