In short: legislatures aren’t always stabilizing. They can also empower rivals and increase coup risk — a double-edged sword for dictators.

The results are striking: about 45% of regimes coded as having legislatures actually fall below the sample mean on my measure. When I re-analyze canonical studies, their results flip — higher legislative power-sharing often predicts shorter leader survival.

I built one of the first continuous measures of legislative power-sharing, covering all autocracies from 1946–2023. Using a dynamic Bayesian latent variable model, I estimate how much real authority rulers delegate to their legislatures.

If legislatures help stabilize regimes by distributing power, we need to measure that capacity directly. Yet existing datasets either use a yes/no indicator or fold legislatures into broad indices of authoritarian rule — losing sight of the mechanism itself.

This paper began with an observation I had a few years ago: scholars argue that legislatures help dictators stay in power by sharing authority with rival elites—but most studies test this with a binary variable, assuming that the mere presence of a legislature equals power-sharing.