The most spectacular advertisement for a research method was a competitor’s failed election forecast. In 1936, the American magazine Literary Digest expected Republican Alf Landon to win the presidency. Franklin D. Roosevelt won instead. George Gallup had identified the correct winner, and the comparison helped establish him as a public authority on opinion research. A smaller, deliberately constructed investigation could outperform an enormous collection of voluntary responses.
For marketing history, however, that episode is only a beginning. Gallup did not emerge from an exclusively political research environment. He had investigated what people read in newspapers and worked in advertising research. His career connected commercial questions about audience interest with the public ambition to make a population’s attitudes visible.
He founded the American Institute of Public Opinion in 1935. The date marks an important institutional development, not the invention of sampling. Earlier statistical work and contemporary research organisations also belong in the account. So does Gallup’s failed 1948 forecast: scientific polling developed through criticism and revision as well as through celebrated success (Harry S. Truman Library and Museum, n.d.; Roper Center, n.d.a).
Before elections came the question of what readers actually read
Gallup’s early work concerned interest in newspaper contents. His 1928 University of Iowa dissertation examined a more systematic way of investigating readers. Editors and advertisers faced a consequential commercial question: which parts of a publication actually received attention?
Susan Ohmer’s CHARM study shows how he subsequently developed this work at Young & Rubicam. Gallup became research director there in 1932 and remained until 1947. His investigations included reactions to visual and persuasive features of printed advertising (Ohmer, 2001).
The shift was from making judgements about audiences to investigating audiences. An editor or copywriter might consider particular material important without knowing how readers used it. Research could expose the gap. It did not replace creative decisions, but it could identify assumptions arising chiefly from producers’ professional perspectives. That connection with advertising is essential to the history: the later pollster had already worked on the problem of discovering whether an audience behaved as its interpreters imagined.
Why advertising agencies wanted empirical evidence
Advertising had to explain to clients why its contribution justified expenditure. An attractive execution and the client’s approval provided limited evidence. Investigating readers’ reactions offered an additional reference point.
Ohmer places Gallup’s agency work within the commercial pressures of the Depression. Research could help substantiate the agency’s usefulness to sceptical customers (Ohmer, 2001). Scientific language consequently operated within economic relationships rather than outside them.
That does not make the work automatically dishonest or useless. It explains why particular forms of evidence were sought. An advertiser asked different questions from an election researcher or a theoretical statistician. Those perspectives encountered one another in Gallup’s career. Methods could travel between assignments, but their suitability had to be reconsidered. A technique useful for investigating reader interest was not thereby a complete account of political choice. Understanding that transfer is more informative than assuming one general method could be applied unchanged to every kind of human judgement.
The American Institute of Public Opinion in 1935
The Truman Library dates the American Institute of Public Opinion’s foundation to 1935. The organisation aimed to measure and report opinion on political and social questions. Findings were distributed to subscribing newspapers and reached a broad readership (Harry S. Truman Library and Museum, n.d.).
The institutional name conveyed scientific ambition; newspaper distribution made the findings a publishing product. Methodological work and public visibility were therefore connected from the beginning.
This is important to the history of professional research. A procedure does not become socially influential merely by existing. Findings need to be produced repeatedly, circulated and recognised as relevant. Gallup’s organisation connected those activities. People encountered polling not simply as an internal aid to business decisions but as recurring news about what other people thought. That created an additional stage on which research could acquire and lose credibility. The history concerns the organisation and communication of knowledge as well as the process through which answers were obtained.
An early question about government spending
Gallup’s own retrospective identifies its first question with an investigation conducted on 10-15 September 1935 about government expenditure for relief and recovery. Nine per cent considered spending insufficient, sixty per cent excessive and thirty-one per cent approximately appropriate (Gallup, n.d.).
The example reveals how attitudes became a public aggregate. A complicated policy issue was expressed through specified alternatives and reported as a distribution. The result was neither an election nor a detailed explanation of each respondent’s judgement.
