Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Gone with the big data: institutional lender demand for private information
(USC Thesis Other)
Gone with the big data: institutional lender demand for private information
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
i
GONE WITH THE BIG DATA:
INSTITUTIONAL LENDER DEMAND FOR PRIVATE INFORMATION
by
Jung Koo Kang
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BUSINESS ADMINISTRATION)
May 2021
Copyright 2021 Jung Koo Kang
ii
Dedication
This dissertation is dedicated to my parents Sung Ki Kang and Myeong Sun Sim, and to
my wife Go Eun Kim
iii
Acknowledgements
I am extremely grateful to my dissertation chair, Regina Wittenberg-Moerman for her
invaluable advice, continuous support and guidance in the development of this paper. I
am also thankful to other members of my dissertation committee, including Clive
Lennox, Rodney Ramcharan, Lorien Stice-Lawrence, and K.R. Subramanyam. This paper
benefited from the helpful comments and suggestions from Karthik Balakrishnan,
Jonathan Craske, Shane Heitzman, Michael Iselin, Shelley Li, Maria Loumioti, Joshua
Madsen, Peter Oh, Vivek Pandey, Leila Peyravan, K. Ramesh, Mani Sethuraman, Chris
Williams, Forester Wong, T.J. Wong, fellow PhD students, and the workshop participants
at the workshops at University of Southern California, Harvard University, Boston
College, Cornell University, McGill University, Northwestern University, Rice
University, Texas A&M University, University of Connecticut, University of Colorado,
University of Rochester, University of Utah, and Yale University. I am grateful to the
Marshall School of Business at the University of Southern California for financial support.
iv
Table of Contents
Dedication ii
Acknowledgements iii
List of Tables v
List of Figures vi
Abstract vii
Chapter 1: Introduction 1
Chapter 2: Background and Hypothesis Development 9
2.1 Satellite Imagery Data 9
2.2 Related Literature and Hypothesis Development 10
Chapter 3: Sample, Data, and Descriptive Statistics 15
3.1 Data Sources and Sample Selection 15
3.2 Descriptive Statistics 16
Chapter 4: Research Design and Empirical Results 18
4.1 Satellite Image Data and Institutional Lending 18
4.2 Falsification Test 23
4.3 Institutional Lenders’ Information Demand 25
4.3.1 Borrower Opacity 25
4.3.2 Early Dissemination of Borrower Private Information 28
4.4 Accuracy of Satellite Image Data 30
4.5 Institutional Lenders’ Demand for Private Information and Borrowing
Terms 32
Chapter 5: Conclusion 35
References 37
Appendix A: Variable Definitions 42
Appendix B: Additional Analyses 45
v
List of Tables
Table 1. Descriptive Statistics 50
Table 2. Satellite Image Data and Institutional Lending 51
Table 3. Falsification Test 53
Table 4. Borrower Opacity 56
Table 5. Early Dissemination of Borrower Private Information 58
Table 6. Accuracy of Satellite Image Data 60
Table 7. Institutional Lenders’ Demand for Private Information and Borrowing
Terms 62
vi
List of Figure
Figure 1. Parallel Trend of Institutional Lending 49
vii
Abstract
I explore whether the value of borrowers’ private information is an important
determinant of institutional lender participation in syndicated loans. Institutional
lenders have been shown to exploit their access to borrowers’ private information by
trading on it in financial markets. As a shock to these lenders’ private information
advantage, I utilize the release of the satellite image data of car counts in store parking
lots of U.S. retail firms. The satellite data provides accurate and near-real-time signals of
firm performance, which undermines the value of borrowers’ private information
obtained through syndicate participation. I find that once the satellite data becomes
commercially available, institutional lenders are less likely to participate in syndicated
loans. Consistent with institutional lenders’ information-demand channel, the effect of
the satellite data coverage is more pronounced when borrowers are opaque or
disseminate private information to their lenders earlier. The satellite data coverage
further attenuates institutional lending when the data is more accurate in predicting
borrower performance. I also show that institutional lenders’ reduced demand for
private information leads to less favorable loan terms for borrowers. Overall, these
findings suggest that big data sources can crowd out the value of private information
acquired through lending relationships.
1
Chapter 1: Introduction
Over the past two decades, the influx of nonbank institutional lenders as
syndicated loan participants has driven the growth of the large corporate loan market
(Ivashina and Sun 2011a; Jiang et al. 2010; Lim et al. 2014; Peyravan 2019).
1
The
outstanding amounts of syndicated institutional loans increased from $100 billion in 2000
to $1 trillion in 2018 (FDIC 2019). The migration of corporate credit risk to institutional
lenders has been facilitated in part by low interest environments and tighter banking
regulations after the global financial crisis (Irani et al. 2018). Importantly, institutional
lenders are not subject to stringent banking regulations and have been shown to exploit
their private information advantages in the equity, bond and credit derivatives markets
by trading on borrowers’ private information gained through lending relationships (e.g.,
Acharya and Johnson 2007; Bushman et al. 2010; Han and Zhou 2014).
2
Moreover, the
informed trading opportunities that are embedded in the lending relationship generate
economically significant profits (e.g., Ivashina and Sun 2011b; Massoud et al. 2011;
Peyravan 2019).
To shed more light on institutional lending, I examine whether
institutional lenders’ demand for valuable private information is a significant
determinant of their participation in syndicated loans.
1
Institutional lenders typically include investment banks, insurance and finance companies, mutual funds,
pension funds, collateralized loan obligations (CLOs), private equity funds, and hedge funds.
2
Over the course of a loan, borrowers regularly provide their lenders with non-public information, which
includes monthly financial statements, covenant compliance certificates, amendment requests and financial
projections (Standard & Poor’s 2007; Carrizosa and Ryan 2017).
2
The primary empirical challenge in estimating institutional lender demand for
borrowers’ private information is that lenders’ information acquisition is not observable.
To overcome this challenge, I take advantage of the availability of alternative data that
undermines the value of borrowers’ private information. Specifically, I employ satellite
image data from Orbital Insight that track the number of cars in store parking lots for a
subset of U.S. retailers. This data has two important advantages. First, the satellite data
provides valuable information about underlying firm performance (Katona et al. 2018;
Kang et al. 2020b). Second, the data is updated on a daily basis, allowing investors with
access to it to obtain timely updates on firm performance before they become publicly
available. These two unique aspects of the satellite data are important in addressing my
research question as they feature key characteristics of the private information exploited
by institutional lenders.
When institutional lenders can access near-real-time information on a borrower’s
performance through this alternative source of information, the value of early access to
borrowers’ performance information through syndicate participation diminishes.
Moreover, the information advantages of institutional lenders relative to other investors
should decline when other investors can also take advantage of the satellite data, thus
reducing institutional lenders’ expected profits from informed trading (Kyle 1985;
Holden and Subrahmanyam 1992; Foster and Viswanathan 1996; Back et al. 2000; Akins
et al. 2012; Katona et al. 2018). As a result, institutional lenders should have a lower
demand for private information acquired through lending relationships, decreasing their
incentives to extend loans to borrowers covered by the satellite data. Therefore, I predict
3
that the probability of institutional lenders participating in a loan syndicate is lower when
the satellite data on a borrower becomes commercially available.
