In this section we provide a brief summary of the available historic evidence on long-

term returns on stocks and shares, as well some alternative investments. To do this
properly, we shall argue that we need a lot of data - ideally looking at returns both over
very long periods of time, and in many different markets. Fortunately both of these are
possible: thanks to the fairly recent efforts of financial economists in building datasets, we
can now look at up to two centuries' worth of data for the USA, and a full century's worth
for a fairly wide range of other countries.

Many years, many countries

The first reason for looking at long runs of data is straightforward. Except for the famed
activities of the 'day traders' during the 1990s boom, investing in stocks and shares is
generally agreed to make sense only for someone with a reasonably long investment
horizon. Most people who save systematically do so because they are saving up for their
retirement. This means that the period over which they save may be anything up to 40
years. Since, as we shall see, there is a lot of short-run volatility in stock returns, you need
a lot of data to get an idea of what longer-term returns look like, once this short-run
volatility washes out.

The second reason for using long runs of data is more subtle. In the previous section, we
argued that one important benchmark against which to assess the value of the stock
market requires you to have an estimate of the return that the 'typical' investor expects for
the future, or would have expected at some point in history: i.e. a reasonably reliable
estimate of the 'Re' that feeds into the Dividend Discount Model in Box 1.1.
Unfortunately, we can never actually measure the returns that investors expected. All that
we can measure is what they actually received.

It is evident that, even over quite long periods, realised returns need not be equal to
expected returns. If they did, the experience of the bull market of the 1990s. would have
implied expected returns of 20% or more for a number of years, followed by a switch to
negative expected returns in the bear market of the new millennium. But this would be
manifestly absurd. There is no evidence that rational investors were expecting to receive
either these very high, or negative, returns in advance. (It's especially easy to rule out the
latter, because in the late 1990s investors always had the alternative of holding a safe
asset that would have yielded quite respectable positive returns, so no rational person
would have held stocks if they were expecting to make losses.)
In practice, of course, realised returns can be broken down into two elements. The first
element is what investors expected to receive; the second is the difference between what
they actually got, and what they expected. One argument for using very long runs of data
is that, over sufficiently long periods, we can hope that the impact of pleasant mistakes
(i.e., underpredictions of returns) will be offset by unpleasant mistakes (i.e.,
overpredictions of returns). If the average error in making expectations is close to zero,
then historic averages of realised returns should be fairly close to revealing the average of
investors' desired returns.

Of course, it is quite possible that expectational errors may not so conveniently average
out to zero, however long the dataset we employ. It is easy to point to examples where
even very long-run historic average returns may still give a some- what biased picture of
expected returns - especially when we look only at returns for markets that have been
relatively successful, such as the US. This provides us with a strong justification for
looking at data, not just from many years, but also from as many countries as possible.

The other, and equally important, reason for looking at returns in a range of markets is
that financial markets throughout the world have become integrated to an ever-increasing
extent. Since virtually any investor can now invest, in principle, in stock markets the
world over, the 'typical investor' should in principle be (and, to a great extent, in practice
is) the same the world over for all stock markets.

2, Two Remarkable features of US stock returns

We start by looking at data for the best documented of all major stock markets: that of the
USA. While we shall argue that it is probably not appropriate to regard the historic
performance of the US market as entirely representative in global terms, nonetheless we
can learn a lot from looking at some of its key features, before going on to compare it
with information from other markets.

Thanks to the dataset constructed by Professor Jeremy Siegel, of Wharton, and

documented in his investment bestseller Stocks for the Long Run, we have data on real
returns on stocks, bonds and bills' (i.e., short-run investments, equivalent these days to
putting your money in a high-interest bank account, or a money market fund) over the
course of nearly two centuries. This dataset allows us to identify two remarkable features
of historic stock returns.
• The first is that the average real return on stocks has been surprisingly stable, at around
6.5% before costs. Since this finding is attributable to Professor Siegel, we have in the
past referred to it as Siegel's Constant. It is worth emphasizing that constants are rather
rare in economics. Siegel's Constant thus deserves more attention that it has yet received.