Those limits belong to its meaning. The percentages describe answers to that particular question. They do not establish precisely which programmes respondents considered or which alternatives they preferred. This became a fundamental issue for market research too: standardisation makes comparison possible while reducing the variety of explanations that people might otherwise give. A structured answer can be useful without being treated as a complete representation of the thought process behind it.
What a sample is intended to do
A sample observes part of a larger group about which a statement is required. Its quality depends on how that part is selected and how the observations relate to the intended population.
Size alone does not resolve the problem. A study interviewing only enthusiastic customers may obtain many responses while missing dissatisfied buyers and people who stopped purchasing. It can describe genuine participants accurately without representing everyone the business wishes to understand.
This is not a difficulty invented by digital questionnaires. Sampling history involved arguments about selection and representativeness long before Gallup’s institute (Brewer, 2013). His public success increased awareness of a principle with a much older theoretical and practical history. It made the question of who participates difficult for a general audience to ignore. The important insight was not simply that fewer interviews could suffice, but that the relationship between the observed people and the target population mattered more than an impressive response total considered alone.
Quota sampling and probability sampling are different
Gallup’s early investigations used quota procedures. Their composition was intended to correspond with selected characteristics of the population. That differs from selection in which inclusion probabilities are established through a random mechanism.
Popular accounts sometimes convert the 1936 success directly into a triumph of contemporary probability sampling. Doing so merges historically distinguishable procedures. Sampling history includes debates between purposive selection, randomisation and later model-based approaches rather than one instantaneous methodological settlement (Brewer, 2013).
Matching distributions on selected characteristics does not guarantee similarity on everything relevant. Within a quota, easily reached people or those especially willing to speak can predominate. A procedure may correct certain differences without controlling every selection mechanism. Scientific ambition should therefore invite examination rather than terminate it. This distinction also protects Gallup’s actual contribution: a historically specific method can be influential without being credited retrospectively with every feature of the methods that later became authoritative.
Literary Digest and the limits of a large total
The famous Literary Digest investigation depended on questionnaires distributed in large numbers and returned voluntarily. The people contacted and those who replied were not equivalent to a balanced representation of the entire electorate.
The failure is often explained entirely through the social composition of the address lists. Peverill Squire’s later investigation also emphasised differential response. The mistaken forecast cannot adequately be reduced to a single defect in the initial list (Squire, 1988).
The general lesson is specific. A large number of observations can reduce random fluctuation without automatically eliminating systematic distortion. If certain opinions enter more frequently, additional responses can stabilise the same misleading picture. This makes 1936 a useful historical case for data-based decisions. Observation volume and the suitability of its origin are different dimensions of quality. Collecting more material from a poorly aligned process does not necessarily move the result closer to the population the researcher intends to describe.
Gallup gained credibility rather than infallibility
Gallup identified the right winner in 1936 while Literary Digest failed. Elmo Roper’s and Archibald Crossley’s organisations also identified a Roosevelt victory. The breakthrough belonged to a wider development of professional polling rather than one individual alone (Roper Center, n.d.a; n.d.b; n.d.c).
Two standards must also remain distinct. Correctly naming the winner is a substantial success. It does not demonstrate that every estimated vote percentage was precise or that all methodological problems were resolved.
A forecast’s public impact can therefore exceed the narrow evidential claim established by its accuracy. It can produce trust, attract clients and change the reputation of a method. A careful history should describe those effects without treating them as guarantees of future correctness. Subsequent events showed that even a successful supplier had to keep testing its assumptions. Gallup’s importance does not depend on believing that one election permanently settled how every future survey should be conducted.
Nonresponse concerns the relationship between answers and silence
Not everyone contacted answers. What matters is not just how many fail to respond but whether those people differ in relevant ways from participants. If they do, distortion can result.
Imagine a company inviting all purchasers to comment. People with exceptionally positive or negative experiences may participate more readily than those with unremarkable experiences. Every answer may be authentic while the distribution remains misleading for the full customer base.
