Affiliated Agents, Boards of Directors, and Mutual Fund Securities Lending Returns

advertisement
Affiliated Agents, Boards of Directors, and Mutual Fund
Securities Lending Returns♣
John C. Adams, Sattar A. Mansi, and Takeshi Nishikawa
February 18, 2013
Abstract
Prior SEC inquiries concerning self-dealing in securities lending programs suggest
potential conflicts of interest when funds employ lending agents that are affiliated with
their sponsor. We posit that the level of self-dealing is potentially greater, and mutual
funds’ securities lending returns are lower, when funds employ sponsor affiliated
lending agents. Using a manually collected U.S. index mutual fund sample, we find that
funds with sponsor affiliated lending agents have lower annual returns on lent securities
and that securities lending returns are significantly higher when funds administer their
own lending programs. We also find that mutual funds boards of directors provide a
monitoring role in the securities lending market. Specifically, we document that multiple
board of director appointments, more director fund ownership, higher board
independence, and lower excess director compensation are associated with higher
lending returns. Overall, the evidence has implications for mutual fund boards as they
consider lending proposals from affiliated agents and for future regulatory actions.
Keywords: Securities Lending; Governance; Affiliated Agents; Mutual Fund Board Structure
JEL Classification: G34, G32, G20
♣
Adams is at University of Texas at Arlington, Mansi is at Virginia Tech, and Nishikawa is at the
University of Colorado at Denver. The authors would like to thank Jason Zweig from the Wall Street
Journal for comments and suggestions that helped improve the paper. The remaining errors are ours.
1
Electronic copy available at: http://ssrn.com/abstract=1947503
I. Introduction
In 2004, the SEC launched an investigation into the management of securities lending
programs for mutual funds administered by a number of investment houses including State
Street, Bank of New York-Mellon, Northern Trust, and JP Morgan. The investigation revealed
potential conflicts of interest when funds employed securities lending agents that were
affiliated with their sponsor.1 It also revealed numerous instances where sponsors manipulated
the bidding process so that mutual fund boards would select affiliated over non-affiliated
lending agents and that most boards failed to adequately oversee securities lending programs.
Further concerns were expressed about off the books side payments not being disclosed to fund
shareholders, lending proceeds not going back to lending funds, and that securities lending
may only be profitable to the fund’s sponsor or lending agent and not the shareholders.2 The
result is a wide-spread criticism and lawsuits alleging that many securities lending agents either
misled investors about the risks of lending securities or inappropriately invested lending
collateral in risky investments that resulted in significant losses during the market crash of
2008.3 Despite these concerns anecdotal evidence points to an increase in the use of securities
lending as a means to enhance returns and the market has grown to over $12 trillion as of 2012.4
In this paper, we examine whether affiliated lending agents and mutual fund boards affect
securities lending returns in a hand collected sample of U.S. index mutual funds. This issue is
important as securities lending income benefits shareholders by offsetting fund expenses. The
extant literature largely ignores agency considerations between beneficial owners and lending
agents. Conflicts of interests are potentially severe in securities lending arrangements because
fund sponsors (or their corporate parents) have financial incentives to favor affiliated lending
agents.5 Although it is customary for mutual funds and lending agents to split lending
1 See Tom Lauricella, SEC Discovers Breaches in Lending Securities, Wall Street Journal (Jan. 29, 2007). The term
sponsor refers to the investment company which manages the fund. The term affiliated refers to sponsors and
lending agents that are part of the same corporate entity.
2 Robert Wittie, SEC Concerns With Securities Lending by Mutual Funds, 12th Annual Beneficial Owners’
Summit on Domestic and International Securities Lending and Repo (Feb. 2006).
3 See, for example, Jason Zweig, Is Your Fund Pawning Shares at Your Expense?, Wall Street Journal (May 30,
2009), Emily Lambert, Securities Lending Meltdown, Forbes (June 22, 2009), and Louise Story, Banks Shared Clients’
Profits, but Not Losses, New York Times (October 17, 2010).
4 See “Securities Lending Best Practices,” by eSEC Lending, 2012 report.
5 It is well documented that funds can increase investor cash inflows by increasing fund performance. One way
to increase fund performance is to maximize securities lending returns. However, what matters is the marginal effect
to the corporate parent (i.e., collecting higher fees versus increased cash flows to the funds they manage).
2
Electronic copy available at: http://ssrn.com/abstract=1947503
proceeds, details on these fee splits are rarely disclosed to investors. This lack of transparency
likely exacerbates agency problems associated with self-dealing behavior.
We posit that the level of self-dealing is potentially greater, and mutual funds’ securities
lending returns are lower, when funds employ sponsor affiliated lending agents. Mutual funds
can select their custodian bank or a third party specialist (i.e., broker/dealer), both of which
may be affiliated with their sponsors, as their lending agent. Mutual funds can also act as their
own securities lending agents. They can utilize an auction service, grant a principal
intermediary or other third party specialist rights to all or portion of their portfolios, or lend
directly to borrowers (e.g., such as hedge funds, market makers, and broker/dealers).
Alternatively, mutual funds can use some combination of custodian, third party specialist, and
self-lending routes to the securities lending market (i.e., seeking competitive prices from various
agents and borrowers for each security to be loaned).6 Self-lending funds likely retain a greater
portion of the proceeds from lending and are able to mitigate the conflicts of interest inherent in
all, and especially affiliated agent, lending agreements. Therefore, we expect that securities
lending returns are higher when funds administer their own lending programs.7
We also examine whether mutual funds boards of directors play a role in monitoring
securities lending programs. The Investment Company Act of 1940 (ICA) charges independent
directors with protecting mutual fund investors from conflicts of interests. The SEC has also
interpreted Rule 38a-1 of the ICA to subject securities lending agents to the review and
oversight of the fund’s board of directors.8 Therefore, we hypothesize that the effectiveness of
board oversight is associated with increased securities lending returns. Tufano and Sevick
(1997) report that as boards become increasingly independent of managers, their monitoring
effectiveness increases.
Del Guercio, Dann, and Partch (2003) and Adams, Mansi, and
Nishikawa (2010) find that board size is negatively related to overall performance. Chen,
Goldstein, and Jiang (2008) and Cremers et al. (2009) find that director ownership in the funds
they oversee is associated with improved shareholder value. Similarly, Khorana, Tufano, and
See the appendix for a detailed discussion on securities lending agreements.
We acknowledge that variation in securities lending returns across funds may be partly driven by collateral reinvestment restrictions that we cannot observe since funds typically do not disclose that information.
8 These guidelines state that boards must approve all components of a fund’s securities lending policies
including who will be the agent, the level of fees, to whom the securities will be lent, loan parameters, and lending
agent compensation – especially how funds and lending agents split the fees earned by lending.
6
7
3
Wedge (2007) examine mergers in the mutual fund industry and find that sponsor interests
often conflict with shareholder interests when directors are highly compensated. Overall, the
evidence indicates that the opaque nature of securities lending programs provides an enhanced
role for boards of directors.
Finally, we examine the relation between lending attributes and the likelihood of operating
a securities lending program. Fabozzi and Mann (2005) note that lender characteristics such as
large portfolios, large holdings of desirable securities, low asset turnover, and organizational
support (e.g., resources to support efficient lending practices) are valued by borrowers.
Sponsors have an incentive to lend so that they (or an affiliated agent) can capture a portion of
lending proceeds. Sponsors favor lending since it enhances performance and this in turn attracts
future cash inflows (see Sirri and Tufano, 1998). Therefore, we expect the likelihood of operating
a securities lending program to increase when funds are large, funds have more desirable
securities, and when mutual fund investors are especially sensitive to performance.
Using manually collected securities lending data of 1,374 fund-year observations covering
the period from 2003 to 2010, we find that securities lending returns are 40% lower than the
sample means when funds use sponsor affiliated lending agents. We confirm these results by
examining securities lending returns during the 2008 financial crisis and documenting the
negative effect of affiliated agent choice is four times larger during the crises than in other years.
We also find that lending returns are negatively associated with excess director pay, and
positively associated with board independence, directors with multiple board appointments,
and director ownership. We further document that the probability of lending increases with
fund and sponsor size and with increased institutional ownership. The latter suggesting that
fund managers lend to increase fund performance and attract institutional cash flows.
Additional testing reveals that securities lending returns are significantly higher when funds
administer their own lending programs.
Our paper is directly related to recent research by Evans, Ferreira, and Prado (2013) on
securities lending in the mutual fund industry. Evans et al. find that lending funds
underperform non-lending funds and this underperformance is caused by investment
restrictions that prevent funds from selling stocks with high borrower demand. They also find
that restricted funds partially offset future stock underperformance by lending securities they
are unable to sell. The next closest studies are those of Kaplan, Moskowitz, and Sensoy (2011)
4
who examine supply shocks and lending fees, Geczy, Musto, and Reed (2002) who examine
borrower returns from numerous strategies that involve short-selling, and Kolasinski, Reed, and
Ringgenberg (2013) who document considerable variation in fees across lending agents and
conclude that lenders benefit from an opaque lending market. Our paper differs from theirs in
that we investigate securities lending returns to the beneficial owners of lent securities rather
than to borrowers and securities lending agents.9
This study makes three main contributions to the literature on the securities lending market
and mutual funds. First, we examine securities lending agency issues from the perspective of
mutual funds who are the actual owners of the lent securities. This issue is of importance
especially in light of the recent SEC investigation into the management of securities lending
programs by various investment banks. We focus on index funds since they are widely held and
provide for natural and active lenders as they often have large positions and follow buy and
hold investment strategies thereby reducing recall risk to borrowers (D’Avolio, 2002; Duffie et
al., 2002). Second, we provide evidence of wealth expropriation from shareholders when funds
employ sponsor affiliated lending agents. This evidence is consistent with recent studies
examining the agency effects of sponsor affiliations on shareholder wealth. Hao and Yan (2012)
find that investment bank affiliated funds underperform unaffiliated funds in part because they
are more likely to hold underperforming securities of affiliated investment banks’ clients.
Bhattacharya, Lee, and Pool (2013) report that affiliated funds of mutual funds provide an
uncompensated insurance benefit to other funds in the family. Third, our results have
implications for mutual fund boards as they consider proposals from affiliated agents, fund
management deliberations as to whether and how to implement a securities lending program,
and for possible future regulatory actions.
The remainder of the paper is organized as follows. Section II discusses the sample, variable
measurements, and descriptive statistics. Section III presents our results on securities lending
returns, securities lending program size, and incidence of securities lending in a multivariate
framework. Section IV concludes.
9 Only one of the twelve lenders in the sample used by Kolasinski, Reed, and Ringgenber (2013) is a mutual fund,
nine are either agent or broker dealers, and two are hedge funds.
5
II. Data and Variable Measures
A. Sample
We use information from the SEC’s EDGAR and Morningstar’s mutual fund databases to
compile the sample. We manually collect securities lending data of index mutual funds from the
annual and semiannual Certified Shareholder Reports, Forms N-CSR and N-CSRS, which are
located in the SEC’s EDGAR database. SEC regulations began requiring registered investment
advisors to certify financial statements and other information in Form N-CSR filings in 2003 in
response to the Sarbanes-Oxely Act of 2002. In addition, we manually gather data from EDGAR
on each fund’s board of trustees (directors) from the Statement of Additional Information (SAI),
which is located in Form 485 of each fund’s prospectus.
We identify index funds from the December 31 editions of the Morningstar Direct mutual
fund database for the period from 2003 to 2010. The Morningstar database contains monthly
class-level returns and total net assets as well as yearly information including investment
objective, 12b-1 fees, expense ratios, management fees, and purchase constraints (e.g.,
institutional share classes). We use the monthly total net assets and fund return information to
compute the monthly standard deviation of funds’ net cash flows from investors (i.e., cash flow
volatility). For each fund, we also gather sponsor-level information including the number of
funds managed and the total net assets from all funds under management. Since many mutual
funds have multiple share classes that differ primarily in expenses, loads, and clientele, we
combine the different classes into a single fund. Specifically, we compute the net asset value
weighted average of the class-level data items. Merging the data and applying these
requirements yields a dataset of 1,374 fund-year observations on 204 funds from 78 sponsors
covering the period from 2003 to 2010.10
B. Measuring securities lending activity
We manually collect securities lending information from the financial statements and
accompanying notes of each fund. For each year of the sample, we record the securities lending
income from each fund’s statement of operations, which is located in Form N-CSR. When
10
We track funds through sponsor level mergers and acquisitions.
6
earnings from securities lending are less than 5% of gross income, Regulation S-X permits funds
to combine securities lending income with the proceeds of other miscellaneous earnings in the
“other income” category. In these instances we record the income from securities lending as
zero.
Forms N-CSR and N-CSRS list the outstanding dollar value of funds’ securities loaned
inclusive of the collateral posted by borrowers as a liability in the statement of assets and
liabilities (Securities Lending Liabilities). Since the level of lending activity can vary throughout
the accounting period, we compute the average of securities on loan at the beginning (prior year
Form N-CSR), middle (Form N-CSRS), and end of each year (current year Form N-CSR).
