Interbank tiering and money center banks

TLDR: In this article, a core-periphery model is proposed to find evidence that interbank markets are tiered rather than flat in the sense that most banks do not lend to each other directly but through money center banks acting as intermediaries.

About: This article is published in Journal of Financial Intermediation.The article was published on 2010-09-23 and is currently open access. It has received 372 citations till now. The article focuses on the topics: Interbank lending market & Financial networks.

Figures

Figure 5: Robustness checks. The figure shows the number of banks in the estimated core (in blue, left axis) and the total error score (in red, right axis) over time for two different experiments. In the first, shown with dashed lines, the two most connected banks (head institutions) are removed from the network. In the second experiment, shown with solid lines, all saving banks and credit cooperatives (pillars) are removed.

Figure 6: German fit against simulated error score densities. This figure compares the total error score from fitting the tiering model to the German interbank network (12.2% of links, shown as an arrow) to the normalized error scores, as defined as in equation (4), from fitting two types of random networks of the same dimension. The red bars show the histogram of error scores from fitting 1000 scale-free networks, whereas the blue bars represent the histrogram from fitting 1000 Erdös-Rényi random graphs.

Figure 4: Transition probabilities over time. This figure shows the transition probabilities for each quarter over the sample period (1999 Q1 – 2007 Q4). The lines trace out the two top rows in equation (7) over time. For instance, prob(C|P), shown in red, represents the frequency with which banks switch from the core to the periphery.

Figure 3: Structural stability over time. The figure shows the size of the estimated core (number of banks, left axis) and the total error score (expressed as a percentage of links as in equation (4), right axis) for the German interbank network on a quarterly basis.

Figure 1: Stylized example of an interbank market. The left panel illustrates a perfectly tiered interbank structure in a stylized interbank market comprising 8 banks. The arrows represent the direction of credit exposure, e.g. bank D lends to A. The middle and right panels depict examples of networks that are not perfectly tiered.

Figure 2: The core as a refinement of intermediation. This Venn diagram illustrates the relationships between various sets of banks in the German interbank market. The majority of banks intermediate, yet only a small subset of intermediaries qualify as core banks.

Frequently asked questions

Q1. What is the way to determine the optimal partition of banks?

The procedure of minimizing the distance between modelM and networkN delivers the optimal partition of banks into core and periphery. 

Q2. What is the reason why the German interbank system organizes itself around a core of money?

If tiering is not the result of random processes but of purposeful behavior, there must be economic reasons why the banking system organizes itself around a core of money center banks. 

Q3. How many possible subsets do the authors need to evaluate for the optimal core?

A medium-sized banking system of some 250 banks already requires on the order of 1078 possible subsets (2n) to be evaluated for determining the optimal core. 

Q4. What does the balance sheet variables help predict?

Balance sheet variables also help predict interbank relations in other studies (Cocco et al. (2009)), with implications for overall market structure. 

Q5. How can the German interbank network be predicted?

The authors also show that the core of the banking system can be predicted by means of a regression that uses only balance sheet variables, which is helpful since most countries do not collect bilateral interbank data. 

Q6. How many links in the network are normalized?

Recall that their measure of distance (4)-(6) normalizes the aggregate error by the total number of links in the observed network, Nij. 

Q7. How do the authors make a structural quality of interest amenable to quantitative treatment?

To make a structural quality of interest amenable to quantitative treatment, the authors formulate a procedure based on blockmodeling techniques fortting a theoretical structure to an observed network. 

Q8. What is the way to determine the optimal set of banks?

This can be thought of as running a regression, but instead of estimating a parameter that achieves the best linear t, one determines the optimal set of core banks that achieves the best structural match between the observed network and a tiered structure of the same dimension. 

Q9. What is the reason why the tiering is not expected to occur in random networks?

As tiering is not expected to arise in such networks, it must be the result of incentives of banks for linking to each other in this particular way. 

Q10. What is the way to determine the optimal set of core banks?

This can be thought of as running a regression, but instead of estimating the parameter that achieves the best linear t, one determines the optimal set of core banks that achieves the best structural match between N and M , a perfectly tiered structure. 

Q11. What is the measure of distance that Doreian et al. (2005) used?

The measure of distance the authors adopt, following the generalized blockmodeling approach of Doreian et al. (2005), is a total error score.