The Problem with Calculating Intrinsic Value
The calculation of intrinsic value has become a forbidding and abstruse practice. It seems reserved for nerds and members of the Warren Buffett cult. As Aswath Damodaran, one of its most elegant and charismatic practitioners, and perhaps the person who has promoted it more than anyone of late, wrote recently, “uncertainty underlies almost every part of intrinsic value.” Damodaran teaches a terrific semester-long course on the subject, and it’s very clear from even a cursory look at it that calculating intrinsic value is a huge time-suck. Various people have tried to automate the process—most notably Simply Wall St. To their credit they have made their calculation procedure public. One can certainly quibble with a lot of the assumptions and values that go into intrinsic value calculation—not just Simply Wall St’s, but also Damodaran’s and the CFA’s versions—and there are dozens of different approaches. But on the whole, few investors actually practice it, despite paying lip service to it. As Buffett said in 1998, intrinsic value is
“the present value of the stream of cash that’s going to be generated by any financial asset between now and doomsday. And that’s easy to say and impossible to figure.”
Charlie Munger chimed in, in 2007:
“There is no one easy method that could be simply mechanically applied by, say, a computer and make anybody who could punch the buttons rich. By definition, this is going to be a game which you play with multiple techniques and multiple models. . . . We throw almost all decisions into the too hard pile, and we just sift for a few decisions that we can make that are easy. And that’s a comparative process. And if you’re looking for an ability to correctly value all investments at all times, we can’t help you.”
Moreover, the conventional methods of estimating intrinsic value simply haven’t workedlately. Underpriced stocks haven’t outperformed basic benchmarks. Aswath Damodaran has, as far as I know, not made millions by investing in stocks whose intrinsic value is far higher than their market value.
If there’s one thing that most intrinsic value procedures have in common, it’s that they’re not evidence-based. Instead they’re based on how accounting and finance should, ideally, work. Ideally, a portion of revenue becomes free cash flow, which is then returned to the shareholder. Ideally, the discount rate one applies depends on a measure of the security’s risk. Ideally, companies have a high-growth stage and then a stable growth stage. And so on.
But practitioners have largely avoided questioning these most basic assumptions and procedures. They’ve avoided actually looking at the evidence and asking the hard questions that result. Is it appropriate to apply the same valuation procedure to Uber (UBER) and Wal-Mart (WMT), as Simply Wall St does? Does valuation have to be so complicated?
What I want to do in this article is to look at intrinsic value from scratch, using actual evidence from today’s markets, and bring some fresh and different ideas to the process.
I’ll warn you in advance, however. My results are not better than Damodaran’s or Simply Wall St’s or Warren Buffet’s. I would never claim that my expertise on this subject is greater than theirs. Damodaran and Buffett are geniuses. I’m a dabbler.
Instead, the purpose of this article is to
- a) show that intrinsic valuation has some basis in real life and is not just an idle practice for finance nerds;
- b) try out an evidence-based method which, I hope, can provide some insights into how the market works; and
- c) present a deliberately naive perspective on intrinsic valuation which might provide present and future practitioners with a few new ideas.
Present Value
The conventional way to think about the intrinsic value of a company’s stock is the present value of future cash flows. This is certainly the way we assign values to bonds. But stocks and bonds are fundamentally different types of investments. A bond has a fixed maturity and a stock does not. A bond is tied to the company’s current performance and a stock is tied to its future performance. A bond pays out its cash flows in the form of interest, but many stocks pay no dividends at all. So the present value of future cash flows may not be the right way to value a company’s stock.
How do we assign a value to a collectible, an artwork, or a bar of gold? We try to figure out how much the object will be worth in a year’s time, or two years, or ten years, and discount that amount back to the present time. But a public company is not a collectible or an artwork or a bar of gold. It is a living, growing entity. And most public companies, if they survive to maturity, eventually produce some sort of shareholder yield in the form of dividends or stock buybacks, while collectibles, artworks, and gold produce none.
Does that mean we can compare pricing a public company to pricing a purebred horse? A horse is a living, growing entity that will produce yield in the form of stud fees and racing winnings. But companies, unlike horses, have no built-in life span. They can survive indefinitely or go broke tomorrow.
No, pricing a public company is unlike pricing anything else. The dividend discount model presents a huge number of problems, but so do all other approaches.
