The Eects of Educational Television
Is educational television an eective teaching aid? The Electric Company was a television programme that ran on US TV from 1971 to 1977. The programme used sketch comedy to provide an entertaining way of helping elementary school children develop their grammar and reading skills. It was widely credited by many teachers in US schools as having important eects on the literacy skills of second, third, and fourthgrade children. In this section, you will analyse data from an experiment that involved randomly assigning classes of children to watch The Electric Company. You will investigate what reading gains, if any, were made classes as part of this experiment.
The unit of analysis in this data is a class of children, and there are 192 classes in the data. Each class was either treated to watch the program or control to not watch the program. The outcome of interest is the average score on a reading test administered at the end of each year called post.score. In addition to the treatment and outcome, the data also contains information on the year grade of the class and the score on the same reading test as adminstered before the treatment took place:
Name
grade
treatment
pre.score
post.score
Description
The school grade of the class 1st through 4th
1 if the class was treated, 0 otherwise randomized
Class reading score before treatment, at the beginning of the school year Class reading score at the end of the school year
The data is stored in electriccompany.csv. Once you have downloaded this file and placed it in the relevant folder, it can be loaded into R as follows:
electricread.csvdataelectriccompany.csv Question 1 16 marks
a. Calculate and interpret the average eect of the treatment on the class reading score at the end of the school year.
b. Explain whether we can interpret your answer to part a as the causal eect of television on student scores. c. Calculate the standard error of the dierence in means. Show your work.
d. Conduct a hypothesis test for the dierence in means. Can we reject the null hypothesis of no eect of the treatment at the 95 and 99 confidence levels?
e. Calculate and interpret the 95 confidence intervals for the dierence in means estimate.
f. Explain the concept of a sampling distribution. What is the shape of the sampling distribution in this
example?
Question 2 10 marks
a. Make a scatter plot which compares student scores at the beginning of the year to student scores at the end of the year.
b. Make a box plot which depicts student scores at the end of the year as a function of the grade they are in.
c. Estimate three linear regression models. The first should predict post.score with only the treatment variable. The second regression model should be the same as the first, but should also control for student grade. The third model should be the same as the second, but should also control for pre.score.
d. Summarise these models in terms of how much of the variation in post.score they explain. What does this tell us about the relationships between 1 the grade a student is in and reading ability, and 2 students prior performance on the test and current performance on the test?
2
e. Are the estimates of the treatment coecient dierent across the three models? Why do you think that is? You may wish to provide evidence from the data to support your argument. You may also wish to refer to your answers to parts a and b of this question.
Question 3 6 marks
Use the grade variable to subset the data, and then use linear regression models to evaluate the eect of treatment within each grade. How does the eect of the treatment dier as grade increases? Comment on both the substantive and statistical significance of these results.
Question 4 6 marks
Write a short paragraph summarising your findings from these analyses. You should write as if you are trying to communicate the results to someone who is interested in the eects of television on learning, but who has not taken a course in quantitative methods. You may wish to create a visualisation to help communicate the findings.
3
Political Parties and Policy Outcomes
Does which political party is in power matter for policy outcomes? This is an important question for political scientists to answer, not least because many theories of voting assume that voters hold governing parties to account on the basis of their performance in oce. If such retrospective voting is to occur, it must be the case that dierent political coalitions have clear and consistent eects on policy outcomes in the time between elections.
To determine whether this is the case, in this section you will use data from the US to analyse the eects of the party in power in US state governments specifically, which party holds the governorship of the state on a number of dierent policy outcomes. The data comes from 864 elections across 50 states in the US, plus the District of Columbia. The variables included in the data are:
Name
demgovernor
demmargin
demmajorityhouse
demmajoritysenate
state
year
unemployment
murder
incometoponepct
houseprices
Description
The party of the governor 1 if Democrat, 0 if Republican
The Democratic electoral margin in the relevant election for governor. This variable is measured as the Democratic candidates vote share minus the Republican candidates vote share, such that negative values indicate the Democratic candidate lost the election, positive values indicate the Democratic candidate won the election.
Whether the lower house of the state legislature is controlled by the Democrats 1 or Republicans 0
Whether the upper house of the state legislature is controlled by the Democrats 1 or Republicans 0
The state.
The year of the governors election.
The unemployment rate, measured 2 years after the governor was elected. The number of murders per 100,000 people, measured 2 years after the governor was elected.
The percentage of income held by the top 1 of the population, measured 2 years after the governor was elected.
Quarterly Housing Price Index, measured 2 years after the governor was elected.
The data is stored in governors.csv. Once you have downloaded this file and placed it in the relevant folder, it can be loaded into R as follows:
governorsread.csvdatagovernors.csv Question 5 6 marks
a. For each of the 4 outcome variables, estimate a linear regression where demgovernor is the only explanatory variable. Present the results in a table.
b. Interpret the regression coecients in each model. Question 6 13 marks
a. Adapt the regression models that you estimated above to include two additional control variables: demmajorityhouse and demmajoritysenate. Estimate these regression models and present the results in a table.
b. Why might it be important to include these additional variables in your regression?
c. Interpret each of your three regressions, paying particular attention to the coecient associated with the
demgovernor variable. Can the coecient be interpreted causally in these models? Explain why or why not. 4
c. Your goal is to identify the causal eect of Democratic governors on these outcome variables. Imagine that you had unlimited time and unlimited budget: describe one variable that you would ideally control for in these models. Why?