• The second remarkable feature is that, although stock returns have been very risky in the
short-term, they have not actually been as risky over long horizons as might have been
expected. To be more precise: if we did not know what the long-term risks had been, but
just had some short-term data, we would assume that the long- term risks would have
been greater than experience has shown them to be. It is this feature, as much as the first,
that has helped to give stocks their desirable properties for the long-run investor. We shall
revert to the surprising safety of stocks as long-run investments in Section 1.6, where we
shall argue that it provides a very strong piece of indirect evidence that the notion of value
makes sense. Figure 1.1 and Figure 1.2 serve to illustrate the first of these remarkable
features.

Figure 1.2

The stability of the real return on stocks

Figure 1.1 shows the total real returns (i.e., both from dividends and from capital
appreciation, after adjusting for inflation) that investors in US stocks have received each
year since 1802.5 It may seem strange to be talking about stability in something that the
chart shows has varied so much. The chart also shows a range of values, between which
the stock return has fallen 90% of the time. As this varies from a positive return of 30%
down to a negative return of 22%, the chart is a reminder of just how variable returns
have been in the short-term.

However, although the one-year returns vary a great deal, the ups and downs will even
out over time, so that we can identify the average return with considerably greater
precision, especially given that we have nearly two centuries' worth of data. Figure 1.1
shows returns again, but with the range of uncertainty about the average return, which is
very much narrower: we can be 90% certain that the average return lies somewhere
between 4.9% and 7.7% (with our best estimate of the actual average being 6.75%).

It is this average return that we refer to as Siegel's Constant. Even after nearly two
centuries, we cannot of course be sure that it really is a constant. And even if it is, we
cannot know with certainty what its true value actually is, but we can say that it cannot lie
too far from our best estimate of 6.75%.

We shall show that our analysis of value can help to explain one thing about stock returns,
which is why returns get pulled back towards this average value more rapidly than might
be expected; but this does not explain the apparent existence of Siegel's Constant itself.
The question as to why it is, or appears to be, so stable is an important challenge, which
needs to be solved before we can have a complete understanding of how capital markets
work. We wish we could say that we have arrived at such a complete understanding, but
we have not. (In our own defence, we should add that this is not a question that the rest of
the economics profession seems yet to have got around to asking, let alone resolving.)

If you are not fully convinced that the stability of historic stock returns is re- markable,
two further charts may help to persuade you.

The first of these (Figure 1.3) provides a comparison between returns on US stocks, over
rolling 30-year investment periods, compared with the returns over the same period on
government bonds and bills', i.e., whatever was the relevant reasonably safe short-term
investment at the time. Taking such a long rolling average inevitably smooths out a great
deal of the variability of the stock return that was visible in the first two charts (though
equally it does not remove it entirely), but the tendency to revert to the reasonably stable
average that we have called Siegel's Constant is quite evident. In sharp contrast, there is
no such tendency for the competing investments of bonds and bills. In the nineteenth
century these offered returns only somewhat lower than stocks; were very much lower
during the middle part of the twentieth century (real returns on bonds and bills in the
inflationary 1960s and 1970s were routinely negative); and then recovered to provide
distinctly more respectable returns in the latter part of the twentieth century.

Figure 1.3

Real returns* on US stocks, bonds and bills since 1831 *Rolling 30-year compound
average return.

Source: Siegel 1801 to 1899 and DMS 1899 to 2010.

Tempting as it might be to dwell on the explanations for why returns on alternative

investments appear to have been much less stable historically, we should not be deflected
from our primary purpose, which is to demonstrate the relative stability of the return on
stocks and shares.7

But we do note in passing that Figure 1.3 provides an important part of the rea- son why
valuing shares in relation to competing assets is pretty much a lost cause. These
competing assets have offered such variable returns, historically, that there is much less
reason to expect their returns to be stable in the future. Without that stability, they cannot
possibly be used to provide a benchmark of value for stock

markets.