But we ought to be able to arrive at a range of possible prices for a stock. We could, as most do, just trust the market to do it. But wouldn’t it be better to try to figure out how the market is doing it, or what the prices that the market assigns stocks are telling us? Can we simply throw up our hands and say that we have no idea what this company is really worth or whether or not it is fairly priced? If we were to do so, then there would be no justification—besides the market’s extremely opaque estimate—for some companies to be valued at hundreds of billions of dollars while others are valued at only a couple of million.
One way of dealing with this problem is pushing it off into the future. Maybe a company’s present market value depends on an estimation of how much it’ll be worth in ten years’ time. This makes some intuitive sense. The only way Tesla (TSLA) can be valued at twelve times the market cap of Ford (F) is for us to assume that in ten years’ time Tesla will be (by some measurement) twelve times as big as Ford—whether that means twelve times as profitable or generating twelve times the revenue or producing twelve times the number of cars. To some folks that may seem unrealistic, but to the true believers it’s an understatement of the magnitude of Tesla’s future success.
But we can play this game indefinitely. Tesla’s current price reflects the expectation of its success ten years from now, and its price ten years from now will reflect the expectation of its success twenty years from now.
Because future prices, however, must always be discounted to arrive at present value, the value of Tesla a hundred years from now will make very little difference to today’s price, while its value two or three years from now matters a lot. So the discounting of future value to arrive at present value, and using infinity as an endpoint, makes a good deal of sense. If you buy Tesla’s stock today, it doesn’t matter if you plan to sell it in five months or in fifty years. In either case, you still need to take into account its expected growth over an infinite amount of time, because its selling price in five months will also reflect that growth.
Some Formulas
Now let’s get into some very elementary math. And let’s treat stocks like collectibles for a moment.
If t is our time frame and r is our discount rate, the present value of a company equals its “worth” after t years divided by (1 + r)t. But we just said that after t years, its “worth” will be its “worth” after t more years divided by (1 + r)t. Moreover, we don’t know what t is.
The way to take care of this conundrum is to define the net increase in a collectible’s value as its dividend. If a collectible increases by 10% in value every year, it is essentially paying a dividend of 10%.
So whether a stock actually pays dividends or simply increases its market price, it pays its shareholders some sort of dividend for holding those shares.
So let’s say the dividend paid by a company in any given year is dt and its present value is PV. Then
What discount rate should we use? Several approaches have been suggested, including the return of an equally risky investment and the cost of borrowing the capital we’re investing. I’ll have a different suggestion shortly. So let’s table that discussion for a moment.
Instead I want to focus for a moment on the difference between d in two consecutive periods. This could be expressed by g for growth.
If g were constant for all periods (it’s not), a simple mathematical reduction would result in the following formula:
where PV is present value, d1 is next year’s dividend, r is the discount rate, and g is the steady growth rate.
This is called the Gordon Growth Model. As g gets closer to r, PV gets closer to infinity, and if g is greater than r, then one cannot place a value on the company. Why? Because if you look at the series, the value of the company gets larger as time passes faster than the discount rate, so you end up with progressively larger numbers over time.
Now what really happens is that g starts off at one number and then changes. Often, when the company is very mature, g becomes a rather low number.
That way g can start higher than the discount rate but eventually be lower than the discount rate. For example, Tesla might have a growth rate of 40% right now, but once it theoretically dominates the earth, its growth rate might be only about 4%.
The conventional way to handle this is to use a two- or three-stage growth model. The first few years, g is at one level, the next few it’s at another level, and at the end it’s at quite a low level. This gets quite complicated. But it’s quite necessary too.
A Close Look at Growth
Over the last 36 years, the median annual sales growth of a company with stock available to buy in the US is around 9.4%. How does that change as a company ages? I decided to take a look.
First, I narrowed my search to companies that have exactly twenty years of annual statements. Then I took the median growth of those companies over each of the last nineteen years. Then I did this again and again, going back year by year to 2003, which is when I exhausted my coverage. I then took the geometric mean of all those median growth numbers and came up with a relatively smooth curve. This is what you can expect as a company grows from year one to year nineteen:
As you can see, a typical company will start with a sales growth of around 19% per year, hold steady at that rate for about five years, then slowly fall to around 6% after fifteen years or so, after which growth will be relatively steady again. This lends itself well to a three-stage growth model. Years 1 through 5 will have an initial growth rate, years 14 through infinity will have a final growth rate, and years 6 through 13 will have a growth rate that steadily drops from the initial to the final growth rate.
