An inquiry into the nature and causes of insider trading Master Thesis Tilburg School of Economics and Management Finance MSc

September 26, 2014

Author: Supervisor: R.V.A. Manders Dr. R.G.P. Frehen s646397 Table of Contents 1. Introduction ...... 2 2. Literature review ...... 4 3. Historical background ...... 7 4. Data ...... 11 Descriptive statistics & preliminary analysis ...... 13 5. Inquiry I: learning from insiders ...... 18 Methodology ...... 18 Analysis of results ...... 19 6. Inquiry II: Insider trading ...... 20 Methodology ...... 20 Analysis of results ...... 23 7. Conclusion ...... 30 Acknowledgements ...... 31 Bibliography ...... 31 Appendices ...... 33 Appendix A ...... 33 Appendix B ...... 34 Appendix C...... 35 Appendix D ...... 36 Appendix E ...... 38 Appendix F ...... 39

1

1. Introduction The academic debate on whether to regulate insider trading has been going on for many years. Plain said: one side calls for regulation, another against.

The side that supports deregulation says that insider trading leads to better security pricing (market efficiency), which is beneficial to society as it improves the allocation of capital. The prime economic argument of those pleading for regulation is that insider trading reduces the appeal of stock markets to investors. They state that because of the unjustified profits to insiders at the expense of investors, people will be more reluctant to participate in the equity market and risk premiums will increase, which would diminish the use of financing through joint stock companies.

Even though fairness is an important issue in this debate, few scholars ask whether the outsider who is harmed by this unfair insider trading had a fair chance to detect the insider activity, let alone whether he is a poor trader who often conducts ill-advised trades, irrespective of insider trading. It is probable that counterparties, as a group, share certain characteristics that explain why they traded with insiders.

There is a widespread belief in the investment community that insiders possess valuable information that is exclusive to them. For that reason the investment community takes an interest in trading by insiders. This might limit profits to insider trading through a price run-up caused by outsiders who learn about insider activity. Therefore, it makes much sense for insiders to hide their trading activities from the market, regardless of regulation. Similarly, it makes sense to outsiders to find out about insider trading. So if counterparties, for example through their broker or market talk, would find out about insider trading and reverse course, this would be a significant limitation to insider trading.

Very little research has been done on insider trading in an unregulated environment and still less with full access to trading data of all insiders and outsiders. Using hand-collected microstructure data from 18th century London, I analyse the stock trading by directors and their counterparties in great detail. Whereby the purpose of this study is to learn whether insiders can freely make trading profits in an unregulated market and how this affects or hurts their counterparties. I find evidence that counterparties of insiders only suffer when trading with insiders, not when trading with fellow outsiders. This is found in a market quite similar to modern markets with respect to privacy of insiders in the market.

This thesis is divided into seven sections. Section 2 gives an overview of relevant present literature on the field of insider trading. Section 3 provides necessary historical background. Section 4 describes the

2 data. Section 5 is the analysis of learning from insiders by the sub question: did counterparties of insiders start to trade in the same direction as the insiders after their ill-advised trade? Section 6 analyses insider trading and its effect on counterparties by the following questions: did insiders engage in profitable insider trading in the unregulated 18th century London market? Did counterparties underperform as a result of insider trading? And if so, what caused this loss? Section 7 contains the final conclusion and recommendations for future research.

3

2. Literature review This chapter is structured according to four themes that are relevant to multiple research questions each. The first part, returns to insider trading, looks into what results could be found depending on market conditions. The second part summarizes literature on how detection of insider trading should work. The third part, literature on the insider trading debate, explains an important concept for the debate on the fairness of insider trading. Fourth, the limits on insider trading in the 18th century markets are discussed.

Returns to insider trading

Most studies on insider trading study the regulated market. There has been contradicting evidence as to whether or not insiders can generate short-run abnormal returns in that sort of market. A somewhat more medium-term approach finds that insider trading predicts aggregate returns 6&12 months ahead (Seyhun, 1992). This notion could be helpful in this study, as like in the contemporary regulated market insiders in 18th century England did not have specific information linked to a clear time window, but rather overall superior information about a company.

Friedrich et al. (2002) find that insiders engage in market timing on the London Stock Exchange and that their trades have slight short-term predictive power in that regulated market. They also find that stocks after insider buy trades show higher abnormal returns than after sell trades. This makes sense, as a buy trade has only one explanation: the stock is a good buy according to the insider. A sell trade can have numerous reasons that are unrelated to the stock; a director might sell shares for liquidity reasons. Friedrich et al. also find that short run insider trading stock returns are statistically significant, but not economically. They find clustered trades and trades of medium size to be more informative. The explanation is consistent with the stealth trading hypothesis (Barclay & Warner, 1993). This says that insiders probably try to conceal their information by trading cautiously. As this study concerns trading by insiders at least 2 months before earnings reports, we may conclude that the returns are not necessarily due to informed trading, but perhaps a capability of insiders to assess the value of the company.

On trade volume, prices, and detection of insider trading

One can see prices as a system to communicate information (Hayek, 1945). From this notion by Hayek, the theory of price decoding emerged that says that when a stock price deviates from its fundamental value, the only logic explanation must be the occurrence of informed trading. And that price effect leads outsiders to discover insider trading. Of course, in a fully efficient market this logic would suffice, but the

4 reality is much less simple (Gilson & Kraakman, 1984). For this study’s purpose, this way to discover insider trading will unlikely explain results, as counterparties of insiders should have noticed a price effect by the time they trade with the insider. We will instead think of market talk, leaking of evidence, or direct observation of insiders as ways of counterparties discovering insider trading.

Meulbroek (1992) investigated illegal and prosecuted cases of insider trading and observed a mean price run-up index of 47.56% towards major events that were (ab)used by insider traders. Half of this run-up came from days on which the insiders traded, the other half from other days. The fact that one half of the run-up comes from days on which insiders do not trade, supports the idea that outsiders can detect insider trading and start trading in the same direction as the insiders. However, Meulbroek’s sample is biased as it only covers those who got caught. Because it takes a lot of time and resources to successfully prosecute insider trading, the SEC does this sparingly. As Dooley (1980) puts it: “Predictably, the SEC commenced investigations in very few of the cases referred by self-regulatory authorities”. There were many cases with a lower run-up towards major events, which were probably harder to detect by the market. Or as Bainbridge (2000) puts it: “empirical studies of SEC case files will be inherently biased towards cases in which insider trading coincident with noticeable price or volume effects”. The impact of insider trading on price or volume may well not occur in most cases. It is therefore very interesting to know whether outsiders learn from insider trading regardless of these price and volume effects.

Easley and O’hara (1987) provide an alternative explanation for the common observation that large trades have a comparatively large price impact. Before their study, the usual explanation for the price effect of large trades was the liquidity effect, which entails that market makers have to adjust prices sharply to make sure their inventory of shares remains stable when confronted with large trade sizes. The explanation given by Easley and O’Hara is the information-effect theory, which says that as informed traders prefer high quantity trades, market makers use the signal of their trading to protect themselves against losses from trading with these informed traders, thus large trades lead to sharper price adjustment by dealers. This is consistent with the stealth theory, in the respect that insiders do well by concealing their activity by trading only medium amounts of stock. More importantly, it gives support to the notion that insider trading can be detected by outsiders.

Literature on the insider trading debate

An important issue in the ethics debate, and to the rewards of security analysis, is if counterparties are induced to trade by insiders through price drifting away from fundamental value.

5

Bainbridge (2000) says that insider trading is unfair insofar insiders induced traders to conduct ill- informed trades by means of a price effect (price run-up). He explains that if this is not the case, it is purely fortuitous that an insider was at the other end of the trade. For example: an (hypothetical) intelligent investor who knows the fundamental value of a security based on public information, sees its price rise, and decides to sell because the security trades above its real value. So if this price increase was fuelled by an insider, he induced the smart trader to trade and directly caused his loss. If on the other hand, the counterparty has no idea of the stock’s value, or decides to buy a stock because it has gone up, it is fortuitous that an insider is at the other end of the trade. From the direction the stock price took before the trade and the direction of the insider trade, we can find out whether the counterparty was harmed by the insider or if it was by chance the insider was at the other side of the trade.

As for the unregulated market, insider trading could reduce investor confidence, leading to lower market participation (Bainbridge, 2000). Also, inside information could render days of equity research useless. It is therefore, on average less profitable to conduct equity research, which in turn may reduce market efficiency.

Limits on insider trading

As Bhattacharya & Daouk (2002) point out, it is not the regulation itself that stops insiders from engaging in insider trading, it is enforcement of regulation that stops insider trading. If there were some sort of guidelines on what was right or wrong to do in 18th century London stock markets and if these were enforced in some way, that would impact the insider trading environment.

Evidence for some kind of enforcement is provided by the case of the broker Charles Blunt. Blunt was involved in stock manipulation of the Sword Blade Company and cut his throat because of this in 1720 (Bell, 2012). This signals that there was some kind of enforcement of ethical values. Apart from this being a limit on insider trading and market manipulation, this makes it more likely that insiders tried to conceal their trading from the investor community. This makes it more interesting to study whether insiders could justly fear detection of their insider trading activity.

