Towards a new “empirical narrative” of the Dutch financial crisis of the early 1920s1

Towards a new “empirical narrative” of the Dutch financialcrisis of the 1920s

A paper prepared for the LSE thesis workshop in economic history on 27 June 2007

Christopher Louis Colvin[1]

1. Introduction

Between 1920 and 1925, the Netherlands experienced a financial crisis which affected over sixty banks and credit institutions. This thesis concerns the causes of this crisis. This particular financial crisis is interesting for threemain reasons: (1) it involved the only classic bank runs experienced in the Netherlands in her history; (2) it marked the end of the Dutch experiment with a system of universal banking that was developed in the build up to and during the Great War; and (3) it is during this period that the Nederlandsche Bank first experimented with modern central banking.

This short paper focuses on the 1924 episode of the crisis in which the RotterdamscheBankvereeniging (henceforth the RBV) – the Netherlands’ second biggest bank bycapitalisation and a pioneer of the universal banking model – came into severe difficulties as a result of close financial and managerial links with a number of failing shipping and mining concerns.There have been a number of previous historical accounts of the unfolding of this particularepisode of the crisis, the most detailed and complete of which is probably that of Johan deVries in his history of the Nederlandsche Bank between 1914 and 1931.[2] However, these previous narratives are predominantly historical and not economically analytical. They fail to address the relative importance of individual developments. Asa result, the unfolding of the crisis in general – and this episode in particular – is not wellunderstood.

My previous research has involved constructing a new narrative of the 1924 episode of thecrisis from previously overlooked primary sources, including the internal meeting minutes,letters and briefings from different layers of management in the RBV.[3] In addition to this“insider” view of the crisis, I incorporated an external viewpoint using commentary from DeKroniek (The Chronicle), a respected contemporary financial journal. Using these qualitativesources, a brief characterisation of the 1924 episode follows along the lines ofMichael Bordo’s crisis recipe.[4] The main observed ingredients were: (a) a sudden change inexpectations as a result of new information made available through the media; (b) fear of theRBV’s solvency by the public; (c) an actual threat to the RBV’s solvency due to her riskyclients; (d) a small run on the bank in the first week of July 1924; and (e) active intervention from the central bank and the state.

An avenue that has yet to undergo any significant exploration is the effect of the crisis oninvestor sentiment. This is interesting for three reasons, outlined as follows: (1) investorsrevise their valuation of companies on an almost daily basis in reaction to new information and events; (2) newspapers reported that the RBV was in serious troubleyears before the crisis proper played out, yet runs on the bank only took place in the firstweek of July 1924; and (3) archival sources reveal that there was active intervention in theprice of RBV share price from investors with insider information.

This paper presents some very early findings for part of the first substantive research chapter of the thesis, the aim of which is the construction of a new “empirical narrative” of the crisis period 1921-1925. This paper is structured as follows. Section 2 discusses two popular explanations for the causes of bank runs and speculates on their relationship with stock market prices. Section 3 outlines the role of the Nederlandsche Bank in the 1924 episode of the crisis. Section 4 describes a new dataset that hasbeen compiled with which to examine the crisis. Section 5 outlines the methodology used to analyse the crisis. Known in the finance literature as Event Study Analysis, this methodology is a wayof measuring the economic impact of (unanticipated) corporate events using asset prices. Section 6 presents some preliminary results. Finally, Section 7 offers some speculative conclusions and a possible extension.

2. Bank runs, stock price crashes and central bank intervention

Charles Calomiris and Garry Gorton argued in 1991 that two distinct theories of bank runs have developed in the modern literature around which research has coalesced.[5] The first approach is the random withdrawal risk explanation, as formalised by Douglas Diamond and Phillip Dybvig.[6]The underlying assumption is that banks are a mechanism of insuring against risk. Agents prefer the higher returns associated with long-term investment, but may unexpectedly wish to consume their capital at an earlier date. Banks exist to allow agents to simultaneously do both.As the result of a “sunspot”, a panic occurs in which agents present their liabilities for redemption, which the bank cannot honour at par (it has fractional reserves). Such a crisis is self-fulfilling: all agents suddenly decide to redeem their claims together in order to avoid being at the end of the line.

The second approachuses the concept of asymmetric information, which focuses on the differences in information available to different parties in a financial contract.[7]Borrowers are said to have an informational advantage over lenders, as they know more about the risks associated with their investment plans. Because of this asymmetry, lenders cannot distinguish between the quality of borrowers. Lenders will therefore provide loans at an interest rate that reflects the average quality. But as in George Akerlof’s classic “lemons” problem,[8] high quality borrowers will be paying higher interest rates than their level of risk warrants and will therefore drop out of the market, leaving only low quality lemons.A banking panic occurs as a result of a sudden, but rational, revision in the perceived riskiness of bank deposits when nonbank-specific, aggregate information arrives.[9] This view holds that bank panics are essentially a natural outgrowth of weak fundamentals arising from the business cycle.[10] For the case of a deflationary cycle, wealth is redistributed from debtor to creditor by the increasing real value of debt, thus reducing the borrower’s net worth. This results in an “increase” in adverse selection, in turn causing a decline in investment and economic downturn.