(I admit that this study necessarily suffers from survivorship bias. It’s possible that companies that didn’t survive for twenty years might have quite different growth trajectories.)
Calculating the Discount Rate
The fact that mature companies grow at a steady rate gives us a way to calculate the discount rate without depending on guesses as to the return of an equally risky investment. We know that the present value of an investment that pays dividends in perpetuity with a constant growth rate equals its dividend divided by the difference between the discount rate and the growth rate. So let’s take all mature companies—companies with fourteen or more years of annual reports—and find out what they’re actually returning to shareholders (shareholder payout). Then we’ll take that as a percentage of these companies’ total market cap (shareholder yield).
The conventional calculation of shareholder yield is the sum of dividends paid, net equity purchased, and debt reduced, all divided by market cap. Including debt reduction is quite problematic, as debt has significant tax benefits and its cost is far lower than the cost of equity, so mature companies very often increase their debt rather than reduce it. Therefore, despite the Modigliani-Miller theorem that capital structure is irrelevant to value, I’m going to include only dividends paid and net equity purchased.
Now we can add shareholder yield to these companies’ total sales growth and thus find out what the discount rate will be. If you don’t follow my logic, here are the equations: v = d/(r – g), where v is value, d is payout, r is discount rate, and g is growth. Solving for r gives you d/v + g: shareholder yield plus growth.
Let’s look at shareholder yield first. I calculate forward shareholder yield by taking the shareholder payout as defined above and dividing it by the market cap at the beginning of the payout period. Since January 1, 1999, this has averaged 3.58% (using a cap-weighted average). However, between 1999 and 2006 it averaged only 2.80% while between 2012 and today it has averaged 4.43%.
Annual sales growth, on the other hand, has the opposite trajectory. Its overall average is 5.84% (again cap-weighted), but its average between 1999 and 2006 is 8.50% while its average between 2012 and today is only 3.37%.
If you add the two together and subtract the risk-free rate (the rate of ten-year US treasury notes), you get a series with plenty of peaks and dips, but one that hasn’t significantly declined or increased this century. This is called the equity risk premium, and it has an average of 5.99% overall.
So we can say that the discount rate should be 6% plus the US ten-year treasury rate. Consider that the treasury rate has averaged 3.44% over this period, and then consider that the total market cap of these companies has increased by 9.62% per year (compounded). It all adds up beautifully. Our discount rate so far this century has averaged 9.43%.
I also thought that one could specify discount rates for specific sectors. But when I tried this, only three sectors had implied discount rates that were more than 10% different from the average (energy, health care, and utilities, all of which were higher than average, most likely because health-care companies grow more and energy and utility companies pay large dividends). Because the implied discount rate varies a great deal from year to year, the difference between sectors is far, far lower than the difference between years. And all three of these sectors face enormous uncertainties in the years ahead as our health-care system and energy use change. So I would conclude that using an overall discount rate of 6% plus the risk-free rate is likely wiser than using sector-specific rates.
(If you’re a Portfolio123 member and want to know how I calculated the discount rate, I created a universe of all stocks that have reported annually for fourteen years or more, have a market cap of $30 million or more, and sell at a per share price of $1 or more. I then took the sum of dividends paid and equity purchased of all these companies over the past year and subtracted the sum of equity issued. I then divided this by the total market cap of all these companies at the beginning of the fiscal year I was measuring. I now had aggregate shareholder yield. To this I added the total sales of all these companies divided by the total sales last year of all these companies and subtracted one to get the aggregate sales growth. I plotted this since 1999, downloaded the chart results, and took the average. I relied primarily on FactSet data, but checked Compustat data as well.)
Should Risky Companies Get a Higher Discount Rate?
Some people think that the discount rate should reflect the risk being taken. Investing in a tech stock is riskier than investing in a utility, so the discount rate should be higher.
Let’s play with this notion a little. Let’s take two companies, a utility and a tech company, with the same shareholder payout and the same growth rate. Let’s say they both pay about $300 million to shareholders and have a growth rate of about 4%, and both are mature companies.
If we assign an 11% discount rate to the tech company and a 7% discount rate to the utility, we come up with an intrinsic value of $10 billion for the utility and $4.3 billion for the tech company. Does this really make sense?
