# Unit of analysis vs. Unit of observation

Having graded another batch of 40 student research proposals, the distinction between ‘unit of analysis’ and ‘unit of observation’ proves to be, yet again, one of the trickiest for the students to master.

After several years of experience, I think I have a good grasp of the difference between the two, but it obviously remains a challenge to explain it to students. King, Keohane and Verba (1994) [KKV] introduce the difference in the context of descriptive inference where it serves the argument that what often goes under the heading of a ‘case study’ often actually has many observations (p.52, see also 116-117). But, admittedly the book is somewhat unclear about the distinction and unambiguous definitions are not provided.

In my understanding, the unit of analysis (a case) is at the level at which you pitch the conclusions. The unit of observation is at the level at which you collect the data. So, the unit of observation and the unit of analysis can be the same but they need not be. In the context of quantitative research, units of observation could be students and units of analysis classes, if classes are compared. Or students can be both the units of observation and analysis if students are compared. Or students can be the units of analyses and grades the unit of observations if several observations (grades) are available per student. So it all depends on the design. Simply put, the unit of observation is the row in the data table but the unit of analysis can be at a higher level of aggregation.

In the context of qualitative research, it is more difficult to draw the difference between the two, also because the difference between analysis and observation is in general less clear-cut. In some sense, the same unit (case) traced over time provides distinct observations but I am not sure to what extent these snap-shots would be regarded as distinct ‘observations’ by qualitative researchers.

But more importantly, I start to feel that the distinction between units of analysis and units of observation creates more confusion rather than more clarity. For the purposes of research design instruction, we would be better off if the term ‘case’ did not exist at all so we could simply speak about observations (single observation vs. single case study, observation selection vs. case selection, etc.) Of course, language policing never works so we seem to be stuck in an unfortunate but unavoidable ambiguity.

## 10 thoughts on “Unit of analysis vs. Unit of observation”

1. Hi, In teaching the difference I found Barbara Geddes’ concept useful (Paradigms & Sandcastles, somewhere in ch. 3): Colloquial use of cases must not be confused with the methodological meaning of a case: a case or observation is a unit within which each variable measured can take only one value or is classified only in one category.

2. Yes, nice pointer, but still – if you have students clustered in schools, your observations are at the student level but the analysis could be at the school level. What do you call the schools then? A case? Second-order observation?

3. Nevermind, I guess what I described really is the difference between the unit of analysis and the unit of observation.

4. what would be the unites of analysis and units of observation in the following study: samples are selected by households in villages, so data is collected at the Household level, but questions are asked only to one individual in the household. I’m guessing units of analysis are individuals within households, and units of observation are households within villages…

what would you say?

• i would say unit of observation is the individual (that’s the level the observations are), but the unit of analysis is the household because that’s the level the inferences are made at

5. The unit of analysis *should be* the lowest level (as in multilevel modeling) at which observations are independent. If you have dyads (2 person groups) and the observations on the dependent variable are nonindependent (i.e., they are correlated within group) then dyad should be your unit of analysis. If you tested for nonindependence and found that dyad members were not similar (did not have correlated outcomes), then you could test your hypothesis using individual as the unit of analysis (i.e., ignoring the dyad level). Most statistical tests require independence of observations for the standard error (and therefore the p value) to be correct.

see Cook, W. L. (2011). Foundational issues in nonindependent data analysis. In Brett Laursen, Todd Little, & Noel Card (Eds.), Handbook of developmental research methods. New York: Guilford Press.

6. Can someone answer this for me? Yin describes two types of Single Case studies – Holistic (single unit of analysis) and embedded (multiple units of analysis). If I organise my survey participants (my units of observation) into groups where I end up with 8 different groups from the same set of participants, do I have multiple units of analysis (8 different groups) or do I have a single unit of analysis (a group) which is analysed 8 times as each version comprises different units of analysis?