Explanations of stock price crashes differ according to theorists’ interpretation of what stock price information displays.[11] Modern finance theory usually assumes that stock prices exhibit a weak form of market efficiency whereby the present price of a stock impounds all the information contained in a record of past prices.[12]This market efficiency assumption may be criticised as being quite strong. However, the argument goes that if this were not true, then financial analysts could systematically make above-average returns by interpreting charts of the past history of stock prices, something for which there is very little evidence; competition among investors ensures that stock prices reflect all information available on past performance. Under this interpretation, a company’s stock will crash if some new, negative, information about its fundamentals becomes known to investors, such as a serious profit warning, or, in the case of a bank, a bank run.

John Maynard Keynes’s alternative view of stock prices holds that they instead reflect how investors think that others value the firm.[13] Even if a particular investor holds (privileged) information about a firm, he may not react in the direction that such information would suggest. This view of investor activity can be used to explain investor “herding” and means crashes can occur at a time when the firm itself is not in trouble. Such explanations also emphasise the identity of the investor; for instance, a firm’s management arguably has an interest in inflating the value of their firm.

The role of central bank intervention in the banking sector is a controversial topic.The mainstream view is thata lender-of-last-resort(LOLR) must be present for two reasons:[14] (1) information asymmetries which make otherwise solvent banks vulnerable to deposit withdrawals and/or problems with interbank lending in times of crisis; and (2) the potential risk to the stability of the financial system as a whole through contagion. Central bank intervention in stock prices is a topic that is much less written about. However, it is an intervention that can arguably be justified along similar lines to LOLR provision,and appeared to bepresent in the policy arsenal of the Nederlandsche Bank in the 1920s.[15]

3. The Nederlandsche Bank and the 1924 crisis: help or hindrance?

According to internal management meeting minutes, between 28 June and 5 July,RBVdepositors withdrew money to the tune of 42.4 million guilders, bringing reserves to a new totalof 96.6 million guilders (a drop of 30.5 per cent).[16] Simply put, this amounts to a run on the bank. Ernst Heldring, a prominent Amsterdam businessman, wrote in his diary on 9 July that “what I can observe leads me to believe that […] people are withdrawing their accounts [from the RBV] and depositing them at other banks”. However, further than this, there is no discussion of this bank run to be found in the existing literature on this crisis.

The question asked here is what caused this bank run. On 2 July 1924, the Nederlandsche Bank issued a communiqué to all major Dutch newspapers. It proclaimed thatthe Nederlandsche was willing to maintain the RBV’s liquidity. A translation ofthe full communiqué is as follows:

The various rumours concerning the financial position of the Rotterdamsche Bankvereeniging have motivated the directors of this institution to turn to the president of the Nederlandsche Bank, whom they are providing with all materials regarding the liquidity of the Rotterdamsche Bankvereeniging.

Following an examination of the provided materials, the president of the Nederlandsche Bank has proclaimed that he is prepared to work with the directors of the Rotterdamsche Bankvereeniging, and if necessary to maintain her liquidity.[17]

De Vries presents thisintervention by the Nederlandsche Bank as an important stabilising force that helped to solve the troubles of the RBV.[18]However, the internal minutes of a meeting of the bank’s supervisoryboard held on 10 July reveal that Willem Westerman – the RBV’s president –may have thought otherwise. Westermanblamed the bank run on the interventionitself:

[…] there was talk with the Nederlandsche Bank about a f 100,000,000 support, but the Nederlandsche Bank is to limit her support to 50,000,000 for the mean time, and that support for the share price, about which the Minister had shown support, was not on the table. In return the Nederlandsche Bank demanded the publication of a communiqué in the newspapers, which our directors conceded to under pressure. This communiqué was published in the morning papers of Tuesday 2 July, with the well-known disastrous consequences.[19]

This apparent contradiction needs to be resolved.The bank was obviously in trouble, as her managers evidently admitted to by seeking assistance from the central bank. However, the question remains as to whether the manifestation of the crisis in the form of a bank run could have been avoided if intervention had been kept secret. Was the communiqué necessary in quelling the public’s fears? Negative stories about the RBV were often in the press, long before the crisis.[20]Despite this press, was the dire state of the bank common knowledge, or was it a surprise to depositors and investors?The aim of this research is to resolve this problem by incorporating shareholder action intothe narrative.

