Securities Lending and Short Selling

Purpose – The aim of this study is to offer the current view on the subject, placing emphasis on studies that use Brazilian data and that could motivate new research.   Design/methodology/approach –A bibliographic review of related studies and a secondary database. Findings – The contribution is to introduce this topic for the academic community in Brazil. We also present the main descriptive information for this market and its economic and policy implications. Originality/value – We contribute to the literature by summarizing the main works in the field, focusing on the Brazilian market, which has very detailed data and great potential for further studies.


Introduction
Securities lending, particularly stock lending, attracts the attention of market participants, academics, and regulators due to its different uses, with short selling being one of the most contentious ones.
The literature on restrictions on short selling, in particular Miller (1977), points out that restrictions on short selling could imply that stocks are overpriced. If the factors that make it unviable for investors to carry out short selling operations last for long, stock prices will remain overestimated and, thus, have low future returns until the overpricing is corrected. So, by identifying stocks that are subject to restrictions on short selling, one can identify the stocks with particularly low future returns.
There are many international studies on the effects of short selling. Some recent papers deal specifically with the Brazilian market. Therefore, the aim of this study is to present the functioning of this market and to serve as an orientation for new studies on this subject to be produced by the academic community for the Brazilian market.
The remainder of this paper is structured as follows. Section 2 presents the economic rationale and operational characteristics of stock lending, highlighting the differences between the centralized model of Brazil and the decentralized model. Section 3 describes the main restrictions on short selling adopted by different countries, especially after the 2008 crisis. Section 4 briefly presents the main studies that address the international market of securities lending, while Section 5 presents the main studies that explore stock lending and short selling using Brazilian data. Lastly, Section 6 presents the final considerations.

E c o n o m i c r a t i o n a l e a n d operational characteristics
Short selling is the process by which an investor sells a stock in the market, for example, in an electronic trading venue, without necessarily having bought the stock. The non-existence of this stock in the investor's portfolio (custody) gives rise to two ways of short selling: normal and naked.
In both cases, the economic rationale is the possibility of repurchasing the stock on the equity market at a lower price and, consequently, profiting from this operation, even after including the cost of the loan.
Normal short selling begins with the loan of the asset, followed by its sale on the equity market, which is operationally a relatively simple process in virtually all countries. In the case of naked short selling, the process is different, as the investor sells on the equity market first and then seeks to borrow the stock to meet the settlement cycle in that market. If the investor is not able to borrow the asset in time to honor the sale, it results in a failure to deliver, which is subject to fines and penalties.
The market configuration of securities lending varies from one country to another. For example, in the US, stock lending is a decentralized process; the lender and borrower agree bilaterally through a brokerage firm on the loan details, including the cost of borrowing, term to termination, and, especially, the amount of collateral that the borrower will have to make available to cover the risk incurred.
In the US market, securities lending is treated as an over-the-counter (OTC) operation, carried out within an institution or between two institutions. The operation is recorded in the accounts of the institutions involved, without any record in a centralized system. As these operations are decentralized, the only limits imposed on loan operations are those imposed by the institutions, considering that these same institutions are the ones that will reap the gains from a greater volume of operations.
In Brazil, the structure is centralized; all stock lending transactions are carried out by clearing and settlement entities authorized by the Securities and Exchange Commission of Brazil (CVM). The current regulatory framework is provided by Resolution No. 3539 of February 28, Securities Lending and Short Selling 2008 of the National Monetary Council, CVM Instruction No. 441 of November 10, 2006, and by the rules and regulations adopted by the entity providing the asset clearing and settlement service.
Historically, securities lending in Brazil, of mostly stocks, came from the need to adopt mechanisms to minimize failures in the delivery of stocks during the settlement cycle of the transactions originating from the equity market. On May 13, 1996, the then São Paulo Stock Exchange (BOVESPA) formed a subsidiary called CALISPA S.A. for the settlement of transactions carried out on the BOVESPA, including those resulting from failure to settle. Months later, the Brazilian Settlement and Custody Company (CBLC) was formed and incorporated CALISPA. This event coincided with the launch of the Securities Lending System of the CBLC, known as BTC, which constituted the securities lending system until 2018, when the B3 1 launched the BM&BOVESPA Securities Lending System (BTB).
The B3 is the only Financial Market Infrastructure (FMI) authorized by the CVM to provide centralized registration and settlement of securities lending transactions. In the exercise of its functions aimed at guaranteeing settlements, the B3 acts as the central counterparty (CCP) for securities lending transactions and calculates the risk associated with the transactions and the margin required from borrowers, along with the maximum limits for lending transactions associated with a given stock.
Although registration in a centralized system is required in accordance with Brazilian regulations, the negotiation phase between the borrowers and the lenders and the consequent determination of the loan fee 2 is similar to the American model, where a broker actively participates in the process to find parties for the transaction. The analysis of Figure 1 shows that there is, in fact, a relationship network between brokers and that some institutions interacted actively from January 2012 to December 2014, while others tend to meet the lending needs of their clients without resorting to other brokers, making use of their own client base.

