Realinsights: Growth Adjusted Dividend Yield

• Rather than being polar opposites, we argue that both value and growth dimensions complement each other in the valuation process. Growth considerations can be incorporated into traditional value investing.
• In particular, this paper focuses on a novel method of incorporating growth considerations into a common value workhorse - the dividend yield factor.
• We show detailed derivations from first principles, and as far as we are aware, this is the first time this methodology has been applied for dividend yield.
• Our new signal, growth adjusted dividend yield, outperforms ordinary dividend yield across most sectors globally based on our empirical modelling.

It is well known that value strategies have had a disappointing run over the last decade. This has led to some market observers claiming that Value is dead. (see our Realinsights response against this claim¹). However this dichotomy between Value and Growth is overly simplistic. The death of “Value” is not true. Paraphrasing from Buffet: what investor does not seek to buy undervalued assets? It is how you define what is undervalued that matters.

We should not paint “Growth” and “Value” to be opposing forces. Instead, we should see them as being two dimensions of the valuation equation, or two sides of the same coin. A simple way of looking at this is by starting off with the Gordon growth model, i.e., the price of a stock today 𝑃 can be determined roughly from expectations of FY1 dividends 𝐷₁, the required rate of return r and long-run growth expectations 𝑔.

Rearranging this simple formulation leads us to,

The expected required rate of return is a linear combination of “Value” (Dividend Yield) and “Growth” (dividend-per-share growth). Whilst this is a simplistic view, it illustrates the point that both factors are important determinants for expected returns. Intuitively, both growth and yield dimensions are considered when valuing a company from bottom up fundamentals. However often and incorrectly, the yield component is highlighted as “value” and the other being categorized as “growth” - something that value investors discard with disgust. We argue it does not have to be this way, and that value signals should also account for growth considerations.

For instance, strong perceived growth prospects can keep an expensive stock expensive for a prolonged period, delaying the inevitable value reversal. Only when growth prospects diminish do value reversals come into play. With more growth being baked into valuations, future cash flows become increasingly long-dated, lengthening equity duration. This in turn makes the stock increasingly more sensitive to growth revisions (numerator shocks) or changes in the discount rate (denominator shocks). A quote from Elon Musk colorfully illustrates this:

We argue that over the last few decades, the importance of growth 𝑔 has risen. We believe this is driven by two key themes:

• The global rise of intangibles assets, and in particular in IT, Consumer Discretionary and Health Care sectors in Developed markets.
  • Intangibles, especially the research component in research & development (R&D) is expensed and not treated as an Asset due to the conservative and backward-looking nature of accounting. However, needless to say, in an increasingly technology dominated society, research expense is a strong indicator of future success.
  • We do not dispute that R&D and technological development has always played a leading role in human history. However, it is only recently, since the advent of modern computing, has intangible assets (data and information), played such an pivotal role across sectors. (For instance, the industrial revolution was largely about an increase in tangible assets.)
• Loose monetary policy and stimulus, especially by the US Federal Reserve, has led to an influx of ‘cheap’ capital.
  • This influx of liquidity in our economy has meant the prices of financial assets have increased. We argue that stocks with more growth ambiguity have more room to accommodate investor imagination, and thus it is easier for them to get bid higher in a low interest rate environment. Their future cash flows are long dated and benefit more when yield curve is ‘flattish’. A piece by Ray Dalio explains this².
  • Furthermore, this unequal distribution of capital also plays on the creditworthiness domain. Investors are more willing to lend to larger capitalization stocks with high creditworthiness, delaying the inevitable reversion. Smaller cheaper value stocks are finding it hard to secure financing at the same rates as their larger cap peers. This has also played out at the individual level. Trickle-down monetary stimulus has led to further wealth inequality. We can see this unfold on a broad macro level. As French economist, Thomas Piketty points out, when the rate of return on capital exceeds the rate of growth on output and income, inequalities naturally emerge.

These have been recurring themes for the Realindex research team. Last year, we updated our core process to incorporate intangible assets into Book value. We have similarly considered incorporating intangible assets into Book Yield³. With regards to the impact of low interest rates, Head of Investments, David Walsh, has also written several Realinsights papers on Zombie firms and how that impacts value orientated strategies⁴.

Not being aware of these themes can lead investors down value traps. At Realindex, we have always been cognizant of this. Our core value-orientated process has a sophisticated multi-factor overlay to mitigate these weaknesses.

In this research paper, we highlight one way we have been able to incorporate growth considerations into our value signals.

Plain vanilla dividend yield 𝐷₁/𝑃 is a staple quantitative value workhorse, that has been employed by almost all equity investors. Realindex is no different - dividend yield is one of the many signals that we look at for large capitalization stocks. Here, we examine if we can modify the traditional dividend yield signal to incorporate individual stock growth considerations under a consistent framework.

