Strength In Numbers: How Polls Work and Why We Need Them. By G. Elliott Morris. W.W. Norton; 224 pages; $28.95 and £21.99
In the 1920s, George Gallup sought to expand the circulation of his student newspaper. To gain readers’ attention, he published a misogynistic article entitled “The Unattractive Women”; it won his student rag so many new fans that the Daily Iowan quickly became a profitable newspaper. Readers claimed to be interested in editorials and news, not comics or gossip columns, but Gallup was right to suspect otherwise.
He began more careful studies, literally looking over readers’ shoulders to observe which pieces truly seized their attention. It was the beginning of his journey from journalist to father of modern opinion polls. G. Elliott Morris, a data journalist at The Economist, has written both a history and a defence of opinion polling; his story about George Gallup hints at many of the topics this lively book explores.
There is a fundamental problem with polls: when pollsters ask questions, the answers they receive may be less than candid. There is also the easily overlooked fact that opinion polling has always been about making money, with discovering the truth as an secondary motive. And there is the disheartening truth that, while opinion pollsters try to discern what people think and feel, what people think and feel can be ignoble. Gallup debased his student newspaper to please his readers, and politicians may debase policymaking to please the voters.
Gallup himself was unabashed. In 1940 he co-wrote “The Pulse of Democracy”, a book which argued that successful governments would be “responsive to the average opinion of mankind”, an opinion which “for the first time in democratic history” could now be continuously and objectively measured. Mr Morris concurs. “Polls are a distillation of the general will first,” he writes approvingly, “and everything else second.”
While most people think of polls as a prediction of election results, Mr Morris shows how they became a constant input to political decision-making. Gallup’s near-forgotten contemporary, Emil Hurja, was famous in the 1930s as the “Wizard of Washington”. His data on public opinion shaped the decisions of Franklin D. Roosevelt’s administration. The Simulmatics Corporation, run by social scientists from the Massachusetts Institute of Technology and Yale, and armed with the latest computers, offered precise (if not always accurate) predictions as to how each of John F. Kennedy’s positions would affect his popularity with different voter segments. Richard Nixon was a prolific procurer of polls.
Pundits then and now worried that if politicians obsessed over opinion surveys, policymaking would become an act of followership rather than leadership. Mr Morris suggests instead that it is better to have a political class that attends to public opinion than one which ignores it, and declares that “public opinion polling has been one of the most democratising forces in American political history.” He even speculates that the Vietnam war might have ended far earlier if only Lyndon Johnson had been as interested in polling as Nixon was.
Mr Morris does not shy away from the horror stories. He eviscerates some influential but misleading surveys of mortality in Iraq and grumbles about partisan push pollsters, who ask loaded questions such as: “Would you still vote for [John] McCain if you knew he had sex with prostitutes and gave his wife venereal disease?”
Given the subtitle of the book, it is surprising that Mr Morris waits until the second half to properly discuss sampling, the most fundamental idea in polling. When he describes the fallout of Donald Trump’s win in 2016 for pundits and pollsters, Mr Morris mixes a vivid journalistic account of modern polling failures with a learned but challenging range of acronyms and technical details.
Polling is flawed, and some of those flaws seem unfixable. But Mr Morris’s repeated refrain is that the critics of opinion surveys overstate their case. If you think polls can mislead, just try understanding the electorate without them. Alas for pollsters, they will always be expected to forecast elections. From an early fiasco in 1936, through Gallup’s “Dewey Defeats Truman” humiliation in 1948, to Mr Trump’s allegedly impossible triumph in 2016, Mr Morris sorrowfully reminds us that pollsters are judged by results. Those results may vary.
Written for and first published in The Economist on 26 August 2022.