This investigation’s working hypothesis is that shareholders, despite rumours in themedia, were unaware of the gravity of the problems experienced by the RBV. They wereunaware of the risky nature of the bank’s clients and their recent troubles. When the centralbank announced that it was to intervene, shareholders reacted by significantly revising theirvaluation of the bank.The alternative hypothesis is that rationally acting shareholders knew that the bank was introuble for years and had already incorporated the possibility of a bank run into theirvaluation of the firm. The announcement of central bank intervention was therefore expectedand no significant revision of the bank’s valuation occurred.

4. Towards a new dataset for Dutch banks in the 1920s

There are no existing studies of share data of Dutch banks for the period of the crisis, orindeed for any period in Dutch banking history. I am currently in the process of collecting anew dataset of daily stock prices from the stock listing pages of the Officiëele Prijscourant of theBeurs van Amsterdam for six Dutch banks for the 1920s (referred to collectively as the big5plus1). These are the Big Five commercialbanks (in order of capitalisation: Amsterdamsche Bank, RBV, Nederlandsche Handelsmaatschappij, Twentsche Bankand Incasso-Bank) and the Nederlandsche Bank.Figure 1 below shows the closing price ofnormal, dividend-yielding, RBV stock over the course of 1923-1924. Note that the Beurs was openfor trade six days a week (Monday to Saturday), apart from during the summer (June toAugust) when it was also closed an Saturdays.

Figure 1. RBV share price, as percentage of nominal face value, 1923-1924

Source: Prijscourant

Note that share prices for this period were quoted as a percentage of their nominal face value. The face value of RBV shares depended on when they were issued. As for most companies at this time, they were in very high denominations: 250 or 500 guilders was not uncommon. Although the ownership of the shares was a secret, their high individual value meant that their ownership must have been confined to a small section of Dutch society. Their high value also meant that the shares were relatively illiquid compared to today’s share offerings.

In order to make stock prices of the different banks directly comparable with one another, logarithmic returns series must be used. These are calculated as follows:

[1]

where pricetis the closing stock price on day t and pricet-1 is the closing price of the previous day. This is calculated for all banks in the sample. A daily returns index for all six banks is also constructed, which is weighted by the traded capitalisation of the six banks, as shown in Table 1 below.

Table 1.Capitalisation of the big5plus1, in guilders

Note: * = The RBV capitalisation for 1924 refers to that following capital restructuring announced atthe Extraordinary General Meeting of shareholders convened on 27 October 1924.

Source: Vereeniging van den Effectenhandel, Gids bij de Prijscourant(1924 and 1925).

5. The Event Study Analysis methodology

Event Study Analysis (ESA) is a popular methodology that is used to measure the effect of an (unanticipated) economic event on the value of a firm.[21]Some examples of events that have featured in the ESA literature include mergers and acquisitions, earnings announcements, issues of new debt and equity and announcements of macroeconomic variables. The standard methodology for carrying out an event study is that of John Y. Campbell et al.[22] It is briefly outlined as follows.

Figure 2. Model set-up

First, the event of interest must be determined. Usually, the event window is defined as the day of this event plus or minus a number of additional days, in order to capture potential insider trading and to take account of delayed reaction from shareholders. In order to determine the event’s impact, a measure of abnormal return is required. This is the actual ex post return of the security over the event window minus the normal return of the firm over the event window, where the normal return is the return that would be expected if the event did not take place. Hence:

[2]

where , Rt, and E(Rt) are the abnormal, actual and normal returns respectively. Xt is the conditional information for the normal performance model. There are a number of approaches to calculating the normal return, broadly classified as “statistical” and “economic”. The normal returns are estimated using a segment separate and non-contiguous of the price series, called the model or estimation window. Abnormal returns are calculated by subtracting the parameter estimates of the normal performance model from the actual returns in the event window. Finally, a test for statistical significance can be carried out on the cumulative abnormal returns over the event window.[23]

6. Preliminary results

A simple OLS regression model was used to estimate normal returns for each of the banks in the sample using Stata. The model took the following form:

[3]

where Rtis the current return, as estimated by equation [1], and It is the weighted returns index of big5plus1. The estimation window was defined as [T-30, T-365], where T is the event date, 2 July 1924. The results are shown in Table 2 below.

Table 2. Estimating normal returns

Note: t-statistics in parentheses.

This is quite clearly an unsatisfactory model;this regression suffers from a large endogeneity problem in that Rtis a component of It. However, given theory and practice, it is not unreasonable to assume that investors base their pricing decision on alternative investment opportunities.Note that this is only an early model and the plan is to increase the constituents of It to more reflect alternative investment opportunities beyond banking, thus reducing the endogeneity problem.

Meanwhile, the above results were used to calculate abnormal returns and cumulative abnormal returns (CAR) in the event window:

[4]