Macro Implementation: Testing the Causal Paths from U.S. Macro Policy to Federal Incarceration

Abstract

Policy implementation is usually studied at the micro level by testing the short-term effects of a specific policy on the behavior of government actors and policy outcomes. This study adopts an alternative approach by examining macro implementation-the cumulative effect of aggregate public policies over time. I employ a variety of methodological techniques to test the influence of macro criminal justice policy on new admissions to federal prison via three mediators: case filings by federal prosecutors, conviction rates in federal district courts, and plea bargaining behavior. I find that cumulative Supreme Court rulings influence the incarceration rate by altering conviction rates in district courts; however, I find only mixed evidence of congressional and presidential influence. The results suggest that U.S. macro policy influences bureaucratic outputs by altering the behavior of subordinate policy implementers; however, the Supreme Court may enjoy an advantage in shaping criminal justice policy.

Europe’s internal migration and public support for income redistribution: The role of social protection

“…By using lagged variables of immigration, the results are evidence of Granger causality whereby if variable (X) causes variable (Y), then a previous value of X should predict a subsequent value of Y (Granger, 1969). While Granger-causality tests are not proof of causality it does suggest “a temporal ordering that is consistent with a causal narrative” (Hall, 2017: 8).…”

Our issue or their issue? Media coverage and framing of the Zika virus epidemic

“…Granger causality helps to determine whether this is true in the data. In short, for a relationship to be considered as signifying Granger causality, the independent variable should have a statistically significant relationship with the dependent variable, but the reverse p-value-a test of the reverse relationship with the dependent variable predicting the independent variable-should not indicate a p-value of less than 0.05 (Hall, 2017).…”