Instructions

• The midterm assessment will be posted on Moodle on 5th November 2021 at 6pm, and is due on 10th November 2021 at 2pm. Please follow all designated SPP submission guidelines for online submission as detailed on the PUBL0055 Moodle page. Standard late submission penalties apply.

• This is an assessed piece of coursework (worth 25% of your final module mark) for the PUBL0055 module; collaboration and/or discussion of the coursework with anyone is strictly prohibited.  The rules for plagiarism apply and any cases of suspected plagiarism of published work or the work of classmates will be taken seriously.

• As this is an assessed piece of work, you may not email/ask the course teaching team questions about the coursework.

• Along with the coursework questions, the necessary datasets for the coursework can be found on the PUBL0055 page on Moodle.

• The coursework should be submitted via the ‘PUBL0055 Midterm Paper - 1000 words (25%)’ link on the course Moodle page.  You will need to click the ‘Submit Paper’link at the bottom of the page. When presented with the ‘Submit Paper’box, the ‘Submission Title’ should be your candidate number, and you should upload your document into the box provided.  Please remember to state ONLY your candidate number on your coursework (your candidate number is made up of four letters and one number e.g. ABCD5).   Your name and/or student number MUST NOT appear on your submission.

• The coursework consists of two separate sections, each with several questions. The marks allocated for each section are indicated in the text. You must complete each question to achieve full marks.

• Unless otherwise stated, answers should be written in complete sentences. Be sure to answer all parts of the questions posed and provide a substantive interpretation of the results.

• The word count for this assessment is 1,000 words. This does not include the code, or any words (or numbers) contained within tables or figures.

•  Please submit your type-written (numbered) answers in a single document (a word file or a pdf file).

You should create an appendix section at the end which contains all the R code needed to reproduce

your results. Your code has to work when we run it. You do not need to include the code that failed to run, but just the well-annotated, cleaned-up version. Failure to include the R code means that the coursework will be marked incomplete.

• You may assume the methods you have used  (e.g. difference in means, linear regression, etc) are understood by the reader and do not need definitions, but you do need to be able to explain what they do and how they apply to answering the question.

• Round all numbers to two digits after the decimal point.

• Do not screenshot or copy and paste any brute R output (e.g. lm(y  ∼ x)) into your answers. Create a formatted table that is easy to read.

• Assign every table and figure a title and a number and refer to the number in the text when discussing a specific figure or table.

Section 1

Populism and COVID-19

Some countries have been more successful than others in terms of public health outcomes during the COVID- 19 pandemic. In particular, deaths associated with the virus have been unevenly distributed cross-nationally. Of the many factors that might be responsible for these differences, some media attention has focused on the idea that countries led by populists may have been less successful at dealing with COVID-19 than countries led by mainstream party politicians.  For instance, a piece by Brett Meyer asked “Did countries with populist leaders suffer more from COVID?”.  Reporting of this sort relates to ongoing debates in the academic literature on rise of populism and the efficacy of populist leaders.

In this section, you will investigate these ideas empirically by examining the relationship between populist

leadership and COVID-19 death-rates in a sample of countries. This exercise is loosely based on papers by Meyer and Bosancianu et al, both of which use data on COVID-19 fatalities and a variety of variables about countries’ social and political context.

The data file you will use, which can be downloaded on the PUBL0055 Moodle page, is titled covid_data .csv and contains information from 210 countries. The data includes the following variables:

You can load the data by using the following command:

covid  <-  read .csv ("data/covid_data .csv")

Questions (68 Marks)

1. Begin your analysis by providing information about missingness in the dataset.  How many missing values does each variable in the dataset have? Explain whether there is unit and/or item non-response.

2. Provide  some  descriptive  statistics  on  three  of  the  main  variables  of  interest  for  this  analysis, electoral_pop, distancing_n and deaths_cum_per_million.

a. Which countries had populist leaders in 2020?   What is the share of countries with populist leaders?

b.  Describe separately the distancing_n and deaths_cum_per_million variable in graphical form (choose one graph each). What do your graphs show?

c. Explain what kind of dataset this is: are we dealing with experimental or observational data?

