Name: ___________________________________________ Date: ________________

E-mail: __________________________________________ Section: ______________

Exercises for Chapter 4: Making Comparisons

1. In this exercise, you will practice creating a cross-tabulation using two categorical variables in the
states dataset. Use unionized.4cat, an ordinal measure of the extent of unionization, as the
independent variable that defines the columns of the cross-tabulation. Use rtw as the dependent
variable that defines the rows; rtw identifies states with “right-to-work” laws.

A. Based on your analysis, fill in the following cross-tabulation. Each cell should report a
percentage as well as a count. Take note of the order of the columns and rows. We’ve set up
this cross-tabulation to help you make comparisons. 1
Unionization Level of State
Low Medium-Low Medium-High High Total
No right-to-work law 68.0%
(34)
Right-to-work law 32.0%
(16)
Total 100.0% 100.0%
(9) (50)

B. Create a line chart that shows the percentage of states with the “Yes” value of the rtw
variable by values of unionized.4cat. By default, the crosstabC function creates a bar chart of
“No” percentages but you can use the function’s plot and plot.response arguments to modify
the plot. Submit the line chart.

1
Refer to Section 4.1 for reference on cross-tabulation analysis. In this exercise, you’re given row labels, column
labels, and some cell values to help you make a proper comparison.

1
 C. Compare the percentages of right-to-work laws in states by level of unionization (focus on
the second row). Do states with higher levels of unionization appear more or less likely to
have right-to-work laws? Reference the percentages you wrote in part A in your answer.

2. Two scholars of comparative politics are discussing possible reasons why some democracies have
many political parties and other democracies have only a few.

Scholar 1: It all has to do with the rules of the election game. Some countries, such as the United
Kingdom, have single-member electoral districts. This militates in favor of fewer and larger parties,
since small parties have less chance of winning enough votes to gain the seat. Other countries, like
Switzerland, have multimember districts. Because voters choose more than one representative per
district, a larger number of smaller parties have a chance to win representation.

Scholar 2: I notice that your explanation fails to mention the single most important determinant of
the number of political parties: social heterogeneity. Homogeneous societies, those with few
linguistic or religious differences, have fewer conflicts and thus fewer parties. Heterogeneous
polities, by the same logic, are more contentious and will produce more parties.

A. Scholar 1’s Hypothesis: In a comparison of democracies, those having single-member

districts will have (select one)
○ fewer political parties ○ more political parties

than democracies electing multiple members from each district.

B. State Scholar 2’s hypothesis:

The world dataset variable enpp3.democ measures, for each democracy, the number of effective
parliamentary parties: 1–3 parties, 4–5 parties, or 6–11 parties. Use enpp3.democ (an ordinal-level
variable) as the dependent variable to test each hypothesis. For independent variables, test Scholar

2
 1’s hypothesis using district.size3, which measures, for each democracy, the number of members
per district: “single-member” districts, more than one but fewer than six members (>1–5), and
countries with “six or more members” per district. Test Scholar 2’s hypothesis using frac.eth3,
which classifies each country’s level of ethnic/linguistic fractionalization as low, medium, or high.

C. In the table that follows, record the percentages of cases falling into the lowest code of the
dependent variable, 1–3 parties. You should find these values on one row of each of your
cross-tabulations. 2
Number of Members per District (district.size3)
Single Member 2 to 5 Members 6 or More
Percentage having 1–3 parties

Level of Ethnic Fractionalization (frac.eth3)

Low Medium High
Percentage having 1–3 parties

D. Which of the following statements best summarizes your findings? (Select one.)
 Scholar 1’s hypothesis is supported by the analysis, but Scholar 2’s hypothesis is not
supported by the analysis.
 Scholar 2’s hypothesis is supported by the analysis, but Scholar 1’s hypothesis is not
supported by the analysis.
 Both hypotheses are supported by the analysis.
 Neither hypothesis is supported by the analysis.

E. Making specific reference to your findings, explain your choice in part D.3

2
Refer to Section 4.1 for reference on cross-tabulation analysis. In this exercise, you’re given column labels and one
row label to help you make proper comparisons.
3
See Section 4.1 for reference on interpreting cross-tabulations.

3
3. What factors determine how people vote in presidential elections? Political scientists have
investigated and debated this question for many years. A particularly powerful and elegant
perspective emphasizes voters’ retrospective evaluations. According to this view, voters who think
the economy got better during the year preceding the election reward the candidate of the incumbent
party. Voters who believe the economy got worse, by contrast, punish the incumbent party by voting
for the other party’s candidate. 4 In 2020, the incumbent president was Donald Trump.