To isolate the effect of changes in the value of borrowers’ private information, I
employ a difference-in-differences approach that compares the probability of
institutional lenders participating in loans to firms with the satellite data coverage
(“treatment borrowers”) and firms without such coverage (“control borrowers”) before
and after the coverage initiation of the satellite data. It is important to note that firms do
not have control over whether they are covered by these data. I focus on institutional
lenders that are not subsidiaries of bank holding companies.
3
These independent
institutional lenders are subject to lower regulatory scrutiny than those affiliated with
bank holding companies and have weaker internal controls with respect to insider
trading activities (Peyravan and Wittenberg-Moerman 2020). I further focus on
institutional lenders that engage in investment businesses, including investment
banking, asset management, private equity, and hedge-fund management. These
investment businesses provide a platform for institutional lenders to extract benefits from
their timely access to value-relevant information about their borrowers. Consistent with
my prediction, I find that institutional lenders are less likely to issue loans to borrowers
when those borrowers are covered by the satellite data, controlling for borrower and loan
characteristics as well as firm, quarter of loan origination, credit rating, and loan type
fixed effects. Economically, the probability that institutional lenders issue loans to
3
I consider U.S. bank holding companies that are identified by the Financial Stability Board (FSB) as global
systemically important banks (G-SIB) which include JP Morgan, Bank of America, Wells Fargo, Citigroup,
Goldman Sachs, Morgan Stanley, Bank of NY Mellon, and State Street.
4
treatment borrowers relative to that of control borrowers decreases by 10% in the post-
coverage-initiation period.
To assess the validity of the parallel-trend assumption in difference-in-differences
estimation, I demonstrate that the probabilities that institutional lenders issue loans to
treatment and control borrowers are not statistically different in the pre-period when the
satellite data is not commercially available. To mitigate concerns that the results may be
affected by other confounding factors, I also control for differences in observable
characteristics across treatment and control borrowers using an entropy balancing
approach. Using this matching technique, I find consistent evidence that the satellite data
coverage decreases institutional lending.
Next, I perform falsification tests using other types of institutional lenders that, in
contrast to independent institutional lenders that engage in investment businesses, are
unlikely to exploit early access to borrowers’ private information for insider trading
purposes. First, institutional lenders affiliated with bank holding companies face higher
regulatory costs than independent institutional lenders and typically have controls in
place to prevent the transfer of sensitive borrower information from loan officers to
traders in other investment divisions (Carey et al. 1998; Peyravan 2019; Kang et al. 2020a).
These lenders therefore are less likely to trade on borrowers’ private information. Second,
independent institutional lenders will have limited demand for private information when
they do not have a trading operation and thus cannot exploit their information
advantages. Consistent with limited information demand of bank-affiliated-institutional
lenders and independent institutional lenders without investment businesses, I fail to
5
find evidence that these lenders have a lower probability of issuing loans to borrowers
covered by the satellite data.
To reinforce institutional lenders’ information demand mechanism, I perform a
number of cross-sectional tests. Although the information demand should incentivize
institutional investors to participate in syndicated loans, I argue that such participation
becomes less valuable to institutional lenders following the coverage initiation of the
satellite data that substitutes, at least partially, for borrowers’ private information.
Therefore, I expect the effect of the satellite data coverage on institutional lenders’
participation to be more pronounced if these lenders had a higher demand for borrowers'
private information in the pre-coverage period. First, I conjecture that institutional
lenders should have a higher demand for private information when borrowers are
opaque. Opaque borrowers provide imprecise public information that encourages
private information acquisition and informed trading (Diamond 1985; Bushman 1991;
Kim and Verrecchia 1991).
4
Second, I expect institutional lenders to exhibit a higher
information demand when borrowers disseminate private information to their lenders
earlier because timely information is more valuable to the lenders’ trading activities.
5
Consistent with these predictions, I show that the satellite data coverage reduces
institutional lending to a greater extent when borrowers are opaque or disseminate
private information to their lenders earlier.
4
I measure a borrower’s information opacity based on its analyst coverage, issuance of earnings forecasts,
and press releases.
5
I measure early dissemination of borrower information based on whether a borrower issues loans with a
higher number of performance (income-statement based) covenants or obtains loans from reputable lead
arrangers (Bushman et al. 2010, Bushman and Wittenberg ‐Moerman 2012; Christensen and Nikolaev 2012).
6
I next investigate whether the satellite data coverage has a greater effect when the
data is more accurate in predicting borrower performance. Increased precision of an
alternative source of information can further crowd out the value of private information
acquired through lending relationships. Therefore, when the satellite data provides more
precise forecasts of borrower performance, I expect institutional lenders to have a lower
demand for private information through loan participation. I indeed find that the effect
of the satellite data coverage on institutional lending is stronger for borrowers for which
the satellite data is more accurate.
6
Lastly, I examine whether institutional lenders’ lower information demand affects
borrowers’ credit outcomes. I find that when institutional lenders stop funding loans to
borrowers in the post-coverage-initiation period, these borrowers pay higher interest
rates, obtain smaller loan amounts, and issue loans with shorter maturities. Such
unfavorable loan terms are consistent with the lower information demand leading to a
decrease in credit supply for borrowers covered by the satellite data, suggesting that
institutional lenders’ information demand is an important factor that shapes loan
contractual terms.
This paper makes several contributions. First, I contribute to the growing literature
on nonbank institutional lending. Recent studies document that institutional lenders
trade on borrower information that they obtain from their lending relationships (Ivashina
and Sun 2011b; Massoud et al. 2011; Peyravan 2019). Moreover, institutional lenders
6
I measure the accuracy of the satellite data using high correlation between car count signals and firm
performance or lower variability of car count signals across the firm’s stores.
7
accelerate the speed of stock-price discovery, especially for borrowers with weak public
disclosure (Bushman et al. 2010), and stimulate greater borrower voluntary disclosure
(Peyravan and Wittenberg-Moerman 2020).
7
While these studies primarily explore the
consequences of institutional lender participation on capital markets, I demonstrate that
the demand for valuable private information is an important factor for institutional
lenders’ decisions to participate in syndicated loans. Relatedly, I contribute to the
literature on the effect of institutional lender participation on loan pricing by providing
evidence on institutional lenders’ lower information demand adversely affecting loan
terms for borrowers (Jiang et al. 2010; Ivashina and Sun 2011a; Lim et al. 2014).
Second, I contribute to the emerging literature on the role of alternative big data
sources in capital markets. Prior studies find that big data is a useful supplementary
source of information that affects price informativeness, managerial actions, and
information asymmetries among investors (Jame et al. 2016; Katona et al. 2018; Zhu 2019;
Kang et al. 2020b). However, there is little work on the effect of big data sources on credit
markets. I demonstrate that the availability of these data sources undermines information
demand of institutional lenders, reducing their supply of credit.
Finally, I also contribute to the nascent literature on the importance of noncredit
sources of income in private lending. Prior studies show that relationship lenders are
7
Peyravan (2019), who primarily focuses on the insider trading activities of dual holders (institutional
investors that simultaneously holds a firm’s loan and equity), also finds that these investors are more likely
to invest in equities of borrowers with weak financial reporting quality. While these findings imply that
institutional lenders tend to pursue opaque borrowers, my study directly examines whether institutional
lenders’ demand for private information is an important determinant of their syndicate lending
participation by utilizing the satellite data coverage as a shock to these lenders’ information advantage.
8
more likely to obtain mandates for their borrowers’ security underwritings and M&A
deals (e.g., Drucker and Puri 2005; Yasuda 2005). These cross-sold products typically
generate substantial income and can enhance the profitability of lending relationships.
