This 2010 article at nobribe.org claims
Measurement of Corruption thus far, has been limited to measuring perception of Corruption [...] Various approaches to measure Corruption directly may be suggested based on how other intractable quantities are measured. One such model to measure petty Corruption based on Inflation, is discussed [...].
(The latter contains a link that should probably point to an article entitled "Obtaining a Driver's License in India: An Experimental Approach to Studying Corruption".)
As for the methodology of Corruptions Perception Index, a little digging on the Transparency International website leads to a "data package" including a "Source description" pdf mentioning 13 different studies as source, some focussed on corruption, some not, often relying on a limited number of "experts" (e.g. 2 per country for study 3, 4200 responding "business executives", apprently in total, for study 7) to "assess" (not: "measure") a question, sometimes longer and detailed, sometimes shorter (e.g. “Bribing and corruption: Exist or do not exist”) on a given scale.