Humana introduced popular Walmart-branded plans. Younger spouses were more likely to pick them if their older spouse joined Medicare when the plan was available.
We study this using a regression-discontinuity design that exploits birth dates just before/after the new plans entered the market.

Over 70% of couples choose the exact same Part D plan. Even among couples whose drug needs are very different, the majority still pick the same plan. We create “placebo spouses” based on observables: <20% overlap.

That's Stata, here's a start in R.

Step 3: eliminated unnecessary ink.

Step 2: all text horizontal. Horizontal text is easier to read than vertical text.

Step 1: labels not legends. This makes the graph much easier to read.

To start, here's a graph with the default Stata options.

ChatGPT can help you deal with tables in LaTeX

Creating slides for class and I broke ChatGPT