Post on 18-Jan-2016
Motor Postcode Zoning Working Party
Duncan Anderson Roger MasseySaideh Charlton Chris SpillerDave Coughlan (Chair) Kim PapeMark Harrison James TanserStephen Jones Richhard Verrall
Aims
• Understand market
• Data availability
• Possible approach
Importance of zoning
• Review of postcodes
• 3rd biggest factor
• Needed to limit scope
• Theft only 10 - 15%
Market Analysis
Investigating market rates
• EL Systems Præmium• "Average" risk considered
for every UK post code district• 18 comprehensive quotes
obtained from 15 insurers• Some possible interactions also examined• Limited
– districts only– no direct writers– off screen discounts
Post code categorisation
InsurerNumber ofcategories
123456789
101112131415161718
1011121415161617171819191920202122
125
Post code categorisation
InsurerNumber ofcategories
Highest premium /lowest premium
123456789
101112131415161718
1011121415161617171819191920202122
125
2.121.972.293.302.202.182.192.672.122.162.701.932.313.172.173.132.762.26
CompanyInteraction 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Age / Sex Age / vehicle group
Area / vehicle group
Area / garaged Age / occupation
Use / mileage Occupation / use Age / marital status / sex
Interactions
Age Factor
17 2.52
18 2.05
19 1.97
20 1.85
21-23 1.75
24-26 1.54
27-30 1.42
31-35 1.20
36-40 1.00
41-45 0.93
46-50 0.84
51-60 0.76
60+ 0.78
Group 1 2 3 4 5 6 7 8 9 10 11 12 13
Factor 0.54 0.65 0.73 0.85 0.92 0.96 1.00 1.08 1.19 1.26 1.36 1.43 1.56
1.00
1.00
1.17
1.00
Group > 1 2 3 4 5 6 7 8 9 10 11 12 13Age v
17 1.36 1.64 1.79 2.09 2.27 2.42 2.56 2.65 3.27 3.71 4.08 4.36 4.8418 1.12 1.31 1.47 1.76 1.84 2.00 2.11 2.19 2.43 2.97 3.29 3.55 3.9019 1.08 1.30 1.46 1.63 1.82 1.91 2.02 2.11 2.53 2.88 3.30 3.35 3.6320 0.98 1.18 1.36 1.54 1.68 1.79 1.83 1.97 2.19 2.66 3.02 3.20 3.38
21-23 0.96 1.13 1.24 1.51 1.65 1.64 1.80 1.85 2.04 2.26 2.55 2.53 2.8924-26 0.82 0.99 1.10 1.31 1.43 1.52 1.51 1.64 1.81 1.93 2.13 2.22 2.4727-30 0.78 0.90 1.07 1.19 1.32 1.39 1.41 1.51 1.65 1.77 1.91 2.01 2.2431-35 0.63 0.78 0.86 0.99 1.09 1.17 1.22 1.32 1.42 1.54 1.66 1.71 1.8836-40 0.55 0.64 0.71 0.85 0.91 0.93 0.99 1.07 1.18 1.29 1.40 1.41 1.5341-45 0.51 0.61 0.66 0.79 0.88 0.88 0.94 0.99 1.09 1.15 1.29 1.31 1.4246-50 0.46 0.55 0.61 0.70 0.76 0.81 0.84 0.92 1.02 1.07 1.12 1.18 1.3151-60 0.40 0.49 0.56 0.64 0.68 0.71 0.78 0.82 0.90 0.99 1.02 1.12 1.20
60+ 0.43 0.52 0.55 0.67 0.72 0.73 0.78 0.83 0.93 0.98 1.04 1.11 1.25
CompanyInteraction 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Age / Sex Age / vehicle group
Area / vehicle group
Area / garaged Age / occupation
Use / mileage Occupation / use Age / marital status / sex
Interactions
Age Factor
17 2.52
18 2.05
19 1.97
20 1.85
21-23 1.75
24-26 1.54
27-30 1.42
31-35 1.20
36-40 1.00
41-45 0.93
46-50 0.84
51-60 0.76
60+ 0.78
Group 1 2 3 4 5 6 7 8 9 10 11 12 13
Factor 0.54 0.65 0.73 0.85 0.92 0.96 1.00 1.08 1.19 1.26 1.36 1.43 1.56
1.00
1.00
1.17
1.00
Group > 1 2 3 4 5 6 7 8 9 10 11 12 13Age v
17 1.36 1.64 1.79 2.09 2.27 2.42 2.56 2.65 3.27 3.71 4.08 4.36 4.8418 1.12 1.31 1.47 1.76 1.84 2.00 2.11 2.19 2.43 2.97 3.29 3.55 3.9019 1.08 1.30 1.46 1.63 1.82 1.91 2.02 2.11 2.53 2.88 3.30 3.35 3.6320 0.98 1.18 1.36 1.54 1.68 1.79 1.83 1.97 2.19 2.66 3.02 3.20 3.38
21-23 0.96 1.13 1.24 1.51 1.65 1.64 1.80 1.85 2.04 2.26 2.55 2.53 2.8924-26 0.82 0.99 1.10 1.31 1.43 1.52 1.51 1.64 1.81 1.93 2.13 2.22 2.4727-30 0.78 0.90 1.07 1.19 1.32 1.39 1.41 1.51 1.65 1.77 1.91 2.01 2.2431-35 0.63 0.78 0.86 0.99 1.09 1.17 1.22 1.32 1.42 1.54 1.66 1.71 1.8836-40 0.55 0.64 0.71 0.85 0.91 0.93 0.99 1.07 1.18 1.29 1.40 1.41 1.5341-45 0.51 0.61 0.66 0.79 0.88 0.88 0.94 0.99 1.09 1.15 1.29 1.31 1.4246-50 0.46 0.55 0.61 0.70 0.76 0.81 0.84 0.92 1.02 1.07 1.12 1.18 1.3151-60 0.40 0.49 0.56 0.64 0.68 0.71 0.78 0.82 0.90 0.99 1.02 1.12 1.20
