Levels Issues II — Do Levels Exist? | Organizations and Markets

Levels Issues II — Do Levels Exist?

4 December 2006 at 5:05 pm Nicolai Foss 8 comments

| Nicolai Foss |

As I indicated in my earlier post on levels issues in social science research, I am confused by these and I suspect that many others are also confused. Perhaps this merely reflects my lamentable lack of serious philosophical training, and it is therefore with very considerable hesitation that I venture into issues of ontology, explanation, and causation that pertain to levels of analysis. (In fact, the following to some extent has the character of a bleg).

Do levels of analysis exist? Well, obviously levels of analysis only exist in our models. Still, there may be some stuff reality that is “like” our analytical levels. If so, is there some kind of mapping from the levels of analysis of our theoretical accounts to the levels (conceivably) existing in social reality? Or, are levels (of analysis) “merely” methodological devices — features of our model — that are not necessarily mirrorred by anything in reality?

Certainly if you are the kind of sociologist who is engaged in “discourses” on “structures” (yes, I know that not all sociologists are like this), then surely “levels” exist. The same goes for Marxists and for political science scholars of the Skocpol variety. In fact, if you are such a person, a “structure” placed at some level above the individual actors may be an independent causal agent. In such a view “downward causation” is something very real.

“Levels” are, of course, phenomena that are somehow ordered in a vertical dimension; they are inherently hierarchical. For a levels distinction to make sense, there must be some qualitative difference at the different levels. Thus, industries are qualitatively different from firms that again are qualitatively different from the input owners that contract with the firm nexus. This seems entirely common-sensical. However, there are subtle differences between the various levels. Thus, industries are not legal persons, and it is pretty meaningless to argue that industries can be causal agents (of course, industries can “exist” as intersubjectively held representations). While firms can be legal persons, they cannot act. Only individuals can act.

The implications of this familiar methodological individualist/reductionist position for levels issues seem to me to be potentially radical:

The idea of structures as causal agents and of downwards causation is highly problematic. (As readers of James Coleman’s Foundations of Social Theory will know Coleman argued that institutions could impact the conditions of individual action (i.e., they are the rules of the game), but this does not seem to me imply any idea that structures are causal agents). As philosopher Daniell Little says in this paper (p.26), “[t]here are no causal powers at work within the domain of the social that do not proceed through structured individual agency.”
Supra-individual entities such as “the industry,” may exist as intersubjectively categories of meaning, but this still does not equip them with causal powers.
When individuals interact under positive transaction costs, externalities arise. For example, interaction among individuals in a market may produce phenomena that “the results of human action but not of human design.” Are these “emergent” phenomena on a level of analysis that is “higher” than that of individual agents? Why exactly? Because they are “emergent”?

Entry filed under: - Foss -, Methods/Methodology/Theory of Science.

You Know You’re Nobody If. . . Best Business Movies

8 Comments Add your own

1. Bo | 5 December 2006 at 12:27 am

I am not sure I follow the argument here. You seem to be implying that only causal agency matters (in reality) and thus only individual levels are real? The reason why different levels of analysis is important (beyond for methodological reasons) is because each level may influence (impact) other levels. Industries and institutional environments DO act in the sense that they put constraints on firms within (and outside) them – hence they influence in a very real way business conduct at the firm level, which in turn, influences how individuals within these firms can (and will) act. If you observe “real” business interactions you will find evidence of the systemic and interdependent nature of these multiple levels at play.

In my opinion, multi-level research is not problematic – it is indeed much closer to reality in allowing us to tease out (separate) the exact influence at each level rather than simply look at gross aggregations as has been the norm.
2. brayden | 5 December 2006 at 1:01 am

I’ll quibble with you on the point that organizations (aka legal persons) cannot act. For all intents and purposes, organizations act all the time. When a worker is fired from GM, we don’t say that Rick Wagoner (GM’s CEO) fired the worker. We say that GM fired the worker. The action of an agent of the organization is, practically speaking, an action of the organization itself. Why should we care if the organization does not have its own mind that propels action? The agent would never have fired someone in the first place if he or she were not acting on behalf of the organization.
3. Peter Klein | 5 December 2006 at 1:19 am

