Банковская работа в классе: сделать смешанные депозиторы снять время

A BANK RUN IN A CLASSROOM:

DO SMART DEPOSITORS

WITHDRAW ON TIME?

Starting with (Diamond & Dybvig 1983) coordination failure in deposit markets has been considered an important source of instability, increasing the risk of bank runs and the subsequent bank defaults. In their model providing insurance against liquidity shock, the standard 2-period on-demand deposit contract adds to the incentives for early withdrawals for those depositors, who are supposed to live for the whole game.  A coordination failure appears as the bank is not able to repay to everyone if they come earlier than scheduled in the contract. Being one of the Nash equilibria a bank run appears when the depositors expect others to withdraw earlier and thus withdraw themselves in order not to come to an empty bank at the end of the game. In contrast to information-based bank runs (Jacklin & Bhattacharya 1988; Chen & Hasan 2006), which are usually efficient in terms of redistributing funds from too risky banks to those who are more reliable, panic-based bank runs are not related to increased bank risks – and therefore deposit redistribution – and even may ruin a stable bank. Financial education and financial literacy are considered tools to increase the degree of rationality of unsophisticated market participant behaviour in different financial markets including that for retail bank deposits. Empirical studies show that financial knowledge and skills can increase participation in the market, as they usually make people more prone to saving strategies (Klapper et al. 2013; Semenova 2011; Beck & Brown 2011), but there is no evidence that they influence information-based bank runs (Brown et al. 2014; Semenova 2012). In this paper we extend this literature and ask whether being smart may prevent depositors from withdrawing deposits when only coordination failure is a problem and no financial deterioration appears in any particular bank. Studying the relationship between financial literacy and depositor behaviour related to withdrawal decisions is a tricky task if empirical data needs to be collected. It seems to be too complicated to collect data about the education of withdrawing depositors, let alone any more complicated measure of financial literacy. As (Dufwenberg 2015) emphasizes, the opportunity to

test hypotheses related to this link appears in an experimental set-up, where various situations and economic conditions can be modelled, and decisions are made by real economic agents, maximizing their profit. In this paper we use unique data from a series of experiments with undergraduate and graduate students from Moscow and Saint-Petersburg, modelling the a-la Diamond-Dybvig deposit market with liquidity shocks, changing macroeconomic conditions and risk-based investment technologies. We combine this data with the student academic  achievement data to examine the relationship between depositor smartness and the propensity to avoid inefficient bank runs. The rest of the paper is organized as follows. The next section describes the literature on bank runs in the experimental set-up, which we contribute to. Then we introduce the experiment design. The next section describes the data and methodology. Then we present our results and conclude

Related Literature

There are several papers studying the depositor behavior via the lens of experiments and providing some proof for the theoretical predictions. (Madiès 2006) was among the first to provide evidence of the a-la Diamond-Dybvig bank runs. He conducted an experiment with simultaneous decision-making and 210 participants, who were university business school students. His results are clearly in line with the theoretical predictions. The bank runs frequently occur as a coordination failure. The suspension of convertibility and “narrow banking” proved to be efficient instruments to prevent them. Deposit insurance is also a solution, but only full coverage provides the depositors with the necessary confidence. (Garratt & Keister 2009), on the contrary, found no evidence that the depositors withdraw voluntarily if there are no forced withdrawals (or liquidity shocks, in Diamond-Dybvig terminology) in the experiment. They show that macroeconomic instability, which causes more liquidity shocks, may also increase the frequency of bank runs. Having more chances to withdraw and observing the withdrawal history, depositors are more likely to run on a bank. (Schotter & Yorulmazer 2009) contradicts (Madiès 2006) to a certain extent, as they prove that deposit insurance, even partial, works well in preventing early withdrawals. The results of their experiment with undergraduate students show that a sequential game provides more stability in the banking sector compared to a simultaneous one, especially provided that detailed information about withdrawals is available. Introducing insiders – depositors knowing more about the quality of the bank – also reduces the severity of bank runs, leaving panic runs aside. A coordination failure may occur as a result of the complicated coordination process. As (Arifovic et al. 2013) suggest, the higher the coordination parameter (the higher the proportion of the depositors who should not withdraw early needed to ensure higher gains for patient depositors compared to the impatient ones) the higher the probability of a bank run. The nature  5 of these bank runs is not based on sunspots. However, people learn over time and, in repeated

games with an increasing degree of coordination required, depositors withdraw less frequently. The role of sequential decisions and the observability of actions is also confirmed in (Kiss et al. 2014a) and their experiment with 48 undergraduate students. Considering 3-period games with depositors being in a line to decide whether to withdraw (two real people and one player being a computer simulation), they show that if the second depositor can observe the actions of the first one (and she is not a computer), this reduces the probability of further bank runs significantly. This is particularly true for the first depositor deciding to be patient. Being aware that she is observed by both followers, the first depositor also withdraws less frequently. (Davis & Reilly 2016) experimentally model the influence of the change in repayment proportions (re-contracting) and information about withdrawal behaviour on the fragility of distressed banks. The experiment was conducted with 240 undergraduate students. The authors show that if contracts are changed in favour of patient depositors, bank stability is less undermined. However this is not true if a sequential game is introduced and the participants observe the withdrawals in the first stage if they are assigned to decide in the second stage. There are a few papers focused mostly on modelling the contagion in the deposit markets in an experimental way. Contagion appears as an informational phenomena in several-step games where some proportion of depositors receive a signal about the deterioration of the bank’s financial condition and all the rest observe their actions and act according to what they see and what they know about the kind of signal the informed depositors may have received. Using the results of the experiment with 200 undergraduate students in a two-bank set-up, (Chakravarty et al. 2014) show that observing the actions of the informed depositors makes uninformed ones withdraw even when their bank is unrelated to the bank of informed ones and there is no correlation with their financial position. In a continuous game the inefficient run is difficult to stop even if the informed depositors do not withdraw intensively. (Brown et al. 2016) study the channels of the contagion transmission when new information on bank fundamentals is revealed and is not promising. Basing on the results of an experiment with 264 undergraduate students participating in a sequential game, they show that if a signal on withdrawals in the first stage is quite informative, depositors withdraw earlier and mostly because they are afraid that others will withdraw, not because their expectations about bank stability suggest doing that (the coordination failure described in (Diamond & Dybvig 1983)). 6 Quite a few studies dealt with students in their experiment: being useful for the students themselves in their understanding of bank runs, as (Balkenborg et al. 2011) suggest. This type of participant provides an excellent opportunity to add the proxy of financial literacy to the analysis, however, as far as we are able to judge there are no papers incorporating student academic achievements into the models of bank runs. This study fills this gap in the literature. Leaving the use of complicated financial literacy measures for further research, we use a very simple proxy for it – the academic achievements. The paper closest to ours is (Kiss et al. 2016), where the authors introduce the cognitive abilities of the depositors proxied by the results of the standard Cognitive Reflection Test (CRT). In their sequential withdrawal experimental set-up they show that depositors with higher CRT withdraw less frequently in situations of strategic uncertainty (when no information on previous withdrawals is available). Our study is different in several ways. First of all, our set-up implies a simultaneous game with multiple independent rounds, modelling several economic scenarios. We aim to detect the coordination failure under different conditions. Secondly, our proxy for financial literacy measures mostly the overall knowledge of the participants and their degree of smartness, while CRT shows the ability of the respondents to go beyond the first – and incorrect – answer coming to mind and to think a bit more about the correct answer. We therefore appear to be closer to the nature of the financial literacy as a characteristic of depositor financial knowledge and skills.