3. Focus on countries in Europe. Graphically explore the two relationships described below. How do you interpret your graphs?

a. Visually represent and comment on the relationship between whether or not a populist leader is in power and the total number of distancing measures

b. Visually represent and comment on relationship between the total number of distancing measures and the total number of deaths associated with COVID-19.

4. Let’s look at all continents again.   Calculate the difference in mean deaths for countries with and without populist leaders.

a. Interpret this difference in means in substantive terms.

b.  Does this reflect the causal effect of populist leadership on COVID-19 deaths? Why or why not?

5. Imagine that the researcher conducting this study is not confident that the difference in means repre- sents the causal effect of populism. Describe one alternative approach/design that the researcher might implement to estimate the desired effect. How does your proposed approach improve on the difference in means comparison?

6. Estimate two regression models with deaths_cum_per_million as the dependent variable.  For the first model,  include  electoral_pop  as the only explanatory variable.   For the second model,  in- clude electoral_pop and three other variables of your choice  (do not use any of these variables: lockdown_bin, distancing_n, distancing_bin). Interpret your results, making sure to explain and

compare

a. the magnitude of the coefficients on populism

b.  the explanatory power of the two models

c. the intercepts

7. In the dataset, we also have information about the total number of distancing measures that had been implemented in a country.

a. Estimate a simple linear regression model with the total number of COVID-19 fatalities per one million people in the population as the dependent variable and the number of distancing measures as the independent variable. Interpret your results.

b. Estimate a second linear model (with the same dependent variable as in the previous model) in which you interact distancing_n with populist leadership. What do you find? Is the relationship between distancing measures and COVID-19 deaths different in countries with populist leaders compared to those without populist leadership?

c. Using the model you estimated in  Section  1,  Question  7. b .,  calculate the fitted values for  1 distancing measure and 20 distancing measures, separately for countries with and without populist leadership. Explain your results and compare the difference in the fitted values between countries with and without populist leadership.

d.  Present the fitted values visually. Describe what the graph shows.

Section 2

Moving Out of Poverty

Can moving out of high-poverty neighbourghoods improve low-income families’ wellbeing?  The “Moving  to Opportunity”demonstration was designed to establish whether neighbourhood conditions affect families’ economic self-sufficiency and subjective wellbeing.  In the 1990s, families in a number of U.S. cities were  enrolled to participate in this demonstration. While some of the enrolled households were randomly chosen  to receive a housing voucher (i.e., rental assistance) that would allow them to move out of high-poverty  neighbourhoods, the other enrolled households did not receive a voucher (but continued to be eligible for  any other government assistance they were entitled to).

In this section, you will analyse data from this experiment.   The data file you will use, which can be downloaded on the PUBL0055 Moodle page,  is titled moving .csv  and contains information from 3263 individuals who were randomly assigned to receive a voucher or not.   The treatment of interest here is not relocating, because whether or not an individual decided to relocate was not controlled by the U.S. Department of Housing and Urban Development that conducted the demonstration.   Instead, what was randomised was whether or not a household received a voucher.

This exercise is loosely based on a paper by Ludwig et al.

The data includes the following variables:

You can load the data by using the following command:

moving  <-  read .csv ("data/moving .csv")

Questions (32 Marks)

1. Describe the data from the experiment:

a. What is the proportion of households that received a voucher?

b.  What is the proportion of households that relocated?

c.  What is the proportion of households that relocated after receiving a voucher?

d. What is the proportion of households that relocated even though they did not receive a voucher?

Answer these questions with a short paragraph. What do these proportions tell you about the effec- tiveness of the vouchers?

2. Provide one measure of central tendency and one measure of spread for wellbeing_zscore. What can you tell us about the subjective wellbeing of enrolled households?

3. Calculate the difference in mean subjective wellbeing between households that received a voucher and those that did not.  Interpret your results.  Does the difference in means estimate the causal effect of receiving a housing voucher on subjective wellbeing? Why, or why not?