6.3: Conceptualization and Operationalization

At the beginning of any observational research study, the researcher must decide what exactly will be observed from the individuals or items in the population. This process is known as conceptualization. Conceptualization is an important part of any research study, and poor conceptualization can often result in weak results that can be easily contradicted by other studies (Blalock 1982). This results in general confusion in the research community about what the important issues really are, and in the broader community, the entire research field can seem unreliable and contradictory. This problem makes it exceedingly difficult for well planned research with important conclusions to be recognized.

Conceptualization can be particularly challenging because many of the variables of greatest interest can be extremely difficult to measure. In many cases research studies deal with abstract concepts for which there is not a specific method for measurement. For example, a researcher may wish to study the anxiety levels of children in a public school. Because there is no direct method for observing the anxiety level of a child, such a measurement must be accomplished indirectly. Other data may not be available in the historical record. For example, we might be able to design a method for measuring the anxiety level of a child, but there is no way to go back and make these measurements on children from previous years. Some types of data may be expensive to collect. In a large study of the effect of exercise on health, it would be cost prohibitive to have every study participant go through a complete medical examination for the sole purpose of the study. There may also be ethical or legal reasons that some data cannot be collected. Studies that consider health related variables, in addition to being expensive, may be hampered by privacy concerns. Still other variables may be difficult to measure because of societal reasons. An invasive health study may offer a small payment in exchange for participating in the study, but such an incentive is unlikely to motivate wealthier individuals who may form an important comparison group to participate in the study.

These problems have profound implications on the application of empirical studies. In some cases, researchers may simply decline to attempt to measure such variables, while in other cases researchers may simply side-step this issue and consider less complicated research questions based on data that is easier to measure. These are unfortunate approaches considering the relative importance of many of the related issues in empirical research. On the other hand, researchers may simply seek out convenient measurement methods, stating that the variable is difficult to measure and so they will choose an arbitrary method for measuring it, which is better than having no study at all. This type of approach is scientifically dangerous as many convenient methods of measurement may not really measure what the researcher is assuming that they measure after all. This can lead to erroneous results. Even in the case where the results may be correct, the perceived reliability of the study can be fundamentally damaged by the lack of care taken with selecting the measurements.

While the best approach for researchers is arguably to face the problem and do the best they can with the data they can observe, the researcher is obligated to defend and justify how abstract concepts are measured. The researcher is also obligated to admit to the shortcomings of the study, and to further explore ways that these shortcomings can be addressed.

In considering the development of measurement systems, all measurement methods, even the ones used regularly in the scientific community, are necessarily indirect measurements of physical phenomenon. For example, the conceptual idea of temperature can be measured using a mercury-based thermometer. Such a device does not measure temperature directly but is rather based on a physical model that describes the physical behavior of mercury in a glass capillary. By observing the physical reaction of the mercury in the vial, we are really measuring—indirectly—the volume of the liquid metal using a known physical model to infer the corresponding temperature. Similarly, the conceptual idea of mass can be measured by how much a spring is compressed when put under the weight of an object.

In the social sciences the physical model is replaced by one that is more abstract and more complicated, and about one for which much less may be known. Therefore, researchers should be cautious. At the most basic level the conceptual idea of a physical and social model is very similar. Consider a basic cause-and-effect type relationship in which there are stimuli acting on a system that evokes a response (Trimmer 1950). This simple idea is diagrammed in Figure \(\PageIndex{1}\). Within the system additional sub-systems could interact with one another to produce a response. In this case researchers may be able to observe the stimuli but may be unable to observe the interactions between the sub-systems. This causes the system to appear to be very complex. It is often the case in social science research that these sub-systems and their interactions may be unobservable, and in this case the system is seen as a type of black box (or in this case a blue box), where stimuli are interpreted by a very complex, invisible process to produce the response.

Figure \(\PageIndex{1}\): Diagram of stimuli acting on a system to invoke a response (Public domain image created by Alan M. Polansky).

This type of black box system is fraught with difficulty for those who wish to make connections between the stimuli and responses. It is quite possible that some, or even all, of the stimuli we can observe do not have a direct bearing on the system but are interacting with unobserved stimuli that do. Similarly, even if we can observe all the stimuli that affect the system, the complex interactions within the unknown system may make it impossible to learn about how the stimuli affect the system. This is made even more difficult by the fact that researchers usually do not control the stimuli. That is, they observe the stimuli simultaneously with the response.

An example of a simple system can help emphasize the context of these issues. Consider a stimulus that can have one of two possible values: 1 or 2. From the black box we then observe whole numbers between 1 and 10. Some example observations from this process are given in Table 6.1. The relevant question is, what does the black box do with the stimulus to get the response?

