Creating and Using Financial Models
By Jeff Greenspan, Managing Director, Financial Modeling Services
George Box famously said that “all models are wrong, but some are useful.” And, though you may not realize it, you use models all the time! The real world is incredibly complex, so we create models in our mind about how the real world operates. For example I use my “model” of my wife whenever I buy her a gift, and I know my model needs adjusting when my gift is not well received. Because we make decisions and take actions based on the models that we create, it is critically important that we understand both the strengths and limitations of models. So, let’s GET STARTED.
The key strength of models and the reason that we build them is because they help us understand the past and predict the future. A good model allows you to take some set of inputs (typically live, historic, or estimated data) and make a prediction about the future. For example, weather models look at multiple data points from thousands of locations to predict the weather, but the real-world weather is so complex that we cannot predict more than a few days into the future1.
When we forget about the limitations of financial models, the consequences can be personally or societally disastrous. The Great Recession of 2008 was triggered by the failure of mortgage backed securities2, which were created by a financial model designed to spread the risk of subprime mortgages. The models weren’t necessarily wrong... we failed to understand their limitations.
There are two sets of limitation that need to be considered: the first set has to do with the model itself, and the key concerns are:
- Are our inputs any good?
- We need to use the right inputs. It you want to predict the weight of a bucket with sand in it, we don’t need the density of water as an input. Our inputs must be predictive!
- Our inputs need to be accurate. It is tempting to jigger a model so that it produces the answer that you want. This is why it is critical to check other people’s models before investing in them.
- Are our calculations correct? We need to use the formula, statistic, or statistical distribution that reflects how reality works.
- Are our statistics and statistical distribution valid? When is the mean more valid than the median? When should we use the normal distribution as opposed to one with a longer tail? While these types of questions don’t require a PhD, some knowledge of probability and statistics is very helpful.
The second limitation and more complex concern is how we as humans interpret models. We are basically a bucketful of biases, so much so that McKinsey has published a series of articles on debiasing3. We can turn to Wikipedia’s List of Cognitive Biases for a comprehensive list, but I like Visual Capitalist's Article on the subject because it focuses on 18 biases that tend to have a “disproportionately large effect on the ways we do business.” Here is their infographic, which I humbly suggest be tacked up on your bulletin board for reference:
Now that we’ve digested the bad, let’s consider some good uses for models in the financial space. They typically support one of two objectives: decision making and budgeting/forecasting.
ProFormas: These forward-looking projections for a business can posit answers to critical questions like the following, but note how important it is that your inputs be reasonable and defensible:
- Should I undertake this business? What are the Net Present Value (NPV) and Internal Rate of Return (IRR) for this business?
- How much capital do I need to get started?
- What do my revenues look like, and what expenses are required to support these revenues?
- What will my cash flows be?
- What is a reasonable capital structure?
M&A Models: Among the most complex models that I’ve worked with, Merger & Acquisition models are used by investment bankers during initial public offerings and leveraged buyouts.
DCF Models: Discounted Cash Flow models help investors understand the value of a series of cash flows to them. See my discussion on your discount rate here.
Net Worth Modeling: What is your net worth today, and how much might you be worth in the future? Financial planners are fond of running Monte Carlo models for this type of analysis, but are often incapable of including non-stock-market assets in those models, making them less informative. [Note: Monte Carlo models provide a future probability distribution for some model output, which can be amazingly informative when the model is well designed.].
Hold or Sell Models: You’ve owned an asset (like real estate) for many years. Are you better off holding it or selling it? If you hold, should you refinance?
Loan Modeling: See my article on Picking the Best Home Loan or Refi.
Budgeting/Forecasting: What do my personal or business cash flows look like?
Financial Modeling Service can help you create a financial model for almost any purpose or serve as a second set of eyes for a model that you have created. A good review will cover the key concerns highlighted earlier in this article.
1 Check out this cool video on the science behind the butterfly effect: https://www.youtube.com/watch?v=fDek6cYijxI.
3https://www.mckinsey.com/business-functions/risk/our-insights/the-business-logic-in-debiasing, https://www.mckinsey.com/industries/financial-services/our-insights/banking-matters/debiasing-in-action, https://www.mckinsey.com/business-functions/organization/our-insights/behavioral-science-in-business-nudging-debiasing-and-managing-the-irrational-mind, and others.