Further support for this idea comes from an earlier study of BoE stock during the South Sea bubble. Merchants, gentlemen, and people in the financial sector, suffered the largest losses as a result of the bubble. Women, people in the food-sector, and soldiers were most profitable (Carlos & Neal, 2006). The BoE directors were members of the upper echelons of London society; some were MP’s, or merchants.

6

3. Historical background This study is conducted in a market that is in some respects different and in others surprisingly similar to present day markets. It is important to shine some light on the mechanics of the market of that time to fully comprehend the results of this study.

Early 18th century London stock market environment

In the early 1700’s stock market trading went through brokers, who commonly went to dealers to get stock quotes. An act in 1697 called "To Restrain the number and ill Practice of Brokers and Stock- Jobbers" regulated the brokers. After an adjustment in 1708 the number of licensed brokers was unlimited, unlicensed brokers could be fined £25, and brokerage fees were limited to ½ percent. The regulation of licensing brokers was not effective, as many operated without license. Dealers (or jobbers) provided brokers with liquidity, for an actively traded stock a dealer would commonly offer a broker both buy and sell quotes; it was then the broker’s job to find the dealer that offered the best acceptable price for his client. Brokers sometimes used personal accounts to fulfil client’s orders, thus acting as principal. Trading was not done exclusively by brokers, as some investors traded directly in the market, thus saving brokerage fees (Cope, 1978).

As the vast majority traders hired a broker, (gave him a limit price) and trusted him to find the best price in the market, we can conclude that the stock market in that respect has not changed that much over the past three centuries. Since the deregulation in 1986, the London Stock Exchange dropped the distinction between jobbers and brokers and began to look more like 18th century markets than the market Cope observed in 1978 (apart from for the digitalization).

The people who traded on the London stock exchange were of diverse social-economic backgrounds. The backgrounds ranged from the nobility and merchants to soldiers, clerks and widows with “the majority of those involved in the market in 1720 did not come from the top end of the social order” (Carlos & Neal, 2006). Interestingly, there was a fair number of women active on the London stock market, who as a group, outperformed men during 1720-1725 (Carlos & Neal, 2006).

Like today, there were tools available to speculate on an increase or decrease in the price of common stock in the 18th century. There were generally three tools, options (both put and call), forwards, and buying on margin. There are no reports of borrowing stock to short. (Cope, 1978).

7

In the 17th century options were already a very common tool used to speculate in the Netherlands, and to some options served as a replacement of shares (Petram, 2011). It is probable that sometimes traders traded options instead of shares when they wanted to trade on inside information.

The timeframe of 1715-1725 is special, because the South Sea bubble of 1720 is right in its middle. For us it is important to know that this affected the stock price of the Bank of England (BoE). A chart of the stock price can be found in Appendix A.

Bank of England

Nowadays we know the bank of England as the British central bank, which is primarily concerned with monetary policy, issuing notes & coinage, and acting as a central bank for commercial banks. In the 18th century, however, the bank was wholly private owned and its main activity was lending money to the government. It accepted deposits, and was by royal charter the only Limited Liability Company that could issue banknotes.

Bank of England stock was “the least speculative stock among the major joint-stock companies whose shares were available to investors” (Carlos & Neal, 2006).

The trading volume of the BoE stock was a few millions of nominal value per year, spread over about 2000 trades, up to a maximum of 7000 trades in the bubble-year (Jan-Nov.) 1720. At the time, the BoE stock was among the top most actively traded, if not the most actively traded stock. To give a further idea of trading habits: the majority of trades had a size of £500 or £1000 in nominal stock, or 5 to 10 shares (a share had a nominal value of £100, but could be split). At £5,559,995, the BoE had the highest share-capital in the London stock market after the South Sea company in 1717 (Scott, 1951).

Another important note is that the BoE did not issue annual or quarterly earnings reports. The bank did, however, pay a biannual dividend.

The Bank of England required minimum shareholdings in order to be allowed to be elected director, deputy-governor, or governor (Carlos & Neal, 2006).

Regulatory background

In the early 18th century, there were no laws limiting insider trading in any form. The corporate charter of the Bank of England also did not restrict insider trading. It is possible that there was self-regulation as the

8 wealthy people in London knew each other and did not accept one taking unfair advantage from one another. Adding to this is the fact that some directors were public figures, like MP’s.

Currency

In 18th century England, the pound sterling was not divided in smaller fractions by decimals, but by an odd system of shillings and pennies. One pound (£1) was worth 20 shilling (20s.), which were worth 12 pence (12d.) each. This system ended on ‘Decimal Day’ in 1971. The purchasing power or value of the pound depends on the desired definition. A typical transaction of £1000 in 1717 would be equivalent to £132,700 today adjusted for inflation, but it would be worth £1,836,000 in terms of today’s labour value1. To put the numbers that will be mentioned in this thesis further in perspective: the median annual salary in the Bank’s accounting office was £50, and the head accountant’s salary was £200.

Transaction costs

During the period examined brokers charged a commission of 1/8 percent on the nominal value of stocks traded. Little is known about bid-ask spreads in those days, but a source from 1733 tells it commonly ranged to of a percent (Cope, 1978).

Settlement

Settlement is the final step of a stock trade. It is the time at which parties actually transact stock for money. In the early 18th century, people distinguished two types of settlement: for money or for ready money. In deal for (ready) money settlement usually took place within two days. Settlement of a deal for time took place after a specified period of time, in a fashion similar to modern day forwards. Trades could not be settled during “shuttings”, which were semi-annual closings of the books to prepare dividends or interest payments. While it was not possible to trade for money during these shuttings, it was possible to deal for time (Cope, 1978).

Dividends

Dividends played a very important part in 18th century capital markets. Especially BoE stockholders had to look towards the dividends for their rewards, as the dividend yield was high while stock volatility comparatively low. For example, in the 10 years from September 29 1715 to September 29 1725, the BoE stock price advanced from 125.5 to 133, which translates to an annual gain of less than 0.6%. Luckily for

1 http://www.measuringworth.com/ 9 the shareholders, the dividend was somewhat more impressive. The BoE paid a biannual dividend of £3 - £4 (per £100 of nominal stock), which provided an approximate annual yield of 5% - 7% during the years 1715-1725. A committee of BoE directors decided on the dividend to distribute during the non-trading periods in March and September, right after the Exchequer paid the BoE. There was an actual ex- dividend day, on which the transfer books were closed, holdings copied, and dividend warrants for each stockholder were prepared (Von Phillippovich, 1911). These dividends were paid before trading resumed in a fashion similar to the picture on the cover on this thesis. An overview of dividend amounts and ex- dividend dates can be found in Appendix B.

Margin on stock

The BoE had decided to let investors get margin on BoE stock and, starting on 10 May 1720 mortgaged 29% of its stock (Carlos & Neal, 2006). The reason given by the court of directors for this initiative was that BoE stock had to compete for capital with South Sea Company stock. It is known who made use of this possibility to mortgage their stock, as the shares were placed in a special account. The court of directors decided to end the programme on the 6th of October 1720 and let it run-off.

10

4. Data For the purpose of this study I created two datasets with account balances and trading data, one with all insiders from 1715-1725, and one with the counterparties of those insiders from 1715 till Sept 29 1720. The insiders were collected separately from the counterparties. An account is categorized as insider if at least one of its holders was on the Court of directors of the Bank during that 5-year period or has been in the previous five years. For example: a director of 1722-1724 is categorized as an insider only for the 1720-1725 period, while a director of 1717 is labelled an insider throughout. The reason for labelling a trader an insider after his directorship is that he could still have friends on the board and perhaps had access to special information through them or other company sources2.

The names of insiders were retrieved from the minutes of the directors of the BoE. In a given year, there were typically 24 directors, 1 deputy governor and 1 governor. There are 50 directors in the sample.

The data on trades and holdings, including dates, identities of traders, buy/sell volume, etc. is hand collected from the Bank of England stock ledgers of 1715-1725. All publicly traded companies kept records of who owned their shares for voting and dividend payment purposes (Carlos & Neal, 2006). There were multiple books/ledgers simultaneously in use, each covered traders within an alphabetical range. From time to time, these books ran out of pages and space and new books had to be commissioned. It was the habit that the accounts were transferred to the new books more or less simultaneously. The exact dates and timing sometimes varied, the books of 1715-1720 had no fixed starting date and used days in May, June and July -1715 as first record dates for accounts. The books starting in 1720 however, did have a shared first date: September 29, 1720. The ledgers contain some (mostly date-related) errors of which many we managed to correct by linking trades to their counterparty through different means. We performed numerous checks to get the data to be reliable, for example we checked all trades that were reported to be on Sundays (there was never trading on a Sunday), Christmas, and during shuttings.

Data on stock prices are from John Castaing’s “The Course of the Exchange and other things”. Castaing was a broker who began issuing his (one sheet) newsletter of stock & commodity prices in 1698. A limitation of the price data is that it is daily, and I therefore use these as closing prices. Because it was possible to deal for time during shuttings, there are prices available for those dates as well.

