Reform Retro!

The Wyman Report noted three major errors in calculating retro subsidies. These were the Double Entry Coding Error, the Occupational Disease Mis-assignment error and the 45 Month adjustment limitation error. Each of these errors increased the size of retro subsidies. But Wyman failed to correct the loss ratios for any of these errors. We will therefore finish the calculations that should have been done by Wyman.

 

 Correcting for Error # 1: The computer coding error:

 

Chart 1: L & I Chart from WCAC meeting presentation, August 2008

The above chart from the August 2008 WCAC meeting minutes (http://www.lni.wa.gov/ClaimsIns/Files/Wcac/Minutes/2008/WcacMtgMin20080825.pdf), was produced by L & I before any of the 3 errors were discovered by Wyman. The chart is intended to show that Retro Claims/ Loss ratios are 17.5% less than non-Retro claims. This chart is now known to be factually incorrect because it was based on at least three errors including the double entry computer coding error and the occupational disease mis-assignment error. 

 

The next chart is from L & I and is based on the L & I assumption of a 2 ½ % over-estimation due to the double entry coding error and about a 7 ½ % difference due to the occupational disease error and a one percent difference due to the 45 month limitation error. Using their estimate, retro refunds were about 45% too high (compared to my estimate that they were 100% to 150% too high). Thus, if retro subsidies were $100 million per year, L & I is claiming that the refunds were only $45 million per year too high.

 

 Chart 2: L & I Chart from WCAC meeting presentation, September 2009

There is a problem in comparing the above two loss ratio charts. This second chart  (deliberately?) does not give actual dates, which makes it difficult to compare to the older “pre-correction chart. Nevertheless the shape of the last Retro “V” makes it possible to line up the graphs. But when we do this, it adds more questions than it answers.

 

For example, it is understandable that the non-retro loss ratios went down (because the claims estimates were lowered by 20% due to the elimination of the computer coding double entry error. This adjustment lowered the non-retro claims and thus lowered the non-retro loss ratios. However, the retro loss ratios should have gone up (because the double entry coding error had no effect on retro claims, but 45 Month adjustment correction should make the Retro Loss Ratio go up a little and the Occupational disease correction should make the Retro Loss Ratio shoot up a lot).

 

So why did the retro loss ratios fall almost as much as the non-retro loss ratios in the above chart???

 

The October 1, 2001 Retro Loss Ratio, shown in this chart as 120%,  magically lowers to 80% in the next chart!!! (This of course is not mathematically possible which is why this is yet another “smoking gun” that someone at L & I is cooking the books).

 

The only possible answer is that someone at L & I  is fudging the data to try to make it look like retro loss ratios are lower than non-retro loss ratios.

 

What would happen if we started with the data from the 2008 chart and adjusted it for the three errors one at a time?

 

Table 1: Comparing 2008 Chart Loss ratios to L & I Website Loss Ratios

 

What would happen if we started with the data from the 2008 chart and adjusted it for the three errors one at a time?

 

Table 1: Comparing 2008 Chart Loss ratios to L & I Website Loss Ratios

 

The following table and chart shows the Retro and Non-retro Loss Ratios from the August 2008 Chart with the estimated PAF’s (not adjusted for any of the 3 errors):

