Rethinking the Float Based Valuation Model
Board: Berkshire Hathaway
The discussion regarding Berkshire's float and how it is invested inevitably leads to the question of whether the "float based valuation" models that were very popular 10-15 years ago ultimately led investors astray and resulted in valuations for Berkshire that, in retrospect, can be considered above fair value.
The float based valuation model is one developed by Alice Schroeder in the late 1990s and published in a 1999 report which is now available on a number of sites online. I think most everyone here is familiar with the methodology. Although not explicitly endorsed by Warren Buffett, many believe that Schroeder's access to Buffett in the late 1990s combined with the fact that she was subsequently selected as his official biographer indicate Buffett's general agreement with her work and methodology.
Although I used this methodology for years as my primary guide to Berkshire's value, I have come to believe that it is of very limited utility and can lead investors to overpay. The main problem with the model is the very extreme sensitivity of the model's results based on relatively small changes to the input variables used to calculate the present value of Berkshire's float: Investment Returns, Cost of Float, Growth of float, and Discount Rate.
There is often debate regarding whether Berkshire's float serves the functional equivalent of equity. What's interesting is that using the float based valuation model, the present value of Berkshire's float usually far exceeds its nominal value based on only moderate input variables. For example, when I wrote a report on Berkshire two years ago, Berkshire had float of $65.8 billion. I computed the present value of the float to be $125 billion based on input variables of 6% for investment returns, -0.7% for the cost of float, 3% for the growth of float, and 6% for the discount rate. Essentially, I was saying that the present value of Berkshire's float was, at the time, nearly twice the actual dollar amount of float. In retrospect, the rate of return on investments was probably too aggressive as was the growth rate of float. However, you can play with the model and even more conservative estimates result in valuing float quite highly.
To Jim's point, if we assume that float will be invested in fixed income and cash (for whatever reasons), then investment returns must be set at a much lower rate based on the current interest rate environment. Buffett has also told us to not expect any significant growth of float going forward. Although the cost of float is likely to remain negative, that is merely one variable in the float valuation model and not the dominant variable if one assumes significant changes to the growth of float and investment returns.
So where am I going with this? Basically, over the past two years since writing the report (and mostly over the past year or so) I have come to the realization that the float based model may be an intellectually "elegant" model but one of limited utility in practice. The assumptions that must be made and the sensitivity of those assumptions to the valuation of float are just too volatile to be consistently useful over time.
I believe that the two-column method is more appropriate for Berkshire and far less likely to lead to errors. The model is also arguably more appropriate for Berkshire as a greater percentage of the company's value originates from non-insurance operations vs. the composition of the company when Schroeder's model was first developed. Buffett has endorsed this model in shareholder letters. The two-column model DOES assign value to float implicitly based on the fact that investments per share are used as one of the "columns" and a portion of the investments are funded by float. Essentially, this model is saying that float is the functional equivalent of equity which is reasonable as long as the cost of float is negative.
It is important to revisit valuation models and opinions from the past to see how they conform to reality over time. Berkshire's stock price has underperformed my expectations in recent years but that is not why I have in recent years come to question the float based model. That has more to do with the overall model making less sense to me today than it did in the past.
The good news is that the two-column method easily justifies a $165K stock price today. The only way I see this model being invalidated in the years to come is if Berkshire's insurance operations suddenly deteriorate to the point where the cost of float gets to be high thereby invalidating the assumption that float has the functional equivalent of equity for shareholders.