Securities lending income increases performance by offsetting operating expenses. To illustrate
the overall economic effect of lending on fund performance, we compute the ratio of securities
lending income to total fund net assets (SL Income to Fund TNA). Securities lending income is
a function of several factors including the supply and demand for lendable securities, rebate
rates, fee splits, lending agent fees, and returns on collateral reinvestments. Collectively these
largely unobservable factors determine the size of a fund’s lending program and the return
earned by lending securities. To better understand how lending agent characteristics, fund and
family factors, and governance structures influence securities lending income in mutual funds,
we decompose the ratio of securities lending income to fund assets as follows:
SLIncome
SLIncome
SLLiabilit ies
=
×
FundAssets SLLiabilit ies FundAssets
(1)
where SL Income is securities lending income, SL Liabilities is securities lending liabilities, and
FundAssets is total net assets. The ratio of securities lending income to securities lending
liabilities measures the return on lent securities (Return on Lent Securities), and the ratio of
securities lending liabilities to total assets captures the proportion of each fund’s portfolio that is
on loan (Securities on Loan).11 As Equation 1 shows, funds can increase securities lending income
by lending more securities at higher rates of return. Funds typically lend several securities per
year at quantities determined by borrower demand, SEC regulations, and fund lending policies.
We report the ratio of securities lending income to net assets (SL Income to Fund TNA) rather than gross assets
to facilitate comparisons with fund expense ratios. Because of differences in the reporting of cash vs. non-cash
collateral we use fund gross assets to compute the proportion of fund assets on loan (Securities on Loan).
11
7
These loans are typically open ended (e.g., no fixed maturity date) but are generally settled
within a few days or weeks. The returns on the individual loans depend on rebate rates that
vary according to supply and borrower demand and the returns on the reinvestment of
borrowers’ collateral (typically 102%-105% of the market value of the securities). The returns on
collateral reinvestments can vary considerably depending on the level of risk and costs of the
collateral reinvestment choice. The return on lent securities is our primary variable of interest
since it captures the overall results of these loans and how lending proceeds are shared between
the lending agent and fund (e.g., the fee split).
We further collect information related to how the securities lending program is structured
from Forms N-CSR and N-CSRS. Specifically, the name of the lending agent, lending agent’s
affiliation to the fund sponsor, type of lending agent, and whether the collateral of loaned
securities is invested in a lending agent affiliated portfolio. It is a common practice for funds to
disclose the details of their securities lending programs in the notes to the financial statement
section of its annual report in Forms N-CSR and NCRS although the level of disclosure varies
greatly by funds. We are able to identify the securities lending agent for about 91% of the cases
using the name or information provided in the filing. The remaining cases either conduct
securities lending programs on their own,12 or we were unable to identify the securities lending
agents. In the latter case, we examine Form 485 in an attempt to identify the lending agents. If
we are unable to precisely identify the lending agent, we classify these funds as unreported.
Using the identity of the lending agent, we record whether the lending agent is affiliated with
the fund sponsor. We also classify lending agents into three types: custodian, broker, or self.
We next record if the collateral for loaned securities is invested in a lending agent affiliated
fund, an unnamed collateral reinvestment pool, another fund offered by the trust family, or in a
fund that is not affiliated with the agent or sponsor. To ensure that the accuracy of our data for
this item, along with the notes to financial statements section, we utilize the information listed
in the Statement of Net Assets for each fund in Forms N-CSR and N-CSRS.
12 For example, Vanguard funds conduct their own securities lending programs. Vanguard 500 index fund states
in its 2006 annual report: “The fund may lend its securities to qualified institutional borrowers to earn additional
income. Securities loans are required to be secured at all times by collateral at least equal to the market value of
securities loaned. The fund invests cash collateral received in Vanguard Market Liquidity Fund.”
8
C. Measuring boards and trustees
We manually collect calendar year end data on each fund’s board of directors from the SAI,
which is included in each fund’s prospectus (Form 485). These include: board size, board
independence, multiple boards of directors appointments held by each trustee, trustee fund
family ownership, and trustee compensation. Board Size is computed as the natural logarithm
of the number of trustees serving on each fund’s board.13 Board Independence is measured as
the proportion of independent or disinterested trustees to total directors, where director
independence is determined in accordance with SEC (2004) regulations. Under these
regulations, independence indicates that an outsider is not an employee, not an employee
family member, not an employee or a 5% shareholder of a registered broker-dealer, and is not
affiliated with any recent legal counsel to the fund. The All Independent Directors is a dummy
variable that equals one if all of the directors on the board are independent.14
Multiple boards of directors appointments (Outside Directorships) is the average number of
other directorships, excluding appointments at non-profits, held by each board member. The
SAI also lists the value of each trustee’s ownership, in specific dollar ranges, of funds in the
fund family. Family Ownership is the proportion of directors with a minimum of $100,000 in
shares of funds within the fund family. We also record the overall compensation received for all
of the funds overseen for each trustee. Our last board structure variable, Unexplained
Compensation, measures each independent trustee’s compensation and is computed as the
residual from the regression model of trustee compensation on the number of funds overseen
by a trustee and the total net assets overseen by a trustee as in Tufano and Sevick (1997).
D. Control Variables
We employ several fund and family specific characteristics as controls in the analysis. These
include performance, size, external monitoring, management and custodian fees, liquidity, and
dummy variables that capture different fund investment objectives. Portfolio return, a proxy for
We rerun the analysis computing board size as the number of independent directors and find similar results.
In unreported analysis we include a dummy variable that equals one when a board’s chair is listed as
independent or disinterested in the SAI document. However, this variable is insignificant and its inclusion does not
substantially alter our findings.
13
14
9
the demand for lendable securities, is computed as a fund’s annual return net of operating
expenses.15 Fund and family size, proxies for the supply of lendable securities and pricing
power with lending agents, are computed as the natural log of total gross (inclusive of borrower
collateral) assets for the fund and total net assets for the fund family, respectively. Expense
ratios are included to control for each fund’s fee structure since lending income directly offsets
fund expenses and is computed as the ratio of net fund expenses (after lending income) to total
net assets. Management and custodian fees collected from Form N-CSR, proxy for the incentive
structure of funds and fund sponsors to engage in securities lending since fund managers and
sponsors can offset higher fees with securities lending programs. These fees are computed as
the proportion of each fee to total net assets.
Asset turnover and cash flow volatility are proxies for the risk that borrowers have to return
securities early. Asset turnover is computed as the lesser of sales or purchases divided by
average monthly total net assets and cash flow volatility is the standard deviation of monthly
investor net cash flows. Our external monitoring measure is computed as the proportion of each
fund’s total net assets that are in institutional share classes.16 We also include dummy variables
to denote whether funds are enhanced or managed since these funds are more likely to pursue
aggressive lending policies than purely passive indexers. Kolasinski, Reed, and Ringgenberg
(2013) note that the lending market is opaque which makes it costly for lenders and borrowers
to find each other in which case funds with investment objectives that hold high demand
securities may find it less costly to administer their own lending programs. Our analysis
includes 26 Morningstar investment category dummies to capture variations in lending costs as
well as the supply and demand for lendable securities that are due to investment objectives.17 A
brief description of the sample variables and the data sources used is provided in Table I.
[Insert Table I about here]
E. Sample description
We also consider each fund’s gross return before expenses and Jensen’s alpha and find similar results.
We also control for clientele effects by noting the presence of marketing costs, namely each fund’s average
12b-1 fees, front end loads, and deferred sales charges. Our results are robust to these alternative specifications.
17 Although not reported, we also employ prospectus investment objective and benchmark index dummies,
available from Morningstar, and find similar results.
15
16
10
Table II provides summary statistics for the securities lending practices in our sample. Panel
A reports the incidence of securities lending activities segmented by investment objective. The
data indicate that the largest category of index funds is US large blend funds (43%), followed by
US small blend (11%), foreign large blend (10%), US mid-cap blend (10%), US intermediate
bond (7%), and US large growth (4%). In terms of securities lending programs, funds with the
highest incidence of lending are concentrated mainly in the US mid-cap blend and US
intermediate bond categories (both at about 84%) and foreign large blend funds (76%), followed
by US large blend, US small blend, and all other categories (e.g. Asian and European growth,
technology, financial, real estate, government bonds, and balanced index funds) of about equal
amounts (in the range of 68% to 69%). For the overall sample of 1,374 fund-year observations,
about 70% report securities lending income.
[Insert Panel A of Table II about here]
Panel B reports the distribution of securities lending across fund TNA, sponsor TNA,
expense ratio, and portfolio return deciles. About 52% of funds in the smallest Fund size decile
report securities lending income while 100% of funds in the largest fund TNA decile report
lending income. In fact, each of the five largest size deciles has a greater incidence of securities
lending than do the smallest five deciles.
The incidence of securities lending is similarly
distributed in the sponsor TNA deciles with the largest fund sponsors lending 86% of the time
and the smallest only about 38% of the time. Securities lending is more prevalent in funds with
low expense ratios, where 81% of funds in the lowest expense ratio decile have a securities
lending program than in funds with the highest expense ratios (in this case only 50% lend).
Panel B also reports a similar incidence of securities lending across high and low return funds.
[Insert Panel B of Table II about here]
Panel C of Table II shows the incidence and medians values for the percentage of portfolio
securities on loan and return on securities lent segmented by securities lending agent type,
whether the lending agent is affiliated with the fund and/or fund sponsor, and how the
collateral is invested. As expected, the majority of the index funds in the sample (about 52%)
use a custodian to implement a securities lending program. However, a large number of funds
11
(about 36%) are their own lending agent while only 11% of the sample has an independent
lending agent or broker to handle all of their lending.
Funds that use a custodian as their lending agent lend about 8.8% of fund assets (median
value) while funds that act as their own agent (self-lending agents) only lend about 2.9% of
fund assets. Interestingly, funds with custodian lending agents only earn about 27 basis points
per year (median value) on the securities they lend while self-lending agents earn a return that
is about four times higher (about 105 basis points). The increased levels of securities lending in
custodial versus self-lending agents are consistent with preferences to borrow from custodians
who are able to lend large quantities of securities since they already hold the securities of
several funds and fund families. The higher returns for funds who self-lend suggest they are
able to obtain better terms from borrowers and do not incur the level of securities lending agent
fees as do funds that employ outside lending agents. Alternatively, self-lending funds are more
likely to have a policy of only lending those securities that are especially profitable to lend (an
industry practice known as value lending) and custodial lending agents earn a lower average
lending return by lending both high and low profit securities.
[Insert Panel C of Table II about here]
Panel C also reports that 29% of the sample employs lending agents who are affiliated with
the fund sponsor. Funds that use affiliated lending agents lend more of their portfolio securities
than funds that use unaffiliated agents (8% vs. 5%) but report a median return on lent securities
lower than funds that use unaffiliated agents. When collateral for lent securities is invested
with a lending agent affiliated mutual fund or with an unnamed pool (the vast majority of
unnamed collateral reinvestment pools are controlled by lending agents) the percentage of
portfolio securities on loan is about 8% versus 4% when the collateral is invested with a money
market type mutual fund in the same family trust as the lending fund. The median percentage
return on lent securities is about 73 and 91 basis points when the collateral is invested with a
family or non-affiliated fund, respectively. However, when the collateral is invested with a
lending agent affiliated fund or reinvestment pool the returns are 37 and 22 basis points.
12
Overall, Panel C reports considerable differences in the amount of securities on loan and the
return on lent securities across lending agent types.18
Table II also reports securities lending income by investment objective for funds with
securities lending income. For ease of exposition, Panel D reports results for the six investment
objectives with the highest levels of lending activity and aggregates the remaining investment
objectives into an ‘other’ category. Consistent with Diether, Lee, and Werner (2011) who note
that stock characteristics explain much of the variation in lending fees, we report significant
differences in securities lending returns when segmenting by fund investment objective. Panel
D reports distributional statistics for the dollar amount of securities lending income, the
percentage of portfolio securities on loan, the return on lent securities, and percentage of
securities lending income to fund TNA. US mid-cap index funds had the highest mean dollar
value of securities lending income, $2.06 million, whereas foreign large blend funds had the
highest median value, $479 thousand. US small blend index mutual funds had the most
portfolio securities on loan with a mean value of 12.03%. In terms of median values, US midcap and US intermediate bond funds had the most securities on loan at 11.31% and 11.02%,
respectively. Foreign large blend funds earned the highest median return from their lent
securities and US large blend funds received the lowest, perhaps reflecting better rebate rates
for more difficult to borrow international securities and the greater supply and consequently
higher rebate rates of large blend securities (e.g., S&P 500 index constituents). The net result of
the amount of portfolio securities on loan and the return of lent securities is the securities
lending income to TNA. US Small blend, US mid-cap and foreign large blend funds had the
highest levels of lending income to TNA with mean values of about 8-12 basis points.19
Panel A of Table III provides fund level summary statistics for the remaining variables used
in this analysis for securities lenders and non-lenders. Included are the mean, median, and
standard deviation values for the fund and governance characteristics. In terms of fund
characteristics, the mean and median fund size for securities lenders is $4.89 billion and $832
million, respectively. The non-lending funds are much smaller with mean and median fund
18 Note that the number of funds reported in collateral reinvestment section of panel C exceeds the overall
number of funds in the sample as funds can use multiple collateral reinvestment vehicles.