The historical discussion of Literary Digest makes the question especially tangible (Squire, 1988). A completed questionnaire is not simply a random fragment of society that happened to appear. Responding is an action with its own conditions. Anyone moving from answers to a wider population needs to consider people whose voices are absent. Silence is not an additional substantive response category, but it can signal the boundaries of a collection process. The problem connects early postal investigations with later research settings without implying that their technologies or solutions were identical.
Wording and response options belong to the measurement
A question does not measure an attitude independently of its formulation. People react to words, examples and offered alternatives. Asking about satisfaction can elicit something different from asking about disappointment even when the same supplier is involved.
The history of Gallup’s field consequently concerns more than improved selection. It also involves translating complicated experiences into questions that people can answer in comparable forms.
The implications are analytical. Two studies using different scales may produce changes partly attributable to the instrument. Combining or renaming categories can limit comparison. Repeated research needs an understanding of what remained constant, not simply another round of interviews. The ease of calculating a percentage should not conceal how strongly the question establishes what the result means. Nor does a familiar response format guarantee that respondents interpret its wording in the same way. Standardisation is an achievement with conditions, rather than an automatic consequence of printing identical questions.
A concrete intersection of consumption and social judgement
In 1935, Gallup asked about local drinking conditions after Prohibition ended. Thirty-six per cent considered conditions better, thirty-three per cent worse and thirty-one per cent approximately unchanged compared with the last Prohibition years. The company’s retrospective dates the investigation to 8-13 October 1935 (Gallup, 2019).
The question connects social arrangements and consumption. It does not measure actual drinking volume or immediate purchase intention for a particular brand. Those objects must remain distinct.
Its historical value lies in that difference. Survey research could investigate a dimension of consumption not directly visible in sales: judgements about a changed everyday environment. A business might use such findings to formulate further questions, but the answer was not a ready-made demand forecast. Social attitudes, individual action and commercial purchasing involve additional relationships requiring investigation. This example shows how research broadened what could be asked about consumers while also illustrating why every answer should be interpreted at the level of the question that produced it.
Hollywood made audience knowledge a production question
Susan Ohmer’s book on Gallup in Hollywood examines his research for studios and producers during the 1930s and 1940s. It includes responses to stars, stories and films around RKO, David O. Selznick and Walt Disney (Ohmer, 2006).
Research entered a field connecting creative decisions with substantial financial uncertainty. The question was not only how to advertise a completed offering. Knowledge of audiences could also affect assumptions about what to produce and how to present it.
The historical significance lies in this widening object of inquiry. Marketing-related research was not restricted to packaged goods or political preferences. It could investigate cultural expectations and reactions. That does not establish that successful films were simply assembled from poll results. Creative production, prediction and subsequent reception remain different activities. Research changes the basis of decisions without making taste wholly calculable. The Hollywood connection demonstrates both the commercial attraction of audience knowledge and the difficulty of converting existing reactions into certainty about a future cultural experience.
Elmo Roper was a co-developer rather than a minor character
Roper and partners established Cherington, Wood, and Roper in 1933. He directed the Fortune Survey from 1935, and his organisation also anticipated Roosevelt’s 1936 victory (Roper Center, n.d.b).
The comparison reveals the connection between commercial research and public opinion observation in another career. Gallup operated within a professional and competitive field, not entirely outside precedent.
This calls for caution about absolute origin claims. A famous research brand can dominate later vocabulary while the initial development involved several organisations. Gallup became a symbol; the historical field was broader. Roper helps make professionalisation intelligible as a partly collaborative, partly competitive process. Different suppliers could investigate similar difficulties and accumulate experience without every technique belonging to one founder. The existence of contemporaries does not dilute Gallup’s influence. It places it within the working environment that helped make opinion measurement a recognisable service rather than an isolated personal experiment.
Archibald Crossley and the connection with media research
Crossley’s career likewise passed through advertising and audience investigation. He worked for an agency and Literary Digest before running his own research firm, which included radio research. In 1936 he conducted election work for Hearst and identified the correct winner (Roper Center, n.d.c).
