Geoff Gannon June 21, 2022

Warren Buffett’s “Market Value Test” – And How to Use It

Someone who listens to the podcast wrote in with this question:

…(in a recent episode) you mention that you want to know if the capital allocation has created value or not. I was wondering how you do this kind of exercise practically? Do you look at the increase in book value/equity over time and compare that to the average ROE? When book value increased far less over a certain period of time compared with the historical average ROE I suppose that is a sign of bad capital allocation, right? Or do you have a different approach?”


There’s no one right approach that is going to work in every situation. The simpler the company and its business model, the easier it will be to see if capital allocation is working. For example, the stock price may tend to follow the earnings per share and the earnings per share may be driven in part by the capital allocation. That would be the case at a company that acquires other businesses for their reported earnings, issues stock, and/or buys back stock. Earnings per share captures all of that.

But, what if you were trying to analyze Berkshire Hathaway (BRK.B) or Biglari Holdings (BH)? In these cases, management might be allocating capital at times to increase earnings per share and at other times in ways where the value received for the capital outlay is not going to appear in the income statement. If capital is allocated to buying stock, land, etc. EPS may not be helpful in evaluating capital allocation. Now, book value would be a good way to analyze those capital allocation decisions. However, at companies like Berkshire Hathaway and Biglari Holdings you have a mix of operating businesses and investments. The operating businesses are held at unrealistic values for accounting purposes – so, an EPS approach works for judging them, but a book value approach doesn’t. And the investments may be held at realistic values for accounting purposes (they’re marked to market) – however, the underlying (“look-through”) earnings won’t show up when judging the EPS growth of the business. As a result, a pure EPS based approach to judging capital allocation will work for part of these conglomerates and fail for the other part. And a pure book value approach will work for judging capital allocation for part of these conglomerates and fail for another part. You need a mixed approach.

Buffett basically suggests this when he used the “bucket” approach for analyzing Berkshire Hathaway. He did this in some past annual letters. You take operating earnings per share (which excludes investment earnings and insurance underwriting). And you take investments per share. Operating earnings per share is a “flow” number. It needs to be capitalized to translate it into a figure that can be combined with investments per share. Investments per share is a “stock” number. You can either look at it as a “stock” number (which makes sense when trying to come up with an intrinsic value) or you can convert it into a “flow” number (by using look through earnings). For our purposes, it’s easier to assume you capitalize operating earnings per share and keep investments per share in the same form.

U.S. corporate tax rate is around 25% (I’m assuming some income is taxed at state level). Long-term average P/E ratio (Shiller, etc.) is something like 16 times. So, 16 times 0.75 (100%-25% = 0.75) equals 12. That’s a good enough number to use when capitalizing pre-tax operating income to turn it from a “flow” number to a “stock” number. So, $1 million of pre-tax operating income is the same as a stock position with a marked to market value of $12 million. In other words, selling $12 million worth of assets (at market value) and buying a business producing $1 million of pre-tax operating income leaves you in about the same place.

One final point specific to Berkshire before I get back to the more general point you’re asking about. The “bucket” approach ignores underwriting profit and loss from insurance. But, it also ignores the liabilities associated with the insurance businesses. I think this is a simplification that works better for judging capital allocation, valuation, etc. at Berkshire better than most analyst models. The value in the insurance operations will be captured by what they allow Berkshire to carry (stocks, cash, operating businesses, etc.) using their float. The negative value (the liabilities) are just an accounting liability which has more meaning in run-off and liquidation than it does for a growing business like Berkshire. If underwriting can approach a combined ratio of 100 over time and float can stay stable or grow over time – then, the value created at Berkshire should all show up in the operating earnings per share and the investments per share with no need to assign a positive or negative value to insurance. In this approach, insurance is used as a source of funding for the productive assets (investments and operating businesses) that Berkshire owns.

Getting back to judging capital allocation, Buffett’s other contribution to the subject is the idea of the “market value test”. The market value test is the idea that one dollar of retained earnings needs to add at least one dollar to the share price of the stock to be justified as a good decision. Basically, stocks go up because each dollar kept inside the business adds more than one dollar to the share price. And, stocks go down because each dollar kept inside the business does not add more than one dollar to the share price. This is also a useful and important concept in valuing a business. If a business is expected to pass the market value test in the future – then, it can be bought whenever the P/E is low enough. If a business fails the market value test – then, it might not be a good purchase even if it has a low P/E ratio. Generally, any stock with a P/E below 13 that passes the market value test should be worth considering adding to your portfolio. This is because 1/13 = 7.7%. That’s the stock’s earnings yield. Assuming the P/E we just cited is the trailing P/E and the stock is able to grow in line with inflation (or better), the actual earnings yield is more like 8-10%. This is in line with the long-term return in stock market indexes in the U.S. generally. So, any stock that currently has a P/E of 13 or less and is going to pass the market value test going forward shouldn’t drag down your returns relative to what the index tends to do. Whether it helps you beat the market or not depends on how much it passes the market value test by and how much below 13 the P/E ratio is. But, this is sort of a dividing line. Look for a stock that passes Buffett’s market value test and that sells at a P/E of 13 or less.

