Silicon Valley is pouring billions into robot cars. Soon – although the time scale keeps shifting – tech manufacturers say driverless cars will replace their traditional counterparts, car parks will become parks again and road fatalities will plummet. People have argued over ethical concerns surrounding the technology, the ensuing job losses and the public’s antipathy to this robot revolution. But the biggest obstacle may well be money.

We have been taking a deep look at the economics of driverless technology. Our conclusion? So-called robocars are unlikely to produce the societal changes tech companies are promising not because they don’t work but because they will cost too much.

Affordability is something self-driving enthusiasts often take for granted. They promote a future where sharing is caring. In this future, society is served by robocabs – for-hire vehicles that operate much like taxis today. The key difference is that software algorithms, rather than humans, are now in control. Without a driver to pay, rides become cheap enough to compete with conventional car ownership. This prompts consumers to abandon car ownership in droves and road safety ultimately soars. At least, that’s the idea. However, the reality is very different.

Driver wages are a key part of taxi fares today. The average cab ride in San Francisco for example will cost you around $13. The driver keeps most of that. There is one caveat, however. Taxis are inefficient – so inefficient in fact that cabbies only spend about half their time earning fares. The rest is spent finding them. This “matching inefficiency” dramatically shrinks a driver’s actual take-home pay. It also cuts the potential savings driverless technology can deliver – and that affects the technology’s affordability.

In fact, assuming current market conditions persist, our work shows that hailing a robocab would actually cost consumers significantly more – on a per-mile basis – than owning a car today. Additionally, even if robocabs were miraculously always occupied (Uber and Lyft fare better than traditional taxi companies in this regard), we find profits would still need to shrink considerably for the technology to be cost competitive with the status quo. Ride-pooling would help. Divvying up fares is an easy way to make driverless technology more affordable. But the concept doesn’t have many fans. Pooling after all, merely trades one set of costs (capital) for another (opportunity).

For all the buzz about self-driving cars it’s worth remembering Silicon Valley’s big bets don’t always pay off. Just look at Facebook’s Aquilla program. Its goal was to build a solar powered airplane that would provide beam internet access to millions. After years of development, Facebook quietly shelved the effort in 2018. But it would be a great pity if the tech companies can’t find an economic rationale for self-driving cars.

Road crashes claim over 40,000 lives in the US annually. Countless more are injured. Most of these crashes are blamed on human error – things like drinking and driving, texting while driving and napping while driving. Driverless cars should fix this. Machines after all, don’t (or more accurately, can’t) get drunk, distracted or drowsy.

Self-driving technology is also particularly important for America’s poor. While this group has experienced an uptick in car ownership, it’s come at a price: high mortality rates. Poorer people are more likely to die on the road compared to those well-off. One reason why is that poorer people often own older vehicles that lack important safety features – things like cruise control, night vision and pedestrian detection, each of which help guard against driver error. Without them, crash risk rises. That makes poorer people more susceptible to the harmful effects of human imperfection. It also means they benefit more than any other socio-economic group from driverless technology. At least they would were it actually affordable.

Addressing these issues won’t be easy. But it is necessary. Socio-economic differences in road death rates haven’t just persisted over time – they’ve gotten worse. That makes driverless technology a solution worth pursuing.

That’s where government comes in. Subsidized access to driverless technology should be offered to America’s poor. Those opposed to such a move must consider that public revenue already pays for a portion of road crashes. If subsidies can deliver an equivalent or greater benefit, using the public purse becomes fiscally prudent. Subsidizing driverless car access is also more effective than improving mass transit. The latter may be greener but it is inconvenient and slow. For the poor, this means being unable to go where they want, when they want, which ultimately impedes upward economic mobility. Conversely, low-income Americans that rely on cars experience less poverty and fare better when it comes to getting and keeping jobs.

That’s the real promise of driverless technology – lifting millions out of poverty without having them pay for it with their lives. It’s a promise worth keeping.