Table 6.1 Values of a stimuli that can have one of two possible values (1 or 2) and the resulting response from a black box.

Stimulus (𝑥)	2	1	2	2	1	1	1
Response	6	4	7	3	3	3	4

From Table 6.1 it is impossible to predict what will happen for each possible value of the stimulus. When the stimulus is equal to 1, we observe 4, 3, 3, 3, and 4, whereas when the stimulus is 2, we observe 6, 7, and 3. The only hint seen in these observations is that when the stimulus is equal to 2, the numbers tend to be larger. However, given the variation in the observations, most people would not have too much confidence in saying that this general trend would hold if we were to have access to further observations. It is simply impossible to say for sure what happens in the black box.

The back box is actually very simple in this case, but we have not been allowed to see all the stimuli. It turns out that there are two other stimuli, which we will call \(y\) and \(z\). What happens in the black box is that if \(x=1\), then the response is simply equal to \(y\). If \(x=2\) then the response is equal to \(x+y\). This can be written in a mathematically succinct way as \[response = y+(x-1)\times z.\] Table 6.2 shows all the stimuli and the related responses. Note that once all the stimuli are known, predicting the response if very simple. However, if these stimuli are unknown, the behavior seems very complex. That is, if only \(x\) is known, then the problem of predicting the response is intractable.

Table 6.2 Values of stimuli and the resulting response from a black box.

Stimulus (𝑥)	2	1	2	2	1	1	1
Stimulus (𝑦)	3	4	3	1	3	2	4
Stimulus (𝑧)	3	2	4	2	5	2	5
Response	6	4	7	3	3	3	4

In practice we can think about how stimuli may operate in a larger complex system to produce the data we observe. For example, we may postulate that race and ethnicity influence income, but of course the relationship is not a direct relationship. That is, all the people with the same race do not have the exact same income. Hence, there must be other stimuli, in this case probably a large number of stimuli, that also affect income. These stimuli could also possibly interact with race. A researcher must carefully analyze what these stimuli may be, determine if they can be measured, and consider whether they may also be related or associated with the levels of the other stimuli. In some cases, researchers may observe a variable that might not have any direct effect on the response but is merely associated with it.

To see how this can happen, consider a second simple black box example. In Table 6.3 we have a potential stimulus for a response. In this case the stimulus \(x\) that can be a whole number between 0 and 5, while the response is a whole number between 0 and 10. While larger values of the response tend to be associated with larger values of the stimulus, it is not possible to predict what the response will be based on the stimulus \(x\). For example, the stimulus is equal to 3 for three of the observations, but the response is equal either to 8 or 7. The black box is again very simple, two unobserved stimuli \(y\) and \(z\) whose numbers are between 0 and 5 that are added to get the response. The potential stimulus \(x\) does not have any effect on the response but is itself also a function of \(y\) and \(z\), as \(x\) is the smaller of these two numbers. See Table 6.4. In terms of prediction, if we knew the values of \(y\) and \(z\), we could predict the response perfectly. If we only know \(x\), then we cannot predict the response perfectly. In this case \(x\) is really a second response, not a stimulus, though in practice we may not be able to always distinguish between the two. More about stimuli that cause, or are simply associated with, responses will be considered later when we study association and causation.

Table 6.3 Values of a stimulus that can be a whole number between 0 and 5 and the resulting response from a black box.

Stimulus (𝑥)	3	3	0	3	1	2	2
Response	8	7	3	8	4	5	5

Table 6.4 Values of the stimuli and the resulting response from a black box.

Stimulus (𝑥)	3	3	0	3	1	2	2
Stimulus (𝑦)	5	3	3	5	1	3	2
Stimulus (z)	3	4	0	3	3	2	3
Response	8	7	3	8	4	5	5

In the research sciences the process of hypothesizing what stimuli and processes are contained in the black box is called conceptualization. If a researcher wants to understand how a response behaves, they need to conceptualize what stimuli may be acting together to elicit the response from the system. This is a complex process that is never complete or perfect, but researchers attempt to do the best they can. Note that the researchers do not necessarily intend to measure all the stimuli, and indeed many stimuli will never be known. The purpose at this point is to understand the system as well as possible from a conceptual viewpoint.

To begin a more formalized development of conceptualization, we begin by classifying the types of observations that are often used in research studies. The factors or observations that researchers can measure are often classified into three categories (Kaplan 2017). The first classification specifies measurements that can be observed directly.

Definition: Direct Observable

A variable is a direct observable if the observation can be measured directly and easily.

