Investing is a tricky but unavoidable business for those looking to secure their financial futures. One answer as to how to invest is to follow the advice of Warren Buffett to his wife that on his death she should invest by purchasing an index fund. Sounds simple, but it assumes knowledge of which index or indices should be tracked. It is also rather disingenuous, as Warren Buffett clearly continues to believe that his concentrated stock approach can outperform a passive investment strategy or surely he would have implemented one within Berkshire Hathaway.
Warren Buffett’s idea was not new. In 1974 the economist Paul Samuelson wrote an article called “Challenge to Judgement” which suggested that overpaid, underachieving professional investors should quit and do something useful instead, like plumbing! He further suggested that someone should set up a cheap index-tracking fund. John Bogle read Samuelson’s article and set up Vanguard, the world’s first index fund business. Some 40 years later roughly half the money invested in US equity mutual funds is in trackers and indexation strategies are estimated to account for perhaps 20% of US equity market capitalization.
In their book Noise: A Flaw in Human Judgement, Kahneman, Sibony & Sunstein argue that there are two types of error in judgement: bias and noise. Bias is systematic deviation. Noise produces random scatter. When these two are combined the result is variability in outcome versus that expected. The great advantage of employing an index-tracking strategy is that it removes bias and minimizes noise. However, what if Samuelson was misguided in his belief that financial markets operate efficiently? What if it is possible to identify professional investors who, at least for a period, systematically outperform their peers?
Investigating Noise
In compiling the ARC Private Client Indices monthly performance data is gathered from around 120 discretionary investment managers on more than 300,000 discretionary private client portfolios. Underlying portfolios are placed into one of four risk profiles based on the volatility of their returns relative to the global equity market. Each contributing manager can then be ranked and changes in rank monitored.
To investigate noise, change in manager rank has been considered over three different time periods: quarterly; annually; and over a 3 year rolling period. The a priori assumption is that as the time frame lengthens the impact of noise will dissipate and evidence of any bias will become clearer. In other words, as the timeframe increases it should be possible to spot those managers exhibiting skill.
The charts below plot the change in manager rank for the ARC PCI Sterling Steady Growth risk category for these three time periods for data ended December 2021. A green bar indicates that the ranking has improved; a red bar indicates that the ranking has declined. Suffice it to say that the data looks very “noisy” and it is difficult to see any pattern.
To make the charts easier to interpret, an assumption can be made that there are no systematic outperformers or underperformers. Rather the manager rankings each period are random. If that is true it is possible to plot an “average expected change” line which can be compared with the average actual change. The chart below presents these results with the grey line representing no bias and the blue line representing the actual outcome.
The three month period ended December 2021 reveals that it was a quiet quarter as regards ranking changes from the previous three month period ended September 2021. For the 12 month period ended December 2021, the movements in the manager ranking table are in line with the expected change assuming no bias. Finally for the 36 month data, compared with the 3 year ranking for the period ended December 2018 the ranking appears to have been far more stable than might have been expected, strongly suggesting that performance bias is dominating market noise.
Evidence of Bias
The final step of the analysis is to generalise these results by analysing outcomes back to December 2003. In the charts below the grey line continues to represent the expected change in rank if manager ranking is only influenced by noise. The blue line plots the actual average ranking change for the period ended December 2021, as shown in the charts above. The red line plots the average ranking change result over time since December 2003. The results are fascinating.
Over 3 month periods, there is little evidence of anything but noise. The average ranking change (red line) is essentially the same as the grey line which represents a random ranking each period. The last quarter of 2021 was unusual in seeing relatively little movement in rankings.
Over 12 month periods, some evidence of bias emerges with the red average ranking change line being somewhat below the theoretical random ranking grey line. It turns out that the 12 month period ended December 2021 saw higher than average rotation in manager rankings.
Increase the time frame to 3 year rolling periods and the evidence of bias remains at roughly the same level as that seen in the 12 month data.
The final chart below shows how evidence of bias tends to strengthen as the timeframe under analysis increases. In this chart the grey line represents the expected ranking change over different time periods where there is no evidence of manager skill and the manager rankings are essentially random. The red line plots the actual average ranking change over different time periods using data from December 2003.
As the chart clearly shows the red line is below the grey line over all time periods longer than three months, providing evidence that manager rankings persist over time and are not random. As the analysis period is lengthened, so the evidence of persistence in manager ranking grows stronger.
Conclusions
Kahneman defined noise in human judgement as “undesirable variability in judgements of the same problem”. There is undoubtedly a great deal of noise when an investor is seeking to select a discretionary investment manager. But there is also evidence that it is worthwhile making the effort to discern where manager skill is being masked by noise.
Over shorter time periods, the noise created by sentiment swings in financial markets tends to drown out the effects of skill bias. But as the time frame lengthens out beyond three years it appears possible to identify discretionary investment managers that are, on average, exhibiting performance persistence, both positive and negative. Find a couple of managers who appear to exhibit skill at different times and the probability of outperforming the peer group increases further.
There is, however, another level of noise that needs to be better understood both by discretionary investment managers and their clients. That is the variability of returns within a manager. The more latitude that is granted to relationship managers by an investment firm to tailor a client’s portfolio, the more scope exists for undesirable variability. Relative performance can become a lottery – pick the right relationship manager and experience top quartile performance; select the wrong relationship manager and experience bottom quartile performance.
Jean-Michel Jarre, musician and pioneer of synthesised rock, said that the difference between noise and music is in what the musician does with the sounds. In other words the same inputs can make a cacophony or a symphony. The noise from financial markets is unlikely to go away but the evidence also points to the possibility of making sweet music!
Skilful discretionary investment managers exist. Skilful relationship managers exist. As Kahneman observed, the challenge is to design noise reduction techniques. Data gathering and organisation is a pre-requisite. Following a predetermined decision making process where human judgement is de-emphasised should deliver noise reduction.
Sign up for more insights here