Subsidies for electric vehicles are a huge mistake. These cars are conceptually the same battery and motor as a fourth-grade science project—not a great innovation. And given high prices for EVs, subsidies are mainly a giveaway to the already well-off. If you add up carbon emissions from manufacturing, daily use and end of life, EVs have total life-cycle emissions 30% lower than gasoline-powered autos. In Silicon Valley, something is considered truly transformational if it’s 10 times better, not a third.

And now there’s a glut of them. Ford is losing billions, and Honda and General Motors have scrapped plans to build affordable EVs. Instead of throwing taxpayer money at EVs, President Biden could have been a hero and helped bring autonomous vehicles to the market faster.

Why? Start with the 42,795 traffic fatalities last year. Cost-benefit analysis involves something called the Value of Statistical Life, and the Transportation Department uses $12.5 million per traffic death. Lowering annual crash fatalities to 10,000 would be worth $400 billion to the U.S. economy every year. Technology seriously reduces driver error, a cause of many accidents.

Another reason? With fully autonomous vehicles, the U.S. would need only half of its nearly 300 million cars. Most sit around doing nothing. Transportation as a service would become reality—no need to own cars; simply click and a driverless one shows up when you need it. That would mean way less emissions and no parking hassle or road rage.

So what’s the problem? Tesla’s Full Self-Driving capability, priced at up to $300 a month, still runs stop signs. The company uses simple CMOS image sensors, like those in an iPhone, instead of more-expensive Lidar pulsed lasers. Teslas get confused and sometimes run into emergency vehicles with flashing lights. I recently rode in a Model 3 in self-driving mode that botched a simple yield.

It has been a bad few weeks for autonomous driving. Self-driving taxis from Cruise and Waymo in San Francisco halt at unexpected construction barriers. Or at least they used to. GM’s Cruise cars had their California driverless permit revoked last month after a pedestrian hit by another car was pinned under a driverless Cruise. A rare case, but it did happen.

But, even with their problems, a joint study (albeit a very early one) by Alphabet-owned Waymo and the insurer Swiss Re showed 95% fewer injuries and 76% less property damage from autonomous driving vs. humans. That’s a preview of 10 times. Auto insurance may force an autonomous shift. Imagine $500 a year for autonomous-car insurance, but $2,000 if you want to drive.

The Society of Automotive Engineers defines basic cruise control as Level 1 autonomous. Tesla’s FSD, which still requires human attention, is Level 2 autonomous. Last month, Mercedes announced Drive Pilot, which uses Lidar and is currently the only Level 3 “conditional driving automation” system approved in the U.S. and Europe. Drivers can take their hands off the wheel and eyes off the road, even surf the web, but only while driving less than 40 miles an hour and following another vehicle in dry conditions during the day. What’s needed for an economic transformation is Level 4: autonomous driving at all speeds in clear weather. Level 5 is self-driving even in rain, snow and fog.

What can government do? Help the imaging systems in cars, which will never be totally accurate, by actively letting them know where things are. Stop signs could emit signals with their GPS locations. Traffic lights could digitally broadcast red, yellow and green. Digital signal transmitters in highway markings—dotted line, solid line, etc.—would digitally paint the road. Car sensors could see the real signs or paint and confirm them digitally. Ambulances and police cars could broadcast their locations and warn others to slow down and stay away. Construction barriers and traffic cones could broadcast their location and indicate how long they will be there to help cars update their maps. Cars could even negotiate who gets to go first at a four-way stop sign. Encrypted signals would prevent hacking.

Don’t confuse this with smart cities or smart roads, the dreams of central planners. The Silicon Valley adage works here: Intelligence moves out to the edge of the network. Make cars smart, and roads dumb but digitally visible.

Expensive? Back-of-the-envelope math: There are 4 million miles of road in the U.S., with perhaps 10 to 25 signs per mile. So we need to update 100 million metal signs and 300,000 traffic lights. I’ll assume the cost of sign transmitters is $100 and road-marking transmitters to paint the lines digitally is $10 for every tenth of a mile. Even if I’m off by a factor of 10, the government could spend less than $400 billion to save 32,000 lives a year, reduce emissions, lower capital costs of transportation, and transform the U.S. economy. That sure beats subsidizing the EV purchases of climate-smug rich folks.

The cost of EVs and batteries was going to fall anyway without handouts. It would have been better to spend that money making autonomous driving viable. What a waste.

Write to kessler@wsj.com.