Teaching tool, not investment advice. In the paper only GDP is statistically significant; the repo rate and CPI effects cannot be told apart from zero. Correlation is not causation.
How the model and data work
The paper’s log-log regression is ln(Nifty) = α + β₁ln(Repo) + β₂ln(GDP) + β₃ln(CPI), fitted on 24 annual observations (March 2001 to March 2024) with adjusted R² of 94.9%. The paper reports the three coefficients but not the intercept α or the raw data, so α here is calibrated so that the model fits the six historical snapshots on average. Because of that, absolute levels are indicative; the sensitivities to each input come directly from the paper.
Snapshot data: Nifty closes and repo rates are 31 March values. GDP is approximate World Bank nominal GDP for the calendar year. CPI is CPI-IW (fiscal-year average) linked to a 1960 base using official linking factors, so treat it as approximate.