I was recently talking to someone about using a discounted free cash flow model for calculating the value of a potential – or actual – holding. They pointed out the usual argument that the process gives you a feeling of exactitude, thereby giving the investor a false sense of confidence in their investment decisions. It got me thinking about the model I’ve used since I first started investing. I decided it might prove useful for myself and my readers to take a look back at my time as the investment officer at Nintai Partners, my former health care consulting firm that had its own internal investment portfolio. It seemed t might be interesting to find out whether the model I used was useful (and effective) in achieving outperformance of the S&P500 over the portfolio’s lifetime.
Discounted free cash mdel: a joyfully short overview
The discounted free cash flow model (DFCFM) simply utilizes future free cash flow projections discounted back to today’s present value by using an estimated cost of capital. Think of it as an accounting version of "a bird in the hand is worth two in the bush." Utilizing a DFCFM has its pros and cons.
- A pro is that you focus on free cash flow and ignore any accounting gimmicks that might be used in calculating earnings.
- A pro is that the DFCFM is most likely to get you closest (that doesn’t infer absolute closest, but relatively closest) of other models.
- A pro is that short-term economic noise can be scrubbed out.
A DFCFM has significant drawbacks that any investor should be aware of if they wish to employ it in their investment process.
- A negative is that estimates get increasingly unreliable over time. Estimated free cash flow in year 10 is substantially less accurate than year 1.
- The discount rate (or WACC: weighted average cost of capital) can be significantly off and produce large variants between estimated and actual valuation.
- Major shifts in corporate strategy, competitive advantages or financing models cannot be taken into account.
There are some attributes that are both positive and negative. For instance, a good DFCFM will make you really think about the company’s competitive advantages, market sizing and growth, and product development. The downside to this is that a DFCFM is the most time-intensive of valuation models.
Measuring for success: how effective was Nintai Partners’ model?
As my friend and I discussed the DFCFM in greater detail, I decided it would be interesting to see how accurate my models have been over time. As most know, when I allocate capital in my investor portfolios, I share both a three to four-page investment case along with a summary valuation spreadsheet (rolled up from a far larger and more detailed model) of the company. I looked at all the companies I had purchased at Nintai Partners between 2002 to 2010[1] and held for at least five years. This would allow me to see how accurate my projections were over an extended period of time.
Evaluation criteria
One of the dangers in designing tests using backward-looking data is building something that will get you the results you want. I thought long and hard about what questions I would want to have answered from this research. They eventually merged into four specific questions.
- What was the overall accuracy of predictions made in the model?
- How often did stock prices correlate with model predictions?
- How often did predicted WACC (sometimes referred to as the discount rate) compare to actual WACC?
- How often did the research lead to outperformance versus the S&P 500?
During the period 2002 to 2010 (including holdings through 2015 if they met the five-year holding requirement), Nintai Partners held a total of 37 stocks. Of these, 29 were held for five or more years.
Research findings
I was surprised by the research findings in several ways – some positive, some negative. A couple of findings were to be expected and are relatively easy to explain. First, as you ran the models out further in time, the less accurate they became. As estimates get further away from historical data, it’s only natural that projections will become less accurate. Second, the accuracy models were company dependent. For a company with a wide competitive moat, projections had a tendency to be more accurate than a company with smaller moat. Company size, returns on capital and equity, and consistent cash flows also increased chances at a more accurate projection rate.
Even with these two broad-based findings, the results were immensely interesting as I delved deeper into each individual holding. Here are the findings for each of the four questions I developed.
Were valuation increases/decreases accurate over a one, three and five-year period?
As quarterly and annual reports come in, I will re-assess valuations. For 10-Qs I will only alter a one-year assessment. I will evaluate numbers out on a two and five-year period after the release of the company’s annual report along with additional research including conversations with management, competitors, customers, industry thought leaders and so forth.
Surprisingly my estimated decreases in valuation were far less accurate predictors than estimated priceincreases. Of the 67 increased valuations, prices tracked my estimates[2] 77% of the time. Not bad! Unfortunately, of the 19 decreased valuations, prices tracked to my estimates by only 38%. A lesson I took from this is that I was overly pessimistic on my estimates that bad news might have on stock prices. Also, I’ve confirmed I should never be a short-seller! In Figure 1, readers can see the distinct difference in predictive value of increasing or decreasing values and wide and narrow moats. When I increased my intrinsic value of a wide-moat company, the price correlated (within 5% positive or negative) roughly 98% of time at one year. Conversely, when I decreased my intrinsic value of a narrow-moat company, the price correlated less than 21% five years from my change.
Estimated Free Cash Flow – Accuracy Rate over 10 Years
A great challenge in using a DFCFM is getting two numbers right over a 10-year period – the estimated growth of free cash flow and the weighted average cost of capital. Let’s start with the first.
As a value investor, I use my estimated free cash flow growth rate as a way to build in a margin of safety. As an example, I might look at the company’s five and 10-year free cash flow growth rates and halve them or even quarter them dependent upon perceived strength of the company’s competitive moat, market growth and so forth. I care less that the free cash flow growth rate matches previous numbers, but rather that it is within 5% of my estimated growth rate. Accuracy is against future expectations, not historical records.
