My thesis contains three chapters which focus heavily on the macroeconomic policies. The first chapter focuses on the effectiveness of monetary policy on firms with different financial constraints. The second chapter addresses on how would the optimal tax policy change the evolution of inequality. The third chapter emphasizes on how to provide a proxy means testing from a welfare perspective to a transfer program. In the first chapter, I study the role of financial constraints in the effects of monetary policy on firm investments. I construct a quarterly textual measure of financial constraints from SEC filings using a deep learning model. It improves the prediction accuracy as compared to a Naive Bayes method by capturing the context information, such as grammatical structure and order of words. Firms classified as highly constrained are younger, smaller, have a higher liquidity ratio and higher leverage ratio. However, popular proxies of financial constraints often do not move monotonically with the level of financial constraints. Particularly for the liquidity ratio, it is high for both the least constrained firms, which have ample of cash, and the most constrained firms, which hoard cash due to precautionary saving motives and the high marginal cost of external capital. Using the constructed measure of financial constraints, the investments of financially constrained firms are persistently less responsive to monetary policy shocks due to high marginal cost of external funds. This implies that monetary policy might be less effective during crisis time due to a larger fraction of constrained firms. My results reconcile previous empirical findings and argue that the seemingly contrary conclusions are, to some extent, consistent with each other. In the second chapter, it intends to address on the question: how would the optimal taxes change the evolution of wealth inequality? This paper studies this question quantitatively under a standard incomplete market heterogeneous agent model. The benchmark model captures the wealth distribution and its evolution from 1967-2010. Optimal tax policy exercise considers an once-and-for-all tax reform at 1967 accounting for the time varying economic environment and transition dynamics. With a utilitarian social planner, the optimal linear comprehensive income tax leads to a higher level inequality in wealth where top 10% and top 1% gain at least 5% more wealth shares at 2010 compared to benchmark. The optimal tax under a parameterized nonlinear tax function implies a highly progressive tax system which is also highly redistributive compared to the benchmark model. The wealth inequality in this case is increasing from 1960s to mid 1990s and then start to decline to its 1960s level or even lower. At 2010, top 10% remains roughly their wealth holdings at their 1967 level while top 1%, 0.1% and 0.01% wealth holding even decrease on average about 2% compared to their low level at year 1967. In the last chapter, I propose a new proxy means testing method with minimizing welfare loss as the target instead of traditional targets such as minimizing consumption loss. In a simple economy with a utilitarian social planner, the welfare approach is equivalent to a weighted logistic regression with inverse consumption as weights. As a result, it focuses mainly on the exclusion error where poor are identified as non-poor and less weights on the inclusion error where non-poor are identified as poor. Using the socio-economic survey data in India in 2011, I compare the targeting performance of the welfare approach to other standard approaches in PMT. It shows that the welfare approach enjoys a lower exclusion error rate by sacrificing the inclusion error rate and does not out-perform the traditional method. It does, on the other hand, provide a welfare foundation for the poverty weighted least square method.