Assignment 6: From Data to Report

Overview

In this assignment, you will use Claude Code to download a dataset, explore it, and produce a focused, economics-quality PDF report. The goal is to practice the full pipeline: data acquisition, exploration, visualization, regression analysis, and report writing – all through the CLI.

Task

Step 1: Get the data

Work with data from week 06 class.

Step 2: The vibe report

Ask Claude Code to “create a nice looking report” with no further guidance. Save the output – you’ll reflect on it later. If you created in class, you can use that.

Step 3: The focused report

Using Claude Code, create a PDF report that answers a specific research question about earnings in your chosen state. Your report must satisfy these constraints:

  • 3 to 5 exhibits (graphs or tables) – no more
  • 600 to 800 words of text
  • At least one regression with appropriate controls
  • At least one well-crafted graph that went through multiple iterations
  • Professional formatting: title, author, date, numbered exhibits with captions
  • All analysis in Python

Good research questions to consider:

  • Is there a gender gap in earnings, and how does it vary by education?
  • What is the return to education, controlling for occupation and age?
  • How do earnings profiles differ across occupations?

Deliverables

Submit:

  1. The PDF report (the focused version) [15p]
  2. A short reflection (100-200 words) comparing the vibe report to the focused report. What was different? What did you have to specify to get quality output? How many iterations did the key graph take? [5p]

Tips

  • Start by downloading the data and creating a variable dictionary – this helps Claude Code understand what it’s working with
  • Don’t accept the first graph. Iterate: add CI, fix labels, adjust colors, add footnotes
  • Be explicit about constraints: word count, number of exhibits, formatting
  • The reflection is about your experience directing the AI, not about the data