Roush and Sood (2023) use a dataset of 162,083 responses to 187 items on 47 surveys to find that partisan gaps are smaller and less frequent than commonly understood. The average is a mere six and a half points and gaps' ''signs'' run counter to expectations roughly 30% of the time. However, one exception is the size of gaps in retrospection items on the ANES, which are considerably bigger. These retrospection items use vague response options, e.g., 'Got better', 'Remained about the same,' etc. Vague response options can inflate partisan gaps by offering partisans the opportunity to interpret the same data differently. For instance, a .2% improvement in unemployment under Democrats can be reasonably interpreted as 'unemployment remained about the same' or as 'unemployment got better.' (The 'reasonableness' comes from the implied variance in the time series, etc. though the point stands independent of the reasonableness standard.) In an experiment, we test whether these vague options inflate partisan gaps. We present partisans data indicating a small improvement in economic indicators and manipulate the partisan tint of the change by manipulating who is responsible for the change.
-
During 2016, (when Barack Obama was president | when Republicans were in control of both Houses of Congress), unemployment decreased from 5.0% to 4.8%, a change of 0.2 percentage points. How would you interpret this change? Would you say that unemployment got better, stayed about the same, or got worse?
-
In 2016, inflation also decreased from 2.1% to 1.9%, a change of 0.2 percentage points. How would you interpret this change? Would you say that inflation got better, stayed about the same, or got worse?
We find that significantly fewer partisans pick the option that '[things] got better' when presented with an out-partisan cue than a co-partisan cue. Our findings suggest that vague options can induce knowledge gaps even when partisans have the same information.
Carrie Roush and Gaurav Sood