What were the shortcomings of Literary Digest’s 1936 presidential poll? How does modern polling correct these problems?

The shortcomings of the Literary Digest's 1936 presidential poll were numerous and significant. Here is a step-by-step explanation of the key issues with this famous poll:

1. Non-Representative Sample: The Literary Digest poll drew its sample from telephone directories, club membership lists, and magazine subscription lists. At the time, not every American household had a telephone or subscribed to magazines, and such amenities were more common among the affluent. This method excluded large segments of the population who were less wealthy and had different political views.

2. Voluntary Response: The poll relied on voluntary response, meaning that individuals chose to send in their ballots. This non-random method of data collection increased the likelihood of response bias because people with strong opinions, particularly those confident in their preferred candidate, were more likely to participate.

3. Overconfidence in Large Sample Size: The pollsters believed that the sheer size of their sample (over 2 million responses) would guarantee accuracy. However, as the sample was not representative of the entire voting population, the size did little to enhance the poll's predictive validity.

4. Failing to Adjust for Bias: The poll did not include any mechanisms to adjust for known biases in its sample, meaning that the results reflected the views of a skewed subset of voters, not the electorate as a whole.

5. Timing: The timing of the poll was also a crucial factor. The Digest sent out its ballots several months before the election, not accounting for any changes in public opinion closer to the election day.

6. Lack of Privacy: Respondents had to mail back their ballots, which were not anonymous. This lack of privacy could have further increased response bias as individuals may have been influenced by the expectation of their choices being publicly associated with them.

These errors led the Literary Digest to incorrectly predict that Alf Landon would win the 1936 presidential election over Franklin D. Roosevelt, which turned out to be one of the most dramatic failures in the history of polling.

Modern polling has corrected these problems through a variety of methodological advancements:

a. Random Sampling: Modern polls use random sampling techniques to ensure that every member of the population has an equal chance of being selected. Random-digit dialing for phone surveys and carefully constructed online panels are examples of this practice.

b. Weighting: Polls now use weighting to adjust their samples to reflect the demographic composition of the population, taking into account factors like age, gender, education, race, and geographical distribution.

c. Privacy: Most modern polls ensure anonymity, which helps to reduce bias as respondents are more likely to answer questions truthfully.

d. Timing: Modern polls are conducted closer to the event or election they are predicting, and tracking polls may even monitor changes in opinion over time.

e. Margin of Error and Confidence Levels: Modern polls report a margin of error and a confidence level, which provide a way of expressing the poll's precision and the probability that the results are within a certain range of the true values in the full population.