Topic 3: Modes of Learning

In this topic we will discuss the mode of learning that is best suited for the governance process. In the previous topic we analyzed the contracting environment between principals and their agents in pension management. We learned that incomplete contracts between principals and agents are strongly non-contractible. The principals do not have observable and verifiable information to evaluate actions and outcomes of the agents. The strongly non-contractible incomplete contracts (SNIC) are embedded in a principal agent relationship characterized heterogeneous expectations and by ‘broken agency’.  In such a contracting environment the governance must be a process that seeks to reduce opportunism as a behaviour by identifying an endogenous learning mechanism that actively considers the attributes of the social and cultural norms (the prior conditioning) and promotes the feeling for the entity. The identification of such a learning mechanism is the goal of the discussion in this topic.

Modes of Learning

Under SNIC, as in pension management, the pension subscribers or principals and their agents or the different functionaries in the pension management industry are in a strategically interdependent relationship. Moulin (1995) identifies three modes of strategic co-operation or learning in strategically interdependent relationships. These modes of co-operation are:

1. The Direct Agreement Mode;
2. The Justice Mode; and
3. The Decentralized Mode

Each of the three modes of co-operation incorporates a learning mechanism. Which one of the three learning mechanisms is appropriate in the context SNIC & ‘broken agency’, to reduce opportunistic behaviour in the relationship between principals and agents in pension management?  This we will evaluate this next.

The Direct Agreement Mode

The direct agreement mode is independent of any institutional context as there is face-to-face transaction. Given SNICs between plan subscribers and financial service providers; heterogeneity of expectations and ‘broken agency’, this mode of learning is unsuitable for pension governance as it will require that the pension subscribers (principals) replicate all management functions by directly negotiating every stage of the pension management process.

The advisor-client relationship in individual pension plans, such as the Registered Retirement Savings Plans (RRSP) in Canada, is the closest example we have of learning in direct agreement mode. However, being free of the institutional context, this mode does not inform us about information access, quality, or decision capability of the contracting parties. This is of crucial importance in a relationship of strategic interdependence characterized by SNIC and ‘broken agency’, as in the pension management process. There are very debilitating challenges to this mode of learning in pension management. First, the agents do not recognize their roles as fiduciaries. The financial advisors subscribe to a suitable standard of care, which is a problem when information exchanged is both non-observable and non-verifiable and there are multiple levels of agents who directly interact only with the financial advisors.  Furthermore, there is no clear evidence to suggest that individual pension subscribers as principals have the required financial capability to monitor and take financial decisions (IBRD, 2008). The evidence on financial capability is that individuals lack the capacity and opportunity to take financial decisions. A meta analysis of financial literacy programs shows that most interventions to promote financial literacy have little or no impact in promoting financial well being.

The Justice Mode

In the justice mode a judge replaces the learning mechanism and decides the fair share of surplus generated by the strategic co-operation among the contracting parties. There are two requirements for the justice mode to work. First, the judge as an arbitrator will have to be accepted as equitable or fair by both the pension subscribers and the pension service providers. Second, the judge will have to be omniscient that is now and forecast the probability of every potential outcome. This is improbable in a SNIC environment; with heterogeneity of expectations and ‘broken agency characteristics of pension governance. There is no room for bounded rationality in the justice mode. The formula for fair and just assessment of expectations and obligations must measure existing and evolving all aspects of similarities and differences in the expectations and obligations of the pension subscribers and the multiple layers of agency in the pension management process. It is difficult to imagine an indivisible collective authority representing all pension subscribers and the agents.

Without a benevolent and omniscient dictator, the only option in the justice mode is to devise mechanical rules. This goes against the focus of pension governance – reduction in opportunistic behaviour. Reduction in opportunistic behaviour requires a feeling for entity which will not be enhanced by the adoption of mechanical rules in the conduct of a strategic relationship between pension subscribers and pension service providers. There is also the issue of acceptability of pension management decisions. The acceptability of decisions will depend on whether the parties understand the decision process. This mode of learning will not reduce opportunistic behaviour.

