Investors like modelling because complicated spreadsheets, detailed calculations, and careful consideration of many variables produces forecasts that appear well-thought out and confident. In reality, the future is almost completely opaque. The more complex a model becomes, the more assumptions and biases it also takes on.
An example of a bias that can occur is anchoring. In this case, not wanting to make outlandish assumptions, the analyst builds assumptions into the model that are only within a few standard deviations of the current or historical data. The reality though is that long term results are often surprisingly strong the direction of positivity or negative, not within a few standard deviations of what appears possible.
Another example pertains to the proper discount rate to use in a discounted cash flow model. Standard academic methodology teaches the CAPM formula, where cost of debt and cost of equity are combined into the weighted average cost of capital (WACC). While cost of debt is relatively straightforward to produce, cost of equity calculations depend on the belief that volatility is used to represent risk in a stock. If you reject this, however, such as would be the case if you belief “risk” refers to the probability of permanent loss of capital, the entire methodological exercise is called into question. This is thus yet again another scenario where complex models project false specificity.
The result of such “scientific” modelling is an overconfidence in “the model” and its predicts and an under-appreciation for the unknown. At the very least, financial analyst being a social task, you are making the thesis more difficult to understand and communicate, again introducing opportunities for error.
Connections
Buffett’s 4 Purchase Decision Criteria
Link Explanation: While not direction mentioned in the linked note, Buffett famously never used excel or relied on complex models to calculate the precise value of a stock. The trickiness, though, is that one of the core purchase decision criteria is that the stock must be attractively priced relative to owner’s value. That is hard to figure out without a discounted cash flow model. I think the better method is to run a discounted cash flow, but keep it very simple and try to understand the input in basic terms. This might be something like “how many units does X business need to sell each year to be worth it? Can they realistically do this based on past growth and the TAM? Then it’s simply a matter of being strict on a deep margin of safety requirement before any purchase decision is made.