The pros and cons of discounted dividend models

Two months ago I wrote a long article about Discounted Dividend Models and how I’ve started to use them to value companies.

I then added a template to my investment spreadsheet so I could quickly build a standardised dividend model for any company which I could then refine and expand upon later.

After that I spent several weeks hammering out one dividend model after another so that I now have reasonably detailed models for all of my portfolio‘s 30 or so holdings.

As I now have a lot more experience of what it takes to build discounted dividend models in the real world, I thought it would be useful to outline some of the major pros and cons I’ve run into while using this approach.

Table of Contents

Let’s start with the positives:

Pro #1: Building dividend models helps you focus on the long-term

Most of the information investors look at is backward looking. Earnings, dividends, revenues, book value, debt; whatever it is, it’s a measurement of something in the past.

But the future is where the action is. It’s where businesses will grow or shrink, it’s where they’ll pay or suspend dividends and it’s where shareholders will get richer or poorer.

Before I started building dividend models I think I spent too much time looking at the hard numbers of the past rather than thinking about the uncertain fog of the future.

It’s not that thinking about the past is wrong, because it isn’t. It’s absolutely necessary. But looking at the past is only useful when it’s used to build your understanding of the future. And by “the future” I don’t mean the next year or two. I mean the next ten years at the very least.

Why bother thinking about where a company might be in ten years?

The answer is simple. When I invest in a company I want to buy it at a discount to its intrinsic value.

The intrinsic value of a business is the present value of all its future cash flows from and to shareholders (e.g. dividends and buybacks), discounted by an appropriate interest rate (I discount future dividends by 10% per year as that’s my target rate of return).

And since the intrinsic value of a business is the sum of all its future cash returns, looking out one or two years is not enough. What happens over the next ten years and beyond is far more important, and building a long-term dividend model is a fantastic way to keep you focused on that sort of time horizon.

Pro #2: Building a dividend model helps you think about markets as well as companies

Once you start thinking about the intrinsic value of a company and how it depends on dividends, buybacks and other cash flows over the next 10 years or more, then long-term industrial and geographic trends become important.

They’re important because the long-term fate of most companies is determined by the growth or decline of their core markets combined with the company’s ability to move into closely related adjacent markets.

Building a dividend model is a great way to force yourself to think about these things (as is reading books like Profit from the Core and Beyond the Core).

For example, one of my holdings is Domino’s Pizza Group (DPG), which runs the Domino’s Pizza franchise in the UK.

DPG is focused on a single subsector (manufacturing and delivering pizzas) and a single geography (the UK), and has no stated intention to expand beyond either, at least for now.

On that basis it would be both realistic and conservative to assume that DPG’s growth over the next decade will be limited mostly by its ability to expand within the UK pizza delivery market. And to answer that, we need to have an opinion on how much the market and DPG’s market share will grow or shrink.

For a long time DPG’s management have said they expect the UK to be able to support around 1,600 Domino’s stores. Currently there are 1,200, so that leaves around 400 more stores before the UK reaches saturation point. That’s a potential 33% increase in the number of stores, and given the company’s historic roll-out rate that could happen within the next decade.

Management also expect “system sales” (total sales across all of DPG’s franchisee and corporate stores) to reach £1.6-1.9 billion per year over the medium-term, up from around £1,350 in 2020. That’s an increase of 20-40%, also within the next decade.

Based on those goals, plus a few other things (such as historic profit margins, returns on capital and so on), it’s fairly easy to argue that a realistic and conservative growth rate for DPG over the next decade would be somewhere around 2-4% per year.

That relatively slow growth rate would likely leave the company with more cash than it could reinvest, and that’s probably why management recently announced a regular buyback program as a way of returning excess cash to shareholders.

The buyback for 2020 has been set at £45m, larger than the 2020 dividend which was a mere £43m, so these buybacks are not trivial. In my dividend model I assume DPG buybacks over the next decade average around 70% of the dividend.

The resulting reduction in outstanding shares would be enough to add another percent or so to the company’s per share growth, leaving DPG with an annual dividend growth rate of around 5% to 2030:

(click to enlarge)

Note: The above model has been extended to take account of buybacks, so it’s slightly more complicated than the template in my investment spreadsheet.

