John Carey opens his book by describing different models of legislative accountability. As agents, legislators need to be accountable to their principals for their actions, particularly for the way they vote in the legislatures. But who are the principals? Older accounts of representative democracy would suggest the principals are the legislators' constituents. In modern political science, accountability is thought to flow from voters through parties: hence the immediate principal of a legislator would be the party itself reified by the party leadership. In recent years, specifically in many Latin American countries where much of the empirics of this book are set, there have been attempts to shift the line of accountability towards constituents through various institutional means. Carey's book considers how we might measure the success of these strategies and evaluates the extent of their success.

Carey's empirical analysis begins with a survey of legislators that asks them the relative importance of their party and constituents when it comes to making decisions. Of course we can only place a certain amount of trust in the answers received through such surveys, but the relative emphasis placed by legislators in different Latin American countries is interesting. Clearly, direct accountability to constituents can only matter if those constituents know what their representatives are doing. In the U.S. voters can easily discern the voting records of their representatives, as they are on record and interest groups publicize them. Not so in many Latin American countries. Where records are kept they are not always published; where they are notionally available it can be extremely difficult to obtain them. In a lovely appendix to Chapter 3, Carey describes his failed attempts to get hold of records in Columbia despite being told repeatedly that the records were public. He also describes how electronic voting machines were installed in various countries yet legislators do not use them, or where they are used they do not record votes for future inspection. Low-tech monitoring by party leaders, however, ensures they have a better idea of how their members are voting. Here it would seem clear that legislators are accountable to their party leaders.

The second half of the book is an attempt to measure the relative accountability of legislators in countries where voting is visible. Carey uses several measures or indices of ‘voting unity’ within parties. Such indices may not get at how responsive legislators are to their constituents, but they can give an idea of the relative accountability of parties: if legislators' voting patterns demonstrate no unity, then party label carries no informational value. Carey concentrates on three indices: how consistently party members vote the same way; how often they win; and how often losses might have been avoided had greater unity been shown. The latter measure is important, since rebellion is cheap if it is known it will not affect the result. RICE (a measure developed by Stuart Rice in 1925) is a measure of cross-voting by party members varying from 0 (equal number vote aye and nay) to 1 (they all vote the same way). Since RICE does not count abstentions, Carey proposes his new measure UNITY where the cross-voting proportions as shares of all members of the relevant party in the legislature. Thus UNITY also varies between 0 and 1 though obviously dips more easily as it takes account of abstentions. By combining these measure Carey creates RICE and UNITY indices for a number of countries and then compares these in a series of graphs plus tables of averages and standard deviations. We see some substantial differences across countries, but also across time within countries and across parties within a country.

What do we learn about institutional design from Carey's careful empirics? He addresses this question in Chapter 6, entitled “Explaining Voter Unity.” There are a number of hypotheses in the literature that attempt to explain party discipline. Parliamentary systems are supposed to have more disciplined parties than presidential ones since the electoral link is clearer in the former; federalism is supposed to create divisions within parties, fostering less discipline; whilst electoral systems that foster competition within parties should create less unity as members vie for the votes of constituents. Of course these generalizations do not always hold; some presidential systems look more like parliamentary ones and political culture, the last explanatory refuge of the institutionalist scoundrel, can account for some divergencies from true form. Drawing a set of hypotheses from the literature, Carey examines them using his data. He finds that electoral rules matter; federalism seems to have no effect; and a popularly elected president does lead to greater disunity for governing parties in parliament but has no effect on opposition parties. It seems presidents create a new principal, partially eclipsing parliamentary leadership.

This book provides an important addition to legislative studies, furnishing new arguments and data at the legislator level to show how parties and legislatures are affected by institutional differences. In that regard it is an important book. Yet, despite a final chapter that addresses the question, I have doubts about the link between the early discussion of lines of accountability and the empirics displayed here. Degree of unity might measure how well the line of party accountability works, but on its own cannot measure direct accountability to constituents; lack of unity might come from many different sources. To examine the constituent legislator link we need more information on ideological elements within parties, and the issues on which we would expect legislators to take more account of their constituents' views. Such information is absent from this book. But all books have their limits. And Legislative Voting and Accountability provides both new data and important insights into the links between legislators, parties, and representation.