RDP 9311: Agency Costs, Balance Sheets and the Business Cycle 2. Agency Costs, Credit Supply and Credit Demand: A Review

In traditional economic theory, specialisation allows efficiency gains through the division of labour. With perfect information, this division of labour can be achieved by one party (the principal) hiring another party (the agent) to perform certain specialised tasks. However, a fundamental problem facing the principal is the inability to costlessly write contracts with the agent that cover every possible outcome. As a result, and in contrast to the frictionless traditional theory, contracts are incomplete and enforcement is costly. This creates a potential conflict of interest between the principal and the agent, with the agent having opportunities to engage in activities that are not in the best interests of the principal. These information and incentive problems give rise to various types of costs that reduce the gains from division of labour. For instance, the principal may be required to spend resources verifying the claims of the agent. These costs are commonly referred to as agency costs.

In the case of a contract between a borrower (the agent) and a lender (the principal), the lender often has limited information about the borrower. In the work of Stiglitz and Weiss (1981), this inability to distinguish between different types of borrowers can lead to the lender restricting credit altogether. In contrast, in Williamson (1987) the lender knows the risk characteristics of the different borrowers, but must verify the claimed outcome of the projects undertaken by borrowers. Just as in the Stiglitz and Weiss model, these verification or monitoring costs can lead to credit rationing.

However, credit rationing is only one response to the information problem. Instead of quantity rationing, the lender may use prices to solve the incentive problems inherent with asymmetric information problems. The more equity or collateral a borrower can commit to a project, the smaller the risk the lender faces, and the lower will be the premium on external funds. This interaction of collateral and interest rates can have macroeconomic implications. More generally, the interaction of agency costs and collateral has implications for how the economy responds to adverse shocks. In this Section we review some of these implications.^[2] Specifically, we examine the interactions between corporate collateral and the cost of external funds; between bank collateral, portfolio allocations and interest rates; and between corporate collateral and management incentives.

2.1 The Role of Corporate Balance Sheets

One of the first models to link firms' balance sheets with the macroeconomy was developed by Bernanke and Gertler (1989). In their model, the borrower (the agent) has information about the outcome of the project that is not freely available to the lender (the principal). Since the lender cannot costlessly observe the project's outcome, the borrower has an incentive to falsely declare the project a failure. By declaring the project as failed, the borrower does not repay the loan principal or interest, and if not caught, is able to retain the entire return from the project. This incentive to declare ‘failure’, gives rise to a ‘costly state verification problem’ first outlined in Townsend (1979).^[3] The lender must decide whether to verify the bad outcome declared by the borrower or to accept the declared bad outcome as the true state of the project. The probability of the bank verifying the borrower is a function of the collateral of the firm. The costs of verification/audit are often referred to as monitoring costs, and can be viewed as a reflection of the principal-agent problem.

When writing the loan contract, the lender must decide on the interest rate to charge on the loan. In doing so, the lender must take into account the expected auditing costs, which increase when the probability of an audit increases. These costs get fed through into higher interest rates. A key factor in the probability of an audit is the amount of collateral the borrower has backing the project.

In the case of complete collateralization, the borrower can guarantee payment even if the project is a failure. The probability of the lender auditing is zero (agency costs are zero) and the interest rate charged contains no added premium. However, in the case of incomplete collateralization, the positive probability induces positive auditing costs. The greater the project's outside funding requirement, the greater are the expected auditing costs. With a low equity contribution, and limited liability, the borrower has less at risk if she falsely claims the bad outcome, and therefore must be audited with a greater probability in order to induce honest behaviour. The lower the contribution of the borrower to the project, the higher is the premium on external funds.^[4] In general, external finance is more costly than internally generated finance, the difference being a measure of agency costs.

This model predicts an income-accelerator effect on investment. A negative shock to (say) productivity might cause the entrepreneur's current period income to be negative. As a result, agency costs will be higher in subsequent periods as the entrepreneur has less collateral for the next investment project. Consequently, investment will be lower than it otherwise would have been. Although firm-specific shocks may tend to average out, systemic shocks to the economy may reduce aggregate investment.

Agency costs are counter-cyclical; increasing in recessions when firms' asset values are most likely to be depressed, and decreasing in booms. Further, shocks may have asymmetric effects through the business cycle. Sharp deteriorations in the economy are more likely than sharp improvements, since there is a limit to the reduction in agency costs in good times. When borrowers experience growth conditions for a number of periods, eventually future investment projects can be self financed and agency costs tend towards zero. Any further improvements in conditions will not lower agency costs, and there will be no further accelerator effects. However, in downturns, some firms that were once fully collateralized will now incur agency costs, exacerbating the downturn.

While the above model provides a theoretical link between corporate balance sheets and the business cycle, the empirical magnitude of the mechanism of the model is questionable. That is, in many cases monitoring (auditing) costs are insubstantial relative to the value of the loan. More recent models of agency costs have attempted to address this problem. Bernanke and Gertler (1990) develop a model where the outcome of the project is common knowledge, removing the need to audit.^[5] Instead, the asymmetry of information arises from the lender not knowing the quality of the investment project chosen by the borrower. The entrepreneur incurs a cost in initial evaluation of a project, and then must decide whether to proceed. If the project evaluated looks like a ‘bad’ project, the entrepreneur can discard it and evaluate another, but only by incurring the evaluating cost again.

