JOURNAL ANALYSIS

Theme
Demand Deposit
Title
Global Games and Demand-Deposit Contracts:
An Experimental Study of Bank Runs _

Author
Alexander Klos, Norbert Sträter
Finance Center Münster

Year
January 2008


INTRODUCTION
Firstly, the global games approach can eliminate the multiplicity of equilibrium present in classic panic based bank run models (Diamond and Dybvig (1983)). Therefore, the probability of a bank run can be calculated, which is a useful number when thinking about policy implications. Secondly, the global games approach allows an integration of two related, but different views found in the literature on financial stability. The panic-based view stresses that bank runs are random events, unrelated to changes in the real economy. Panic-based bank runs might even occur when the economic environment issue efficiently strong. In contrast to this, the fundamental view assumes that bank runs are a natural of the business cycle and only occur in connection with negative real shocks. In the global games approach, it is possible that fundamental runs occur, if the fundamentals of the economy and therefore the signals which agents receive are very bad. However, in other situations, fundamentals may be good enough so that the prevention of a bank run is desirable from the viewpoint of the depositors, but runs still occur due to strategic uncertainty about others' beliefs. Moderate signals about the fundamental state of the economy may lead to the belief that other agents withdraw with a high probability. In such a situation it is rational to withdraw also. In that sense, moderate expectations about the fundamental state of the financial system can trigger panic-based bank runs.

PURPOSE
Our paper is related to the recently started experimental investigation of bank runs. Schotter and Yorulmazer (2005) investigate in a policy oriented study the factors that affect the severity of fundamental bank runs. In their setup a bank run occurs by any means; they investigate how quickly the depositors withdraw their money. Their results indicate that partial deposit insurance and the existence of insiders may mitigate bank runs.
RESEARCH METHODOLOGY
In laboratory scenarios inspired by theoretical models, subjects receive a noisy private signal about the true fundamental state of the banking system. Subjects employ threshold strategies and low rates of deviation from threshold strategies.

Reviewing the theory generates a number of hypotheses which follow directly from the intuitions mentioned in the introduction. We focus on two of them:
Hypothesis 1: Subjects use threshold strategies.
Hypothesis 2: Banks become more vulnerable to bank runs when they offer more risk sharing.

RESULTS AND ANALYSIS
that higher thresholds for higher repayment rates can be observed in between- and withinsession comparisons. This result contradicts the view that within-designs point subjects to the differences the researcher investigates, and are therefore not as reliable as between-subjects experiments (see Camerer (2003), page 41 _ 42, for a brief discussion).
To test the statistical significance of the results, we take two different approaches. In light of the discussion of different matching protocols in section 2, the most conservative way to analyze the data is to analyze estimated thresholds on the session level.

Evidence on the session level
We have two observations per session (TS), one for r1 = 1.25 and one for r1 = 1.5, resulting in a total of 16 observations, eight for model A and eight for model B. We run linear regressions separately for both models to infer the influence factors of session thresholds. Dummy variables for the degree of risk sharing (RD) and the ordering of risk sharing conditions (OD) are used as explanatory variables.
TSi = const + _1 • RD + _2 • OD + _i
It turns out that the dummy for a low degree of risk sharing is significantly negative for both models as predicted by the theory. The ordering of risk sharing conditions does not have a significant impact.

Evidence on the individual level
Stronger evidence for the key role of offered risk sharing can be provided by the analysis of thresholds estimated on the individual level. As already mentioned above, we estimate an individual threshold for decisions under the conditions r1 = 1.25 and r1 = 1.5. Therefore, we have two matched observations per subject. We apply a non-parametric Wilcoxon test, separately for every session, consequently based on thirty matched observations of individual thresholds.

CONCLUTION
Our results suggest that the global games approach applied to bank runs leads to experimentally valid predictions with respect to the comparative statics. Both hypotheses are supported by the data: Our subjects use threshold strategies and increased risk sharing leads to increased thresholds at the individual and the aggregate level.
However, the rather small reaction of thresholds to changes in the repayment rates is at odds with the theory. This result is important with respect to the socially optimal repayment rate. One major advantage of the global games approach is the ability to trade off the benefits of high repayment rates (increased risk sharing) with its costs (increased probability of a bank run). Such a social optimization requires valid predictions about how people react to changes in the repayment rates. At least in our experimental setup, this assumption is questionable.