Why Roll-Call Votes Don’t Always Reveal What Lawmakers Believe

In a new paper published in the journal Political Analysis, Anthony Fowler, the Sydney A. Stein, Jr. Professor at Harris Public Policy and his coauthor Jeffrey B. Lewis introduce a new way of analyzing congressional voting that accounts for moments when lawmakers vote against bills they actually prefer.

Their study, “Accounting for Protest Voting in the U.S. Congress,” examines roll-call votes in the U.S. House from 1889 to 2022 and shows how these non-ideological votes (what the authors call “protest voting”) can affect how researchers measure the ideological positions of members of Congress.

For decades, scholars have estimated the ideology of members of Congress by analyzing roll-call votes. These models generally assume that legislators vote for the option they prefer, supporting a bill if they like it more than the status quo. In practice, however, that assumption does not always hold.

Members of the majority party sometimes vote against legislation that they would otherwise support. According to Fowler and Lewis, these small number of votes (approximately 1%-3%, and only within the majority party) may reflect frustration that a bill does not go far enough, dissatisfaction with party leadership, or other strategic considerations unrelated to ideology.

“We often assume that votes in Congress directly reveal ideology,” said Fowler. “But sometimes legislators vote ‘no’ as a way of signaling dissatisfaction or sending a different message. Our goal was to account for those moments so we can better measure what legislators actually believe.”

To address this issue, Fowler and Lewis developed a new statistical model that allows for the possibility of protest voting. They then used the model to estimate the ideological positions of members of the U.S. House across more than a century of congressional history.

The results suggest that failing to account for protest voting can lead to misleading conclusions about where legislators fall on the ideological spectrum.

One striking example involves U.S. Representative Alexandria Ocasio-Cortez (D-New York) and other progressive lawmakers sometimes referred to as “the Squad.” Traditional roll-call models identify them as relatively moderate Democrats, a result that appears inconsistent with their public rhetoric and policy positions. Standard roll-call models rank Squad members between the 56th and 187th most liberal members of the House. When the researchers account for protest voting, however, these lawmakers emerge as some of the most liberal members of Congress (six of the seven most liberal members of Congress), much closer to what observers would expect. This likely reflects voting against policies that they believe do not go far enough to deal with policy challenges facing them.

The adjusted measures also line up more closely with other indicators of ideology that do not rely on roll-call votes.

The study also implies that the trajectory toward modern polarization may have begun earlier than previously recognized, even if the change was gradual and less visible in conventional measures. When Fowler and Lewis account for protest voting, they find that ideological polarization in Congress is even greater than previous estimates suggested. Their results also indicate that polarization began rising earlier than many scholars previously thought, well before the period often described as the “textbook Congress” of the mid-20th century (roughly the 1940s to the 1970s).

At the same time, the research highlights an important distinction: polarization does not necessarily mean strong party discipline. Highly ideological lawmakers (especially in recent decades) may be more likely to cast protest votes against their own party’s proposals. This means Congress can appear somewhat less polarized in traditional voting models even when ideological differences between parties are quite large.

The researchers also examined how protest voting affects other questions commonly studied in political science.

Using their adjusted measures of ideology, they find that voters appear to impose a stronger electoral penalty on ideological extremism than previously estimated, which relates to Professor Fowler’s larger body of work on the power of moderates and the median voter in electoral politics. The study finds no evidence that protest voting itself is rewarded at the ballot box.

In fact, conditional on ideology, members who engage in more protest voting tend to raise less campaign money, the paper finds.

Although protest voting rarely determines the outcome of a bill, the authors show that even a small number of these votes can influence how scholars interpret congressional behavior. Their findings suggest that researchers studying ideology, elections, and representation may benefit from accounting for protest voting when analyzing legislative data.

“Roll-call votes remain an incredibly valuable source of information about congressional behavior,” Fowler said. “But our findings show that accounting for a small number of non-ideological votes can give us a clearer picture of where legislators actually stand.”