This thesis contributes to the study of the role of information in elections and public policy formation. Its main focus is on information acquisition and voting behaviour. Chapter 1 discusses the motivation of this research and presents a survey of related literature. Chapter 2 focuses on electoral turnout, Chapter 3 on public policy, and Chapter 4 on mass media. Chapter 2 studies the impact of information on electoral turnout. Since incentives to be informed are correlated with other incentives to participate in public life, a model of information acquisition and turnout is introduced to isolate potential instrumental variables and try to establish a causal relation. Results are tested on the 1997 General Election in Britain. It is shown that information, as well as ideology, matters for turnout. It also contributes to explain the systematic correlation of turnout with variables like education and income. Voters' knowledge of candidates and of other political issues is also substantially influenced by mass media. Chapter 3 presents a model that links the distribution of political knowledge with redistributive policies. It argues that voters can have private incentives to be informed about politics and that such incentives are correlated with income. Therefore redistribution will be systematically lower than what the median voter theorem predicts. Moreover, more inequality does not necessarily lead to an increase in redistribution and constitutional restrictions might have unintended consequences. In Chapter 4 it is argued that instrumentally motivated voters should increase their demand for information when elections are close. In supplying news, mass media should take into account information demand, as well as the value of customers to advertisers and the cost of reaching marginal readers. Information supply should therefore be larger in electoral constituencies where the contest is expected to be closer, the population is on average more valuable for advertisers, and the population density is higher. These conclusions are then tested with good results on data from the 1997 General Election in Britain.