Reform Retro!

The State recently hired an independent expert, Oliver Wyman Actuarial Consulting, Inc to examine the policies and procedures of the Department of Labor and Industries. In particular, the study was to determine whether the L & I Retro program was fair and equitable. The 100 page report found several major problems with the formulas used to calculate retro refunds, but then strangely concluded that the L & I Retro program policies and procedures were fair and equitable. Sadly, the Wyman Retro Report, released in August 2009, was also loaded with errors.  Here are just a few of the errors in the report and found by the report:

 

1. The report assumed, based entirely on unchecked statements from L & I that Retro programs had lower costs than non-retro programs. For example, on page 22, the report states “Retro employers generally have better than expected claims experience relative to non-retro programs. Although the report was 100 pages long, it never tried to identify how much that difference was or how that difference was calculated. Nor did the report provide any actual underlying data to support the assertion. Nor did the report attempt to analyze whether the difference was merely cherry picking (also called claims management), or whether the difference between retro and non-retro improved or got worse over time as an indication that retro did or did not improve job safety.

 

2. On page 21, the report referred to the Performance Adjustment Factor (PAF) as a Policy Adjustment Factor. The authors noted that no other State uses a Performance Adjustment Factor. Perhaps this was why the authors of the report (who are located in New York) made so many mistakes in attempting to interpret how different data would affect the Performance Adjustment Factor.

 

3. For example, Footnote 23 on the bottom of page 22 was factually incorrect. The footnote claims: Historically, PAF’’s have ranged from 0.6 to 1.1, although they have been below 1.0 prior to the last few years. The very low PAF’s in the early 2000’s were a direct result of low rate levels, and therefore high loss ratios, during this period of time. More recently, rate levels have increased (in part due to significantly reduced investment income). With higher rate level and lower loss ratios, PAF’s have increased significantly.

The reason the above statement is incorrect is that PAF’s are independent of rate levels or investment income (because both rate levels and investment income affect retro and non retro equally). PAF’s are a relative ratio of only retro to non-retro claims. They are independent of rate level or actual losses going up or down. Only the relative ratio matters.  Since the premiums of retro and non-retro programs are initially set the same, the only data which can affect the PAF is the relative ratio of Retro claims to non-retro claims. The authors likely confused PAF’s with Loss Ratios which do go up and down with a variety of factors. It is very troubling that the authors of this report failed to understand this simple fact about PAF ratios and calls into question the over-all credibility of their report.

 

4. On page 26, the report simply accepted L & I’s claim that the computer coding error only increased retro refunds by 10%: “The correction of this issue may potentially reduce the average retrospective refund by approximately 10%, according to L & I.  The authors of the report noted they were not doing an audit of L & I, but they were supposed to check L & I calculations for potential Bias. Therefore they should have at least asked to see the formula used by L & I to determine that the correcting the coding error would only reduce the retro refunds by 10%. This is particularly true in light of the fact that it was the Wyman report authors requests for underlying information which led to the discovery of the computer coding error in January 2009. In fact, as we have shown above, the coding error accounts for almost 100% of the retro refunds and not merely 10% as claimed by L & I.

 

 

5. The Occupational Disease Computer “Coding Error”

Another major L & I  “error” uncovered by Wyman - which should have raised a giant RED FLAG – was that for the past 15 years, L & I has been handling the distribution of occupational disease claims in what Wyman referred to as an “actuarially unsound” manner. Beginning on page 35, the Wyman Retro report discusses “Treatment of Occupational Disease Data. The problem with occupational disease is that an employee might have worked for both Retro and non-Retro employers during their career. The question is how to fairly divide up claims?

The following diagram shows occupational disease claims for 2003 through 2007:

The green area ($116 million) was known to be solely due to retro employers. The darker blue area ($153 million) was known to be solely due to non-Retro employers. As non-retro employers are slightly more than 50%, it is expected they would have slightly more than 50% of claims.  The light blue area of $112 is attributable to either retro or non retro or both.

The white area ($135 million) cannot be attributed to any employer.

 

The logical and equitable way to divide up the $247 million in unclear claims would be to use the ratio of $116 million over ($116 million plus $154 million) = 43% or $106 million would be charged to Retro and 57%or $141 million of the unclear costs would be assigned to non-retro.

 

But this is not what L & I did.  Their formula was to assign all of the “6750000” coded claims to Non-Retro and to completely ignore the “Not charged” claims. Thus, Retro claims remained $116 million while non-retro claims rose to $265 million. The resulting ratio of Retro to total claims was  $116/($116 + $265)  = 30% Retro with non-retro going up to 70% of claims.