The nes dataset allows us to test the retrospective voting hypothesis. Researchers asked respondents
whether they thought the economy got better or worse during the prior year (the econ.lastyear
variable) and who they voted for in the 2020 presidential election (the presvote2020 variable). 5

A. Create a cross-tabulation to compare vote choice by assessment of the economy. Be sure to

weight observations using the wt variable. Use your results to complete the table below. 6
Did the Economy Get Better/Worse in the Last Year? Total
Much Somewhat About Somewhat Much
Better Better Same Worse Worse
Joe Biden
Choice 2020

Donald Trump

B. Create a line chart that allows you to assess the retrospective evaluation theory. Since the
theory generates a hypothesis about voters’ assessment of the incumbent president (in 2020,
Donald Trump), plot the percentages voting for Trump by values of the econ.lastyear
variable. Submit the line chart.

4
As political scientist V. O. Key once famously put it, the electorate plays the role of “rational god of vengeance
and reward.” V. O. Key, Politics, Parties, and Pressure Groups, 5th ed. (New York: Crowell, 1964), 568.
5
See Section 4.1 for guidance on cross-tabulations.
6
The nes presvote2020 variable includes votes for minor-party candidates. To keep this analysis simple, we’ve
omitted those rows and the column totals from the cross-tabulation.

4
 C. What do you think? Are the data consistent with the retrospective evaluation hypothesis?
Write a paragraph explaining your reasoning.

4. In this exercise, you’ll make comparisons with an interval-level independent variable by collapsing it
into simplified categories and then making a cross-tabulation. This is a good way to make broad
comparisons and visualize the results.

The states dataset contains the variable medicaid.expansion, a nominal-level variable that indicates
what legislative action each state took on Medicaid expansion under the Affordable Care Act. The
dataset also contains the interval-level variable dem.stateleg, which is the percentage of Democratic
state legislators in each state. Let’s compare the Medicaid expansion decisions of states by level of
Democratic representation in their state legislatures.

A. Start by creating a new variable in the dataset named dem.stateleg.3cat that classifies states’
Democratic representation by three equal-sized tiers. (There are forty-nine partisan state
legislatures.) After you create the new ordinal-level dem.stateleg.3cat variable based on the
interval-level dem.stateleg variable, execute the command freqC(states$dem.stateleg.3cat)
and complete the following frequency distribution table for dem.stateleg.3cat. 7
Value of dem.stateleg.3cat Count Percent
17 34.69
16 32.65
16 32.65
Total 49 100.00

7
Part A asks you to practice a skill we covered in Section 3.3.

5
 B. Produce a cross-tabulation of the existing medicaid_expansion variable and the new
dem.stateleg.3cat variable. Since medicaid_expansion is the dependent variable, it defines
the rows; the independent variable, dem.stateleg.3cat, defines the columns. Use the results of
your analysis to complete the following table. Fill in the dem.stateleg.3cat variable’s labels.
Each cell should have a percentage and a count. 8
Democrats in State Legislature
Medicaid Expansion Total
Not adopted 63.3%
(31)
Adopted

Total 100.0%
(49)

5. In this exercise, you’ll examine how Americans get political campaign news. Newspapers are a
traditional source of campaign information but newspaper usage may be changing over time, with
younger people turning away from newspapers The nes dataset includes a set of variables that
identify sources that respondents mentioned when researchers asked them how they get political
campaign news. The value of the campaign.news.papers variable is either 0 (“not mentioned”) for
those who don’t use newspaper or 1 (“mentioned”) for those who say they read newspapers for
campaign news.

A. Use respondents’ ages as the independent variable in this analysis. To make broad
comparisons by age, create a new variable named age.6cat by collapsing the interval-level
age variable into six ordered categories. Use the transformC function with type=“cut” and
cutpoints=c(25, 35, 45, 55, 65, 75). To check your work on age.6cat, complete the following
frequency distribution table (be sure to weight observations).
Value of age.6cat Count Percent
[18,25) 10.9
[25,35)
[35,45)
[45,55)
[55,65)

8
Section 4.1 covers cross-tabulation analysis. In this exercise, you’re given row labels, column labels, and some cell
values to help you make a proper comparison.

6
 [65,75)
[75,80]
Total 7,956.7 100.00

B. Use the crosstabC function to compare the use of newspapers for political campaign
information by age group. Use the function’s w argument to weight observations with the
nes dataset’s wt variable. Be sure to organize the table correctly and submit it below.

C. Create a chart that displays the percentage of respondents who mentioned using newpapers
for political campaign news by age group. Keep in mind that the independent variable,
age.6cat, is ordinal when deciding whether to create a bar chart or a line chart. Label the
chart, the x-axis, and the y-axis appropriately and submit the chart with these exercises.

7
D. Do the data support the claim that younger people are less likely to use newspapers for
political campaign news compared to older people? (Select one.)
○ Yes ○ No
Explain your answer.