Figure 1. Relationship network
Comment: Each circle represents a brokerage house involved in the stock lending process and the arrows show the direction in which the relationship occurslender and/or borrower. Only those brokers that traded more than 1 billion reais from Jan 2012 to Dec 2014 are displayed. The size of each circle represents the financial value that the brokerage house represented in the securities lending process for the period.
As previously mentioned, the economic rationale for short selling is the possibility for the investor to buy back a stock in the future at a price lower than the sale price and, consequently, make a profit. Studies have investigated investors' informational capacity to anticipate a decline in a stock price, as will be discussed in sections 4 and 5.
However, when the analysis is extended, a trading strategy that actively uses stock lending is the market making of options. For the investor, the lending of shares does not presuppose the expectation of a future downturn but rather a risk management strategy for his/her portfolio.
It should be noted that the stock lending process involves the temporary transfer of ownership to the borrower. Thus, the borrower is entitled to all rights, such as the right to vote, if he/ she has not sold the stock. However, with respect to proceeds, such as dividends and interest on net equity, the borrower is obligated under the terms of the stock lending agreement to reimburse the lender. The same reimbursement process is also applied in the case of corporate events, such as dividend payments, interest on net equity (JCP), mergers, and splits.
Thus, it can be said that the third-most regular use of stock loans is associated with a temporary transfer of rights. In this sense, the study of Mota (2017) shows that stock lending was a strategy that was extensively exploited by institutional investors motivated by regulatory arbitrage (in this case, taxes) until 2014. Table 1 provides an overview of the Brazilian loan market, describing its evolution from 2000 till 2011:  Table 1 shows that the stock lending market has grown in numbers over the last decade-the financial volumes have grown, the number of transactions has reached 1 million, and the number of shares is approaching 300showing that the lending market has been actively used not only to address settlement failures but also for carrying out investment strategies.
According to the aggregated data provided by the B3, the composition of the three main types of lenders in September 2018 is: local investment funds, corresponding to 45.3% of the value traded by lenders; international investors, corresponding to 29.5%; and individual investors, corresponding to 20.5%.
When observed in disaggregate form, the data show high heterogeneity among the rates in the loan market, as described in Table 2.  Table 2 presents the 10 companies with the largest amount of shares lent during the period 08/21/2018-09/13/2018. During that period, the share borrowed the most was CIEL3, of Cielo SA, a company that acts as an acquirer in the means of payment industry.
In addition to the number of shares loaned and their corresponding financial value, Table 2 describes the fees of lenders and borrowers during the period, with the difference between the two fees representing the commission charged by intermediaries, again showing a high degree of heterogeneity among the shares.
To illustrate the effect between stocks over time, Figure 1 presents the average rate received by lenders for stocks that had records for every day during the period and where the total financial value is over 100 million reais.

Figure 2. Average Rates Received by Lenders
Comment: Evolution of the stock lending market for the most recent period disclosed by the B3 (08/21/2018 to 09/13/2018). The chart shows the companies that had transactions for every day of the sample and that had a financial value of over R$ 100 million. The left axis shows the average rate received by lenders for BBDC4, BOVA11, ITUB4, and PETR3 shares. The right axis shows the average rate received by lenders for RADL3 and CIEL3 shares. Source: B3 (http://www.b3.com.br) Figure 2 shows that there are two distinct groups of assets: stocks with loan fees below 1% (BBDC4, ITUB4, and PETR3) and stocks that are considered hard-to-borrow 3 , with rates above 10% (RADL3 and CIEL3).
As observed, the fees paid by borrowers are different from those received by lenders. This difference can be seen in Figure 3 for the same group of stocks. There is variability between the stocks and over time, with the differential being proportional to the fee level. Stocks with low loan rates are easily found and, consequently, there is less space for the commissions charged by the broker, while stocks that are hard-to-borrow end up with a bigger differential because of the service provided by intermediaries for finding a lender willing to lend that asset.

