Learning Objectives
By the end of this section, you will be able to:
- Understand what quantitative methods are
- Learn Steven’s Four Scales of Measurement
- Master the differences between cases, coding, and variables
As mentioned in Chapter Two, quantitative methods are defined by Flick (2018) as “research interested in frequencies and distributions of issues, events, or practices by collecting standardized data and using numbers and statistics for analyzing them”. Again, what this means is that political scientists solve puzzles using mathematical analysis or complex mathematical measurement. This differs from using qualitative methods where the main source of evidence used to solve a puzzle is the use of words. When using such methods, we often turn to appraising evidence in the form of words. As mentioned in Chapter Seven, we can use interviews and focus groups, archival research, and even digital ethnographies to understand the world. Given this, quantitative methods are simply the use of numbers to draw conclusions rather than words.
In political science, statistical analyses of datasets are the preferred quantitative method. This mostly developed from the behavioral wave in political science where scholars became more focused on how individuals make political decisions, such as voting in each election, or how they may express themselves ideologically. This often involves the use of surveys to collect evidence regarding human behavior. Potential respondents are sampled using a questionnaire constructed to elicit information regarding a subject. When using voters as an example, we may develop a survey that asks citizens if they are registered to vote, if they intend to vote, and which candidate for an office they might vote for. Respondent choices are then coded, usually using a scale of measurement, and the data is then analyzed often with the use of a statistical software program. Scholars probe for correlations among the constructed variables for evidence in support of their hypotheses on the topic.
However, quantitative methods extend beyond statistical analyses of survey datasets. Formal models are one such method. In formal models, political scientists attempt to understand representations of political institutions and political choices in the abstract. Relying on logic and causality, these scholars express relationships among concepts and variables in mathematical terms. They often use precise statements, written as equations, where the results can be replicated, almost always through a mathematical proof. Modeling the behavior of individuals or institutions has proven quite helpful in political science, particularly in the applied side of political science: public policy making. In this field, elected officials and subject matter experts work together to develop programs that can benefit society. Often the effect of a program is not discernible until the program has been implemented. However, formal models could help in projecting or predicting the effects of the program before implementation, which can help policymakers immensely.
Given that quantitative methods in political science often includes the analysis of data sets, it is often referred to as large-n analysis, where the “n” stands for number. Thus, we have an analysis of a large number of cases, again often assembled as sets of data. Cases are the people, places, things or actions (subjects) that are being observed in a research project. They are often also the unit of analysis. Units of analysis the “who” or the “what” that you are analyzing for your study. So, for large-n analyses of surveys, each case could consist of one respondent to the survey, or one person. Alternatively, cases could include the recording of individual actions taken. For quantitative analyses of institutions, cases could include people, such as senators or representatives, or the decisions made by lawmakers and/or policymakers.26
Keep in mind that cases and data are intertwined, but not the same thing. Each case can produce numerous data points. For example, each respondent in a survey can answer multiple questions, which could lead to large amount of data collection. In addition, in observational studies, where researchers observe and record the actions of individuals, there can also be a plethora of data points(Diez, Barr, and Cetinkaya-Rundel 2012).27
As statistical analyses of datasets are the popular quantitative method in political science, it is good to understand how such analyses work. First, it is important to understand that in some analyses, words must be transformed into numbers. By this we mean that any responses provided in surveys must be converted to numerical expressions, or value, for an analysis to take place. We often refer to this as coding. Coding is essential for the creation of variables to analyze in any quantitative research. A variable is defined by Hatcher (2013) as having “some characteristic of an observation which may display two or more values in a data set”.
In other analyses, there may be no need to code. The data itself is already in numerical form and forms the variable without any changes. An example could be a survey instrument that asks a respondent if they donated money to a campaign and what was the amount. As campaign donations is measured in dollars, there may be no need to code as the amounts represent individual data points for that variable.28 In other examples, respondents might be asked to rate themselves or some item/activity on a scale of 1-5. Consequently, each response and corresponding number could also be brought in directly, such as the campaign donations above. Or researchers can recode the data points, in some cases changing the way the variable is analyzed, or even create new variables entirely.
To better understand how variables work, we reference the four scales of measurement often used by statisticians. In his book on data analysis, Hatcher (2013) recounts this classification system, which is partially reproduced below in Table 8.1. These scales help researchers determine which statistical techniques would be the most proper to use to analyze the relationships between variables, which are all measured, coded and constructed differently,
Table 8-1: Steven’s Four Scales of Measurement
| Type of Measurement |
Definition* |
Example |
| Nominal Scale |
Identifies the groups to which a participant belongs; does not measure quantity or amount |
This is a variable that classifies a respondent. An example could include political party identification, where the distance between the variables is unimportant |
| Ordinal Scale |
Subjects are placed in categories, and the categories are ordered according to amount or quantity of the construct being measured. However, the variables are not equidistant from each other. |
This is a variable that is constructed from an ordinal scale, or a ranking of variables. An example could include asking students on a scale of 1-5 how liberal they might be. Ordinal variables are normally constructed from just one survey question, or a single item. Thus, the distance between the choices (1 through 5) are not necessarily equal. |
|
Interval Scale
|
A quantitative variable that possesses the property of equal intervals, but does not possess a true zero |
This is a variable that is constructed from a Likert- scale, or when several survey questions are used to create a score, or multiple items. An example would be asking students to complete a number of survey questions regarding their ideology on scales of 1-5. The responses are totaled and divided by the number of questions, providing a single score on where the student is positioned ideologically. |
| Ratio Scale |
An interval quantitative variable that displays a true zero |
This is a variable that has equal intervals between the responses or scores, but also includes a zero option which indicates that no amount of the construct has been measured. |
*definitions taken directly from Hatcher (2013)