Therefore, banks take borrowers’ cross-selling potentials into account when initiating
new lending relationships (Kang et al. 2020c). I complement these studies by showing
that the potential trading benefits embedded in the lending relationship can significantly
influence institutional lenders’ incentives to retain these relationships.
The next section presents the hypothesis development. Section 3 describes data
and sample selection. Section 4 reports main results, and Section 5 concludes.
9
Chapter 2: Background and Hypothesis Development
2.1 Satellite Imagery Data
Satellite images are photos of Earth’s surface collected by remote sensing satellites
operated by government programs or commercial entities. Satellite images have detailed
and high-spatial resolution; therefore, they are huge in size and mostly in an unstructured
format, which is often referred to as “big data.” Recent advancements in machine learning
and cloud computing techniques have made it feasible to parse out vast quantities of
satellite images across the globe and extract useful information from them each day,
which enables investors to “explore the world in real-time.” Investors can receive real-
time updates on various economic activities measured based on satellite images of store
parking lots, manufacturing centers, oil refineries, petrochemical plants, agricultural
land, and mining operations among others. These data help, for example, gauge a
country's fuel supply, predict crop yields, estimate damages from natural disasters, and
track flows and disruptions along supply chains.
In this paper, I employ satellite image data provided by Orbital Insight that tracks
the number of cars in parking lots for a subset of publicly listed U.S. firms. Orbital Insight
was founded in 2013 and the car count data became commercially available to its clients
in the third quarter of 2015. At the end of each day, Orbital Insight collects satellite images
from its various providers including Landsat (a joint program by NASA and U.S.
Geological Survey), DigitalGlobe, Airbus, and Planet Labs. Once these satellite images
are gathered, Orbital Insight counts the number of cars in each parking lot using a
proprietary computer vision and machine learning algorithm which includes procedures
10
to enhance accuracy of the car count data. For example, if multiple stores share the same
parking lot, the algorithm identifies the area of the parking lots in front of each store’s
entrance and records the number of cars specific to the store. In this case, the algorithm
provides information on a contamination level of each store’s car count data based on the
probability of inaccurately counting the number of cars. Moreover, the car count data is
adjusted based on the time stamp of the satellite images to ensure comparability over
time. Satellite images taken outside of operating hours and for stores with covered
parking lots are excluded. Orbital Insight provides the car count data to its clients the
following morning.
2.2 Related Literature and Hypothesis Development
Over the past two decades, nonbank lenders have played an increasing role in
supplying credit to corporations. In the syndicated loan market, nonbank lenders have
grown their share from 40% in 2000 to 60% in 2014 (Peyravan 2019). Low interest
environments and tighter banking regulations after the global financial crisis facilitated
the migration of corporate credit risk to nonbank lenders (Irani et al. 2018). Recent studies
examine characteristics of nonbank loans and show that they have higher interest
spreads, have more flexible covenants, and are more likely to be secured while nonbank
borrowers are smaller, less profitable, and have few financing alternatives (Lim et al.
2014; Chernenko et al. 2019; Loumioti 2019).
11
Nonbank lenders include a growing number of institutional lenders who also
engage in investment businesses in financial markets (e.g., investment banks, investment
managers, hedge funds, private equity funds, and collateralized loan obligation
managers). As syndicate participants, these lenders have access to borrowers’
performance information before it is publicly disclosed to market participants. Over the
course of a loan, borrowers typically provide information to lenders on a monthly basis,
including financial performance updates, covenant compliance reports, amendment
requests and financial projections, and allow lenders to visit their sites (Standard & Poor’s
2007; Carrizosa and Ryan 2017, Gustafson et al. 2020). Given the non-public nature of this
information, lenders are prohibited by U.S. laws from trading on material borrower
private information.
8
Despite these regulations, prior studies show that institutional
lenders exploit their information advantages by engaging in insider trading in the equity,
bond, and credit derivatives markets (Acharya and Johnson 2007; Bushman et al. 2010;
Ivashina and Sun 2011b; Massoud et al. 2011; Han and Zhou 2014; Peyravan 2019). In
addition, insider trading generates economically significant profits. For example,
institutional lenders can make abnormal profits of around $5 million by short-selling
borrowers’ stocks during the 20-day window around negative credit events (Massoud et
al. 2011) or achieve 5% to 8% excess annual returns by trading borrowers’ stocks (Ivashina
and Sun 2011b; Peyravan 2019).
8
The Securities Exchange Act of 1934 and the Insider Trading Sanctions Act of 1984 are two federal laws
that regulate insider trading.
12
I examine whether the value of borrowers’ private information is an important
determinant for the institutional lenders’ incentives to have lending relationships with
borrowers. Empirical evidence on this topic is limited because lenders’ acquisition of
private information is not directly observable. Recent studies use Freedom of Information
Act (FOIA) requests to identify private information acquisition. Glaeser et al. (2020) show
that information asymmetry between managers and outsiders promotes private
information acquisition measured by FOIA requests submitted to the U.S. Securities and
Exchange Commission. Down et al. (2020) find that lead arrangers strategically withhold
negative private information from their participant lenders when lead arrangers are more
informed based on prior FOIA requests submitted to the Food and Drug Administration.
FOIA requests reveal private information because such information is unknown to
outside individuals and is costly to obtain as these requests are fulfilled with a
considerable delay. However, institutional lenders value information available without a
time lag because timely information is critical to their instantaneous trading activities. To
overcome this issue, I take advantage of the satellite image data from Orbital Insight that
provides daily updates on the number of cars in store parking lots. This data has two
important advantages. First, the car count data provides valuable information by
accurately predicting firm performance (Katona et al. 2018; Kang et al. 2020b). Second,
the data is updated daily; therefore, investors can purchase the data to obtain timely
updates on firm performances before they become publicly available. These two unique
aspects of the data comprise the key characteristics of the private information exploited
by informed traders in financial markets.
13
When institutional lenders can access the satellite data that can substitute for
borrower private information, the value of private information acquired through lending
relationships declines. Moreover, when other investors can also acquire the satellite data
and trade on timely information about firm performance, institutional lenders expect
greater competition in financial markets. The competition among informed investors
reduces expected profits from their informed trading and discourages private
information acquisition (e.g., Holden and Subrahmanyam 1992; Foster and Viswanathan
1993, 1996; Back et al 2000; Akin et al. 2012). As a result, institutional lenders should have
a lower demand to acquire private information by extending loans to borrowers covered
by the satellite data. Building on these arguments, I predict that the probability that
institutional lenders participate in syndicated loans is lower after the satellite data on a
borrower becomes commercially available.
However, there are a number of factors that may confound this prediction. First,
factors other than the value of private information can dominate institutional lenders’
incentives to have lending relationships with borrowers. For example, prior studies
suggest that institutional lenders pursue syndicated loans because these loans offer high
interest rates (Lim et al. 2014). Second, the satellite data may be less informative than
what institutional lenders can directly learn through syndicate relationships. Third, the
satellite data may complement rather than substitute the private information of
institutional lenders. For example, the satellite data may help institutional lenders to
better interpret and trade on the private information about borrower performance (Kim
and Verrecchia 1994; McNichols and Trueman 1994). Fourth, the costs associated with
14
both acquiring and processing the satellite data can be prohibitive to investors (e.g.,
Blankespoor et al. 2020). Therefore, institutional lenders may continue to demand early
access to borrower information through their lending relationships. For these reasons,
whether the availability of the satellite data attenuates institutional lending remains an
open question.