2 The, director account of Richard Houblon has been moved to the counterparty population. This is this was done because he was an insider in 1719 only, in which he visited few meetings and received an inheritance, which was his only trade that year. In April 1720 he was a counter trader of an insider, and all this makes me conclude that Houblon was not part of the Bank of England old-boys/insider network. 11

The data on dividend amount and meeting dates (announcement of dividend) was retrieved from the minutes of the court of directors. These were then combined with available price data to identify the exact ex-dividend date. There is no clear data on when the shuttings began and how long they lasted, but from the available trading data it is possible to distil a reliable picture of what these dates were.

A remarkable note to the data is that, for the first part of the 18th century, all of England used the Julian calendar, which mainly differs from the Gregorian calendar in respect to the date on which a new year begins, which is March 25th on the Julian calendar. To distinguish between the two systems, Gregorian calendar years will be followed by the letter ‘G’ (e.g. 1720 G).

For every statistical, descriptive, or graphical purpose, I removed (trades with) the mortgage account John Hanger & others from the data. The account is part of the insider dataset and all trades with this account were meticulously traced back and labelled as such.

The BoE issued a substantial amount of stock in 1723. The folios of these issues to the public are missing, so is other vital information on this SEO. The issue took place over a long period of time and it is uncertain who was eligible to buy, at what price and with what kind of delay. Therefore, stock transfers relating to stock subscription trades are omitted for every statistical purpose.

Within a given week, stock trading was very concentrated. Most trading took place on Fridays (29% of sample volume) and Tuesdays (26%), with little trading on Mondays (2%) and Saturdays (1%). Tuesdays and Fridays coincided with the semi-weekly issues of Course of the Exchange by Castaing. The BoE transfer office was always closed on Sundays.

12

Descriptive statistics & preliminary analysis Before looking into the data using more advanced statistical and analytical tools, this section is meant to give a basic understanding of the data and does some exploration towards answers to the research questions.

Table 1 below lists some basic properties of the data this study works with. The year 1720 is singled out because this is a bubble year, in which there was much more trading than in regular years and more opportunities for insider trading because there was material non-public information available to insiders with regards to the government’s debt for equity trading program.

Table 1 - Trading data Panel A: Sell trades Buy trades Insiders N directorship trades 374 417 1715-1725 - of which 1720 109 130 - 1720 in % 29% 31% Average trade size 1534 1333 Clustered trades 32 48 - of which 1720 10 22 N trades in sample 673 620 Panel B: Counterparties N counterparty trades 6201 6398 1715-1720 - of which 1720 1368 1605 -1720 in % 22% 25% Average trade size 1053 1068 Total volume 6.529.653 6.833.064 - of which 1720 1.367.092 1.550.860 - 1720 in % 24% 20% Notes: N directorship trades are the number of trades done by directors while in office. Clustered trades are trades that are preceded by two net trades in the same direction within the preceding 5 days; so every clustered trade represents at least 3 trades over one week time. N trades in sample is the total number of trades in the insider sample, it includes the trades of directors before and after they were in office. N counterparty trades represents the total number of trades in the counterparty sample.

The average trade size is a lot higher for directors than for counterparties. This has two plausible explanations: one being that insiders are wealthy people, thus trade larger amounts of stock, another being that informed insiders want to trade larger amounts, like the information-effect theory predicts.

What is even more striking is the outsized trading volume of insiders during the bubble year. In relative terms, insiders trade 1.7 times more than their counterparties in those 6 months. This might lead one to think there was a large amount of valuable information to trade on.

13

Because trading by Bank directors during the bubble is remarkably heavy, it deserves special attention. The chart below displays net trading flows (in nominal value) from January to December 1720 G.

Monthly trading by insiders in 1720 G 40000 180% 35000 30000 160% 25000 20000 140% 15000 10000 120% 5000 0 100% -5000 1 2 3 4 5 6 7 8 9 10 11 12 -10000 80% -15000 -20000 60% -25000 Net volume (lhs) Price (rhs) -30000 40%

Chart 1: Price is the monthly average stock price for transactions. The price is indexed to the closing price of Dec 31 1719, which was £150.5s (£150.25) per share. Net volume is buy minus sell volume. The chart includes trades by retired directors.

As the chart markedly shows, insiders bought heavily in February, March, and April, just before the bubble took off. These months also had high buy/sell volume ratios, with 34.25 for April, and infinity for March, as no insider sold shares that month. Another remarkable month is September, which had a very low buy/sell ratio of 0.12 just before the price slid during September’s shutting.

A calculation, using the cash flows and transaction volumes sums profit insiders made on their trading activities (disregarding their prior stock) in 1720 G to £10,811 (of which £3,155 dividends). On an average capital employment of almost £80,000 that makes a return of 13.6% which is quite good for a year in which a buy-and hold strategy would have yielded 3.2%3. This excludes the pre-existing balance worth over £ 375,000 which yielded them £ 11,875 (3.2%), bringing total the profit for 1720 G to £ 22,686. In simple terms: the stock trading almost doubled the cumulative insider profits.

It might well not be a coincidence that when in April 1720, the court of the BoE came with the idea of the BoE lending money to its shareholders, the directors were heavily invested and still investing in BoE stock. Their possible benefits from the plan were twofold: firstly, it would increase the demand for BoE

3 Price 12/31/1719: 150.25, price 12/31/1720: 147.5. Dividends 1720 G: 7.5 per share 14 stock, which increased the value of the director’s holdings; second, the mortgaging increased their capacity for insider trading by providing funds to buy more shares. Six directors actually did use the brand-new opportunity to mortgage their stock in May and June 1720.

Insiders tried to temporarily aid the value of their stock by upping the dividend to £4 in September 1720. Before that time, the dividend had been £3½, down from £4. The insiders had just begun their sell-off and the bubble was about to burst. So they hiked the dividend to provide a cushion on which they were able to sell their stock after the end of shutting – which was what they actually did. Of course, the £4 was a drain on the Bank’s capital, and contributed to the abrupt dividend cut to £3, six months later.

It is puzzling why the insiders did not sell-off their recently purchased shares during the summer of 1720, or at least in September, when they must have had an idea it was a temporary bubble. Instead, they sold most of their shares only after the price declined to a 5-month low in October. Maybe they were being cautious to cash-in on their superior information because they feared other people would find out about their trading activities. But it is more likely that they really did not have a clue about the upcoming crash, as they still bought shares in July and August. An investigation of the learning effect later on will tell us if fear of publicity could have been reasonable. Lack of possible counterparties to sell to cannot be a plausible cause because July’s and August’s trading volume of the counterparty population is roughly twice and four times the average of the previous 2 years, respectively. To top that, these (former) counterparty traders were net buyers in July and August, so selling would have been possible. Counterparties In contrast to the insiders, the counterparties did not fare well during the first nine months of 1720 G.

Monthly trading by counterparties in 1720 G Chart 2: The stock prices and 80000 180% volumes are the monthly averages for the full 60000 160% counterparty group. The price is indexed to the 40000 140% closing price of Dec. 31 1719, which was £150.5s. 20000 120% (£150.25) per share.

0 100% 1 2 3 4 5 6 7 8 9 -20000 80%

-40000 60% Net volume (lhs) Price (rhs) -60000 40%

15

At crucial points, the counterparties made different choices than the insiders, note February, March, and September. At some other points however the counterparts did seem to trade like the directors. Charts showing sample volume by insiders and counterparties can be found in Appendix D. Table A3 of the appendix shows interesting patterns in the stock holdings of insiders in relation to stock price.

The bubble was clearly a time in which insiders traded especially heavy and showed interestingly undivided trading patterns. It is therefore interesting to look into how the trading profits of the insiders and their counterparties compare for that period.

Table 2 below summarizes trading activity of insiders and counterparties for the first three quarters of 1720 G. This calculation assumes a begin balance of 0, and purely tracks the additional performance from trading in that period. A full overview of trading cash flows can be found in Appendix C.

Table 2 - 9 Months Activity Statement 1720 G Insiders Counterparties Other traders Net cash flows -£ 81,719.00 -£ 189,092.00 £ 270,811.00 Dividends £ 3,155.00 -£ 630.00 -£ 3,785.00 Ending stock £ 59,775.00 £ 81,605.00 -£ 141,380.00 Value ending stock £ 113,572.50 £ 155,049.00 -£ 268,622.00 Total gains/losses £ 35,008.50 -£ 34,673.00 -£ 1,596.00

Total nom. stock traded £ 252,975.00 £ 4,561,888.00 N/A Average balance £ 42,733.00 -£ 17,856.00 -£ 24,877.00 Buy-and-hold return £ 20,191.34 -£ 8,436.96 -£ 11,754.38 Extra return 73% -311% 86% Average price £ 136.71 £ 231.72 £ 191.55 Notes: This table summarizes the financial consequences of the trading activity (market timing) by both the insiders and their counterparties during the first part of the bubble (1/1/1720 G - 29/9/1720 G). Net cash flows represents net cash flow from trades. Dividends is the cumulative dividend received on trading balances during the period. Ending Stock is nominal amount of stock net purchased or sold in the period by the corresponding group at Sept. 30, 1720. Value ending stock is the value of the ending stock at the prevailing price of Sept. 30, 1720 of £190 per share. Total gains/losses is the sum of value ending stock, dividends, and net cash flows. Average balance is the average amount of nominal stock holdings (buys minus sells, so it can be negative). The buy-and-hold return is gross return if the group would have bought their ‘average balance’ of stock on December 31 1719 and sold September 30 1720 (prices are £150.5s & £190 on those dates, respectively). Extra return is gains/losses as a percentage of B&H return. Average price is the average share price paid (insiders and counterparties) or received (other traders) for the ending stock.