Examples of variables that are direct observables include data that can be obtained by visual inspection. For example, the number of applicants who show up on time for a job interview. When meeting with an individual participating in a study, an interviewer may be able to determine gender identity, hair color, skin tone, and the height of the individual. Any data that must be obtained through additional processes, but that can still be observed by the researcher, are known as indirect observables.

Definition: Indirect Observable

A variable is an indirect observable if the observation can only be measured through an additional process.

Indirect observables are usually inferred from secondary measurements. Nearly everyone has witnessed a partial solar eclipse, but few have observed it directly. Unless one has specialized eye protection, such as a welder's helmet, the usual method of observing solar eclipses is through a pinhole apparatus where one observes the light projected onto a flat surface (see Figure \(\PageIndex{2}\)). When using this method, the eclipse is not observed directly; rather, the effect of the eclipse is observed by the shape of the projected light.

Figure \(\PageIndex{2}\): Diagram of a simple device for observing a solar eclipse (Public domain image reated by Alan M. Polansky).

Radiocarbon dating provides another example. This method is used to approximate the age of an object that contains organic material. The actual age of the object is not observed but is inferred by the amount of carbon-14 remaining in the object.

In research studies indirect observables may include variables such as gender, ethnicity, and age. For example, gender may be self-identified by an individual, which is directly observed. In other cases, gender identification may be inferred by observing the manner of dress and social interactions of an individual, which is indirectly observed. Note that there is a chance of misidentification with this type of indirect observation. In the social sciences indirect observables refer to any observations taken from secondary material. For example, using birth records to determine the age of individuals in a survey.

Some types of data cannot be observed at all. These types of data are called constructs.

Definition: Construct

Constructs are variables that cannot be observed either directly or indirectly.

Constructs are often abstract ideas whose approximate values may be inferred from direct and indirect observables. We previously considered measuring microaggressions against Asian Americans, and the effect that these microaggressions can have on the victims’ health. Both variables, microaggressions and health, cannot be observed. However, as we discussed, both variables can be inferred to some extent from other measurements.

Moving from a construct to more practical matters for research brings us to the idea of a concept. A concept is essentially a construct with the addition of a well-defined meaning. For a construct, this well-defined meaning must be based on direct or indirect observable measurements. The process of moving from the theoretical abstraction of a construct to the well-defined abstract concept is called conceptualization.

Definition: Concept

A concept is a construct that has been clearly defined in terms of ideas that can be observed either directly or indirectly. The process of developing a construct into a concept is called conceptualization.

For example, the construct of healthiness is meant to convey the theoretical abstract idea of the general health of an individual. The process of conceptualization develops this abstract idea into the concept of a self-evaluated measure of health. This is a concept that is observable. Initially, the concept of microaggressions can be conceptualized to a self-assessment of perceived microaggression encounters over a specified period.

Conceptualization can be a difficult and complex process, and the result can be complex as well. It is important to consider how a construct is conceptualized, and the researchers publishing the results of a study should provide some details about the conceptualization of a construct. When researchers consider complicated concepts, there may be many specific aspects of a concept that are considered. That is, there may be many parts of the concept that need to be considered separately. Each aspect of a concept is called a dimension.

Definition: Dimension

A particular aspect of a concept is called a dimension.

If we think about conceptualizing general health, we might consider the health of an individual to have two aspects: physical health and mental health. Each of these aspects is a dimension of health. A researcher would have to decide what indicators there may be for each of these dimensions. For example, physical health might include indicators for high blood pressure, body mass index, alcohol and drug use, and high cholesterol levels. Similarly, indicators of mental health might include depression, anxiety, and self-destructive behavior.

Definition: Indicator

A particular aspect of a dimension is called an indicator.

As another example, consider a study that looked for differences between Mexican American and Anglo-American families in their attitudes toward money (Medina et al. 1996). For this purpose, they used the modified money attitude scale (Yamauchi and Templer 1982). The first dimension measures the perceived power and prestige that an individual attaches to money. The second dimension measures how one feels about saving money and making and enacting financial plans. The third dimension measures the amount of anxiety and distrust that money causes is one's life. The fourth dimension measures whether the individual perceives that money is associated with quality.

Within each dimension of a concept, indications about the value of the dimension are likely to exist. This is essentially a list of what characteristics would contribute to a high or low measurement for the dimension. In some cases, indicators consist of yes or no evaluations, so that an individual either has the indicator or they do not. In other cases, indicators are often based on a Likert-type scale.

In the example considering the study that looked for differences between Mexican American and Anglo-American families in their attitudes toward money, four dimensions were identified to measure the concept of attitude toward money. The first dimension measures the perceived power and prestige that an individual attaches to money. This dimension is defined to have eight indicators as shown in Table 6.5.

Table 6.5 The eight indicators of how an individual views power and prestige with respect to money (Medina et al. 1996).

Indicator	Statement
1	I tend to judge people by their money rather than their deeds.
2	I behave as if money were the ultimate symbol of success.
3	I find that I show more respect to those people who possess more money than I do.
4	I own nice things in order to impress others.
5	I purchase things because I know they will impress others.
6	People tell me that I place too much emphasis on the amount of money I have.
7	I enjoy telling people about the money I make.
8	I try to find out if other people make more money than I do.

One can observe from Table 6.5 that the indicators are represented as questions to which a respondent can register an amount of agreement. Indicating more agreement to any of the indicators provides evidence that the individual considers money to be associated with high power and prestige. These statements include how the individual views their own prestige, as well as the prestige of others as associated with wealth. The more indicators that an individual agrees with is an indication that the individual considers those with more wealth to have more power and higher prestige.

The second dimension measures how one feels about saving money and making and enacting financial plans. This dimension is defined to have seven indicators, as shown in Table 6.6. Note that most of these indicators ask whether the individual tends to save money, plan on financial matters, and use a budget. The statements are designed so that the higher scores from the indicators mean the individual considers planning and careful wealth management to be important.

Table 6.6 The seven indicators of how an individual views saving and planning with respect to money (Medina et al. 1996).

Indicator	Statement
1	I put money aside on a regular basis for the future.
2	I do financial planning for the future.
3	I save now to prepare for my old age.
4	I have money available in the event of an economic depression.
5	I follow a careful financial budget.
6	I am prudent with the money I spend.
7	I keep track of my money.

The third dimension corresponds to the amount of anxiety and distrust that money causes is one's life. This dimension is defined to have eleven indicators, as shown in Table 6.7. These indicators tend to assess the individual's feelings about purchasing with an emphasis on the cost of items. The statements are designed so that the higher scores on the indicators mean that monetary transactions tend to cause distrust and anxiety for the individual.

Table 6.7 The eleven indicators of the amount of anxiety and distrust that money causes is one's life (Medina et al. 1996).

Indicator	Statement
1	It bothers me when I discover I could have gotten something for less elsewhere.
2	I complain about the cost of things I buy.
3	I show worrisome behavior when it comes to money.
4	I worry about not being financially secure.
5	When I make a major purchase, I have suspicion that I have been taken advantage of.
6	I show signs of anxiety when I don't have enough money.
7	After buying something I wonder if I could have gotten the same thing elsewhere for less.
8	I hesitate to spend money, even on necessities.
9	It is hard for me to pass up a bargain.
10	I automatically say “I cannot afford it.”
11	I am bothered when I have to pass up a sale.

The fourth dimension measures whether the individual perceives that money is associated with quality. This dimension is defined to have five indicators, as shown in Table 6.8. These statements measure the willingness of the individual to spend money to gain the associated perceived quality. Higher values of the indicators indicate that the individual considers high cost and good quality to be synonymous.

Table 6.8 The five indicators of whether the individual perceives that money is associated with quality (Medina et al. 1996).

Indicator	Statement
1	I am willing to spend more to get the very best.
2	I buy top of the line products.
3	I buy name brand products.
4	I pay more for some things because I know I have to in order to get the best.
5	I buy the most expensive items available.

The final phase in developing methods for gathering information about a construct or a concept is operationalization, where the researcher develops the procedures that will result in the empirical observations that will be used in the study.

Definition: Operationalization

Operationalization is the process of developing the procedures that will result in the empirical observations representing the concepts.

Through the process of operationalization, the researcher will decide how to score the responses from a survey, what the range of the observations will be, and what level of measurement will be used. For concepts with multiple dimensions and multiple indicators, the researcher must make important decisions on how a composite measure of the concept will be computed from the individual indicators and dimensions.

Returning to the study of attitudes toward money, the operationalization of the four dimensions and their associated indicators are taken together to form what is known as the Money Attitude Scale. The implementation of this schedule relies on everyone responding to each indicator statement using a Likert-type scale with a numerical range from 1 to 7, where 1 means the behavior “never applies” and 7 means it “always applies.” The midpoint value of 4 indicates that the behavior sometimes applies to the individual. The responses from each of the dimensions are summed to get a total score for that dimension. These values indicate how strongly each dimension applies to the individual, and the total score from all the dimensions indicates the relative importance of money for the associated individual.