Coordination Problems in the Theory of the Firm
28 December 2010 at 11:49 am Nicolai Foss 21 comments
| Nicolai Foss |
Many textbooks (e.g., this one or this one) begin by noting that there are two fundamental problems of economic organization, namely the coordination problem and the motivation problem — and then devote 95% of the space to the latter problem. (In a paper published in 1993 (but written in 1989), I proposed that extant theories of the firm could be understood as taking either a PD (-like) game or a coordination game as the fundamental underlying structure of interaction. In my reading, capabilities theories were about coordination problems, while mainstream organization economics fundamentally started from PD-like situations; this paper develops the argument a little bit).
Important work has been done on coordination problems in the context of the theory of the firm by Colin Camerer and Mark Knez (e.g., here), Phanish Puranam and Ranjay Gulati (here), Luis Garicano (e.g., here), Birger Wernerfelt (e.g., here), and, of course, co-blogger Dick Langlois (check his CV here on O&M — most of his stuff on economic organization is about coordination). One could also make the point that large parts of traditional organizational design theory (of the information processing/contingency variety, including Marschak & Radner’s team theory) are really about coordination problems rather than motivation problems. Dick Langlois has long argued that Coase (1937) is fundamentally about coordination rather than motivation.
This is definetely something; however, compared to the enormous outpouring of work on motivation problems, it is fair to say that coordination problems are neglected, although there are reasons to suppose that they are quite important: There are plenty of examples of highly motivated people utterly failing with respect to organizing and coordinating.
I just came across an excellent paper, “Coordination Neglect: How Lay Theories of Organizing Complicate Coordination in Organization,” that deals with a number of obstacles to coordination rooted in heuristics (“lay theories”) that individuals apply, for example, when setting up a division of labour in an organization. Notably, individuals systematically neglect task interdependencies. They also fail to communicate sufficiently because of knowledge bias and they are poor at translating problems for others. There are plenty of useful illustrations and anecdotes in the paper, making it excellent as a companion to a traditional motivation/incentive-focused textbook in a theory of the firm class. Highly recommended!
Entry filed under: - Foss -, Papers, Recommended Reading, Theory of the Firm.
Organizations and Markets 
1.
teppo | 28 December 2010 at 3:16 pm
Very interesting. Had not seen that piece…
2.
Avinash | 29 December 2010 at 1:09 am
Economics as a Coordination Problem by Gerald P. O’Driscoll
http://oll.libertyfund.org/?option=com_staticxt&staticfile=show.php%3Ftitle=110&chapter=35676&layout=html&Itemid=27
3.
Thomas | 29 December 2010 at 3:36 pm
I like this observation. I also think we are too fixated on motivation over coordination (both in org theory and in org practice). It seems similar to a fixation on production over distribution. Imagine a world in which as much time was spent coordinating our efforts and distributing the spoils as is spent motivating us to be more productive (whereafter the spoils are quietly extracted and concentrated).
4.
Scott Masten | 29 December 2010 at 7:46 pm
Realizing potential gains from trade requires undoubtedly requires coordination. But coordination is a problem only because of bounded rationality. And in Williamsonian TCE, economizing on bounded rationality has always been a goal of organization. Now, it is true that Williamson and (Ben) Klein and those that followed their lead have emphasized factors that exacerbated opportunism. Peter’s point in the previous post — with which I agree — was that Coase has long and consistently emphasized the bounded rationality/coordination sources of transaction costs. I’m all for exploring those sources more. But I’d maintain that you cannot have a meaningful theory of organization as governance — as distinct from organization as allocation — without opportunism or motivation or purposeful action or however we term that side.
5.
srp | 29 December 2010 at 8:49 pm
Down at the microlevel of task interaction, mismatches are costly. People tend to label any mismatch as a coordination problem, but that’s too imprecise–the mismatch is a painful symptom of some proximate cause, which might be coordination but also might be a domain knowledge gap, a differing interpretation of “official” priorities, or a conflict of interest.
A workable theory of organization has to encompass the various integrating devices brought to bear on these proximate causes in order to reduce the damage they induce. Firm boundaries and incentives bear almost exclusively on conflicts of interest, but there are lots of other structural and process aspects of organization that attack the other types of proximate causes ( including coordination failures).
6.
Nicolai Foss | 30 December 2010 at 4:24 am
Scott — I disagree that coordination is a” problem only because of bounded rationality.” The simplest example is a symmetrical coordination games in which players just cannot coordinate in spite of common knowledge. Or think of Bacharach’s work on “hi-lo games.”
Re bounded rationality and Williamson, I think we can agree that he only makes use of a bare bones version of BR. It basically only serves to make contracts incomplete and that’s all.
On the need for opportunism in a theory of the firm (and point both you and Steve make), I am not so sure anymore that it is strictly necessary assumption. I am impressed with Wernerfelt’s work on adjustment costs in the theory of the firm which does not appeal to opportunism.
7.
Scott Masten | 30 December 2010 at 12:01 pm
Nicolai — Let me respond to your points in reverse order.
