What's for dinner?
I demand a bailout for Chinese restaurants
I’m entering my 8th week of lockdown/quarantine/pause, and it’s been pretty surreal. When you don’t leave your house for days at a time, you don’t realize everyone else is also locked up. I will try to keep this post brief because there is too much content to consume out there and “newletters are now just writers self-therapizing,” to paraphrase a friend.
We are all coping in our own ways. For a while, I was running and biking, slowly growing a beard (the male equivalent of cutting bangs). Then I crashed my bike and the bike shop was closed, so I started eating carbs again. I guess I’m also writing again.
The TLDR today: while Americans have significantly increased their purchases for food delivery, and nearly all cuisines have seen massive growth in delivery popularity, demand for Chinese delivery does not show statistically significant growth. Also, we are all cooking our own food a lot more, let’s see how long that sticks for.
I was curious about what people are eating, or at least searching for, during this time. Nowadays, just loading the dishwasher feels like a battle, let alone hand cleaning the pots and pans. I suspect that lots of people have been eating like shit: pizza, fried chicken, ice cream, and hot pockets are easy and delicious but, um, not very balanced.
One of the most interesting datasets early on in the pandemic was the OpenTable "State of the Industry” data, which showed global restaurant reservation volume dropping by region. Which, you know, lines up with the global pandemic and shelter in place orders from (roughly) the middle of March 2020. Now, however, that data has no entropy or information value (the technical term is “boring AF”):
People are still searching Google, and we can use Trends data to estimate, in a very hand-wavy way, what people are looking for.
Unsurprisingly, a lot of people just don’t know how to cook.
Note that the red lines refer to roughly March 8th, which lines up with when OpenTable noted the first -10% YoY changes. Shelter in place orders would mostly follow a week later, but consumer demand had already dropped before official orders and will likely lag official reopening orders. My last meal out before the lockdown was at the House of Prime Rib, and I would do just about anything for another King Arthur Cut (and that salad!), but “sit in a densely populated serving room with 200 other strangers” is not really one of those things.
But we still gotta eat. So what are we ordering?
Pizza is a good illustration of the data and methods we have available.
We can see that pizza delivery search volume is fairly seasonal (New Years is the biggest day of the year), and that overall volume has increased over the last few years. We can observe that search volume dropped right before the first week of March and spiked after, but can we quantify that effect?
Using the CausalImpact library (and closing our eyes a bit), we can estimate that pizza delivery search volume has increased by about 8% (the 95% CI is [0.12%, 16%]) over what we would expect at this time of year, if there were no pandemic. There are a lot of caveats to this estimate, e.g. that the trends data is normalized in opaque ways, the timeseries model that CausalImpact uses makes a bunch of assumptions, we only have a few data points, customers might start on aggregators instead of Google, states shut down at different times, these bands are often obscenely wide, etc. If you’re curious, here’s the R code that I used; should be easy to check other search terms.
Other credit card-derived data show roughly a small growth in sales for takeout-focused pizza restaurants, while dining experience-focused chains have seen massive dropoffs. It’s hard to estimate the specific effect of delivery from this data, and companies are not really breaking that metric out either. The best we can say is that the numbers don’t look terribly wrong.
We could compare “pizza delivery” vs “pizza” search trends. Overall search volume for pizza is down 1%, but with statistically insignificant results; you could say that there isn’t evidence that search volume for pizza has changed, but searches have mix-shifted more towards delivery (+8%) or cooking (+121%).
As a side note, I attempted to make some pizza and it was a disaster. San Marzano tomatoes, organic cheese, dry-aged meat, gourmet mushrooms, easy. Measuring out the correct amount of yeast, flour, and water?? Literally impossible. Anyways there are a lot of other people also messing up their pizzas.
I attribute pizza’s resilience to a few factors:
Consistency: you know what you’re going to get, whether you’re ordering Domino’s, Delfina’s, or a Safeway frozen pizza.
Strong chain presence: “consistency” might be a subset of this result, but around 50% of the US pizza market is chain vs independent shops. Chains should have better balance sheets and operational leverage with respect to marketing and delivery.
Stores well: pizza the next day is great cold, or easily reheated.
Nostalgia: it’s such an American Dish!
Who’s coming to (deliver) dinner?
But what about other dishes or cuisines that don’t hold up as well on these fronts?
Immediately, I think of pho or ramen: absolutely delicious, but short shelf-life and don’t travel well because of the soup base. Pho and ramen also have much higher variance in their quality, or maybe I don’t know how to appreciate bad versions, like I can appreciate bad pizza (usually involves alcohol). My assertion is that there is one great pho restaurant in SF and no great ramen places. The last time I wrote about that, though, some white dude mocked me in Wired magazine;
Pho also is very cyclical, it’s a warm noodle soup that you want to drink when it’s cold out. But man, that dropoff, after 5 years of increasing popularity.
All of this got me curious about what the breakdown was, between what type of cuisine is getting delivered, what people are cooking, and so on. I did a more systematic analysis of some cuisines and their associated behavior, comparing the post March 8th behavior with the predicted behavior. Note that these numbers are point estimates, the methodology has plenty of problems, but directionally everything looks reasonable.
In general, we see that overall searches for “x cuisine” and for restaurants are down, while recipe and delivery searches are way up. People are cooking and ordering more, they’re eating out less.
The one outlier here is Chinese food delivery: it’s basically unchanged versus before coronavirus, while every other cuisine has seen large jumps in search volume. That surge lines up with credit card receipt data, which indicate that spending on food delivery has grown between 10-30%.
If restaurants as a whole are struggling to survive, despite increased delivery volume, how are Chinese restaurants doing with no increased delivery volume and no dine in service? Is that related to the president continually calling coronavirus a “Chinese disease” on national TV, or to these continuous bat soup jokes?
Or is it because they’re closed out of fear? A Michelin-starred Korean restaurant in NYC was graffitied with “stop eating dogs.” So it’s not a huge surprise that “Kim hasn’t been offering delivery from his shop — one of the only ways that restaurants can continue to operate during the shutdown — in part because his staff has felt unsafe commuting to work on public transportation due to fear over anti-Asian harassment.” Oh, also, “Tang and Ng said that they submitted applications for the small business relief program rolled out by Mayor Bill de Blasio last month, but hadn't received any updates from the city.” Yeah I’m not holding my breath.
This is another reason why the commissions charged by platforms such as Uber Eats are so insidious: taking 30% or more from an order makes the order unprofitable for many independent restaurants, especially Chinese restaurants. There will always be well-capitalized entities attempting to “disrupt restaurants,” but none of them know how to cook.
If you can, please support your local Chinese restaurants (skip the middleman). And stay safe.