I studied engineering in university. As someone who knew nothing about finance, accounting or economics, I never thought I would work in investments. I was an undergraduate student when the stock market crashed in 1987. But I wasn't very interested about it. I only knew that was an important headline news on BBC six o’clock news.
After I graduated, my first job was to write computer codes and spreadsheets on exchange traded options. I was hired purely because I seemed to know the complex maths. There were some similarities, especially option pricing models which used some of the theorems in communication engineering (mainly relating to noise). That was in the 1990s, when many physicists, engineers and mathematicians were hired by big investment banks to build financial derivatives trading models. But for mainstream investments, I still thought the skills required were from accountants, financial statement analysts and economists.
I started studying for the CFA level 1 exam when I returned to London in the mid-1990s. I was recommended to read Capital Ideas: The Improbable Origins of Modern Wall Street. The book introduced modern portfolio theory to me, with examples on how they were used in real life investments. It also introduced two important people to me: Peter Bernstein, the author of the book. And Harry Markowitz, a Nobel Economics Prize winner whose PhD thesis was (allegedly) thought to have nothing to do with economics by Milton Friedman.
I met a few Nobel Prize economists at CFA annual conferences, like Daniel Kahneman and Eugene Fama. But I never met Harry Markowitz. But for someone who had no training in finance and had just started working in investments in 1990s, the concept of portfolio construction using optimisation really appealed. I know the maths (it’s kind of borrowed from linear programming and operational research). And I realise that I can build a portfolio without understanding financial statements, valuations and economics: I simply need some statistical forecasts on each security – expected return, volatility, and correlation with other securities – to build a portfolio of a certain risk in which the expected return is maximised.
Some institutional clients of the company I worked for used Markowitz optimisation model for asset allocation. I was very pleased to be involved in the analytical work. I learnt how to use off-the-shelf software or built my own spreadsheets to run these optimisations. But I also learnt the weaknesses of Markowitz’s model. Your ‘optimal’ asset allocation will depend a lot on what you use as expected (or forecast) returns, volatility and correlation. Academic studies showed that historical volatility and correlation are quite reliable forecasts. But that is not the case for expected returns. After a decade of low interest rates globally, will historical returns in bonds be reliable enough for use in the optimiser? Will you have confidence in that asset allocation? So, Friedman might be partly right: investments still need input from economists (and also analysts who can understand a company’s financial statements).
Markowitz’s model also cannot handle financial instruments with asymmetric risk/return profiles. Options and some derivatives are examples. The volatility and correlation statistics do not capture these important features, where you can have limited downside risk and unlimited upside potential (or limited returns with full downside risk if you sell options). There are sophisticated theoretical models that can handle such complexity in an optimisation process, but I haven’t seen them being put into a working software or system yet.
The model proposed by Markowitz was an equilibrium model. The returns of each security and hence the portfolio may fluctuate over a mean (or average). That mean value will be reached over the long term when the effects of interactions of many variables finally settle down for an equilibrium. But how long is that: 5 years, 10 years, 30 years? Or like what John Maynard Keynes said, ‘in the long run we are all dead’? Here, economists can provide useful input based on economic or business cycle analysis.
Markowitz passed away a few weeks ago. Despite the above limitations, his model is still important. He quantified the importance of diversification. His work also contributed to the foundation of passive investing. There are improvements and variations of Markowitz’s model, including focusing on risk premium with a better understanding that there are other risk factors, not just volatility, that can drive returns. We wouldn’t have smart asset allocation models, factor investing and maybe even passive investing without Markowitz. And like many other people, his work inadvertently affected a young person who never thought he could work in the investment industry.
James Chu CFA
Head of Investment Solutions