So, the market value test is a sound idea theoretically. And it’s a useful idea for a stock picker to apply when hunting for stocks. They don’t have to be all that cheap as long as you have a high degree of confidence the business will pass the market value test going forward. The problem is applying the market value test. One, you can only apply it to the past record. And it’s the future capital allocation record – not the past capital allocation record – that’s going to determine your returns in the stock. Two, price multiples are a huge part of stock returns. So, a business that retains $1 when trading at a P/E of 5 and then ends up at a P/E of 15 just 5 years later is going to look like it created value even if it destroyed it. If the $1 of retained earnings added just 40 cents of “intrinsic value” – that’s still going to show up as having created value because: 15 divided by 5 equals 3. And 3 times 40 cents is $1.20. It’ll look like the stock created $1.20 in value for just $1 of retained earnings. However, this increase in the stock price was due to a multiple expansion that was big enough to overwhelm poor capital allocation. Multiple expansions and contractions of this size are very common. So, attempts to apply the “market value test” over short periods of time aren’t able to differentiate good capital allocation from bad capital allocation.

What about over a 15-year time period? It’s still a problem. A multiple expansion of 3-5 times over 15 years can be as much as like an 8-12% annual return contribution. So, capital allocation that added no value at all (but also didn’t destroy any) can look like a decent business (an 8-12% a year performer in terms of share price) over even as long as 15 years. Therefore, you have to apply some common sense. If a stock started at a P/E of 5 and went to 15 or 25 – you need to adjust for this in your mind. The same would be true if the P/B went from 1 to 3 or 1 to 5. This would be obvious when using something like I recommend using a site like that. Try to use a 15-year record when possible ( has some 10 year info available for free and full 20 year info available if you subscribe).

I think a 15-year numerical record when combined with your common sense overlay is enough. Some would say 10 years is enough. It’s possible. If the industry is clearly non-cyclical – 10 years would be enough. However, long-cycle businesses can run in the 15-20 year range. So, a 10-year record is short enough that you could be capturing purely a cyclical effect where the business was early cycle on the start date you’re using and is now late cycle. I think 15 years is a decent compromise for companies that have been public for a long time. It’s also a good idea to always use moving comparisons. So, don’t just use 2007-2022. Also, take a look at 2004-2019 and 2001-2016. A few observations (3 at a minimum) taken some years apart is best. The market value test is very sensitive to strange readings caused by unintentionally unrepresentative start and end dates.

None of this is necessary if you understand the business model, the industry, etc. well and the company is a lot simpler than a Berkshire Hathaway or a Biglari Holdings. On the podcast I mentioned CBIZ (CBIZ). Operating margins in that industry are stable enough that simply asking whether $1 of retained earnings has created 1 divided by “x” dollars of additional sales where “x” is the P/S multiple you think is appropriate would work as a good enough market value test. It’s not an exact science. For ad agencies, I’ve always suggested that 1.5 is a fine P/S multiple to use. So, an ad agency that pays $100 million to buy something and gets $67 million ($100 million /1.5 = $67 million) isn’t far from neither creating nor destroying value. Of course, this isn’t as simple as saying that any acquisition done at 1.4 times P/S is a value add and anything done at 1.6 times destroys value. The issue is the long-term trend in sales for the total company. If you acquire at 1.4 times and growth in the acquired business is poor over time – the acquisition might’ve been a mistake. If you acquire at 1.6 times and it grows nicely – the acquisition was a success.

So, why use P/S instead of just the share price? For companies like CBIZ (or the ad agencies example I gave), the advantage would be cutting down on noise. During a business cycle, operating margins might vary a bit. And then during a stock market cycle, the multiples put on earnings per share might vary a bit. By tracking just sales per share you can filter out both these sources of noise and get closer to what intrinsic value creation might be. However, this requires your belief that there is a strong long-term relationship between sales per share and intrinsic value. Basically, that requires a stable free cash flow margin over time.

For banks and insurers, you can use P/B where you believe the return on equity can be stable over time. It won’t be exactly stable. But, if you believe you can assume some sort of “normal” return on equity over a full cycle – this will work. Again, this might work better than using the actual share price. You can look at changes in book value over time just as you can look at change in sales per share over time.

