How long is long enough when looking at asset class histories?

Greg Flash of Alphen Asset Management writes that "mean reversion" or "regression to the mean" has also been widely accepted as a powerful force in the fields of finance and economics (although it applies to many other fields like genetics and physics).

The law of compounding returns is often referred to as "a law of nature". Einstein apparently stated that it was the eighth wonder of the world (although I can't find a good reference to this). Although not referred to as a law, the theory or phenomenon of "mean reversion" or "regression to the mean" has also been widely accepted as a powerful force in the fields of finance and economics (although it applies to many other fields like genetics and physics).

The four most dangerous words in finance are said to be "this time is different". Although they have been used to comment on situations that do not involve mean reversion, the words are likely most dangerous when referring to it. At PSG Alphen Multi-Management we have often written about how we use "Back Testing" and "Strategic Asset Allocation" as tools in the management of assets. When back-testing asset classes, two issues are important: a) the quality of the data and b) how much of the data to use. Without good data, our models are useless. As the saying goes 'rubbish in, rubbish out'. Assuming reliable data, the question is how much to use.

We have data on the six asset classes: domestic cash, domestic bonds, domestic equity, US cash, global bonds and global equities going back to 1926. In addition, we have exchange rate data for US dollar/rand going back to the same period. Prior to the financial crash of 2008, many people that we interacted with would comment that using data going back that far was useless because things were different. Their arguments that the world going off the gold standard in the early 70s and the emergence of the New South Africa in the 90s meant that financial data older than 30 years was no longer applicable. In addition, when we said that we included the period of the Great Depression, we were told we were crazy because THAT would never happen again. Not so strangely, we don't hear these comments anymore.

Today I will focus on the annualised returns of the six asset classes mentioned above (all in rands) both in nominal and real (adjusted for inflation) terms. Below are two graphs with data tables attached of the nominal and real annualised returns of the six asset classes all in rands. I have presented the annualised returns for each asset class for the periods 5, 10, 20, 40, 60 and 84.2 (beginning of data) years.  

Points from Figure 1 above:

• Equities (domestic) over the last 5 and 10 years have delivered higher returns than the longer periods (except for the last 40 years).
• Bonds and cash (both domestic) over the last 5 years have delivered returns similar to the long term periods (60 and 84.2 years)
• The global asset classes (in rands) all saw the greatest performance over the last 40 years. This is largely attributed to the collapse of the rand in the early 80s. Consequently, domestic inflation over the last 40 years is also the highest.

• With the effects of inflation removed, equities (domestic) over periods 20 years and longer are very constant at around 7% to 8% real.
• Real returns for bonds and cash (domestic) for periods 40 years and longer are also fairly consistent and low (less than 2%).
• Global equities in real terms have returned 4% to 6% for periods 20 years and longer
• Global equities over the short term (5 years and 10 years) are far below the long term averages.