The Decentralized Mode

In the decentralized mode decision making power is distributed among both principals and agents. Co-operation takes place in strategic interactions based on conjectures about each other’s expectations. The basic framework of expectation formation in the relationship between pension subscribers and the pension services providers is the property rights view. Pension subscribers as owners of pension assets can choose institutional structures or pension design, governance and investment structures, that allow them the freedom to pursue their objectives. In effect the pension subscribers as principals can adopt a dominant strategy. In a two-person game, strategy L by a player A is a dominant strategy if no matter how the other player responds, strategy L will maximize player A’s pay off. Strategy L is unambiguously the best strategy for player A even if he or she does not have the slightest idea of how the other player(s) will react. Thus, player A’s behaviour will be insensitive to the amount of information possessed by other players or their preferences.

The use of quarterly performance data or a similar checklist of performance indicators in the hiring and firing of fund managers or other functionaries in the pension management process are examples of the use of dominant strategy.  The learning mechanism in the dominant strategy is that the plan subscribers, through quarterly assessments of plan performance, hire and fire managers and thereby unilaterally impose sanctions on the agents in the pension management process. By providing information in this form, pension subscribers (principals) aim to establish expectations and thus elicit co-operation from managers (the agents) and ensure returns on their invested pension assets are maximized.

Such governance structures will elicit a response from agents or managers/advisors involved in the pension management process that will seek to minimize their own human capital risk. Psychoanalytic responses of managers faced with such a dominant strategy is one of unfairness, subjectivism and irrationality about the event. Managers or agents will have a significantly reduced feeling of entity. This feeling for the entity is one of the determinants of opportunistic behaviour. It increases the demands on the governance structure of pensions. Opportunistic behaviour on the part of agents will become endogenous and a function of the governance structure. Instead of reducing opportunistic behaviour the pension governance structure will exacerbate the problem of opportunism. In a ‘broken agency context’, this will be a further motivation to book profits in the short run and postpone costs.

We find  that none of the three modes of learning can be used directly to set up the endogenous learning mechanism for the governance of SNICs between plan subscribers and financial service providers. The direct mode cannot be implemented, given SNICs; heterogeneity of expectations; ‘broken agency’ The learning mechanisms in both, the justice mode and the decentralized mode of strategic co-operation will increase the problem of opportunism in pension management relations. However, a mixed mode called the   Procedural Justice Mode based on a combination of the justice and decentralized mode can be used to set up the framework for learning between principals and agents in pension management.  We find that a mixed mode learning will minimise both opportunism as a behaviour and opportunism as an attitude by providing the prior conditioning and promoting the feeling for the entity. This we discuss in the next section.

A Mixed Mode: Learning in the Procedural Justice Mode

In the Procedural Justice Mode learning is a process in which principals and agents construct approximate models of the decision environment that is updated as more information becomes available (Salmon, 1995). This mode of strategic co-operation can enhance the feeling for entity and minimize the scope for opportunism as an attitude as well as behaviour. However, minimization of the scope for opportunism will depend on how learning is specified in the procedural justice mode. In economics, learning in the procedural justice mode is incorporated using statistical algorithms like the Bayes’ rule . This approach to learning does not incorporate the role of the decision environment in shaping strategic co-operation. In the case of pension management with SNIC and ‘broken agency as attributes, learning in the procedural justice mode will have to move beyond statistical algorithm of learning from ex post data and revising probabilities of outcomes for the next outcome.  In the context of pension management, bounded and ecological rationality provides a more appropriate framework for specifying the learning mechanism. In bounded and ecological rationality learning is shaped by the interaction between the individual and the decision environment and is not the product of a statistical artifact. Such a specification of procedural learning will be conducive to the feeling for entity and have the effect of reducing opportunism as a behaviour.

Rawl’s (1971) analysis of procedural justice allows for the incorporation of bounded and ecological rationaity in the procedural justice mode of learning. He distinguishes between three modes of procedural justice; pure, perfect and imperfect. Pure procedural justice emphasises justice of the procedure independent of the outcome. There is no independent criterion for the right result. There are however, procedures where the output is considered correct or fair, whatever it is, provided that the procedure has been properly followed. Perfect procedural justice on the other hand, is an independent standard for deciding which outcome is just and determining a procedural guarantee to lead to it. Finally, Imperfect procedural justice occurs when there is an independent criterion for the correct outcome but no feasible procedure that guarantees to lead to it (Gustafasson, 2002). This differentiation of the procedural justice mode is the basis for a more detailed specification of the learning mechanism that can be used in pension governance.