This approach to valuing companies is very different to just extrapolating Domino’s historic 10% growth rate out into the dark eternity of the future and then assuming that somehow it will find enough pizza lovers to keep growing that quickly. Perhaps it can, but that seems optimistic and I’d rather be realistic and conservative, at least when it comes to investing.

If I didn’t have to build a dividend model it might have been easier to gloss over the company’s long-term growth prospects, and the resulting valuation would have been less robust.

In summary then, companies operate in markets that are in long-term decline or growth, they use distribution channels which are about to be boosted or decimated by the internet and some of them have acres of room for growth within their core markets while others have long since reached the limits to growth.

All of this can be factored, realistically and conservatively, into a discounted dividend model.

Note: If you’re wondering about the impact of Just Eat and Deliveroo on Domino’s in the UK, my assumptions are that (a) food delivery aggregators will help grow the overall food delivery market, (b) Domino’s has a far more efficient vertically integrated supply chain which is a material advantage and (c) aggregators will undermine Domino’s market share. I assume the net effect of all this is highly uncertain but unlikely to be terrible for Domino’s, so I’m currently estimating the net effect as zero until reality proves otherwise. And if it does, I’ll adjust my model.

Pro #3: Building dividend models helps you think about business units and not just whole companies

As I’ve built dividend models for my 30 or so holdings over the last few weeks, I found that many of them can’t be modelled sensibly without deconstructing the business into its major business units.

There are limits to this of course, and I’m not going to deconstruct Unilever (which I hold) into all of its hundreds of brands. The data isn’t available anyway and it would probably be a massive waste of time anyway.

But breaking Unilever down into Beauty & Personal Care, Home Care and Food & Refreshment and then tracking how revenues for each business unit evolved over the last decade; that’s probably quite sensible. And you could do the same thing by its major geographic regions as well, e.g. Asia, The Americas and Europe.

If you laid all that out in a spreadsheet it would become clear that Beauty & Personal Care has grown by a third over the last decade, while Food & Refreshment has shrunk by about 15%.

You could then look at how much exposure the company has to each, how these areas are expected to grow (or shrink) in the future, how that all ties in with the company’s stated growth strategy, and then you could put together a realistic and conservative estimate of what the future might look like.

Okay, so to some extent I did deconstruct companies like this even before I started building dividend models, but having a dividend model as a concrete outcome makes it clear how this sort of information feeds into your final valuation.

Admiral is another example of a business where an analysis of its component parts may be useful.

Historically Admiral was primarily a UK car insurer, and it still is. But only just, because it has several smaller but faster growing businesses operating in different insurance and geographic markets.

Its international insurance businesses (including L’Olivier in France, ConTe in Italy and Balumba in Spain) have mostly been operating for more than a decade and have started to become consistently profitable in recent years (insurance is a price-driven business so it’s all about economies of scale, which means building scale is more important than making profits in the start-up and scale-up phases).

These international insurance business have grown aggregate turnover at a compound rate of 20% per year for the last decade, although of course that has started to slow as they’ve become significantly larger.

Their average combined ratio (the ratio of losses and expenses to premiums) has fallen almost every single year, from 164% in 2011 to 108% in 2020. Once it falls below 100% (meaning costs are less than premiums) these companies will produce material underwriting profits which will hopefully fuel additional dividend growth. For some context, Admiral’s UK combined ratio is usually around 90%.

Outside of car insurance, Admiral has successfully expanded into UK home insurance, with consistent profits now flowing from that business and with many cross-selling opportunities ahead.

Admiral has also recently moved into travel insurance (which obviously wasn’t great in 2020) as well as pet and home insurance in Europe.

It’s also started looking outside insurance with Admiral Loans, where it has has a toe in the water of the UK unsecured lending market, leveraging its expertise in pricing risk.

If I’m going to estimate Admiral’s dividends over the next decade and beyond then I need to have at least a reasonable understanding of the company’s various activities, which bits are growing and how fast, and how that might pan out over the years ahead.

This doesn’t necessarily mean you have to do a ten-year revenue and profit forecast for every part of a business, but it might.

In Admiral’s case I haven’t. I’ve just pulled out and analysed the historic data and used that to give me some “feeling” for what a realistic and conservative growth estimate might be for the next decade and beyond.