7. HI there,

In my opinion the “Unit of Observation” is confused with the “Level of analysis.” In addition to this there are other dualism in social science that have had the students confused for decades, for example, Micro and Macro; Agency and Structure; Individual and Society; Heart and Mind; Subjective and Objective; Explaining and Understanding etc.

I will try to explain the Unit of Analysis, Unit of Observation and Levels of Analysis/Explanation below and do look forward to your comments:

Unit of Analysis: is the fundamental unit of interest in an investigation. This is the question that you want to have your investigation answer (not to be confused with the research question). This can be: “why did people voted for Obama?” The unit of analysis is the voting behavior. This behaviour can be analysed at different levels of aggregations, or the Levels of Analysis (this should in fact be called “Level of Explanation” as it is the level of aggregation at which a phenomenon is to be explained).

Therefore, we can investigate voting behaviour at the individual level, or at the group level. Since the “voting behaviour’ is an unobservable asocial phenomenon our Unit of Observation will have to be something that is observable, so it can give us a handle on measuring the Unit of Analysis. If our Unit of Observation is individual then we can interview the individual. The point here is that since we can not get to what we what to know directly, which is “voting behaviour”, we will have to choose a variable that can stand as proxy for our Unit of Analysis. Unit of observation is an indirect way of measuring the Unit of Analysis. If the unit of analysis is human feelings the unit of observation can be the individual or group etc. We can roughly measure human feelings by asking individuals questions about their feelings or we can take a magnetic resonance image (MRI) of their brain to see what they feel when they are exposed to different situations. For instance, in some countries, Pedophile that have served their sentences are not directly released from precision. They are, instead, monitored for their sexual tendencies in special detention centers.

One of the method that’s used to measure their sexual tendencies is by placing a very sensitive rubber band that is connected to a sensor around subject’s penis. The subject is then shown suggestive images of children. If these images arouse him the rubber band will record change in his penis size and the investigators will record him as being “effected” by the images, therefore, not safe for release into the society. In this example the Unit of Analysis is subjects sexual tendencies toward children and the Unit of Observation is subject’s erection.

Another example is Small and Singer’s COW dataset. They define “War as organised mass violence” and then they go to define ‘violence’ as “the taking of human life”. Here the Unit of Analysis is “organised mass violence” and Unit of Observation is “the taking of human life”.

Now what will be the Unit of Analysis depends on the research question. If i am investigating “the economic effects of US civil war”, then my Unit of Analysis will be the US Civil War. Which I will define in terms of temporal and spatial details before I set out to investigate it. The killing of people is of no direct interest to me here, instead my Unit of Observation, can be some economic variable, which will account for the economic changes due to the Civil War. From this point I can look for economics changes at household or individual levels or at the city or state level or at the country level. This will be my level of Explanation. Whatever Level of Explanation I chose for my investigation I will still have to identify appropriate Units of Observations for my investigation.

However, if i am investigating the causes of Civil War, I might look at the decisions of Confederates State governments, or the decisions made by the individuals. These are again different Level of Explanation. The Unit of Analysis can be individuals, or the groups.If the Unit of Analysis is an individual then my Unit of Observation can also be the individual (if they are still alive) or it can be recorded testimonies or previous interviews etc. These can be business level data that can explain how the abolition of slavery meant economic loss to the Southerners, therefore, under some theory about elites mobilize masses for war may guide our research.

Now how is Unit of Observation different from the Levels of Analysis/Explanation?

As explained above, the investigator may want to know how; people become sexual predators? What are the effects of Civil War? Why people go to war? These question can be answered at the individual level, as Weber would or at the group or structural level as Durkheim would define them in terms of human collectivities. Therefore, the choice of Unit of Analysis and the Levels of Explanation depends on the researcher’s social ontology. The Unit of Observation is not a question of choice of Levels or the unit of Analysis that the investigator chose to investigate a phenomenon. It is simply the question of looking for proxies that can stand for unmeasurable Unit of Analysis.

I hope this will help. Any critique?