Most studies have focused on the effectiveness of training programmes in impacting surrogate outcomes such as the knowledge and attitudes of participants after compared to before attendance at the said programme. Study designs are generally weak, as is usually the case in education research in general, I suppose.
I am aware of a single study conducted in Nigeria that attempted to answer the question you posed, but only secondarily. In their amazing study, Okonta and Rossouw  managed to survey 133 Nigerian researchers, 91 (69%) of whom admitted to have ever committed scientific misconduct. When they cross-classified whether these respondents had "education in ethics" in the past, they found no statistical significant association. Furthermore, prior education in ethics was not associated with having had committed the following specific actions:
- falsifying data
- intentional protocol violation related to subject enrollment
- intentional protocol violations related to procedures
- selective dropping of data from 'outlier' cases
- falsification of biosketch, resume, reference list
- disagreements about authorship
- Pressure from study sponsor to engage in unethical practices
Take from it what you will.
 Okonta P, Rossouw T. Prevalence of scientific misconduct among a group of researchers in Nigeria. Dev World Bioeth. 2013 Dec;13(3):149-57. doi: 10.1111/j.1471-8847.2012.00339.x.