15
Chapter 3: Sample, Data, and Descriptive Statistics
3.1 Data Sources and Sample Selection
I obtain loan characteristics from DealScan and borrower characteristics from
Compustat and CRSP. I collect analyst coverage data from I/B/E/S, press release data
from RavenPack, and borrower credit ratings from Compustat and Mergent FISD.
Satellite data coverage and store-level car count data are from Orbital Insight. I select
loans issued by U.S. borrowers in the same industries as borrowers covered by the
satellite data, resulting in 6,907 loan packages over the 2011 and 2019 period. I eliminate
borrowers with missing Compustat identifiers, resulting in 2,684 loan packages. I match
this sample to Compustat and further eliminate loans with insufficient borrower and loan
characteristics. The final sample contains 98 treatment borrowers with the satellite data
coverage and 546 control borrowers without the data coverage, corresponding to 2,129
loan packages syndicated by 677 lenders.
To identify institutional lenders, I first classify lenders as either commercial bank
lenders or nonbank lenders. Following Lim et al. (2014), I identify a lender as a
commercial bank lender if its lender type in DealScan is “US Bank,” “Foreign Bank,”
“Thrift/S&L,” “African bank,” “Asian-Pacific Bank,” “Eastern Europe/Russian Bank,”
“Middle Eastern Bank,” or “Western European Bank.” I also classify a lender as a
commercial bank lender if a lender’s SIC 4-digit code is between 6011 and 6082, or 6712,
6719. For each lender identified as commercial bank, I manually check whether the lender
mainly engages in commercial banking business and exclude lenders if they do not
16
mainly accept deposits and extend individual or business loans.
9
I classify all remaining
lenders as nonbank lenders.
Next, I further classify nonbank lenders as bank-affiliated-institutional lenders
and independent institutional lenders based on lenders’ business description from the
company website, annual report, Bloomberg, or Capital IQ. I classify a lender as a bank-
affiliated-institutional lender if it is a subsidiary of the U.S. bank holding company. For
each independent institutional lender that is not affiliated with banks, I next identify
whether it engages in investment businesses based on DealScan lender type. An
independent institutional lender is considered to have an investment operation if its
lender type in DealScan is “Inst. Invest. Prime Fd,” “Inst. Invest. Prime Hedge Fd,” “Inst.
Invest. Prime CDO,” “Investment bank,” “Mutual Fund,” “Distressed (Vulture) Fund.”
In addition, I check each lender’s business description to determine whether the lender
engages in investment businesses, including investment banking, asset management,
private equity, and hedge fund management. Finally, I classify remaining lenders as
independent institutional lenders that do not engage in investment businesses. These
lenders include captive finance companies, lease companies, and farm credit institutions.
3.2 Descriptive Statistics
Table 1 presents descriptive statistics of the main variables used in the analyses. A
total of 16.5% of the loans in the sample are issued with at least one independent
9
For example. I exclude from commercial bank lenders Goldman Sachs Group, ORIX USA Corp and
Pilgrim Group.
17
institutional lender engaging in investment businesses (Inst. Lender).
10
A total of 33.2% of
loans are issued after the third quarter of 2015 when the satellite data becomes
commercially available (Post). Treatment borrowers issue 19.4% of sample loans
(Treatment Firm). My sample borrowers are relatively large (Assets) and have an average
leverage ratio (Leverage) of 0.332. The mean market-to-book ratio (MTB) is 3.23, the mean
sales growth (Sales Growth) is 0.148, and the mean interest coverage ratio (Interest
Coverage) is 65.5.
11
The sample borrowers have an average return on asset (ROA) of 0.027,
an average Altman Z-score (Altman Z) of 3.546, an average age (Age) of 24 years, and
show 5.5% of stock return before the loan issuance (Past Return). With respect to loan
characteristics, the mean value of loan size (Amount) suggests that sample loans are
relatively large (USD $3.2 billion), have an average maturity (Maturity) of approximately
four years, and have an average all-in-drawn spread of 195 bps (Interest Spread). Around
49% of loans are secured (Secured) and 9% of them have a guarantor (Guarantor). Detailed
variable definitions are reported in Appendix A.
10
I note that 24.8% and 5.9% of the loans in the sample are issued with at least one bank-affiliated-
institutional lender (Inst. Lender Bank Affiliated) and independent institutional lender without investment
operations (Inst. Lender No Investment).
11
The median value of Interest Coverage is 8.4. Main results are robust to winsorizing it at 95% level.
18
Chapter 4: Research Design and Empirical Results
I organize my empirical analyses as follows. First, I examine the effect of the
satellite data coverage on institutional lender participation in syndicated loans. Next, I
further explore the information demand channel by investigating whether the observed
effect is stronger when institutional lenders are expected to have a higher demand for
borrowers’ private information in the pre-coverage period. Lastly, I examine whether
institutional lenders’ information demand affects borrowers’ credit outcomes.
4.1 Satellite Image Data and Institutional Lending
I begin my analyses by investigating whether the probability that institutional
lenders issue a loan is lower after the satellite data on a borrower becomes commercially
available. My empirical strategy exploits the fact that a subset of U.S. borrowers is
covered by the satellite data after the third quarter of 2015. I employ a difference-in-
differences analysis using control borrowers in the same industries (SIC 4-digit) as the
treatment borrowers covered by the data.
12
Specifically, I estimate the following model:
Inst. Lender = β0 + β1Treatment Firm × Post + Controls + Fixed Effects + ε, (1)
In Model (1), the dependent variable (Inst. Lender) is equal to one if the loan is
issued with at least one institutional lender that is not affiliated with bank holding
12
Main results are robust to using SIC 3-digit industries.
19
companies and is engaged in investment businesses (institutional lender hereafter), and
zero otherwise. The variable of interest is Treatment Firm × Post, where Treatment Firm is
equal to one if the borrower is covered by the satellite data during the post-coverage-
initiation period (and zero otherwise) and Post is equal to one if the loan is issued after
the satellite data becomes commercially available (and zero otherwise). If institutional
lenders have a lower probability of participation in loans to borrowers after those
borrowers are tracked by the satellite data, I expect a negative and significant coefficient
on Treatment Firm × Post.
I control for borrower characteristics that can influence institutional lending
decisions, which include a borrower’s size (Assets), liquidity (Current Ratio), leverage
(Leverage), market price (MTB), sales growth (Sales Growth), interest coverage (Interest
Coverage), profitability (ROA), credit risk (Altman Z), age (Age), and stock performance
(Past Return). I also control for loan characteristics, including loan amount (Amount),
maturity (Maturity), and whether a loan is secured (Secured) or has a guarantor
(Guarantor). I include firm fixed effects to control for unobservable time-invariant
characteristics of each firm. I also include year fixed effects to control for time-varying
factors common to all sample firms.
13
I estimate the Model (1) using both a logit and OLS
model. In the OLS model, I substitute year fixed effects with year-quarter fixed effects
13
It is important to note that I include year fixed effects using indicator variables for trailing 12 months
ending in September of each year to ensure that Post does not have within-year variances. Thus, the
coefficients on both Treatment Firm and Post are not estimated because they are perfectly collinear with year
and firm fixed effects.
20
and further include credit rating and loan type fixed effects.
14
I cluster standard errors at
the firm level.
I present my main findings in Table 2. First, I report the results of univariate tests
in Panel A. I find that when investors can access the satellite data (Post = 1), 8.7% of loans
are issued with institutional lenders for treatment borrowers, compared to 19.4% of loans
for control borrowers. In contrast, when the satellite data is not available to investors
(Post = 0), 16.0% of loans are issued with institutional lenders for treatment borrowers,
compared to 16.3% of loans for control borrowers. I find that the difference-in-differences
((8.7% - 19.4%) – (16.0% - 16.3%)) is statistically significant at the 1% level. This evidence
is consistent with my prediction that institutional lenders are less likely to issue loans to
borrowers after these borrowers become covered by the satellite data.
Next, I present estimation results of Model (1) in Panel B of Table 2. In column (1)
(columns (2) and (3)), I employ a logit model (OLS models). I find a negative and
significant coefficient on Treatment Firm x Post for all specifications.
15
Economically,
relative to the control borrower, the probability of institutional lenders issuing a loan to
the treatment borrower is lower by 13.8% after the coverage-initiation of the satellite data.
I measure economic significance based on the OLS specification in column (3), where I
14
Due to issues regarding a large number of fixed effects in nonlinear models (e.g., Maddalla 1987; Greene
2004), I include only year and firm fixed effects in the logit model.
15
With respect to controls, the negative and significant coefficients on ROA and Past Return suggest that
borrowers with higher profitability or higher prior buy and hold return are less likely to obtain loans from
institutional lenders. The negative and significant coefficient on Altman Z indicates that borrowers with
higher credit risk attract institutional lenders. Institutional lenders are also more likely to participate in
secured loans (Secured).
21
include firm, year-quarter, credit rating, and loan type fixed effects. These findings
reinforce my prediction that the probability of institutional lenders issuing a loan is lower
for borrowers tracked by satellite data.
16
The key identifying assumption of the difference-in-differences analysis is the
parallel-trend assumption that institutional-lending trends would be the same for both
treatment and control borrowers in the absence of the satellite data coverage. In other
words, it assumes that control borrowers provide the appropriate counterfactual of the
trend that treatment borrowers would have followed if they had not been covered by the
satellite data (Angrist and Pischke 2008). To examine whether the parallel-trend
assumption holds, I estimate the following model:
Inst. Lender = β0 + β1Treatment Firm × Trend t=-3,-4 + β2Treatment Firm × Trend t=1,2 +
β3Treatment Firm × Trend t=3,4 + Controls + Fixed Effects + ε, (2)
In Model (2), I replace Treatment Firm x Post variable in Model (1) with separate
interaction variables between Treatment Firm and trend variables. Each trend variable is
equal to one for every two-year sample period before and after the initiation of the
satellite data coverage (and zero otherwise). I exclude from the trend variables the last
16
I employ several alternative specifications to ensure that my results are not sensitive to research design
choices. First, I re-estimate Model (1) using continuous dependent variables of the proportion (%) or the
number of institutional lenders in the loan package and find robust results (reported in Panel A of
Appendix B). Also, a subset of treatment borrowers in my sample are covered by the satellite data provided
by RS Metrics. Main results are robust to using Post RM as the main variable of interest, which is equal to
one if the loan is issued after the satellite data from either RS Metrics or Orbital Insight become
commercially available, and zero otherwise (results are tabulated in Panel B of Appendix B).
22
two-year period immediately prior to the release of the satellite data (from 4
th
quarter in
2013 to 3
rd
quarter in 2015); therefore, this period serves as a benchmark period. In Figure
1, I graphically depict the estimation results of Model (2). I find that the counterfactual
treatment effect in the pre-coverage period (i.e., the coefficient on Treatment Firm ×
Trendt=-3,-4) is statistically indistinguishable from the benchmark period while treatment
effects in the post-coverage-initiation period (i.e., the coefficients on Treatment Firm ×
Trendt=1,2 and Treatment Firm × Trendt=3,4) are significantly different from the benchmark
period. These results provide support for the parallel-trend assumption.
17
While firms cannot self-select to be covered by the satellite data, I recognize other
factors that may confound my results. For example, if treatment borrowers and control
borrowers differ in many dimensions, satellite data coverage may be endogenous with
respect to these differences. In Panel C of Appendix B, I compare firm characteristics
between the treatment and control borrowers. Treatment borrowers are more profitable
and older while they exhibit lower sales growth and have lower credit risk relative to the
control borrowers. Although I control for time invariant firm characteristics by including
firm fixed effects in all the analyses, to further alleviate this concern, I employ the entropy
balancing approach. This matching technique achieves covariate balance between
treatment and control observations by re-weighting control observations, which ensures
that the mean and the variance are identical along the matching variables for both
17
I also plot the probability of institutional lending for treatment and control borrowers separately during
the sample period (untabulated). I visually check that these univariate trends do not indicate a violation of
the pre-trend assumption.
23
treatment and control samples. Moreover, the entropy balancing reduces bias from
nonlinear relationships between observable characteristics and the dependent variable
(Hainmueller 2012; McMullin and Schonberger 2020). In Panel C, Table 2, I present the
estimation results using the entropy balanced sample and continue to find a negative and
significant coefficient on Treatment Firm x Post, consistent with the satellite data coverage
curbing institutional lending.
4.2 Falsification Test
In this section, I perform falsification tests to provide additional support for
institutional lenders’ information demand mechanism. I suggest that when the satellite
data provides accurate and near-real-time signals on firm performance, institutional
lenders have a lower demand to acquire borrower private information for insider trading
purposes. Therefore, if the information demand is instrumental to the relationship
between the satellite data coverage and institutional lending, my main results should not
hold or be much weaker for other types of lenders that are unlikely to exploit early access
to borrowers’ private information by engaging in informed trading.
Institutional lenders affiliated with bank holding companies are less likely to trade
on borrower information obtained through lending relationships because these lenders
are subject to stringent banking regulations and face higher regulatory oversight (Carey
et al. 1998; Peyravan 2019). Moreover, bank-affiliated lenders tend to be larger
organizations and have controls in place to prevent the transfer of sensitive borrower
24
information from loan officers to traders in other investment divisions who may exploit
it (Carey et al. 1998; Peyravan 2019; Kang et al. 2020a; Peyravan and Wittenberg-
Moerman 2020). Therefore, I focus on loans issued with institutional lenders that are
subsidiaries of bank holding companies. Inst. Lender Bank Affiliated is equal to one if the
loan is issued with at least one bank-affiliated-institutional lender but is not issued with
independent institutional lenders (and zero otherwise).
To exploit private information advantages for potential insider trading, lenders
need to engage in investment businesses which can provide a platform to extract benefits
using value-relevant information about their borrowers. Using loans issued with
independent institutional lenders, I further identify those loans issued with lenders that
do not engage in the investment businesses. Inst. Lender No Investment is equal to one if
the loan is issued with at least one independent institutional lender not engaging in
investment businesses but is not issued with independent institutional lenders that
engage in investment businesses (and zero otherwise).
I perform the falsification test by re-estimating Model (1) with each of these variables
as the dependent variable. I report the results in Panel A, Table 3. Consistent with my
prediction, I failed to find a significant coefficient on Treatment Firm x Post across all
specifications where either Inst. Lender Bank Affiliated or Inst. Lender No Investment is the
dependent variable. Next, I re-estimate Model (1) using a multinomial logit model. For
this analysis, I create a dependent variable that takes the value of one if Inst. Lender Bank
Affiliated is equal to 1, two if Inst. Lender No Investment is equal to 1 and, three if Inst. Lender
25
is equal to 1, and zero otherwise. As reported in Panel B, Table 3, I failed to find significant
coefficients on Treatment Firm x Post when the dependent variable takes the value of one
or two, which suggests that the satellite data coverage does not affect loans issued with
bank-affiliated-institutional lenders or independent institutional lenders without
investment operations. These results are consistent with these lender types having low
demand for borrowers’ private information.
4.3 Institutional Lenders’ Information Demand
To further support the information demand mechanism, I investigate whether the
effect of the satellite data coverage on institutional lending is more pronounced if
institutional lenders had a higher demand for borrowers’ private information in the pre-
coverage period. While higher information demand can stimulate institutional lenders to
participate in syndicated loans, this participation becomes less valuable in the post-
coverage-initiation period when the satellite data substitutes, at least partially, for
borrowers’ private information. Thus, I predict the effect of the satellite data coverage on
institutional lending to be stronger for borrowers that attracted higher institutional
lenders’ information demand prior to the initiation of the data coverage.
4.3.1 Borrower Opacity
I perform several analyses that exploit cross-sectional variances in institutional
26
lenders’ information demand based on borrower characteristics in the period before the
satellite data becomes available. First, I examine whether the effect of the satellite data
coverage on institutional lending is more pronounced for opaque borrowers. Opaque
borrowers provide imprecise public signals; therefore, traders have more heterogeneous
beliefs about them, which encourages private information acquisition and informed
trading (Verrecchia 1982; Diamond 1985; Bushman 1991). Moreover, when a borrower is
opaque, lenders are endowed with greater information advantages, which increase the
value of borrowers’ private information. Therefore, institutional lenders should have a
higher information demand for opaque borrowers when alternative information sources
are not available.
To measure a borrower’s information opacity, I begin with a borrower’s analyst
coverage. As an important information intermediary, financial analysts actively engage
in private information production and provide accurate and timely information about
firm performance to investors (Brown et al. 1987; Fried and Givoly 1982; Healy and
Palepu 2001). Moreover, increased analyst following reduces the likelihood of insider
trades and discourages insider purchases (Frankel and Li 2004). No Analyst Coverage is
equal to one if the borrower does not have equity analyst coverage in the pre-coverage
period (and zero otherwise).
As another measure of borrower opacity, I consider a borrower’s disclosure
choices – decisions to issue earnings forecasts and press releases. Public disclosures may
preclude costly private information acquisition (Diamond 1985, Verrecchia 2001) and are
27
important determinants for a firm’s information opacity (e.g., Beyer et al. 2010). No
Earnings Forecast is equal to one if the borrower does not issue earnings forecasts in the
pre-coverage period (and zero otherwise). Low Press Releases is equal to one if the average
number of press releases by the borrower, measured in the pre-coverage period, is less
than the sample median (and zero otherwise).
Using these borrower opacity variables, I assign loans to the high and low opacity
partitions and re-estimate Model (1). In Panel A of Table 4, I find a negative and
significant coefficient on Treatment Firm x Post in the low analyst partition (i.e., No Analyst
Coverage = 1). Importantly, I show that the magnitude of the coefficient on Treatment Firm
x Post is statistically higher in the low analyst partition than in the high analyst partition.
Next, in Panel B of Table 4, the coefficients on Treatment Firm x Post are negative and
significant using both low disclosure partition (i.e., No Earnings Forecast = 1) and high
disclosure partition (i.e., No Earnings Forecast = 0). However, the magnitude of the
coefficient is statistically higher for the low disclosure partition, consistent with non-
guider borrowers attracting higher information demand. Lastly, as reported in Panel C
of Table 3, I find that the coefficient on Treatment Firm x Post is negative and significant
in the low press release partition (i.e., Low Press Releases = 1) and its magnitude is
statistically higher than the magnitude in the high press release partition (Low Press
Releases = 0). Economically, using the low press release partition, the probability that an
institutional lender issues a loan to the treatment borrower is lower by 18.7% relative to
the control borrower after the coverage-initiation of the satellite data. Overall, these
results suggest that satellite data coverage attenuates institutional lending to a greater
28
extent when borrowers are opaque.
4.3.2 Early Dissemination of Borrower Private Information
To strengthen the information demand mechanism, I next perform additional cross-
sectional tests to determine whether the satellite data coverage has a stronger effect when
borrowers disseminate private information to their lenders earlier. Prior studies suggest
that timely access to borrower information facilitates informed trading by incumbent
lenders (Bushman et al. 2010; Carrizosa and Ryan 2017). Because timely information is
more valuable for instantaneous trading activities, I expect institutional lenders to have
a higher information demand when borrowers disseminate their information to lenders
on a timely basis.
I first measure early dissemination of borrower information based on whether a
borrower issues a higher number of performance covenants (Bushman et al 2010,
Christensen and Nikolaev 2012; Christensen et al. 2016; Carrizosa and Ryan 2017).
18
Performance covenants are based on earnings and cash flow metrics and are frequently
set tightly relative to the underlying performance variables. Moreover, these covenants
often obligate borrowers to provide information about their current performances to
lenders more frequently. Therefore, performance covenants enable lenders to monitor
borrowers efficiently, which accelerates timely acquisition of private information about
18
Following Christensen and Nikolaev (2012), I classify cash interest coverage ratio, debt service coverage
ratio, level of EBITDA, fixed charge coverage ratio, interest coverage ratio, ratio of debt to EBITDA, and
ratio of senior debt to EBITDA covenants as performance covenants
29
borrowers (Bushman et al. 2010; Carrizosa and Ryan 2017). High Perf. Covenants is equal
to one if the average number of performance covenants for loans issued by the borrower,
measured in the pre-coverage period, is greater than the sample median (and zero
otherwise).
Next, I use lender reputation as another measure of timely dissemination of
borrower information to lenders. The reputation of a lead arranger reflects its expertise
and commitment to monitor borrowers (e.g., Diamond 1989; Boot et al. 1993; Chemmanur
and Fulghieri 1994). Reputable lead arrangers have been documented to collect greater
private information about borrowers and communicate it earlier to syndicate participants
(Bushman et al. 2010; Bushman and Wittenberg-Moerman 2012).
19
Therefore, I expect that
institutional lenders have a higher information demand when they participated in loans
syndicated by reputable lead arrangers in the pre-coverage period. High Reputation is
equal to one if the borrower obtains loans issued with one of the top five lead arrangers
in the pre-coverage period (and zero otherwise).
I partition sample observations based on these measures of timely dissemination
of borrower information and re-estimate Model (1). As I report in Panel A of Table 5, the
coefficient on Treatment Firm x Post is significant in the high covenant partition (i.e., High
Perf. Covenants = 1) but not in the low covenant partition (i.e., High Perf. Covenants = 0). I
also show that the coefficient on Treatment Firm x Post indicates statistically higher
magnitude in the high covenant partition than in the low covenant partition. Further, in
19
Also, reputable lenders experience higher reputational losses if they withhold important private
information about borrowers from participants (Down et al. 2020).
30
Pane B of Table 5, I show a negative and significant coefficient on Treatment Firm x Post
in the high reputation partition (i.e., High Reputation = 1) but not in the low reputation
partition (i.e., High Reputation = 0). In addition, the coefficient on Treatment Firm x Post
has a significantly higher magnitude in the high reputation partition. Economically, using
the high reputation partition, relative to the control borrower, the probability of an
institutional lender issuing a loan to the treatment borrower is lower by 29.4% after the
coverage-initiation of the satellite data. Taken together, these findings suggest that the
effect of satellite data coverage on institutional lending is more pronounced when the
flow of borrowers’ private information to lenders is faster, which further supports
institutional lenders’ information demand channel.
4.4 Accuracy of Satellite Image Data
In this section, I investigate whether the satellite data coverage has a greater effect
on institutional lending when the data is more accurate in predicting borrowers’
performance. When alternative sources of information provide signals with higher
precision, traders can generate higher profits from their informed trading (Grossman and
Stiglitz 1980; McNichols and Trueman 1994). Therefore, more precise satellite data can
further crowd out the value of private information acquired through lending
relationships, leading to the reduction in institutional lender participation.
I measure the accuracy of the satellite data for each borrower using its store-level
car counts. A borrower has more accurate satellite data when the correlation between its
31
car count signals and firm performance is higher or the variability of its car count signals
across stores is lower. Treatment Firm High Corr (Treatment Firm Low Corr) is equal to one
if the average correlation between quarterly changes in car counts and quarterly changes
in sales of the borrower is greater (lower) than the sample median, and zero otherwise.
Treatment Firm High SD (Treatment Firm Low SD) is equal to one if the average standard
deviation of quarterly changes in car counts across stores is greater (lower) than the
sample median, and zero otherwise. Using each of these car count accuracy variables, I
estimate the following model:
Inst. Lender = β0 + β1Treatment Firm High Accuracy × Post +
β2Treatment Firm Low Accuracy × Post + Controls + Fixed Effects + ε, (3)
In this model, I replace Treatment Firm x Post variable in Model (1) with separate
interactions between Post and high (or low) accuracy of car count variables.
20
In Panel A
of Table 6, I present results of the analysis using Treatment Firm High Corr (and Treatment
Firm Low Corr). I find a negative and significant coefficient on Treatment Firm High Corr x
Post but do not find a significant coefficient on Treatment Firm Low Corr x Post, and the
difference between these two coefficients is statistically significant. I also find similar
results using Treatment Firm High SD (and Treatment Firm Low SD). As I report in Panel B
of Table 6, the coefficient on Treatment Firm Low SD x Post is negative and significant
20
Accuracy of the car count signals cannot be measured for control borrowers whose car count data is not
available.
32
across all specifications, and its magnitude is significantly higher than the coefficient on
Treatment Firm High SD in OLS specifications. Economically, when the car count signal
exhibits lower variability, the probability that an institutional lender issues a loan to the
treatment borrower is lower by 18.4% after the coverage-initiation of the satellite data.
Overall, these results suggest that institutional lenders have a lower demand to acquire
private information by extending loans to borrowers when more precise satellite data
further crowds out the value of borrowers’ private information. These findings not only
further support the information demand mechanism but also provide evidence on the
validity of the satellite data coverage as a proxy for the value of borrowers’ private
information.
4.5 Institutional Lenders’ Demand for Private Information and Borrowing Terms
Thus far, I provide robust evidence that the information demand for borrowers’
private information is an important factor for institutional lenders’ decisions to issue
loans to borrowers. In the last set of analysis, I explore whether the information demand
of institutional lenders influences borrowers’ credit outcomes. When the satellite data
coverage reduces credit supply from institutional lenders, borrowers may obtain less
favorable loan terms (e.g., Ivashina and Sun 2011a; Lim et al. 2014). On the other hand,
the satellite data provides useful information about borrower performance, which may
help mitigate adverse selection concerns of syndicate participants and incentivize them
to supply more credit (Bushman et al. 2016; Kang et al. 2020c). In this case, borrowers
33
may obtain more favorable loan terms despite the lower information demand of
institutional lenders after the initiation of the satellite data coverage. To investigate this
question, I estimate the following OLS model:
Loan Term = β0 + β1Treatment Firm × Post No Inst. Lender × Had Inst. Lender +
β2Treatment Firm × Post Inst. Lender × Had Inst. Lender + Main Effects +
Lower Order Interactions +Controls + Fixed Effects + ε, (4)
Where the dependent variable of Loan Term is one of the following three borrowing
terms: the natural logarithm of the all-in-drawn spread (Interest Spread), the natural
logarithm of loan amounts (Amounts), and the natural logarithm of the loan maturity in
months (Maturity).
21
To estimate the effect of the reduction in institutional lenders’
participation, I identify the following loans issued in the post-coverage-initiation period:
loans issued without institutional lenders participation (Post No Inst. Lender), loans issued
with institutional lender participation (Post Inst. Lender), and loans issued to a borrower
who had institutional lender participation in its loans issued before the satellite data
became available (Had Inst. Lender). The main variable of interest is the triple interaction
term - Treatment Firm × Post No Inst. Lender × Had Inst. Lender. This variable measures
loans issued by treatment borrowers (Treatment Firm = 1) who do not obtain loans from
institutional lenders in the post period (Post No Inst. Lender = 1) but had lending
21
I include all main effects and lower order interactions of each triple interaction variable in Model (4) but
do not specify them for brevity.
34
relationships with institutional lenders in the pre-coverage period (Had Inst. Lender = 1),
which indicates that these borrowers experience a reduction in information demand from
institutional lenders.
As I report in Table 7, I find a positive (negative) and significant coefficient on
Treatment Firm × Post No Inst. Lender × Had Inst. Lender in the Interest Spread (Amounts or
Maturity) specification. The results indicate that when institutional lenders stop issuing
loans to borrowers following the initiation of the satellite data coverage, borrowers pay
higher interest rates, obtain smaller loan amounts, and issue loans with shorter
maturities.
22
These unfavorable loan terms are consistent with the reduced information
demand leading to lower credit supplies from institutional lenders. Overall, these
findings suggest that institutional lender demand for borrowers’ private information is
an important factor that can influence the contract outcomes of syndicated loans.
22
I restrict sample to the treatment borrowers and re-estimate Model (4) and continue to find a positive
(negative) and significant coefficient on Post No Inst. Lender x Had Inst. Lender in Interest Spread (Amounts
or Maturity) specification (reported in Panel D of Appendix B).
35
Chapter 5: Conclusion
I show that the value of borrowers’ private information is a significant determinant
for institutional lenders’ participation in syndicated loans. As a shock to institutional
lenders’ private information advantages, I utilize the release of the satellite image data of
the car counts in store parking lots of U.S. retail firms. I predict that accurate and near-
real-time information on borrower performance through the satellite data diminishes the
value of borrowers’ private information; therefore, institutional lenders have a lower
demand for the private information obtained through lending relationships. Consistent
with my prediction, I find that institutional lenders are less likely to participate in loan
syndicates when the satellite data on a borrower becomes commercially available.
Supporting the information demand argument, I further show that the satellite data
coverage further attenuates institutional lending when borrowers are opaque,
disseminate private information to their lenders earlier, or when the satellite data
provides more accurate forecasts of borrower performance. Lastly, I find that the lower
information demand of institutional lenders leads to unfavorable credit outcomes for
borrowers.