The connection is especially relevant to marketing history. Public polling and commercial audience research did not develop as completely separate worlds. People, methods and commissions crossed their boundaries.
The comparison also qualifies an uncomplicated story of technical progress. Someone formerly employed by an organisation later associated with failure could contribute to more successful work. Institutional affiliation was not a permanent quality guarantee. What mattered was the procedure used for a question and what practitioners learned from experience. Crossley’s history demonstrates why development cannot be reduced either to one inspired idea or to the inevitable victory of a particular type of organisation. The field changed through careers and applications as well as through published methodological arguments.
Earlier international foundations: Kiær and Bowley
The sampling debate substantially predates 1935. Anders Nicolai Kiær promoted representative partial investigations in the later nineteenth century. Ken Brewer’s history also identifies Arthur L. Bowley’s theoretical work of 1906 and his Reading investigation, published in 1912 (Brewer, 2013).
These examples arose in settings different from Gallup’s American institute. Precisely that difference matters. Statisticians and social investigators debated the legitimacy of examining a part rather than enumerating everyone well before election forecasting made the problem familiar to a broad readership.
Gallup’s achievement was therefore not the first discovery that a part could provide information about a whole. He contributed to the publicly visible, commercially organised application of surveys. The older international history limits an invention claim without denying the institution’s influence. Theoretical development, practical investigation and public establishment are different historical contributions. A responsible account can recognise all three without forcing them into one date or attributing them to the best-known name.
Neyman’s 1934 contribution beside Gallup’s institutional beginning
Jerzy Neyman’s 1934 paper explicitly distinguished two aspects of the representative method: stratified sampling and purposive selection. It preceded the American Institute of Public Opinion and belongs among the important foundations of modern sampling discussion (Neyman, 1934).
Proximity in time does not establish an unsupported direct genealogy. The paper’s existence does not prove that Gallup’s early quota procedures already implemented all its theoretical implications. Academic literature and commercial practice do not necessarily develop simultaneously.
Their parallelism is historically informative. While new suppliers attracted public attention, basic questions about selection and uncertainty were being clarified theoretically. Popularising a procedure and supplying its mathematical justification were connected but distinct. A careful account can value both developments without making either actor the inventor of every element. This also prevents later statistical standards from being used as though they described exactly what successful commercial polling already did in 1935 or 1936.
Britain and international diffusion
The British Institute of Public Opinion was established in 1936. After experimental work, it began publication in 1937; News Chronicle acquired publication rights in 1938. Questions from 1938-1946 now available through the Roper Center cover politics, war, economics and everyday conditions (Roper Center, 2020).
The breadth demonstrates that opinion research was not merely an election machine. It investigated how people evaluated change, shortages and social arrangements.
International development required more than translating a questionnaire. Political institutions and lived circumstances altered the significance of a question. American experience could provide a starting point without reproducing the same social environment. British polling therefore belongs within the account as a process of adaptation. The later growth of an international research name should not erase local institutional differences. Diffusion involved deciding what could be transferred, what required revision and which questions mattered in a new setting, rather than simply exporting a complete instrument into an identical society.
The failed forecast of 1948
On 1 November 1948, Gallup’s institute issued a forecast favouring Thomas E. Dewey. Harry S. Truman won. The Truman Library documents the releases as well as subsequent explanations and proposed improvements (Harry S. Truman Library and Museum, n.d.).
A Miller Center account made available by the library reports the final Gallup estimate as 49.5 per cent for Dewey and 44.5 per cent for Truman. It distinguishes the earlier collection of answers from publication immediately before the election (Miller Center, n.d.).
The difference matters. A publication date is not a fieldwork date. A freshly released number may describe older responses. Nor should the failure be attributed exclusively to one point in the timetable. Selection, likely participation and changes in preference also deserve investigation. A memorable explanation is not a substitute for methodological examination. Including 1948 alongside 1936 prevents a history of research from becoming an account of reputation that retains successes and edits out the occasions when confidence proved misplaced.
