Visualizing Insurance Losses

What’s in this post?
Discover a new way to visualize multi-dimensional data and track the performance of an insurance portfolio over time.

Introduction

Performance of a pool of insurance policies can be tracked with one important metric: loss ratio, which is given by dividing the amount paid for claims by the premium earned for these policies during a given time span.

(In reality, Loss Ratios can be broken down in many parts. Check our article on loss ratios!)

Cohort table — Loss ratios of an insurance portfolio

Some Insurance jargon

• Accident year: 12-month period for which incidents are taking place.
• Development lag: Years passed between an incident and the loss being paid. They exist because determining losses can be hard (e.g. legal disputes) or claims have to be paid over long periods of time (e.g. long-term disabilities).
• Insurance portfolio: Pool of insurance policies, usually covering the same type of risk (e.g. auto insurance).
• Loss triangle: Term used by actuaries to describe a table with cohort data.
• Incurred loss: Insurance company estimate of losses, paid and unpaid yet.

Visualizing loss ratios over time

Visualizing paid loss ratios developments over 30 years

In addition to tracking loss ratios, Insurers have to estimate what their losses could be in 10 years in order to figure out how much capital they need.

Incurred Loss Ratios Explained

Benchmarking model predictions

We have also added the ability to compare stochastic models used in the industry to determine reserves of capital. These predictions are the results of tens of thousands of simulations and give us confidence intervals on the probability of where ultimate losses are heading.

Interactive Data Manipulation

Time travel

Simply slide to explore data as if you were an observer of the past.

Explore details

Scroll to zoom, click & drag to see what you want to see.

Conclusion

We believe there are better ways to learn complex concepts with our visual cortex and with interaction for immediate feedback. The visualization shown in this post has more than 1,100 data points across several time dimensions and data types.

In order to achieve transparency to promote investing in insurance risk, we need to make even greater progress using the best tools web technologies and data science can offer.

Visit us https://ledgerinvesting.com/

Resources:
https://en.wikipedia.org/wiki/Cohort_analysis
https://d3js.org/
https://www.irmi.com/term/insurance-definitions/loss-triangle