John Matsusaka has posted this draft on SSRN. Here is the abstract:

Direct democracy backsliding occurs when a state alters its laws to hinder the use of initiatives and referendums. This study quantifies the prevalence of direct democracy backsliding over the period 1960-2022, and investigates its causes. I find a continuous chipping away at direct democracy throughout the period; legislatures proposed 2.3 amendments restricting direct democracy every two-year electoral cycle on average, and there were four amendments restricting direct democracy for one expanding it. Contrary to media speculation, the amount of such activity is not unusually high recently. Using time series and cross-sectional variation with state and year fixed effects, I identify two factors that triggered anti-direct democracy proposals: Republican control of the state legislature, and successful initiatives and referendums. I develop a theoretical framework to characterize the motives for direct democratic backsliding, and provide evidence suggesting that strategic motives – restricting direct democracy to induce favored policy outcomes – may be less important than philosophical preferences over processes, including objections to citizen lawmaking in principle. Anti-direct-democracy efforts were largely driven by elites – ordinary voters did not appear to share the negative orientation of their elected representatives.