The number of Americans killed in traffic accidents fell 10 percent last year, making the death rate per passenger mile the lowest in the nation's history.

The year's 46,300 automobile, motorcycle and truck deaths--5,200 less than the year before--were the fewest since 1975.

The last major drop occurred between 1973 and 1975, when the Mideast oil embargo, gasoline shortages and lowered speed limits cut travel significantly and deaths fell 18.5 percent.

But the reason for last year's drop, actually one that began in early 1981, is largely a mystery, according to the American Public Health Association.

There may have been a cluster of contributing factors, including new wars on drunk driving and a profound change in America's demography. "The number of persons aged 16 to 19 declined by 3 percent last year," said James Hedlund, chief of mathematical analysis for the government's National Highway Traffic Safety Administration.

Yet the single most important factor was probably bad times, most analysts agree. Motor vehicle deaths usually climb in a good economy and drop in a bad one.

Still, miles driven actually rose 1 percent in 1982 despite the troubled economy and unemployed drivers. It was the combination of an increase in miles driven and drop in deaths that lowered the motor vehicle death rate to 2.95 deaths per 100 million miles.

What must have changed is the driving pattern, and this could be somehow related to the economy, said Brian O'Neill, senior vice president for the Insurance Institute for Traffic Safety, Hedlund and Alan Hoskin, statistical manager of the National Safety Council

"How, we don't know--we don't know what there was about the economy that caused the change," O'Neill said.

For example, he said, in hard times "we'd expect less social driving and less drinking and driving." This should mean fewer night crashes and single-vehicle crashes. Yet all types of auto fatalities went down at the same rate.

One thing that was expected for years did not happen: smaller cars have not so far caused an overall increase in deaths.

It is established, said O'Neill, that "for every 10,000 small cars that replace 10,000 large cars, you double occupant deaths." The number goes from 1 to 1.5 deaths per 10,000 full-size cars to between 2 and 3 deaths per 10,000 small ones.

This grim fact made some safety experts predict a few years ago that the growing small-car fleet and rising speeds would by now have caused 60,000 to 70,000 highway deaths every year. As it turned out, O'Neill said, "We've still got a large number of large cars out there. Small cars are not yet predominant."

"I have no proofs," Hedlund said, but "many things" besides the economy may have helped produce last year's major change.

He pointed to the crackdowns on drunk drivers, with at least 23 states either hardening penalties or increasing the drinking age from 18 to 21 or doing both. Unfortunately the effect of such crackdowns is usually temporary, O'Neill said, since drivers soon learn that "their risk of being caught is minuscule, about one chance in 2,000."

There have been some improvements in enforcement, Hoskin said.

"The key to reducing drunk driving is not tougher penalties, it's tougher enforcement," O'Neill agreed. "When people think there's a real chance of being apprehended, they change."

We are also driving somewhat safer cars, the result of several improvements, Hoskin added. Many, like energy-absorbing steering columns and laminated windshields, were dictated by stronger federal standards of past years.

O'Neill said the decline is not the result of the increased use of seat belts, either. "There's only about 11 percent use, " he said, and over the years use has declined. "When there were only lap belts, 20 percent used them," he explained, but many people who fastened lap belts shun shoulder belts, and lap and shoulder belts are now connected.

The danger, O'Neill said, is that with a quickened economy, car deaths will increase again.

The latest government figures show a leveling off of the number of traffic deaths during January and February and only a slight decline in deaths in March. Bad weather in March reduced driving nationwide.