Brayden, this is precisely the point at which economists and sociologists tend to disagree. To a methodological individualist (i.e., to most economists), the point is precisely that GM _doesn’t_ act. We may say, “GM fired the worker,” but this is just a figure of speech. Likewise, the statement “Germany invaded Poland” is shorthand for a complex set of decisions by individual actors. Such language may be useful, as a loose description, but not as a means of explanation. We can’t fully understand WWII — what it was, what were the causes and consequences — without reference to the beliefs, objectives, and actions of individuals.

Of course, individual actors are embedded in a social context, and that context must be taken into account. Private Schmidt wouldn’t have crossed the border if he weren’t part of the German Army. But we can’t possibly understand the phenomenon without taking people like Private Schmidt into account.
4. levels of analysis « orgtheory.net | 5 December 2006 at 2:49 am

[…] Nicolai at O&M has a post on levels of analysis-related matters. The topic is central in organization theory (and management research more generally) – no matter what phenomenon one is interested in – identity, creativity, advantage, learning, knowledge – one has to be sensitive to underlying assumptions about 1) the individuals that compose the organization, 2) the unique organizational component (if any? emergent? – more on this critical, missing component later), and 3) the environmental and contextual (industry, society, culture) matters that need to be considered. […]
5. brayden | 5 December 2006 at 10:55 am

I hate to just agree to disagree, and so I put a response of sorts in the comments to Teppo’s post.
6. Bo | 5 December 2006 at 12:39 pm

Peter – it seems to me that what you are saying is exactly that levels DO matter in reality. Although individuals act they do so within a context – this higher level also needs to be accounted for in our analyses (in fact most of the time this is the focus of our analyses). Thus – multi-level issues are real phenomena that needs attention.
7. Nicolai Foss | 5 December 2006 at 3:47 pm

Bo,
I think that upon reflection you will realize that “industries and institutional environments” do in fact NOT act. These collective notions are shorthand for complicated patterns of individual action and interaction. Think of a firm contemplating entry. It may come to the conclusion that the “industry may fight a price war in the case of entry,” but of course this is just a shorthand way of representing the explicit or implicit agreement among incumbent firms that a price war should be fought.
To be sure, industries can (therefore) exist as mental representations, and these can very well be shared among many people (more or less, Searle’s notion of an institution). But representations of this kind do not have causal powers.
Contrary to you I do think there are certain important issues that need to be sorted out in connection with multi-level analysis. The idea that there is a vertical relation (i.e., levels) is one that I am increasingly having problems with. In what sense — other than a trivial aggregation sense — is an industry on a “higher level” than a firm? We may be conflating representations of reality with reality here. Perhaps all that really exists on the social domain is one big horizontal web of individuals, their relations and interactions, and their interpretations and representations thereof.
But I guess what is needed is a serious analytical philosopher to deal with these issues.
8. Bo | 6 December 2006 at 1:52 pm

Perhaps we are not so far from each other after all. I do AGREE (in priciple) that institutions etc do not act – only individuals act in this sense – however, that does not mean that institutions etc are not having an impact/effect on individuals and vice versa – this is what multi-level theorizing and testing may help us sort out (I think). In terms of the vertical metaphor – you may be right that it does not necessarily make sense to talk about industries and clusters etc as a “higher” level except if we define this as a higher level of aggregation (which is indeed what most people do). However, since the methodologies are still lacking behind reality (and theorizing) in this domain, we may need to use this abstraction for now – HLM simply needs you to think of levels hierarchically!

The argument for an individual web of interactions is interesting but does not really convince me (yet). Institutions do exist and they DO put restrictions on individual decision making (e.g. managerial discretion literature). The corporate governance literature, for instance, clearly establishes different roles and responsibilities of different agents – depending on insititutional influences above and beyond the individual relations and interactions. Of course, if you really wish to reduce everything to the individual level you can in the extreme but I question the “reality” in this? If you are right then we all need to simply study cognitive psychology to answer all questions…