Updated: Feb 18, 2020
Our clients often ask us for an opinion on the future returns on US stocks. We thought it might be helpful to answer this question by highlighting some examples of the research and analysis that we use when estimating future equity market returns.
Spoiler alert: We believe that the likely return for US equities over the next is 4% to 5%. It could range from as high as 7% to as low as 1%. Please read on to find out more.
It’s impossible to predict what the stock market will do in any given year.
The chart above shows the annual percentage change in the S&P 500 from December 1927 to 5th February 2020. As you can see, there’s no reliable pattern or sequence that can be discerned from year-on-year returns.
Just because we can’t predict next year’s returns doesn’t mean that we can’t make meaningful and useful assumptions about future returns. To do this, we have to lengthen our time horizon. This is because fundamental relationships take time to play out.
Take valuation as an example. Clients often point to the stock market’s price-to-earnings ratio or the Cyclically- Adjusted-Price-Earnings (CAPE) ratio. They note that these measures of valuation are currently high.
Source: JP Morgan
What does a high-valuation mean for next year’s return? As the chart above shows, not that much. The correlation between valuation and next year’s returns is very low. In contrast, the correlation between valuation and returns over the next five years is significant.
Valuation becomes a more reliable indicator of future returns only when we lengthen our investment horizon.
The same is true of earnings. Earnings growth is extremely volatile on a year-on-year basis. In fact, when it comes to estimating future returns, earnings growth is the toughest assumption to get right.
Source: JP Morgan
Once again, extending the investment horizon can help. Over the long-term, real earnings growth tends to track real-per capita GDP growth. This makes intuitive sense. If earnings grew faster than the economy, the stock market would eventually become the entire economy! The long-term rate of economic growth acts as a natural speed limit on long-term earnings growth.
That’s not to say that earnings growth can’t depart from this long-term trend. But it helps to explain why taking a long-term perspective helps to even out a lot of the short-term noise.
The economist John Maynard Keynes wrote that investment returns were made up of two components, an investment return and a speculative return. The investment return is the earnings of a business (including dividends) and the growth of these earnings over time.
The speculative return is what an investors are willing to pay for the business. It changes with investor sentiment and is driven by crowd psychology; hence the “speculative” label.
In 1991, Jack Bogle, the founder of Vanguard, drew inspiration from Keynes. He used Keynes’ concepts of investment and speculative return as follows:
1. Current dividend yield (i.e. owners share of earnings)
2. Earnings growth
3. Change in valuation
The first two variables are Keynes' “investment” return and the third variable is the “speculative” return.
Bogle used this framework to do two things. Firstly, he decomposed the historical return of US stocks into these three key drivers of returns. Secondly, he used these three variables to estimate future long-term returns.
Bogle pioneered the technique that most asset allocators have used to estimate long-term returns ever since. Every asset allocator has their own tweaks (for example, to include or exclude stock buybacks), but nearly everyone relies on the same conceptual framework (derived from the dividend discount model).
Source: JP Morgan
The chart above shows the returns for US, emerging market (EM), European and Japanese stocks broken up into Bogle-esque categories.
You’ll notice that sentiment (i.e. change in valuation) is the dominant driver over shorter periods and that earnings and earnings growth become more important as the time horizon gets longer. You’ll also notice that variability (i.e. the range of outcomes) gets narrower.
Again, this is because a long-term investment horizon helps to tune out a lot of the short-term noise.
Here’s an example of a fund manager, AQR, using a similar framework to estimate medium-term (in this case 5-10 years) equity returns.
The 5-10 year expected return for U.S equities has also fallen from 4.3% in 2019 (lighter colours) to 4.0% in 2020 (darker colours). These are real (i.e. inflation-adjusted) return estimates. So, assuming inflation of 1-5 to 2%, the estimated return would be 5.5% to 6% per year.
The bars in the above chart cover a 50% confidence range. This range is 6% for US equities. In other words, AQR expect that there's a 50% chance that 5-10 year real returns for US equities will be between 1% and 7% (or 3% and 9% if you assume 2% inflation).
It’s important to think of future return estimates as a distribution – i.e. a likely scenario, surrounded by a wide range of possibilities. This is because several of the key inputs – valuation and earnings growth – can vary significantly over time.
It’s impossible to be precise with these sorts of estimates. Investors need to consider the range (i.e. the high and low extremes) and not only the most likely outcome.
Does that mean that this analysis is a waste of time? No, it can still be very useful. Here’s an example. Let’s assume that I asked you to select the football team that’s most likely to win next year’s championship as well as how many games they will win in the home-and-away season. It’s very difficult to do this. You may get lucky one year, but what about the next?
Suppose instead that I asked you to pick the top and bottom five teams at the end of the home-and-away season. Chances are, if you know something about football, you’ll probably get 3/5 of the top and bottom teams correct.
What lessons can we draw from this example:
· Luck is less likely to be a factor over an entire season (long-term) vs the finals (short-term)
· Ranking and sorting (especially if its excluding poor performers) is easier to do than picking the winner.
We encourage our clients to think of asset allocation as a ranking and sorting exercise. No investor is able to consistently pick THE winner. But we can create wealth for our clients and manage risk by ranking and sorting.
The table above summarises expected real returns for global equity markets. AQR use the dividend discount model as a baseline for estimating real returns. In this model, the expected real return is simply the sum of the dividend yield, expected trend growth in real dividends or earnings per share (EPS) and expected change in valuation. The source report applies this methodology by taking the average of two slightly different approaches.
1: Earnings Based
10-year average inflation adjusted earnings divided by today’s price multiplied by the US long run dividend payout ratio, add real earnings growth (approximately the US long run geometric average.
2: Payout Based
This method estimates net total payout yield (NTY) as the sum of current dividend yield and smoothed net buyback yield. Then an estimate of long-term real growth of aggregate payouts (including net issuance) is added. This growth assumption is an average of the smoothed historical geometric aggregate earnings growth and foretasted GDP growth.
For US equities, we observe an expected 5-10Y return of 3.3% under the earnings-based methodology and 4.6% for the payout-based methodology. The combined expected return figure is 4% real per annum.
This doesn’t mean that US stocks will earn 4% after inflation every year for the next 5 to 10 years. As the first chart in this post shows, it’s impossible to predict the exact path returns will take over time.
Some readers might be thinking, a 4% return net of inflation is hardly worth it. To these readers we respectfully point out:
· Every asset class is priced for lower future returns (relative to history)
· Equities arguably represent the best risk/return currently on offer
I challenge any reader who doubts the superiority of equities as a source of long-term income to READ THIS.
Long-term return estimates such as the estimate above can also be used as a reality-check on our performance expectations for active management. Let’s suppose that a particular strategy has historically beaten the S&P 500 by 5% per year. That’s 50% better than the S&P 500, assuming a long-term return of 10%.
It would arguably be unreasonable to expect 5% per year out-performance in an environment where US stocks are likely to deliver 6% per year. In this case, the expected out-performance is 83% better than the market return.
The lesson is that active return expectations need to be scaled back to more realistic levels in an environment where the returns across all asset classes are likely to be lower.