What am I up to now?

April 6, 2024

December, 2023


  1. Updates
  2. Reading
  3. Music
  4. Previously


It doesn’t snow in Oxford. I mean, I’m sure it does sometimes, I’ve seen the postcards, but this is my third winter here and I don’t expect to see any flurries.

A high-pressure front arrived in Oxford the last week of November, clearing the skies and plunging the thermometers. To get below freezing in the Thames River Valley, you have to accept blue skies. This is a trade I’ll accept, after a dreary and soaking Michaelmas term. A few Oxfordians missed the rain and rushing Thames, especially the blue heron who fishes on Isis weir each morning, but the town feels more cheerful in the seven hours of hoary sunlight we get each day.

The immediate benefits came for sports — the river level dropped enough for rowing to resumeThe Thames/Isis was too fast for rowing for most of the academic term. This might have been the most visited site in Oxford this month. and fields dried enough for soccer and baseball. The Oxford baseball team is undefeated, putting us on track for the national championships in Slough in March.

Academically, Michaelmas is wrapping up, and it was as boring as I expected. I’ve registered for my exams, which doesn’t mean anything at all, it’s barely a commitment at all.

I’ll be back in the US mid-December through mid-January. Definitely reach out if you’d like to meet up!


Related to what I wrote about last month: my dad texted me last week, asking if I knew what was going on in Sierra Leone. No, I don’t, and any analysis will probably age badly.Reminding me of this old post on Afghanistan, which aged not all.

I’m skeptical of any insight I feel to have about this situation, and you should be too.To answer another question, I would be surprised if the Russians had anything to do with this. They wouldn’t mind the chaos, of course, but it’s out of Wagner’s way, for a pretty minimal reward. Timothy Garton Ash provides a good metric for this sort of analysis:

If you want to judge anything written by a foreigner about a country, you need to know when the writer first went there. Was it in the bad old days? Or perhaps for him they were the good old days? Was it before the revolution, war, coup, occupation, liberation or whatever the local caesura is? Of course the writer’s own previous background and current politics are important too. But so often the first encounter is formative. Emotionally and implicitly, if not intellectually and explicitly, it remains the standard by which all subsequent developments are judged.

I spent 13 months in Sierra Leone; my only interactions with the national government were at customs and army checkpoints. The Covid years, late 2020 into 2021, were when the national government flexed its state capacity for the first time since the reconstruction period. The lockdowns and curfews were enforced by the army; Bio’s motorcade reportedly doubled in length; vaccines were transported in armed convoys. None of this felt particularly authoritarian — Sierra Leone has put a lot of effort into having an effective military, it makes sense to rely on it. I didn’t observe an election while I was there, but both the major parties — Bio’s SLPP and the opposition APC — were everpresent on the streets of Freetown. We never really got closure on July coup plot; it was probably downstream of the presidential election this past summer — reelecting Bio. I, again, have no particular insight into this but my instinct is that even if the election wasn’t fair, it was probably free.

We have some other good elections coming up this year. The big one, of course, is the DRC general on 20 December. Last year, it looked like the opposition candidate — Martin Fayulu — had a chance to unseat President Tshisekedi, but the resurgent conflict in the Kivus makes me less optimistic.

Also exciting is the 3 December Venezuelan referendum over Guayana Esequiba, the western two-thirds of neighboring Guyana.Note this is not a secession referendum — it’s the people of Venezuela voting about whether or not they respect the old treaties defining their eastern border. If they vote the way the questions are leading them, then oops, I guess we’re going to war to regain our old territory and oil. This is a clever but somewhat blatant oil-grab by Nicolás Maduro. Venezuela’s military has been massing on their border with Guyana, and Brazil’s military is making noise too, so we might see another slight scuffle while the superpower(s) are busy elsewhere.

Complaining about papers

This is actually a re-hash of an old complaint, but it’s a new paper. David Chan, David Card, and Lowell Taylor (hence CCT) ask if there’s a VA advantage — and find persistent benefits.

Last time I talked about this identification strategy, I focused on how patient preferences for certain hospitals messes up the exogenous assignment of healthcare to patients. I still think this is an issue,But, as I wrote, an easily solvable one! Data on patient facility choice is available. but CCT deal specifically with VA/military hospitals — something else I have experience with.

When asked, I used to say that I grew up “outside of DC.” I’ve stopped doing this because a surprising amount of people knows where Bethesda is, even in Oxford. If they do, it’s often because of either NIH or Walter Reed (WRNMMC), the Naval medical center. Both of these facilities have their own emergency services. NIH has an ambulance and three fire trucks;For the nerds: a 2019 Pierce center pivot, a 2003 Pierce rear pivot, and a 2005 Pierce custom HazMat. Their ambulance is a sick custom Freightliner with a full bio-hazard containment unit. Walter Reed has a paramedic ambulance and a couple of engines. While they’re dedicated to the campuses they’re based on, these units can and do respond to 911 calls in Bethesda. My fire station was right next to both campuses, and our response areas overlapped.This often left us to respond to calls inside the campuses, if their ambulance was busy elsewhere. My very first call as an EMT was dispatched for NIH, for a scientist whose finger was bitten off by a monkey, offended at being used as a test subject. Since I left Bethesda, the NIH units have cut back on external calls, but the WRNMMC paramedic ambulance still sorties often.

Now, the WRNMMC ambulance isn’t the only military-affiliated ambulance in the country which services both civilian and non-civilian populations. Some VA hospitals have ambulances which respond to There’s heterogeneity even within civilian ambulance services. My station developed a reputation at one nursing home for veterans — Knollwood, in NW DC — for always taking our patients their destination of choice (often WRNMMC), relative to other nearby stations.This would often lead to absurd situations, like nurses telling 911 not to bother upon hearing which ambulance was dispatched.

