Millions of people in the United States interact with the criminal justice system each year. The severity of these interactions varies dramatically ranging from (seemingly) minor misdemeanor convictions to long-term prison sentences. How do these interactions affect political participation? Are more severe punishments the most demobilizing, or is punishment severity outweighed by other factors? How might the mark of a criminal record as opposed to the sentencing effects of a conviction (e.g., incarceration) differ in their effects on voting? These questions underscore a critical policy feedback issue: if interactions with the criminal justice system promote withdrawal from political life, the individuals and communities most affected by these policies will become increasingly unable to represent their interests. Moreover, I contend that studying policy feedback within the context of the criminal justice system enhances our understanding of policy feedback more generally. While policy feedback theory typically emphasizes how policy characteristics shape feedback effects, I use the uniquely sequential nature of criminal justice interactions (e.g., arrest, conviction, incarceration) to show that individuals’ prior experiences with the state can alter their responsiveness to subsequent policies. In this way, my dissertation demonstrates that policy feedback effects are not solely shaped by the design of a particular policy, but also by one's cumulative experiences with the state. The three papers in this dissertation pair causal inference and computational methods with large administrative datasets to evaluate the effects of different interactions with the criminal justice system on political participation. The results demonstrate that more severe punishments like prison sentences have negligible effects on participation while less severe punishments such as misdemeanor convictions have strong, enduring demobilizing effects. Put differently, those with little prior experience with the criminal justice system are far more susceptible to demobilization than those with extensive experience who have likely been previously demobilized through earlier encounters. These findings update policy feedback theory by showing that one's prior interactions with the state, not just policy characteristics, are key determinants of the potency of feedback effects. In the first paper, I use a regression discontinuity design based on felony sentencing guidelines in North Carolina to study the effect of prison sentences on voter turnout. I find negligible effects of imprisonment on turnout. I argue this result is explained by a “saturation” mechanism where individuals facing the most severe criminal justice punishments are unlikely to be demobilized by these interactions because they have been demobilized through extensive prior interactions with the criminal-legal system. In the second paper, I use the as-if random assignment of judges to misdemeanor cases in Tampa Bay, Florida to estimate the effect of misdemeanor convictions relative to deferred adjudication on turnout (a finding of guilt without a conviction). I find that convictions lead to significantly lower turnout. This effect does not dissipate over time and is driven by defendants with no prior convictions suggesting that the results are driven by the mark of a conviction. I also show that misdemeanor convictions also increase the likelihood of future felony convictions so demobilization from misdemeanors may be particularly strong in states with strict criminal disenfranchisement laws. These results point to misdemeanor convictions as pervasive and lasting barriers to voter participation among justice-involved citizens. In the third paper, I use a similar regression discontinuity design based on custody classification score cutoffs in Pennsylvania prisons to estimate the effect of incarceration severity on turnout. I find qualified evidence that assignment to stricter custody levels (medium versus minimum) reduces turnout, but this effect is small and inconsistent across model specifications. Collectively, this dissertation demonstrates that punishment severity is a poor predictor of demobilization. Rather, a relatively weak punishment—a misdemeanor conviction for first-time defendants—has the largest, most persistent demobilizing effect reported in existing research on criminal justice interactions and turnout. This makes an important contribution to policy feedback theory by demonstrating a need to look beyond policy design and also consider how one's cumulative experiences with the state shape feedback effects. My findings also provide systematic evidence of massive turnout gaps between those with criminal involvement and the general public and point to the mark of a misdemeanor criminal record as a key contributor to this gap. At the same time, these findings show promising evidence that front-end policies designed to limit entanglement with the criminal justice system have considerable potential to shrink this participatory deficit.