闪电代写 -代写CS作业_CS代写_Finance代写_Economic代写_Statistics代写_代码代做_IT代写_加急帮助

Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit

ECON 191b: Econometric Methods for Causal Inference

Research Paper — Spring 2022

This document contains instructions for the preparation of a research paper for ECON 191b in Spring 2022. On LATTE, you will ﬁnd three papers with their as- sociated replication packages (in ZIP format), which contain, among other things, Stata datasets and do-ﬁles that enable replicating the estimation results in each paper. The papers are as follows:

● Aizer, Anna, Shari Eli, Joseph Ferrie, and Adriana Lleras-Muney. 2016. “The Long-Run Impact of Cash Transfers to Poor Families.” American Eco- nomic Review 106(4): 935–971.

●	Belenzon, Sharon, Aaron K. Chatterji, and Brendan Daley. 2017. “Epony- mous Entrepreneurs.” American Economic Review 107(6): 1638–1655.

● Voigtl¨ander, Nico, and Hans-Joachim Voth. 2012. “Persecution Perpet- uated: The Medieval Origins of Anti-Semitic Violence in Nazi Germany.” Quarterly Journal of Economics 127(3): 1339–1392.

Your job is to choose one (and only one) of these papers and extend its empirical analysis in a direction of your choice, perhaps using the tools that we have dis- cussed in class. You may consider using different estimation methods than those considered by the authors of your selected article; estimating different parameters than those studied by the authors; conducting a sensitivity and/or a heterogeneity analysis; and/or assessing the sensitivity of estimation results to the violation of assumptions that are necessary for identiﬁcation. (You can also do something else entirely, as long as it involves one of these papers, although in this case I would recommend consulting your idea with me before pursuing it.)

Your paper is due by 1 pm on May 2, that is, just before our last class. Please submit the paper (in DOC/DOCX or PDF format) on Dropbox, together with your Stata do-ﬁle and log. The appropriate Dropbox link is posted on LATTE. Please make sure that your do-ﬁle only references (“uses”) datasets with names that cor- respond to those in the original replication packages. This will ensure that I am able to rerun your code on my computer if I choose to do so.

Your paper is required to be ﬁve pages or fewer (double-spaced, twelve-point Times New Roman font with one-inch margins). Additionally, you can include

up to two pages with tables that contain estimation output (for a total of seven pages maximum). The formatting of these tables should follow the conventions in the paper you are replicating. It is particularly important that you report only the relevant part of your estimation output1 and that you report standard errors whenever appropriate. Your paper should include a brief summary of the paper you are replicating and should otherwise focus on your extension(s) of the original analysis. You should be transparent about whether your results agree with the conclusions of the authors of the original paper. If you disagree with them in some ways, you should make it very clear.

In the remainder of this document, I will make some speciﬁc comments about the problems you may encounter when using the replication packages associated with each of the papers. (In any case, these are the problems I have encountered; it is, of course, possible that you may run into some other problems, too.)

●

Aizer et al. (2016): The most important estimation results of this paper are reported in Table 4. The most important do-ﬁle in the associated replication package is Results.do. You should be able to run this code in its entirety without any problems, although parts of the code towards the end of the do-ﬁle are rather slow. (It took me about one hour to run this do-ﬁle in its entirety on my computer.)

Belenzon et al. (2017): The most important estimation results of this paper are reported in Table 4. The most important do-ﬁle in the associated replica- tion package is EPO MAIN REG FILE AER.do. You should be able to run most of this do-ﬁle without any problems. The code is rather slow, however, which is due to the fact that the dataset is very large. I have been unable to replicate columns 4 and 6 in Table 5 (“City-FEs” and “Nonlinear effects”) because my computer has been running out of memory. You should feel free to ignore these estimates as well as others that you may be unable to replicate.

Voigtl¨ander and Voth (2012): The most important estimation results of this paper are reported in Table VI. The most important do-ﬁle in the associated

replication package is QJE Replicate Paper.do. To successfully execute this

do-ﬁle, you need to make two changes. First, you need to change the name of the dataset in line 9; the correct ﬁle name is not Dataset QJE Replicate.dta but Dataset QJE Replicate with Cities.dta. Second, you need to com- ment out the code in line 520 (reg hephep pog1349...); I do not understand why – but it does not run.