mostly-true

Code to accompany "Most claimed statistical findings in cross-sectional return predictability are likely true" (Journal of Finance: Insights and Perspectives). Working paper version: https://arxiv.org/abs/2206.15365.

This code omits the bootstrap simulation "proof" that standard FDR estimates work in cross-sectional asset pricing. For that evidence, see the long-version branch of the code and the previous draft of the paper.

The code downloads all the data you need, including the Yan-Zheng (2017, RFS) data-mined strategies. Special thanks to Sterling Yan and Lingling Zheng for sharing their data.

Repo Contents

Excel Table

• ez-table.xlsx: used to create Table 1: Easy Bounds on the FDR
  • The bounds are easy, after all. No need for real code.
  • The tex input is created manually using excel2latex plugin.

R Scripts

Main Workflow

• main.r: Runs all scripts in order.
• config-and-functions.r: Provides project configuration utilities and shared helper functions

Data Preparation

• 01-prep-data.r: Downloads and processes predictor return data from Google Drive and Dropbox; creates cleaned datasets for Chen et al., CLZ, and Yan-Zheng predictors; also runs the CLZ bootstraps
  • Creates: data/emp_data.Rdata (required by most other scripts)
  • Creates: data/bootact.Rdata (bootstrap samples for CLZ returns)

Make Exhibits

• 02-easy-bounds.r: Makes Figure 1: Conservative Extrapolation from Published t-stats
  • Requires: data/emp_data.Rdata
  • Output name: hlz-intuition.pdf
• 03-visual-bounds.r: Makes Figure 2: A Visual Bound on the FDR
  • Requires: data/emp_data.Rdata and data/bootact.Rdata
  • Output names: dm-viz-storey-err.pdf and dm-viz-ez-err.pdf
• 04-factor-adjustments-yz.r: Makes Table 2: FDR Estimates Controlling for Value Weighting and Factor Adjustments
  • Requires: data/emp_data.Rdata
  • Output name: yz-fdr.tex
• 04alt-factor-adjustments-clz.r: Makes an alternative version of Table 2, but is not used in paper.
  • Requires: data/emp_data.Rdata
  • Output name: vw-ffn-visual-raw.tex
• 05-hlz-simpler.r: Makes Figure 3: Two Interpretations of Harvey, Liu, and Zhu's (2016) SMM Estimates
  • Requires: None (fully independent, generates own simulations)
  • Output names: hlz-simp-standard.pdf and hlz-simp-post-truth.pdf

Environment

• R version 4.5.0 (2025-04-11) — "How About a Twenty-Six"
• Package versions used:
  • here 1.0.1
  • tidyverse 2.0.0
  • data.table 1.17.8
  • ggplot2 4.0.0
  • extrafont 0.19
  • latex2exp 0.9.6
  • kableExtra 1.4.0
  • googledrive 2.1.2
  • haven 2.5.5
  • foreach 1.5.2
  • doParallel 1.0.17
• Run on MacBook M4 Pro
  • 24 GB RAM
• Run time is very roughly 5 minutes