 

This second computer coding error caused L & I to mistakenly believe that non-retro “occupational disease” claims were twice as high as retro occupational disease claims! According to the Wyman report (page 38), this error caused retro refunds to be 20% higher than they would have been had an equitable formula been used.

 

As total Retro refunds were about $1.3 billion, this single error could have resulted in an additional $260 million being improperly given to retro groups. Alternately it can be stated as costing non-retro groups $260 million more than they should have been charged.

 

 

 

On the last page of a paper released on September 21, 2009 (2010 Proposed Rate Discussion) L & I estimated that the “occupational disease assignment formula” caused retro refunds to be $30 million per year too high. This converts to $300 million more over the past 10 years.

 

The L & I response to the revelation of this problem has been grossly inaccurate. Instead of immediately correcting the error, and re-capturing excess retro refunds for the past 3 years, they have proposed a “rule change” which might or might not take effect in 2010.

 

But can the mis-assignment of hundreds of millions of dollars in occupational disease claims honestly be dismissed as merely an innocent computer coding error?

 

Given that it obviously benefits retro groups by artificially raising the cost of non-retro claims, it does not appear to be an error at all. Instead, it is more accurately described as a mathematical manipulation deliberately intended to stack the deck in favor of retro programs. 

 

This math manipulation also occurred in the early 1990’s when the double entry coding error occurred. It therefore calls into question whether the double entry was an “error” or part of a pattern of fraudulent conduct deliberately intended to rig the PAF in favor of artificially increasing Retro refunds. 

 

The past and present handling of the occupational disease claims is clearly a “smoking gun” confirming that someone at L & I was deliberately trying to inflate non-retro claims in order to artificially inflate retro refunds.

 

6. Inequity of limiting claims adjustments to 45 months.

 

Wyman also noted that the way long term claims adjustments were handled was “actuarially unsound” and resulted in artificially inflating retro refunds. On page 77 of the Wyman report, they note that the current limitation on retro claims adjustments at 21, 33 and 45 months results in retro refunds being 4% too high. The reason for this is that retro long term costs (beyond 45 months) are slightly higher than the long term costs of non-retro claims. The Wyman report recommends adding two more claim adjustment periods to the current 3. This would add a significant administrative cost and complexity to the system. About the same result can be achieved by adding one more claim adjustment at 60 months (5 years).

 

It is known that long term claims are about the same amount as short term claims. Thus, develop claims are about double initial claims. So if limiting claim adjustments to 45 months results in a 4% error, one would predict that total Retro claims must be about 8% higher than non-retro claims.

 

The Wyman report did not address the question of how it could be possible for short term retro claims to be 20% less than non-retro claims, yet long term Retro claims be 4% greater than non-retro claims. Statistically, this would seem to be impossible.  Instead, what is more likely is that short term retro claims are 8% higher than short term non-retro claims and long term retro claims are also 8% higher than long term non-retro claims.

 

No other Workers Comp system in the country ends adjustments after only 45 months. Most other States continue adjustments until there are no other adjustments to be made. The fact that L & I set up an adjustment policy which clearly maximizes retro refunds is a third “smoking gun” discovered at the scene of the crime.

 

 

8. Wyman “Evidence” that Retro programs cost less than non-retro programs:

Beginning on page 46 of their report, Wyman presents their “evidence” that retro programs cost less than non-retro programs.  This evidence is broken down by occupational categories using the following occupational category codes:

On page 54, Wyman presents the following chart showing that between 2003 to 2007, for nearly all occupational categories, retro groups have slightly lower loss ratios:

Because loss ratios are directly related to claims cost, the above chart leads to the conclusion that Retro claims as a percent of premiums were lower than non-retro claims. However, we know from L & I’s reported PAF calculations for this time period that – even without adjusting for the computer coding errors - Retro claims were higher than non-retro claims:

 

Year (3rd quarter only)	2003	2004	2005	2006	2007	Ave
Reported PAF (Deflated due to coding error)	0.99	1.09	1.13	1.06	1.08	1.07
Adjusted PAF (corrected for coding error)	1.02	1.12	1.16	1.09	1.11	1.10

 

 

The L & I reported PAF for this period is 1.07 and the adjusted PAF is even higher.

So how can the PAF ratios be above one and yet the claims loss ratios be below one??? The answer is that they cannot.

Instead, one of these sets of numbers is wrong.

Wyman got this above chart directly from L & I. Neither Wyman nor L & I bothered to correct  for the double entry computer coding error, or the occupational disease computer coding error or the claims adjustment shortage error. Had all three of these adjustments been made it is certain that retro claims costs would be significantly higher than the non-retro costs.