We limit the scope of this research in focusing on adjusting dividend yield for growth, however, we acknowledge that similar techniques to what we describe below can also be applied to other value signals. For example, incorporating growth considerations into P/E yields the PEG ratio. For interested readers, we refer to the work of Ohlson and Juettner-Nauroth (2005)⁵.

Before we begin, we can ask ourselves: what’s the easiest way of incorporating growth into dividend yield?

We could, for instance, go with the cost of equity equation shown earlier in our introduction. This simply adds dividend yield with forecast growth in dividends. We are against this approach for a variety of reasons. Firstly, proxying for long term 𝑔 is hard. One could, for instance, use analyst consensus on long term EPS growth, however, the variability in 𝑔 will drown out the 𝐸₁/𝑃 component almost entirely. Furthermore, the linear construction of the two terms throws away potential interactions that may occur between yield and growth.

Ultimately, what we envisage is a new dividend yield signal constructed in such a way that the old dividend yield signal 𝐷₁/𝑃 is nested within it, and that a growth dimension 𝑔 is also embedded in a manner in which it does not completely dominate the value component, given we know that 𝑉𝑎𝑟(𝑔) > 𝑉𝑎𝑟(𝐷₁/𝑃). Of course, the new construction also needs to be theoretically sound, and not some ad hoc aggregation.

The signal we propose is:

This is not as daunting as it first looks. When growth 𝑔 is zero, our signal collapses to the traditional dividend yield signal. And thus, we have the nest property we desire. The relationship between traditional dividend yield 𝐷₁/𝑃, growth 𝑔 and growth adjusted dividend yield may not be so clear in the formula above, and therefore we show it interactively below (please see online version for interactive chart).

At the boundaries we note:

• When DY is zero, Growth Adjusted DY also remains at zero.
• When growth is zero, Growth Adjusted DY is the same as DY.

When DY and growth are both positive, we note that our new signals slightly rewards stocks with higher growth.

In this section, we provide a step by step derivation of growth adjusted dividend yield. It explains our choice for 𝑔, and why the formulation is as such.

We base our model framework off Ohlson and Juetter-Nauroth (2005) “Expected EPS and EPS Growth as Determinants of Value”, Review of Accounting Studies.

We begin with two equations that are always true.

First, there is the standard dividend discount model:

Next is an identity introduced by Ohlson and Juetter-Nauroth (2005).

This is true for any 𝑦_𝑡 , as everything on the right-hand side cancels out.

By adding the two ‘always true’ equations, we get something that is also always true:

Rearranging this yields us,

So far, we have not made any assumptions. We have only created a few notational definitions, and thus this equation above is still always true.

Let us visualize future cash flows in the chart below. In Panel A below, we illustrate a very simple discounted cash flow. The price is the sum of all the future cashflows (increasing due to growth) after discounting them. In Panel B, we shade in orange of what has been incorporated by 𝑦₀=𝐷₁/𝑟. As you can see, we’re only discounting future cash flows assuming no growth.

The maths here is quite simple, the net present value of the orange bars in Panel B is

Notice how the numerators are all the same. And if we multiply both sides of the equation by (1+𝑟) we get

And thus the difference between the two equations yields 𝑦₀=𝐷₁/𝑟.

By the same logic, you can see what 𝑍₂, 𝑍₃, … represents graphically in Panel C. And thus, the traditional dividend discount model can be re-expressed as,

In effect, it is simply a different way of slicing and summing up discounted cash flows. Traditionally, we slice vertically, and here we are slicing horizontally.

Notice how 𝑍₂ is the ‘present’ value of all future 𝐷₂−𝐷₁ sliver of dividends at time 𝑡=1, and 𝑍₃ is the ‘present’ value at time 𝑡=2, … Therefore, they all need to be further discounted to time 𝑡=0.

Now we are ready to make our first assumption. We assume that the 𝑍_𝑡+1 evolve by 𝛾. By this, we mean,

If 𝛾=1, then we are assuming that the increase in dividends is a constant dollar amount over time. This means the growth rate in DPS asymptotes towards zero as time extends to infinity. If 𝛾>1, then we are assuming that the increase in dividends over time is by more than a constant dollar amount.

Using the same useful trick we showed earlier, the infinite sum can be expressed as,

Since 𝑍₂ = (𝐷₂−𝐷₁)/𝑟, let’s remove 𝑍₂ altogether from our pricing equation.

Let’s divide both sides by 𝐷₁,

Here we introduced 𝑔 as the percentage growth in dividends, i.e. the short term growth rate 𝑔 = (𝐷₂−𝐷₁)/𝐷₁. 𝛾 modulates longer term growth.

A bit of rearranging yields us this quadratic equation,

Taking only the positive root, our expected return for a stock with short term 𝑔 and long term 𝛾 growth considerations.

