Financial market globalization, nonconvergence and credit cycles

Aghion et al. (2004) also study the role of financial sectors as a source of instability in small open economies. However, fluctuations are driven by external shocks in their model. Other notable works in this line of research include studies by Suarez and Sussman (2007), Kikuchi (2008), Kikuchi and Vachadze (2015) and Kikuchi et al. (2016). Although cycles also emerge in their models, the mechanism that our model relies on to generate cycles is different from theirs.

These 44 countries are the US, Japan, Germany, UK, France, Netherland, Italy, Switzerland, Korea, Sweden, Canada, Finland, Brazil, Australia, Norway, Belgium, South Africa, Spain, Mexico, Turkey, India, Ireland, Taiwan, Austria, Israel, Argentina, New Zealand, Hungary, Denmark, Hong Kong, Malaysia, Chile, Philippines, China, Colombia, Singapore, Thailand, Indonesia, Poland, Portugal, Czech Republic, Greece, Peru and Pakistan.

There are several reasons for working with a small open economy model. First, it is technically more tractable. Second, it is well known that endogenous volatility may arise in Boyd and Smith’s (1997) case due to interest rate movements in the world financial market. Therefore, compared with a two-country case, a small open economy case can help to highlight the mechanism we propose to generate cycles. Third, there are very few, if any, analytical treatments of a small open economy version of Boyd and Smiths (1997) study in the literature. This paper aims to partially fill this gap.

We show that the threshold is smaller than the output elasticity of capital when the production function is in the Cobb–Douglas format.

Some careful readers may notice that this mechanism shares some important features with that in Matsuyama (2007). Specifically, Matsuyama (2007) studies the macroeconomic implications of credit market frictions in a closed economy environment with two heterogeneous investment projects (type-I and type-II projects), which differ in a number of ways. In one case, he assumes that a type-I project produces less physical capital, generates a less pledgeable rate of return and requires less external funding to implement than a type-II project. As a result, credit cycles may appear when a rise in net worth shifts credit to the less productive (type-I) project. In the present paper, whereas type-I projects are based in the poor countries, type-II projects are based in the rich countries. The mechanism we propose to generate cycles can be broadly considered as an open economy application of Matsuyama’s (2007) study.

Based on this assumption, the monitoring cost, \(\gamma (k^{i}_{t})^{\beta }\), is increasing in proportion with the investment project size, \(q(k^{i}_{t})^{\beta }\). When \(\beta = 0\), both investment project size and the monitoring costs are constant independent of \(k^{i}_{t}\) and the environment outlined here collapses to the one in Boyd and Smith (1997).

See Krasa and Villamil (1992) for the case where the intermediary’s incentive is considered.

When \(\beta > \theta \), \(\mu ^{i}_{t}\) decreases with \(k^{i}_{t}\), which implies that borrowers in rich countries are less likely to obtain external funds than those borrowers in poor countries. To view this model implication through the lens of development economics, consider poor countries consisting of mainly traditional light industries, such as textiles that requires relatively small initial start-up funds and rich countries dominated by modern heavy industries, such as steel and electrical equipment that need large initial expenditures.

When \(w_{t} > qk^{\beta }_{t}\), wage income is not fully utilized in producing capital. We can either assume that the unused part of wage income is simply wasted or there exists a default technology that can store the consumption goods from current period to next period of time. The major findings carry over with either assumption.

Since \(\frac{\partial ^{2}\psi (k_{t})}{\partial k^{2}_{t}} = \frac{[(1-\beta )k^{\theta -\beta }_{t} - q(1 - \theta +\beta )](\theta -\beta )k^{\theta -\beta -1}_{t}}{[q-(1-\theta )k^{\theta -\beta }_{t}]^{2}}\frac{k_{t+1}}{k_{t}}\), q cannot be large to ensure \(\psi ^{''}(k_{t}) > 0\).