With standard data sets (that is, data sets without trader identifiers) manipulation is, in my opinion, hardly measurable. You may be able to identify misvaluation (for examle, misvaluation wil lead to return reversals, which can be measured). However, it will be almost impossible to trace the misvaluation to manipulative activities. This is because the misvaluation can have other reasons (e.g. investor sentiment, rumors, momentum trading, ...).
I agree with Erik. But if you want to use standard data, you can assume that earnings management phenomena can exists in some firms for price manipulation. Try searching about earnings management in accounting literature. Patricia Dechow and Mary Barth are good options for that.
Sorry, the issue was share price manipulation. I don't see why one has to assume earnings management here, when one could use direct proxies for share price manipulation.
Sorry. Now I understand your context.. and now I saw your profile and recognized you're an expert in accounting... my mistake. Trying to help you, I am still thinking it's difficult to measure SPM directly.. If you know, or find out a way, please share with us. In finance literature we can find something related to executive compensation and algorihtymic trading, but they are all indirect proxies for measuring share price manipulation in my opinion.. hope help you in some way.
A proxy for share price manipulation could be, in my opinion, the measurement of the volume (abnormal) of shares traded compared to the past volume and compared to other stocks. The well-informed traders (the "rumors" come from traders who hold confidential information not yet disclosed to the public), will Implement an abnormal strategy of "buy or sell" than other stocks and compared to the average of the past exchange of identical shares.