The second piece of evidence for the relative stability of the stock return comes from
broadening out our data to include a range of stock markets. Figure 1.4 shows evidence
on returns since the end of 1899 in 14 different stock markets, taken from Elroy Dimson
et al's definitive dataset, first summarised in their book Triumph of the Optimists (2002)
and subsequently updated. The key features to note in this chart are:

While there has been a very wide range of historical experience in the countries covered
(a number of which have gone through major dislocations, such as wars and
hyperinflations), the range of historic average stock market returns is not actually all that
wide. Only two countries have had average real returns in local currency of less than 2%,
and only one of over 7% - a range of experience not all that different from the range of
uncertainty as to the true value of Siegel's Constant that we noted in relation to US data.
Most of the major, and more stable, markets were in a distinctly narrower range.

For virtually all countries returns have been fairly similar, whether expressed in local
currency or in a common currency such as sterling. Thus, for example, UK investors who
had invested in a portfolio made up of investments in each of these countries could have
earned a distinctly more stable return than if they had only invested in one, or a few.8

The chart also shows that the US experience of an average return of close to 7% in
Dimson et als sample has been rather better than the weighted average of all the markets
shown (where the weight on each markets is given by its size: thus small markets such as
Belgium have a much lower weight than large markets such as the US or the UK), which
has been somewhat below 6%. But this should not come as a great surprise. The relative
success of the US economy over the course of the twentieth century was not predicted in
advance. Thus investors in US shares received, on average, more pleasant surprises than
in other markets. This gives us grounds for thinking that Siegel's Constant, based on US
data alone, may be something of an overestimate of the true expected return of a typical
global investor over this period."

• The chart also shows the amount by which the average return on investments in stocks
exceeded that on investments in bills over the twentieth century. The fact that this
estimate of the equity risk premium' actually shows more variation across different
markets than the stock return itself is due to the fact that, in many markets, real returns on
bonds and bills were more risky than on stocks and shares, owing to the impact of
inflation. We have not included the data for Germany, for which the true riskiness of
bonds and bills cannot be shown, since, during the course of the hyperinflation of the
early 1920s, investors in bonds and bills effectively lost all their money. But, more
generally, it is worth noting that countries with relatively poor average stock market
returns typically had a rela- tively high equity premium - implying that bonds and bills
were hit by the same bad news as stocks, but were typically hit even harder.

Figure 1.4

Compound average real returns in local currency Compound average return in dollars

Excess return vs bills

Global equity returns and premiums, 1900-2009

Source: Dimson et al (2002); updated by the authors.

The key lesson that you should take away from this section is that there is quite a lot of
evidence that realised returns on investing in the stock market over long periods have
been fairly stable both over time and across different countries. Of course this does not
tell us that the returns investors will expect in the future will be the same as they expected
in the past, but it does tell us that if you had made this assumption in the past you would
not have gone too far wrong. For this reason, we have a reasonable basis for using historic
average realised returns as a benchmark against which we can compare both actual
historic returns over shorter periods and prospective returns in the future.
This section discusses the importance of using long-term data to assess the returns on
stocks and shares and other investments. The first reason for doing so is to account for
short-term volatility and get a clearer picture of what longer-term returns look like. The
second reason is that we need to estimate the expected return of investors in order to
assess the value of the stock market, but we can only measure what they actually
received. Using long runs of data can help to offset the impact of expectational errors. It is
important to look at data from many countries as financial markets have become
increasingly integrated.

The article discusses the two remarkable features of historic stock returns in the US
market. The first feature is that the average real return on stocks has been stable at around
6.5% before costs, which is referred to as Siegel's Constant. The second feature is that
although stock returns have been very risky in the short-term, they have not been as risky
over long horizons as might have been expected. The article provides charts and data to
support these features, and highlights the challenge of understanding why Siegel's
Constant has remained stable over time.

The article presents evidence of the stability of historic stock returns by comparing them
with returns on government bonds and bills over rolling 30-year investment periods.
While alternative investments have offered variable returns historically, the return on
stocks and shares has been relatively stable. The evidence also shows that, despite a wide
range of historical experience in different countries, the range of historic average stock
market returns is not actually all that wide. This suggests that using historic average
realised returns as a benchmark against which to compare actual historic returns over
shorter periods and prospective returns in the future is reasonable.