Definition of cluster analysis
What is cluster analysis?
Cluster analysis is a technique used to group together sets of objects that share similar characteristics. It’s common in statistics. Investors will use cluster analysis to develop a cluster trading approach that will help them build a diversified portfolio. Stocks with high correlations in returns fall into one basket, slightly less correlated ones into another, and so on, until each stock is placed in a category.
If done correctly, the different clusters will exhibit minimal correlation to each other. In this way, investors benefit from all the virtues of diversification: reduction of losses on the downside, preservation of capital and the possibility of carrying out riskier transactions without increasing the total risk. Diversification remains one of the pillars of investment and cluster analysis is only one channel to achieve it.
Key points to remember
- Cluster analysis helps investors develop a cluster trading approach that builds a diverse portfolio of investments.
- Cluster analysis allows investors to buy and bundle assets with associated returns that correspond to different market segments.
- One of the benefits of cluster analysis is that it helps protect the investor’s portfolio against systemic risks that could make the portfolio vulnerable to losses.
- A criticism of cluster analysis is that clusters with a high correlation of returns sometimes share similar risk factors, meaning that poor performance in one cluster could translate into poor performance in another.
Understanding cluster analysis
Cluster analysis allows investors to eliminate overlap in their portfolio by identifying securities with related returns. For example, a portfolio made up entirely of tech stocks may appear safe and diversified on the surface, but when an event like the Dotcom bubble occurs, the entire portfolio is vulnerable to large losses. Buying and pooling assets that correspond to different market segments is essential to increase diversification and protect against such systemic risks.
Stock selection and trading based on cluster analysis
The technique can also uncover certain categories of stocks like cyclical and growth stocks. These specific strategies fall under smart beta or factor investing. They attempt to capture better risk-adjusted returns from specific risk premia such as minimum volatility, growth and momentum.
In a way, smart beta or factor investing embodies the concepts of grouping and categorization advocated by cluster analysis. The logic of grouping on a single common behavior mirrors the basic methodology behind factor investing, which identifies stocks that may present similar systemic risks and share similar characteristics.
It is not always the case that the assets of a cluster live in the same industry. Often, clusters own stocks from multiple industries such as technology and finance.
Critique of cluster analysis
One obvious disadvantage of cluster analysis is the level of overlap between clusters. Clusters that are close in distance, which means a strong correlation of returns, often share similar risk factors. So a day of downtime in one cluster could result in equally poor performance in another cluster. For this reason, investors must find and bundle stocks with a large distance between them. In this way, the clusters are impacted by different market factors.
That said, large market downturns like the 2008 recession will stifle the entire portfolio, no matter how built. Even the most diverse clusters would find it difficult to weather the headwinds of the recession. Here, the best clustering can do is minimize extreme losses.