WASHINGTON (AP) _ Farmers have been increasing their productivity for a long time. Horse power helped, and so did the mechanical reaper and the tractor.

But the last half-century has produced changes that even Agriculture Department statisticians find difficult to describe. They keep trying, however.

One example, which is popular with agribusiness speech writers, is the number of people a single American farmer feeds. It's a statistical exercise and results in some fractions that are not very humanlike.

In 1947, according to a new Economic Research Service study, one farm worker produced enough to feed 14.1 people, including the worker. Those included 12.6 in the United States and 1.5 in other countries.

By 1989, the most recent year included in the report, a single farm worker supplied 98 people, including 73.7 at home and 24.3 abroad.

It's more complicated than these numbers, however. And the report said that it ''would be misleading'' to suggest that the labor of one farm worker - a term that includes farm operators, hired workers and unpaid family members - produced the food, fiber and tobacco consumed by 98 people.

Many jobs have moved to the non-farm sector, the report said. Thus, many non-farm workers now perform tasks previously done by farm workers.

''For every worker on farms, five to six non-farm workers provide resources and services in producing, processing, transporting, wholesaling and retailing farm products,'' the report said.

There are many other ways of looking at farm production and efficiency, and the Economic Research Service provides a detailed view of some.

Increases in farm production and the efficiency with which crops and livestock are produced are the foundation blocks of much of the nation's economy and social structure.

The migration of millions of people from the land in the years after World War II - a process that continues - prompted the survivors to enlarge their farms. Get bigger or get out, it was said.

That led to larger tractors, combines and other machines that enabled fewer people to till more acres. It saw the increased use of chemical fertilizers and pesticides, erosion of fragile land and the rise of many of today's environmental problems.

Trade in agricultural commodities surged as farmers in other countries joined the parade to supply rising demand.

But productivity growth had a negative side as surpluses grew larger than demand or exceeded the capability of needy countries to buy them.

In one way or another, much of the farm legislation passed by Congress in the last half century or more is linked to the sensational rise of productivity of American farmers.

So the economists and statisticians in USDA annually try to put some new spin on the numbers by updating and revising them in a five-part series called Economic Indicators of the Farm Sector. The latest is Production and Efficiency Statistics, 1989.

Basically, the idea is to assign an index number - a base of 100 in 1977 is used - to show the comparative changes in certain agricultural indicators. Production is a major item, and so is the amount of things needed to produce.

Economists like to talk about the ''inputs'' of agriculture, which include land, labor, machinery, chemicals, feed, seed and just about everything used to produce crops and livestock.

These inputs are also divided between ''non-purchased'' items such as a farmer's own labor and that of his family, and ''purchased'' items, which are everything else required on the farm.

For describing what has happened on the farm, few comments can equal the stark, dry numbers in the input category.

In 1947, just two years after World War II, farmers used 4 percent more total inputs than they would in 1977, the year used as the comparative base. Non-purchased inputs were 74 percent greater than the base, while purchased inputs were 41 percent less.

The 1947 inputs included about triple the use of farm labor than in 1977, indicating that human muscle power was still a highly important factor in postwar farming.

While total inputs in 1949 reached a peak of 8 percent more than the base year, the general trend was down slightly, generally hovering at just under the 1977 level until the late 1970s, when total inputs actually increased for a time.

Since then, the trend has been downward, and in 1989 total inputs used by American farmers were 12 percent less than the 1977 total.

More remarkable, perhaps, has been the decline in non-purchased inputs, from 74 percent over the baseline in 1947 to 21 percent below it in 1989.

At the same time, purchased inputs, which were only 59 percent of the base in 1947, were reported at 98 percent in 1989. Those actually rose to more than 100 percent in the expansion years of the late 1970s and early 1980s.

Labor, the other big item for inputs, plummeted from nearly triple its 1977 level in 1947 to 24 percent below the baseline in 1989.

Meanwhile, since 1947 the actual production of American farmers rose from its postwar level of 42 percent below the baseline to 14 percent above it by 1989.

Thus, with fewer total inputs, including less labor, U.S. farm production increased dramatically in less than half a century. And that translates into a higher rate of productivity and efficiency.