It appears from the returns that the insiders made a colossal profit, 73% over and above the buy-and- hold return on their average balance, while the counterparties lost their average net exposure twice. Add to this the sizable balance of the counterparties, and the fact the stock price plummeted after the

16

September shutting, and it gets really hard to imagine how the counterparties could have escaped the year with a profit. Also note how both insiders and counterparties were net buyers during the bubble. It is somewhat strange that both ended up with completely different returns due to market timing. Note how the loss of the counterparties is of roughly of the same size as the profits of the insiders.

The third column of Table 2 contains third party traders who sold 1414 shares (£ 141,380.- nominal stock) to the counterparties at an average price of £ 191½ each, which was a good trading result, accounting for the fact that the bubble was about to burst.

The overview of trading cash flows in Appendix C shows that in the period from 1715 to 1719 insiders were net sellers and received an average of £ 158 per share. Over that same period, counterparties, as a group, purchased shares at an average of £ 139. This is a bit puzzling, as one might expect those to be equal. But of course, the difference is explained by counterparties trading with third parties (the counterparties of the counterparties). A quick calculation learns that counterparties of insiders bought shares from third parties at an average of £ 132½ per share, while the average price in the 1715-1719 period was £ 142, and £ 146 weighted for volume4. This leads us to think that counterparties were not such bad traders, as the counterparties did better than third party traders. Surprisingly, wealth was transferred from third parties to insiders, with the ‘counterparties’ acting as intermediaries. The bubble is a notable exception during which this was not the case. But by no means can the trading between insiders and their counterparties be seen as advantageous to the counterparties.

Trading after directorship

Interestingly, directors also did profitable trades after their directorship, , Esq. a former director and relative of the contemporary director Sir John Ward, also seemed to be an informed trader. He bought 30 shares in April 1720, sold 10 in May for a profit of £ 400, and he sold his last 20 shares on the last day of trading before the September’s shutting at a handsome profit of £ 1460. Conceivably John Ward, Esq. was informed by the director Sir John Ward, who had not traded on his own account in 1720.

4 Buy and sell volumes are calculated per month and summed to get the monthly volume. The average price per month is multiplied by the corresponding fraction of total trade volume, and that total is the price weighted for volume used in the calculation. 17

5. Inquiry I: learning from insiders Did counterparties of insiders start to trade in the same direction as the insiders after their ill-advised trade?

Methodology There is one main way in which I intend to measure learning by counterparties. This is done by directly observing the trades by the counterparties of insiders. When they consistently reverse course after trading with an insider, this is a strong signal that learning takes place. The reversion (buy after sell or sell after buy) should take place after a short while, because using a longer term approach would include too much noise. Actual days will be used instead of trading days because the spreading of a rumour depends on time, not on trading days. Because days of the week can have significant effects on results (e.g. people meet in church each Sunday), leaps of 1 week are used.

I will define reversal as the cumulative net buy volume following a sale to an insider or the net sell volume after a buy from an insider. Note that this number can be negative in case the counterparty continues its old behaviour after an insider trade. The reversal effect is grouped according to the original trade being a purchase or sale. This gives 4 samples: 7-day buy and sell and 14-day buy and sell.

I will run t-tests on the reversal for trades with insiders paired with average reversal by counterparty.

Counterparties with less than 3 recordings (trades and begin/end -balances) will not be included in the sample due to passive trading behaviour.

To use returns in the test is pointless because returns would cover for the full reversal period while the next trade could have taken place anywhere in that period. This would then only clutter the picture.

18

Analysis of results Table 3 below displays the reversal effect by which we measure whether there has taken place a change in traders’ behaviour after trading with insiders. As the table shows, the reversal values are negative, meaning that counterparties of insiders on average continue buying after buys and keep selling after sells. When looking at the two week period, the behaviour of counterparties is barely affected by insider trading. For the one week period, there is no clear pattern, and insofar there is a difference, it is statistically insignificant.

Table 3 - Copying of insiders Sell Buy 7 days 14 days 7 days 14 days Reversal -111.68 -345.68 -348.92 -324.43 Normal reversal -278.93 -379.05 -209.18 -335.94 Difference 167.25 33.37 -139.74 11.51 T-stat difference 1.0569 0.2447 -1.335 0.0715 Notes: this table tests the reversal effect using a paired t-test, as laid out in the methodology. It only contains trades by the counterparty population. Reversal is the average (per trade) trade volume in a different direction (change) minus the trade volume in the same direction when trading with insiders. Normal reversal is the same as mean reversal, except that these are trades with non-insiders (the control group). Difference is the difference between mean reversal and normal reversal.

From these results we can conclude that counterparties of insiders did not start to trade in the same direction as the insiders after their ill-advised trade. This leads to the conclusion that information did not spread through the market.

This inquiry provides a foundation for the rest of the study by showing that insiders could trade freely without raising suspicion. In that way, the 18th century market is similar to modern markets, where in the absence of regulation, insiders are completely anonymous.

There is one more factor that could be added to this inquiry, which is stock return after insider trading. An insider trade should cause short-term price movements if it is identified as such by the market. These short-term price movements are properly studied in Inquiry II on page 26 and 27 with figures and explanation respectively.

19

6. Inquiry II: Insider trading Did insiders engage in profitable insider trading in the unregulated 18th century London market? Did counterparties underperform as a result of insider trading? And what caused this loss?

Methodology It is key to first find out if trades by members of the BoE court of directors had predictive power over various periods of time.

Predictive power

To find out whether or not the insiders and their copycats could make money, a good first step is to take a look at the predictive power of the trades; a buy must be followed by a price increase, while a sell should be followed by a decrease in stock price. For finding predictive power, I follow the method used by Temin & Voth (2004). Returns from day t over the next τ days are calculated as ln( ). For the event study, I use a trading horizon of 5, 10, and 20 trading days. The drawback of using trading days is that this omits shutting periods. A typical shutting took three to four weeks, and for stocks that were traded just before such a shutting, this seriously increases the actual days in the trading day period during which events can happen that affect the stock price.

The following OLS regressions for buy and sell trades are estimated under Newey-West consistent standard errors with lags equal to horizon:

ln( ) = C + β1Dbuy,t + β2Dshutting,t + εt+ β3ln(

ln( ) = C + β1Dsell,t + β2Dshutting,t + εt )

The first variables Dbuy,t and Dsell,t are dummies that equal one on days of net insider buying and selling, respectively, measured in number of trades and zero otherwise. The second variable Dshutting,t is the shutting dummy that equals unity when a shutting occurs within the return horizon, and zero otherwise. I chose to base buy and sell dummies on the net number of trades to avoid days with having both dummies set to one, and to filter trades that took place between insiders. The shutting dummy controls for the increased number of regular days within a trading day period. This regression only uses insiders during their directorship only. Newey-West standard errors control for heteroscedasticity and for the autocorrelation resulting from rolling returns. The results of the two regression equations above can be found in panels A and B of Table 4.

20

To check for predictive power of clustered trades I constructed a buy dummy for each day on which a trade was preceded by at least 2 net trades in the same direction in the prior 5 trading days. I use the following OLS regression with Newey-West consistent standard errors:

ln( ) = C + β1Dclusterbuy,t + β2Dclustersell,t + β3Dshuttingτ,t + εt

Where Dclusterbuy,t equals one for net insider buy days with at least 2 more buy than sell trades in the preceding 5 trading days, and Dclustersell,t equals one for net insider sell days with at least 2 more sell than buy trades in the preceding 5 trading days. Dshuttingτ,t equals one for an upcoming shutting within the next τ days, and zero otherwise. For this equation, regressions (1), (2), and (3) will be constructed using a return horizon of 5, 10, and 20 days (τ) respectively. This regression uses insiders during their directorship only. The results of these three regression equations can be found in panel C of Table 4.

Using the same horizons (τ), I estimate the regressions (2), (4), and (6) to estimate the predictive power insider volume has. For the insider volume regressions, the years 1723-1725 will be omitted due to a big stock issue in 1723 and the immense concentrated selling related to this issue. For trading volume I will use gross volumes by full sample insiders, instead of net volumes by current insiders in order to show the predictive power of insider volume. Its use and results should also show that the previous short-run regressions are not designed to prove significance. The regression again under Newey-West is as follows:

ln( ) = C + β1Buyvolumet + β2Sellvolumet + β3Dshuttingτ,t + εt

Where Buyvolumet is the insider buy volume in thousands of pounds nominal stock and Sellvolumet is the insider sell volume in thousands. The Dshuttingτ,t equals one for an upcoming shutting within the next τ days, and zero otherwise. This regression uses insiders during their directorship only. The results of the above regression can be found in panel C of Table 4.

