Statehouse Democracy is a very important book for anyone interested in connections between public opinion and policy. It actually fulfills the oft-trumpeted hope of using American states as a research "laboratory" for studying questions of broad theoretical interest. Indeed, Erikson, Wright, and McIver have produced the most full and convincing empirical findings to date concerning democratic representation.

The book is not a breezy read; bringing together, updating, and extending material from a series of journal articles, it presents a mass of tables and graphs, including abundant regression analyses that range from the straightforward to the subtle and intricate. But the effort required to work through the data analysis is richly rewarded. And the writing is clear. The shaded maps are illuminating. The many scatter plots, in which each state observation is labeled by name, nicely illustrate the strength of relationships and identify outliers (e.g., heavily Mormon Utah, whose policies are much more liberal than citizens' conservative ideology would predict).

The heart of the book is chapter 4, where the mean liberal-conservative self-identification of citizens from each state (cleverly measured by the state-level aggregation of some 142,000 responses to CBS/New York Times random-digit-dialing telephone surveys conducted between 1976 and 1988) is related to the liberalism or conservatism of state policy (measured, ca. 1980, as a composite of eight issues -- education spending, scope of Medicaid, scope of Aid to Families with Dependent Children, consumer protection, criminal justice, legalized gambling, Equal Rights Amendment ratification, and tax progressivity). Across states, the bivariate correlation between public opinion and policy is a hefty .82, which rises, when measurement error is corrected, to a remarkable .91 .

The strong claim that "public opinion is the dominant influence on policy making in the American states " (p. 244) is further supported by the fact that controls for state socioeconomic variables attenuate the opinion-policy relationship only a little. The much heralded impact of economic development upon state policies appears to work almost entirely through citizens' attitudes. Moreover, a two-stage least squares analysis (using religious fundamentalism as a key instrumental variable, postulated to affect citizen ideology but not policy directly) indicates that causation is not reciprocal: opinion genuinely affects policy rather than vice versa.

To be sure, one can offer various quibbles, some of them noted by the authors. Since the domain of opinion and policy studied is restricted to generalized liberalism-conservatism, we cannot be sure to what extent specific policy preferences account for specific policies. Nor is the bear of causal inference totally tamed. Spurious opinionpolicy relationships could result from such factors as fundamentalist religious organizations (outside the model) affecting both policy and opinion. And it is plausible that, over the years, selective migration (also outside the model) may help the liberalism-conservatism of citizens come into harmony with the policy of the states they live in, so that policy reciprocally affects opinion. (Who moves to Texas? Who stays?) The possibility of manipulation of opinion by officials or elites is not really ruled out, even setting aside issues like foreign affairs that fall altogether outside the state politics context.

Still, this research is methodologically much more solid than the old representation-in-Congress studies, with their dyadic focus and their poor measurement of district opinions. It outdoes most national-level, over-time studies that are subject to causal inference nightmares of spuriousness and/or reciprocal causation. This is state-of-the-art research. It has profound (though not extensively discussed) implications for democratic theory. Even in the American states-arguably the least representative of our political institutions-citizens' preferences apparently have powerful effects upon policy.

Statehouse Democracy goes a long way toward illuminating the mechanisms or processes by which opinion affects policy. State-level measures of citizens' party identifications and the liberal-conservative ideology of state officials and activists permit a thorough examination of "responsible party" versus Downsian processes. Downs generally comes out ahead. For example, states with Democratic legislatures do not particularly tend to enact liberal policies, not just because of the anomaly of the South but because voters elect more Democrats where party elites are centrist rather than liberal, and, additionally, because Democratic legislators trim away from their liberalism in making policy.

The book is also full of intriguing findings concerning what influences mass ideology (standard demographics, religious fundamentalism, and state "culture"), variations among Elazar's state types ("traditionalist" southern states tend to resist public opinion, "individualistic" states follow Downsian processes, and "moralistic" northeastern states resemble responsible party systems), stability and change in state ideology and representativeness, and voting in state elections. Even scholars who have memorized the articles should buy this book. It is a landmark.