Sebastien Lleo is Associate Professor of Finance at NEOMA Business School (France)


Analyst views and expert opinions matter. They are an invaluable complement to market data when it comes to formulating relevant capital market expectations and to strengthening risk management models and practices. But watch out for behavioral biases!

“Garbage in – garbage out!” Every investment management professional has heard the warning that poorly formulated capital market expectations will get portfolio optimisers to produce inefficient, unrealistic, and even outright dangerous portfolios.

Thus, considerable efforts have taken place to turn available economic and market data into accurate capital market expectations. These lead to the development of slick statistical methods, effective econometric techniques, and powerful machine learning algorithms.

Opinions can also be an invaluable source of insights to construct accurate capital market expectations.

What are the types of opinions on financial markets?

Opinions take multiple forms in financial markets. They include analyst views, opinions from political and economic experts, super forecaster predictions, and investor polls.

Moreover, opinions abound on financial markets. Consultancy Quinlan & Associates reported that the bigger banks and brokerages emailed over 40,000 pieces of research every week in 2016, despite continuing job cuts in the financial sector. Social media also contribute to the spread of opinions: according to the financial website, there are at least 839 active financial blogs published in English.

Why should I use expert opinions?

Opinions have three key benefits.

First, opinions can be a crucial complement to traditional economic, corporate and financial market data to construct realistic capital market expectation, and keep those up-to-date. This statement is especially true in times of heightened uncertainty, such as market bubbles and financial crises, when traditional data fail to provide an accurate assessment of market conditions.

Second, opinions can strengthen risk management models and practices. Opinions can widen the range of scenarios considered in portfolio optimisation and risk management. Dissenting opinions provide a cornerstone for the construction of meaningful stress test scenarios.

Third, we can use opinions, even when traditional data are not. For example, assessors evaluate insurance claims, and appraisers estimate the value of illiquid assets, such as real estate and collectables, periodically.

How easy is it to collect opinions?

The inclusion of opinions requires extreme care.

Let’s look at analyst views and expert opinions. We all know that not all experts or forecasters are equally accurate. A widely reported study by CXO Advisory Group LLC tracked 6,582 forecasts for the U.S. stock market published by 68 experts between 2005 and 2012. The study found that average accuracy across experts was 47.4%, with individual accuracies ranging from a low of 21% to a high of 68%.

Therefore, investment management teams need to implement a process to guarantee the relevance of the opinions used in their models. This process, known as “elicitation,” is described in abundant literature. The books by O’Haghan (2006) and by Meyer and Booker (2001) are an excellent place to start. Essentially, the elicitation process helps to construct views that are specific, explicit, and structured. Opinions need to focus on a specific variable or parameter, such as the price of a given asset or the mean of a distribution. Opinions need to explicitly provide a mid-point or most-likely scenario, a confidence interval, and to relate the confidence interval to a probability distribution. Finally, opinions need to be structured to provide a transparent and auditable trail.

What are the implementation challenges?

Three main implementation challenges need addressing.

The first and most dangerous challenge is that opinions are often subject to the behavioral biases. Behavioral biases, in particular overconfidence, excessive optimism, conservatism, confirmation bias, and groupthink play an essential role in how finance professionals perceive and process information, and on how they form their forecasts. Recently, in a simulation study, Davis and Lleo (2020) recently found that the presence of biases explained nearly 70% of excess risk-taking. Therefore, it is crucial to debias forecasts before using them in any model.

Second, expert opinion models are Bayesian and therefore require the specification of a prior distribution. We can overcome this difficulty with some original thinking, as with Black and Litterman’ reverse optimisation exemplifies.

Third, aggregating of multiple expert opinions is considered an essential conceptual and computational problem because it requires engineering a joint distribution out of a collection of univariate distributions.


How can I integrate opinions in my portfolio selection model?

Currently, several families of portfolio selection models use opinions as input. The best-known and oldest is the Black and Litterman (1992) model, which uses analyst views to generate capital market expectations in a Markowitz-style single-period optimisation framework. This approach has been extensively discussed and developed in a large number of subsequent papers and chapters.

However, the Black-Litterman approach has two fundamental limitations. First, it is static, meaning that it locks portfolio managers into a “buy-and-hold” strategy, ignoring the possibility that portfolio managers may shift their asset allocation as financial market conditions change. Second, it ignores the presence of behavioral biases in expert opinions.

To address the first limitation, Frey et al. (2012) and Davis and Lleo (2013,2020) proposed two closely-related dynamic portfolio management models. Although both models are developed in continuous time, we can transpose them to a multiperiod discrete-time setting.

The second limitation has proved more elusive. At the moment, Davis and Lleo (2020) is the only dynamic portfolio selection model that addresses for behavioral biases.



Black, F., Litterman, R., 1992. Global portfolio optimisation. Financial Analyst Journal 48 (5), 28–43. Davis, M., Lleo, S., 2013. Black-Litterman in continuous time: the case for filtering. Quantitative Finance Letters. 1 (1), 30–35.

Davis, M., Lleo, S., 2020, Debiased expert forecasts in continuous-time asset allocation. Journal of Banking and Finance. 113.

Frey, R., Gabih, A., Wunderlich, R., 2012. Portfolio optimisation under partial information with expert opinions. International Journal of Theoretical and Applied Finance 15 (1). O’Hagan, A., 2006. Uncertain Judgments: Eliciting Expert’s Probabilities. Wiley.

Meyer, M., Booker, J., 2001. Eliciting and analysing expert judgment: a practical guide. ASA-SIAM Series on Statistics and Applied Probability. Society for Industrial and Applied Mathematics.

Explore more