In addition I will look into whether returns influence trade decisions and estimate returns before buy and sell trades by insiders and counterparties. I will do this by the following OLS regression:

ln( ) = C + β1Dbuy,t + β2Dsell,t + εt+

Where ln( ) is the natural logarithm of the return over the preceding (τ) days till day t, Dbuy,t is a

dummy for net buy days, and Dsell,t is a dummy for net sell days. This regression uses insiders during their directorship only. The result of this regression is to be found in panel B of Table 5.

21

Lastly, I want to test if trades by insiders or counterparties break a trend, using the following regression:

ln( - ) = C + β1Dbuy,t + β2Dsell,t + εt+

Where ln( - ) is the next 10 day return minus the previous 10 day return, where τ is 10,

and Dbuy,t is a dummy for net buy days and Dsell,t is a dummy for net sell days. This regression uses insiders’ trades during their directorship only. The result of this regression is to be found in panel C of Table 5.

Long-term performance measures

Apart from the short-term predictive power, a look at longer run and actual profits of the 50 insiders will tell us whether being an insider is beneficial over a prolonged period of time. I will use a t-test to test for significant positive profits over and above what one would expect using a buy-and-hold strategy.

In addition to a check for long run profits of insiders, I will construct a sample using all counterparties for whom it is possible to calculate profits5, and measure their returns using the same method as with the insiders. In addition to the t-test, I will test whether extra profits were earned using the following OLS regression:

Extra returni = C + β1B&Hi + β2B_tradesi + β3S_tradesi + β4Dfam,i + β5ITradesi + β6Fraction_insidersi + εi

Extra returni is profit minus B&H profit for an individual counterparty. For the independent variables:

B&Hi is buy-and-hold profit; B_tradesi is number of buys; S_trades is number of sells; Dfam,i is a family dummy that is equal to unity for family names corresponding to the name of an insider; ITradesi is the number of trades with an insider by counterparty; Fraction_insidersi is fraction of insider trades; also an error term and constant are added. The buy-and-hold profit variable is meant to correct for the greater extra profit variation of larger portfolios. Average balance does not work because Extra returni can be negative, while average balance cannot. For each counterparty, 1 will be subtracted from his/her number of insider trades. This only affects the constant and provides a clearer view on the marginal effects of insider trading within the sample of counterparties. The results of this regression are displayed in Table 6.

5 This is essentially random; the selection is based on alphabetical position of first names which is only a slight bias. 22

Analysis of results Predictive power

The picture that appears from panel A&B in Table 4 is that insiders’ buys can be used to predict the stock performance over the next 5-20 days. The coefficient for 5 days seems small, but is actually quite large when measured against the constant. An insider buy dummy increases return by 8-fold (0.0088 versus 0.0011). This 8-fold return takes the annualized return from 6% to 60%. For the 10-day horizon, buys yield 0.92% more than dates on which insiders did not buy. However larger than the 5-day return, the relative difference to the standard return is lower for 10-day returns, with buy days achieving annualized returns of 37% while the stock otherwise would return 7.5% annually.

Table 4 - Returns following insiders' trades 1715-1725 5 day return 10 day return 20 day return Panel A: C Buy C Buy C Buy Performance after Return 0.0011 0.0088 0.0029 0.0121 0.0076 0.0255 net buys Difference 0.0077** 0.0092* 0.0179 (2.34) (1.95) (1.43)

Panel B: C Sell C Sell C Sell Performance after Return 0.0022 -0.0055 0.0042 -0.0047 0.0096 -0.0018 net sells Difference -0.0077* -0.0089 -0.0114 (-1.69) (-1.32) (-1.41)

Panel C: 1 2 3 4 5 6 Volume and clustered trade - Bdum 0.0159* 0.0294** 0.0669* regressions (1.81) (2.30) (1.73) Sdum -0.0212 -0.0227 -0.0293 (-1.32) (-1.05) (-1.24) Bvol 0.0016 0.0034** 0.0055* (1.31) (2.25) (1.94) Svol -0.0007 -0.0018 -0.0038 (-0.61) (-0.99) (-1.35) Notes: t-values in parentheses. *, **, *** Denote significance at the 10%, 5%, and 1% level respectively. This table is based on formulas as presented in the ‘predictive power’ section of the methodology. Bdum is equal to unity if the net number of insider buyers on that trading day was at least one and if the total net number of buyers in the past 5 trading days was more than one; Sdum is equal to unity if the net number of insider sellers on that trading day was at least one and if the total net number of sellers in the past 5 trading days was more than one. Bvol is the number of shares bought in thousands of pounds nominal value; Svol is the number of shares sold bought in thousands of nominal pound sterling value. C denotes the constant of a regression.

It is somewhat odd that coefficients increase with return horizon for buys, while statistical significance falls.

23

The regressions using clustered trades in panel C of Table 4 generally tell the same story as the regular buy and sell days in panel A&B. The largest difference lies in the larger coefficients. Insider clustered sales are not statistically different from 0, but are significantly lower than buys, which is essentially what we look for as well. The buys on the other hand, are all significant at the 10% level and large with a return of 6.7% for 20-day horizons. This would amount to annual returns of 475%, which would not be of very much use, as there are on average only 5 such opportunities (cluster buy dummies) a year. But if one supposes these 5 opportunities are always at least 20 trading days apart, one can make 38% over the course of a year by copying clustered buy trades.

It is no surprise that trading volume has an effect, as it proxies buy and sell dummies. For a 10-day horizon, a day with £ 2700 insider buy volume has the same effect as one net buy dummy day. This is not as poor a result it may seem; the dummies are constructed as net dummies, while volume includes all gross volume from full-sample insiders. It is also encouraging that the coefficients increase with horizon.

All that said, it is clear that even though statistical significance is not very high, insider trades have predictive power. Purchases and sales deliver contrasting subsequent returns and their differences increase with horizon.

Momentum check

However important, interesting, or significant these results may be, they might be solely caused by the bubble, which was an event that not only created excessively useful insider information but might also have given opportunity for momentum trading. This is interesting, because the BoE stock is positively auto-correlated to the point that 17% of future 10 day return can be explained by the past 10 day return. To check whether insiders engaged in momentum trading, I perform a regression of the return over the past 10 trading days on buy and sell dummies. This exercise yields a statistically significant negative coefficient for buys at the 5% level and nothing significant (at the 10% level) for sales. This leaves us the conclusion that insiders did nothing like short-term momentum trading and liked to buy after the stock went down, which is consistent with earlier findings (Friederich, Gregory, Matako, & Tonks, 2002). Adding a past return variable would slightly increase size and significance of results in Table 4.

Additional checks excluding bubble

Yet, the concern remains that the results could be explained by the bubble, as this was clearly a time of high volatility and the bubble covers many of the trades in the sample. When removing the data of 1720

24 from the dataset and running the same regressions, all significance in the results vanishes. What is even more puzzling is that some signs of buy/sell coefficients change. Unsurprisingly, the predictive power during 1720 G is very high; a version of Table 4 that only uses the bubble year is available in Appendix E. These results are quite spectacular, with clustered buys yielding over 14% over a 20 day horizon.

There are ample possible explanations available to explain the sole existence of predictive power in 1720. The most important explanation is that the BoE stock is very stable over time; its volatility is low and corporate announcements were limited to biannual meetings on which dividends were declared. This leaves little room for insiders who were not on the dividend committee to use information to their advantage. Also, the nature of BoE stock makes it so that its value is mostly dependent on how the British treasury interest rates and creditworthiness are doing. As these are macro-economic factors, this information was not exclusive to BoE directors. The bubble, on the other hand, highly increased stock volatility and the insiders had information to trade on.

The fact that all predictive power of insiders disappears when excluding the bubble means that all predictive power is explained by insiders’ timing skill during the bubble. However, the insiders were net buyers during the bubble period, as is evident from shares bought during 1720 (Appendix C), and the pattern in Chart A3 of Appendix D. It seems to be inconsistent that they had timing ability during the bubble, but were not so intelligent investors, as they did not sell their shares in the bubble period. Surprisingly, the findings are consistent with earlier findings by Frehen et al. (2012), who found that Stad Rotterdam directors had some timing ability during the bubble, but were net buyers for that time period.

Table 5 on page 26 contains information on the timing of insiders and counterparties outside 1720. Panel A shows the lack of predictive power of insider trades. Panel B is somewhat more interesting and shows price changes prior to trades by both insiders and counterparties. For buy trades, the two sides show almost perfect contrasting behaviour: insiders buy after a dip, counterparties buy after an upswing. The insiders’ counterparties seem to really like trading on momentum, as is obvious from their trading pattern of buying after price rises and selling after declines.