19 Panel D also reports the first and third quartile values and for each measure of securities lending considerable
interquartile ranges exists. This suggests that some funds are more aggressive in pursuing securities lending
opportunities than others.
13
total net assets of $560 million and $232 million. A similar pattern holds for family size. The
larger fund and family sizes are consistent with larger funds and fund families having the
resources to manage a lending program and borrowers’ preferences for borrowing securities
from a single source.
[Insert Panel A of Table III about here]
Non-lenders have higher mean and median portfolio returns than lenders, however lenders
have higher measures of Jensen’s alpha. Securities lending funds charge lower expense ratios
and management fees than non-lenders. Median custodian fees are about 1 basis point for
lenders and 2 basis points for non-lenders. This is consistent with anecdotal evidence
suggesting that custodians offer lower fees in exchange for the opportunity to serve as lending
agents. This is also consistent with an economy of scale effect as lenders are on average larger
than non-lenders. Average asset turnover is higher for non-lenders, but median asset turnover
is lower than for lenders. The reverse pattern holds for cash flow volatility. However,
institutional ownership mean and median values are much greater in funds that lend. This is
consistent with institutional investors being more knowledgeable than retail investors and
seeking the lowest fees. Since securities lending income directly reduces expense ratios lenders
are more likely to charge lower fees and attract institutional investors than non-lenders.
In terms of fund governance, lenders and non-lenders have similar board sizes with about
eight directors each. Lenders appear to have slightly more independent directors than nonlenders with median values of 86% and 80%, respectively. The average number of outside
directorships per trustee is greater for lenders than non-lenders. Family ownership levels are
higher in funds with securities lending income with mean and median values of 86% and 100%
versus non-lender values of 81% and 89%. The mean and median unexplained compensation
for directors is higher for lenders than non-lenders by about $24,000 and $8,800, respectively.
Overall, Panel A reports considerable variations in fund and governance characteristics of
lenders and non-lenders.
Panel B reports Pearson correlations among the variables in the sample. The correlation
between the return on lent securities and the amount on loan is negative and significant. The
return on lent securities is also negatively correlated with the presence of an affiliated lending
14
agent. The correlation between the return on lent securities and the dummy variable that
captures whether a fund acts as its own lending agent is positive and significant. The remaining
fund measures (fund and family TNA, independent directors, and director family ownership)
are also significantly positively correlated with the return on lent securities. In contrast, the
amount on loan is positively correlated with affiliated lending agents and unexplained
compensation and negatively correlated with fund and family TNA. However, the amount on
loan and the return on lent securities are both positively correlated with independent directors.
[Insert Panel B of Table III about here]
III. Empirical Results
We conduct cross-sectional analysis to examine relation between returns on lent securities
and lending related governance mechanisms while controlling for fund-specific measures as in
Adams, Mansi and Nishikawa (2010). To test the hypotheses that the returns to lending
programs are influenced by lending agent type, whether the agent is affiliated with the fund
sponsors, and that effective boards mitigate conflicts between agents and fund shareholders, we
apply the following specification using fund level clustered standard errors
3
6
j =1
j =4
SecuritiesLending ei,t = β0 + Σ (Affiliatedi,t ) + Σ (Lending Agent Typei,t )
12
20
+ Σ (Board Characteristicsi,t ) + Σ (Fund Characteristicsi,t )
j =7
j =13
46
64
j = 21
j =46
+ Σ (Investment Objectivei,t )+ Σ (Timet ) + εi,t,
(2)
where Securities Lending is measured as the return on lent securities. Note that securities lending
return calculations are vulnerable to measurement error as the amount of securities on loan is
only observable at the beginning, middle, and end of each fiscal year while loan durations
typically range from a few days to a few weeks. This measurement error inflates lending return
variance and biases the t-statistics downward. We mitigate this potential measurement problem
in two ways. First, we limit our sample to funds that report securities on loan at least twice
during the fiscal year (i.e., lending liabilities at either the beginning and end of the year, middle
15
and end of year, etc.). Second, we trim observations with securities lending returns that are
below or above the first and 99th percentiles (i.e., a 1% trim).20 Affiliated are dummy variables
that indicate whether the lending agent is affiliated with the fund sponsor and if securities
lending collateral is invested in an agent affiliated fund or an unnamed investment pool.
Lending Agent Type consists of dummy variables to indicate whether a fund self-lends, uses a
broker or dealer, or if lending agent type is not reported. Board Characteristics include board size,
independent directors, all independent directors dummy, outside directorships, director
ownership levels in the fund family, and director compensation. Fund Characteristics include
portfolio return, family TNA, fund TNA, institutional ownership, fund turnover ratio, cash flow
volatility, enhanced fund dummies, custody fees, and the percentage of portfolio securities on
loan. The variables Investment Objective and Time represent dummies for each fund’s
Morningstar investment objective category (large blend, large growth, large value, mid-cap
blend, foreign blend, etc.) and time (year and fiscal year end month) dummies, respectively.
A. Securities Lending Returns
Table IV presents the results from regressing securities lending returns on fund, family, and
lending program characteristics. We employ six models to minimize multicollinearity effects.
Model 1 reports regression coefficients for the specification that focuses on the effect of fund
characteristics on securities lending income. Model 2 is similar to Model 1 but adds custodian
fees. Since custodian fees are not uniformly reported the number of observations in Model 2 is
less than that reported for Model 1.21 Model 3 incorporates dummy variables to capture lending
agent type.22 Model 4 includes affiliated agent and collateral investment measures.23 Models 5
20 In unreported analysis we obtain similar results with 0%, 0.5%, and 2.5% sample trimming and from requiring
one or three lending observations during the fiscal year.
21 Custodian fees are often bundled with other services and not separately reported.
22 The Vanguard funds represent about 27% of the overall sample and about 48% of the self-lending sub-sample.
Because Vanguard does not retain any securities lending proceeds from the funds they sponsor and follows a value
oriented policy of only lending high yielding securities, we repeat our analysis excluding Vanguard funds and find
similar results.
23 Mutual fund managers are employees of sponsors and not fund shareholders so fund managers who
administer their own lending programs are affiliated with fund sponsors. That is, all self-lending agents are affiliated
with fund sponsors but not all sponsor affiliated agents are self-lending agents. Our use of the term affiliated lending
agent applies to external agents (custodians, broker/dealers, etc.) and not internal agents (i.e. fund managers).
16
and 6 add lending program size (the percentage of each fund’s portfolio that is on loan) to the
affiliated agent and lending agent type specifications.
[Insert Table IV about here]
Model 1 reports a positive and significant (at the 1% level) relation between fund size and
securities lending returns, which suggests that larger funds are able to manage their portfolio
lending more effectively than smaller funds, possibly because managers of larger funds have
greater abilities to generate lending returns than managers of smaller funds. Alternatively,
borrowers may seek larger lending portfolios in order to reduce costs associated with locating
additional sources of borrowable securities. If so, they may be willing to pay a premium to gain
access to larger pools of securities. Our finding that family size is also positively and
significantly related to lending returns is consistent with both explanations. Cash flow volatility,
our proxy for the risk of early redemption, is positive and statistically significant (at the 1%
level) because the characteristics of the underlying fund assets that lead mutual fund
shareholders to make buy and sell decisions are also relevant to short sellers. Alternatively,
higher redemption risk is associated with higher returns simply since, on average, riskier
securities offer higher lending returns. The remaining fund characteristics are statistically
insignificant.
Model 2 reports the estimated coefficient for family size is similar in size, sign, and
significance to that reported in Model 1. Interestingly, Model 2 finds that none of the other
variables, including custodian fees, are statistically significant. Model 3 reports estimated
coefficients for three types of non-custodian lending agents. Funds who act as their own
lending agents (Self Lending Agent), funds that employ a 3rd party agent, a broker and/or a
dealer and funds that do not disclose their lending agent in the N-CSR or Form 485. The
coefficient for the self-lending agent dummy is economically large, positive, and statistically
significant at the 1% level. The results from Model 3 suggest that securities lending returns are
about 1.3% higher in funds that administer their own lending program rather than using their
custodian to lend securities. Furthermore, this value is almost equal to the overall sample
average return on lent securities of about 1.5%. Applying the sample mean values for the
amount of securities on loan and fund TNA, the estimated benefit of self-lending versus agent
17
lending is over $5 million per fund.
Since most fund complexes offer several funds the
economic incentives to administering lending programs internally can become large. These
results suggest that funds can earn higher lending income and charge lower expenses to their
shareholders by effectively administering their own lending program. This finding also
indicates potential conflicts in securities lending arrangements between funds and affiliated
lending agents where arm’s length bargaining is potentially problematic.24
Model 4 provides results from including measures intended to capture potential agency
conflicts between fund shareholders and sponsors via sponsor affiliated lending agents. The
most interesting finding is that the estimated coefficient on the sponsor affiliated lending agent
dummy variable (Sponsor Affiliated Agent) is negative and highly statistically significant (at the
1% level). The result is also economically large since returns to lent securities are about 0.6%
lower when the lending agent is affiliated with the sponsor. Applying the sample average
annual return on lent securities of 1.5%, our results indicate that affiliated lending agents are, on
average, associated with a 40% reduction in lending returns. This finding suggests self-dealing
behavior by some affiliated lending agents. Model 4 also reports that investing the collateral
that borrowers commit to secure securities loans with lending agents (Collateral Invested with
Lending Agent Fund and Collateral Invested in Unnamed Pool) are associated with lower lending
returns (at the 1% and 5% levels). These results indicate that, on average, lending agents do not
provide competitive collateral reinvestment returns. These results also suggest that lending
funds should carefully consider whether the benefits of bundling lending and collateral
reinvestment services, a common industry practice, outweigh the costs.
Models 3 and 4 present considerable evidence that securities lending agent choice
influences the return mutual funds earn on lent securities. This is important as securities
lending returns increase mutual fund returns. However, lending returns are correlated with the
24 Self-lending funds likely incur additional staffing and related costs from administering their own lending
programs while funds that employ lending agents incur little additional lending expenses. That is, securities lending
income for agent lending funds is generally reported net of expenses while the reported securities lending income for
self-lending funds may not include all lending related expenses. If lending costs are not deducted from lending
income but are included in overall fund expenses our findings could overstate the true benefits of self-lending. In
unreported univariate and cross sectional analysis we examine fund gross expense ratios (e.g. what expenses ratios
would have been without lending income) and do not find any evidence that self-lenders have higher operating
expenses. In addition, the size of the self-lending benefit is much larger than we anticipate any additional staffing
cost to be, especially when considering that lending staffing costs are likely spread over several funds in the complex.
18
portion of assets on loan since some funds follow a value oriented policy of only lending when
lending yields are relatively high while others follow a wholesale approach and lend as many
securities as possible (LaBarge, 2011). Funds that follow a value orientated lending policy earn
relatively more of their securities lending income through lending fees and less by reinvesting
borrowers’ collateral since they typically have fewer (but higher yielding) securities on loan. In
contrast, wholesale lenders focus on earning securities lending income via reinvesting borrower
collateral so they have incentives to lend more and invest the collateral in higher risk, higher
return investment vehicles. Panel C of Table II notes considerable variation in lending policies
across lending agent type (e.g. custodian vs. self-lending) and lending agent affiliation status
and suggests that self-lending funds, on average, follow a value lending policy while custodian
and affiliated lending agents tend to take a wholesale approach. In addition, funds that
administer their own lending programs may have fewer lending opportunities than funds
which employ custodian lending agents.
Models 5 and 6 repeat the analysis of Models 3 and 4 while controlling for the portion of
fund assets on loan in order to capture possible variations in securities lending returns due to
differences in lending strategies. As in the previous specifications, Models 5 and 6 include fund
investment objective category dummies to capture borrowing demand across different types of
securities. All of the control variable coefficients that were significant in Models 3 and 4 are also
significant in Models 5 and 6. The estimated coefficients on the percentage of portfolio loaned
are economically large, negative, and statistically significant at the 1% level. The estimated
coefficient on the self-lending agent dummy in Model 5 remains positive and statistically
significant at the 1% level.
Similarly, Model 6 reports that affiliated lending agents are
significantly (at the 1% level) and negatively associated with lending returns after controlling
for the level of lending activity.
Overall, the results in Table IV suggest that securities lending returns are lower when funds
use sponsor affiliated lending agents. It also demonstrates how mutual funds that administer
their lending programs, either by acting as their own agent or employing their custodian, effects
securities lending returns. As such, the results provide compelling evidence that conflicts of
interests exist in securities lending programs. In addition, Table IV reports that securities
lending returns are negatively associated with lending agent affiliated collateral reinvestment
choices and to the percentage of fund assets on loan.
