In thinking about India, a big constraint lies in the statistical system. Many key pieces of information are mis-measured, and appropriate skepticism needs to be applied when consuming official statistics. As an example, in my experience with learning macroeconomics in an Indian setting, it took many years to gain comfort about what time-series one is going to take seriously. Mundane issues like measurement of inflation, which ought to be effortless in a mature market economy, are a complex challenge in India.
There are three datasets where I feel there is a big gap between perception and reality: the National Sample Survey (NSS), the Annual Survey of Industries (ASI) and "State Domestic Product" (SDP). I find myself being extremely skeptical about these three. Today's Business Standard has a good article by Sunil Jain about the measurement problems of NSS (and NAS).
Many academic research papers merrily proceed on trying to use these datasets. When I read these papers, I generally end up shaking my head saying "I just don't trust any of this". It's a shame, for these datasets really ought to be done better. But it's also a waste when researchers indiscriminately proceed to expend man-months of time on weak foundations. I think anyone working on India should be a little more thoughtful about difficulties of measurement. To a discriminating reader, weak data foundations lead to a fatal loss of persuasion, and if a research paper won't persuade, why bother writing it?