The RAND 2016 Presidential Election Panel Survey (PEPS)

The RAND Presidential Election Panel Survey (PEPS), a subsample of the RAND American Life Panel (ALP), offers a unique perspective on voter attitudes and intentions as we near the 2016 presidential election.

The RAND Corporation kicked off the PEPS with an initial survey covering late December 2015 and early January 2016, and will continue surveying through the months leading up to the presidential election in November. The PEPS is designed to examine voter attitudes, intentions, and choices, and how these change throughout the 2016 presidential election cycle using a representative sample of around 3,000 U.S. citizens over the age of 18. The PEPS will ask participants not only about their voting intentions and attitudes toward and preferences for potential candidates, but also opinions on a wide range of topics such as political and other issues in the news, and detailed information about the respondents themselves. RAND developed the survey instrument in collaboration with political scientists John Sides of George Washington University, Lynn Vavreck of UCLA and Michael Tesler of UC Irvine.

Why This Poll Is Different

First, it allows us to ask the same people for their opinion repeatedly over time. In comparison to most polls, this leads to much more stable outcomes; changes that we see are true changes in people's opinions and not the result of random fluctuations in who gets asked the questions.

Second, we may be more accurately capturing the likely votes of a greater number of voters in the crucial "middle" (i.e., not closely aligned with either candidate) by allowing respondents to more precisely assign their own numerical probability (or percent chance) to both the likelihood that they will vote and the likelihood that they will vote for a particular candidate. By comparison, traditional polls may not be fully capturing the intentions of these voters because they rely on less precise qualitative metrics (such as somewhat likely and somewhat unlikely) when asking respondents to indicate for whom they may vote and the likelihood that they will vote.

How the Poll Works

We have recruited 3,000 ALP respondents to our baseline survey, and will use this same sample throughout subsequent surveys. Because of the importance of the baseline data for assessing individual-level change over time, we will not include new respondents in subsequent waves. Our sample is limited to U.S. citizens, as we are examining presidential election voting behavior.

This baseline sample will be invited to take at least five additional surveys at key points throughout the 2016 election cycle, as shown in Figure 1.

Figure 1. Key Election Dates and Timing of PEPS Survey
Key Election Dates and Timing of PEPS Survey

For Waves 1-3, the baseline sample will be divided into three random subsamples of 1,000 members who will be invited to participate in a short survey of 10 or fewer questions on a rolling basis - for each wave, one-third of the sample will be invited during the first week, one-third will be invited during the second week, and one-third will be invited during the third week. Each survey will remain in the field for two weeks for each subsample, in effect creating a continuous poll over the course of one month. Each subsample of 1,000 is sufficiently large to power analysis independently, as well as when combined. This strategy has the added benefit of potentially capturing the effect of significant unexpected events on voter opinions. Wave 4 will be fielded to the entire baseline sample simultaneously because of the relatively short time between the final debate and election day. The post-election survey will be fielded immediately after the election.

We will assess voting intentions during the first four waves, in addition to the baseline.

Voting intentions are assessed in the baseline survey with two questions. First, respondents were asked:

Q1. We'd like you to ask you to think about the upcoming Presidential election in 2016.

What is the percent chance that you will vote in the Presidential election? _____%

The percent chance can be thought of as the number of chances out of 100. You can use any number between 0 and 100. For example, numbers like 2 and 5 percent may be 'almost no chance', 20 percent or so may mean 'not much chance', a 45 or 55 percent chance may be a 'pretty even chance', 80 percent or so may mean a 'very good chance', and 95 or 98 percent chance may be 'almost certain'.

Q2. If you do vote in the election, what is the percent chance that you will vote for a Democrat? And for a Republican? And for someone else? Please provide percent chances in the table below.

Democrat _____%
Republican _____%
Someone else _____%
Total _____%

The order of the Democrat and Republican rows were randomized. Totals are automatically summed as people type in their responses. If responses do not sum to 100, an error message is shown that says: "Your total does not add up to 100%, please return to the previous question and fix your answers." Only responses between zero and 100 will be allowed, to avoid problems faced in the previous presidential election poll, where respondents gave inappropriate responses such as 1100% and -1000%. In subsequent waves where the party nominees are known, Democrat and Republican will be replaced with their names instead.

Weighting for Reported Probabilities

By asking respondents for probabilities (i.e., "What is the percent chance..."), we acknowledge the fact that respondents may not be 100-percent sure who they will vote for or whether they will even vote. We use both of these probabilities to weight their responses in the final poll data. This approach was developed by Adeline Delavande (University of Essex, United Kingdom) and Charles Manski (Northwestern University).

Other Weighting

As with nearly all survey data, we weight our sample to best match the population of interest (including factors such as sex, age category, race-ethnicity, education, household size, and family income). Based on the premise that the best predictor of future voting behavior is past voting behavior, we also reweight each daily sample separately such that its voting behavior in 2012 matches known population voting behavior in 2012.

Full description of our methodological approach, including specific details on our weighting techniques »

For the latest data analysis and updates based on the PEPS, see:

For more information, contact Michael Pollard, a sociologist at RAND who is running the PEPS, at

RAND ALP and MMIC are proprietary capabilities of the RAND Corporation. They were developed by RAND with its own funds and with the support of numerous clients and grantors who have commissioned social-science and economics research and analysis at RAND. Use of the RAND ALP and MMIC is available to outside researchers for a fee. The use of RAND ALP and MMIC for the continuous presidential election poll was initiated and financed by RAND in order to demonstrate their distinctive capabilities to a broader audience. All RAND research products are subject to rigorous peer review to ensure that they meet high standards for research quality and objectivity. The RAND Labor and Population review process employs anonymous peer reviewers, including at least one reviewer who is external to RAND. The methods and findings presented here benefited from the technical reviews of two reviewers.