Levels of analysis

Consider if you were a photographer who has the perfect zoom lens and a great wide-angle lens. You can choose to photograph things in extreme detail or you can choose to photograph a big picture that provides a broader context. Both views highlight different aspects of what you are trying to capture. And both views in combination provide a more complete picture of what you want to capture. At the same time, you can choose a level of zoom in between, depending on what you want to capture. In research, the availability of this choice is referred to as levels of analysis.

Choose your lens

Similarly, any phenomenon of interest can be studied at different levels. For example, organizations and people within the organization can be studied at the organizational level (how they perform as an organization), at the department level, at the team level, or individual level. In the same way, a social phenomenon like racism can be studied at different levels. It can be studied at the individual level, organizational level, institutional level, constitutional level, and societal level. Every level provides a different insight into the issue of racism and thus makes our understanding of the problem better.

In general, a social science study can be set at the micro level when individuals are analyzed. Or it can be set at a higher meso level where aggregates of individuals such as households, wards, precincts, firms, and neighborhoods are studied. Or at an even more macro level such as communities, counties, provinces, states, or nations are analyzed. What you analyze becomes the unit of analysis and the structural level, spanning the range from most micro to the most macro, at which a social scientific investigation is carried out is called the level of analysis (micro, meso, or macro).

The obvious reason to study any phenomenon at multiple levels of analysis is to gain a holistic and interrelated understanding of a complex phenomenon compared to the one that is obtained from the perspective of a single level of analysis. Quantitative methods are generally well suited for a more macro level of analysis whereas qualitative methods might produce greater insights at a more micro level. A mixed-method research design can be a useful methodological tool to analyze and integrate multiple levels of analysis.

Examples of levels of analysis

Here is an example of substance (ab)use that can be studied at various levels of analysis starting from the molecular level to societal, cultural, and political level.

Level of Analysis	Example of research study for substance use
Molecular and/or genetic	Genetic determinants of substance use
Cellular	Tracing metabolic pathways of substance use
Organ	Effects of substance use on the kidneys, nervous system, skeletal organs, muscular systems, respiratory system, immune system
Behavioral and/or psychological	Effect of substance use on dependency (reducing important activities because of use of the substance) and abuse (driving while intoxicated, getting into trouble because of intoxication), Designing effective treatment programs
Social and/or environmental	Family and social network norms on substance use, social support, socioeconomic status, role of poverty, stigmatized identity, availability and access to illicit drugs, exposure to discrimination and/or racism
Cultural and/or political	The cultural context of illicit drug usage – Lack of protective norms, media portrayal of substance use, references in music, myths, potential legal consequences

Another example could be leadership. Leadership is a multifaceted concept that can be studied at the individual level, team or group level, dyadic level, organizational level, and even at the national level. Individual level of analysis may look into personal traits, behaviors, and all sorts of individual differences that distinguish the leader. Dyadic relationships focus on superior-subordinate, leader-follower, or interpersonal relationships. Group or team-level leadership studies may focus on groups as units of analysis and the emergence of leadership qualities that enable teams to achieve their stated goals. Organizational level and national levels studies of leadership might focus on a leader’s ability to initialize and maintain organizations/movements to accomplish its stated mission.

Plurality

Different levels of analysis produce different insights essentially integrating multiple explanations and producing multiple models fostering pluralism instead of reductive explanations of a phenomenon of study. This points towards the fact that there are multiple ways of representing (even if it is partial) a phenomenon thus making knowledge generation a dynamic and emergent process. This is because each level within the system might be governed by its own laws. That is, at the macro level systems might have properties that their subsystems (micro level) might lack. Thus each level of analysis can provide complementary information about the phenomena under study.

Level of analysis and Unit of analysis

Determining the level of analysis is usually straightforward. Your unit of analysis (what you are studying) determines your level of analysis. If you are studying individuals, your level of analysis is at the individual level. If you are studying relationships between team members, your level of analysis is groups or teams. For example, to study teamwork in organizations you can survey individual team members in different teams and average their individual scores to create a team-level score for team-level variables like conflict, goal clarity, influence, participation, and so on.

Cross-level inference – Ecological fallacy

The plurality of models/theories also raises concerns about whether we can combine different levels of analysis, often framed as individual and interpersonal versus structural levels. There is a great interest in the interrelationship between these levels of analysis. In simple terms, we can distinguish our level of analysis by our unit of analysis. But can we draw inferences from one level of analysis to another? This is what is referred to as cross-level inferences.

Cross-level inferences draw conclusions at one level of analysis from research done at another level of analysis. The most common is to conclude a macro or aggregate level of analysis to a micro or individual level of analysis. However, relationships at the aggregate level might not hold true at the individual level. Such an error of inference is called an ecological fallacy.

Ecological Fallacy
Ecological fallacy is the failure of logic or logical fallacy in the interpretation of statistical data where inferences about the nature of individuals are deduced from inference for the group to which those individuals belong.

For example, in the USA, wealthy residential districts are positively correlated with better school outcomes. This association can give rise to ecological fallacy if we conclude that wealthy residents contribute to better school outcomes. When in fact it just means that wealthy residents are more likely to choose to live in residential districts where school outcomes are already better. By appropriating the school-level data (aggregate) to wealthy residents (individual) (different units of analysis) we commit the ecological fallacy.

Since school-level data is an aggregate measure of groups of students by districts, the individual outcomes of students might have disappeared. Additionally, such grouping may add its own effect (such as preexisting differences in the school outcome) over and above the individual variable. Here residential wealth might serve as a proxy for better outcomes. Thus one needs to be careful against making inferences about the analysis of relationships from one level to another level. Cross-level interpretive errors can go both ways if we assume individual-level data always describes aggregate-level analysis. Such an error is referred to as the “individual differences fallacy”. One way to avoid such fallacies is to check whether you are inferring about the wrong unit of analysis.

Additionally, assuming that societal processes are mere composites of individual behavior, is also incorrect. Assuming this constitutes a compositional fallacy. Another example is stereotyping. Stereotyping is a way of representing and judging people based on group characteristics rather than being viewed as individuals with their own personal features and qualities.

The most interesting studies involve multiple levels of analysis.

From this discussion, it might seem like the researcher must choose one level of analysis and most likely stick with it. On the contrary, the most interesting studies involve multiple levels of analysis. For example, we might want to ask an individual-level question about residents’ economic wealth and an aggregate-level question about school outcomes. This might provide an opportunity to combine questions are different levels of analysis as well as their interaction.

Bibliography

Dooley, D. (2001). Social research methods (4th ed.). Prentice Hall.

Richards, J. M. (1990). Units of analysis and the individual differences fallacy in environmental assessment. Environment and Behavior, 22(3), 307-319. https://doi.org/10.1177/0013916590223001

Hansen, H. (Spring 2023 Edition). Fallacies. In E. N. Zalta & U. Nodelman (Eds.), The Stanford Encyclopedia of Philosophy. Available at: https://plato.stanford.edu/archives/spr2023/entries/fallacies