KOBAYASHI Keiichiro: This time, in addition to introducing works of other economists, I would like to briefly report findings from a study I conducted with my research collaborator, Kengo Nutahara, in an attempt to build on theories developed in these other works. As was the case for my similar attempt in Part Five, the findings were rather disappointing. Yet, I believe there is a possibility for further developing our theory and I would be very much interested in and grateful for any ideas and comments from readers of this column.
One of theories about business fluctuations focuses on delays in propagation caused by credit market frictions. A typical example is the financial accelerator theory put forward in Bernanke and Gertler (1989), an overlapping generation (OLG) model that assumes economic agents live for only two periods. An OLG model is sufficient for the purpose of describing theoretical results but becomes a problem when we try to compare the performance of the theory with that of other business cycle theories.
Econo-kun: Is that because business cycle theories usually use dynamic general equilibrium (DGE) models in which economic agents are assumed to live indefinitely?
KOBAYASHI Keiichiro: Yes. The Carlstrom-Fuerst model was developed by implanting the Bernanke-Gertler model into a DGE model. In this model, information asymmetry, a costly-state-verification mechanism, generates collateral constraints on borrowers so that economic agents with few collateral assets cannot expand their business as much as desired. Under this situation, any shock on the economy propagates slowly, rather than being immediately transmitted to the whole economy. Also building on the Bernanke-Gertler model, Matsuyama (2004) presents a model in which financial factors cause endogenous fluctuations while the economy is still expanding.
Matsuyama, Kiminori (2004). "The Good, the Bad, and the Ugly: An inquiry into the causes and nature of credit cycles." Mimeo. Northwestern University.
Like Bernanke-Gertler, Matsuyama (2004) uses an OLG model, but with two types of production technology (three if storage technology is included). The first type of technology, "the good," produces goods with the input of labor, whereas the second type of technology, "the bad," increases the volume of goods with the input of such goods. The storage technology, referred to as "the ugly," is a technology that simply stores or hoards the goods produced. Both the good and the bad are subject to information asymmetry (of a costly-state-verification type) and thus cannot be used by economic agents unless they have sufficient assets. The volume of assets an economic agent needs in order to be able to use the good technology is assumed to be smaller than that for the bad technology. (In other words, the degree of information asymmetry is smaller for the good than for the bad.) Meanwhile, the profitability of the good is assumed to be lower than that of the bad. Under these conditions, an economic agent gradually begins to use the good technology as the volume of its assets increases. Because the good technology intends to produce goods with the input of labor employed by the economic agent, its increased use boosts labor demand and wages, which in turn results in further increase in the agent's assets (revenue generated by the input of labor in the previous term). Through this positive feedback, the economy slowly continues to expand. Up to this point, the mechanism is no different than what is observed in the Bernanke-Gertler and Carlstrom-Fuerst models. However, when there is a further increase in assets, the economic agent begins to use the bad technology in the Matsuyama model. Because the bad technology does not require the input of labor, the increased use of this technology does not lead to an increase in wages. The result is the breakdown of the positive feedback loop, whereby the volume of assets held by the economic agent begins to decrease. Negative feedback then begins, triggering endogenous deterioration of the economy, referred to as the "credit reversal effect."
As such, Matsuyama (2004) is a theory built and expanded on the financial accelerator mechanism presented by Bernanke-Gertler with the addition of the credit reversal effect.
It may be possible to embed Matsuyama's OLG model in a more general DGE model as Carlstrom-Fuerst did with the OLG model of Bernanke-Gertler. As one such generalization attempt, I considered the following expanded model and compared its numerical computation performance with that of the original model.
The Carlstrom-Fuerst model assumes cost is required for the lender in a financial transaction (money-lending) to monitor the situation of the borrower (success or failure of a project). It also assumes that the monitoring cost is calculated as an amount equivalent to μ units of "final good." In the expanded model, I assumed that the monitoring cost is defined as an amount equal to μ units of "labor," (w_t*μ units of final good, provided that w_t is the wage rate).
Econo-kun: Why did you use this assumption?
KOBAYASHI Keiichiro: When the economy expands and the wage rate increases, the monitoring cost also rises, halting the economic expansion. So, I expected that not only would the economic expansion propagate slowly but a reversal of the economy would also occur at some point. That is, I expected that this expansion of the model would bring the credit reversal effect. This is a very simple expansion, thus only slight modifications to the original Carlstrom-Fuerst model were needed in performing numerical computation. Ultimately the expected reversal of the economy did not occur. Instead, the propagation of the shock (productivity growth shock) becomes slightly slower by only one period than in the original model. In other words, numerical computation results of the expanded model were not so different from those of the original model.
This particular expanded model might have been too simplistic. But by somehow embedding the Matsuyama model in a DGE model, we can pave the way to developing a new type of model for describing a business cycle, and I believe this would have very high theoretical value. It seems to me that as a mechanism explaining a business cycle with financial factors, the credit reversal effect is quite realistic. (It greatly resembles the logic explaining the X-inefficiency (see note) often discussed in business cycle theories.) Developing such a theoretical model and then simulating it by means of numerical computation would help deepen the understanding of business cycles. I hope that inventive attempts will be made by theorists in the future.