19
B. Additional Tests
B.1 The Financial Crisis
For robustness, we examine our results using an exogenous shock provided by the financial
crisis. The credit and liquidity crisis that reached a peak around the Lehman Brothers
bankruptcy of 2008 presents a unique opportunity to examine returns generated by self and
affiliated lending agents. Although often portrayed as a low risk strategy to enhance fund
returns, funds can lose money when lending securities. Funds face two main risks: counterparty
risk, the concern that the borrower may not return the lent securities, and collateral
reinvestment risk, the possibility that the value of the collateral investment is insufficient to
repay the borrower when the lent securities are returned. Both counterparty and collateral
reinvestment risks increase as funds lend more. When lenders need cash to repay borrowers’
collateral (or liquidity in general) they may be forced to sell the collateral investments at steep
discounts during a credit or liquidity crisis. The discounts will be greater for longer maturity
and riskier reinvestment pools. One way that lending agents can engage in self-dealing
behavior is to offer less favorable fee splits to mutual funds while lending more securities.
Although this strategy can result in overall lending revenues that are comparable to strategies
that feature more favorable fee splits to the funds with fewer (higher yielding) securities on
loan, it also exposes funds to greater risk. Lending agents are less concerned with securities
lending risks since, as agents, they typically do not share in lending losses. While some agents,
most notably custodial lending agents, do indemnify funds against counterparty risk it is rare
for agents to indemnify against collateral reinvestment risks. We hypothesize that agency
problems in lending programs are exacerbated during periods when counterparty and
reinvestment risks are particularly high.
Table V presents the results from repeating the specifications of Models 5 and 6 in Table
IV.25 The specifications include the amount of securities on loan (and by extension the collateral
reinvestment levels) to account for variations in securities lending returns due to value and
wholesale lending approaches and are segmented by the year 2008. Models 1 and 2 provide
25
For ease of exposition only selected variable coefficients are reported.
20
results on the impact of agent type (e.g., self-lending vs custodial lending). Models 3 and 4
report coefficients estimates for affiliated agent and collateral reinvestment status.
The estimated coefficients for the self-lending agent dummy variable are positive and
statistically significant (at the 1% level) in Models 1 and 2, suggesting that funds that administer
their own lending programs outperform agent lenders regardless of market conditions.
Furthermore, these results are consistent with those previously reported in Table IV and do not
appear to be driven by differences in value and wholesale lending strategies as both models
control for the amount of a fund’s portfolio on loan. The magnitude of the self-lending
estimated coefficient in Model 1 is about twice as large as the coefficient in Model 2 (a difference
that is statistically significant at the 1% level). This difference indicates that self-lending funds
were better able to manage the risks associated with the credit and liquidity crisis than were
agent lenders.
Models 3 and 4 report results for the affiliated agent lender specifications. The results for
the control variables are consistent with those reported in Models 1 and 2. Model 3 reports, for
2008, that sponsor affiliated agents are associated with a 0.8% decrease in securities lending
returns, results that are statistically significant at the 5% level. This is about twice the negative
estimated coefficient for affiliated agent lending reported in Model 4 (all other years). The
estimated coefficients for collateral reinvestment choices are statistically significant in both
models but there is considerable variation in economic significance. Specifically, Models 3 and
4 report that funds which elected to invest borrowers’ collateral in lending agent affiliated funds
experienced significantly negative securities lending returns in all years, but the negative effect
in the year 2008 was about four times that of other years (estimated coefficients of about -1.5%
vs-0.4%). Similarly, the negative relation between securities lending returns and unnamed
collateral reinvestment pool choices is much greater during the financial crisis (again about four
times). This is consistent with the industry practice of lending agents usually not indemnifying
funds against collateral reinvestment losses. Furthermore, as in Models 1 and 2, these results
account for the percentage of portfolio assets on loan so the results are unlikely to be driven by
differences in value and wholesale lending strategies.
Table V reports that when funds administer their own lending programs and collateral
reinvestments their securities lending returns are higher than the returns of funds that employ
agent lenders, especially during periods of increased lending market uncertainty. Table V also
21
reports that affiliated agents are associated with lower lending returns when collateral
reinvestment risk is particularly high. As such, the results presented in Table V are consistent
with the hypothesis that agency costs inherent in securities lending programs are elevated
during periods of heightened counterparty and collateral reinvestment risks. The results are
also consistent with conflicts of interest not only resulting in lower lending returns but also
subpar risk/return tradeoffs.
B.2 Endogeneity and Sample Selection Bias Tests
Our analysis suggests that the choice of securities lending agent (e.g., affiliated vs. nonaffiliated) causes lending returns to be relatively high or low. However, the opaque nature of
securities lending markets makes any determination of causality problematic. Rebate rates, fee
splits, and collateral reinvestment policies, are important but unobservable determinates of the
returns earned on any lent security. If agent choice is correlated with these unobservable
factors the affiliated and self-lending agent dummies are endogenous. While it is not possible
to completely mitigate endogeneity concerns, in addition to employing investment objective
category, year, and fiscal year end month fixed effects, we employ a two stage least squares
approach and use sponsor organization type (e.g., whether the investment company that
sponsors the fund is publicly traded or privately held) as an exogenous first stage instrument.26
Sponsor ownership structure is likely determined before funds are created and any
securities lending policies are set. Adams, Mansi, Nishikawa (2010) and Ferris and Yan (2009)
demonstrate that shareholders of mutual funds sponsored by publicly traded investment
companies suffer more from conflicts of interest than funds sponsored by privately held firms.
If so, we expect public sponsors to be more likely to engage in self-dealing behavior such as
favoring affiliated agents that don’t offer the best fee splits or execution and to avoid selflending where gains accrue to the mutual fund shareholders and not to the sponsor or its
affiliated agent. In addition, public sponsors are usually part of large corporate parents who are
26 In unreported analysis we use the number of share classes offered by each fund as a proxy for operational
complexity. Our rationale is that fund sponsors and/or managers with the ability to manage complex fund
operations are more likely to internally manage lending programs (i.e., act as self-lending agents) and that any
influence of operational complexity on securities lending returns is via the decision to self-lend. Our results are
robust to the use of this alternative instrument.
22
more likely to have affiliated subsidiaries that offer lending services (e.g. custodial banks).
Rebate rates are market determined and U.S. offered index funds generally invest securities
lending collateral in money market type investment vehicles so sponsor organizational
structure only matters to securities lending returns via its role in influencing lending agent
choice. As such, sponsor organizational type is a plausible instrument since it satisfies both the
relevance (e.g., public sponsors are more likely to favor affiliated agents) and the untestable
exclusion (only influences returns via its effect on agent choice) conditions.
Table VI reports results from two stage instrumental variable regressions as in Wooldridge
(2002).
Models 1 and 4 report the first stage estimated coefficients obtained from probit
regressions of the self-lending and affiliated agent dummy variables, respectively, on the public
sponsor dummy instrumental variable and fund characteristics. The remaining models report
second stage ordinary least squares regression results. The self-lending Models 2 and 3 are
comparable to Models 3 and 5 of Table IV and the affiliated agent Models 5 and 6 are
comparable to Models 4 and 6 of Table IV.
[Insert Table VI about here]
Consistent with our expectations, Model 1 of Table VI reports the estimated coefficient on
the public sponsor dummy is negative and statistically significant at the 5% level. Model 2
reports that the self–lending agent dummy is positive and statistically significantly related to
the return on lent securities. Model 3 of Table VI reports that the estimated coefficients for the
self-lending dummy and the percentage of portfolio loaned are both statistically significant,
results that similar to those reported in Table IV. Consistent with the requirement that
instrumental variables be relevant with regards to the endogenous variable, the public sponsor
dummy coefficient is positive and significant in Model 4. Similar to the results reported in
Table IV, the sponsor affiliated agent and collateral invested with unnamed pool dummy
variables are both negative and statistically significant in Models 5 and 6.
However, the
estimated coefficient on the collateral invested with lending agent fund dummy is insignificant
in Models 5 and 6. Finally, similar to previous models the percentage of portfolio loaned is
significant in Model 6.
23
We also consider potential sample selection bias arising from selective disclosure of
securities lending income by mutual funds (e.g. window dressing). Regulation S-X permits
funds to put securities lending income in the “other income” category on income statements
when earnings from securities lending are less than 5% of gross fund income. This allowance
makes it difficult to identify funds with poorly performing lending programs. We address the
issue of sample selection bias by implementing the Heckman (1979) approach using our original
data set that includes funds that do not report securities lending income in the first stage to
calculate the inverse Mill’s ratio which we use as an additional explanatory variable to control
for sample selection bias in the OLS models. The results of this alternative specification are
reported in the appendix, Table A.1, and are similar to those reported in Table IV.
Overall, the results reported in Tables IV, V, VI, and Appendix A.1 provide strong evidence
that the choice of securities lending agent (e.g., affiliated vs. non-affiliated) is an important
driver of variation across securities lending programs and suggest conflicts of interest result in
lower securities lending returns for mutual fund shareholders. We next consider the impact of
boards of directors, who are responsible for awarding contracts to lending agents and
monitoring lending programs on behalf of mutual fund shareholders, on securities lending
returns.
C. Boards of Directors and Securities Lending
Tables IV and V strongly suggest that agency problems are potentially severe in securities
lending arrangements. We next examine whether mutual fund boards mitigate conflicts
between sponsors, lending agents, and fund shareholders. Table VII repeats the analysis of
Table IV while including mutual fund board and director characteristics. Models 1 and 2
provide results for regressions that include different measures of board independence: the
percentage of independent directors on a fund board (Independent Directors) and a dummy
variable that takes on a value of one if all of the directors on a board are classified as
independent (All Independent Directors). Models 3 and 4 include the agent type and affiliated
agent variables, respectively, to the specification presented in Model 1.
The significantly negative coefficients for the percentage of portfolio loaned in Table VII are
comparable to those reported in Table IV. Models 1 and 3 report that boards comprised of more
24
independent directors are associated with statistically significant (at the 10% level) positive
lending returns. These findings are consistent with Tufano and Sevick (1997), who report lower
expense ratios when boards have greater percentage of independent directors. Model 2 reports
higher securities lending income when mutual fund boards have all independent directors,
however the result is not statistically significant. In unreported analysis we find a negative and
insignificant independent chair coefficient which calls into question the efficacy of independent
board leadership. This finding is also consistent with Lutton et al. (2011) who report lower
index mutual fund fees for independent board chairs.
Table VII also considers individual director attributes. Directors with more outside
(multiple) board appointments (Outside Directors) are associated with higher lending returns
(results that are significant at the 1% level), perhaps because more in demand directors are
better monitors. Consistent with the findings of Chen, Goldstein, and Jiang (2008) that director
ownership in the funds they oversee is associated with improved shareholder value and/or
governance, Models 1 and 2 report that director mutual fund ownership (Family Ownership) is
positively associated with securities lending returns. In addition to being statistically significant
(at the 5% level), fund family ownership by directors is also economically significant. Our
analysis indicates that annual securities lending returns are, on average, about 1% higher when
directors own shares in the funds they monitor (which represents an increase of about 66% over
the sample mean securities lending return).
Interestingly, Models 1 and 2 indicate that
relatively higher paid directors (Unexplained Compensation) are associated with lower returns
suggesting highly compensated directors may fear losing their positions and therefore may not
adequately monitor sponsor lending program proposals (see Khorana, Tufano, and Wedge,
2007).
The sign and significance on the self-lending dummy coefficient reported in Model 3 are
similar to those presented in Table IV, although the economic impact is slightly lower. Similarly,
Model 4 shows that sponsor affiliated agents are associated with lower securities lending
returns. However, the sponsor affiliated agent estimated coefficient is statistically insignificant;
suggesting that boards of directors can mitigate some of the agency problems in securities
lending programs. Similar to Table IV, the agent affiliated collateral investment choices (agent
fund and unnamed pool) are negative and significant in Model 6. Overall, the results presented
in Table VII show that lending agent collateral reinvestment choices are associated with lower
25
returns on loaned securities even after controlling for boards of directors. This finding is of
particular interest given the recent concerns that boards of directors may not be adequately
monitoring securities lending programs. Table VII also supports our earlier finding that funds
which administer their own lending programs is robust to the inclusion of board and director
characteristics. Finally, Table VII reports that the associations between board of director
attributes and securities lending returns are generally economically and statistically smaller
after controlling for lending agent choice. This finding is consistent with board influence on
lending returns is primarily related to lending agent choice.