What this comes down is, there are patient fundamentals which affect their hospital assignment, beyond and separate from the ambulance fundamentals. With this in mind, let’s see how CCT approach their identification:

We use the ambulance design proposed by Doyle et al. (2015) to study the causal effect of receiving emergency care at the VA versus a non-VA facility. Our approach compares veterans sharing key characteristics—zip code of residence, prior VA and non-VA health care utilization, and location of pickup (e.g., home residence, nursing home)—who receive the same dispatched level of ambulance service (i.e., advanced versus basic life support) from different ambulance companies. […] Specifically, we study veterans who arrive at a hospital via a 911-dispatched ambulance, comparing veterans from the same zip code who could have been transported by different ambulance companies with different propensities to transport patients to a VA hospital.

Honestly, I think that the CCT paper is “right”, in the important way; I even believe the Doyle strategy is useful in an important way. But this data-generating process is messier than the authors give it credit for. I wrote in my original blog post:

Joseph Doyle, at MIT, wrote the first paper to use the ambulance identification strategy I described above — if he wanted a ridealong with Boston EMS, we could have hooked him up! If Doyle had heard a grandmother arguing that she needs to go to Tufts and not BMC because her cardiologist is at Tufts, and if he had seen the look on the face of the paramedic who hadn’t slept since Wednesday, his identification strategy would have been strengthened. The paper would have survived learning about what it means for an ambulance to take a patient to a hospital.

I’ll stand by that. To any academic working with ambulance data,Or more realistically, their RA. email me and I’ll find you an ambulance house near you to go see what it’s like. Just buy the crew pizza at the end of the day.

Another mini-complaint, not about a paper, but about the world, highlighted by a paper. Here’s an AER from last year I recently discussed with CM; a free version. I remember seeing an older version of this, and dismissing it, but CM talked me into reading it last week, and I’m upset that the world and CNNs are weirder than I thought. Night lights have been on their way out for a while;LH shared a good summary here. This paper, Khachiyan et al., isn’t a dataset which you can download and start working with tomorrow, but it shows that space-based economics still works.

The authors use deep learning on multi-spectralThis is a random sidenote Landsat imagery to estimateThey use the word predict. Another thing I’m angry about. household income at the (roughly) square mile level. They throw these images into a CNN and achieve ridiculously accurate levels and deltas for household income. This paper is a reminder of two things I particularly don’t want to be reminded of.

First, we’re a lot less subtle than we think we are. As a species, we’re obvious and oblivious about our effect on the planet. Landsat is in low-Earth orbit — meaning a few hundred miles up — and takes low-resolution imagery, about 30 meters, depending on the spectral band. This is good enough to distinguish a football field from a baseball field, but not much better. And yet — there’s enough data in these blobs of ultracolor that two square miles of America with a 10% difference in household income are obviously different from each other.

Second, with all of this information in front of us, we don’t know what it means. If you showed me hundreds of low-resolution squares of America, and asked me to rate their wealth, I could come up with some heuristics. More golf courses, richer; industrial parks, poorer. Just looking, I could probably distinguish Brockland, MA from Wayland, MA. But that’s cross-distribution. I expect I would have no chance distinguishing a 40th percentile square mile from a 60th percentile square mile. It’s a stark and unintuitive finding that the massive yet granular imprint of our wealth is so plainly visible. I’m annoyed that I don’t understand it better.


I’m usually pretty strict about only writing about books I’ve read here, but I’ll break that rule here for Zach and Kelly Weinersmith’s new book, A City on Mars.

A couple of years ago, my friend AC asked a great question, one I recommend asking to any reader in your life: if you could be the literary executor to any living writer, whom would you choose? I surprised myself with my answer — Zach Weinersmith. But I love it all. I love his weird political manifesto. I love his poetry. I love his children’s books; his version of Beowulf is better than any except maybe Headley’s, if you’re looking foor fun.I also just read Weinersmith’s Augie and the Green Knight aloud and it was great! Weird and rough and beautiful. His last name is way cooler than Will Macaskill’s. Out of all of Bryan Caplan’s opinions, Weinersmith found the correct one, and illustrated it beautifully.

And we haven’t even gotten to the comics yet. Weinersmith’s SMBC is xkcd with higher variance in quality and more sex.
Yesterday 2023-11-29’s xkcd made me angry. I’ve read most of the archives of xkcdsucks.blogspot.com and think that guy has lost his head but are you seriously going to look at me and hit publish? I’m sorry Randall we both know you’re better than this. It’s not that it’s not funny — it is — it’s just a joke you’ve made before and actually no, it’s not funny; it was funny fifteen years ago, when all those t-shirts with triangles and here it is ok deep breath.
More importantly, you get a much better sense of Weinersmith through his comics than you get of Munroe through his.

All of this adds up to: if this man and his wifeWho has discovered a cool wasp! write a book about space economics which turns out to confirm all of my views precisely — I’m going to read it.

Another book I haven’t read yet is my good friend Savannah Mandel’s upcoming Ground Control. As these two end up in the same spot (“let’s not go to space”) for *very *different reasons, I think a joint book review could be a lot of fun. Ask me about that if you haven’t heard anything in six months.



November, 2023

October, 2023

September, 2023

August, 2023

July, 2023

June, 2023

May, 2023

April, 2023

March, 2023

February, 2023

December, 2022

November, 2022

October, 2022

September, 2022

August, 2022

November, 2021

October, 2021

September, 2021

July, 2021

June, 2021

May, 2021

What am I up to now? - April 6, 2024 - Joseph Levine