Figure 3. Differential between the fees paid by borrowers and received by lenders
Comment: Evolution of the stock leading market in the most recent period disclosed by the B3 (08/21/2018 to 09/13/2018). For each share, the value represents the spread between the fee paid by borrowers and the fee received by lenders. Source: B3 (http://www.b3.com.br) Theoretically, this differential can be attributed to the search cost of finding a stock that the borrower wants to sell but which is not available internally at the brokerage firm of origin. As shown by Chague, De-Losso, De Genaro, and Giovannettet (2017), the heterogeneity can be best seen when the fees are analyzed at the investor-broker level. In this sense, using a database that identifies only those involved in the transactions (lenders, borrowers, and brokers), finding a stock is easier for a borrower who has a good relationship with brokers who, in turn, have good relationships with frequent lenders of the stock. Based on this, a borrower is said to have low search costs if he/she is well connected to brokers who are well connected to active lenders.

Short selling and its prohibitions
As pointed out by Bris et al. (2007), short selling has been the object of bans and restrictions for almost as long as stock exchanges have existed. Regulators have argued that short selling magnifies declines in asset prices, and therefore should be limited or prohibited during times of high stock market volatility.
There have been several episodes where short selling was prohibited. The stock market crisis of 1929 or the 2000 technology stocks crisis are examples of these events. Although their impacts were significant at the time, these prohibitions were limited to the US stock markets. The real estate market crash, which began in 2008, is one example where regulators across the board restricted or banned short selling.
As the US was the first country to feel the impact of the financial crash, it was also the first to restrict 4 short selling. The first ban was in effect from July 21, 2008 until August 12, 2008 and affected all naked short selling transactions. The second and more restrictive measure adopted by the Securities Exchange Commission (SEC) was the temporary prohibition of any form of short sale of approximately 1,000 shares from September 19 to October 8, 2008, to stabilize the fire-sale pressure on the equity market.

L i t e r a t u r e r e v i e w w i t h international data
Since Miller's seminal paper (1977) The empirical evidence that verifies the effects of short selling on stock prices is not unanimous. Although there is some evidence that short selling restrictions lead to stock overpricing, for example that of Jones and Lamont (2002), Ofek and Richardson (2003), Ofek, Richardson and Whitelaw (2004), and Cohen, Diether, and Malloy (2007), other studies find that short Of course, one of the main challenges when studying this issue is to identify causal relationships when the fundamentals of the company, short selling, and stock prices are almost always determined simultaneously. In this sense, the study of Grullon, Michenau, and Weston (2015) used an experiment, resulting from a regulatory change, which reduces restrictions on short selling in a random sample of US companies to test whether capital market frictions have an effect on stock prices and corporate decisions.
The authors found that allowing short selling caused prices to fall and smaller companies reacted to these lower prices by reducing new issues of shares as well as the level of investment.
These results not only provide evidence that short selling restrictions affect asset prices but also confirm that short selling has a causal impact on financing and investment decisions. Based on these results, the authors concluded that short sellers are informed investors (IH) but, as they generally cannot short sell as much as they wish to, the prices do not reflect all the information present in the market (OH). b) Well-connected short-sellers pay lower loan fees: A market-wide analysis -Chague et al.

