260
G. Marti et al.
• Reference [127] suggests that one may design a new set of Ricci curvature network
based-strategies in statistical arbitrage (e.g. for mean-reverting portfolios).
• Reference [103] finds the existence of significant relations between past changes
in the market correlation structure and future changes in the market volatility.
• In [68], authors suggest the use of clustering to build statistical classification of
quantitative trading alphas for which there is no analog of “fundamental” industry
classifications such as the GICS, BICS, ICB, NAICS, SIC, etc.
• References [27, 76] describe a behavioral bias: investors overly rely on the standard
industry classifications (e.g., SIC, NAICS). Corporate managers can exploit this
behavioral bias to steer their company towards a more favorable industry and
therefore benefit from a lower cost of capital. The article [27] does not discuss
this behavioral bias from the investor point of view, but it can pay to have a
different (statistical) view of the standard industry classification to avoid such
opportunistic window dressing. In [76], authors claim that long-short strategies
exploiting mispricing due to the industry categorization bias generate statistically
significant and economically sizable risk-adjusted excess returns.
• In [61], authors show that considering alternative industry classifications (for
example, text-based ones) can enhance the returns of well-known quantitative
strategies such as the industry momentum one which is driven, according to the
paper, by inattention to less visible horizon peers.
• The same authors find in [60] that their text-based classification of peers significantly outperform traditional industry peers in explaining firm valuations and peer
comovement in the stock market: Firms with peers whose products more closely
match theirs have higher stock-market comovement; Firms that have more unique
products relative to their peers have higher stock market valuations.
• In [153], the author investigates the supply chain network using several centrality
measures. He finds that “the stock performance of supplier central portfolios tends
to predict the movements of the overall stock market.” More generally, the author
suggests that “shocks to the supply chain can generate ripple effects, which can
show up potentially as lead-lag predictive relations in security returns, earning
surprises, default probabilities, and/or distinctive term structure patterns in realized
and option implied volatilities and credit default swaps.”
• Authors of [87] suggests using hierarchical clustering to estimate two parameters
required by their “False Strategy” theorem, namely the number K of effectively
independent tests, and the variance of the Sharpe ratios across the K effectively
independent tests. This number K corresponds to the number of clusters of strategies found considering a distance based on the correlation of their returns. Using
these estimates, they can conclude how likely a strategy is spurious.
• In [31], authors monitor the correlation network statistics of the constituents of a
stock index, and find that an abrupt change in the network is predictive of significant
falls in the price of this index. Reference [134] aims at predicting such abrupt
network changes.
G. Marti et al.
• Reference [127] suggests that one may design a new set of Ricci curvature network
based-strategies in statistical arbitrage (e.g. for mean-reverting portfolios).
• Reference [103] finds the existence of significant relations between past changes
in the market correlation structure and future changes in the market volatility.
• In [68], authors suggest the use of clustering to build statistical classification of
quantitative trading alphas for which there is no analog of “fundamental” industry
classifications such as the GICS, BICS, ICB, NAICS, SIC, etc.
• References [27, 76] describe a behavioral bias: investors overly rely on the standard
industry classifications (e.g., SIC, NAICS). Corporate managers can exploit this
behavioral bias to steer their company towards a more favorable industry and
therefore benefit from a lower cost of capital. The article [27] does not discuss
this behavioral bias from the investor point of view, but it can pay to have a
different (statistical) view of the standard industry classification to avoid such
opportunistic window dressing. In [76], authors claim that long-short strategies
exploiting mispricing due to the industry categorization bias generate statistically
significant and economically sizable risk-adjusted excess returns.
• In [61], authors show that considering alternative industry classifications (for
example, text-based ones) can enhance the returns of well-known quantitative
strategies such as the industry momentum one which is driven, according to the
paper, by inattention to less visible horizon peers.
• The same authors find in [60] that their text-based classification of peers significantly outperform traditional industry peers in explaining firm valuations and peer
comovement in the stock market: Firms with peers whose products more closely
match theirs have higher stock-market comovement; Firms that have more unique
products relative to their peers have higher stock market valuations.
• In [153], the author investigates the supply chain network using several centrality
measures. He finds that “the stock performance of supplier central portfolios tends
to predict the movements of the overall stock market.” More generally, the author
suggests that “shocks to the supply chain can generate ripple effects, which can
show up potentially as lead-lag predictive relations in security returns, earning
surprises, default probabilities, and/or distinctive term structure patterns in realized
and option implied volatilities and credit default swaps.”
• Authors of [87] suggests using hierarchical clustering to estimate two parameters
required by their “False Strategy” theorem, namely the number K of effectively
independent tests, and the variance of the Sharpe ratios across the K effectively
independent tests. This number K corresponds to the number of clusters of strategies found considering a distance based on the correlation of their returns. Using
these estimates, they can conclude how likely a strategy is spurious.
• In [31], authors monitor the correlation network statistics of the constituents of a
stock index, and find that an abrupt change in the network is predictive of significant
falls in the price of this index. Reference [134] aims at predicting such abrupt
network changes.