Lazy Network: A Word Embedding-Based Temporal Financial Network to Avoid Economic Shocks in Asset Pricing Models

Table 5

Portfolio returns comparison: monthly excess returns (returns minus risk-free rate), 3-factor alphas (market, size, value), and 5-factor alphas (market, size, value, momentum, and liquidity) for the long-only portfolios, “Lazy Prices” [2] and “Semantic Similarity Portfolio,” or “SSP” as referred to by [4]. To make comparisons easier, we also repeat the results from Table 3 for the proposed portfolios that use the eigenvector and closeness centrality. All these portfolios use similar rebalancing as well as similar datasets as input.

Portfolio	Equally weighted			Value-weighted
	CAPM alpha	3-factor alpha	5-factor alpha	CAPM alpha	3-factor alpha	5-factor alpha
Eigenvector t-stat	1.07 (4.57)	0.92 (4.35)	0.96 (4.69)	0.64 (2.62)	0.56 (2.30)	0.62 (2.52)
Closeness t-stat	0.84 (4.32)	0.70 (4.03)	0.72 (3.80)	0.94 (3.97)	0.90 (4.14)	0.92 (4.22)
Lazy Prices [2] t-stat	0.96 (3.05)	0.24 (2.76)	0.23 (2.70)	1.00 (3.00)	0.44 (2.78)	0.43 (2.81)
SSP [4] t-stat	0.87 (2.56)	0.53 (2.50)	0.60 (2.69)	0.87 (2.75)	0.64 (2.27)	0.48 (1.59)

The t-statistics are in parentheses and displayed under the estimates. , , The significance at the 1%, 5%, and 10% levels, respectively.