Given the incomplete contracts in pension governance, characterized by SNIC; heterogeneity of expectations and ‘broken agency’, only the pure procedural justice mode can be a mode of strategic co-operation between shareholders and managers. Under pure procedural justice the interacting parties have equal opportunities to exchange messages and there is equal influence of messages of all involved in the strategic relationship. The outcome chosen is in the decentralized mode which can involve termination of the relationship with the managers/advisors or the agents. Such decision environments are treated as just and independent of the outcome by all in the strategically interdependent relationship.

The emphasis of the pure procedural justice mode is on the decision process. Procedures can be an end in itself; irrespective of the outcome. Procedures will be the best guarantee for the realization of self-interest (Thibut and Walker, 1978); as an indicator of group value (Tyler and Lind, 1990) and in the perception of dual obligation (Folger, 1993). Procedures will contribute to the perception of equality which is critical for pure procedural justice. It will also promote the feeling for entity, which reduces the scope for opportunism as behaviour.

Communication Behaviour in the Pure Procedural Justice Mode

Strategic co-operation between principals and agents under pure procedural justice mode requires equal opportunity and influence in the exchange of messages. The outcome chosen from the decentralized behaviour of contracting parties is considered just even if the decision has adverse consequences for one party. Several studies on procedural justice have found a positive association between communication behaviour and perceptions of procedural justice (Greenberg, 1994; Korosgaard, Scweiger and Sapienza, 1995; Miles and King, 1997; Sapienza and Korosgaard, 1990). In the analysis of relationships that cannot be fully specified or controlled in advance of their execution and where underlying expectations can vacillate in unforeseeable ways, the legal literature draws similar conclusions. For the management of such relationships, it is concluded that the learning mechanism will not be limited to processing of information but will be have a more pro-active stance based on sustained engagement between the contracting parties in the strategic relationship. Such engagement will depend on relational assets like favourite prior belief, trust and goodwill (Salbu, 1995). Thus, the operative part of the pure procedural justice mode is that there must be equality in opportunity and influence of messages. Procedural justice mode requires a learning mechanism with extensive information processing capabilities based on trust, goodwill and equal opportunity for exchange of information.

Mix of Structured and Unstructured Communication – Deliberative Decisions

Kim and Mauborgne (1998) list three requirements for learning under a pure procedural justice mode in strategic decision procedures: explanation; engagement and clarity of expectations. The pattern of communication that meets the requirements of explanation, engagement and clarity of expectations will require decisions with a mix of structured and unstructured exchange of information and periodicity and intensity of communication. Examples of structured information can be formally identified channels like financial reports, actuarial assessments of pension plans, etc. Unstructured information exchange can be through annual pension meetings, board and committee meetings, social exchanges, professional gatherings, etc. The communication or deliberation between pension subscribers and the manager/advisors will have two attributes:

1. Unstructured communication will be the main mode of interaction or strategic co-operation under pure procedural justice mode.
2. Financial communication and other forms of structured communication will be used as ‘fire alarms’ to reduce the cost of ongoing communication or deliberation.

Deliberating for a decision is a cognitive process in which the decision maker engages as the decision is framed, as goals and plans are adopted or rejected and as implementation is monitored and plans and goals are modified with new information.  This conception of learning identifies the sources of deliberation as the decision-makers own knowledge of (pension plan) organizations; suggestions of support persons; examples offered by outsiders; and existing rules and regulations (Beach, Mitchell, Paluchowski and van Zee, 1992). A decision-making environment based on unstructured communication exchange will not only reduce the agency problem but also improve the quality of decisions as it will provide the basis for voluntary co-operation. Such co-operative behaviour improves the decision frame as there is a greater flow of idiosyncratic information associated with the strategically interdependent relationship, as in a pension plan.

Financial reports are the traditional structured channel of communication. Such forms of information must be simple and standardized to be understood and interpreted by all concerned. Such standardized information is termed as ‘focal points’ by shareholders (Kreps,). Standardized information can be ignored or there are there are two possible other ways in which ‘focal points’ could be used. Standardized information can be used as a dominant strategy, as ‘triggers’ to hire and fire managers. If financial reports and other structured information are used as ‘triggers’ to hire and fire managers, this will go against the requirements of the procedural justice mode. There is in effect no reciprocity in the exchange of information. Pension plan subscribers or their representatives in organizational pensions plans will in effect adopt a dominant strategy. Such use of ‘focal points’ will lead to a strong perception of unfairness by the agents – the mangers/advisors and lack of faith in the authority of principals – the pension plan subscribers or their representatives. This will elicit response or be used as a justification for opportunistic behaviour by agents such as financial and pension advisors, to preserve their human capital. The norm in the management of pension assets is to either ignore the focal points and decide on prevailing market sentiments or indexing or to use them as triggers as is in the case of earnings reports.

