The main thesis in this book is that social choice theorists, and especially Bill Riker and his Rochester “school,” have cast a “long dark shadow over democratic politics” (the title of Chapter 1). Gerry Mackie then proceeds to dispel this dark shadow by an excessively argumentative interpretation of an impressive number of intellectual accomplishments in the study of social choice, public choice, and what is called in political science “formal theory.” His main attack is directed against the work of Riker, especially that author's work on disequilibrium in majority rule.

Democracy Defended contains an impressive survey of social choice theory. It is worth reading by someone familiar with the basics of this theory, but I find much of the book tedious. But one basic matter of preference orderings that Mackie presents us with is his personal view about the political and economic world. His world seems to be shaped by his upbringing in a small Oregon town and his experience as “founder and elected leader of a large forestry worker's cooperative movement” and his related political activities (p. 3). In Mackie's world, “if the question is about the public interest, then individual preferences are naturally sufficiently similar to one another's to avoid cycling most of the time. And if the question is fixed sum redistribution, then destructive self-seeking preference should be excluded from public consideration (as taught in kindergarten)” (p. 94). The first sentence of this quotation is a statement about political behavior that is not supported with any evidence, even if we can agree about a definition of “most.” The second sentence is obviously normative. It does not apply to the election and voting behavior of the Executive Committee of the Department of Government at the University of Texas at Austin where the Exec Com determines salary increases and other allocations by voting.

As Mackie points out, social choice theory is about aggregation of preferences. The work stemming from Arrow's Theorem deals with rank order preference orderings. Most of this work is about aggregation and not voting, but the applications to voting involve very simple voting schemes that are used in committee choice. The preferences of members of a voting committee are often influenced by arguments made as the committee deliberates, but sometimes the preferences are determined by pure self-interest. The study of how individual preferences are determined is a complex and difficult problem. I am not convinced about the value of studies of group choice by college students facing group decisions about small amounts of money.

The main theoretical paradigm for the study of mass politics is the spatial theory of electoral competition. The best-known result in spatial theory is the Median Voter Theorem. This theorem fails to hold if the votes are a simple probabilistic function of distances to the choices, if there are more than two choices under consideration, and if there is abstention due to alienation or indifference.

The author falls into the trap about cycles set by the early social choice theorists, who applied their results to simple committee voting. Although I have encountered what appeared to be cycles in academic committee meetings, the games played within and among legislative committees is complex. There is usually one committee that controls the bills that are voted on. In some cases, a single person, the speaker of the house, determined the agenda. The agendas are not the simple paired comparisons that I use in class to show how strategic voting is the sensible game play. The typical legislative agendas are made by the introduction of alternative bills and amendments to bills. Since information is valuable in a game, the players will attempt to disguise or hide their preferences. If there is a well-developed political space for alternative bills and legislators' preferences for bills, then the most preferred position of legislators will probably be known. Once again this is an empirical issue. The trade-offs implied by the shape of individual preference functions will not be known. Since all legislators know that they have to pass a budget and deal with a number of issues in a finite time using the structure that they agree to use as part of the game, the choices they face are limited, and the political game is played out behind closed doors and rarely in an open vote.

Mackie dismisses vote trading and other forms of logrolling as unstable due to the lack of knowledge of the preferences of others. The knowledge set that the vote organizers possess is very important to the stability of a coalition that has to predict future behavior. The assumption of common knowledge (all players know the probabilities of all outcomes in the game) that Austen Smith and other such game theorists use is crucial for the result quoted at the bottom of page 162. Common knowledge is unlikely since probabilities are estimated and people have different priors even if they have the same information. The assumption of common knowledge in game theory has not been tested by experiment or empirical analysis of game behavior. I believe that the reality of most political games is that the play is strategic but with lots of errors and confusions.

Turning to mass politics, elections are infrequent. We have presidential elections in this country every four years and congressional elections every two years. Since politicians have to raise money for their campaigns, they have to convince contributors that they will support their vital interests. Many of the contributors to campaigns are business and labor political interest groups. Business abhors political instability since such instability jeopardizes investments. Contributors are concerned about what an election outcome will imply for their interests. Thus, the most likely outcome of an election is a change in the leadership in Congress and in the Executive that is unlikely to result in large changes in public policies. This is also true for most proportional representative democracies, but there are exceptions.

The AK Party won a controlling majority in the Turkish parliament as a result of the 2002 election and several special elections held afterward. The present government is attempting to weaken the power of the army and to move the country toward an Islamic direction on the main axis of what Professor Ali Çarkoglu and I have found to be a two-dimensional political space, using respondents' political attitudes from two face-to-face national surveys of Turkey in 2001 and 2002. We have shown that the voters have not shifted on the religious dimension but that the AK Party won because it was able to convince many of the moderates that they were the most credible party in the election to bring Turkey out of its depression. Turkey has a 10% threshold for election into their parliament. If the AK Party moves to change the Turkish constitution to make the country an Islamic country, there may be a civil war. Almost all Turks are Sunni Moslems and most are moderates, but there are serious differences of opinions about the role of religion and nationalism in political and social life in Turkey. Turkey's democracy is more typical then the idyllic model of Mackie's Oregon town.

Although Mackie presents some of the two-dimensional spatial voting results, he relies heavily on the unidimensional empirical results of Keith T. Poole and Howard Rosenthal based on legislative roll-call data. The information content in roll calls is limited, and the data and methods cannot statistically test the dimensionality of the political space. My collaborative work on attitudinal data from U.S., German, and Chilean surveys indicates that the political spaces of those countries are two-dimensional with a third valence dimension for candidate or party competence.

The final third of this book is a rebuttal to Riker's historical analysis where he claimed to have found voting cycles. Since I did not buy Riker's arguments when he published them, I basically agree with Mackie's analysis. I felt that Riker did not provide enough evidence about the preference orderings of the players in those voting games.

I have some special criticisms of this book. The material on pages 25 and 26 are inappropriate. There is no reason to mention that poorly researched polemical article in the New Republic. Who cares what goes on in Harvard's Government Department? Also, who cares about what economics students in Paris want or do not want? These pages pander to the haters of mathematics and the use of logic in political analysis. The term “Starship Rochester” is also juvenile. As a matter of fact, Bill Riker was doing some experiments in political game theory when Otto Davis and I as graduate students introduced the first multidimensional spatial theory of elections in 1965. The history of the development of formal theory and public choice is more interesting than the superficial treatment presented in this book.