Enrollment Period Starting	Retro Loss Ratio 2008 data	Non-Retro Loss Ratio 2008 data	PAF’s From 2008 data	Retro Premiums From L & I Website ($ millions)	Retro Claims From L & I Website ($ millions)	Retro Loss Ratio From site	PAF’s From  L & I site	Refund from L & I site ($ millions)
1/1/2002	1.10	1.30	0.85	73	73	1.00	0.66	13
4/1/2002	.85	1.16	0.73	35	30	.86	0.72	8
7/1/2002	.85	1.04	0.82	315	299	.95	0.82	73
10/1/2002	.78	0.92	0.85	26	25	.95	0.88	3
2002								$97 M
1/1/2003	.80	0.88	0.91	100	107	1.07	.96	13
4/1/2003	.72	0.90	0.80	49	50	1.01	.99	7
7/1/2003	.66	0.84	0.79	400	364	0.91	.99	102
10/1/2003	.76	0.80	0.95	35	38	1.08	1.04	4
2003								126M
1/1/2004	.64	0.76	0.84	102	110	1.08	1.07	14
4/1/2004	.56	0.74	0.76	40	37	0.92	1.07	8
7/1/2004	.58	0.72	0.81	480	460	0.96	1.09	108
10/1/2004	.64	0.70	0.91	40	42	1.12	1.12	3
2004								131M
1/1/2005	.58	0.68	0.85	122	120	.98	1.09	25
4/1/2005	.40	0.66	0.61	35	26	.74	1.11	15
7/1/2005	.50	0.64	0.78	532	493	.93	1.12	134
10/1/2005	.60	0.62	0.97	36	41	1.14	1.09	2
2005								176M
1/1/2006	.56	0.61	0.92	121	108	1.06	1.08	15
4/1/2006	.57	0.60	0.95	4	4	1.06	1.07	1
7/1/2006	.48	0.58	0.83	567	553	0.98	1.08	117
10/1/2006	?	>	?	28	31	1.13	1.06	1
2007								134M

 

The first problem with the above table is that the Loss Ratios posted on the L & I website are much higher than the loss ratios shown on the August 2008 L & I graph. As the 2008 graph was posted before the errors were known and as the website data is also from 2008 or sooner reported data, there is no way to account for this difference. The data should have matched, but in most cases does not. If fact, in some cases, the Retro loss ratios on the L & I website are more than twice as high as the loss ratios indicated on the 2008 L & I graph.

 

Even worse, the non-retro claims and loss data are not even listed on the L & I website. So there is no way to double check either the non-retro data or the PAF’s calculated from it. This makes it difficult to have any confidence in anything reported by L & I.

Nevertheless, for the sake of consistency, we will use the 2008 graph data to show how the 3 errors reported by Wyman should have changed the graph. The next Table corrects for the double entry coding error by lowering the Non-retro loss ratio by 10% and leaving the retro loss ratio untouched.

 

Table 2: Loss Ratios Adjusted for Coding Error, OD Error and 45 M Error

Enrollment Period Starting	Retro Loss Ratio 2008 data	Non-Retro Loss Ratio 2008 data	Non retro Loss ratio Adj for coding error (-10%)	Non retro Loss ratio Adj for OD error  (-10%)	Retro Loss ratio adj for OD & 45M errors (+15%)	PAF’s Adjusted for all three errors
1/1/2002	1.10	1.30	1.17	1.06	1.27	1.20
4/1/2002	.85	1.16	1.05	.95	1.08	1.14
7/1/2002	.85	1.04	0.95	.85	1.08	1.27
10/1/2002	.78	0.92	0.83	.75	.92	1.23
2002
1/1/2003	.80	0.88	0.79	.71	.94	1.32
4/1/2003	.72	0.90	0.81	.73	.82	1.12
7/1/2003	.66	0.84	0.76	.69	.75	1.09
10/1/2003	.76	0.80	0.72	,65	.88	1.35
2003
1/1/2004	.64	0.76	0.68	.61	.73	1.20
4/1/2004	.56	0.74	0.67	.60	.65	1.08
7/1/2004	.58	0.72	0.65	.59	.67	1.13
10/1/2004	.64	0.70	0.63	.57	.73	1.28
2004
1/1/2005	.58	0.68	0.61	.55	.67	1.22
4/1/2005	.40	0.66	0.59	.54	.46	0.85
7/1/2005	.50	0.64	0.58	.52	.58	1.12
10/1/2005	.60	0.62	0.56	.50	.69	1.38
2005
1/1/2006	.56	0.61	0.55	.50	.65	1.30
4/1/2006	.57	0.60	0.54	.49	.66	1.34
7/1/2006	.48	0.58	0.52	.47	.55	1.17

 

The fourth column in the above table takes care of the $10 to $15 million per year problem with the double entry coding error. The fifth column adjusts for the $30 million occupational disease problem. This is a little harder because this is a growing problem whereas the first problem was a consistent problem. While the second problem is currently about 3 times bigger than the first problem (according to L & I), we will only adjust for it as if it were twice as big in order to take into account the fact that it used to be much smaller. Thus, the next adjustment will be to lower non-retro loss ratios by 10% to take into account the fact that they were artificially inflated by essentially doubling the occupational disease claims attributed to non-retro programs.

