Econometrics Project — Impact of Sales Meetings on Saving Behavior

Overview

This repository contains an applied econometrics project based on Case 1, which studies the impact of a salesperson meeting on clients’ saving behavior.

The objective is to evaluate whether meeting a salesperson affects:

• total savings
• retirement savings

The project is divided into three parts, following the progression of the course:

1. Data exploration and research question
2. OLS / Probit analysis
3. Impact evaluation with Difference-in-Differences (DiD)

---

Project Context

This project uses a panel dataset of bank clients observed over multiple time periods.
A treatment (meeting with a salesperson) is introduced during the observation window, which makes the dataset suitable for a causal analysis framework.

The project includes:

• data exploration
• variable transformations
• regression analysis (OLS, Probit)
• causal impact evaluation (DiD, fixed effects)
• interpretation and discussion of results

---

Repository Structure

econometrics-saving-behavior-impact/
│
├── data/
│   └── bank_clients_panel.csv
│
├── scripts/
│   ├── 01_data_exploration.R
│   ├── 02_ols_probit_analysis.R
│   └── 03_did_impact_eval.R
│
├── reports/
│   ├── case_description.pdf
│   ├── econometrics_case1_part1.pdf
│   ├── econometrics_case1_part2.pdf
│   └── econometrics_case1_part3.pdf
│
├── feedback/
│   └── feedback_summary.md
│
├── .gitignore
└── README.md

---

Data

• File: data/bank_clients_panel.csv
• Panel dataset of clients observed over several periods
• Main variables include:
  • client identifier (id)
  • time period (time)
  • income
  • savings
  • retirement savings
  • treatment indicator (meeting)
  • demographic controls (e.g. age, gender)

The original case description is available in:

• reports/case_description.pdf

---

Methods Used

Part 1 — Data Exploration

• Research question and hypothesis formulation
• Initial descriptive analysis
• Variable inspection and first transformations

Part 2 — OLS / Probit Analysis

• Descriptive statistics by treatment/control groups
• OLS regressions
• Probit model (for treatment assignment / balance checks)
• Interpretation of coefficients and statistical significance

Part 3 — Causal Impact Evaluation

• Difference-in-Differences (DiD)
• Fixed effects panel regression
• Placebo test
• Heterogeneity analysis (e.g. by gender)

---

Main Learning Outcomes

This project helped me better understand:

• how to move from a research question to an econometric strategy
• how to work with panel data
• the difference between correlation and causal interpretation
• how to use OLS, Probit, and DiD in a coherent workflow
• how to interpret regression results in an economic context

It also helped me improve the quality of my work across the three submissions (structure, methodology, and interpretation).

---

Reproducibility

The analysis is written in R.

To run the scripts, place the dataset in the data/ folder and run the scripts in order:

source("scripts/01_data_exploration.R")
source("scripts/02_ols_probit_analysis.R")
source("scripts/03_did_impact_eval.R")

Note: some scripts may require package installation (e.g. dplyr, ggplot2, fixest, margins, etc.).

---

Notes on Feedback and Improvements

A summary of the feedback received during the project and the improvements made between parts is available in:

• feedback/feedback_summary.md

This helps document the progression and the methodological refinements made throughout the project.

---

Author

Ibrahim Fousfos

---

License

This repository is shared for academic and portfolio purposes.