Having incurred the cost, the borrower has an incentive to pass off the bad quality project to the lender as a good project, as the evaluating costs are sunk and cannot be recouped should she decide against proceeding with the project. The lower the net worth of the borrower, the higher is the incentive to pass off ‘bad’ projects as ‘good’ projects. As a result, the probability that any loan is repaid is lower. This in turn leads lenders to increase interest rates. Finally, the higher interest rates reduce the willingness of entrepreneurs to evaluate and undertake projects. Again, this model predicts that the business cycle will be amplified if adverse shocks affect firms' equity.

The models described above are all single period models. If financial institutions recognise the existence of business and asset price cycles, changes in equity may have smaller effects on the price and availability of intermediated finance.

Gertler (1992) extends the single-period contract models by examining multi-period financial relationships between borrowers and lenders. Unlike the single period models, lenders now take into account the net present value of the expected ‘lifetime’ earnings of the borrower's investment project when deciding on the cost of external funds. If the borrower declares a ‘bad’ outcome in one period, the lender may be willing to reschedule debts if the borrower has sufficient expected ‘collateral’ in future periods. Because the debt gets rescheduled rather than written off, there is less incentive for the borrower to falsely declare the state as failed. A positive shock to a firm's profits in the first period increases output at the end of the first period, strengthening the firm's balance sheet and reducing the incentive problem for investment in period two. Consequently, the firm faces a lower interest rate in period two. Thus, shocks to a firm's balance sheet in one period can influence production in future periods even though the shocks themselves are serially uncorrelated.

In addition, a more favourable expectation about the profitability of future projects reduces the agency cost today. Even if the bad state was declared at the end of the first period, the expectation of high future earnings allows a larger debt burden to be rolled over. Conversely, a negative expectation about future economic conditions will increase the premium on loan interest rates today, because there is less ‘collateral’ in future projects, and less debt will be allowed to be rescheduled.

Finally, in a continuing relationship between borrower and lender, the past performance of the borrower can influence the size of the premium on external funds. The borrower realises that the moral hazard problem of attempting to pass off bad projects as good projects increases the premium on external funds. However, by always telling the truth, the borrower gains a good reputation that allows the lender to reduce the size of the premium. Hence, agency problems may be more pronounced for new firms than for established firms.

The above models imply that a deterioration in corporate equity through say an adverse demand shock or a reduction in asset prices may cause the cost of external funds to increase. This increase will amplify the contractionary effects of the initial shock. Further contractionary effects are likely to be experienced if the reduction in corporate equity is translated into a reduction in the equity of financial institutions. It is to this issue that we now turn.

2.2 The Role of Banks' Balance Sheets

Just as a lender faces some asymmetric information when lending to a firm, a depositor faces asymmetric information when lending to a bank. Using this insight, Bernanke and Gertler (1987) develop a model that shows how factors that affect the health of the financial sector can have macroeconomic effects. The general asymmetric information problem is the same as before, but now it arises between the bank (the agent) and the depositor (the principal).

Banks are assumed to have a cost advantage in evaluating and monitoring potential projects.^[6] However, the information gained from project evaluation remains private to the bank; only the bank knows the true outcome of its loan investments. The contract between the depositor and the bank cannot be a state-contingent contract because the states cannot be verified by the depositor. Therefore, debt contracts arise as the optimal arrangement between the parties.^[7] Further, it is assumed that the bank cannot obtain a sufficiently high number of differentiated projects to perfectly diversify away risk. As a result, depositors face a potential risk similar to that outlined in Diamond and Dybvig (1983). However, the bank is able to overcome the information problem by holding bank capital, which acts as a form of collateral for the depositor.^[8] Greater amounts of capital allow the bank to secure more deposits, ceteris paribus, and allow the bank to allocate a larger proportion of its lending portfolio to risky loans.^[9]

Thus, the bank's net worth and the quality of the bank's illiquid assets affect the quantity of deposits that it can attract, and hence the amount of lending it can conduct. In turn, this affects investment and output. A deterioration in the bank capital or deterioration in the value of the collateral implicit in illiquid bank assets forces a contraction in a bank's illiquid investment projects, with resources being shifted to liquid assets (in their model, storage) and safe projects. Bernanke and Gertler suggest that this is what happened in the banking crisis of the 1930s. Some projects, especially those that were information-intensive and relied on bank credit, found that they were not able to obtain funds under any conditions.

There are implications for interest rates as well. The deposit rate is sensitive to the expected return on the bank's portfolio. The higher the bank's collateral, the greater the proportion of the bank's assets that can be invested in risky illiquid assets, and the higher the deposit rate the bank can offer. In the case of a deterioration in the bank's capital, the deposit rate falls, and the bank substitutes its assets towards the safe sector.

