Which Factor is Truly a Best Idea?
Novus tracks the most popular stocks among hedge funds along four dimensions. Let's discuss the top performers YTD: Consensus and Conviction.
Clients often ask us which of our 4C Indices is the clearest reflection of hedge fund best ideas. Is it Conviction, a measure of capital allocation, or Consensus, a measure of agreement? One may demonstrate value at risk while the other reflects the wisdom of the crowds. Let’s walk through these two indices and unpack how each provides a lens into the value-add behind hedge fund stock picking.
At Novus, we believe that portfolio construction and its careful measurement are key ingredients in generating alpha. This is especially the case today where alpha is harder to come by. Conviction is a measure of belief through portfolio construction rather than sheer stock picking. Case in point: an equally weighted portfolio of 100 securities demonstrates no divergence in conviction. A portfolio of 20 of those names varying from 1,000 basis points to 50 basis points is entirely different. Conviction is a means of capturing the amount of capital one allocates to securities in a portfolio rather than the simple choices at play. The Novus Conviction Index is a distillation of the twenty highest conviction holdings across our Novus Hedge Fund Universe.
Consensus measures the movement of crowds, an increasingly important phenomenon given the large growth in hedge fund assets over the last decade. Hedge funds are characterized as notoriously clubby. This may stem from their common training, lineage, or socialization. This is especially the case nowadays. While a pack of hedge funds centered on a single idea may have been a contrarian indicator 20 years ago, today it is indicative of a normative view in capital markets.
So how do the two compare nuts and bolts? First, let’s start with a measurement of security-level overlap, a foundational way of understanding the similarity of two portfolios. As we can see from this overlap graphic, there is significant cross-ownership across the two as of the most recent data period. 60% of constituents are common across the two indices. Perhaps it is the case that the most convicted securities across hedge fund managers are often the most normative:
Indeed, the returns over the past fifteen years for these portfolios are directionally similar, with an average correlation of .90. Since the lead-up to the financial crisis, that rolling correlation has indeed been elevated!
But correlation is hardly the whole story. We can clearly see this over a time-series of performance. Conviction demonstrably outperforms Consensus, which anecdotally underperforms the market. This is a huge difference for two indices with elevated overlap and correlated returns.
Why is this the case? It could be because Conviction is a clearer proxy for true best ideas than Consensus, which is increasingly a proxy for the median. While that may be, one might also ask why does Consensus underperform the broader market? It could be that the safest ideas for fund managers to crowd into are riskier than they assume.
One way to think about that risk is through the factor-betas of these indices. When we decompose both indices through the Novus risk-factor model, we see a few stark differences. First, the Conviction Index has more sensitivity to equity market dispersion. Dispersion is a measurement of how the returns of all 3,000 of the Russell’s securities are different. There must be divergence in security-level performance in order for there to be outperformance, and Conviction has a stronger correlation to those dispersion environments. Second, there is less relative value and momentum beta. Consensus is more heavily negative those two factors, and so less of the Conviction exposure is explained by these traditional factors. Finally, the alpha term of the residual is larger. While both models fit well to these indices (r^2 > .8-.9), we see that Conviction is an exposure that is less explained by traditional factors. This is perhaps owing to catalysts for value creation that cannot be patterned through historical constituent-level returns, and so conviction holdings may be less susceptible to quants and factor investors eating their lunch.
In short, while Conviction and Consensus both feel similar and are somewhat correlated with high levels of overlap, their behaviors diverge over long periods of time. Conviction expresses a higher amplitude of stock picking alpha over time and Consensus a reversion to the mean. Do you know your portfolio’s exposure to each?