With Admiral this means I’m more comfortable estimating a higher perpetual growth rate (beyond ten years) because it has successfully expanded outside the UK and outside of its core car insurance market.

So rather than estimating perpetual growth of 3% (which is my default for high quality but UK-only companies) I’m currently estimating 4% for Admiral’s long-term growth rate. And if Admiral continues to successfully expand into new markets then I might up that to 5% which would then increase my estimate of Admiral’s intrinsic value.

(click to enlarge)

Pro #4: Dividend models give your investment research focus

So far I’ve said that building discounted dividend models can help you think about the long-term; specifically the long-term evolution of markets and the underlying components of a business.

This is much better than looking at a company’s short-term headline results, such as whether revenues, earnings and dividends are up down this year.

However, it isn’t like I wasn’t doing all of that before I started building dividend models. I was. I was thinking about the long-term, I was thinking about industrial and geographic markets and I was thinking about business units as well as whole companies.

What’s changed is that the work I do to understand a company, its markets, its business units and its long-term potential now has a much more concrete outcome. And that outcome is the dividend model.

I now analyse the evolution of a company’s major components because I want to build a good dividend model. I now look at how relevant geographic and industrial markets have evolved in the past and might evolve in the future because I want to build a good dividend model. And I now think about the long-term prospects of a business because I want my dividend models to be more than pure fantasy.

In other words, building dividend models is now absolutely central to everything I do as an investor. They guide my buy and sell decisions and my position sizing decisions.

And this doesn’t end once I invest in a company. Instead, the dividend model retains its central position. Every time a company puts out new quarterly, interim or annual results, I review and, if necessary, update the model. If a company announces a major acquisition or disposal, I update the model. If anything happens which is likely to materially change the company’s future, I update the model.

This turns all of the research from qualitative statements like “this company has strong competitive advantages” and “management have a good strategy for expanding into Asia” into a concrete and tangible model which hopefully encapsulates everything I think about a company’s prospects.

So those are the initial benefits I’ve seen from my first few months as a dividend modeller. What about the downsides?

Con #1: Building dividend models can take a lot of time

The only notable downside I’ve seen from building dividend models is that it can take a lot of time. Obviously it doesn’t take long to punch a few numbers into a spreadsheet. What takes time is the research which underpins those numbers.

Whether this is really a downside depends on how much you like analysing companies and how much time you have to do it.

If you hate analysing companies and you have no spare time, then building dividend models is probably a bad idea. But if you enjoy analysing companies and if you have a few hours here and there to spare, then I’d say building dividend models is definitely worth the additional effort.

Having said that, building dividend models for 30 companies over the last month or two was not always a fun experience and I wouldn’t generally recommend it. Far better to spread the analyses out so you don’t build more than, say, one model per week from scratch.

I’ve also found that holding 30 companies is too many, because at certain times of the year (such as during results season) there’s just too much work to do analysing all the results and updating all the models.

The obvious fix is to hold a more concentrated portfolio of higher quality companies. That’s the direction I’ve been headed since late last year anyway, having ended up with 35 holdings after an excessive buying spree during the 2020 market crash.

I now have a target of getting down to 20-25 holdings by the end of 2021, with perhaps 20 being the goal beyond that.

That may sound very concentrated to some people, but as John Maynard Keynes once said:

“It is a mistake to think that one limits one’s risk by spreading too much between enterprises about which one knows little and has no reason for special confidence”

“[A portfolio should hold a] careful selection of a few investments (or a few types of investment) having regard to their cheapness in relation to their probable actual and potential intrinsic value over a period of years”

John Maynard Keynes

Author: John Kingham

I cover both the theory and practice of investing in high-quality UK dividend stocks for long-term income and growth.

14 thoughts on “The pros and cons of discounted dividend models”

  1. Hi John,

    Thanks for your nice posts regarding discounted cash flow analysis. A useful reminder to be forward looking!

    With regards you CAPE updates I have missed seeing the 6mthly updates and it should make for interesting reading when you publish the next one!

    Keep up the good investing, we are certainly in interesting times.


    1. Hi Johan

      You’re welcome. I’ve basically been super-busy for the last few months and haven’t really had time to write many new articles. CAPE is near the top of the list so should be out sometime in April.