 

The Wyman report stated (based on unverified information from L & I), that adjusting for the OD error would reduce retro refunds by about 20%. In the September 21 data, L & I estimated that adjusting for this error would reduce 2010 projected retro refunds from $120 million to $90 million (which is a 25% reduction). But the actual effect of this mis-assignment of occupational disease claims is not only to increase retro claims (which is what lowers retro refunds), but also to reduce non-retro claims (thereby reducing the non-retro loss ratio). A reasonable estimate of the effect of this adjustment is that retro loss ratios will be raised by about 10% and non-retro loss ratios will be lowered by about 10%. Thus, in the sixth column, we must raise the Retro loss ratios by 10% to take into account that they were artificially lowered by failing to include their fair share of occupational disease unknown claims. In addition, we need to add 5% to the retro loss ratios to take into account the erroneous 45 month limitation on claims adjustments.

Once all of these corrections are made, we can calculate more accurate PAF’s done by dividing the retro loss ratios by the non-retro loss ratios.

 

Error #3: Correcting for the 45 Month Limitation

Currently, there are only 3 retro adjustments with the final one occurring at 45 months. No other State has such a short final adjustment period. Most States go out 10 years or more. The Wyman Report noted that if there was a 4^th Adjustment at 60 months, it would reduce retro subsidies by 4% and a 5^th adjustment would reduce retro subsidies one more percent.

This means that currently retro loss ratios are 5% too low. (Since non-retro is based on actual costs, there is no time limit to non-retro and the non-retro loss ratios are not affected by the 45 month retro adjustment limitation error). The following table increases retro loss ratios 5%:

 

Adjusting for low data cells

Note that the Retro loss ratios in the above chart bounce around quite a bit. This is because the second and fourth quarter Retro enrollment periods are much smaller than the first and third quarter Retro enrollment periods. In fact, the third quarter Retro enrollment period accounts for over half of the total (Including all of the BIAW enrollments). Eliminating every other Retro cell produces a more accurate comparison.

 

The following Table corrects for all three errors and eliminates the small data (even numbers) cells

Period Starting	Non retro Loss ratio corrected	Retro Loss ratio corrected	PAF’s Adjusted for all three errors
1/1/2002	1.06	1.27	1.20
7/1/2002	.85	1.08	1.27
1/1/2003	.71	.94	1.32
7/1/2003	.69	.75	1.09
1/1/2004	.61	.73	1.20
7/1/2004	.59	.67	1.13
1/1/2005	.55	.67	1.22
7/1/2005	.52	.58	1.12
1/1/2006	.50	.65	1.30
7/1/2006	.47	.55	1.17

 

 

Below is the chart generated from the above table:

 

Chart 4: Retro versus Non-retro Loss Ratio Comparison

Before Loss Development Adjustment (after adjusting for all three major errors)

The above chart confirms that, after adjusting for all known errors, non-retro programs have had lower loss ratios than retro programs since at least 2002. In fact, they have always had lower loss ratios as is confirmed by the long term difference of retro claims costs being at least 10% higher than non-retro long term claims costs. This chart is consistent with all the other research concluding that retro claims costs are slightly higher than non-retro claims costs.

 

Retro versus Non Retro Performance Adjustment Factors (PAF) corrected for all three errors:

The above chart shows the corrected PAF’s for 2002 though 2006. Thus, during this time period, retro agencies had costs that were about 10% to 20% greater than non-retro agencies.

 

Thus, instead of having been given refunds of $1.3 billion dollars (20% of premiums) during the past 10 years, retro agencies should have been charged additional assessments of $0.7 billion (10% of premiums) during the past 10 years. The combined cost to tax payers was $2 billion dollars or $200 million dollars per year for these “errors.”

 

So why was it that neither L & I or Wyman bothered to correct the data that they knew was wrong and instead put out a report in August 2009 based upon data which they knew to be factually incorrect??? The only possible reason is that L & I is still trying to hide the truth about Retro from the public. Thus, the charts in this section are yet another “smoking gun” that at least someone at L & I is fudging the data in order to hide the fact that retro programs do NOT save money and never have.