In the 1990s, India undertook large scale market-based reforms. This was expected to bring about convergence in growth rates across different states. However, the bulk of empirical literature on India suggests that there has been increased divergence in the post reforms period. We re-examine the convergence debate of India’s economic growth and question the validity of existing estimates of convergence rates, most of which are based on either cross-section analysis or non-spatial panel data analysis. We provide the first estimates of convergence rates in India from spatial panel data models. Our analysis reveals a significant influence of neighbouring states’ growth on per capita income of Indian states. The impact of initial income on growth is much smaller than earlier anticipated once we control for spatial dependence. This suggests that many of the earlier estimates of convergence rates and the impact of initial income on growth may be inaccurate and biased. Our results confirm that the “β” convergence coefficient is positive and significant suggesting income divergence. In addition, we find evidence of spatial dependence among the states in India. Apart from the state’s own initial income, what matters is how rich or poor its neighbours are. This occurs even though the neighbours may have different growth drivers and lack a set of common public policies. This has implications for growth policy-making in India, especially due to the ongoing institutional changes underway in the country.