1. In Modern Portfolio Theory - there is a lot of valuable information about factors and style funds where they position within the mean variance framework. I feel that it is silent for stock picking fund managers. How do we try to make sense or articulate our thinking or understanding of a fund manager who has consistently beaten the benchmark through time by DCF valuations analysis?
Asset pricing isn't directly covered in my LFS courses. In my Factor Modelling for Investment Management course, I discuss methods to identify priced factors in markets and discuss research that attempts to correlate identified factors with DCF methods. I also highlight how factor contribution and attribution can aid in understanding ex post returns and risks (i.e. understanding the performance of your persistently outperforming manager). In the portfolio construction course, I explain how active stock managers can build efficient active portfolios. However, I treat their valuation analysis as a given input.
2. A lot of sub-modules discussed here. Is there sufficient time in the course to deal with each of these adequately?
Especially in Asset Allocation and Portfolio Construction course, where there are numerous open-ended questions, time can be limited. I address this by custom-tailoring the presentation to the audience's priorities. Having presented on these topics for many years and given my practical involvement, I am very flexible. I can switch from a more top-down presentation focusing on the larger picture to more specific implementation details. All the concepts in my courses come with spreadsheet examples, enabling participants to engage with their own data and generate further questions.
3. What is the current trend in asset allocation in the western market, i.e. in the developed market?
Currently, the dominant trend, as I perceive it, revolves around addressing the myriad of macroeconomic uncertainties, such as changing rate environment, sectorial shocks (energy prices) and similar. Methodologically, there's a rising demand for scenario-based analysis. This involves crafting relevant macroeconomic scenarios, comprehending asset risk and return characteristics in these scenarios, and pinpointing allocations that enable successful navigation through these scenarios.
4. MPT: Nicolas Taleb has eviscerated MPT because it’s based on Gaussian statistics. Okay, got it. But I’ve argued with Dr. Taleb, ‘’Okay, we see the failures of MPT, but what do you have as a substitute?’’ Question: Is a substitute for MPT truly available in any practical form?
The assumption of a normal distribution for asset returns in MPT isn't critical. For those concerned about downside risk, there are multiple portfolio construction approaches which work well for non-normal asset returns. I'm unaware of an alternative that provides as comprehensive a framework as MPT. As I emphasized in my presentation, the need isn't to replace MPT but to thoroughly understand its core components. MPT continues to evolve, and it's widely acknowledged that its basic textbook iterations aren't immediately applicable in practical settings. Through my courses, I delve into various techniques available to practitioners, transforming MPT from abstract theory to a valuable tool in practical investment decision-making.
5. Do you think that factor investing is something commonly used by the industry nowadays or is it left aside because of the machine learning (ML) theories?
Factor models are more available in both quantity and quality than ever before. They serve various roles, from ex post descriptive purposes in reporting and analysis to ex ante applications in portfolio construction. Some of which I discuss in my Asset Allocation and Portfolio Construction and Factor Modelling for Investment Management courses.
I do not see ML as a substitute for factors, but ML has the potential to maybe uncover more and “better” factors from ever more and better data. ML complements existing techniques that are being used in empirical financial market research already for decades. Personally, I do not see much potential for major breakthroughs in empirical factor research. On the contrary, I see concerns that data mining has already reached unhealthy levels a while ago in empirical factor research (see “factor zoo”). My prediction is that ML/Data Science will affect other areas of the overall investment process more than asset allocation and portfolio construction, and probably also factor identification. I think incremental advances are more realistic than major departures with factors.
6. Correlations of bonds and shares in recent years have started to turn positive. When constructing strategic asset allocation, how would you address this relationship in an optimiser for asset class returns?
The bond and equity correlation is highly scenario/regime dependent, this has been known for decades and I have explicitly pointed it out in my courses for many years. In Asset Allocation and Portfolio Construction course, we need forward-looking correlations: therefore regime-based approaches to forecasting correlation is my answer to your question. In case confidence intervals of the forecasts get too large (correlations are tricky), robust methods are available which suggest working with worst-case assumptions. Especially for SAA purposes, avoiding overly optimistic assumptions about future diversification potential is important, in order not to avoid unrealistic expectations of clients and other stakeholders.