Interestingly, buy trades by both insiders and counterparties break a trend, as shown in panel C of Table 5. For insiders this price trend was downward before buys, but flat afterwards (thus resulting in a net positive sign), for counterparty buys this trend was upward but stabilized after purchase date. Neither side had predictive power, but both had a different response to certain price movements.

25

Table 5 – Predictive power 1715-1715, excluding 1720 G 5 day return 10 day return Panel A: Buy Sell Buy Sell Returns following Insi ders -0.0002 0.0005 0.0002 0.0007 trades (-0.20) (0.53) (0.11) (0.57)

Panel B: Buy Sell Buy Sell Returns preceding Insiders -0.0022** -0.0005 -0.0048*** -0.0004 trades (-2.23) (-0.48) (-3.42) (-0.26) Buy Sell Buy Sell Counterparties 0.0019*** -0.0008 0.0035*** -0.0019* (2.74) (-1.08) (3.59) (-1.90)

Panel C: Buy Sell Change in Insiders 0.0051*** 0.0012 returns (2.60) (0.58) Buy Sell Counterparties -0.0031** 0.0027* (-2.27) (1.86) Note s: t-statistics are in parentheses. *, **, *** Denote significance at the 10%, 5%, and 1% level respectively. Displayed coefficients are to be added to the constant. Change in returns is the future 10-day return minus the past 10-day return. Panel A shows returns following trades following the same methodology as panels A&B of Table 3. Panel B shows returns preceding trades as its formula in the methodology shows. Panel C displays the change in returns using a dependent variable that subtracts future 10-day return from past 10-day return as shown in the methodology section.

As found by Friedrich et al. (2002), buying against the trend is common for insiders in absence of specific events to trade on (being in a regulated market in that case). Logically, the strategy of the insiders makes a lot of sense for trades who know the fundamental value of the business. They paid less for their stock over time. The contrarian behaviour is less strong for retired directors and directors who were no director yet.

It is unlikely that insider trading induced counterparties to trade by pushing the price away from the fundamental value (as perceived by counterparties) as the stock was tradable at an even more attractive price 10 days before their trade. In fact, the countertrades were momentum traders who had been fortunate that insiders pushed prices in the right direction for them. It makes no sense to buy a stock because it has gone up. That is especially true in the case of the counterparties, who did not see an appreciation of their stock after their purchase.

Of course, the situation was much different in 1720 G, when insiders did have somewhat reliable information on timing of important events and directly caused immediate losses to their counterparties.

26

The results in Table 5 also support the view that insiders’ counterparties did not find out about insider trading. The lack of price movements after insider trades is further support for the notion that insiders could operate without their trading being detected. If the market knew about insider trading, the stock price would have responded to this knowledge.

Although it is not the case that governance prevented insider trading, it did something for the market. Insiders could have spread false rumours about the stock, or otherwise deliberately manipulate the stock price away from its true value to cash-in on short term gains. From the meagre short-run returns outside 1720, it appears that no significant manipulation had taken place.

Long-term performance

The results of the predictive power tests are somewhat unsettling, there was so little information available to the public, and so much at the disposal of the directors that there must have been profits to be made by insiders. Due to the lack of corporate announcements and stock volatility, short-run returns might not be the best way to discover returns on inside information.

For this purpose I calculated the exact profits of each insider and compared this to the BoE stock return over his individual holding period. The full table with detailed information can be found in Appendix F. On these results I ran tests to check whether the returns are statistically significant. A paired t-test checking whether profits are larger than buy-and-hold profits yields a t-statistic of 3.12, meaning that the average £ 2360 profit from trading activity (market timing) is positive at the 0.5% level. It is important to note that less than 10% of insider trading profits can be explained by 1720. Testing profit per insider on directorship dummies for each year (equalling one for being a director, and zero otherwise), in an OLS regression yields a negative, non-significant result for being an insider in 1720.

The counterparties did not fare that well. Their average profit from trading activity was -£ 334, but with a t-stat of -0.6245 not statistically significant from 0. However, the £ 2700 gap between insiders and their counterparties is significant and large. The full results of the OLS regressions on the long-run counterparty sample can be found in Table 6 on page 28. The most important part of this is that each additional trade with an insider decreases profit from trading by over £ 1000, which is half the average insider trade size. That is five times the average of the additional £ 200 take on a simple stock purchase6.

6 Buys produce better returns than sells because the stock traded at a bubble price at the end of the sample period. 27

Also, an additional 10% fraction of insider trades on total trades decreases returns by £ 224½. Take note of the high adjusted r-squared; this is a really good model to predict long-run trading profits.

Table 6 - Long-run trading profits counterparty traders OLS # Insider Fraction Adjusted Constant B&H profit # Buys # Sells D family model trades insider trades R-squared (1) 1191.9 -0.7428 201.3 -167.4 2379.4 -1063.3 - 0.8363 (5.10) (-27.88) (6.80) (-6.36) (1.95) (-4.18) -

(2) 1623.5 -0.7415 129.7 -131.5 2322.3 - -2245.3 0.8283 (4.88) (-27.16) (5.09) (-5.15) (1.85) - (-2.36)

(3) 1690.4 -0.7406 195.6 -165.7 2269.6 -1034.5 -2051.1 0.8389 (5.24) (-28.00) (6.64) (-6.35) (1.87) (-4.10) (-2.22) Notes: t-statistics are in parentheses. All coefficients are significant at the 1% level, except for fraction insider trades which is significant at the 5% level and D family which is significant at the 10% level. The dependent variable is extra return for a (counterparty) trader over his entire time in the sample and it measures his absolute profit over and above the B&H profit. The B&H profit is the profit counterparties would have made, had they purchased their average stock holding at first date and sold at the last date. The first date is the latest out of first purchase date and summer 17157; last date is the first of last sell-out date and Sept. 29, 1720. # Buys is the number of buy-trades for the trader. # Sells is the number of sell-trades for the trader. Dummy family indicates whether the trader is related to an insider. # Insider trades is the number of times the trader traded with an insider. Fraction insider trades is a traders’ fraction (0 to 1) of trades conducted with insiders.

There could be a bias here: traders who traded more often with insiders were poorer traders and also lost money on their trades with non-insiders. But there is still no excuse for losing money on both each additional insider trade and the fraction of insider trades. The fraction of trades with insiders is a better proxy for bad trading habits than # Insider trades, as the number of insider trades is controlled for by #Buys and #Sells. The fact that number of trades with insiders has a bigger influence on results than the mere fraction makes it credible that insiders cause direct harm to their counterparties with each informed trade.

Being related to an insider8 increases extra return by over £ 2200, which fascinatingly is about equal to the average extra profit of insiders. This £ 2200 is an incredible sum: 11 times the annual salary of the Bank’s head accountant, but it is especially huge when compared to the counterparties’ average extra

7 The first record date of accounts varies in the studied stock ledgers, but all are between April and August, 1715. 8 The family dummy equals unity for counterparty traders having the same surname as any insider. 28 return of 09. However high the economic significance may be, the statistical significance has not reached the 5% level because the sample contained only 12 people who were possibly related to a director.

Concluding remarks

Insiders did insiders engage in profitable insider trading in the unregulated 18th century London market. The evidence that accounts of insiders were more profitable than those of their counterparties is overwhelming. Especially the timing of trades, how clustered trades earned higher returns and the trading during the bubble, fuels the finding of insider trading. Insiders made huge long-term profits, their counterparties earned slightly less than they would have, had they employed a buy-and-hold strategy.

The predictive power and short-term success of insider trades is disputable. However, for 1720 it is crystal clear that the insiders gained at the expense of their counterparties. Third party traders do not appear to be harmed by insider trading during the bubble, but were harmed in other periods. Insiders were long-term investors; they made profits slow, not quick. This was due to their contrarian strategy, which did not work immediately, but clearly did work over time. Their counterparties, on the other hand, were likely to buy after the stock went up and to sell after it went down.

Counterparties did underperform partly as a result of insider trading. The empirical analysis confirms that counterparties, as a group, consistently and significantly underperformed when insiders were at the other end of the trade, even in periods where insiders did not book short-term gains on their trades. These losses were caused by the gains achieved by insiders who had superior knowledge on the stock’s long-term value, knowledge that counterparties did not have. But what is important to remember, is that it was not a price push by insiders that induced counterparties to trade with them.

9 It is 0 due to lack of statistical significance of the £ -334 negative average extra return. 29

7. Conclusion

From the overwhelming evidence, we can conclude that being an insider in an unregulated market pays- off, regardless of specific knowledge about corporate events or announcements. Insider trading is profitable to insiders, especially over a longer horizon, mostly because insiders know the fundamental value of the company and often exploit this knowledge using a contrarian trading strategy.

The counterparties of insiders lose money to insiders, directly in the event of specific inside information on events, and indirectly by their habit of momentum trading, which is opposite to the strategy the insiders employ. These trading habits match insiders with counterparties by design, so it cannot be claimed that insiders induce counterparties to trade in such a situation.

