research articles

MELISSA J. MARSCHALL
Rice University

ANIRUDH V.S. RUHIL
Ohio University

Of Models and Methods: A Response to Matsusaka

Do instruments of direct democracy affect policymaking and, if so, how? The political science literature is rife with increasingly sophisticated empirical efforts to answer these questions (Matsusaka 2004; Lupia and Matsusaka 2004). Having expended much energy over the past two decades studying the initiative’s effects on state and local policy, Matsusaka (1995, 2004) is convinced that initiative states spend and tax less than states without the initiative (2004, 3; 1995). Agnostic about whether initiative states spend and tax more or less than noninitiative states but puzzled by scholars’ failure to account for the endogeneity of the initiative in their models, we sought to determine whether the fiscal conservatism ascribed to the initiative remained after explicitly modeling states’ initiative status. Our study, published in this issue of State Politics and Policy Quarterly contradicts Matsusaka’s conclusions. In commenting on our study, Matsusaka (2005) argues that our results are fragile and can be rejected on both theoretical and empirical grounds. To support this argument, he offers an alternative model that both tweaks our fundamental specification and introduces a quaint variable that ostensibly brings the estimation of initiative effects in line with the “rest of the literature” (2005, 356). In this rejoinder, we demonstrate that neither of these claims is true. First, we show that the empirical evidence on the fiscal conservatism effect of the initiative in the scholarly literature is far from unanimous. Second, we demonstrate that our key finding regarding the endogeneity of the initiative holds up under alternative model specifications, including the one Matsusaka (2005) proposes in his Model 2.1 Regarding the alternative specifications Matsusaka reports in his comment, despite considerable effort and a good deal of creativity, we were unable to reproduce his Model 2 findings, especially the negative coefficient on the initiative dummy. As for Model 3, by specifying the initiative as a function of initiative status in 1920, Matsusaka has, for all intents and purposes, rendered the question of endogeneity moot. Given the lack of variation in initiative status between the two time points, initiative status in 1920 is a nearly perfect predictor of current initiative status. This result not only renders the assignment equation in his Model 3 superfluous, but it also contributes nothing to our understanding of what explains initiative status.2