I decided to test this. I divided my universe of mature companies into two groups: those with a five-year beta greater than one, and those with a five-year beta less than one. But the stocks with a higher beta got a lower discount rate. In general, high-beta stocks seem to be paying less and growing less than stocks with low betas, though there are time periods in which the opposite is the case. The difference between them is small, but not insignificant. I also tested a more extreme version: stocks with a beta less than 0.8 versus stocks with a beta greater than 1.2. The results were even more extreme: low-beta stocks deserved a discount rate almost 2% higher than high-beta stocks.
When you think it through, high risk does not equal high reward. Yes, risk and reward are correlated up to a certain point, but beyond that point, the higher the risk, the lower the reward. You can read more about this here.
Finding the Right Variables
Now in performing the above exercise, I used revenue growth and shareholder yield as my proxies for g and d in the formula
But what if we use EPS or EBITDA growth for g? What if we use free cash flow or net income for d? Wouldn’t the discount rate change significantly? Perhaps.
But when trying to calculate the discount rate, I think it’s safest to be conservative. Revenue growth is the most conservative measure of growth. It doesn’t jump around nearly as much as other measures. The same holds true for shareholder yield. These numbers are not nearly as subject to accounting tricks as the others. There are probably twenty different ways to measure free cash flow, and companies regularly report earnings and EBITDA numbers that differ significantly from those mandated by generally accepted accounting principles (GAAP). In addition, some people think R&D expenses should be capitalized, and others don’t, which will seriously affect earnings and EBITDA figures. Sticking to revenue and shareholder payout brushes all that aside.
Cash generated by profits, or free cash flow, can pay down debt, generate growth, be returned to shareholders, increase executive compensation, be invested outside the company, or just sit there. If a company generates growth or returns the cash to shareholders, that will show up in this conservative exercise. When a company chooses one of the other options, why should that affect its return to shareholders—and therefore its intrinsic value? If a company puts all its profits into paying down its debt, that will not benefit shareholders unless it’s in danger of bankruptcy. If the market sees that a company’s profits are squandered on executive compensation rather than fueling growth or being returned to shareholders, it will not place a premium on that company’s stock.
What I’m saying is that we really can boil down a mature company’s intrinsic value to two basic things: shareholder payout and revenue growth. Everything else—profitability, return on capital, earnings growth, free cash flow generation, asset turnover, accruals, and so on—are simply the steps between those two. All of them, of course, have to be in place before revenue growth can be successfully converted into shareholder payout, and some of them figure a great deal in how a company converts its assets into sales growth. They are indispensable figures and should never be ignored. But the basics of valuation lie in those two items.
Predicting Shareholder Payout
I compared shareholder payout in one year to the previous year’s possible indicators of shareholder payout. By far the best predictor was the previous year’s shareholder payout, which is exactly what one might have guessed. Surprisingly, however, free cash flow was an extremely poor predictor of shareholder payout. After the previous year’s shareholder payout, the best predictors I found were, in order of best fit, operating income (EBIT), EBITDA, operating cash flow, old-fashioned cash flow (after-tax income plus depreciation and amortization minus preferred dividends), net income, and gross profit. Because EBITDA and old-fashioned cash flow are very highly correlated with (in other words, very similar to) operating income but inferior in fit, we can eliminate those, leaving us with five significant data points in predicting shareholder payout. Performing multiple regression after trimming outliers can give us a formula, which I’ll provide at the end of this article.
This formula, like all the formulas in this article, will give you a very rough estimate, and should not be used widely, especially when trying to establish an intrinsic value for a particular company. Instead, it’s a nice basis for doing some wider data analysis. The key point is that in calculating future shareholder yield, past shareholder yield should not be your only data point. You also need to take into account income (before and after taxes), cash flow, and perhaps even gross profit.
This in turn can give us some insight into what will be useful for estimating the intrinsic value of immature companies: their ability to convert revenue not only into shareholder payout but also into income and cash flow.
Predicting Sales Growth
I’ve already discussed this at length, giving a number of factors to take into account. I came up with a different formula based on multiple regression and winsorizing outliers that avoids ranking. It takes into account price momentum, analyst estimates of next-twelve-month sales growth, asset growth, the ratio of net operating assets to total assets, the median sales growth over the last five years, and analyst recommendations. (The actual formula is at the end of the article.) By this formula, Apple’s (AAPL) projected sales growth is 14%, Microsoft’s (MSFT) is 13%, Amazon’s (AMZN) is 25%, Google’s (GOOGL) is 15%, Facebook’s (FB) is 19%, and WalMart’s (WMT) is 3%.