Another possibility is that the ‘focal points” could be used as ‘fire alarms’ to signal the need for gathering more (qualitative) information.   The gathering and communication of structured information serves a ritualistic purpose indicating to the manager/advisors in the pension management process that a proper attitude towards decision making exists. Information is not simply a basis for action but a representation of competence. Thus, the gathering of simple information reflects credible decisions and will contribute positively towards perceptions of procedural justice.

Why and how financial information should be used as ‘fire alarms’? Deliberation or unstructured information is a costly exercise. High intensity deliberation cannot be sustained for long. Conceptually, it will amount to the adoption of the direct agreement mode of strategic co-operation between plan subscribers and the managers/agents and a near replication of all functions in the pension management process.  A cost-effective alternative would be to use ‘focal points’ as ‘fire alarms’; as signals for initiating deliberation (McCubbins & Schwartz, 1987). Should deliberation be initiated only when ‘fire alarms’ are sounded and when the ‘focal points’ suggest poor performance? The monitoring authority in our case is the pension subscribers either at the individual or at the organizational levels. Given SNIC in the context of ‘broken agency’, the interpretation of structured information requires a level of awareness. This is only possible if there is ongoing and continuous communication. There is also the issue of reliability of information. Intervention by pension plan subscribers or their representatives in organizational pension plans can be quickly affected if there is a live data base available from continuous low-key deliberation. The framework for information exchange and governance that is proposed is one of low-key deliberation with ‘focal points’ as ‘fire alarms’ signalling the need for more intensive deliberation.

The discussion in the learning mechanism leads to the following inferences:

1. A mixed mode called the Pure Procedural Justice mode will be the most appropriate mode for endogenous learning. Strategic co-operation in strongly non-contractible incomplete contracts and the ‘broken agency between principals and agents in pension management;
2. Learning in the pure procedural justice mode will be based primarily on unstructured communication;
3. Structured communication will be used as ‘fire alarms’ to economize on the costs of gathering unstructured communication and for effective learning. Use of structured communication as ‘triggers’ will increase endogenous opportunism.

The pure procedural justice mode of strategic co-operation as an endogenous learning mechanism is also consistent with behavioural and ecological rationality as it allows for adaptive cognition; information gathering and fairness in the exchange of information between principal and agents.

The use of the procedural justice mode of learning and emphasis of governance as a process has implications for pension design and investment. Process driven governance requires relationship-based governance so that qualitative and quantitative information may be gathered. Significant information processing capability is required to engage in a deliberative process driven governance where structured information is not used as ‘triggers’ for hiring and firing decisions of agents. Given the level of engagement and information processing capabilities required to sustain the procedural mode of learning, only organizations with considerable internal management resources will have the capability to participate in such a mode of learning. At the individual level the evidence on financial capability and financial literacy suggests that most individuals do not have the capacity to interact with agents and engage in deliberation.

This has implications for the design of pension systems. Pension systems will have to be designed so that the target rate of income replacement in retirement is achieved by pension plans that are organization-based and large enough to have significant resources to develop internal management capability to engage in deliberative mixed mode of learning that combine structure and unstructured communication in the procedural justice mode. This does not rule out choice-based individual pension plans but they cannot be primarily relied upon to provide old age security.

Even at the organizational level, scale is an important consideration in the effectiveness of pension organizations. Surveys of governance practices show that there is a lack of expertise and focus in the functioning of the pension board and pension investment committees. (Ambachtsheer, 2016, p.90). There is unwillingness to engage and there is an overreliance on agents or advisors who do not recognize their roles as fiduciaries. The pension board and investment committees spend a disproportionate amount of time on administrative and routine compliance reporting and not on strategic issues.  Capacity building in boards and investment committees for a deliberative, process driven governance framework requires investment in training on an ongoing basis. This expense is more readily incurred by large pension organizations.  These inferences find substantiation in a recent report on Canadian pension plans.

The examples of good practice discussed in this report are large organization and multisector pension plans.