Examining failure was part of professionalisation
The mistaken forecast prompted systematic investigation. In 1949, Frederick Mosteller and co-authors published The Pre-Election Polls of 1948 for the Social Science Research Council. The existence of this report demonstrates that the failure became an object of organised methodological work (Mosteller et al., 1949).
Its importance does not require assuming that publication resolved every difficulty. It lies in comparing findings with an external outcome and making the research process itself subject to examination.
That matters particularly for commercial inquiry. An inaccurate claim can often be protected by retrospective explanations. An election supplies an unusually visible comparison. Even that comparison has limits because stated intention and actual participation can differ. It nevertheless forces distinctions between measurement, forecast and outcome. The 1948 episode is consequently not an embarrassing footnote to the method’s real history. It is part of that history: professionalisation includes the capacity to investigate why a trusted procedure did not deliver the expected result.
A margin of error does not cover every source of error
Under simplified assumptions of a simple random sample, a thousand respondents produce an approximate uncertainty of plus or minus 3.1 percentage points for an estimate near fifty per cent at the conventional 95 per cent level. This calculation illustrates a model; it does not describe Gallup’s 1935 procedure.
Under those assumptions, the number concerns random sampling variation. It does not automatically cover misleading wording, systematic nonresponse or the wrong target population. Other designs can produce different uncertainty.
The historical scientific claim therefore needs restrained interpretation. An interval is not a general guarantee against error; it depends on conditions. Methodological competence includes understanding and communicating those conditions. Quoting a mathematical result without its assumptions creates an appearance of precision unsupported by the actual procedure. The issue is not whether numbers should be used, but whether the statement accompanying them accurately describes what the calculation can establish about the particular investigation.
What changed for marketing and advertising research
Gallup’s career connected readership, advertising, public attitudes and cultural audiences. Surveying became a conspicuous means by which organisations could examine their assumptions about people.
Its contribution extends beyond the number of responses collected. It affected the relationship between producers and audiences. What interested buyers, readers and viewers should not be decided solely from the producer’s position. Systematically collected feedback could confirm, qualify or contradict those expectations.
This did not replace every other method. Sales data show something different from attitudes; observation differs from remembered experience; qualitative inquiry can explore reasons left unresolved by a structured questionnaire. Fullerton’s account of later motivation research emphasises contributions from several social-scientific traditions (Fullerton, 2013). Gallup belongs within that wider field rather than serving as a reason to compress its development into one procedure. Research became influential through an expanding combination of questions and methods, not through the disappearance of everything that could not be expressed in a poll percentage.
The limits of stated purchase intention
A person may find a product appealing without subsequently buying it. Prices may change, the item may be unavailable or another purchase may take priority. Asking about intention therefore observes something different from recording a later purchase.
This does not invalidate attitude research. It limits the conclusion. A question supplies evidence about a response under particular conditions. A revenue forecast requires additional assumptions about availability, circumstances and actual choice.
The historical connection between polling and marketing is most useful when it sharpens that distinction. Gallup’s successes made a form of knowledge credible. His errors showed that a relationship between stated preference and future action remained to be investigated. The fact that they can be related does not make them interchangeable in every setting. A valuable finding may identify what to ask next rather than supply the final forecast. That is a practical contribution to decision-making even when it falls short of predicting behaviour with certainty.
The enduring importance of 1935
The American Institute of Public Opinion made public-attitude measurement a regularly distributed service. Alongside other pioneers, Gallup gave surveying exceptional visibility. Several activities were subsequently brought together in the Gallup Organization in 1958 (Roper Center, n.d.a).
His historical significance lies neither in inventing every sampling procedure nor in infallible prediction. It lies in connecting research practice, institutional organisation and public communication. Advertising was an important part of that experience rather than an incidental prelude.
