by. Dan Price, Twenty Twenty Analytics
Behavioral finance is an approach to finance based on the observation that psychological variables affect and often distort individuals’ decision making. Behavioral finance is driven by a number of different behavioral biases. There are many different types of behavioral biases, but the common themes are that people tend to be somewhat illogical which fuels impulsive, irrational decisions.
For example, studies have shown that when presented with two independent proposals:
- Gain Proposal
- Receive a $500 guaranteed, or;
- 50/50 chance of receiving $1,000 or receiving nothing
- Loss Proposal
- Lose $500 guaranteed, or;
- 50/50 chance of losing $1,000 or losing nothing
The expected value for each option is $500, so a rational person would be indifferent to the choices within each proposal. However, individuals with a Gain Proposal overwhelmingly took the guaranteed option while individuals with a Loss Proposal overwhelmingly took the risky option. This suggests people tend to avoid losses and accept gains without considering the magnitude of gain or loss (Kahneman and Tversky, 1979).
This is interesting to the credit union industry when making the decision to roll out a new loan product, especially if the product holds higher risk than current offerings. Fear of increasing losses could halt a new offering that may be both beneficial to your members and result in a higher yield product.
Anchoring bias suggests individuals develop initial expectations about an outcome, but do not sufficiently adjust their initial expectations for changing circumstances. Joyce and Biddle (1981) studied this bias by asking two groups to estimate the prevalence of fraud in the accounting industry.
- Group A was asked if they thought fraud was present in more or less than 10 in every 1,000 companies audited by (formerly) Big Eight Accounting Firms.
- Group B was asked if they thought fraud was present in more or less than 200 in every 1,000 of these companies.
- Both were then asked to estimate the actual number of companies per 1,000 where fraud was present.
The first and second groups estimated 16.52 and 43.11 instances per 1,000 companies, respectively. The dispersion between the groups illustrates that insufficient adjustments were made to the anchor set in the first question to arrive at their estimates in the second question.
This can be applied to credit union operations when considering changes in pricing on a loan product or risk tier. Initial rates were either benchmarked against a peer or qualitatively decided upon. Subsequent performance of the product may suggest an opportunity to raise or lower rates, but by how much?
There are a number of other biases affecting human rationality. Mitigating the effects of biases involves:
- Documenting quantifiable performance objectives, and;
- Consistent, objective evaluation of performance.
This requires analytics.
Analytics is a funny word, both because if I look at “Analtyics” enough times in one day I start to think its spelled wrong and because of its ambiguity. Analytics are what you make of them. For example, I came across a website that listed different beers and their respective alcohol and calorie contents. I imported the data into a spreadsheet to figure out what beers had the most alcohol per calorie. Analytics…
There are a few questions you should ask yourself when developing analytics:
- What question(s) am I trying to answer?
- What data is available?
- How can I use the data to answer my question(s)?
Continuing with the beer example, the question was how I could enjoy a beer or two with the smallest impact on my waistline. Using the data available (beer type, calories per 12 oz., alcohol content) I calculated how many calories I was taking in for incremental changes in alcohol content.
I won’t leave you with a cliffhanger. The analysis showed that light beer was the superior in this respect. Interestingly, super-light 64 calorie beers fell behind due to reduced alcohol content.
Developing analytics for your credit union is no different. Here’s an Example:
- What question(s) am I trying to answer? What are our expectations for 12-month losses incorporating credit quality?
- What data is available?
- Loan balances by credit tier and loan type as of December 31, 2011
- Charge offs by credit tier and loan type for the year ended December 31, 2012, and;
- Loan balances by credit tier and loan type as of December 31, 2012
- How can I use the data to answer my question(s)? You can calculate the charge off percentages in each category and apply those disaggregated loss percentages to the new balances.
You can continue to build onto your analysis. You could then estimate interest margin by incorporating a calculation of interest income and backing out your estimated charge offs. Using that information, you could evaluate the bottom line margins on the loan products and risk tiers to determine if a change in rates on any of these segments is warranted.
It is also important to monitor the results of your expectations. Compare your expectations to actual results. You may find that your expectations were spot on, way off base, or that the assumptions used are appropriate for some loan segments, but not others. Because each credit union’s membership is unique, the assumptions for successful analytics will be too.
There are challenges to creating (and outsourcing) an analysis. The challenges are so robust that the OCC released Bulletin 2011-12 “Sound Practices for Model Risk Management”. The bulletin details two types of model risks:
- Risk of bad data
- Risk of bad calculations
Either can significantly affect the quality of your analysis. It is important to understand how your analysis works and what data is being included.
These behavioral biases are nothing new. The examples cited are older than most credit unions. What’s new is the data and technology to mitigate these biases is now available. Incorporating technological innovations in data storage and analytics into your operations is no different than incorporating innovations in payment processing or social media. The information to make better decisions is here, and there is no better time to start using that information.