3. Re: opportunism. I like Berger’s adjustment cost theory but (i) I don’t see it as eliminating opportunism/self-interest; his 1997 article defines “direct costs of governance” as consisting of communication costs (including negotiation costs) required to adapt and resolve incentive conflicts raised by adaptation;” (ii) as far as I can tell, his definition of “the employment relation” does not distinguish between (true) employees and independent contractors who are paid on an hourly basis for services, and in this sense doesn’t address the organization as governance question.
2. To be fair, Williamson does do more with bounded rationality than just motivate incomplete contracts; his discussion of the limits to firm size (in relation to the impossibility of selective intervention), for example, turns on bounded rationality considerations. But clearly, determinants of variations in internal organization costs have not received as systematic a treatment in the theory as the determinants of limits of market/external forms.
1. You have lost me here. Players in PD and coordination games know the structure of the game, but they, by construction, don’t know other players’ choices when they chose their strategies. The phrase “don’t know” implies bounded rationality to me.
8.
Nicolai Foss | 30 December 2010 at 12:16 pm
A quíck reply:
1: You are right that Birger includes the communication costs of resolving incentive conflicts, but his basic idea would seem to stand even if the relevant communication costs had nothing to do with incentive conflicts.
2. When Williamson talks about the limits to firm size in terms of the impossibility of selective intervention it seems to me that he talks about 1) measurement costs (in the 1985 book) and 2) credible commitment (in the 1996 book), none of which necessitates bounded rationality.
3: “Don’t know” certainly does not imply bounded rationality to me. Asymmetric information is consistent with maximizing, non-bounded rationality. Re the game example think of a game with 1,1 and 1,1 in one diagonal and 0,0 and 0,0 in the other one (and the standard epistemic assumptions obtain). There is no rational (in the standard sense) ground for preferring either one of the 1,1 equilibria. Thus, the players confront a coordination problem, even in the presence of hyper-rationality.
9.
Scott Masten | 30 December 2010 at 12:43 pm
Quick reply reply:
2. Inability to credibly commit (to intervene only selectively) is what prevents replicating the benefits of non-integrated exchange, but the consequences of that inability to commit include greater demands on manager attention and decision making and greater reliance on bureaucracy. But measurement costs don’t necessitate bounded rationality?….
3. The latter and your statement that asymmetric information is consistent with non-bounded rationality leads me to conclude that we mean different things by bounded rationality. In my understanding, asymmetric information is a result of prohibitive information costs, and information costs, like all costs, are the result of scarcity. And what is the scarce resource that causes information costs to be positive if not awareness, attention, recollection, etc.? And is it not those cognitive limitations that define bounded rationality?
10.
Nicolai Foss | 30 December 2010 at 12:54 pm
I agree that we probably mean different things by BR. However, the consequence of your view is that only a all-knowing being (with perfect awareness, attention, recollection, etc. is non-bounded rational. That is, IMHO, making the concept pretty vacuous. I think it is best to think of knowledge and rationality as two different things. (On the level of theories, your view implies that agency theory is based on bounded rationality which most would think it is not (because it starts off from complete contracting)).
11.
Scott Masten | 30 December 2010 at 1:49 pm
Vacuous? I would say central! Complete contracting theory presumes far greater capacity to anticipate contingencies than is likely realistic in most cases. But noncontractibility/nonverifiability/asymmetric information all turn on someone not being able to learn something because the cost of doing so exceeds the benefit of having the information. But why is the cost of learning one state of the world prohibitive and another free, and why to one agent but not another? Knowledge (and therefore “information cost”) is endogenous, and what keeps us from acquiring it is our limited cognitive capacities. If that is not what BR describes, I am unsure what it is. Not behavioral anomalies, I think you’d agree. And computers have more or less removed computational limitations for all but the most complex problems.
Yes, agency models assume bounded rationality, but they do so implicitly and in an ad hoc fashion. Recognizing that BR is ubiquitous and is the ultimate source of information, measurement, and communication costs holds the prospect, at least, of making the specification of information costs less arbitrary.
12.
srp | 30 December 2010 at 8:40 pm
I hesitate to get into the middle of Nicolai and Scott when they’re having so much fun, but in the interests of clarification…
A. As far as standard definitions go, game theory emphatically agrees with Nicolai about rationality. Imperfect information (defined as not knowing another player’s strategy when you choose your own) is not bounded rationality. Neither is incomplete information (not knowing another player’s payoff schedule or “type”). These are part of the structure or rules of the game; hidden cards and moves in draw poker are not symptoms of BR, for example.
B. BR could affect coordination in ways not captured in standard game theory, for much the same reason that chess is not trivial and people exert effort to solve Sudoku puzzles. Because we are not “logically omniscient” we can’t costlessly deduce all the consequences of our knowledge, even when dealing with purely logical problems. So coming up with a good critical path for scheduling a project, even with perfect information, may not be trivial.