As it was the case that insiders had good cover and leaking of information was kept to a minimum, social enforcement of governance rules was no pressing issue to insiders. They must have felt mostly unrestricted in trading on insider information, which showed in their trading behaviour around the South Sea bubble. Another consequence of outsiders not learning about insider trading was that insiders could cautiously spread trades across time to avoid causing price movements, without being bothered by the market catching up. For the counterparties, the absence of learning about insider trading was not so amusing; they fell right into the trap of informed traders and could not recoup their losses as they were not nearly on a level playing field with the insiders regarding information.

These findings support the argument that insider trading is undesirable as it is unfair and hurts other traders. The effects are especially unfair to outsiders in case of insider trading on major events that cause volatility in the stock price. Thus, current regulation that allows insiders to trade, but not on material information, and forces them to make their trades public, looks very sensible for the well- functioning of markets.

Suggestions for future research

A worthwhile extension to present research would be a more extensive study on who the victims of insider trading really are. Are they intelligent investors who are scammed by vicious insiders, or are they a group of unskilled investors whose capital is better off in more capable hands? It is very important for the well-functioning of the capital market to know what sort of investors is hurt by insider activity.

30

Acknowledgements I would like to thank Dr. Rik Frehen for giving me the opportunity to work on this topic, his guidance and supervision, and for taking the thousands of pictures of stock ledgers in the BoE archives necessary for this study. I would also like to thank Harrie, Esmee, and Sjors, with whom I collected the data. Without the help of Dr. Frehen and my three fellow students, this thesis would have been a futile endeavour.

Bibliography Bainbridge, S. M. (2000). Insider Trading: An Overview. Encyclopedia of Law and Economics.

Barclay, M. J., & Warner, J. B. (1993). Stealh trading and volatility Which trades move prices? JFE, 281- 305.

Bell, S. (2012). ‘A masterpiece of knavery’? The activities of the Sword Blade Company. Business History, 623-638.

Bhattacharya, U., & Daouk, H. (2002). The World Price of Insider Trading. The Journal of Finance, 75-108.

Carlos, M. A., & Neal, L. (2006). The micro-foundations of the early London capital market: Bank of England shareholders during and after the South Sea bubble 1720-1725. Economic History Review, 498-538.

Cope, S. R. (1978). The Stock Exchange Revisited: A New Look at the Market in Securities in London in the Eighteenth Century. Economica, 1-21.

Dooley, M. P. (1980). Enforcement of Insider Trading Restrictions. Virginia Law Review, 1-84.

Easley, D., & O'Hara, M. (1987). Price, Trade Size, and Information in Securities Markets. Journal of Financial Economics, 69-90.

Finnerty, J. E. (1976). Insiders and market efficiency. The Journal of Finance, 1141-1148.

Frehen, R., Goetzmann, W., & Rouwenhorst, G. (2012). New evidence on the first financial bubble. Journal of Financial Economics, 585-607.

Friederich, S., Gregory, A., Matako, J., & Tonks, I. (2002). Short-run Returns around the Trades of Corporate Insiders on the London Stock Exchange. European Financial Management, 7-30.

Gilson, R. J., & Kraakman, R. H. (1984). The Mechanisms of Market Efficiency. Virginia Law Review, 549- 644 (572-579).

Hayek, F. (1945). The Use of Knowledge in Society. American Economic Review, 519-30.

31

Koudijs, P. (2013). Those who know most: insider trading in 18th c. Amsterdam. National Bureau of Economic Research.

Leland, H. E. (1992). Insider Trading: Should It Be Prohibited? Journal of Political Economy, 859-887.

Meulbroek, L. (1992). An Empirical Analysis of Illegal Insider Trading. The Journal of Finance, 1661-1699.

Morse, D., Elliot, J., & Richardson, G. (1984). The Association between Insider Trading and Information Announcements. The RAND Journal of Economics, 521-536.

Petram, L. (2011). The World's Frist Stock Exchange. UvA-DALE.

Scott, W. R. (1951). The Constitution and Finance of English, Scottish and Irish Joint-Stock Companies to 1720. New York: Peter Smith.

Seyhun. (1992). Why Does Aggregate Insider Trading Predict Future Stock Returns. The Quarterly Journal of Economics, 1303-1331.

Seyhun, H. N. (1986). Insiders’ Profits, Costs of Trading, and Market Efficiency. Journal of Financial Economics, 189-212.

Temin, P., & Voth, H.-J. (2004). Riding the South Sea Bubble. The American Economic Review, 94(5), 1654-1668.

Von Phillippovich, E. (1911). National Monetary Commission. History of the Bank of England. Washington: Government Printing Office.

Miscellaneous sources

Cover image: Dividend Day at the Bank of England by George E. Hicks. Photo credit: Bank of England

Information used for section 3. Historical background: The day Big Bang blasted the old boys into oblivion. (2006, October 29). Retrieved June 15, 2014, from The Independent: http://www.independent.co.uk/news/business/analysis-and-features/the-day-big- bang-blasted-the-old-boys-into-oblivion-422005.html

Bank of England. (n.d.). Timeline. Retrieved June 14, 2014, from Bank of England: http://www.bankofengland.co.uk/about/Pages/history/timeline.aspx

32

Appendices

Appendix A Bank of England daily stock price chart from January 1715 G to Dec 1725 G

Chart A1 - Storck price 1715-1725 G 280

260

240

220

200

180

160

140

120

100

33

Appendix B Table A1 - Dividend schedule Bank of England Shutting period Dividend Dividend meeting date September 1715 £ 4 22-9-1715 March 1715 £ 4 22-3-1715 September 1716 £ 4 20-9-1716 March 1716 £ 4 21-3-1716 September 1717 £ 4 19-9-1717 March 1717 £ 4 15-3-1717 September 1718 £ 4 18-9-1718 March 1718 £ 4 19-3-1718 September 1719 £ 3½ 17-9-1719 March 1719 £ 3½ 24-3-1719 September 1720 £ 4 29-9-1720 March 1720 £ 3 23-3-1720 September 1721 £ 3 14-9-1721 March 1721 £ 3 15-3-1721 September 1722 £ 3 20-9-1722 March 1722 £ 3 21-3-1722 September 1723 £ 3 19-9-1723 March 1723 £ 3 19-3-1723 September 1724 £ 3 17-9-1724 March 1724 £ 3 18-3-1724 September 1725 £ 3 16-9-1725

Note: The director meeting upon which the dividend was announced took place on the first or second Thursday after the start of the shutting. Up to and including 1720, the ex-dividend date coincided with the meeting date and on the Saturday after the meeting thereafter.

34

Appendix C Table A2 - Trading flows excluding dividends Insiders: Trading cash flow Shares bought (sold) Average price 1715 -1719 £ 108,622 -£ 68,705 £ 1.58 1720 -£ 48,036 £ 40,890 £ 1.17 1721-1725 -£ 49,167 £ 59,267 £ 0.83 Full sample £ 11,419 £ 31,452 N/A Co unterparties: 1715 -1719 - £ 376,136 £ 270,621 £ 1.39 1720 -£ 188,152 £ 71,005 £ 2.65 Full sample -£ 564,287 £ 341,626 £ 1.65 60 month sample: Counterparties 1715 -1720 - £ 549,096 £ 330,123 £ 1.66 Insiders 1715-1720 £ 25,241 -£ 4,335 £ 5.82

Notes: All numbers are cumulative for their respective time period. Average price is the price per £1 nominal stock calculated as cumulative cash flows divided by net acquisition or disposition of shares. The full-sample average price of trades of insiders could not be calculated because they essentially received ‘free’ shares valued at over £43.325 and some extra cash. 60 month sample covers the 10 years starting Sept. 30, 1715, and ending Sept. 29 1720.

35

Appendix D Chart A2 - Monthly stock trading 1715-1722 100000 250

80000 220

60000 190

40000 160

20000 130

0 100 1715 1716 1717 1718 1719 1720 1721 1722 -20000 70

-40000 40

-60000 Insider volume (lhs) counterparty volume (lhs) Price (rhs) 10

Notes: The year marks are placed in September of each year. September is the 6th month of the year for the Julian calendar. The year 1723 and thereafter were omitted because the year 1723 contains a stock subscription which causes quite extreme trading volumes (both buy and sell) from that point onwards that would divert attention from the rest of the graph.

36

Chart A3 - Insiders' holdings and prices 380000 280

360000 260

240 340000

220 320000

200 300000 180

280000 160

260000 140

240000 120

220000 100

Stock holdings (lhs) price (rhs)

Notes: The chart is adjusted for new insiders entering the sample in September 1720; Chart covers period from September 30, 1715 until September 1721. Year labels are based on the Gregorian calendar.

