While investigating patents, you learn that China started a program to incentivize patents in 1997. From that year on, the government would give the authors of successful patents a large cash reward. You decide to investigate the causal eect of this policy on the number of patents awarded. 1. 2. Why might 61 from the regression N U M PATENTSCMW = g + l 1{t > 1997} + comm; be an inconsistent estimator for the causal effect? Say you also have patent data from Japan. Under what assumption will this data enable you to estimate the causal effect of the Chinese program on patents awarded? How could you test this assumption (hint: draw a graph)? . Given that the assumption you stated above holds, what regression would you run to estimate the causal effect of the Chinese program on patents awarded (i.e. what is the difference-in- di'erence specication)? . Say the causal effect of the program is some value g. Give an expression for 33 in terms of population moments. Also, give a consistent estimator for [33 in terms of sample moments. . Show how to test the null that the causal effect is 0 versus the alternative that the causal effect is different from 0 at the 5% level (hint: testing for difference-in-difference works the same as with any other regression)