2.01: Concepts to Measurements

OBJECTIVES:

Understanding how to narrow from general concepts to specific measurements

As we discussed in a previous lecture, concepts are just general categorizations of things we use to understand phenomena. Concepts exist within the conceptual framework, and help us to identify patterns and relationships. However, we as researchers need to narrow down these general concepts into more specific measurements in order to specific the models we are actually testing.

If you remember from that lecture, we go through a process of narrowing down from the conceptual framework, to a theory, to a hypothesis, to a model. At the same time we do this we are narrowing down from a concept, to a conceptual definition, to a particular variable, to a specific operational definition and measurement.

Narrowing Down: Concept to Measurement

Define the Concept
1. Example: what do we mean when we say political participation?

i. Voting? Protest? Contacting a public official?

Select the Variables of Interest
1. If voting:

i. Which aspect of voting are we most interested in?

ii. Is it whether or not someone votes? Or which party or candidate they vote for?

iii. Are we interested in individual behavior or aggregate behavior?

iv. National elections? Mid-term elections? State elections? Local elections?

When selecting variables of interest we are slightly narrowing down the possible choices. We are also thinking about what level of analysis (and/or cases) best addresses our research question.
Identify the evidence that tells us about the variables: Operationally Define the Variable
1. What are some common ways that other academics measure variables you have chosen?

i. We find these things by looking through our sources/references, and identifying what those researchers have done.

ii. For example, if you are conducting an individual-level analysis of survey data of whether or not someone voted, then a simple question asking “did you vote in the [latest] election” with the choices [no], [yes], [don’t know, refuse to answer, etc.] is the most common way of measuring a variable of individual voting behavior.

We will discuss reliability and validity in another lecture, but note that when your operational definitions and data sources are in line with the existing literature, you have an easier time arguing validity and reliability.
In this process you are generating an operational definition of your concept. An operational definition is simply identifying the common way that this variable is measured—specifically for your unit of analysis.

i. Look back at the discussion of units of analysis.

Example: Using National Survey Data:
1. Individual respondent data: Individual level
2. Aggregated to county median/mean: County Level
3. Aggregated to state median/mean: State Level
4. National median/mean: national level

ii. Your operational definition should accurately reflect the content of the conceptual definition AND should be identified by its level of analysis.

For surveys it is a survey question and response; for aggregated data it may be a percentage, a rate, a raw number (as a particular level of analysis)
See the map on this page. Essentially, it is giving us bits of aggregated information about counties in the United States. The information is the % of a county population (some of these % are sub-populations)
Which type of measurement will accurately reflect the variable as we defined it in the operational definition?

i. You want to choose a measurement that reflects the conceptual definition and operational definition of your variable.

ii. If the concept is political participation, then you do NOT want to use a measurement of whether or not someone worked on a community project (that is measure of community participation).

Identify the actual data we will use to measure the variables: choose a measure from a data set
1. Where an operational definition identifies the type of measure, the measurement we choose is specific to a particular data set.
2. Surveys? Reported turnout?
3. Where will we obtain the data? Who created the data set?

2.011_concpet_to_measure.jpg

Identifying the Concept

Concepts

Have no objective meaning
They are labels we put on phenomenon
- These labels help us to link observations of cause and effect
- They also enable us to make generalizations about things such as power, participation, etc.

Similar to the Specification Process from Theory to Hypothesis

In going from a concept to a measurement, you are going from a general idea to a specific aspect of that idea

Conceptual Definition (Academic Definition)

An agreed upon meaning of a concept
- Among professionals, politicians or academics
- How do we know what is the agreed upon definition of a concept?
  - The literature review: we look to how the people in our sources have defined these concepts.
  - By doing so, we are “standing on the soldiers of giants” and using established definitions.
  - We are not reinventing the wheel here, we are using the existing knowledge to test our ideas
  - How well does the definition of your concept address the research question?
    - It ultimately influences:
    - The choice of variables (dependent and independent)
    - The choice of measures (data)

Variables of Interest

There are many variables that fall within a given concept. Which one are you interested in examining.
While we may have a conceptual framework for political participation, there are many types of political participation. We choose one of those—for example, voting.

Operational Definitions

Now that we have a variable of interest selected, we narrow down even further into a specific level of analysis and operational definition.
As with the conceptual definition, look to the literature for common operational definitions of your variable of interest.
For example, with a variable like voting
- We can use a survey question of whether or not someone voted
- We can use aggregated turnout at the district, city, county, state, or national level
- Our choice is determined in part by the level of analysis we are conducting. In some cases, there are multiple operational definitions of a concept at a given level of analysis
  - For example, what if our variable of interest was the strength of teachers’ unions in US states?
    - One could use the amount of money teachers’ unions/associations donated to state political campaigns, or the number of members belonging to the state group, or a measure created using a survey of state policymakers asking them about group strength, or some other measure we create with a combination of these.
    - NOTE: It is important to understand why this all matters. The choice we make here influences our results later. From the discussion of ways to measure group strength, we may get different results based on which of those we have chosen.
      - This is why looking to the literature is particularly important. We want to be aware of any problems past researchers have identified with each possible measure. We want to establish that our choice is the best one possible.

Measures (Operational Definitions):

Operational definitions are the criteria past researchers have used to select their measurements, measurements are the specific pieces of data that they use from a specific data set.

How are the variables going to vary?
Which type of measurement is most appropriate for my research question?

Are there multiple types of measures that are appropriate?
If so, do I have a preferred type of measure

Why measurement choices matter:

They influence the available statistical tests
They influence results of statistical analysis

You need to understand the type of measurement you are using and how it is coded in order to correctly evaluate your statistical relationships.