The Uses of Diversity. David Ellerman

      Figure 2.2 Breadth-First Search versus Depth-First Search.

      The choice is between the parallel-oriented breadth-first strategy and the series-oriented depth-first strategy. “The ideas of breadth and depth are in competition throughout the whole history of combinatorial optimization” (Strang 1986, 609).

      The main work on this topic in human affairs is Albert Hirschman’s (1970) development of the contrast between the parallel-oriented logic of exit (exit the branch to try other branches) and the series-oriented logic of voice, loyalty, and commitment (stay loyal and committed to the given branch by searching further along it).²

      As a variation on the tree model, we could think of options with characteristics. Suppose one has an option with unsatisfactory characteristics. Does one treat the characteristics as fixed and then seek improvement by exiting the option to find a better one? Or does one stick with the given option and try to change the characteristics for the better?³ Hirschman referred to the change-the-characteristics strategy as voice: “Voice is here defined as any attempt at all to change, rather than to escape from, an objectionable state of affairs” (Hirschman 1970, 30). These two logics of exit or voice (commitment) are quite common in human affairs. Does the “unhappy camper” fold his tent and look for a better campsite or does he work to make the given campsite better? Every potential migrant faces the question: Exit to find a better home or commit to making home better?

      Managers constantly face similar decisions. When a team of workers is not performing satisfactorily, the manager has two choices. One choice is to take the capabilities of the team members as fixed so that people need to be shuffled in and out of the team until the right team characteristics are obtained. Or a manager might proceed with more commitment to the team members and then try to work with them to better develop their capabilities until the team performed satisfactorily.

      The two logics are quite ubiquitous. I will call the parallel-oriented approach in an organizational or institutional context, the logic of exit. Decline and dissatisfaction leads to exit and replacement. I will call the series-oriented approach using voice in an institutional context, the logic of commitment.⁴ Decline and dissatisfaction leads to renewed commitment and the attempt to transform the characteristics (e.g., through the exercise of voice).

      Five Points on the Two Logics

      My first point is simply the ubiquity of the two logics—as I hope will become clear in the course of the argument. I will focus on cases where the two system logics are incompatible rather than cases where they need to be blended to find the best system.

      The second point is that sometimes there seems to be an awareness of only one logic, for example the belief that improvement can only come through exit and replacement. For a trivial example, young people might think that the only response to a dull razor blade is to throw it away and replace it with a new sharp blade. But in “the old days” there were straight razors; when the blade got dull, one would sharpen it. This is an example of the two logics as the replace versus repair strategies: buy cheap replaceable items and replace as necessary or buy more expensive quality items and repair as necessary.

      The third point is that often the desired performance can be obtained by using a system based on either the logic of exit or the logic of commitment. Both ways are possible; the best system will depend on the particular circumstances.

      The fourth point is that there are often interdependencies in a system based on one of the logics so that hybrids tend to be lethal rather than vigorous. If one tries to “mix and match” exit-based components with commitment-based components, then the system will most likely malfunction.⁵

      For instance, in an exit-based system of conflictual labor relations, large inventories help to mitigate the holdup problem of workers going on strike. A just-in-time inventory system fits together with cooperative labor relations. But if a company with conflictual labor relations decides to try a just-in-time system to save on inventory costs, then any labor conflicts (e.g., truckers) will lead to costly holdups for the company.

      Or in an exit-oriented system of arms-length finance, companies need low leverage and a large equity buffer since when the company hits some turbulence, their financiers may not be willing to extend or rollover their loans. But in a commitment-oriented system of relational finance, firms can have more leverage since their financiers are expected to stay committed in times of trouble.⁶

      One of the (partial) explanations for the East Asian crisis in the late 1990s was that the globalization of finance tempted highly leveraged firms in countries that historically had relational finance (e.g., Korea) to “take advantage” of the cheap and easy finance available from international (arms-length) financiers. But then small shocks to the new “hybrid” system were soon amplified to a crisis when arms-length financiers did not rollover loans and firms had too much leverage to ride out the shocks. In such a system that got stuck halfway between the two system logics, there are two basic policy options. One option is to make a wholesale conversion to an exit-based system of arms-length finance. The other option is to curtail the impact of the exit-based features to maintain functionality in a modernized commitment-based system, for example, using capital controls to install “speed bumps” on the “hot money highway” of globalized arms-length finance.

      The fifth point is that while one system logic may be appropriate in one case and the other logic in another case, there is however a “worst of both worlds” hybrid which combines the fixed characteristics of the one logic with the fixity of the option in the other logic. If both the option and the characteristics are fixed, then one has an unviable “lethal hybrid” that cannot change in the face of adversity and decline. Unfortunately, some attempts to create hybrids with the virtues of both logics may end up with the worst case that only combines their vices.⁷ And some attempts to jump over the chasm by converting a system based on one logic into a system with the other logic may stall halfway across and fall into the chasm to get the “worst of both worlds.”

      I am focusing on cases where the two logics tend to be antagonistic or mutually exclusive (i.e., hybrids are more lethal than vigorous) so that the system needs to be based on one logic or the other. If the two logics are on the two axes, then we are interested in cases where the best solution is a corner solution.

      There is a whole range of cases where the two logics are complementary rather than antagonistic. The logics appear as two “moments” that need to be optimally combined to have the best solution. Then the best case is not a corner solution but some convex combination of the two moments. For instance in complex adaptive systems, the parallel-oriented moment is called exploration (or variation) and the series-oriented moment is exploitation (or selection).⁸ If the goal of the model is to reach a higher altitude on a rugged and cloudy “adaptive landscape,” then exploitation means staying committed to the given hill and climbing higher on it, while exploration means jumping, traversing, or otherwise exiting to another hill that may be higher. In such a system, a corner solution would not work. A system with all exploration would jump in an agitated manner from hill to hill without settling down to climbing higher on any hill. A system with all exploitation would only climb the hill it is on so it would tend to get stuck on a low hill.⁹

      Two Ways to Put Eggs in Baskets

      One of my theses is that often there seems to an awareness of only one system logic to the exclusion of the other. This is probably nowhere truer than in neoclassical economics which is based on