The central questions posed by this thesis are: what are the effects of labour market regulations pertaining to job security in India, and why are these regulations so difficult to reform. The thesis finds that job security regulations have a negative effect on both efficiency and equity. They have a significantly negative impact on employment in all categories. They benefit a small minority of highly educated and high human capital workers, while excluding the large majority of the labour force from secure, protected work. They also have a negative impact on output, as they discourage investment. This is shown through a ranking of twenty four Indian states according to the strictness of job security regulations. Highly labour regulated states have lower levels of investment, leading to a negative impact on output, employment and real wage. In this way, these regulations harm both efficiency and equity. In saying this, this thesis supports the distortion view of job security regulations as held by the World Bank, and refutes the institutional view as held by the International Labour Office (ILO). The findings of this thesis show that the result of high levels job security regulations do not cause a necessary trade-off between efficiency and equity (sacrificing the former to get more of the latter), but that the result is a negative impact on both efficiency and equity. The thesis then asks why policies that reduce both efficiency and equity are so difficult to reform in a democracy like India. It explores this by doing an inter-state analysis of policy reform in ten Indian states, considering each state as a separate democracy. It finds conclusive evidence that political factors influence the capacity and motivation to carry out labour policy reform, and it analyse what factors these might be. We use a multi-pronged political economy approach in this thesis. We use extensive historical and institutional analysis, combined with fairly simple, but powerful, empirical analysis. Most of our empirical analysis relies largely on simple and straightforward ordinary least squares (OLS). We are encouraged by the fact that we use four different datasets, and all four give us the same significant result. This gives us confidence in the strength and robustness of our findings.