CHAMPAIGN, Ill. — A study from University of Illinois economics professors demonstrates a new method to analyze the relationships among voters’ issue preferences, the candidates’ policy positions and voter behavior.
Estimating the distribution of voter preferences and the extent of policy divergence between the candidates’ platforms, economics professors Stefan Krasa and Mattias Polborn are able to separate observed changes in voter behavior into those driven by voter radicalization versus those caused by increased policy differences between the two parties.
“We have a long series of data from the American National Election Survey, which is an extremely detailed opinion poll administered by the National Science Foundation over several decades that asks a representative sample of voters for their vote in the presidential election, as well as their opinions on different political issues such as abortion rights, or whether the government should guarantee a certain standard of living to all people,” said Polborn, also a professor of political science at Illinois. “We group these issues into ‘economic’ and ‘cultural’ policies and calculate an economic and a cultural-ideological position for each voter.”
According to the research, if candidate positions are very similar, then voters choose candidates primarily based on non-policy attributes.
“This is true even if voters care a lot about policy – without meaningful policy differences between candidates, voters cannot express the direction or intensity of their policy preferences through the act of voting for one of the candidates,” Krasa said. “By contrast, if candidate positions are very polarized, the election presents a stark choice for voters, and most will vote for the candidate who is closest to their own favorite policy position.”
For example, the model shows that the relationship between a voter’s economic and cultural-ideological leanings and who they ultimately cast their ballot for depends on how far the economic and cultural-ideological positions of the candidates are apart from each other.
“If parties had exactly the same economic position, but differed in their cultural-ideological positions, a voter’s choice to vote for the Democratic or Republican candidate does not depend on her economic issue preferences, because she gets the same economic policy, independent of who wins, while her cultural-ideological views matter a lot because she has a meaningful choice between the candidates on these issues,” Polborn said. “A social liberal would likely vote for the Democrat, whereas a social conservative would likely vote for the Republican.”
By contrast, if the candidates differed on economic policies, but not on cultural-ideological ones, voters would condition their vote choice on their economic policy preferences, while their ideological positions would not matter.
“If parties differ on both economic and cultural issues – and this is generally true in real life – what matters more becomes a relative question,” Krasa said. “For example, if parties are very differentiated economically, and not very much cultural-ideologically, then a voter’s economic preferences are very predictive of his vote choice, while his cultural preferences are only moderately good predictors.”
By applying this logic to voter behavior observed in the data, the model allows the researchers to recover how the difference between the two parties’ positions has changed from 1972 to 2008.
Krasa and Polborn discovered that, relative to 1976 (the year with the smallest policy differences, as perceived by the voters), the distance between the cultural-ideological positions of Democrats and Republicans has approximately quadrupled.
“This increase started with Ronald Reagan’s ‘Conservative Revolution’ in 1980 and the beginning involvement of evangelicals in the Republican Party, and has proceeded very steadily since then,” Polborn said. “Many former conservative Democrats and liberal Republicans have switched parties since then, further reinforcing the initial ideological differentiation.”
According to Polborn, the economic policy positions of the parties also diverged relative to 1976, but to a much smaller extent, and not steadily, but rather sometimes increasing and sometimes decreasing from election to election.
“As a consequence of the relatively smaller change in economic policy differences between the parties, a voter’s cultural-ideological position is much more predictive of his vote choice today, relative to his economic position, than it was in 1976,” he said.
The researchers say that the core of the study’s contribution is methodological.
“The research shows that we can recover both the extent and the direction of policy divergence from the voting behavior of voters with different ideal positions,” Krasa said.
But the study is also of substantive interest because it helps to address some of the central questions in American politics, the researchers say.
“Do voters increasingly split along ideological lines, and if so, what does their behavior tell us about the underlying fundamental causes?” Krasa said. “Are voters today more polarized than they were a generation ago, or do they just appear more polarized because they face more polarizing choices?
“These questions can only be answered with the guidance provided by a structural model framework because one needs to separate the causal effects of changes in voter preferences and changes induced by the political parties.”
The researchers also say that the methods outlined in the paper are applicable to other data sets – namely, the questions of policy divergence and polarization in other countries.
“It would be particularly interesting to analyze whether the developments that we identified for the U.S. in the last generation – policy divergence between parties, and stronger divergence on cultural issues than on economic ones – are also reflected in other countries, and in other voting systems such as proportional representation, or whether the experience in the U.S. is unique in this respect,” Polborn said.
“Evidently, these fundamental questions will require a lot more work, but we hope that the instruments that we have developed in this paper will prove useful in this long-term project.”