I'd model a criminal conviction as indicative of one of four possibilities: 1) Defendant is guilty and the police did their job right, bringing strong evidenc

I'd model a criminal conviction as indicative of one of four possibilities: 1) Defendant is guilty and the police did their job right, bringing strong evidence to the prosecutor. 2) Defendant is innocent, but there's a weight of misinterpreted evidence that his defense fails to overcome. 3) Defense is exceptionally shoddy in the face of an average-to-weak prosecutorial case. 4) Corruption. To disentangle the effects of gains in (1) & (4) on increases in the conviction rate, you need a metric of police efficiency that's unaffected by corruption. For instance, a country with DNA or fingerprint forensics will have a higher conviction rate than one without DNA/fingerprints, but as a result of solid police work, not as a result of corruption.