Leslie Kish; Pioneering Statistician

University of Michigan statistician Leslie Kish, who pioneered what is now called margin of error in surveys, died Oct. 7 in Ann Arbor, Mich. He was 90.

Kish introduced the idea that non-responses must be counted to produce the most accurate probability sample. His technique was validated in 1948 when he and his colleagues, using a small national sample of fewer than 1,000 U.S. households, found President Harry Truman running slightly ahead of Republican challenger Thomas Dewey. Other polls predicted a Dewey landslide, but Truman won a narrow victory.

In a classic 1949 paper, Kish became the first to develop a technique for random selection of poll respondents within a household. His method for choosing one adult in a household depended on careful questioning to determine the age and sex of each household member and relationship to the household head.

His 1965 book, “Survey Sampling,” is a classic text still in use.

Born in 1910 in the Austro-Hungarian city of Poprad (now in Slovakia), Kish immigrated to New York in 1925. He interrupted his undergraduate studies at City College of New York to fight for the Spanish Loyalists as a volunteer in the International Brigade, returning in 1939 to earn a degree in mathematics.

He helped create a survey research program in the U.S. Department of Agriculture before serving in the U.S. Army Air Corps as a meteorologist during World War II. After the war, he joined the faculty of the University of Michigan, where he was a co-founder of the Institute for Social Research.

He conducted statistical work full time while obtaining his master’s in mathematical statistics in 1948 and a doctorate in sociology in 1952. He trained statisticians from more than 100 countries through a summer program he initiated at the university in 1948. He was a past president of the American Statistical Assn. and a fellow of the American Academy of Arts and Sciences.