Once again, this is a very, very rough calculation, and should be used for overall data purposes rather than figuring out the sales growth of a particular company. The key takeaway is that in calculating future sales growth, one needs to look at a wide variety of factors, not only past sales growth.
Classification of Companies
Before we get into multi-stage analysis, let’s contrast young and mature companies. We shouldn’t look only at how many annual statements a company has filed to determine this; instead we should also use a company’s characteristics. There are plenty of companies that have reinvented themselves and gone from old age to infancy in terms of their growth rates.
So I’d like to propose five things that separate young from mature companies.
- A young company will likely not have a shareholder payout; a mature company will.
- A young company will have high sales growth; a mature company will have low sales growth.
- A young company may not yet have figured out how to transform revenue into earnings and cash flow; a mature company will have solid and stable earnings and cash flow.
- A young company may have accelerating growth; a mature company’s growth will be either decelerating or steady.
- And, of course, a young company will not have many annual statements in its history while a mature company will have a lot.
I came up with a rather complicated formula and a ranking system to classify companies as mature (top 50%), infant (bottom 25%), or in-between. But it turned out to be an idle exercise, not worth presenting in detail. Suffice it to say that for infant companies like Uber (UBER) and Peloton (PTON), I’ve concluded that the kind of present-value valuation I’m advocating in this article is nearly impossible, for reasons I’ll give shortly. For adolescent companies, a two-stage valuation will work, but is somewhat complicated. For mature companies, however, a very rough present-value calculation can be relatively simple.
Valuation of Mature Companies
For mature companies—companies with steady but low sales growth and consistent shareholder payout—intrinsic value is basically the expected shareholder payout divided by the difference between the discount rate and expected revenue growth. It’s therefore impossible to calculate for companies whose expected growth exceeds or comes close to the discount rate.
If we use 9.43% as our discount rate, a mature company with 0% expected revenue growth will be worth about 11 times its expected shareholder payout while a mature company with 8% expected revenue growth will be worth about 70 times its expected shareholder payout. This gets across pretty forcefully why you can’t value a company without taking its growth into account! It also explains why conventional measures like P/E and price-to-sales and price-to-book tell us very little when used on their own, without any consideration paid to growth potential.
Intrinsic Value Is a Very Rough Approximation
In Security Analysis, Benjamin Graham and David Dodd wrote,
“[The] concept of intrinsic value, as . . . definite and ascertainable, cannot be safely accepted as a general premise of security analysis. . . . The essential point is that security analysis does not seek to determine exactly what is the intrinsic value of a given security. It needs only to establish either that the value is adequate—e.g., to protect a bond or to justify a stock purchase—or else that the value is considerably higher or considerably lower than the market price.”
Accordingly, I think it’s best to consider any company whose intrinsic value is between one-half and twice its market cap to be fairly priced. It’s those companies whose intrinsic value is greater than twice their market cap that you want to seriously consider investing in.
Valuation of Immature Companies
Immature companies often have little or negative shareholder payout, so there’s no dividend payment to put into the present-value equation. But we can’t simply substitute a percentage of revenue. Some companies have high revenue growth but poor prospects of actually converting that revenue into shareholder payout.
The essential thing to examine is a company’s margin and whether it is likely to increase. Margin is the percentage of revenue that remains after various costs have been deducted. Just as there are five components to predicting shareholder payout, I recommend looking at five margins: gross margin (gross profit to sales), operating margin (operating income to sales), net margin (net income to sales), cash flow margin (operating cash flow to sales), and shareholder margin (shareholder payout to sales). By looking at a company’s current margins and whether those are likely to increase or decrease over the years, and by projecting revenue growth, we can, through some relatively complicated mathematics, come up with the “dividends” needed for a two-stage analysis of an “adolescent” company.
As for those companies in “infancy”—companies with very few years under their belts, high sales growth rates, and negative projected shareholder yields—I would venture that trying to estimate their intrinsic value is a fool’s game.
An Example: Calculating the Intrinsic Value of NVR
NVR (NVR) has a projected revenue growth of 16.5%, making it impossible to value using the mature company approach. Its projected shareholder payout is $453 million.
Here’s my rough calculation of its intrinsic value.
We require as inputs the projected shareholder payout, which I base on the company’s present shareholder payout, its net income, its EBIT, its cash flow, and its gross profit; its recent annual sales; its projected sales growth; its median payout margin growth over the last five years; and the standard discount rate.
The calculation proceeds as follows.
- The sales growth diminishes linearly from year to year down to a final value of 8.3%, which is a somewhat arbitrary number I use for high-growth companies.
- The sales increase by the sales growth.
- The payout margin starts at 5.43%, since that’s the projected shareholder payout divided by the year-one sales. It decreases from there to a final value of 2.18%, which is based on a formula that takes into account projected shareholder payout as well as the growth of shareholder payout over the last five years.
- The shareholder payout is the payout margin times the sales, so for year 1 it’s exactly the payout in the input.
- And then each year’s payout is discounted by 9.43%.
- For the terminal value, I took the eleventh year’s payout and divided by (9.43% – 8.3%) before applying the discount rate.
When I added up all the present values of the future dividends, I obtained an intrinsic value of $22 billion. That’s a little more than NVR’s current market cap of $16 billion. So I would conclude that NVR is fairly priced or slightly underpriced.
Now obviously it would be crazy to use this calculation for every company. Some companies have terrible margin growth but show signs of turning that around. For others you may want to put in different numbers for sales growth. If you were to do this kind of valuation for a company like Uber, you’d end up with a huge negative number, since its current margin is not only negative, but is getting more negative every year.
Approximating Intrinsic Value
In all, this formula includes the following measures. In calculating projected sales growth, we looked at 52-week relative price change, average recommendation, asset growth, average sales growth, the ratio of net operating assets to total assets, and analysts’ estimates of the next twelve months’ sales growth. In calculating our base for sales, we looked at the last three years’ sales. In calculating projected shareholder payout, we looked at equity purchased, equity issued, dividends paid, net income, operating income, operating cash flow, and gross profit. And in calculating margin growth, we looked at how shareholder payout has changed from year to year over the last five years. So our approximation of intrinsic value is a pretty holistic measure. It doesn’t take into account a lot of things—there’s no asset turnover, no consideration of debt, no return on equity or assets or capital, no free cash flow or capital expenditures, no enterprise value. All these factors are of great value, and if you’re serious about calculating intrinsic value, you’ll take them into account somehow. But for a very rough approximation that can be relatively easily automated, this is what I’ve come up with. I’m not claiming that it’s better than what others (e.g. Simply Wall St. or Aswath Damodaran) have come up with. But it suits my personal style. To be perfectly honest, I don’t think this measure is necessarily any better than a good, solid combination of relative value ratios. But maybe you can play around with it some and come up with something better.
I’ve made all my formulas public for Portfolio123 users who want to play with them. The later ones rely on the earlier ones, so you’ll have to copy all of them to get the rough intrinsic value formula to work. Some of them use a custom formula called $r, which is the discount rate you choose to use, expressed as a decimal (i.e. 0.094 rather than 9.4). Here are the links:
Projected sales growth; Projected shareholder payout; Terminal sales growth; Terminal value of a company's sales; Terminal payout margin; Rough intrinsic value.
For non-Portfolio123 users, I’ve also created a Google sheet with all my formulas. You can see it here. I think the language in the P123 formulas is relatively easy to decode, but for more difficult items there's a glossary there too.
Intrinsic Values of Large American Companies
Right now, according to my rough calculation of intrinsic value, the ten most valuable companies in the US are, in order, Apple (AAPL), Amazon (AMZN), Alphabet (GOOGL), Microsoft (MSFT), Facebook (FB), AbbVie (ABBV), Unitedhealth (UNH), Costco (COST), T-Mobile (TMUS), and Centene (CNC). In terms of value, the company that really sticks out from this list is Centene, whose market cap is far lower than those of the others.