37

Appendix E Table A3 - Returns following insiders' trades 1720 G 5 day return 10 day return 20 day return Panel A: C Buy C Buy C Buy Performance Return 0,0008 0,0345 0,0045 0,0497 0,0176 0,1080 after net buys Difference 0.0337*** 0.0452*** 0.0904*** (2.71) (2.90) (2.88)

Panel B: C Sell C Sell C Sell Performance Return 0,0134 -0,0267 0,0197 -0,0263 0,0197 -0,0512 after net sells Difference -0.0401**** -0.0460** -0.0709*** (-2.83) (-2.26) (-2.87)

Panel C: 1 2 3 4 5 6 Volume and clustered trade Bdum 0.0291* 0.0644*** 0.1417*** -regressions (1.89) (3.23) (3.54) Sdum -0,0698 -0,0775 -0.08002* (-2.24) (-1.38) (-1.88) Bvol 0.0056* 0.0100*** 0.0141*** (1.66) (2.80) (2.94) Svol -0,0023 -0,0053 -0.0102* (-0.73) (-1.14) (-1.72) Note: t-values in parentheses. *, **, *** denote significance at the 10%, 5%, and 1% level respectively.

38

Appendix F

Table A4 - Profitability overview of insiders 1715-1725 Annualized Annualized Additional profit Holding Profit Name return B&H return B&H profit from trading Average Balance period # trades £ 72.201 Peter Delme 5,9% 5,5% £ 65.244 £ 6.957 £ 89.353 10,25 232 £ 33.927 Gilbert Heathcote 12,2% 5,5% £ 10.963 £ 22.964 £ 15.088 10,21 11 £ 27.400 Justus Beck 10,6% 5,6% £ 12.714 £ 14.686 £ 29.097 6,61 240 £ 21.863 William Scawen 7,7% 6,3% £ 16.974 £ 4.889 £ 26.700 8,05 36 £ 15.740 Richard Cary 11,9% 5,5% £ 5.331 £ 10.409 £ 7.286 10,24 15 £ 15.591 William Humfreys 12,7% 5,5% £ 4.708 £ 10.882 £ 6.443 10,29 14 £ 15.401 (John) Francis Fauquier 8,6% 5,5% £ 8.397 £ 7.004 £ 11.607 10,23 50 £ 14.723 10,1% 5,5% £ 6.415 £ 8.308 £ 8.758 10,28 35 £ 13.839 Richard Chiswell 12,1% 5,6% £ 4.675 £ 9.164 £ 6.171 10,29 20 £ 10.828 Randolph Knipe 11,8% 5,6% £ 3.843 £ 6.985 £ 5.005 10,30 155 £ 10.051 23,2% 4,3% £ 1.288 £ 8.763 £ 5.600 4,93 10 £ 9.351 Christopher Lethieullier 5,6% 5,5% £ 9.217 £ 134 £ 12.429 10,30 6 £ 8.642 Gerald Conyers 6,3% 5,4% £ 7.178 £ 1.464 £ 9.781 10,39 12 £ 7.900 William Joliffe 5,6% 5,3% £ 7.228 £ 672 £ 10.271 10,38 12 £ 7.828 Joseph Eyles 14,5% 5,9% £ 2.060 £ 5.768 £ 2.577 10,29 54 £ 7.038 Nathaniel Gould 6,3% 5,9% £ 6.436 £ 602 £ 8.122 10,19 2 £ 6.310 John Ward 6,4% 5,4% £ 5.077 £ 1.233 £ 7.093 10,27 6 £ 5.860 Robert Bristow 14,5% 4,6% £ 1.292 £ 4.568 £ 2.981 8,02 10 £ 5.549 John Lordell 5,2% 5,6% £ 6.090 -£ 541 £ 8.081 10,29 16 £ 5.451 John Gould 4,8% 5,6% £ 6.604 -£ 1.154 £ 8.956 10,22 20 £ 5.121 William Dawsonne 6,7% 6,1% £ 4.515 £ 606 £ 5.699 9,85 9 £ 5.023 John Shipman 13,5% 4,8% £ 1.320 £ 3.703 £ 3.182 7,47 6 £ 4.990 25,1% 3,2% £ 421 £ 4.569 £ 2.629 4,75 19 £ 4.906 William Thompson 5,6% 5,5% £ 4.851 £ 55 £ 6.548 10,28 18 £ 4.695 Philip Jackson 6,3% 5,6% £ 4.005 £ 690 £ 5.345 10,29 29 £ 4.485 Thomas Abney 7,8% 5,9% £ 3.126 £ 1.359 £ 5.065 8,44 13

39

Table A4 (continued) Annualized Annualized Additional profit Holding Profit Name return B&H return B&H profit from trading Average Balance period # trades £ 4.424 5,4% 5,7% £ 4.725 -£ 301 £ 6.152 10,30 15 £ 4.061 John Edmonds 4,9% 5,2% £ 4.430 -£ 369 £ 6.489 10,25 9 £ 3.555 Robert Atwood 6,2% 5,6% £ 3.062 £ 493 £ 4.125 10,25 20 £ 3.457 George Thorold 4,9% 4,9% £ 3.428 £ 29 £ 6.924 8,41 14 £ 3.454 Theodore Janssen Bar 6,3% 5,8% £ 3.123 £ 330 £ 3.952 10,29 30 £ 3.121 John Cope 6,1% 5,6% £ 2.781 £ 340 £ 3.721 10,24 26 £ 2.879 John Hanger 5,2% 5,6% £ 3.192 -£ 313 £ 4.279 10,21 13 £ 2.838 John Ward 19,6% 7,4% £ 827 £ 2.010 £ 1.833 5,22 12 £ 2.540 Josiah Diston 5,9% 4,8% £ 2.018 £ 522 £ 4.563 7,75 18 £ 2.306 Matthew Howard 11,9% 12,1% £ 2.344 -£ 37 £ 7.440 2,40 2 £ 2.074 Moses Raper 5,0% 5,6% £ 2.389 -£ 315 £ 3.158 10,25 17 £ 1.874 John Heathcote 5,9% -3,3% -£ 683 £ 2.557 £ 2.315 10,30 8 £ 1.695 5,2% 3,0% £ 945 £ 750 £ 6.300 4,73 9 £ 1.379 Francis Forbes 10,7% 12,4% £ 1.609 -£ 230 £ 6.500 1,88 3 £ 720 Nathaniel Gould Junior 9,0% 9,0% £ 720 £ - £ 2.000 3,58 1 £ 406 William Fawkener 9,8% 9,8% £ 406 £ - £ 2.500 1,61 1 £ 343 Bryan Benson 1,5% -6,9% -£ 1.338 £ 1.680 £ 4.440 5,00 8 -£ 407 -1,7% -10,4% -£ 2.148 £ 1.740 £ 5.099 5,00 6 -£ 1.079 Henry Herring -1,4% -10,4% -£ 6.459 £ 5.380 £ 15.333 5,00 3 -£ 1.858 Delillers Carbonnel -15,9% -6,9% -£ 964 -£ 894 £ 3.200 5,00 3 -£ 5.213 John Olmius -28,5% -6,9% -£ 1.930 -£ 3.284 £ 6.405 5,00 42 -£ 5.584 Barrington Eaton N/A 5,6% £ 3.803 -£ 9.387 £ 5.090 10,23 10 -£ 7.119 Humphrey Morice -21,9% 5,4% £ 5.560 -£ 12.679 £ 7.729 10,29 40 -£ 14.446 Richard Du Cane -16,5% -10,2% -£ 9.761 -£ 4.685 £ 28.500 3,91 45 £ 350.129 Total £ 232.052 £ 118.077 £ 457.964 1405

40

Table A5 - Profitability of shared accounts 1715-1725 profit from Average Holding # Profit Shared accounts B&H profit trading Balance period trades £ 387,00 Nathaniel Gould & Sir Alexander Cairnes Bar £ 421,52 -£ 34,52 £ 564 10,30 4 £ 365,00 John Hartopp & Samuel Holden £ 323,33 £ 41,67 £ 1.333 1,66 1 £ 348,00 Nathaniel Gould & Thomas Prime £ 348,00 £ - £ 800 8,13 1 -£ 427,50 Nathaniel Gould & John Gould junior -£ 427,50 £ - £ 500 0,67 2 -£ 604,88 Nathaniel Gould & John Monteage -£ 342,88 -£ 262,00 £ 10.550 0,49 12 £ 272,00 Humprhey Morice & Samuel Edwards £ 272,00 £ - £ 800 3,31 1 £ 240,00 Sir Thomas Willys & William Willys & Samuel Holden £ 240,00 £ - £ 3.000 0,88 2 -£ 1.619,75 William Scawen & Thomas Gibson -£ 1.619,75 £ - £ 6.200 5,00 0

Notes for tables A4 & A5: The tables are calculated as follows: Profit is the actual total mark over the holding period. B&H profit is the sum of dividends received plus the average balance multiplied by price at last holding date minus price at start date. For example: average balance is 100 (one share), dividend is 5, price at first recorded holding date is 140, price at last date is 150 makes 1.00*(150-140) +5 = 15. Average balance is the average of stock balances on dividend dates. Profit from trading is profit minus B&H profit. Holding period is the number of years between the opening date of the account and the closing date of the account. Annualized return uses average balance, profits and holding period as inputs. These tables use continental separators.

Bear in mind that these people each were the ‘millionaires’ of their time and that while, the investment profit of, say, Bryan Benson, might seem negligible at £343 over 5 years, a house cleaner had to work 34 years to earn that amount of money.

41