Nearly 60 percent of Uber riders never tip, about 1 percent always tip, and those who tip leave an average of $3, according to a new analysis of the company’s rollout of in-app tipping in 2017. And if you loved those gratuity factoids, rest assured these are just the tip of the iceberg.

We all have theories or stereotypes about who tips best and which workers earn the most, but honest-to-goodness tipping facts are rare. Economists struggle to get data on gratuities, even though they total about $36.4 billion each year in the U.S. and form a load-bearing pillar of many Americans’ livelihoods.

But this week, we got a truckload of new tip factoids in an analysis of more than 40 million Uber trips circulated as a working paper by the National Bureau of Economic Research.

(Three of the paper’s four authors, Bharat Chandar of Stanford University, John List of the University of Chicago and Ian Muir of Lyft, worked with Uber at the time the data was collected. They retain equity in Uber, where the post-IPO blackout on selling shares will end in early November. The other, Uri Gneezy of the University of California at San Diego, isn’t affiliated with the company.)

The data was collected during four weeks in August and September of 2017, soon after Uber added tip support in June and July of that year. The economists analyzed location, rating, ride quality, and characteristics of the driver and rider, but couldn’t track outside variables like cash tips or personal interactions.

Cornell University psychologist Michael Lynn, a widely cited tipping researcher who was not involved in the study, says the sheer volume of data is a significant step forward for a discipline that often relies on small-scale studies involving a handful of workers and, at best, thousands of interactions.

“It’s not an easy problem. It’s hard to get the granularity of information that we see on Uber,” said Chandar, who also shared the paper’s results in a majestically wonky blog post. “We can observe lots of things about the experience that other datasets can’t do very well,” he added later.

Lynn, now in his fourth decade of tipping research, agreed. “This is so much more data and it’s such a unique context that this kind of stands on its own,” he said.

And that’s exactly what we’ll let these factoids do: stand on their own.

If you know who the rider is, you can guess the tip regardless of trip or driver quality. Rider’s identity accounts for three times more of the difference in tip size than driver identity does, the authors found. Like the other figures we’ll list, this has been adjusted for the time and location of the rides.

Men tip more often on Uber. Women tip 14.3 percent of the time. Men tip 17.0 percent of the time. Men’s tips are also bigger. But other research shows the gender tip gap varies depending on context, Cornell’s Lynn said.

“There are circumstances where I find men are bigger tippers, but there are many circumstances where I do not find that, so I do not believe men are better tippers across the board,” Lynn said.

It seems plausible, given complaints of harassment, that women on Uber tip less because they’re getting a worse experience in the vehicle.

“We don’t see what’s happening literally in the car,” Chandar said. “We don’t observe the conversation between rider and driver.”

An Uber spokesperson noted the app had added several safety features since the data in the study was collected in 2017.

Men tip 12 percent more if their driver is a woman, but that’s entirely because they give more money to the youngest female drivers. The premium men pay to women behind the wheel shrinks as the women get older. By the time the drivers are age 65, it has virtually vanished. Women also tip other women more, but they don’t significantly change their tips based on the driver’s age.

Drivers who use the Uber app in a different language get 30 percent lower tips. Users can change the Uber app’s default language to Spanish, Chinese, French and several other languages. Researchers used this setting to estimate whether someone was an immigrant, or at least didn’t speak English as a first language. Riders who use the app in a different language also leave lower tips, the authors found.

Drivers from areas with lower incomes or larger Hispanic populations get tipped less. The researchers weren’t able to measure rider or driver ethnicity directly, but they can guess at some characteristics based on their home addresses. Riders from higher-income, whiter ZIP codes tip more.

In a 2005 Yale Law Journal essay, Ian Ayres, Fredrick Vars and Nasser Zakariya analyzed more than 1,000 tips from New Haven, Conn., taxi drivers and found black drivers were tipped about a third less than white ones, and black and Hispanic passengers tips about half as much as white ones.

In a 2014 Sociological Inquiry analysis, Lynn and Zachary Brewster of Wayne State University surveyed almost 400 restaurant customers and found both black and white customers tipped black servers less, and that the disparity wasn’t due to differences the quality of service provided.

An Uber spokesperson said the company was committed to detecting and mitigating bias on the app, and that almost $2 billion in tips had passed through the Uber and Uber Eats apps since the feature was implemented.

Riders leave higher tips if their driver is on time. They tip less if a driver accelerates too sharply, brakes suddenly or speeds during the ride. They also tip slightly less if the driver picks them up in older-model car, which the authors define as one manufactured before 2009.

If you see a driver again, you’ll tip 27 percent more the second time. You’ll tip even more the third time. Based on tipping patterns, the researchers rule out factors such as conversation and game theory and conclude the additional tip is likely due to guilt or social connection.

Riders rated a perfect five stars tip twice as often as those with 4.75 stars. And when they leave a tip, it’s about 14 percent higher. Riders can’t tip until after they’ve been rated, and drivers can’t see a rider’s past tipping behavior before or during a ride. Drivers with five stars are almost 50 percent more likely to get tipped than those with just 4.75 stars, and their tips are about 5 percent higher.

Tips are highest in small cities and the middle of the country. Riders in California and the Northeast weren’t great tippers. This is true both as an absolute amount and as a percentage of the fare. The economists couldn’t include New York City in the study because tipping was handled differently there.

Tips are highest between 3:00 a.m. and 5:00 a.m. There are quite a few trips to the airport and other business-related destinations during these hours, and it’s reasonable to assume that riders on an expense account are more generous with their tips. Tips also jump on Friday and Saturday evenings.