Question 7 18 marks
In the paper on which this example is based, the authors use a regression discontinuity RD design. In this design, the authors use the Democratic candidates electoral margin variable to make comparisons between states that narrowly elected a Democrat to states that narrowly elected a Republican for governor. In this question, you will replicate parts of the original RD analysis.
a. Write a short paragraph discussing why using a regression discontinuity design of this type might be better than simply comparing states that have Democratic governors to states that have Republican governors. Explain also one disadvantage of using a regression discontinuity design in the context of this study.
b. Use the demmargin variable to compare policy outcomes between states that narrowly elected a Democratic governer and states that narrowly elected a Republican governor. Report and interpret the regression discontinuity treatment eect for all four outcome variables.
c. Produce four plots that depict the regression discontinuity design graphically. Each plot should depict the relationship between the Democratic electoral margin and one of the policy outcomes. Your plot should include two lines depicting the relationship on either side of the cuto, and a vertical line to show the location of the cuto on the xaxis.
d. Write a short paragraph which compares your findings from the regression discontinuity design analysis here to your findings from the regressions that you estimated in questions 5 and 6. What do you conclude about whether political parties have important eects on policy outcomes?
5
Religion and the Electoral Success of the Nazi Party in 1932
In Weimar Germany, the Catholic Church vehemently warned ordinary parishioners about the dangers of extremist parties. During this period, the church in Germany was particularly active in discouraging Catholics from supporting the Nationalsozialistische Deutsche Arbeiterpartei NSDAP, which is commonly known in English as the Nazi party. Alerted by the Nazis sudden success at the polls and afraid of anticlerical movements within the party, Catholic bishops took an explicit antiHitler stand in the autumn of 1930. Historians have long contended that this antiNazi position from Catholic religious leaders had consequences for the level of support amongst Catholic citizens, particularly in the context of the Reichstag elections in 1932.
In this section, you will investigate whether Catholic areas of Germany were less likely to support the Nazi party in the elections held in November 1932. The data includes election results from 982 counties, and includes a number of variables:
Name
nsdapshare
pctcatholic
pctjewish
pctotherrel
pctwomen
logpopulation
femaleworkforce
urbancounty
unempwhitecollar
unempbluecollar
unempdomestic
industrymanufacturing
industrytrade
industryservices
industrydomestic
distancetoberlin
literacyrate
bavaria
logincomepercapita
religiosity
Description
The NSDAPof the vote in the county
The Catholic percentage of the population in the county
The Jewish percentage of the population in the county
The percentage of the population in the county of another Religion The female percentage of the population in the county
The total log population of the county
The female percentage of the workforce in the county
If the county was urban 1 or rural 0
The unemployment rate amongst whitecollar workers in the county The unemployment rate amongst bluecollar workers in the county The unemployment rate amongst domestic servants in the county The percentage of the countys workforce in manufacturing industries
The percentage of the countys workforce in trade and commerce The percentage of the countys workforce in public and private services
The percentage of the countys workforce employed in domestic service
Distance of the county to Berlin KM
The literacy rate in the county
If the county was located in the state of Bavaria 1 or not 0 County log income per capita
A binary indicator for whether the county is above average in terms of religiosity, defined on the basis of how many citizens regularly participate in religious festivals. 1more religious, 0less religious.
The data is stored in reichstag.csv. Once you have downloaded this file and placed it in the relevant folder, it can be loaded into R as follows:
reichstagread.csvdatareichstag.csv Question 8 25 marks
Your task in this section is to investigate the relationship between the share of Catholics in a district and the NSDAP vote share in that district in the election in order to answer the research question outlined above. In particular, you should implement two linear regression models with nsdapshare as the dependent variable.
6
In the first model, the only explanatory variable should be the pctcatholic variable. For the second model, you should build a model whichin addition to the pctcatholic variableincludes exactly three additional explanatory variables that you think might be useful to include from the supplied dataset. You should explain why you think these particular variables are important to include, given that our main interest is in the relationship between Catholicism and Nazi vote share. Please note that, for the second model, you should not estimate several dierent models and present the results, but rather you should argue theoretically why you chose certain variables.
You should write up the results of these models as if they were to be published in a political science journal article with a focus on communicating the substantive meaning of your results. In your discussion of these models, you should focus on communicating the substantive implications of the regression that you implement, paying particular attention to the relationship between the Catholic population of a district and Nazi vote share in the election. You may wish to focus on the following:
Provide descriptive statistics andor plots to provide the reader with an overview of the dependent variable and the important explanatory variables that you intend to use.
Provide a wellformatted table of regression output which includes the key information about the models you have estimated.
Discuss both the statistical and substantive significance of the relationships that you illustrate.
Discuss model fit, using appropriate statistics.
Discuss whether or not we should consider the estimates you present to be causally identified.
Discuss weaknesses of you analysis, and potential alternative analysis designs that you might use given
dierent data to evaluate this research question.
7
Reviews
There are no reviews yet.