C. Scott is still correct that without some conflict of interest there is no way to distinguish among governance forms. Communication and instruction can occur entirely independently of contracts or authority relations. Such binding constraints on decisions can only play a functional role when allowing discretion by the constrained party would result in him or her knowingly doing something bad for the team.
13.
Scott Masten | 30 December 2010 at 10:18 pm
Steve (?), The more the merrier, I say!
Yes, I know that game theorists won’t agree, but the concept of bounded rationality didn’t originate with game theorists so I don’t see why we need to defer to their definition. The information structure of a game, including what each player knows about other players actions and when, still ultimately involves assumptions about information costs. In the real world, I can change a simultaneous game to a sequential one by peeking at your cards (to use your analogy). Game theory essentially assumes that the cost of learning another player’s action (or his value in a private value auction, for example) before you have to take your action is infinite (or prohibitive). But the real world the costs of peeking are often not prohibitive, nor zero, but rather vary continuously. As I said before, information costs are just costs, and all costs reflect a scarcity of some type. If the source of information costs is not limited cognition (aka bounded rationality), what is the scarce resource from which information derives its scarcity value?
Although I think Williamson has a claim to priority (after Simon, of course) in the use of the term bounded rationality, I’d be willing to give up the term for something else in the interest of terminological harmony if we can agree that information costs derive from whatever it is we decide to call the source from which they derive.
14.
David Hoopes | 31 December 2010 at 12:19 am
I think Postrel and I need to market our work better.
15.
srp | 31 December 2010 at 4:41 am
You can have a game where one player has the option to buy information as part of its strategy space. But in the real world I don’t think it makes sense to require, say, industrial espionage to be a part of every analysis. We simply say that the implicit costs of cheating, while finite, are too high to be considered–I wouldn’t recommend sneaking a peak at other players’ cards when you’re playing poker, at least in most venues. (Nor do I think information costs are due to scarcity in the same way as
goods that are rival in consumption, but that’s a big digression.)
Back to the substance, though, there’s no way, no matter the information structure, to get any advantage from preventing or deterring an agent rather than instructing him, unless that agent has a conflict of interest with the team. So you really can’t derive governance, which is fundamentally about restricting agents’ discretion and controlling agents’ incentives, without assuming something like opportunism.
16.
Avinash Mulye | 31 December 2010 at 6:20 pm
I tried to find a definition of the coordination problem.
A statement of “what constitutes coordination problem”.
In an “over-simplified” form … coordination problem is stated as –
who will do what?
In other words – who has which skills and willingness to perform a specific task ? The discovery process.
For instance, A buyer within a firm has 2 choices – 1)buy it from market, 2)build it internally. (Firm is, in this sense, an alternative to market.)
If the firm does not have internal capacity and if buyer has to buy it from market then he has to search for the provider/seller.
This leads to the search and information costs. First of the 3 TCs.
And then once the coordination problem is dealt with … the problem of motivation begins. Why would he/she do it – and it involves second of the remaining 2 TCs – Bargaining and negotiation & Decision making costs. This results in the contract.
And once these 2 are dealt with … then governance problem remains. Are all parties doing what they were supposed to do per the contract? This leads to Monitoring and policing/enforcement costs.
My question is – what changes should be made into my characterization of the coordination problem in order to make it more correct ?
17.
Fred Thompson | 1 January 2011 at 12:44 am
Coordination is a trivial problem where everyone knows what they are doing and who can and will do what. It seems that you are greatly simplifying the collective aspect of search and discovery, which is what coordination is about. Indeed, bargaining and negotiation, monitoring and enforcement are also collective processes, where people usually follow the rules, except where they don’t. Scott says that “coordination is a problem only because of bounded rationality.” Yes, but ignorance is fundamental to the human condition, and it’s protean.
18.
Scott Masten | 2 January 2011 at 1:25 pm
Avi, This looks like something that would best be discussed in person. Why don’t you stop by for a cup of coffee if you are around the school one of these days.
Best, Scott
19.
Thomas Esmond Knox | 7 January 2011 at 5:22 am
Interesting but not profound.
I have two copies of “The Mythical Man-Month” in my library.
One of them says:
“Dedication of the 1975 edition
To two who especially enriched my IBM years:
Thomas J. Watson, Jr.,
whose deep concern for people still permeates his company,
and Bob O. Evans,
whose bold leadership turned work into adventure.
Dedication of the 1995 edition
To Nancy,
God’s gift to me.”
I am under-impressed with the argument in this article.
It smacks of “I have never done it, so I am an expert.”
Kind Regards, & Sleep Well.
20.
David Hoopes | 9 January 2011 at 12:17 am
You are unimpressed with “The Mythical Man Month”? Fred Brooks certainly did plenty of programming.
21.
Complete Contracts: Roomate Agreement Edition « Organizations and Markets | 15 November 2011 at 12:12 pm
[…] with an extreme low-probability event (time for Nicolai and Scott to resume their debate over bounded rationality?). GA_googleAddAttr("AdOpt", "1"); GA_googleAddAttr("Origin", "other"); […]