Another way to use these metrics is to look at the trajectory of rates of change to see if the capital allocation machine has been speeding up, staying the same, or slowing down. A sign that the capital allocation “flywheel” could be stalling out (or the runway shorter than you anticipated) is an increase in retained earnings intensity. Basically, is the company having to retain more and more earnings to drive the same increases in equity per share, sales per share, earnings per share, etc.? An even worse issue would if debt per share is having to accelerate more and more vs. your “intrinsic value proxies” like sales per share or book value per share. So, if debt per share had always been growing at 15% a year and sales per share was also growing at 15% a year – that’s business as usual. But, if you notice that debt is continuing to grow at 15% a year this decade (same as last decade) but sales per share is only growing at 10% a year, that could be a problem. Actually, a more common problem for a management intent on continuing the compounding is the reverse. The sales (and EPS) growth stays the same, but the use of additional capital rises. So, maybe sales per share has always grown at 10% a year while debt per share has too – but, now you notice debt per share growing at 15% a year while sales per share is still growing at just 10% a year. More debt is being used to drive the same growth as before. This might mean capital allocation is getting worse.

When that happens, it doesn’t automatically mean management is to blame. As Warren Buffett has said – his returns (which are the result of his capital allocation) have gotten worse every decade since he started in the 1950s. That’s because opportunities have become scarcer as the capital he’s needed to allocate has grown bigger. The same thing will happen with a lot of businesses. As more capital has to be allocated, returns are poorer on the marginal unit of allocated capital relative to the past average unit of allocated capital. This will show up in long-term averages – but, only gradually. A business that was once a compounding machine can boast of very high CAGRs since inception for a long time after the capital allocation engine starts sputtering and even stalls out. The way CAGR math works, you’re going to still have a very high CAGR over the lifetime of a company if you had amazing returns upfront and then mediocre returns in the most recent years. However, as an investor – you only benefit from future returns.

This is the main point I’d make about analyzing the capital allocation record. I don’t think it’s as important as other investors do. The past track record – even if very long – is only of use to you to the extent you can project it into the future. Understanding the philosophy of management toward capital allocation, the broad strokes of what capital will be used for (is the company a cannibal that buys back stock, does it never issue stock, does it use debt for financial engineering purposes, is it a serial acquirer within its industry, is it some sort of conglomerate, etc.) and the runway left for additional capital allocation is what I’d be focused on. A great track record is fine. But, it doesn’t necessarily mean much. If you had a fund manager who started with a $10 million fund and ran it for 15 years and now it’s a $10 billion fund – how much should you care that his CAGR has been 30% since inception? That would be an amazing track record. But, there’s no similarity in available opportunities to allocate cash towards when you have a $10 million AUM versus when you have a $10 billion AUM. So, it’d matter how the record was created. Was it all done in big cap stocks even when the fund had $10 million under management? Well, that’s repeatable.

The past capital allocation record – no matter how good – only matters to an investor to the extent it’s repeatable. But, it doesn’t have to be exactly repeatedly. That’s where the philosophy of management, incentives, etc. would come into play. Management won’t be able to use the exact same playbook that determined capital allocation in the past. Eventually, the cycle will turn or the company will get too big or the opportunities it took advantage of will get arbitraged away by copy cats. But, the style of capital allocation can stay the same from a philosophical perspective. If management is focused on free cash flow per share – they can keep that focus as a $100 million company and a $100 billion company. If they have a strong bias toward never issuing shares or never paying a dividend or never buying back stock – those kinds of bias tend to be maintained for a really long time. A good example is just a total aversion to issuing shares, options, etc. This is such a small thing, but it can easily add 1% a year to your returns versus what you’d get in other stocks. And this tends to be something that is maintained over time. If a company kept the share count flat for the last 10 years – it’ll probably keep it flat for the next 10 years. If it bought back stock aggressively in the last couple downturns, it may do so in the next one. These are things you can observe and use in your stock selection even though they aren’t directly a part of the track record of the company’s capital allocation. For example, if a company in an entertainment or restaurant business aggressively bought back stock during COVID – that actually may not have added much to the track record you can judge them on in terms of share price results. Some of these companies are still quite cheap. And some have taken longer to recover from the COVID shutdown then they expected. But, this information is still really useful. They may have misjudged how long COVID would last. But, they were willing to buy into a crashing stock price. If they did it then, they’d probably be willing to do it under different circumstances. And those different circumstances might offer a much better payoff. The next crash won’t be due to a pandemic. It’ll be due to something else. So, the same action taken at different times will have different payoffs. It’s not always possible to judge management on the payoffs their capital allocation got in the past. But, it usually is possible to judge their past actions and get some idea what their future actions will be. For this reason, I’d focus more on how the past track record was created than on how to measure the past track record.