2.3 The Demand for Credit: the Role of Managerial Incentives

The models discussed in Section 2.1 and 2.2 imply that a reduction in the equity of the corporations and financial institutions will lead to lenders supplying fewer loans at any given interest rate; that is, the supply curve for intermediated finance moves to the left. In practice, movements in the demand curve may be equally important. A deterioration in the equity of the firm may well affect managers' decisions concerning production and investment levels. As a result, falling internal net worth may see both the supply of, and demand for, funds decline.

The work of Jensen and Meckling (1976) represents one of the first attempts to link management incentives and firms' capital structure.^[10] They argue that when there is asymmetric information between managers (agents) and owners (principals), managers have an opportunity to transfer some of the firm's resources to themselves without the knowledge of the owner. For example, managers may build plush offices and buy corporate jets. Even if the owners see the manager's actions, it is difficult for them to know if these actions are in their own best interests, since managers know more about the firm than the owners (this is the main reason for hiring a manager in the first place). If the manager owned the firm, or owned a large proportion of the firm's equity, then this incentive problem would disappear.

In Jensen and Mecking's model, managers are assumed to be approximately risk neutral. They are not concerned with possible adverse effects on their reputation should their actions ultimately bankrupt the firm. In practice, such considerations may be important. The firm's owners can diversify risk by holding shares of many companies. In contrast, the managers are likely to have sunk effort and skills in a particular firm that cannot be easily used at another firm. If the manager should lose her job, she suffers a high cost. Thus, the manager may behave in a risk-averse manner taking on projects that do not maximise the value of the firm.

Greenwald and Stiglitz (1993) develop a model in which internal equity has a role in influencing the evolution of the business cycle. The firm, through its manager, is assumed to maximise expected profits less expected bankruptcy costs. Output decisions must be made before the price of the output is known, and are therefore risky. Also, expected bankruptcy costs are a decreasing function of equity. As a consequence, the lower is the equity of the firm, the lower is its output.

In this model, bankruptcy costs are added to the objective function in an attempt to capture management's dislike for bankruptcy. This type of specification does not address the question of how the type of management contract effects the manager's decision and how risk aversion by management interacts with the contract. These factors are potentially important. In practice, the type of contract that is offered to managers can have important implications for their decisions. Agrawal and Mandelker (1987) find evidence that holdings by management of common stock and stock options lead to an increase in firm risk. Using information about the variance of a firm's share price before and after an announcement of an investment, they find that managers with a relatively high proportion of shares in their compensation package select investments that result in variance increases. This supports the hypothesis that the managers' holdings of shares induce them to make investment decisions that are in the interests of the firm's owners. Along similar lines, Jensen and Murphy (1990) find that there is a small but significant positive relationship between performance schemes and firm performance.^[11]

The costs incurred by a manager when her firm experiences ‘financial distress’ are documented by Gilson (1990). He finds that, on average, less than half of the directors and managers of a firm that experiences financial distress keep their jobs. Also, the directors who have resigned from these firms are less likely to serve as directors of other companies. Jensen (1988) also suggests that a takeover, or a potential takeover, can have significant incentive effects on managerial behaviour. Evidence shows that roughly 50 per cent of a firm's top level managers are gone within three years of acquisition.

These studies of managerial compensation are generally divorced from the business cycle and the internal equity of the firm. In the model developed below, we attempt to capture this interaction between contracts, the internal equity of firms and investment.

Footnotes

See also the reviews by Gertler (1988), Hansen (1992) and Stiglitz (1992). [2]

Other papers to examine the implications of the costs of evaluation and monitoring include Gale and Hellwig (1985) and Williamson (1987). [3]

The problem disappears if firms can costlessly raise equity to finance the project. However, similar information problems are also present in the equity market. Consequently, it is assumed firms are equity rationed. See Greenwald, Stiglitz and Weiss (1984) for details. [4]

Calomiris and Hubbard (1990) also model the role of collateral in mitigating adverse selection problems arising from information-intensive borrowers, but do not rely on monitoring costs. [5]

This is a common assumption in the literature. Diamond (1984) shows that there is a cost advantage of individual investors (depositors) delegating monitoring duties to a financial intermediary. Perfect diversification by the bank means the depositor does not need to monitor the bank. [6]

A deposit is a loan contract between the depositor and the bank. The depositor receives a fixed payment on the deposit, regardless of the outcome of the bank's investment project. Thus, this contract is very similar to the debt contract between the bank and a borrower. [7]

The bank's capital consists of equity obtained from the bank's owners or from retained earnings. [8]

Central banks may offer explicit or implicit deposit guarantees which act to remove the risk to depositors and increase the bank's moral hazard incentives. However, even in the unlikely case that the guarantee held exactly, the bank may still wish to hold capital if the bank is risk averse. For example, Stiglitz (1992) argues that if bank's objective functions are characterised by decreasing absolute risk aversion, the size of a bank's loan portfolio is positively correlated with the bank's net worth. [9]

Harris and Raviv (1991) provide a good survey of the theoretical and empirical literature on incentive problems and the determinants of capital structure. [10]

Australian anecdotal evidence (Kavanagh and McBeth, 1988) suggests that although bonus schemes are used in the majority of companies, they have become expected, and seen as part of the manager's total remuneration package. However, many firms are now adopting performance – based criteria in addition to the bonus scheme. [11]