I have been doing research for a potential youtube video on DCFs. And while doing a deep dive into DCF math and logic I came back to a common belief that intuitively felt wrong to me and now I am much more confident is wrong.
Limiting terminal growth rate to 2-3% is incorrect and illogical.
The common logic is as follows, GDP growth since the inception of the metric has been 2-3% CAGR. Thus if a company is assumed to grow faster than 3% into infinity it will become larger than the entire economy.
There are some logical problems with this and some practical problems with this.
..................................................................
A. Logic Errors
The first and most glaring error is this, the growth of cash flow and earnings of the US market has been closer to 5.5% over the last 100 years. This is because GDP is an inflation adjusted number, the GDP grows 2-3% ON TOP OF INFLATION. So a company can grow at the rate of the GDP growth plus the inflation rate and not become any larger relative to its peers, and can do so into infinity without becoming a larger portion of the economy.
The second logic error is conflating cashflow, earnings, or revenue growth with GDP growth. GDP instead tracks gross value added. A company that doubles all or any of these numbers doesn't necessarily double its contribution to the GDP. If a company were to lay off or offshore a large portion of its staff it would increase earnings and cashflow, but actually lower its GDP contribution. If a company with an inelastic good like a lifesaving drug instead makes half as many and quadruples the price, its revenue will double but its GDP contribution would be cut in half.
A third logic error is to assume all of the company's growth comes from within the US, entering foreign markets and offshoring will increase a company's earnings and revenue and cashflow without contributing to the GDP of the US.
A fourth logic error is to assume part trends can always be extrapolated into the future. GDP growth has been decelerating due to the law of large numbers but that is not set in stone and it is fully possible that AI and robotics can in the future re-accelerate things as it can provide easy access to labor. Also AI and robotics may be drastically deflationary, and as GDP is inflation adjusted it would also be deflation adjusted which could cause GDP to grow faster than earnings if the dollar gains value.
A fifth logic error is the fact that these issues arise from the fact we are running any growth number into infinity. Infinity does not play nice with reality. The universe will not exist infinitely, the world will not exist infinitely, the USA and its markets will not exist infinitely, and most certainly the companies we invest in will not exist infinitely.
..................................................................
B. Exit Multiple = Terminal Growth rate
So personally it would seem to me that picking an exit multiple simplifies things and avoids dealing with infinity in my formula and the insanity it adds to the equation. But people will say that picking an exit multiple is the exact same as picking a terminal growth rate.
Yes according to the math formula these can be substituted, a terminal growth rate and discount rate can be converted into an exit multiple. So lets look at some of these conversions.
With a 10% discount rate (10% desired rate of return) a 0% terminal growth rate = 10x FCF exit multiple, 2% TGR = 13x FCF exit, 3% TGR = 15x FCF exit, 4% TGR (mathematically impossible) = 17x FCF. 5% TGR (super impossible) = 21x FCF, 6% (impossible including inflation) FCF = 27x FCF. If you want a 10% return and expect to sell your stock at 17x cashflow in 10, 20, or 30 years you are apparently expecting your stock to be the entire world's economy some day.
With an 8% discount rate: 0% TGR = 12.5 FCF, 2% TGR = 17x FCF, 3% TGR = 21x FCF, 4% TGR = 26x FCF, 5% TGR = 35x FCF, 6% TGR = 53x FCF.
So if you desire 8% annual returns and expect to sell your shares for 26x FCF in 5, 10, 20 years you are once again projecting the company will become the entire world's economy some day.
..................................................................
C. Practical problems: Backtesting Failures
So lets go and backtest, we can do a DCF on an actual company's cash flow for the last 20 years using their real year-by-year cash flow numbers, and set our discount rate to the desired rate of return. This will tell us what price we would need to have bought the company at 20 years ago to get our desired rate of return. But we will use a terminal rate of 3% as anything higher than that means the company will become the entire world's economy some day...
Lets look at Walmart. When plugging in its real cashflow numbers from 2005-2025 and using a 10% discount rate and a 3% terminal growth rate shows we would need to buy Walmart stock at $24.42 per share in 2005 to get 10% annualized returns, and the stock was trading between $43 and $53 that year, so with perfect knowledge of the next 20 years of cash flow we can see that walmart stock was twice as expensive as it should be to get 10% annualized returns...
Except if you bought at the low price that year you would get 10.8% annualized returns, at the high price 9.6% annualized returns, at the mid price 10.2% annualized returns. If you bought at the price the perfect knowledge DCF says you would actually get 15% annualized returns...
So why is this? The terminal growth rate is wrong, Walmart today trades at a Price to Free Cash Flow of 71x right now, you would need to use a terminal growth rate of 8.5% to accurately model that exit multiple and actually perform our "perfect knowledge" backtested DCF on Walmart. Yet apparently such a thing is 3x higher than the highest possible number we are allowed to assume without Walmart taking over the entire world's GDP someday. I also checked their 3 year normalized FCF in case this was an irregular year, still a P/FCF of 62x which as an exit multiple is equivalent to a terminal growth rate of 8.3%.
This is true for any mature company trading much above 20x FCF today, any past DCF done on the company even if it guessed the future cashflows perfectly would be unable to accurately predict its value because they limit themselves to a terminal growth rate of 3% because of logic errors, WMT came to mind as a clear example but any company with a P/FCF of 25+ would run into this same issue, perfect cash flow knowledge would still lead to massive underestimation of value/returns because an exit multiple above 20 is more or less impossible in traditional DCF math.
In summary I think limiting your DCF to only a 3% terminal growth rate or exit multiples in the teens is begging for underestimating businesses. Even if you are perfectly correct about every other assumption it can lead to massive underestimation of a company's value.
Some may say that this adds margin of safety, but that is not what it is there for. The point of a DCF is to come to a comfortable conclusion on the total future returns. If you pick a lower exit multiple than you think is realistic, a lower growth rate, a higher discount rate, lower margins, then look for margin of safety on top of that you are compounding your underestimation of the company. I would say you should try to run an accurate and reasonable DCF and adjust the discount rate for the risk of the investment and look for a margin of safety on the final output, not apply a margin of safety to every single ingredient within the DCF. Also running multiple DCFs with multiple scenarios. But a 3% discount rate or 15x-20x exit multiple is actually a pretty conservative assumption, especially for shorter DCFs, and should not be treated as the upper limit.