As any cyclist knows, headwinds can make a ride challenging: every meter gained can be a battle and the effort has all your attention. Curiously, the tail wind on the return is appreciated initially, but is quickly forgotten and you assume a feeling of general well being, even that you are stronger and fitter than you thought. Arriving home, you might mention the dreadful headwinds that slowed your progress, but rarely the helpful tailwinds that brought you back with the pace of an Olympian.
And so it is in life: headwinds are bad luck, perhaps overcome with fortitude, but tailwinds easily morph into talent.
Distinguishing between talent and luck is difficult, especially give the natural human tendency to attribute success to talent rather than luck. However talent is defined, most would agree that the distribution of talent amongst individuals is congregated around an average, with some less talented and some obviously more talented. The distribution of talent in a population can be represented by the familiar bell curve, as shown below.
Since firms are collections of people we might expect some firms to be more talented than others and, if driven solely by talent, to have a distribution of success following the bell curve. Company CEOs and company reports are always keen to cite the talent of their employees as the most important factor in the success of their business, leading to the erroneous conclusion that all firms are above average in the talent ranking.
For publicly traded companies, their growth in stock price is an obvious measure of success. Consequently, a stock portfolio, over time should evolve so that the prices of the stocks are also distributed according to the bell curve, with the most talented firms exhibiting the greatest investment returns.
The observed compounded returns are different from what would be expected from the bell curve2. A typical portfolio of stocks will have a distribution of cumulative returns such that 20% of the stocks will contribute 80% of the portfolio gains, leaving a long tail of underperformers. This is commonly called a Pareto distribution3 and occurs across a range of economic and natural processes: the wealth distribution within a country, the size distribution of sand particles on the beach, and the probability of machine failure in a factory.
When observations don’t agree with a model, the implication is that the model is missing an important factor. Even if it is often downplayed, we know that luck can often be a factor in success – could it be sufficiently important to attribute most of the portfolio gains to only few stocks?
To explore the role of luck and talent in a controlled environment, Italian researchers recently created a very simple digital world populated by agents with different degrees of talent. As time progresses, the agents are exposed to a series of lucky or unlucky events. Agents are endowed with a store of success which is enhanced in proportion to their talent when they encounter good luck, and reduced by half when they encounter bad luck4. The simulation was run over a number of years and the researchers gathered data on which agents were the most successful and why. This artificial world is clearly and deliberately designed to focus on the interplay between luck and talent, without any other extraneous factors.
The simulation results had the following properties:
The model, although simple, demonstrated the common experience that great talent is not sufficient to guarantee success and that less talented individuals can be very successful if helped along with doses of good luck. Importantly, the researcher’s conclusions were robust under a variety of different starting conditions.
Having developed a model that seemed to replicate real world behaviour, the researchers were eager to use it to make predictions. Being academics, it is not surprising that they were interested in the most effective funding strategy when grant awarding bodies had to commit limited resources to competing grant submissions from researchers. They considered a range of scenarios: an equal distribution of funds amongst all researchers, an even distribution to a random subset of applicants, and a selection of the best (as measured by past performance). They also examined combinations of these strategies.
The conclusions were that rewarding researchers based on past performance was a poor strategy, with the results worsening as the selection was narrowed. Conversely, the best strategy was to allocate equal amounts across all applicants. The underlying dynamics were that providing even small resources to a wide pool of researchers and embracing the serendipitous nature of research, outweighed rewarding past success. There is an obvious resonance with stock selection where the evidence is clear that buying individual stocks or funds based on past success is a poor strategy, compared to an allocation across the entire market as a passive indexing suggests.
Telling successful people or the CEOs of successful firms that they, or their firm, might just be modestly talented but very lucky will always be a hard slog. To see what you might have to deal with, consider this segment (starts 13.30 minutes into the video) from Fox cable news interviewing Professor R.H. Frank5. We have a human bias towards believing our own efforts are responsible for our success, and circumstances outside our control were responsible for poor outcomes. It is a bias investors should avoid.
If you feel the need to test your own willingness to recognize the role of luck, try this simple experiment. Write a list of five stocks that you think will outperform over the next year. Write a second list of five that you think will underperform. Label each clearly, Winners and Losers, and put them in a drawer for a year. If you are skilful and stock selection is a game of skill (like chess), then the chances are that your predictions will be correct. If stock selection is dominated by luck (like Snakes and Ladders) then there is a good chance that your Losers will outperform the Winners. Good luck.
1 Born Under a Bad Sign, recorded by Albert King (1967). Lyrics by Booker T Jones & William Bell.
2 This is a complicated topic but recent research is summarised here.
3 A Pareto distribution is the best known example of the more general type of power law distributions, but we don’t dwell on this distinction. See here for further details.
4 This article provides a summary for the general reader of the original research paper.
5 Author of “Success and Luck: Good Fortune and the Myth of Meritocracy”, 2016