60+ 0.43 0.52 0.55 0.67 0.72 0.73 0.78 0.83 0.93 0.98 1.04 1.11 1.25
Low
High
1.00
1.01
1.02
1.03
1.04
1.05
1.06
Vehicle Group
Area
High
Low
Example interaction
Loadings by postcode
• For each company, "average loading" for a district defined as premium for district divided by unweighted average of premiums across all districts
• Districts ordered by average"average loading"
• Average loading plotted against district
60%
80%
100%
120%
140%
160%
180%
District
Loadings by postcode
60%
80%
100%
120%
140%
160%
180%
District
Average Minimum Maximum
Loadings by postcode
Comparison of categorisations
244
274
307
343
385
432
484
543
609
685
770
299316334352369387404422440457475495518542632
0
20
40
60
80
100
120
140
Num
ber
of d
istr
icts
Company F
Company I
244
274
307
343
385
432
484
543
609
685
770
299316334352369387404422440457475495518542632
0
20
40
60
80
100
120
140
Num
ber
of d
istr
icts
Company F
Company I
Comparison of categorisations
275
311
325
352
396
439
525
606
318340
366385
426447
467521
555590
683734
0
50
100
150
200
Num
ber
of d
istr
icts
Company A
Company G
2/8
Comparison of categorisations
318
340
366
385
426
447
467
521
555
590
683
734
299330
357372
399426
442477
507549
591
0
50
100
150
200
Num
ber
of d
istr
icts
Company G
Company N
3/8
Comparison of categorisations
265
352
374
402
423
453
503
563
638
713
293302308314319326332343353361369382393405428442458475508633
0
50
100
150
200
250
Num
ber
of d
istr
icts
Company B
Company M
4/8
Comparison of categorisations
0
397
434
478
517
563
617
687
768
182208234284292299307315328342376409439471484495513530554579
0
20
40
60
80
100
120
140
160
Num
ber
of d
istr
icts
Company D
Company K
5/8
Comparison of categorisations
275
311
325
352
396
439
525
606
244259274290307324343363385407432457484513543575609647685726770
020
40
60
80
100
120
140
160
Num
ber
of d
istr
icts
Company A
Company F
6/8
Comparison of categorisations
275
311
325
352
396
439
525
606
265341352363374387402413423441453473503528563598638675713
0
20
40
60
80
100
120
140
160
Num
ber
of d
istr
icts
Company A
Company B
7/8
Comparison of categorisations
265
352
374
402
423
453
503
563
638
713
312353373386401425443460483501525550580614651685
0
20
40
60
80
100
120
140
160
Num
ber
of d
istr
icts
Company B
Company P
8/8
Comparison of categorisations
Data, Method and Results
Information available
• Complex
• Appropriate
• Up to date
• Higher level view
Data
• Internal data
• External
• Combined/Grouped
Interesting views
• Drive around and look
• Social deprivation index
• Subjective views
Methods
• Size of exercise
• Consider residual risk
• Three main types of approach- Smoothing- Credibility- Market related
Smoothing
“Based on underlying assumption thatareas “close” each other have similar risk”
Adjacency?
Credibility
“The risk associated with a postcode is a mixture of its own experience and some other estimate of expected”
Appropriate?
Market Related
“everybody doing it, so I should too”
Realistic?
Approach
• Smoothing - “weighted distance”
- “Spatial models”
• Credibility - “Credibility Method”
• Market - “Average Market”
- “External provider”
• Other - “Modern Heuristic”
Approach
• Split data 70/30
• Standardise data
• Use residual risk to zone
squares as test statistic
Simple results
• All proved better than null model
• Little to choose between results
• Combination of “credibility and
weighted distance” is the best
Comparison of methods
Implementation
• Solid risk cost
• Improved
understanding
Implementation - issues
• Competitive position
• Existing customers
• Systems
Conclusion
• Market is not uniform in
approach
• Suggest approach as
mixture of
methods
Motor Postcode Zoning Working Party
Duncan Anderson Roger MasseySaideh Charlton Chris SpillerDave Coughlan (Chair) Kim PapeMark Harrison James TanserStephen Jones Richhard Verrall