2017
The main contribution of this study is that it is the first to examine the relationship between loan rates and search cost at the borrower level. Measuring search cost that is specific to the borrower is empirically challenging, since one must measure the importance of each lender in the market, as well as the relevance and/or intensity of relationships among borrowers, brokers, and lenders.
The authors test two hypotheses: To empirically evaluate the hypotheses, the authors constructed a specific search cost measure for the borrower based on the theoretical model of Duffie, Garleanu, and Pedersen (2002).
To reproduce the loan dynamics of a stock, it is known that in a typical transaction, a possible short seller contacts his/her broker requesting a specific stock for loan. The broker then searches for a potential lender within his/ her own brokerage (own inventory) or turns to another brokerage house. Hence, obtaining a share is easier for a borrower who has a good relationship with brokers who, in turn, have good relationships with the active lenders of the share.
Based on this transactional flow, a borrower is said to have low search costs if he/she is "well connected" to brokers who are "well connected" to active lenders. A borrower is said to be well connected to a broker if he/she is a major client of that broker. Similarly, a broker is well connected to a lender if he/she is responsible for a high share of the lender's business.
Since the data set allowed each market participant to be tracked over time, the authors were able to calculate: (a) how well connected each borrower was to each broker, (b) how well connected each broker was to each lender, and (c) how active each lender was in the loan market of each share. The authors also included non-linear effects by separating borrowers into three groups (high, medium, and low BC) and compared the average lending rate in each group. Firstly, the results show that borrowers from the low BC group pay 14.5% higher loan rates than borrowers from the high BC group.
Secondly, direct measures of loan fee dispersion (standard deviation of loan fee and deals range for the same share) are used to test whether loan rate dispersion is higher among low connection borrowers. Again, the authors found that the standard deviation of the loan fee and loan fee magnitude among borrowers in the low BC group are, respectively, 46% and 135% higher than the rates among borrowers in the high BC group.
Finally, the analysis was refined to study the in-broker variation in the loan rates. The same regressions described above were estimated, but using only those deals closed within a single broker-the largest in terms of transactions. The conclusions are the same as before: it was found that on the same day, for the same share, the broker has different loan rates for different types of borrowers.  (2017) In this study, the author measured the impact of short selling on the future return of stocks through a natural experiment. This is a very interesting approach, because it made it possible to identify an exogenous factor that affected the loan fee, as well as the quantity of short positions in loan operations.
As seen in Table 1 This abnormal activity is related to a tax loophole that lasted until 2014, where mutual funds were exempt from the 15% tax on the payment of interest on net equity by companies.
Thus, the dynamics that drove stock lending were supported by the following logic: if a stock was loaned to a mutual fund during the distribution of JCP, the fund would receive the total dividend, regardless of the fiscal status of the lender of the stock. When the funds borrowed shares from lenders that were subject to withholding taxes, they had to repay only the net amount of the taxes that the lender earned, retaining most of the value of the taxes. These stock lending operations generated significant gains for the funds at the expense of the government, generating an abnormal increase in stock lending activity around JCP payment dates.
The increase in stock lending activity caused by these tax arbitrage opportunities restricted the supply of short selling loans during those days. Since these tax arbitrage activities were unrelated to the investors' expectations of stock returns, they provided an almost natural experiment of an exogenous variation in the availability of stock loans for short selling in the equity market.
The results showed that the increase in short selling restrictions for a stock around its dividend payment date caused a considerable increase in the price of the stock, corroborating the overpricing hypothesis also found by Chague et al. (2014). Furthermore, the author found that the average loan rate jumped fivefold, from an average of 2% of the notional value during normal periods to 10% during periods of JCP distribution, while the short positions increased from an average of 2.2% to 3%.

Final comments
The stock lending market and short selling strategies are a topic of interest to investors, regulators, and academics especially, since they can be empirically validated with available data. Saffi and Sigurdsson, 2011), (ii) they make the price discovery process less efficient (Saffi and Sigurdsson, 2011), and (iii) they distort corporate decisions, such as regarding higher investments and the issue of new shares (Grullon, Michenau, Securities Lending and Short Selling and Weston, 2015). These results allow us to conclude that the negative impacts of short selling bans and restrictions on the efficiency and good functioning of the markets are clear.
The debate on stock lending and short selling is relatively new in the Brazilian context, but the same is relevant for the results found, especially in regards to policy implications, as found by Chague et al. (2017), where the opacity of the stock lending market was responsible for significant frictions that made stock lending more costly for the investor than it would have been if it were done through an electronic trading venue.
We also believe that the subject has great potential for future studies in the Brazilian academic community, in particular due to the model adopted in Brazil, where the loans take place centrally, with the identification of investors and participation of a central counterparty (CCP).

Notas
1 B3 is the company resulting from the acquisition of CETIP by BM&FBOVESPA in 2017. 2 The loan fee that represents the total cost to borrow a share is thus the loan rate (which includes broker commission), plus a B3 annual fee of 0.25%. 3 The expression 'hard-to-borrow' (HTB), although not widely used in Brazil, describes those shares that are not easily lent. In the US, brokers usually inform their clients about the shares that are on this list and have a higher loan rate or may be compulsorily used (buy-in) by the broker as a way of ending the settlement cycle or meeting the demand (recall) of the original donor. This particular item was studied by De Genaro and Avellaneda (2018) for a set of leveraged ETFs traded in the US. 4 As part of an effort to update short sale regulations, the US Securities and Exchange Commission (SEC) put into force Regulation SHO on January 3, 2005. Regulation SHO requires, among other things, that the broker has a reasonable belief that the equity to be short sold can be borrowed and delivered to a short seller on a specific date before short selling can occur. Further details and amendments are available at https://www.sec.gov/ spotlight/shortsales.shtml. 5 https://www.esma.europa.eu/regulation/trading/shortselling 6 http://www.cvm.gov.br/legislacao/instrucoes/inst530. html