D. Logistic Analysis on the Incidence of Securities Lending
We next provide multivariate logistic analysis to examine whether fund characteristics are
related to the incidence of securities lending. We apply the following logit model and compute
fund clustered errors as in Petersen (2009).27 That is
1 + exp( − β 0 − β1−6 ( FundCharacteristicsi ,t −1 )



E ( SecuritiesLending i ,t ) =  − β 7−11 ( BoardCharacteristicsi ,t



 − β12−37 ( Investment Objectivei ,t ) − β 36−55 (Timet )) 
−1
(3)
Where SecuritiesLendingi,t is a dummy variable that takes on a value of one if a fund reports
securities lending income in its financial statements that are located in the annual report, if a
fund self lends, or if a fund employs an affiliated agent. All models include FundCharacteristics
such as portfolio return, family TNA, fund TNA, institutional ownership, asset turnover ratio,
cash flow volatility, and enhanced fund dummies.28 Board Characteristics include board size,
independent directors, outside (multiple) directorships, family fund ownership of directors, and
unexplained director compensation. As in the earlier models, the variables InvestmentObjective
27 Although we compute fund clustered standard errors as in Petersen (2009) in all of the cross-sectional
analyses, this methodology can be problematic when the number of observations within clusters is small (Donald
and Lang, 2007). As such, we repeat the analysis using robust standard errors and find similar results. For
completeness, we also compute fund family and time clustered standard errors. We report t-statistics using fund
clustered standard errors since securities lending returns within family and time clusters are often negatively
correlated (e.g., the securities of some family funds can be in high demand while the securities of other family funds
may be in low demand) which in turn biases the t-statistics upward. In general, the t-statistics obtained from fund
clustering are similar or smaller than those obtained from family and time clustering.
28 In unreported analysis we include expense and management fee ratios and find similar results.
26
and Time represent dummies for each fund’s benchmark investment objective category, year,
and month dummies.
Table VIII provides results for four model specifications. Model 1 reports estimated
coefficients for the primary specification that includes each fund’s annual portfolio return, fund
and family total net assets, institutional ownership, cash flow volatility, and asset turnover.
Model 2 incorporates board characteristics, Model 3 uses the self-lending dummy as the
dependent variable, and the dependent variable in Model 4 is the affiliated lending agent
dummy. Models 1 and 2 both report positive and significant relations between prior year
portfolio returns and the likelihood of reporting securities lending income. The results for
Models 1 and 2 indicate a positive and significant relation (at the 1% and 5% levels) between
fund size and the likelihood of a fund reporting securities lending income. This finding suggests
that economies of scale are important considerations in operating a securities lending program.
Similarly, the estimated coefficients on fund family size are positive and significant (at the 1%
level). This is consistent with larger sponsors being more able to provide the expertise to lend
securities than their smaller competitors.
[Insert Table VII about here]
Model 1 reports that institutional ownership is associated with an increased likelihood of
funds having securities lending programs (at the 1% level). This finding is consistent with
sophisticated investors seeking higher returns and lower expenses, which in turn provide
incentives for funds to engage in securities lending in order to improve fund performance.
Models 1 and 2 also report positive and marginally significant (at the 10% level) estimated
coefficients for cash flow volatility. None of the board characteristic variables are significant in
Model 2. The results presented in Models 1 and 2 suggest the decision to lend securities
depends on fund characteristics and not board oversight.
In contrast to Models 1 and 2, the estimated coefficient on portfolio return reported in
Model 3 is negative and statistically significant at the 1% level. The differences in the sign and
significance of the portfolio return coefficient may be due to lending strategies adopted by
funds after deciding to become lenders.
However, since securities loans typically have
durations that vary from a few days to a few weeks and borrower demand for any securities
27
likely varies throughout any year it is difficult to interpret how annual returns are related with
lending agent choice. The estimated coefficients for family size is negative and significant (at the
10% level) in Model 3 but positive, perhaps because larger families are more likely to have an
affiliated custodian or other lending agent. The fund size coefficient in Model 3 is positive and
significant (at the 1% level) which is consistent with larger funds having the resources necessary
to administer their own lending programs. In terms of board structure, Model 3 reports funds
with larger boards and funds with relatively high director pay are less likely to self-lend, results
that are significant at the 1 and 5% levels. Model 3 also reports that self-lending is more likely
when directors own funds they oversee (Family Ownership) and when they have outside
(multiple) directorships, results that are statistically significant at the 1% level.
Similar to Models 1 and 2 and unlike Model 3, the estimated coefficient on portfolio returns
are positive and statistically significant (at the 10% level). Model 4 reports larger funds are less
likely to employ affiliated lending agents. This finding is consistent with the positive coefficient
reported for fund size in Model 3 for self-lending and suggests larger funds have the necessary
human capital to manage securities lending programs. Model 4 reports a statistically lower
likelihood of funds employing affiliated lending agents when directors have more outside
(multiple) appointments and when directors own funds in the family complex. In contrast,
Model 4 reports the estimated coefficient on unexplained director compensation is economically
and statistically significant (at the 1% level) indicating that relatively high paid directors are
more likely to approve affiliated lending agent service contracts.
IV. Conclusion
Recently, a great deal of attention has been given to the management of securities lending
programs and whether conflicts of interests exist when funds employ securities lending agents
that are affiliated with their sponsor. Central to this issue is the question of whether boards of
directors adequately oversee securities lending programs. In this paper, we attempt to shed
light on these issues by examining the relation between affiliated lending agents and securities
lending returns in a hand collected sample of U.S. index mutual funds.
Our analysis suggests that the SEC concerns of self-dealing in some securities lending
programs may be warranted. Specifically, we find that sponsor affiliated lending agents are
28
associated with lower annual returns on lent securities and that securities lending returns are
lower when borrower’s collateral is invested with sponsor affiliated agents. In terms of
governance, we find that excess director compensation is negatively associated with securities
lending returns, while director ownership in the funds they oversee is negatively related to the
incidence of sponsor affiliated lending agents. This suggests that conflicts of interest in the
mutual fund industry are an important consideration in the lending of securities.
We also provide a first look at the characteristics of the mutual fund securities lending
market. We document that about 70% of funds engage in some level of securities lending and
the typical fund lends about 8% of its portfolio. Fund and family TNA are also positively related
to the likelihood a fund having a securities lending program. Funds with institutional
ownership are also more likely to lend, suggesting that fund managers lend to increase fund
performance which in turn attracts institutional cash flows. Finally, we report significant
differences in securities lending returns when the data is segmented by fund investment
objective. Overall, the results have implications for mutual fund boards as they consider
lending proposals from affiliated agents and for possible future regulatory actions.
29
References
Adams, J., S. Mansi, and T. Nishikawa, 2010. Internal governance mechanisms and operational
performance: Evidence from index mutual funds. Review of Financial Studies 23, 12611286.
Bhattacharya, U., J. Lee, and V. Pool, 2013. Conflicting family values in mutual fund families.
Journal of Finance 68,173-200.
Chen, Q., I. Goldstein, and W. Jiang. 2008. Directors' ownership in the U.S. mutual fund
industry. Journal of Finance 63, 2629-2677.
Cohen, L.,K. Diether, and C. Malloy, 2007. Supply and demand shifts in the shorting market.
Journal of Finance 62, 2061-2096.
Cremers M., J. Driessen, P. Maenhout, and D. Weinbaum. 2009. Does skin in the game matter?
Director incentives and governance in the mutual fund industry. Journal of Financial
and Quantitative Analysis 44, 1345-1373.
D'Avolio, G. 2002. The market for borrowing stock. Journal of Financial Economics 66, 271-306.
Diether, K., K. Lee, and I. Werner. 2011. Short-sale strategies and return predictability. Review
of Financial Studies 22, 575-607.
Del Guercio, D., L. Dann, and M. Partch. 2003. Governance and boards of directors in closedend investment companies. Journal of Financial Economics 69, 111-152.
Donald, S. and K. Lang. 2007. Inference with difference-in-differences and other panel data.
Review of Economics and Statistics 89, 221-233.
Duffie, D., N. Gârleanu, and L. Pedersen, 2002, Securities lending, shorting, and pricing. Journal
of Financial Economics 66, 307-339.
Evans, R., C. Geczy, D. Musto, and A. Reed, 2009. Failure is an option: Impediments to short
selling and options prices. Review of Financial Studies 22, 1955-1980.
Evans, R., M. Ferreira, and M. Prado, 2013. Equity lending, investment restrictions and fund
performance. Working Paper, University of Virginia.
Faulkner, M., 2006. An introduction to securities lending., 3rd Edition, Spitalfields Advisors.
Fabbozi, F. 2008. Handbook of Finance, Volume 1, Financial Markets and Instruments. John
Wiley and Sons, Inc.
Fabozzi, F. and S. Mann, 2005. Securities finance: Securities lending and repurchase agreements.
John Wiley and Sons, Inc.
Ferris, S. and X. Yan, 2009, “Agency costs, governance, and organizational forms: Evidence from
the mutual fund industry,” Journal of Banking and Finance 33, 619-626.
30
Geczy, C., D. Musto, and M. Reed, 2002, Stocks are special too: An analysis of the equity lending
market. Journal of Financial Economics 66, 241-269.
Hao, Q. and X. Yan, 2012. The performance of investment bank affiliated mutual funds:
Conflicts of interest or informational advantage? Journal of Financial and Quantitative
Analysis 47, 537-565.
Heckman, J., 1979. Sample selection bias as a specification error. Econometrica 47, 153-161.
Kaplan, S., T. Moskowitz, B. Sensoy, 2011. The effects of stock lending on security prices: An
experiment, Chicago Booth Research Paper No. 09-39.
Khorana, A., P. Tufano and L. Wedge, 2007. Board structure, mergers, and shareholder wealth:
A study of the mutual fund industry. Journal of Financial Economics 85, 571-598.
Kolasinski, A., A. Reed, and M. Ringgenberg, 2013. A multiple lender approach to
understanding supply and search in the equity lending market. Journal of Finance,
forthcoming.
LaBarge, K., 2011. Securities lending: Still no free lunch. Vanguard Research.
Lutton, L., K. Rushkewicz, K. Liu, and X. Ling, 2011. Mutual fund stewardship grade research,
Morningstar, Inc.
Nagel, S., 2005. Short sales, institutional investors and the cross-section of stock returns. Journal
of Financial Economics 78, 277-309
Petersen, M. 2009. Estimating standard errors in finance panel data sets: Comparing
approaches. Review of Financial Studies 22, 435-480.
Sirri, E. and P. Tufano. 1998. Costly search and mutual fund flows. Journal of Finance 53, 15891622.
Tufano, P. and M. Sevick. 1997. Board structure and fee-setting in the U.S. mutual fund
industry. Journal of Financial Economics 46, 321–355.
Wooldrige, J. 2002. Econometric analysis of cross section and panel data. MIT Press, Cambridge.
31
Table I
Variable Definitions
Variable
Securities Lending
Securities Lending Income
Securities Lending Liability
Portfolio Securities on Loan
Return on Lent Securities
Securities Lending Inc. to Assets
Custodian Lending Agent
Self-Lending Agent
3rd Party/Broker Agent
Unknown/Unreported
Affiliated Lending Agent
Collateral Inv. Lending Agent
Fund
Collateral Invested in Unnamed
Pool
Fund Characteristics
Portfolio Return
Jensen’s Alpha
Fund Assets
Family TNA
Expense Ratio
Management Fee
Custodian Fee
Institutional Holding
Enhanced Index Funds
Cash Flow Volatility
Asset Turnover
Definition
Data Source
Annual income from securities lending program as reported in statement of operations
Two year average of borrowers’ collateral as reported in the statements of assets and
liabilities.
Ratio of securities lending liability to fund assets (%)
Ratio of securities lending income to securities lending liability (%)
Ratio of securities lending income to fund assets (%)
Dummy variable that equals 1 when lending agent is the fund’s custodian.
Dummy variable that equals 1 when fund is its own lending agent.
Dummy variable that equal 1 when lending agent is 3rd party or broker
Dummy variable that equals 1 when lending agent is not identified in Form N-CSR
Dummy variable that equals 1 when the lending agent is affiliated with fund sponsor,
excludes self-lending agents
Dummy variable that equals 1 if borrower’s collateral is invested in a fund offered by
lending agent
Dummy variable that equals 1 if borrower’s collateral is invested in an unnamed
collateral reinvestment pool
N-CSR
N-CSR/S
Annual fund return net of expenses (%)
Annualized Jensen’s alpha computed from S&P 500 over 24 months (%)
Log of total assets of fund ($MM) at fiscal year end
Log of total net assets for all funds managed by sponsor ($MM)
Expense ratio of fund (%)
Percentage of fund assets used to pay management and administrative services
Percentage of fund assets used to pay custodian
Percentage of institutional class holdings in fund
Dummy variables that equal 1 when fund attempts to outperform index using leverage
or other techniques or is managed to minimize taxes
Standard deviation of monthly net cash flows computed over the observation year
Trading activity/change in portfolio holdings computed as the lesser of sales or
purchases divided by average monthly total net assets in percentage
32
N-CSR/S
N-CSR/S
N-CSR/S
N-CSR/Form 485
N-CSR/Form 485
N-CSR/Form 485
N-CSR/Form 485
N-CSR/Form 485
N-CSR/S
N-CSR/S
Morningstar
Morningstar
Morningstar
Morningstar
Morningstar
Morningstar
Morningstar
Morningstar
Morningstar
Morningstar
Morningstar
Governance Characteristics
Board Size
Independent Directors
All Independent
Outside Directorships
Family Ownership
Unexplained Compensation
Log of number of directors on fund board
Proportion of directors who are classified as outsiders (independent)
Dummy variable that equal 1 when all directors are independent
Average number of outside/other directorships of directors
Proportion of directors who have ownership in funds in family
Tufano and Sevick’s (1997) measure of relative independent director compensation
Form 485
Form 485
Form 485
Form 485
Form 485
Form 485
Note: This table provides variables definitions for the sample used in the analysis. The data covers the period from 2003 to 2010 for 204 funds
comprising 1,374 fund-year observations. N-CSR/S is the annual/semiannual Certified Shareholder Report and Form 485 contains each fund’s
prospectus and Statement of Additional Information. Both reports can be found in the SEC’s EDGAR database.
33
Table II
Summary Statistics
Panel A: Incidence of Securities Lending by Investment Objective
US Large Blend
US Small Blend
Foreign Large Blend
US Mid-Cap Blend
US Intermediate Bond
US Large Growth
Other
Overall
Observations
(%)
594
157
142
132
90
61
198
1,374
43.23
11.43
10.33
9.61
6.55
4.44
14.41
Funds with
Securities Lending
Income
(%)
384
109
108
111
76
33
136
957
69.65
69.43
76.06
84.09
84.44
54.10
68.69
69.65
Note: Panel A reports the distribution of six investment objectives and categories of the funds in our
sample. The sample consists of 1,374 fund-year observations, representing index mutual funds offered for
sale in the U.S. market, covering the period from 2003 to 2010. The investment and category designations
are obtained from Morningstar. Variable definitions are provided in Table I.
Panel B: Distribution of securities lending
Fund
Assets
(%)
Sponsor
TNA
(%)
Expense
Ratio
(%)
Overall Sample
69.65
69.65
69.65
69.65
Smallest 1
2
3
4
5
6
7
8
9
Largest 10
52.17
59.85
51.82
56.93
70.07
71.74
82.48
78.99
83.82
100.00
37.68
59.71
55.47
74.26
81.02
86.43
66.67
63.57
97.08
86.47
80.88
70.07
84.56
77.37
74.45
83.44
69.64
55.15
59.12
50.00
77.21
63.50
76.81
75.91
66.67
71.53
70.07
68.12
66.42
71.53
Deciles
Portfolio
Return
(%)
Note: Panel B lists the percentage of funds with material securities lending programs in the overall sample
and in each decile. The data covers the period from 2003 to 2010.
34
Panel C: Securities Lending by Agent Type
Obs.
(%)
Portfolio
Securities
on Loan
(%)
Return on
Securities
Lent
(%)
Custodian
Self-Lending Agent
Broker/3rd Party
Unreported/Unknown Agent
496
347
108
6
51.83
36.26
11.29
0.63
8.814
2.921
6.343
11.650
0.268
1.051
0.435
0.138
Affiliated Agent
Nonaffiliated Agent
282
675
29.47
70.53
8.246
5.480
0.348
0.482
Is the Collateral Invested with
Fund of Lending Agent
Unnamed Pool
Fund in Trust Family
Nonaffiliated Fund
350
384
474
31
36.57
40.13
49.53
3.24
8.168
8.854
4.268
4.019
0.372
0.215
0.729
0.912
Note: Panel C reports the median distribution of securities lending agent type and affiliation status of the
funds in the sample. The sample includes 957 fund-year observations with securities lending activity
covering the period from 2003 to 2010. The lending agent type and affiliation status are obtained from the
Certified Shareholder Report (Form N-CSR). Variable definitions are provided in Table I.
35
Panel D: Securities Lending Income by Investment Objective
Obs.
Mean
Standard
Deviation
First
Quartile
Median
Third
Quartile
US Large Blend ($000)
Portfolio Securities on Loan (%)
Return on Lent Securities (%)
Securities Lending Income to TNA (%)
384
1,571
6.327
1.334
0.025
7,009
5.669
4.479
0.067
32
1.843
0.117
0.005
132
5.214
0.218
0.011
497
9.575
0.731
0.027
US Small Blend ($000)
Portfolio Securities on Loan (%)
Return on Lent Securities (%)
Securities Lending Income to TNA (%)
109
1,609
12.031
1.231
0.120
5,173
8.689
1.519
0.132
52
4.785
0.358
0.035
139
10.061
0.775
0.076
528
18.376
1.295
0.146
Foreign Large Blend ($000)
Portfolio Securities on Loan (%)
Return on Lent Securities (%)
Securities Lending Income to TNA (%)
108
1,579
7.019
1.771
0.076
3,117
4.899
2.440
0.074
48
2.749
0.579
0.026
479
6.454
1.056
0.059
1,747
10.557
1.965
0.090
US Mid-Cap Blend ($000)
Portfolio Securities on Loan (%)
Return on Lent Securities (%)
Securities Lending Income to TNA (%)
111
2,058
11.341
1.101
0.082
4,719
7.464
1.584
0.089
54
4.243
0.207
0.022
263
11.309
0.550
0.047
1,082
16.809
1.278
0.121
US Intermediate Bond ($000)
Portfolio Securities on Loan (%)
Return on Lent Securities (%)
Securities Lending Income to TNA (%)
76
487
10.908
0.520
0.035
997
7.195
1.030
0.055
55
4.462
0.128
0.004
167
11.016
0.229
0.017
415
16.628
0.526
0.039
US Large Growth ($000)
Portfolio Securities on Loan (%)
Return on Lent Securities (%)
Securities Lending Income to TNA (%)
33
442
8.444
2.205
0.021
959
9.402
3.371
0.030
15
0.252
0.110
0.004
92
7.524
0.590
0.009
322
18.047
2.985
0.025
Other ($000)
Portfolio Securities on Loan (%)
Return on Lent Securities (%)
Securities Lending Income to TNA (%)
136
2,612
7.962
2.887
0.051
6,629
9.807
5.371
0.071
29
0.602
0.156
0.010
291
3.564
0.898
0.024
1,856
10.764
3.111
0.068
36
Overall ($000)
Portfolio Securities on Loan (%)
Return on Lent Securities (%)
Securities Lending Income to TNA (%)
957
1,656
8.306
1.531
0.053
5,686
7.487
3.769
0.085
40
2.200
0.160
0.008
166
6.526
0.438
0.024
Note: Panel D provides the securities lending income by investment objective. The data covers the period from 2003 through 2010.
definitions are provided in Table I.
37
772
12.719
1.296
0.065
Variable
Table III
Descriptive Statistics (Fund Level Data)
Panel A: Securities Lenders vs Non-Lenders
Funds with Securities Lending Income
Mean
Median
Std. Dev.
Funds w/o Securities Lending Income
Mean
Median
Std. Dev.
Differences
Mean
Median
Fund Characteristics
Assets ($000,000)
Family TNA ($000,000)
Portfolio Return (%)
Jensen’s Alpha (%)
Expense Ratio (%)
Management Fee
Custodian Fees (%)
Enhanced (%)
Asset Turnover (%)
Institutional Ownership (%)
Cash Flow Volatility (%)
4,889
270,170
9.199
0.176
0.478
0.213
0.024
9.269
42.542
36.303
63.859
832
39,971
10.728
-0.710
0.386
0.200
0.010
0.000
16.000
19.555
2.138
14,355
430,364
21.626
4.240
0.396
0.199
0.056
29.000
104.838
39.301
1,768
560
95,364
11.387
-0.997
0.700
0.319
0.038
21.891
78.125
21.876
14.989
232
11,534
13.246
-1.556
0.574
0.250
0.020
0.000
12.000
0.000
2.486
792
252,529
22.306
5.336
0.546
0.255
0.116
41.402
247.887
35.916
94.071
4,329a
174,806a
-2.188c
1.173a
-0.222a
-0.106a
-0.014a
-12.622a
-35.583a
14.427a
48.870
600a
28,437a
-2.518
0.846a
-0.188a
-0.050a
-0.010a
0.000a
4.000c
19.555a
-0.348a
Fund Governance
Board Size
Independent Directors(%)
All Independent
Outside Directorships
Family Ownership (%)
Unexplained Compensation($)
8.378
84.559
13.683
0.859
86.288
6,593
8.000
85.714
0.000
0.875
100.000
-8,257
2.169
8.803
34.385
0.540
23.750
55,872
7.803
80.880
4.975
0.670
80.853
-17,392
8.000
80.000
0.000
0.556
88.889
-17,065
2.368
9.747
21.770
0.594
26.355
62,379
0.575a
3.679a
8.708a
0.189a
5.435a
23,985a
0.000a
5.714a
0.000a
0.319a
11.111a
8,808a
957
957
957
417
417
417
No. of Observations
Note: This panel provides descriptive statistics for the sample segmented by securities lenders and non-lenders. The notations a, b, c denote
statistical significance at the 1%, 5%, and 10% levels, respectively. The data covers the period from 2003 to 2010 for 204 mutual funds. Variables
definitions are provided in Table I.
38
Panel B – Pearson Correlations
% Sec.
on
Loan
%Return on Loan
%
Return
on
Loan
%
Return
to TNA
Fund
Assets
Family
TNA
Port.
Return
Inst.
Own.
Self
Lend.
Agent
Aff.
Lend.
Agent
Indep.
Dir.
Outside
Director
Family
Own.
-0.291
% Return to TNA
0.393
0.092
TNA
-0.138
0.205
-0.027
Family TNA
-0.230
0.346
0.026
0.486
Portfolio Return
-0.025
-0.001
-0.255
0.008
0.036
Institutional Ownership
0.218
-0.029
0.088
-0.011
-0.113
-0.016
Self-Lending Agent
-0.168
0.284
-0.046
0.277
0.511
0.032
-0.197
Aff. Lending Agent
0.137
-0.135
0.068
-0.094
-0.155
0.017
0.077
-0.458
Independent Directors
0.079
0.107
0.140
0.092
0.164
-0.068
-0.005
0.130
0.034
Outside Directorships
-0.118
0.175
0.040
0.168
0.360
0.088
-0.025
0.362
-0.013
0.146
Family Ownership
-0.037
0.155
0.053
0.156
0.322
-0.024
-0.130
0.335
-0.061
0.162
0.129
Unexp. Compensation
0.118
-0.002
0.106
0.005
0.017
0.009
0.024
0.012
0.292
0.244
0.517
0.253
Note: This panel provides Pearson correlations for selected variables used in the analysis. The data covers the period from 2003 to 2010 for 204 mutual funds.
Variables definitions are provided in Table I. Correlations with significance at the 1% level are presented in bold.
39
Table IV
Securities Lending Returns and Agents
Dependent Variable = Return on Securities Loaned
Portfolio Return
Fund Assets
Family TNA
Institutional Ownership
Turnover Ratiox100
Cash Flow Volatilityx1000
Custody Fee Ratio
Primary
Specification
(1)
-0.003
(0.51)
0.179b
(2.12)
0.221a
(2.72)
0.003
(0.98)
0.054
(0.55)
0.019a
(3.35)
Custodian
Fees
(2)
-0.001
(0.16)
0.126
(1.35)
0.257a
(3.09)
0.002
(0.45)
0.019
(0.16)
-0.477
(0.14)
2.451
(1.42)
Self-Lending Agent
Lending
Agent Type
(3)
-0.002
(0.38)
0.072
(0.92)
0.161b
(2.14)
0.005
(1.27)
-0.011
(0.14)
0.021b
(2.55)
Affiliated
Agent
(4)
-0.002
(0.40)
0.141c
(1.80)
0.154c
(1.86)
0.005
(1.51)
0.059
(0.56)
0.014c
(2.76)
Unreported/Unknown Agent
Sponsor Affiliated Agent
-0.557a
(3.14)
-0.679a
(2.76)
-0.475b
(2.57)
Collateral Invested in Lending Agent
Fund
Collateral Invested in Unnamed Pool
Percentage of Portfolio Loaned
-0.081a
(6.24)
40
Affiliated
Agent and
Amount
Loaned
(6)
-0.002
(0.44)
0.095
(1.33)
0.111
(1.45)
0.006c
(1.97)
-0.033
(0.34)
0.001
(0.32)
1.246a
(4.50)
0.133
(0.56)
-0.165
(1.17)
1.300a
(4.11)
0.124
(0.44)
-0.172
(1.07)
3rd Party Agent/Broker
Lending
Agent Type
and Amount
Loaned
(5)
-0.002
(0.44)
0.023
(0.32)
0.109
(1.61)
0.007b
(2.20)
-0.101
(1.17)
0.007
(0.97)
-0.468a
(3.00)
-0.639a
(2.73)
-0.352b
(2.05)
-0.076a
(6.15)
Inv. Objective Dummies
Year Dummies
FYE Month Dummies
Adjusted-R2
Model p-value
Observations
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
0.387
0.00
841
0.423
0.00
722
0.430
0.00
841
0.423
0.00
841
0.475
0.00
841
0.462
0.00
841
Note: This table provides results from regressing the percentage return on securities loaned on various fund characteristics. The data covers the period from 2003
to 2010 for 204 funds. Independent variables include: annual fund return (Portfolio Return), fund total net assets (Fund TNA), investment company total net assets
(Family TNA), percentage of institutional class holdings in the fund (Institutional Ownership), standard deviation of monthly cash flows in and out of the fund
computed over the prior twelve months (Cash Flow Volatility), percentage of fund assets used to pay custodian (Custodian Fees), dummy variables that equals 1
when the lending agent is a custodian, 3rd party/broker, and self-lending agents, dummy variable that equals 1 when the lending agent is affiliated with fund
sponsor, excludes self-lending agents, and a dummy variable that equals 1 when borrowers’ collateral is invested with agent or unnamed pool. All models
include year, FYE month, and investment objective fixed effects (including dummy variables for enhanced, tax managed, or leveraged funds). T-statistics derived
from fund-level clustered robust standard errors are in parentheses. The notation a, b, c denotes statistical significance at the 1%, 5%, and 10%, respectively.
41
Table V
Securities Lending Returns and Risk During the Financial Crisis
Dependent Variable = Return on Securities Loaned (%)
Lending Agent Type
Affiliated Agent
All Other
All Other
Year=2008
Years
Year=2008
Years
(1)
(2)
(3)
(4)
0.005
0.133
0.005
Portfolio Return
0.117
(1.61)
(0.62)
(1.55)
(0.60)
Fund Assets
0.106
0.002
0.219
0.064
(0.42)
(0.02)
(1.09)
(0.86)
0.235
0.097
Family TNA
0.206
0.085
(1.21)
(1.25)
(1.27)
(1.19)
Self-Lending Agent
2.426a
1.075a
(2.79)
(4.60)
0.310
0.079
3rd Party Agent/Broker
(0.51)
(0.35)
Unreported/Unknown Agent
-0.196
-0.078
(0.53)
(-0.50)
Sponsor Affiliated Agent
-0.833b
-0.441a
(2.18)
(2.80)
-0.426b
Collateral Invested in Lending Agent
-1.549a
(2.98)
(2.11)
Fund
-0.267c
Collateral Invested in Unnamed Pool
-1.052c
(1.66)
(1.78)
Percentage of Portfolio Loaned
-0.086b
-0.081a
-0.094b
-0.076a
(2.44)
(6.46)
(2.42)
(6.23)
Inv. Objective Dummies
Y
Y
Y
Y
Year Dummies
N
Y
N
Y
FYE Month Dummies
Y
Y
Y
Y
Adjusted-R2
Model p-value
Observations
0.547
0.00
121
0.441
0.00
720
0.579
0.00
121
0.423
0.00
720
Note: This table provides selected results from regressing the percentage return on securities loaned on various fund
characteristics segmented by the year 2008. The data covers the period from 2003 to 2010 for 204 funds. Independent
variables include: annual fund return (Portfolio Return), fund total net assets (Fund TNA), investment company total
net assets (Family TNA), percentage of institutional class holdings in the fund (Institutional Ownership), standard
deviation of monthly cash flows in and out of the fund computed over the prior twelve months (Cash Flow
Volatility), percentage of fund assets used to pay custodian (Custodian Fees), dummy variables that equals 1 when
the lending agent is a custodian, 3rd party/broker, and self-lending agents, dummy variable that equals 1 when the
lending agent is affiliated with fund sponsor, excludes self-lending agents, and a dummy variable that equals 1
when borrowers’ collateral is invested with agent or an unnamed pool. Some variable coefficients are not reported to
ease exposition. All models include year, FYE month, and investment objective fixed effects (including dummy
variables for enhanced, tax managed, or leveraged funds). T-statistics derived from fund-level clustered robust
standard errors are in parentheses. The notation a, b, c denotes statistical significance at the 1%, 5%, and 10%,
respectively.
42
Table VI
Securities Lending Returns and Agents: Tests of Endogeneity
Public Sponsor
Portfolio Return
Fund TNA
Family TNA
Institutional Ownership
Turnover Ratiox100
Cash Flow Volatility
Self Lending Agent
3rd Party Agent/Broker
Unreported/Unknown Agent
1st Stage
Probit
Dependent
Variable =
Self Lending
Dummy
(1)
-0.486b
(2.01)
-0.002
(0.53)
0.194b
(2.08)
0.212a
(2.59)
-0.832a
(2.62)
-0.001
(0.01)
-0.001
(0.53)
Lending Agent Type
2nd Stage OLS 2nd Stage OLS
Dependent
Dependent
Variable =
Variable =
Return on
Return on
Securities
Securities
Loaned
Loaned
(2)
(3)
0.010c
(1.78)
-0.104
(1.08)
0.226a
(2.67)
0.515c
(1.75)
-0.003c
(1.83)
0.001c
(1.81)
2.344a
(3.56)
0.585
(1.59)
0.054
(0.21)
0.005
(0.94)
-0.146c
(1.70)
0.205a
(2.69)
0.691a
(2.81)
-0.001
(1.20)
-0.001
(0.10)
1.884a
(3.28)
0.390
(1.22)
-0.088
(0.40)
1st Stage
Probit
Dependent
Variable =
Affiliated
Agent
(4)
0.541b
(2.08)
0.002
(0.56)
-0.171c
(1.77)
0.216c
(1.93)
0.444
(1.44)
-0.004
(1.10)
-0.001a
(3.94)
Sponsor Affiliated Agent
Collateral Invested in Lending Agent Fund
0.414c
(1.71)
-0.407c
Collateral Invested in Unnamed Pool
43
Affiliated Agent
2nd Stage OLS 2nd Stage OLS
Dependent
Dependent
Variable =
Variable =
Return on
Return on
Loaned
Loaned
Securities
Securities
(5)
(6)
0.001
(0.09)
-0.052
(0.47)
0.399b
(3.68)
0.572c
(1.68)
-0.001
(0.83)
-0.001
(0.95)
-0.003
(0.39)
-0.060
(0.60)
0.304a
(2.79)
0.686b
(2.29)
-0.001
(0.24)
-0.001b
(2.03)
-2.680b
(2.14)
0.094
(0.34)
-0.637a
-2.197c
(1.91)
-0.184
(0.72)
-0.519b
(1.67)
Percentage of Portfolio Loaned
Inv. Objective Dummies
Year Dummies
FYE Month Dummies
R2
Model p-value
Observations
Y
Y
Y
Y
Y
Y
-0.084a
(5.67)
Y
Y
Y
0.321
0.00
841
0.325
0.00
841
0.405
0.00
814
(2.65)
Y
Y
Y
Y
Y
Y
0.212
0.00
841
0.198
0.00
841
(2.37)
-0.070a
(3.62)
Y
Y
0.296
0.00
841
Note: This table provides results from two stage instrumental variable regressions. The data covers the period from 2003 to 2009 for 226 funds. Independent
variables include: annual fund return (Portfolio Return), fund total net assets (Fund TNA), investment company total net assets (Family TNA), percentage of
institutional class holdings in the fund (Institutional Ownership), standard deviation of monthly cash flows in and out of the fund computed over the prior twelve
months (Cash Flow Volatility), percentage of fund assets used to pay custodian (Custodian Fees), dummy variables that equals 1 when the lending agent is a
custodian, 3rd party/broker, and self-lending agents, dummy variable that equals 1 when the lending agent is affiliated with fund sponsor, excludes self-lending
agents, and a dummy variables that equal 1 when borrowers’ collateral is invested with agent or in an unnamed pool. All models include year, FYE month, and
investment objective fixed effects (including dummy variables for enhanced, tax managed, or leveraged funds). T-statistics derived from fund-level clustered
robust standard errors are in parentheses. The notation a, b, c denotes statistical significance at the 1%, 5%, and 10%, respectively.
44
Table VII
Securities Lending Returns and Boards of Directors
Portfolio Return
Fund TNA
Family TNA
Institutional Ownership
Turnover Ratiox100
Cash Flow Volatility
Percentage of Portfolio Loaned
Independent
Directors
(1)
-0.001
(0.17)
0.063
(0.77)
0.032
(0.22)
0.004
(1.25)
0.001
(0.43)
-0.001
(0.17)
-0.092a
(4.34)
All
Independent
(2)
-0.001
(0.22)
0.061
(0.72)
0.059
(0.42)
0.467
(1.31)
0.001
(0.65)
-0.001
(0.29)
-0.092a
(4.30)
Self-Lending Agent
3rd Party Agent/Broker
Unreported/Unknown Agent
Type/Board
Structure
(3)
-0.001
(0.02)
-0.010
(0.11)
0.052
(0.38)
0.591
(1.63)
-0.001
(0.40)
-0.001
(0.26)
-0.089a
(4.34)
1.138a
(3.54)
0.234
(0.84)
-0.849
(1.90)
Sponsor Affiliated Agent
0.611
(1.16)
2.516c
(1.92)
-0.233
(1.04)
-0.625b
(2.15)
-0.355c
(1.81)
0.535
(1.05)
1.258
(1.00)
0.724a
(3.19)
1.028b
(2.05)
-0.591a
(2.94)
Y
Y
Y
0.368
(1.05)
0.700a
(3.15)
1.091b
(2.07)
-0.602a
(3.07)
Y
Y
Y
0.361c
(1.84)
0.299
(0.73)
-0.352
(1.61)
Y
Y
Y
0.416c
(1.93)
0.733
(1.58)
-0.491b
(2.12)
Y
Y
Y
0.465
0.00
700
0.460
0.00
700
0.480
0.00
700
0.474
0.00
700
Collateral Invested in Lending Agent
Fund
Collateral Invested in Unnamed Pool
Board Size
Independent Directors
0.620
(1.17)
2.364c
(1.68)
All Independent Directors
Outside Directorships
Family Ownership
Unexplained Compensation (per
$100,000)
Inv. Objective Dummies
Year Dummies
FYE Month Dummies
Adjusted-R2
Model p-value
Observations
Aff./Board
Structure
(4)
-0.001
(0.11)
0.067
(0.82)
-0.013
(0.09)
0.005
(1.55)
0.001
(0.68)
0.001
(0.29)
-0.085a
(4.21)
45
0.587
(1.08)
Note: This table provides results from regressing the percentage return on securities loaned on various board, agent,
and fund characteristics. The data covers the period from 2003 to 2010 for 204 funds. Independent variables include:
annual fund return (Portfolio Return), fund total net assets (Fund TNA), investment company total net assets (Family
TNA), percentage of institutional class holdings in the fund (Institutional Ownership), standard deviation of monthly
cash flows in and out of the fund computed over the prior twelve months (Cash Flow Volatility), dummy variables
that equals 1 when the lending agent is a custodian, 3rd party/broker, and self-lending agents, dummy variable that
equals 1 when the lending agent is affiliated with fund sponsor, excludes self-lending agents, dummy variables that
equal 1 when borrowers’ collateral is invested with agent or in unnamed pool, log of number of directors on fund
board (Board Size), proportion of directors who are classified as outsiders (Independent Directors), proportion of
directors with outside board appointments, proportion of directors who have ownership in funds in family (Family
Ownership), Tufano and Sevick’s (1997) measure of relative independent director compensation (Unexplained
Compensation). All models include year, FYE month, and investment objective fixed effects (including dummy
variables for enhanced, tax managed, or leveraged funds).. T-statistics derived from fund-level clustered robust
standard errors are in parentheses. The notation a, b, c denotes statistical significance at the 1%, 5%, and 10%,
respectively.
46
Table VIII
Likelihood of Fund having a Securities Lending Program
Portfolio Returnt-1
Fund TNAt-1
Family TNAt-1
Institutional Ownershipt-1
Cash Flow Volatilityt-1
Asset Turnovert-1
Boards:
Dependent
Variable=1 if
Fund Lends
(2)
0.016c
(2.34)
0.339c
(2.27)
0.700a
(3.01)
0.018
(1.60)
-0.001c
(1.75)
-0.033
(0.37)
-0.566
(0.63)
3.156
(1.36)
0.053
(0.10)
-0.851
(0.79)
-0.643
(1.63)
Self-Lending:
Dependent
Variable=1 if
Fund Self Lends
(3)
-0.018a
(2.74)
0.585a
(2.81)
-0.387c
(1.77)
-0.013c
(1.66)
-0.006
(0.26)
-0.011c
(1.73)
-2.413b
(2.56)
-2.110
(0.70)
1.580a
(3.00)
4.897a
(3.54)
-1.780a
(4.66)
Affiliated
Agent:
Dependent
Variable=1 if
Fund Employs
Affiliated
Agent
(4)
0.016c
(1.79)
-0.431b
(2.32)
0.546c
(1.89)
-0.008
(0.87)
0.005
(1.02)
-0.018b
(2.21)
0.794
(0.72)
-0.715
(0.27)
-3.231a
(4.03)
-2.297c
(1.94)
4.340a
(7.92)
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
0.368
0.00
1,086
0.304
0.00
841
0.421
0.00
614
0.507
0.00
614
Primary
Specification:
Dependent
Variable=1 if
Fund Lends
(1)
0.013b
(2.02)
0.463a
(3.43)
0.432a
(3.22)
0.017a
(2.90)
0.001c
(1.83)
-0.033
(0.36)
Board Size
Independent Directors
Outside Directorships
Family Ownership
Unexplained
Compensation (per
$100,000)
Inv. Objective Dummies
Year Dummies
FYE Month Dummies
Pseudo-R2
Model p-value
Observations
Note: This table presents logit regressions modeling impact of fund characteristics on the likelihood of a fund having
a securities lending program. The data covers the period from 2003 through 2010 for 204 funds. Independent
variables include: annual fund return (Portfolio Return), fund total net assets (Fund TNA), investment company total
net assets (Family TNA), percentage of institutional class holdings in the fund (Institutional Ownership), standard
deviation of monthly cash flows in and out of the fund computed over the prior twelve months (Cash Flow
Volatility), and the percentage of fund holdings that changed over the preceding year (Asset Turnover), log of
number of directors on fund board (Board Size), proportion of directors who are classified as outsiders (Independent
Directors), proportion of directors with outside board appointments, proportion of directors with outside board
appointments, proportion of directors who have ownership in funds in family (Family Ownership), Tufano and
Sevick’s (1997) measure of relative independent director compensation (Unexplained Compensation). All models
include year, FYE month, and investment objective fixed effects (including dummy variables for enhanced, tax
47
managed, or leveraged funds)and are lagged one period where so noted. T-statistics derived from fund-level
clustered robust standard errors are in parentheses. The notation a, b, c denotes statistical significance at the 1%, 5%,
and 10%, respectively.
48
Appendix 1. The Market for Securities Lending
Subsequent to the 1940 Investment Company Act which allowed mutual funds to lend
securities, the SEC developed operating guidelines to regulate securities lending practices of
funds in a series of no-action letters.29,30 These guidelines address board oversight, loan
collateral, termination of the securities loan, returns, fees, voting rights, and loan limits. For
each securities loaned the mutual fund must receive eligible collateral from the borrower at the
time of the loan (e.g., Nagel, 2005). The collateral must be marked to market daily to account
for any increases in the values of the securities loaned and/or decreases collateral value. In
practice, most lending agreements call for 102% collateralization for domestic loans and 105%
collateralization for foreign securities.31 The higher collateral requirement for foreign securities
is intended to protect funds from foreign currency fluctuations.
In addition to the initial
margin, the lending agreement specifies the minimum margin level to be maintained
throughout the life of the loan. The overwhelming majority of securities loans in the U.S. are
facilitated with cash collateral while other forms of collateral (e.g., corporate and government
bonds, equities, letters of credit, warrants, and delivery by value) are more common in the U.K.
and European markets.32
Mutual funds in the U.S. can terminate a securities loan at any time and if the borrower does
not return the securities the mutual fund can use the borrower’s collateral to repurchase the
securities (Evans et al., 2009). Likewise, it is common for securities borrowers in the U.S. to
maintain the right to return securities early. In other jurisdictions, mutual funds can enter into
fixed term lending agreements although loans are usually open (e.g., the maturity dates are
non-binding and either party can terminate at any time) especially for equity loans. How
mutual funds generate returns from securities lending depends on how much securities are on
loan, the supply and demand for lendable securities, lending agent fees, and collateral type
(cash vs. non-cash). Obviously, larger securities lending programs will generate higher
expected lending income. The SEC has interpreted Section 18(f) of the Investment Company
Act of 1940 as a limit on the size and scope of U.S. domiciled mutual funds’ securities lending
programs. Specifically, mutual funds may not lend securities in excess of 1/3 of their TNA.
If a mutual fund receives cash as collateral in exchange for lending securities the lending
fee is computed using a market reinvestment interest rate (typically an overnight rate). A
portion of the reinvestment rate, the “rebate interest rate”, is paid by the lending fund to the
borrower (similar to the repo rate in the bond lending market). If the fund decides to invest the
collateral at the reinvestment rate, the lending fund’s profit is the reinvestment interest earned
less the rebate interest paid to the borrower and lending agent fees. Lending funds can elect to
invest borrowers’ collateral in more risky investments in order to receive higher expected
29 See Dechert on Point, March 2007, Issue 11 for a more complete discussion on the legal environment
surrounding securities lending by mutual funds, and Faulkner (2006) for details of the securities lending market.
30 Securities lending regulations in a number of jurisdictions are similar to those in the U.S. In addition, most
securities lending agreements utilize standardized contracts such as the U.S., Global, and European Master Securities
Lending Agreements.
31 The overcollateralization increased by as much as 20% during the Fall of 2008 (Spitalfields Advisors, 2008.
“The Impact of Recent Regulatory Changes and Economic Events on Beneficial Owners Participating in a Lending
Programme.” Available at www.spitalfieldsadvisors.com/pdfs/Impact_on_beneficial_owners.pdf.)
32 U.S. rules limit securities lending collateral to cash, U.S. Treasuries, and letters of credit.
49
lending returns.33 However, borrowers are entitled to receive the rebate interest based on the
agreed upon reinvestment rate regardless of the actual returns on the invested collateral. U.S.
mutual funds generally invest collateral at the market interest rate or similar low risk
investment vehicle. For securities loans not collateralized with cash borrowers pay lenders an
outright fee based on the length of the loan.
Securities loan fees are a function of borrower demand and lender supply and can vary
dramatically (Cohen, Diether, and Malloy 2007).34 For example, D’Avolio (2002) finds that S&P
500 constituent stocks are in excess supply and lending fees are less than 1% per annum while
stocks with less supply and greater borrower demand (commonly referred to as ‘special’ or
‘hot’) have significantly higher lending fees. When borrowing demand for a securities is high
and the securities is not widely available to lend the laws of supply and demand drive lending
fees up and for cash collateralized loans the rebate rate can become negative (e.g. the borrower
pays the lender additional interest and does not receive a rebate). Finally, funds that employ
lending agents or intermediaries compensate them by either splitting the lending fees (more
common) or by making an outright payment. U.S. regulations require mutual funds to receive
no less than 50% split of gross lending fees. However, it is difficult to observe fee splits in
practice as mutual funds are not required to disclose this information and few do.
Mutual funds and other securities lenders have four general alternatives when it comes to
selecting an agent or principle to facilitate their securities lending programs. They can select
their custodian bank, a third party specialist such as a broker/dealer, lend directly to
borrowers, or some combination of the aforementioned lending choices (e.g. acting as their own
lending agent). Very often, the securities lending agent employed by a fund is affiliated with
the firm that sponsors the fund.
Custodian banks are the most commonly selected lending agent type (about 51% of the
funds in our sample use custodian lending agents) as the custodian provides an easy entry into
the lending market since (i) presumably the mutual fund board has already performed due
diligence on the custodian, (ii) the custodian already has custody of fund assets, (iii) are able to
accommodate small portfolios by pooling their securities with other funds, and (iv) custodians
are in a position to handle all of the operational details associated with lending (Fabozzi, 2008).
In addition, as banks, custodial lending agents may (but not necessarily) provide lenders
indemnities and manage borrowers’ collateral. Given the reduced levels of effort and risk
associated with using custodial lending agents, revenues are usually lower than if the mutual
fund selects another way to access the securities lending market (see Fabozzi and Mann, 2005).
Securities lenders may also opt to appoint a third party specialist agent or principle,
typically a broker/dealer, to facilitate their lending programs. These third party intermediaries
focus on securities lending and do not provide the wide range of services offered by custodial
banks. Third party specialists are generally smaller than custodian agents and are not
constrained by allocation rules that determine how much of each securities a fund can lend
33 We use the term lending return to describe the combined fee and collateral reinvestment income earned by the
mutual fund.
34 Motivations to borrow include the need to cover a short position (i.e. settlement coverage, pairs trading, naked
shorting, market making), a desire by borrowers to lend cash as part of a financing transaction (i.e. prime brokerage
units supporting hedge fund activities may borrow to manage their inventory and balance sheets), and to facilitate
arbitrage opportunities (e.g. tax, trading, and dividend arbitrage).
50
when the supply of lendable funds exceeds demand. As such, third part specialists have the
potential to increase the amount of securities on loan and offer higher lending returns. In
addition, a principal third party intermediary allows lenders to effectively lend to organizations
they would not lend to directly because the borrowers may not be well recognized, regulated,
and/or do not have good credit ratings. An example of principal third party intermediation is a
prime broker who borrows from mutual funds and lends to hedge funds The main
disadvantage to using a principal third party instead of an agent third party or custodian is that
principal lenders do not receive a portion of the earning from lending securities, but instead
benefit by borrowing at the lowest possible cost. Therefore, mutual fund managers must have
the expertise to negotiate lending fees with principal third parties. Figure 1 illustrates one of the
many ways mutual funds can lend securities.
51
Figure 1
Figure 1 illustrates one of many ways a mutual fund can access the securities lending market. A
customer of the broker/dealer sells a particular security short to a long investor. The broker/dealer
locates and borrows securities to facilitate the short sale using the proceeds of their customer’s short sale
to collateralize the borrowed securities. For this example the reinvestment rate is 4.00% and the lending
fund/agent and broker/dealer agree to a 1.00% rebate rate. The lending fund approves the lending
agreement and instructs the lending agent to invest the collateral in the agent’s collateral pool to earn the
reinvestment rate (for the sake of simplicity we assume the lending fund does not choose an alternative
investment vehicle for the collateral). The lending agent and the mutual fund split the net interest profit
of 3.00% (4.00%-1.00%) with at least 50% of the proceeds going to the lending mutual fund. The
broker/dealer uses the borrowed securities to make settlement with the long investor. If the short selling
customer earns 50 basis points on the cash proceeds from the short sale the broker/dealer earns the
remaining 50 basis points. Mutual funds can increase their profits by searching for borrowers and/or
agents who are willing to receive the lowest rebate rates and fee splits. Likewise, the remaining parties to
the security lending arrangement seek the most favorable terms.
52
Appendix Table A.1
Securities Lending Returns
Dependent Variable = Return on Loaned Securities (%)
(1)
(2)
Portfolio Return
-0002
-0.002
(0.38)
(0.39)
0.091
Fund TNA
-0.014
(0.09)
(0.59)
Family TNA
0.144
0.140
(1.53)
(1.34)
0.527
Institutional Ownership
0.516
(1.57)
(1.55)
Turnover Ratiox100
-0.001
0.001
(0.28)
(0.45)
0.001a
Cash Flow Volatility
0.001b
(2.55)
(2.87)
Lambda
—0.539
-0.469
(0.54)
(0.42)
Self Lending Agent
1.316a
(3.63)
0.136
3rd Party Agent/Broker
(0.48)
Unreported/Unknown Agent
--0.205
(1.17)
Sponsor Affiliated Agent?
-0.562a
(2.82)
Collateral Invested in Lending Agent
-0.691b
(2.48)
Fund?
Collateral Invested in Unnamed
-0.473b
Pool?
(2.40)
Inv. Objective Dummies
Y
Y
Year Dummies
Y
Y
Adjusted-R2
Model p-value
Observations
0.429
0.00
719
0.459
0.00
841
Note: This table provides results from regressing the percentage return on loaned securities on various fund
characteristics using the Heckman (1979) approach, where we first model the dummy variable that takes a value of 1
if a fund reports the securities lending income and otherwise zero, using the probit model to calculate the inverse
Mills ratio. The independent variables of this first probit model includes organizational status (public or private),
fund total net assets (Fund TNA), investment company total net assets (Family TNA), investment objective dummy,
and year dummy. The data covers the period from 2003 to 2010 for 204 funds. Independent variables include: annual
fund return (Portfolio Return), fund total net assets (Fund TNA), investment company total net assets (Family TNA),
percentage of institutional class holdings in the fund (Institutional Ownership), dummy variables that take a value of
1 when the fund is an enhanced, leveraged, or tax managed index fund, standard deviation of monthly cash flows in
and out of the fund computed over the prior twelve months (Cash Flow Volatility), dummy variables that equals 1
when the lending agent is a self-lending agent, 3rd party/broker, and unknown, dummy variable that equals 1 when
the lending agent is affiliated with fund sponsor, dummy variables that equal 1 when borrowers’ collateral is
invested with a lending agent fund or an unnamed collateral reinvestment pool, and lambda that is estimated from
the first probit model. T-statistics derived from fund-level clustered robust standard errors are in parentheses. The
notation a, b, c denotes statistical significance at the 1%, 5%, and 10%, respectively.
53
Download