Predictive Regressions: A Present-Value Approach

Predictive Regressions: A Present-Value Approach#

Last updated: Mar 16, 2026, 3:43:43 PM

Table of Contents#

Pipeline Charts 📈

Pipeline Dataframes 📊

Pipeline Specs#

Pipeline Name

Predictive Regressions: A Present-Value Approach

Pipeline ID

PR

Lead Pipeline Developer

Moxiao Li & Yilun Cai

Contributors

Moxiao Li & Yilun Cai

Git Repo URL

Pipeline Web Page

Pipeline Web Page

Date of Last Code Update

2026-03-16 13:22:51

OS Compatibility

Linked Dataframes

PR:replication_data

About this project#

This repository contains our final replication project for FINM 32900: Full-Stack Finance at the University of Chicago.

We replicate key empirical results from:

van Binsbergen, Jules H., and Ralph S. J. Koijen.
Predictive Regressions: A Present-Value Approach.
Journal of Finance (2010).

The project focuses on replicating the dividend growth predictability results in the paper and comparing the replicated statistics with the values reported in the original publication.

Using an end-to-end reproducible analytical pipeline, we reproduce:

  • Summary statistics of dividend growth

  • Predictive regression specifications

  • Key tables used to validate the replication

We also extend the data sample through 2024 to examine how the results change in more recent periods.

The project is implemented using a fully reproducible pipeline with:

  • Python

  • PyDoit task automation

  • Jupyter notebooks

  • LaTeX report generation

Replication objects#

Our replication produces the following outputs:

  • Table 1 – Summary statistics of dividend growth

  • Specification tables for predictive regressions

  • Side-by-side comparisons between replicated and paper results

All tables are generated automatically from the pipeline and compiled into a LaTeX report.

Pipeline overview#

The pipeline executes the following steps:

  1. Pull raw data from CRSP

  2. Clean and organize the data into tidy datasets

  3. Construct dividend growth series

  4. Compute summary statistics and regression inputs

  5. Generate tables and figures

  6. Convert notebooks into LaTeX

  7. Compile the final PDF report

The entire pipeline can be executed automatically using PyDoit.

Quick start#

1. Install LaTeX#

You must install a LaTeX distribution such as TeX Live or MacTeX.

Mac: https://tug.org/mactex/

Windows: https://tug.org/texlive/windows.html

2. Create a Python environment#

python -m venv .venv
source .venv/bin/activate
Contents