Machine learning algorithm vs actuarial science:

In today’s fast paced world of technology, an insurance industry executive or aspiring leader has to think about the future, investor expectations, scaling and leveraging technology, and this new question of “Machine learning algorithm vs actuarial science?”

ML vs. AS – What’s the difference?

Machine learning (ML) is a type of artificial intelligence (AI) that provides computers with the ability to learn without being explicitly programmed. Machine learning focuses on the development of computer programs that can teach themselves to grow and change when exposed to new data.

Actuarial science (AS) is the discipline that applies mathematical and statistical methods to assess risk in insurance, finance and other industries and professions. Actuaries are professionals who are qualified in this field through intense education and experience.  Actuaries use statistical inference to help predict the future.   Actuaries help underwriters understand the “why” in the data, they bring tremendous value to underwriters and organizations that are trying to understand risk via the statistical data.

Machine learning algorithm vs actuarial science is a continuous debate by insurance industry leaders.  Eventually an Actuary is going to team up with a Data scientist to build an algorithm and tool that can do both.  Until then – we will focus on the idea that these two fields compete with each other.

Do underwriters need to know the why? If they can just get the result via ML algorithm?

Would you or your shareholders trade optimal return on risk or equity for increased scale, relevance, and speed?  The question of this generation of insurers/reinsurers.

Is making a 10% ROE using traditional actuarial science sustainable if you can make 8-9% ROE using ML algorithm and reduce your expenses and scale at the same time?  What if the ML algorithm produced the same ROE ~10%?

These examples highlight a fundamental question – does the actuarial process provide enough value to support the time, costs, and restraints? Or should industry executives be focusing on using a ML algorithm to build a similar return & leverage the technology for increased scale and relevance to customers/brokers/the market.

If not, are you willing to return a Beta of the underwriting result using algorithm vs. generating alpha using actuary?

Benefits of the Machine Learning algorithm approach

Expense savings

If less actuaries are required, the speed is improved and underwriters can scale their portfolios easier – the more expense savings can be found.

If these savings are used to hire more claims staff.  The money saved in acquisition expenses used to evaluate the price of the risk (actuarial salaries) could be reallocated to the claims staff/resources.  As discussed here, we believe claims as one of the best ways to allocate resources to improve results – focus should be on the loss costs side.  This should be the focus of the insured and insurer – and how we add value to customers over time.

By hiring additional claims expertise, the insurer would have savings to the bottom line.  For example, the impact of saving $2-5m on the loss ratio is comparable to writing $10-25m of premium- depending on product line loss ratio.  Increased claims staff will help you when you look at the impact to the bottom line.

Time speed of analysis

If the algorithm is processing the submission information, the speed of the transaction and analysis can improve.  Therefore, offering improved turnaround time to underwriters for coverage and risk analysis.


The nature of ML algorithm will allow for increased scale and growth for insurers/reinsurers.  This scale provides a host of benefits – as mentioned helps with expenses and expense ratios.  The scale will also help you maintain preferred relationships with core customers and the ability to absorb shock losses.

Is making $30m in margin on a $100m portfolio as good or better than making $50m in margin on $150m portfolio?


We predict this debate will go on for quite some time, as the industry is historically slow to embrace change.  However, we predict the following to occur:

  • Insurance industry will need Data engineers
  • Insurance industry will need Data scientists
  • Insurance industry will need less actuary only resources
  • Software Developers will build Platforms that will produce easy to use output for non data scientists
  • The data and employees that Insurers and Reinsurers have will become their most valued asset to generate Alpha
  • We predict these new tools will produce different types of future Insurers/Reinsurers, along with different expectations from investors.
    • Two different types of insurers
      • Quick quote and small lines – that will produce a beta return of market
      • Slow quote larger lines uses Actuarial science – generates alpha
    • Two different types of reinsurers
      • Quick quote and small lines – that will produce a beta return of market
      • Slow quote larger lines uses Actuarial science- generates alpha
  • Underwriters will continue to sell product, evaluate risk,  and change coverage – however will have to become versed in the data science field

Related Articles:

About the author

Arnold Smith

Leave a Comment