How digitalization supplants old insurance models
Today, credit scores, such as FICO, are a seemingly intractable and fundamental component of insurance pricing models in most states. However, these are likely to disappear, and faster than most insurance carriers are prepared or planning for, thanks to several trends.
First, credit scores fluctuate with the economy, so if the US is to experience a market correction (which economists forecast for 2019), the power of credit to predict losses will degrade and put carriers’ profitability at risk. Second, indefensible biases are catching the eyes of state legislators who are regulating against credit-based insurance and models that lack transparency, like FICO. And lastly, alternative data products are emerging that compete with the predictive power of credit, where there were few before.
Insurers can do better than credit
Credit scores are intractably linked to the economy. As people are doing well, earning more, buying more on credit, and paying off debt, their credit scores go up. However, when market conditions change, as was the case with the Great Recession, FICO credit scores declined by three points on average in one year, 2009.
Dynamic credit scores that fluctuate not based on changes to the underlying risk, but rather on the macro-economy, disrupt actuarial models built on the premise that people with poor credit are more likely to experience losses that justify higher premiums.
To illustrate how out of whack credit-based insurance is when it comes to pricing risk, in a 2017 study published by the Arkansas State Insurance Department found that a person with poor credit pays two times more annual premium (an additional $1000) for personal automobile insurance on average than a person with excellent credit and a DUI.
In the near future, carriers’ ability to predict losses and price risk using credit scores could deteriorate quickly and catastrophically if one of two conditions occurs: If credit scores become extremely volatile among random groups of people or, more likely, the population of people with low credit is so large that a pricing model loses variation. The latter is frighteningly possible, as the New York Federal Reserve Bank documents a generation of millennials default on auto loans, rack up student debt that is unabsolvable in bankruptcy, and struggle to gain employment with limited job experience in a tight labor market.
Regulators are losing patience
In three states, California, Hawaii, and Massachusetts, it is illegal to provide credit-based insurance prices. This is because in most studies that are not published by FICO, researchers have found that credit scores disproportionately negatively affect minorities, immigrants, and young people, making it more expensive for these groups to gain access to affordable insurance.
In an era of increasing scrutiny of machine learning models in insurance that can not be deconstructed or explained to consumers, the long-standing mystery of what components go into a credit score and in what combination, will likely draw even more negative attention by regulators to credit-based insurance.
New technologies are a threat
New technologies are also threatening the long-term viability of credit-based insurance. Carriers are increasingly seeking out data and building predictive models that will prove more powerful and profitable, even during periods of economic volatility.
For example, my company, ODN has shown it is possible to extend more policies at more affordable rates to people with poor credit, by pricing risk based on where people drive, rather than who they are.
To remain competitive and profitable in the long-term, underwriting and actuarial teams need to pay attention to the dynamics of credit-based insurance today and plan for a future that simply may not include FICO.
Carriers should be asking, what will happen to our pricing models if economic conditions or regulators make credit-based insurance irrelevant? What new technologies need to be in place to continue pricing risk and remain competitive in a world without FICO?
Asking these questions is an important exercise now, before it is too late.