When I was a congressional reporter, it was my job to cover the activities of the House of Representatives, especially what happened on the floor. When you do that, you quickly come to realize how important congressional staff are to the operation. Lawmakers are so scheduled, and spend so little time in actual hearings and debate, that they have to rely on the people they hire for a lot of their official work.

Which makes looking into how they hire those staffers an interesting line of inquiry.

As with most hiring processes, with Congress we mostly only see the outputs: who actually fills a role (made a bit easier by the House Office Expenditure Data I highlighted in a previous newsletter). But for years, the House of Representatives has been publishing a weekly listing of jobs with members and committees. You have to sign up to get those emails, but anyone can do that, so of course I did at least eight years ago.

One of the good things about long-running government processes is that the format usually doesn’t change that much. That’s true in this case: each week, the House Vacancy Announcement and Placement Service sends out an email with two PDF attachments: one listing jobs and the other listing internships. A recent jobs one looks like this:

A couple of years ago, I put out a call asking if anyone had more archives of these listings, trying to flesh out my collection. Luckily, I’m not the only one who gathers weird stuff, and as a result I’ve got more than 1,200 listings. So far, so good. But these listings are still PDFs.

That’s a solvable problem. The GitHub repository of these job and internship listings contains not just the original PDFs but text representations of them, extracted using one of my favorite command-line tools: pdftotext. Here’s the latest one, for example.

Many of these listings follow a somewhat consistent format, but there doesn’t appear to be a template that all of them are working off, so there’s enough variation to make parsing a bit of an issue. Until LLMs, that is. I started parsing them into JSON data structures using Google Gemini (example here) and have lately been using Qwen 3.5’s cloud model (same listing here), which I think has produced better results.

Some important caveats: these job and internship listings are not comprehensive. That is, many House offices do not use the official job listing service, preferring ones organized within party caucuses or informal networks. Any conclusions from the data need to acknowledge that fact. It also appears to have many more Democratic listings than Republican ones. Still, I think there are some good uses for this data.

Among them is a broad view of the kinds of skills that congressional offices are looking for in staff. Splitting the listings into two periods (2013-2018 and 2019-2026) I looked at changes between them in desired skills:

You won’t be surprised that digital and video production skills are more in demand now, but mentions of “attention to detail” - which should always be a critical thing for members of Congress - have increased pretty dramatically in job listings. Also, every skill seems more important these days. Good luck out there, congressional job-seekers. Pay attention.

Other People’s Data

Sure, I make a lot of my own data, but I’m constantly coming across other people’s efforts and wanted to highlight some of those as well.

• White House Aliens Data by Kate Sills. The White House produced a site with aggregate totals for immigration enforcement arrests that on May 27 had more than 400k of them listed. After a WIRED story reported that more than 700 of them were U.S. citizens, the administration “updated” the site, dropping more than 270k arrests from the interactive map. Sills has the before and after.

• Catholic Mass Times Scraper by Ben Ashkar. If you wanted to maintain a list of worship service times for more than 23k Catholic churches in the U.S., Ashkar has you covered with this codebase (not the data, alas, but you could reproduce it). Bonus: it also extracts weekly bulletins!