The strongest lesson is consequently more demanding than saying small samples are better than large datasets. What matters is how observations arise, which question they answer and what generalisation they can support. Gallup helped make those issues publicly visible. His achievements and mistakes belong together within that contribution to marketing history. Remembering both preserves the value of inquiry while resisting the temptation to turn scientific ambition into a promise that no commercial or public decision will ever again encounter uncertainty.
References
Brewer, K. (2013) ‘Three controversies in the history of survey sampling’, Survey Methodology, 39(2), pp. 249-262. Available at: https://www150.statcan.gc.ca/n1/pub/12-001-x/2013002/article/11883-eng.htm (Accessed: 21 October 2026).
Fullerton, R.A. (2013) ‘The birth of consumer behavior: Motivation research in the 1940s and 1950s’, Journal of Historical Research in Marketing, 5(2), pp. 212-222. https://doi.org/10.1108/17557501311316833.
Gallup (n.d.) ‘Gallup Poll quiz’. Available at: https://news.gallup.com/poll/9970/gallup-poll-quiz.aspx (Accessed: 21 October 2026).
Gallup (2019) ‘Gallup Vault: Little thirst for restoring Prohibition in 1930s’, 28 October. Available at: https://news.gallup.com/vault/267830/gallup-vault-little-thirst-restoring-prohibition-1935.aspx (Accessed: 21 October 2026).
Harry S. Truman Library and Museum (n.d.) ‘American Institute of Public Opinion records’. Available at: https://www.trumanlibrary.gov/library/personal-papers/american-institute-public-opinion-records (Accessed: 21 October 2026).
Miller Center (n.d.) ‘Harry S. Truman: Campaigns and elections’. Reproduction at the Harry S. Truman Library and Museum. Available at: https://www.trumanlibrary.gov/public/1948Campaign_MillerCenter.pdf (Accessed: 21 October 2026).
Mosteller, F., Hyman, H., McCarthy, P.J., Marks, E.S. and Truman, D.B. (1949) The Pre-Election Polls of 1948: Report to the Committee on Analysis of Pre-election Polls and Forecasts. New York: Social Science Research Council. Bulletin 60. Bibliographic record: https://openlibrary.org/books/OL6050843M/.
Neyman, J. (1934) ‘On the two different aspects of the representative method: The method of stratified sampling and the method of purposive selection’, Journal of the Royal Statistical Society, 97(4), pp. 558-625. https://doi.org/10.2307/2342192.
Ohmer, S. (2001) ‘Gallup meets Madison Avenue: Media research and the Depression’, Proceedings of the Conference on Historical Analysis and Research in Marketing, pp. 109-116. Available at: https://ojs.library.carleton.ca/index.php/pcharm/article/view/1729/1559 (Accessed: 21 October 2026).
Ohmer, S. (2006) George Gallup in Hollywood. New York: Columbia University Press. Publisher’s description available at: https://cup.columbia.edu/book/george-gallup-in-hollywood/9780231511285/ (Accessed: 21 October 2026).
Roper Center for Public Opinion Research (n.d.a) ‘George Gallup’. Available at: https://ropercenter.cornell.edu/pioneers-polling/george-gallup (Accessed: 21 October 2026).
Roper Center for Public Opinion Research (n.d.b) ‘Elmo Roper’. Available at: https://ropercenter.cornell.edu/pioneers-polling/elmo-roper (Accessed: 21 October 2026).
Roper Center for Public Opinion Research (n.d.c) ‘Archibald Crossley’. Available at: https://ropercenter.cornell.edu/pioneers-polling/archibald-crossley (Accessed: 21 October 2026).
Roper Center for Public Opinion Research (2020) ‘British Institute of Public Opinion polls from 1938-1946 in Roper iPoll’, 28 January. Available at: https://ropercenter.cornell.edu/blog/british-institute-public-opinion-polls-1938-1946-roper-ipoll (Accessed: 21 October 2026).
Squire, P. (1988) ‘Why the 1936 Literary Digest poll failed’, Public Opinion Quarterly, 52(1), pp. 125-133. https://doi.org/10.1086/269085.