This is the generalized formula. Estimating 𝛾 for each individual stock is problematic. Therefore we make the simplifying assumption that 𝛾=1. This is not unreasonable, because we’re not enforcing all future DPS growth to be zero, but rather slowly over time, DPS growth asymptotes towards zero from current levels 𝑔.

When 𝛾=1,

This is our growth adjusted dividend yield formula. It can also be rearranged as

Notice when growth 𝑔 = 0, i.e., no near term growth, our signal collapses to,

which is simply the standard FY1 dividend yield signal.

The model derivations show clearly why growth 𝑔 needs to be defined as DPS growth. We have also empirical tested other formulations for growth, such as earnings growth. We find that expected analyst consensus dividend growth is superior to consensus forward earnings growth in this context as the latter tends to be significantly more unstable. The way we formulate 𝑔 in our signal is such that 𝑔 ≥ 0. It is not uncommon to see negative earnings growth by analysts, whilst it is relatively more rare for analysts to predict negative dividend growth.

In the next section we will run a horse race between existing FY1 DY and our ‘new and improved’ growth adjusted signal across developed and emerging markets. We will also examine performance improvement across sectors and correlations.

In the charts below, we show recent performance of dividend yields for developed and emerging markets. This is constructed using High-low fractile performance (i.e., Q5-Q1), based on market cap weighted fractiles. The signals are also region sector normalized.

We find significant improvements in fractile performance in Global, Developed and Emerging Markets after adjusting for growth considerations. As a standalone factor, growth adjusted dividend yield is still quite problematic with regards to its recent performance. However, it is clear that by incorporating growth, we are able to construct a better value signal. The correlation between dividend yield and growth adjusted dividend yield is circa 0.8, making the latter a reasonable replacement for the former. The adjusted signal also exhibits lower volatility than the original signal.

Fractile Performance Charts

Source: Realindex, from Jan 2005 to March 2021

Fractile Performance Charts

Source: Realindex, from Jan 2005 to March 2021

Fractile Performance Charts

Source: Realindex, from Jan 2005 to March 2021

It is well known that dividend yield works better in certain sectors than others. For instance, they perform better in REITs (where 90% of earnings is paid out), Utilities and Energy sectors; and worse in IT and Heath Care.

By adding our growth adjustment, we see improvements across most sectors. Adding growth to dividend yield did not reduce the information coefficient (IC) for traditionally effective sectors such as Utilities and Energy (in fact we see substantial improvement), but it also improved the efficacy in sectors such as IT and Health Care.

In the tables below we show average sector ICs from 2005.

Global

Developed

Emerging

Finally, we have a look at stock examples to see how the growth adjustment have individually affected dividend yields. When applying growth adjusted dividend yield, we generally apply at least sector normalization, and it some circumstances, region-sector normalization. Due to normalization, looking at individual stock differences may not a particularly meaningful. Nevertheless, it provides readers with a ‘feel’ of the growth adjustment (GDY_NTM) to original dividend yields (DY_NTM).

Developed

Largest cap dividend paying stocks in Developed Markets

Source: Realindex, Factset as at June 2021

Top 10 changes in Developed Markets

Source: Realindex, Factset as at June 2021

Emerging

Largest cap dividend paying stocks in Emerging Markets

Source: Realindex, Factset as at June 2021

Top 10 changes in Emerging Markets

Source: Realindex, Factset as at June 2021

At Realindex, we believe that trading signals need to be continually evolving as markets and societies change. In particular, value signals definitions need to be continually reviewed and updated. Last year, we wrote about incorporating intangible assets into Book Value (and also adding back intangible expenses in Cash Flows) in our accounting-weighted core process. We saw it as a necessary change to reflect the growing percentage of intangible assets globally, and especially in developed markets. In this paper, we tackle the dividend yield signal, which had a lacklustre run over the last ten years. We derive a theoretical model that incorporates growth considerations, and we show that this has been able to improve performance globally in both developed and emerging markets across sectors whilst remaining largely faithful to the original signal in terms of correlations.

1. _{https://www.firstsentierinvestors.com.au/au/en/adviser/insights/latest-insights/value-is-dead-long-live-value.html}
2. _{Dalio, Ray (2019) The World Has Gone Mad and the System Is Broken, https://www.linkedin.com/pulse/world-has-gone-mad-system-broken-ray-dalio/↩︎}
3. _{https://www.firstsentierinvestors.com.au/content/dam/web/australia/realindex/realinsights-incorporating-unrealised-intangibles.pdf}
4. _{https://www.firstsentierinvestors.com.au/au/en/adviser/insights/latest-insights/zombie-companies-affecting-markets.html}
5. _{Ohlson and Juettner-Nauroth (2005) Expected EPS and EPS Growth as Determinants of Value, Review of Accounting Studies 10, 349-365}