Of the largest companies in the US by market cap, the ones that don’t make the list are also interesting. Berkshire Hathaway (BRK.B), which is the fifth largest, comes out 12th; Tesla (TSLA) comes out 17th, ahead of other US car companies, but behind Volkswagen (VWAGY) and Toyota (TM); Visa (V) comes out 58th because it’s really overvalued; Walmart (WMT) comes out 11th; and Johnson and Johnson (JNJ) comes out 20th. At the very bottom are the companies with the largest negative value: Uber (UBER), Peloton (PTON), and Zoom Video (ZM). All of these are clearly “infant” companies, which, as I’ve stated before, are not fit subjects for intrinsic valuation due to their enormous negative projected shareholder payouts and their history of decreasing margins.
The most underpriced companies in the S&P 500 are, according to my calculations, Centene, Amerisource Bergen (ABC), BorgWarner (BWA), ViacomCBS (VIAC), Best Buy (BBY), NRG Energy (NRG), McKesson (MCK), Humana (HUM), Cardinal Health (CAH), and L Brands (LB). But all of these companies require a far closer look than I’ve given them before determining whether they’re truly undervalued.
Performance of My Intrinsic Value Measure
Here’s a chart illustrating the performance of my intrinsic value method on the S&P 500 compared to some common relative value methods. These results were obtained by looking at the performance of all stocks with an intrinsic value greater than twice the company’s market cap (an average of about 25 to 30 stocks out of the 500) if one bought those stocks every week and held them for a year. I then compared that to buying about the same number of stocks per week by using relative valuation methods. Maybe intrinsic value is a little better than the relative value methods, but only barely. It’s hard to draw conclusions from numbers like these.
Here’s another chart, showing how this intrinsic value method compares to price to sales and EV to EBITDA over the last twenty years:
You’ll see that at the beginning of this century, EV/EBITDA was a great way to generate excess returns, and by now it’s really not; intrinsic value’s outperformance, though small, has remained pretty good throughout this entire time.
Please note that I did not backtest my system before coming up with it to see whether it would actually make a profit. My sole aim in designing the system was to see whether I could approximate a company’s worth without looking at its market cap. The fact that it works on S&P 500 stocks is a nice bonus, but I wouldn’t bet a dime that this particular automated system will continue to work.
As I said at the outset, doing intrinsic value by an automated method, which is what I’ve done here, is almost certainly not going to make you wealthy. But the fact that an automated system isn’t a complete failure gives me hope that applying the principles of Benjamin Graham and Warren Buffett properly—not by using relative value but by calculating intrinsic value, taking practically everything you can think of about a company into account—might end up working even in the twenty-first century.
Conclusions
I came away with my study of intrinsic value with the following conclusions.
- Intrinsic value calculation is, as everyone who has practiced it seriously has pointed out, far from exact. It is a kind of guessing game; it’s irrelevant for extremely high-growth and low-payout companies; and at best it can only give us a very rough idea of what a company is worth. However, it is not just an impractical fantastical game for nerds. The basis of intrinsic value calculation is logical, sound, and inescapable. A company has to be worth something besides what the market estimates it’s worth. And intrinsic value is the best guess we can possibly make.
- The four pillars of intrinsic value analysis are shareholder payout, revenue growth, payout margin (and its growth), and the discount rate. But practically everything else about a company should be taken into account when estimating these numbers. The discount rate, however, should probably be applied across the board equally to all companies, and should probably be in the range of 7% to 11%. Most importantly, risk measures should not affect the discount rate. High-beta companies do not deserve a higher discount rate than low-beta companies.
- Companies do go through stages: a stage of extremely high but steady growth, a stage of declining growth, and a stage of low but steady growth. Companies in the first stage are almost impossible to assign an intrinsic value to, but using a two-stage valuation process for the others can give you an approximation of what they might be worth.
- Estimating growth is of paramount importance to calculating value. A few percentages difference in growth has a larger effect on an intrinsic value calculation than almost anything else about a company. The growth/value dichotomy that so many people talk about is a false one when it comes to intrinsic value. That is because intrinsic value is concerned with the far distant future. So a company that is not growing is worth very little.
- Trying to come up with an automated way to calculate intrinsic value is probably not a great idea. There are simply too many factors to take into account. If you want to judge a company holistically in an automated (quantitative) fashion, I suggest using multifactor ranking systems instead. But if you don’t mind rolling up your sleeves and tweaking the numbers an intrinsic valuation system gives you, you might find some success.
My CAGR since 1/1/2016: 40%.
My top ten holdings right now: FKWL, STRT, AVNW, RNDB, DVD, JAGGF, PRGX, PLPC, CTG, LMB.
