We study how voters allocation attention in a multi-issue election. In our model, two parties with di§erent positions on two issues com- pete for the votes of a Önite number voters. The party receiving the most votes is elected, and each voter receives a payo§ equal to the sum of the winning partyís policy payo§s on the two issues. We assume that the issues di§er in terms of importance. SpeciÖcally, one issue is more important than the other in that the payo§ di§erence between the better and worse party for that issue is larger than the di§erence for the other issue. Hence, voters prefer the better party on the more important issue even when it is inferior on the other issue.

Before casting her vote, each voter can choose to acquire a free and independent signal about one of the issues. The signal for the important issue is less informative than the one for the other issue.

We Önd that votersí information acquisition decisions are con- nected strategically through a negative cross-issue inference e§ect. Ex ante, the issues are uncorrelated, and each party is equally likely to be better on each issue. However, conditional on the vote being tied, it is likely that voters following di§erent issues are supporting di§er- ent parties. Hence, when a voter learns that one party is likely to be better on one issue, she will infer that, when she is pivotal, the other party is likely to be better on the other issue. The negative inference will lower the value of both signals. But importantly, it diminishes

1

the value of the signal about the less important issue to a greater ex- tent, and the e§ect is stronger when voters are closer to equally spilt between the two issues.

As a result of this endogenous cross-issue negative correlation, there may be multiple equilibria. In one equilibrium, a majority of the voters will focus on the important issue and equilibrium outcome is asymptotically e¢ cient. In another equilibrium, all voters focus on the less important issue. This equilibrium is ine¢ cient even asymp- totically because the party better on the less important issue almost surely wins when the number of voters becomes large. However, with probability 1/2, this candidate is worse on the important issue. Thus, our results show that in a democracy, voters may sometimes focus on more concrete concerns while ignoring issues that may hold greater long-term importance.