  2. Hi John

    Obviously very good ideas. I very much agree with number 4 in that it makes you think harder about the specific inputs to be made as opposed to some qualitative statements e.g. strong moat etc.

    How do you handle new prospects in this regard, e.g. is there a separate model for each of them?

    Are all your new experiences and ideas like building specific (dividend) models or taking the long term lease / rental liabilities into consideration to name just a few revolutionary enough to make you write an update of your book 🙂

    I’ve been following your blog for some time, keep up the good work!

    Best regards


    1. Hi Lubo

      Some people build multiple models which gives multiple intrinsic valuations, e.g. optimistic (O), baseline (B) and pessimistic (P). They then assign probabilities to each and then add up the probability-adjusted amounts. So if we estimate that outcome O has a 10% chance of occurring, B a 70% chance and P a 20% chance, then the intrinsic value of the business is O10% + B70% + P*20% (I think!).

      I don’t do that. I prefer to come up with a model is that is (a) conservative and (b) realistic. A conservative model is one where reality is very likely to turn out better than the model, perhaps 80% chance that reality is better. It also means there should be almost no chance of reality being much worse than the model. A realistic models means I can’t just say that Admiral is going to pay a 1p dividend next year and never pay another dividend. That’s (very) conservative but it isn’t realistic and being unrealistically conservative will hurt my returns because few if any companies would ever look attractively valued.

      So I just have one conservative model, and if the future for a company is so uncertain that a single model doesn’t capture the range of possibilities, then I probably won’t want to invest in that company (hence my preference for firms like Admiral or Unilever).

      As for the book, I am working on a second edition which I hope to release in 2021, but the way things are going I think it will probably be 2022 before its done. Also, I want to get some more experience with dividend models and active position sizing before putting them into a book because I want to be fairly confident that they actually add value and are usable in the real world by real people!

  3. Hi John

    Love the adoption of the dividend discount model which with the QualityValueDefensive considerations provides not just indicative Buy and Sell prices but a target allocation as a percentage of the portfolio.

    I have been a subscriber for some years and admit that in the past I have confined myself to the monthly Newsletter and blogs. But with the introduction of the new dividend model I actually braved a look at the Investor spreadsheet – a wonder to behold! I shall certainly be looking at this much more closely in the future and I would urge other subscribers, even if they are numerically challenged like myself, to take a look!

    1. Hi Jonathan, thanks.

      I’ve generally had good feedback from subscribers about the new approach to valuation and active position sizing. I certainly think it’s much better than the old “hands off” approach to position sizing.

      As for the spreadsheet, I was in two minds about whether I should put all of the detailed numbers into it in case it looked bewilderingly complicated.

      I know that’s a risk, but I thought it was better to be as transparent as possible and to provide all of the underlying data so subscribers at least have the option of using it or ignoring it as they see fit.

  4. You probably want to revisit your Dominos analysis or maybe was just a misquote. They actually have operations in some foreign countries which they are actively divesting and that will have an effect in the company.

    1. Hi L, DPG does have some international operations, but they’re non-core and they’re being sold off so that the company can focus on the UK and Ireland. It has a more material investment in the German Domino’s franchise, but DPG is a minority shareholder and doesn’t run the German business. As far as I can tell management have no interest in international expansion until they’re happy that the UK business can cope with Just Eat, Deliveroo etc.

      1. Hi John
        I was sure you were aware of it, I just thought of mentioning because from you text I read something different then I went to check again. I’m still studying the UK branch as I am more familiar with the US operations. Thanks again for the excellent reading as always.

  5. I prefer a discounted free cashflow model as it can apply to companies which reinvest earnings, and more important to companies that do buy backs.

    Some people become so focused on the formulas and they do not realise the business is disrupted or shrinking.

    For example, Admiral which in my opinion is well overvalued. The problem with this type of business is that with introduction of EVs, insurance will be cheaper, and less profitable. Plus that firms like Tesla etc would like to dominate this market and disrupt it. In the end, they can monitor what you do with the car!

    To give you an example, you can invest in a better insurer – Progressive Corp in the U.S., paying 4.8% dividend and pay for a share 9.4 times last year net earnings. It has a is higher dividend than Admiral, better dividend coverage, and half the price per share.

