Transcript Experience Rating Plan Off

Experience Rating
Current Challenges
Presented by
Tony DiDonato, NCCI, Inc.
2003 CAS Seminar on Ratemaking
San Antonio, Texas
March 28, 2003
WCP-2
NCCI Perspective - Outline
I. ER Basics
II. Off-Balance in NCCI States
III. ERA – Experience Rating Adjustment
IV. Performance Testing Results for NCCI States
2
NCCI Plan - ER Basics
The formula:
Actual / Expected
Some Qualifications:
• Premium thresholds
• Actual Losses limited
• Mods limited
• ERA
Ap + Ae(W) + Ee(1-W) +B
E+B
A = Actual
E = Expected
p = primary
e = excess
B = Ballast
W = Weight
3
Countrywide* Off-Balance Factors
1.00
0.989
0.99
0.98
0.964
Off-Balance
0.97
0.954
0.96
0.95
0.941
0.937
0.94
0.924
0.93
0.931
0.926
0.924
0.923
1997
1998
1999
0.92
0.91
0.90
1993
1994
1995
1996
2000
2001
2002p
Year
* NCCI states
4
Unanticipated Frequency and Severity:
How is Off-Balance Impacted?
Theoretical Off-balance
Total Unanticipated
% Change in Frequency
<-- Total Unanticipated % Change in Severity -->
-20%
-12%
-4%
0%
4%
12%
20%
-20%
-12%
-4%
0%
4%
12%
20%
0.900
0.920
0.940
0.951
0.961
0.981
1.001
0.916
0.938
0.960
0.970
0.981
1.003
1.025
0.931
0.955
0.978
0.990
1.002
1.025
1.049
0.939
0.964
0.988
1.000
1.012
1.036
1.061
0.947
0.972
0.997
1.010
1.022
1.047
1.073
0.963
0.989
1.016
1.029
1.043
1.069
1.096
0.978
1.006
1.034
1.049
1.063
1.091
1.119
Theoretical off-balance increases by about 0.002 - 0.003 for every percentage point of unanticipated severity change.
Theoretical off-balance increases by about 0.0025 - 0.0035 for every percentage point of unanticipated frequency change.
5
Sensitivity of Off-Balance to:
Total Losses vs. Primary/Excess Split
Theoretical Off-balance
Total Unanticipated
% Change in
Total Losses
<-- Total Unanticipated Percentage Change in D-ratio -->
-20%
-5%
-1%
0%
1%
5%
20%
-20%
0.843
0.915
0.934
0.939
0.944
0.963
1.035
-5%
0.871
0.956
0.979
0.985
0.990
1.013
1.098
-1%
0.878
0.967
0.991
0.997
1.003
1.027
1.116
0%
0.880
0.970
0.994
1.000
1.006
1.030
1.120
1%
0.882
0.973
0.997
1.003
1.009
1.033
1.124
5%
0.890
0.984
1.009
1.015
1.022
1.047
1.141
20%
0.917
1.025
1.054
1.061
1.068
1.097
1.205
Theoretical off-balance increases by about 0.002 - 0.004 for every percentage point of unanticipated losses.
Theoretical off-balance increases by about 0.005 - 0.007 for every unanticipated additive percentage point in the D-ratio.
6
D-ratio Changes Over Time
• The primary change is due to severity trend
• Since severity generally increases over time, the
d-ratio generally decreases
• ERA allows for a trended split point which would
tend to stabilize the d-ratio, but trend has not yet
been used on the split point
7
Total Losses and Primary/Excess Split:
Forecast vs. Actual Results
Intrastate Risks – NCCI States
Experience
Rating Years
Total Actual /
Total Expected
Primary Actual
/ Total Actual
(Actual D-ratio)
Primary Expected Difference Between
/ Total Expected
Actual D-ratio Average
(Expected D-ratio)
Expected D-ratio
Mod
1996
0.84
0.30
0.28
0.02
0.96
1997
1998
0.85
0.87
0.30
0.30
0.29
0.30
0.02
0.00
0.95
0.95
1999
0.84
0.27
0.28
(0.01)
0.94
2000
0.88
0.26
0.28
(0.02)
0.94
2001
0.89
0.25
0.28
(0.02)
0.95
8
Experience Rating Plan Off-Balance–
Historical Analysis
• Countrywide (NCCI states) Off-Balance has increased
in each of the last three years.
• Total actual losses that were less than total expected
losses is the major reason for the relatively low OffBalances from 1996-1999.
• While the split between claim frequency and severity
did not drive the low Off-Balances, it did dampen the
Off-Balance increases in the last three years.
• Unanticipated claim frequency has slightly more of an
impact than does unanticipated severity on Mods.
9
What Should the Experience Rating
Plan Off-Balance Be ?
• Manual loss ratios for the smallest premium sizes and
for unrated risks are higher than the all-risk average.
• If the off-balance is 1.00, then there is no standard
premium price differential between experience rated
and unrated risks.
• Having an off-balance less than 1.00 can partially
address the difference.
• The indicated standard premium level is still correct
even if there is a net off-balance.
10
$500,001
and Above
$300,001 $500,000
$200,001 $300,000
$150,001 $200,000
$100,001 $150,000
$50,001 $100,000
$25,001 $50,000
$10,001 $25,000
$5,001 $10,000
$2,001 $5,000
$1,501 $2,000
$1,001 $1,500
$501 $1000
$0 - $500
Developed Manual Loss Ratios by
Policy Manual Premium
(State X, Five Recent Policy Years)
200%
160%
120%
80%
40%
0%
11
Developed Manual Loss Ratios
(State X, Five Recent Policy Years)
120%
100%
80%
60%
40%
20%
0%
1995
1996
1997
Unrated
1998
Intrastate
1999
1995-1999
Interstate
12
Simplified Hypothetical Illustration of
Experience Rated vs. Unrated
Impact on Off-Balance
Unrated Risks
Rated Risks
Average
% of Manual Premium
10.0%
90.0%
100.0%
Manual Loss Ratio
83.6%
55.3%
58.1%
Actual Loss / Expected Loss
1.439
0.951
1.000
Mod Needed to Equalize
Standard Premium Loss Ratios
1.000
0.661
0.695
Mod if Unrated Risks
are Subsidized
1.000
0.951
0.956
13
Summary of ERA Changes
ERA was designed to increase the incentive for employers
to report small med-only claims and to improve the
performance of the Plan. This was accomplished by the
following three changes to the Plan:
(1) Using only 30% of med-only claims in the Experience
Rating formula
(2) The Weighting Value (W) was increased
(3) The primary/excess split point (currently $5,000) will
be adjusted over time
The effective date of ERA varies by state. The earliest effective date
is 7/1/98, which is applicable in several states.
14
ERA Impact on Med-Only Losses
• Compared changes in the proportion of med-only
claims in states adopting ERA vs. states that had not
• Changes in the average severity of med-only claims
were also reviewed
• This review did not reveal a significant impact on the
reporting of med-only claims due to ERA
15
Med-Only Changes in Late 1990s
From policy period beginning 11/95 to policy period beginning 11/98
ERA States
State
A
B
C
D
E
F
G
H
I
% Chng in
Med-Only
+ 2.3%
- 0.1%
- 0.7%
- 0.7%
+ 1.2%
+ 1.2%
+ 1.8%
+ 0.3%
+ 1.6%
Non-ERA States
% Chng in
Med-Only
Severity
+ 6.1%
+ 23.5%
+ 25.3%
+ 25.9%
+ 17.9%
+ 12.4%
+ 14.4%
+ 33.2%
+ 21.5%
State
1*
2
3
4
5
6
7
8
9
% Chng in
% Chng in Med-Only
Med-Only
Severity
+ 0.1%
+ 14.6%
+ 1.1%
+ 21.0%
+ 8.5%
+ 15.1%
+ 1.2%
+ 24.0%
+ 3.7%
+ 9.9%
+ 3.4%
+ 21.9%
- 4.4%
- 19.5%
+ 1.3%
+ 18.9%
+ 1.3%
+ 26.3%
* State 1 reflects a 2-yr change
16
Performance Testing: GERT vs. ERA
• 1998 mods were calculated for intrastate risks in the 9
states adopting ERA on 7-1-98
• Initially all rated under GERT, then all re-rated under ERA
• ERA showed slightly more accurate results
17
Performance Testing: GERT vs. ERA
GERT
Quintile
Stratum
Determined
by Mod
(1)
Actual
Subsequent
Losses
Divided by
Manual
Expected
(2)
1
0.70
2
3
4
5
Mean or Total
Squared Deviation
from Mean of (2)
x 10,000
(3)
Squared Deviation
from Mean of (4)
x 10,000
(5)
0.93
0.85
0.94
1.09
1.48
900
225
36
81
2,304
1.00
1.04
1.07
1.12
121
16
0
9
64
1.00
3,546
1.04
210
0.97
1.04
1.04
1.07
1.09
64
1
1
4
16
1.05
86
Test Statistic: (5)/(3)=
ERA
Actual
Subsequent
Losses
Divided by
Modified
Expected
(4)
1
0.72
2
3
4
5
0.88
0.93
1.08
1.44
784
144
49
64
1,936
Mean or Total
1.00
2,977
Test Statistic: (5)/(3)=
0.059
0.029
18
Brief Summary of Quintile Testing
• Col (2): The actual losses are unlimited losses from WCSP data,
generally at a 2nd report or subsequent.
• The rates/loss costs in effect during the appropriate time period were
used as a proxy for expected losses. The column (2) ratios have been
normalized to 1.00 to minimize differences between the actual and
expected losses related to development, expenses, etc.
• Col (3) shows the deviation of each quintile group from the overall total.
• Col (4) reflects the normalization from Col (2), but after application of the
Mod (in the denominator) the results were NOT re-normalized. This has
no impact on the result in Col (5). The mean value shown is an
intermediate step in the calculation and has no particular meaning.
• Col (5) shows the deviation of each quintile group from the overall total.
• The test statistics shown at the bottom of each analysis are key. A
statistic less than 1.00 is expected from an Experience Rating Plan.
Lower values of the statistic indicate better performance.
19
Performance Testing: GERT
GERT
Individual State Quintiles
Actual / Modified Expected Loss
1.6
1.4
1.2
1.0
0.8
0.6
Correlation = 0.654453
0.4
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
Actual / Expected Loss
20
Performance Testing: ERA
ERA
Individual State Quintiles
Actual / Modified Expected Loss
1.6
1.4
1.2
1.0
0.8
0.6
Correlation = 0.543165
0.4
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
Actual / Expected Loss
21
Countrywide* Quintile Testing Results
• Includes Interstate and Intrastate Risks
• Policy periods 7/1/97–6/30/98 and 7/1/98–6/30/99
were reviewed
• Two quintile groupings were tested:
– quintile groups which equalized risk count
– quintile groups which equalized expected losses
* NCCI states
22
Countrywide* Quintile Testing Results
ALL RISK SIZES
Policy Effective Period 7/1/97 - 6/30/98
Quintile
Stratum
Determined
by Prior Mod
(1)
Actual
Subsequent
Losses
Divided by
Manual
Expected
(2)
Squared Deviation
from Mean of (2)
x 10,000
(3)
Actual
Subsequent
Losses
Divided by
Modified
Expected
(4)
Squared Deviation
from Mean of (4)
x 10,000
(5)
Percentage of
Expected Loss
in Quintile
(6)
Percentage of
Risks Count
in Quintile
(7)
20.0%
20.0%
20.0%
20.0%
20.0%
100.0%
8.9%
20.0%
30.3%
18.3%
22.6%
100.0%
33.4%
14.1%
13.8%
21.4%
17.3%
100.0%
20.0%
20.0%
20.0%
20.0%
20.0%
100.0%
Expected Losses Uniformly Distributed Among Quintiles
1
2
3
4
5
Mean or Total
0.66
0.82
0.95
1.13
1.43
1.00
1,129
324
28
163
1,880
3,524
1.04
1.03
1.07
1.14
1.13
1.09
Test Statistic: (5) / (3) = 0.028
26
34
3
22
14
100
Risk Count Uniformly Distributed Among Quintiles
1
2
3
4
5
Mean or Total
* NCCI states
0.72
0.88
0.99
1.13
1.47
1.00
772
141
0
173
2,234
3,319
1.04
1.05
1.10
1.12
1.13
1.09
Test Statistic: (5) / (3) = 0.022
29
16
1
9
17
73
23
Countrywide* Quintile Testing Results
ALL RISK SIZES
Policy Effective Period 7/1/98 - 6/30/99
Quintile
Stratum
Determined
by Prior Mod
(1)
Actual
Subsequent
Losses
Divided by
Manual
Expected
(2)
Squared Deviation
from Mean of (2)
x 10,000
(3)
Actual
Subsequent
Losses
Divided by
Modified
Expected
(4)
Squared Deviation
from Mean of (4)
x 10,000
(5)
Percentage of
Expected Loss
in Quintile
(6)
Percentage of
Risks Count
in Quintile
(7)
20.0%
20.0%
20.0%
20.0%
20.0%
100.0%
8.9%
19.7%
30.4%
19.1%
22.0%
100.0%
33.7%
14.9%
11.8%
21.9%
17.7%
100.0%
20.0%
20.0%
20.0%
20.0%
20.0%
100.0%
Expected Losses Uniformly Distributed Among Quintiles
1
2
3
4
5
Mean or Total
0.67
0.82
0.96
1.12
1.43
1.00
1,074
329
16
138
1,826
3,382
1.07
1.04
1.10
1.12
1.12
1.10
Test Statistic: (5) / (3) = 0.016
10
30
0
7
8
55
Risk Count Uniformly Distributed Among Quintiles
1
2
3
4
5
Mean or Total
* NCCI states
0.72
0.89
1.00
1.13
1.45
1.00
765
112
0
168
2,076
3,121
1.05
1.07
1.12
1.12
1.13
1.10
Test Statistic: (5) / (3) = 0.015
23
6
5
6
8
48
24
Individual State Quintiles
ALL RISK SIZES
Policy Effective Period 7/1/97 - 6/30/98
Actual / Modified Expected (normalized by state)
1.6
1.4
1.2
1
0.8
0.6
Correlation = 0.620352
0.4
0.25
0.5
0.75
1
1.25
1.5
1.75
2
2.25
Actual / Expected
25
Individual State Quintiles
ALL RISK SIZES
Policy Effective Period 7/1/98 - 6/30/99
Actual / Modified Expected (normalized by state)
1.6
1.4
1.2
1
0.8
0.6
Correlation = 0.502986
0.4
0.25
0.5
0.75
1
1.25
1.5
1.75
2
2.25
Actual / Expected
26
Quintiles by Risk Count
Test Results by Size of Risk
Expected
Loss Range
Quintile* Test
Statistics for
Policy Period
7/1/97 - 6/30/98
Quintile* Test
Statistics for
Policy Period
7/1/98 - 6/30/99
$ 0 - $ 7,500
0.151
0.083
$ 7,500 - $ 15,000
0.051
0.060
$ 15,000 - $ 40,000
0.063
0.067
$ 40,000 - $ 100,000
0.040
0.027
$ 100,000 +
0.039
0.015
* The risk count underlying each policy period and range is uniformly distributed among quintile stratum.
27
Quintiles by Expected Losses
Test Results by Size of Risk
Expected
Loss Range
Quintile* Test
Statistics for
Policy Period
7/1/97 - 6/30/98
Quintile* Test
Statistics for
Policy Period
7/1/98 - 6/30/99
$ 0 - $ 7,500
0.155
0.080
$ 7,500 - $ 15,000
0.053
0.062
$ 15,000 - $ 40,000
0.064
0.066
$ 40,000 - $ 100,000
0.039
0.024
$ 100,000 +
0.040
0.009
* The expected loss volume underlying each policy period and range is uniformly distributed among quintile stratum.
28
Performance Testing: Next Steps
• Currently reviewing the 30% factor for Med-only claims
under ERA
• Reviewing treatment of small risks
• Alternative performance measures are being reviewed
• Continued monitoring of ER Plan and performance
29