Overview
Must use stata to complete questions！！！！！！！！！
！
！
!
！
note: You can talk to each other about this assignment but I want individual papers, do-files, and
submissions. You should be providing your own analysis and thoughts.
1 Summary of the Assignment
This paper will involve two parts: reading and summarizing an academic article and writing a paper
using some of the data from the academic article. The paper should be 5-7 pages long 1.5-spaced
(including graphs and figures) and is meant to get you experience working with actual research
practices. Note that you are not trying to copy the analysis in the paper, although your own
analysis can be inspired by ideas in the paper.
2 Requirements
There are two major sections to the paper. There is no specific length requirement for any section,
but please do not let your summary exceed about 30% of the paper!
You will submit the paper which is 5-7 pages, including any associated pictures or drawings you
deem appropriate to include (though if they are too big or you include too many I reserve the right
to dock points). You will also submit a do file for any code you wrote to generate your statistical
tests, graphs, or regressions.
The paper you will be reading is ”Neighborhoods Matter: Assessing the Evidence for Place Effects”
by Eric Chyn and Lawrence F. Katz from the Fall 2021 edition of the Journal of Economic Per
spectives. The idea is you will read this paper, relate their findings to what we studied in class, and
then perform your own analysis using some of the data, which I will provide to you.
2.1 Summary
You should summarize, in your own words, the major findings of the paper across a few paragraphs.
Rather than just copying the abstract, you should try to relate their findings to concepts from class
as much as possible. You will lose points if you do not explicitly draw connections between
the paper and between things we studied in class.
Some examples of how you should try to relate the material to class:
1• Do the authors run any regressions? What are the major dependent and independent vari
ables?
variables, polynomials, logs, interaction terms, etc.
like violations of iid sampling or omitted variable bias?
for key variables or parameters?
If the techniques they use are more complex than what we do in class, explain to the best of your
ability what they are and why they use them. Please do not just write a book report that
does not talk about concepts from class!
2.2 Analysis
In this section, we will try to answer a research question using a series of models from class. The
ideas is that you will start simple and gradually use more complex econometric methods, while
explaining each choice you make.
2.2.1 The Data
First I will explain the data: you are going to be using some of the Opportunity Atlas data from
opportunityatlas.org. I have provided a ”commuting zone” (think of these as metropolitan areas,
slightly bigger than cities) level dataset that lists many key social and economic variables. The
dataset and a pdf explaining how the variables are named is on Canvas, though note that I have
removed many variables from the original dataset for ease of use. There is a provided codebook,
and it is your responsibility to figure out what the included variables are and which ones you would
like to use.
This dataset is mostly focused on things like inequality and opportunity. At the commuting zone
level, it has many variables broken out by race and gender, but also pooled across the entire CZ. For
example, ”has dad natam female mean” represents the percentage of Native American households
in a CZ where a female child has a father present at home. ”grad white pooled mean” is the
percentage of white children who grew up in an area who eventually received a graduate degree,
pooled across gender.
2.2.2 What you need to do
The idea here is you are building your own analysis in the vein of what we have learned in class.
We will start with a ”main” variable that you are interested in, where you will report some basic
statistics, and then you will do your best to build a causal model to explain relationships between
this variable and other variables in the data.
2Step 1: Pick a variable to focus on and report its mean and variance. There are about
400 variables in the included dataset, so I want you to give this some real thought. Try to pick a
variable you are interested in and one that you think could have an interesting relationship with
other variables in the dataset. Please do your best to interpret the mean and variance that you get.
Step 2: Run a one sample t test of this variable. It is up to you to pick your null hypothesis.
Please explain the choice you made, and interpret your results.
Step 3: Run a two sample t test of this variable, splitting your sample by a binary
variable that you generated based on another variable in the data
. The idea here is
you’re getting a two sample t test for the variable you are concerned with, across two groups of the
sample. Please explain your choice of groups, and interpret your results. The question here should
be something like ”is my variable of interest different across the two groups I chose?” You may find
the ”gen” command from earlier problem sets helpful here.
Step 4: Run a bivariate regression using your variable of choice as the dependent
variable, and another variable you think would be am important causal determinant
as the independent variable (X). In addition, provide a scatterplot of the two variables
Please justify all choices you make and interpret your results, mainly for the slope parameter and
R2 . Don’t forget to use robust standard errors!
Step 5: Turn this into a multivariate regression by adding ONE key control variable.
Please explain why you added this control variable in particular and why or why not this variable
needed to be included in the regression. Do your results support that the bivariate regression was
suffering from omitted variable bias? Is the new control variable a significant determinant of your
dependent variable?
Step 6: Please provide at least 3 other regression results with more control variables
included. The idea here is we have our first two regressions and 3 others to compare them to
to tell a story. Please do not exceed 5 total control variables in any one regression. Which set of
control variables do you favor and why? Please also run at least one F test and report the results.
Step 7: Run at least one nonlinear regression. Whether you want to do a polynomial, a
logarithm, or something else, that is up to you. Please explain any choices you made and interpret
your results.
Step 8: Reach an overall conclusion! Write at least one paragraph that summarizes your
results and report your overall conclusion!