    I think even Progressive Corp is expensive and I am going to sell it soon, because I do not think it will be able to get hold of the sort of AI information it needs to be able to underwrite correctly and remain profitable.

    1. Hi Eugen

      A good discounted dividend model should take account of buybacks too, although it does make the model more complicated and I’ve had to do this for several of my holdings which consistently make meaningful buybacks.

      In the case of a company that ONLY does buybacks, such as Berkshire Hathaway, it’s still true that at some point it must pay a dividend (i.e. actual cash to shareholders).

      That’s because there is a limit to buybacks because a company cannot buy back 100% of its shares. If it did it wouldn’t have any owners. If the company bought back every share but one, the only way it could “buy back” that share would be to close its operations, convert its net assets into cash and then return that cash to the last remaining shareholder. And that is effectively the company’s terminal dividend (although I guess it could split the share and carry on doing buybacks, but even then unless you think the company is going to exist forever it will still fold at some point and hopefully return its remaining net assets to shareholders as a terminal dividend).

      But personally I like dividends so the added complexities of valuing a non dividend paying company like BH isn’t an issue.

      As for Admiral and Tesla, EVs won’t be a material part of the total pool of road vehicles for several decades, so even if EVs do have a big impact on car insurers (I’m not sure why they would as they’ll still crash and run people over) then Admiral has several decades to adapt. Also, Admiral is already expanding into insurance for homes, pets and other non-car items, and it has an expanding unsecured lending business. It already calls itself a financial services company rather than an insurance company, so perhaps car insurance won’t even be its main business in 30 years.

      And Admiral and many other insurers can already monitor how you drive either with a black box or more recently via a smart phone app.

      Having said all that, you’re right to raise AI and EVs as potential disruptors, and that’s exactly the sort of thing that investors and companies should be thinking about. In this case though, I think Admiral already has a reasonable grip on AI (risk pricing through data analysis is a core competency of a good insurer) and self-driving cars (such as its relationship with Ford where – I think – Admiral underwrites Ford’s fleet insurance in the UK, the idea being to get closer to manufacturers in case insurance moves to the manufacturer with self-driving cars).


      1. John, given that a share could be split in 5, 10 or 1,000 shares, the potential to buy backs is unlimited.

        Re Admiral, I hope you are right. There is so much disruption going on now, I do not think there would be 20 years available for these old companies to adapt.

        Starting an insurer from scratch now is a lot easier, access to capital is a lot cheaper, and there is no need to have all those complicated systems from the past.

        Pay as you go is an interesting insurance service, and the increase in automation where we would not own the car, platforms like Tesla or Uber using “captive insurance”, are new models that would not do good for existing insurers.

        Insurers make profits because the market is fragmented, each person has to arrange his/her own insurance, but with self-driven cars, this is not needed anymore.

        Firms like Tesla, Waymo (Google) could get a professionals specialised in procuring involved which has to insure only certain risks, as many others could be now kept on the balance sheet (captive insurance), and premiums paid to reinsurers drop to less than 10% what insurers were getting. And there is no place for an insurer anymore, just a broker to arrange some big risks insured.

      2. Hi Eugen,

        On the buyback point I agree, splitting shares allows buybacks to go on “forever”, and if that means the company never ever pays a dividend or a cash return until the day it folds then I would still say the company never had any economic value. From a more practical point of view, dividend models are typically used to value dividend-paying companies (that’s what I use them for), so valuing a company only making buybacks is not really an issue.

        Your points on insurance are all valid and worth pondering. For now my position is that disruption is one possible future outcome, but the timing and magnitude of the risk are uncertain enough that I’m still comfortable investing in traditional insurers for now.

      3. On the buyback topic again, if a company only does buybacks and never dividends, then you can value the company using a DDM by looking at the whole company rather than on a per share basis.

        So imagine you’re thinking about buying a whole company that only does buybacks. If you bought it, would you still do buybacks, buying back shares that you already own? No, because you’d be both buyer and seller. So instead you would probably just take the cash used to make buybacks and pay them to yourself as dividends.

        On that basis I think it would be reasonable to value a company like Berkshire by imagining cash used for buybacks are instead paid as dividends and then using a DDM.

Comments are closed.

%d bloggers like this: