Moderator and mediator in research Úvod
The continuous examination of the relationship between cause and effect has aimed throughout centuries toward humanity's effort to gain deeper insight into the world and its laws. At the scientific level, the question of cause and effect is also addressed by researchers in the field of human sciences.
By testing causal hypotheses, researchers not only verify their hypotheses or theories regarding the respective scientific phenomenon, but also find answers to practical questions about whether an intervention has or does not have the expected effect. However, as knowledge matures, researchers go beyond the scope of a "mere" cause-and-effect relationship and attempt to understand what bridges the causal connection and what determines the magnitude or direction of the causal relationship (Frazier et al., 2004, Rose et al., 2004).
The most common questions in research are formulated as follows: Does variable X predict or cause variable Y? This is the so-called basic type of research, the basic type of research question. However, progress requires going beyond "ordinary" findings in research and dealing with more complex relationships and connections between variables, thereby elevating our research.
A number of authors (for example Frazier et al., 2004, Rose et al., 2004, Verešová, 2004, MacKinnon, 2008, and others) in recent years in their socio-psychological or pedagogical-psychological research devote great attention to examining variables of a moderator and mediator nature. The conceptualization of these two terms differs among some authors. In the following subsections, we present a brief overview of the conceptualization of the terms moderator and mediator from various authors in their research works.
3. Moderator-type Variable
Generally speaking, a moderator is a qualitative (e.g., gender, race, class) or quantitative (e.g., level of reward) variable that affects the direction or strength of the relationship between other variables (MacKinnon, 2008).
Fig. 1 Model describing a moderator
According to MacKinnon (2008)
A moderator is a variable that changes the direction or strength of the relationship between a predictor and an outcome. In reality, this involves an interaction, that is, the influence of one variable on another that is dependent on a different variable. For example, if we are not only interested in determining the influence of social support on the level of depression occurrence, but also in identifying whether it differs depending on whether the person is male or female. Specifically, within the analysis, a moderator is a third variable that affects the relationship between two other variables. A moderator can be represented as an interaction between the central independent variable and a factor that determines the appropriate conditions for its functioning (Baron, Kenny, 1986).
Fig. 2 Influence of a moderator
According to Baron, Kenny (1986)
Figure 2 provides a description of how we examine each type of effect step by step. Researchers can use multiple regression when examining the effects of a moderator. For example, predictively, whether a moderator-type variable is categorical (for example gender or race) or continuous (for example age). When both the predictor and the moderator-type variable are categorical, we can use analysis of variance.
2. Mediator-type Variable
A mediator-type variable brings the so-called "message" into the relationship between two other variables. A mediator is often conceived as a mechanism through which one variable (the so-called predictor) influences other variables (the so-called criterion). When testing the effect of a mediator, the researcher can examine whether a third variable could represent or explain the relationship between these variables (Rose et al., 2004).
Baron and Kenny (1986) explain that the mediation model in statistics is presented as a mechanism or process that seeks to explain, name, or describe the identified relationship between the independent and dependent variable through the inclusion of a third explanatory variable. It is well known and often also called an intermediary of variables. The mediator variable serves to explain the relationship between independent and dependent variables.
Fig. 3 Model describing a mediator
According to MacKinnon (2008).
Frazier et al. (2004, according to Baron, Kenny, 1986) define mediation as a process that leads from the independent variable to the dependent variable. In other words, in a simple mediation model, the independent variable is the cause of the mediator, and subsequently the mediator acts on the dependent variable. For this reason, the mediation of an effect is also referred to as an indirect effect, surrogate effect, intermediary, or as a side participant in influence. MacKinnon et al. (2002, In Amery, Zumbo, 2007) described the following example of mediation: a drug abuse prevention program (the so-called independent variable), resistance to drugs (as a mediator), and subsequently drug-related behavior/drug use (the so-called dependent variable). In the given context, Verešová (2004a, 2004b) draws attention to the description of individual mediators of drug use, defining mediators of drug use as factors that act as intermediaries in a certain direction of polarized behavior, for example pro-drug vs. abstinent (i.e., also risky vs. non-risky). By mediator she understands, during ontogenesis, a changing factor predominantly of a psycho-social and personality nature, which is significant from the standpoint of directing behavior oriented in a certain direction within the relationship of an individual to a certain element of reality, and simultaneously the action of this element on the experience and behavior of the individual. Modifiable risk factors are then determinants of the course of development of drug addiction and the effectiveness of preventive action or intervention in the school environment. Verešová illustrates the mediation model in drug addiction prevention as follows (figure 4):
Fig. 4 Mediation model in drug addiction prevention (Verešová, 2004)
Rose et al. (2004) and Wengener and Fabrigar (2000, In Amery, Zumbo, 2007) present mediation as a causal model that explains the process of "why" and "how" causes and effects occur.
According to Baron and Kenny (1986), the general test for the existence of a mediator is to examine the relationship between the mediator and the criterion variable, the relationship between the mediator and the variables, and the relationship between the mediator and the criterion variables. All of these correlations should be evident. The relationship between the mediator and the criterion should be reduced (down to zero) after controlling for the relationship between the mediator and the criterion variables.
Another way to think about this question is that a moderator is a variable that has an influence on the strength of the relationship between two other variables. Thus, a mediator is the variable that explains the relationship between two other variables. Rose et al. (2004) explain a mediator as a message in the relationship between two other variables. It is often conceived as a mechanism through which one variable (the so-called predictor) influences other variables (the so-called criterion). When testing the effect of a mediator, the researcher can examine whether a third variable could represent or explain the relationship between these variables.
4. Moderator versus Mediator
Frazier et al. (2004) and Rose et al. (2004) state that the mediator and moderator are two tools whose understanding is associated with a kind of puzzle. A mediator is a third variable that has a relationship to both the effect and the cause. A moderator is a third variable that adjusts the cause of the effect. A causal model refers to a theoretical hypothesis about changes in one variable and the resulting changes in another. Testing a causal hypothesis requires examining whether the causal conclusion is that X causes Y and is viable (Wegener and Fabrigar 2000, In Amery, Zumbo, 2007). The scientists explicitly stated that there are three types of common causal hypotheses: a direct causal effect, a mediated causal effect, and a moderated causal effect. Amery and Zumbo (2007) define mediators and moderators as third variables whose purpose is to enhance, deepen, and more precisely understand the causal relationship between the independent variable and the dependent variable.
MacKinnon (2008) describes the relationship between moderators and mediators as examples of third variables in research. Most research focuses on the relationship between two variables — the independent variable X and the outcome Y (dependent variable) — which is statistically variable. Examples of two different effects are the correlation coefficient, odds ratio, and regression coefficient. With two variables, there is a limited number of possible causal relationships between them: X causes Y, Y causes X, and X and Y are mutually connected. With three variables, the number of possible relationships between variables increases. Essentially, this means: X may cause a third variable Z to cause Y; Y may cause both X and Z, and the relationship between X and Y may differ for each value of Z. Mediators and moderators are two types of third-variable effects. In the case where a third variable Z is an intermediary in the causal sequence such that X causes Z and Z causes Y, then Z is a mediating variable; it is in the causal sequence X → Z → Y. In the case where the relationship between X and Y differs at different values of Z, then Z is a moderating variable. The primary difference between mediation and moderation is that a mediator-type variable determines the causal sequence in that the mediator carries the causal effect of X onto Y, but a moderator-type variable does not determine the causal relationship, only that the relationship between X and Y differs at various levels.
Based on the authors' statements, we can conclude that the difference between a moderator and a mediator lies in the fact that a moderator-type variable determines when certain effects are achieved, while a mediator-type variable addresses why and how such effects appear. However, the same variables can serve either as a mediator or a moderator, or both, depending on the research question.
Author: Mgr. Lucia Foglová References
AMERY, D., ZUMBO, B. D. (2008). Understanding and Using Mediators and Moderators. [Electronic version]. Social Indicators Research Series, 87:367–392, DOI 10.1007/s11205-007-9143-1 BARON, R. M., KENNY, D. A., (1998). The Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations, [Electronic version]. Journal of Personality and Social Psychology. Vol. 51, No. 6, 1173-1182, 0022-3514/86/$00.75 BARON, R. M., & KENNY, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173-1182. MACKINNON D. P. (2008). Introduction to statistical mediation analysis. Mahwah, NJ: Erlbaum. MACKINNON D. P. (2008). Introduction to Statistical Mediation Analysis. Mahwah, NJ: Erlbaum. ROSE, B. M. et al. (2004). Mediator and Moderator Effects in Developmental and Behavioral Pediatric Research. [Electronic version]. Developmental and Behavioral Pediatrics. 0196-206X/00/2501-0058 VEREŠOVÁ. M. (2004a). Mediátory užívania drog : cesta k efektívnej prevencii drogových závislostí. Nitra : UKF, 394 s. ISBN 80-8050-767-8. VEREŠOVÁ. M. (2004b). Mediátory užívania drog – most od predisponovanej osobnosti k prodrogovému vs. protidrogovému správaniu. In. Sociálne procesy a osobnosť. Bratislava : SAV, ISBN 80-88910-16-1, s. 505-511