I decided to go back and utilize the same 29 stocks that were held for five years or more within the Nintai Portfolio. Of these stocks, 13 (or 45%) of the portfolio holdings achieved free cash flow growth of 20% or greater over the next-to-last five-year period[1]. Nine companies achieved greater-than-15% but less-than-20% growth over the next-to-last five-year period. The remaining seven companies achieved between 10-15% growth in free cash flow over their next-to-last five-year period.
To build in a margin of safety, I reduced each category’s growth rate by a significant percentage. The highest 13 I reduced the FCF growth rate by 60%, the middle group by 50% and the lowest group by 40%. So how did it work out? Overall, I was pleased to see my estimated margins of safety were too large. As you can see in Figure 2, the greater the historical growth, the larger the error in the margin of safety.
I should point out the variance wasn’t based on just the growth rate. For instance, companies with wider competitive moats, higher returns, stronger financials and so forth generated higher outperformance than those with lower performance in these categories.
Estimated WACC versus actual WACC over 10 years
Compared to the estimated growth rate in free cash flow, estimating the weighted average cost of capital (or discount rate) can have an even greater impact on your estimated valuation. I’ve written previously about the impact the Federal Reserve can have on valuations as it pushes different monetary levers that can raise or decrease the 10-year discount rate (the interest rate banks charge each other to borrow money). I use that 10-year rate as the basis for generating a company’s WACC. I then add on percentage points for the company’s financial strength, size, competitive advantage and so forth. When these numbers are all added up (the 10-year Treasury rate plus company specific attributes), I get the company’s WACC or discount rate.
I look for companies where management’s return on capital far exceeds its weighted cost of capital. Imagine if you were invested in the stock market and achieved a return of 8% annually. If you are using excess profits from your salaried income, then the cost of your capital is roughly what you would earn by investing in a 10-year Treasury. At a rate of 3%, your return on capital (8%) would generate a 5% return. But let’s assume you were earning the same amount on stock investments, but you used a cash advance from your credit card to obtain the money to invest. Your weighted average cost of capital might be 16%. Here you return on capital (8%) would be swamped by your cost of capital (16%) generating a negative (-8%) return. If both people came to ask you to invest in their business, who would you prefer?
During the period of 2002 to 2010, the goal of Nintai Partners was to invest in companies that had a net return on capital of at least 20%. To calculate this, I would subtract the weighted average cost of capital from the companies' return on capital. Utilizing the same 29 stocks from previous parts of the study, I calculated returns as seen in Figure 3.
As you can see the vast majority of holdings in the Nintai Partners portfolio generated return on capital greater than 30%. I estimated that those companies – on average – would produce a weighted average cost of capital of 9.8%. I was off by roughly 2%, making the net ROC much higher (and hence the companies much more profitable) than I originally estimated. The results were similar across the board. Underestimating profitability isn’t a problem I would complain about too much.
Underperform/outperform S&P 500 over five years
The ultimate measurement of success of your DFCFM is whether the portfolio holdings under or outperform the S&P500 (or some other appropriate proxy). Up until now, I have been quite pleased the model has picked companies that outperformed in key measures. In three of the four questions posed at the beginning of this article, research has proven that a well-designed DFCFM could meet its targets.
A key part of answering this question was whether I should use just the 29 stocks that were held in the portfolio for greater than five years or the total 37 stocks owned in the portfolio for the period 2002 to 2015. In Figure 4, for comparison's sake, I have chosen to do both. This way readers can get a look at the numbers from both perspectives. One thing that jumps out right away: The stocks that we held longest generated the highest annual returns.
Conclusions
Over time, I’ve come to believe that a well designed discounted free cash flow model -- combined with certain investment selection criteria -- can provide investors with an opportunity to outperform the general markets in the long term. By building in tools that create significant margins of safety, I believe a DFCFM can help identify companies that have the ability to generate highly profitable growth that can be held for an extended (five to 10 or more years) period of time. This allows the investor to keep trading costs to a minimum, reduce emotional buy-and-sell decisions, and let corporate management generate outstanding returns on capital. Nothing is guaranteed in investing (as a required disclaimer: past performance does nor gurantee furure results!), but reducing overpaying for assets, remaining patient in your activity, and using hard data to make decisions can greatly improve your odds at meeting your investment goals. A well-designed discounted free cash model is an essential tool to help get you there.
As always, I look forward to your thoughts and comments.
Disclosure: None
[1] I used the next-to-last five-year period to see how my estimates worked over the last five-year period before Nintai ceased operations.
[2] We ended the analysis of companies purchased in 2010 as Nintai Partners was closed in 2015. The closing date dictates that with a five-year holding maximum, the last purchase date be 2010.
[3] I considered it a success if the price tracked to my estimate by +/-5% over 1 year, +/-10% over three years, and +/-20% over five years.
