Quantitative Profiles: Occam’s razor for quants: get rid of 30% of your factors

研报英文原文证据摘录

Quantitative Profiles: Occam’s razor for quants: get rid of 30% of your factors

Exhibit 2: Momentum Leadership Becomes More Diversified

3-month Momentum sector composition change as of 21 July vs 30 June

8%

4%

0%

-4%

-8%

-12%

-16%

Source: BofA US Equity & US Quant Strategy. FactSet

BofA GLOBAL RESEARCH

Occam’s razor for quants: 31% of factors are duplicative

In our latest primer, we found the average number of signals used by quants is around

22. But in our analysis of variance inflation (a measure of how much of a factor’s

variation can be explained by other factors, or VIF) and alpha tradeoffs, we found that

using just a slim percentage of the factors we track could deliver superior results.

Exhibit 3: Nearly one-third of factors Exhibit meaningful redundancy (VIF > 5)

VIF scores withing the group (higher VIF score indicate higher factor redundancy in the group)

Group Factor VIF

Corporate Cash Deployment Dividend Yield 1.0

Corporate Cash Deployment Dividend Growth 1.1

Corporate Cash Deployment Share Repurchase 1.1

Growth Proj. 5yr EPS Growth 1.3

Growth Earning Momentum 1.5

Growth EPS Estimate Revisions 1.6

Miscellaneous Neglect - Institutional Ownership 1.0

Miscellaneous Neglect - Analyst Coverage 1.1

Miscellaneous Short Interest 1.1

Miscellaneous Size 1.1

Quality 5y ROE 3.3

Quality 1yr ROE 4.2

Quality 5yr ROE Adj 5.0

Quality ROA 7.8

Quality 1yr ROE Adj 12.3

Quality ROC 14.5

Risk Variability of Earnings 2.0

Risk Beta 2.3

Risk Low Price 2.9

Risk Estimate Dispersion 3.1

Technical/Momentum Most Active 1.3

Technical/Momentum Price Return - 12-mth and 1-mth Reversal 3.8

Technical/Momentum Price Return - 3-mth Perf. 4.1

Technical/Momentum Price Return - 12-mth and 1-mth Performance 8.0

Technical/Momentum Relative Strength - 30wk/75wk MA 9.6