My study is not without limitations. My sample is restricted to retail firms because
satellite images of store parking lots are only available for those firms. Although I believe
that information demand of institutional investors is an important determinant of their
participation in loan syndicates, I caution against generalizing my results to firms in other
industries. I leave it for future research to explore whether the information demand
36
significantly influences institutional lending and credit outcomes for non-retailer
borrowers. In addition, future research can further identify other sources of big data and
examine how institutional lenders’ information demand varies with unique features of
these data.
37
References
Acharya, V., and T. Johnson. 2007. Insider trading in credit derivatives. Journal of Financial
Economics 84 (1):110-141.
Addoum, J., and J. Murfin. 2020. Equity price discovery with informed private debt. The
Review of Financial Studies 33 (8):3766-3803.
Akins, B., J. Ng, and R. Verdi. 2012. Investor competition over information and the pricing
of information asymmetry. The Accounting Review 87 (1):35-58.
Angrist, J., and J. Pischke. 2008. Mostly harmless econometrics: An empiricist's
companion: Princeton University Press.
Back, K., C. Cao, and G. Willard. 2000. Imperfect competition among informed traders.
The Journal of Finance 55 (5):2117-2155.
Beyer, A., D. Cohen, T. Lys, and B. Walther. 2010. The financial reporting environment:
Review of the recent literature. Journal of Accounting and Economics 50 (2-3):296-343.
Blankespoor, E., E. deHaan, and I. Marinovic. 2020. Disclosure processing costs,
investors’ information choice, and equity market outcomes: A review. Journal of
Accounting and Economics Forthcoming.
Boot, A., S. Greenbaum, and A. Thakor. 1993. Reputation and discretion in financial
contracting. The American Economic Review:1165-1183.
Brown, L., R. Hagerman, P. Griffin, and M. Zmijewski. 1987. Security analyst superiority
relative to univariate time-series models in forecasting quarterly earnings. Journal of
Accounting and Economics 9 (1):61-87.
Bushman, R. 1991. Public disclosure and the structure of private information markets.
Journal of Accounting Research 29 (2):261-276.
Bushman, R., A. J. Smith, and R. Wittenberg ‐Moerman. 2010. Price discovery and
dissemination of private information by loan syndicate participants. Journal of
Accounting Research 48 (5):921-972.
Bushman, R., and R. Wittenberg ‐Moerman. 2012. The role of bank reputation in
“certifying” future performance implications of borrowers’ accounting numbers.
Journal of Accounting Research 50 (4):883-930.
Bushman, R., C. Williams, and R. Wittenberg ‐Moerman. 2016. The Informational Role of
the Media in Private Lending. Journal of Accounting Research 55 (1): 115-152.
Carey, M., M. Post, and S. Sharpe. 1998. Does corporate lending by banks and finance
companies differ? Evidence on specialization in private debt contracting. The Journal
of Finance 53 (3):845-878.
38
Carrizosa, R., and S. Ryan. 2017. Borrower private information covenants and loan
contract monitoring. Journal of Accounting and Economics 64 (2-3):313-339.
Chemmanur, T., and P. Fulghieri. 1994. Investment bank reputation, information
production, and financial intermediation. The Journal of Finance 49 (1):57-79.
Chernenko, S., I. Erel, and R. Prilmeier. 2019. Nonbank lending. Working paper.
Christensen, H., and V. Nikolaev. 2012. Capital versus performance covenants in debt
contracts. Journal of Accounting Research 50 (1):75-116.
Christensen, H., V. Nikolaev, and R. Wittenberg-Moerman. 2016. Accounting
information in financial contracting: The incomplete contract theory perspective.
Journal of Accounting Research 54 (2):397-435.
Down, A., C. Williams, and R. Wittenberg-Moerman. 2020. Strategic syndication: Is bad
news shared in loan syndicates. Working Paper.
Diamond, D. 1984. Financial intermediation and delegated monitoring. The Review of
Economic Studies 51 (3):393-414.
Diamond, D. 1985. Optimal release of information by firms. The Journal of Finance 40
(4):1071-1094.
Diamond, D. 1989. Reputation acquisition in debt markets. Journal of Political Economy 97
(4):828-862.
Diamond, D. 1991. Monitoring and reputation: The choice between bank loans and
directly placed debt. Journal of Political Economy 99 (4):689-721.
Drucker, S., and M. Puri. 2005. On the benefits of concurrent lending and underwriting.
The Journal of Finance 60 (6):2763-2799.
Fama, E. 1985. What's different about banks? Journal of Monetary Economics 15 (1):29-39.
FDIC. 2019. Bank and nonbank lending over the past 70 years. FDIC Quarterly:31-39
Foster, F., and S. Viswanathan. 1993. The effect of public information and competition on
trading volume and price volatility. The Review of Financial Studies 6 (1):23-56.
Foster, F., and S. Viswanathan. 1996. Strategic trading when agents forecast the forecasts
of others. The Journal of Finance 51 (4):1437-1478.
Frankel, R., and X. Li. 2004. Characteristics of a firm's information environment and the
information asymmetry between insiders and outsiders. Journal of Accounting and
Economics 37 (2):229-259.
Fried, D., and D. Givoly. 1982. Financial analysts' forecasts of earnings: A better surrogate
for market expectations. Journal of Accounting and Economics 4 (2):85-107.
39
Glaeser, S., B. Schonberger, C. Wasley, and J. Xiao. 2020. Private information acquisition
via freedom of information act requests. Working Paper.
Greene, W. (2004). The behaviour of the maximum likelihood estimator of limited
dependent variable models in the presence of fixed effects. The Econometrics Journal,
7 (1), 98-119.
Grossman, S., and J. Stiglitz. 1980. On the impossibility of informationally efficient
markets. The American Economic Review 70 (3):393-408.
Gustafson, M., I. Ivanov, and R. Meisenzahl. 2020. Bank monitoring: Evidence from
syndicated loans. Working paper.
Hainmueller, J. 2012. Entropy balancing for causal effects: A multivariate reweighting
method to produce balanced samples in observational studies. Political Analysis 20
(1):25-46.
Han, S., and X. Zhou. 2014. Informed bond trading, corporate yield spreads, and
corporate default prediction. Management Science 60 (3):675-694.
Healy, P., and K. Palepu. 2001. Information asymmetry, corporate disclosure, and the
capital markets: A review of the empirical disclosure literature. Journal of Accounting
and Economics 31 (1-3):405-440.
Holden, C., and A. Subrahmanyam. 1992. Long ‐lived private information and imperfect
competition. The Journal of Finance 47 (1):247-270.
Irani, R., R. Iyer, R. Meisenzahl, and J.-Peydró. 2018. The rise of shadow banking:
evidence from capital regulation. Working Paper.
Ivashina, V., and Z. Sun. 2011a. Institutional demand pressure and the cost of corporate
loans. Journal of Financial Economics 99 (3):500-522.
Ivashina, V., and Z. Sun. 2011b. Institutional stock trading on loan market information.
Journal of Financial Economics 100 (2):284-303.
Jame, R., R. Johnston, S. Markov, and M. Wolfe. 2016. The value of crowdsourced
earnings forecasts. Journal of Accounting Research 54 (4):1077-1110.
Jiang, W., K. Li, and P. Shao. 2010. When shareholders are creditors: Effects of the
simultaneous holding of equity and debt by non-commercial banking institutions.
The Review of Financial Studies 23 (10):3595-3637.
Kang, J., C. Lennox, and V. Pandey. 2020a. Client concerns about information spillovers
from sharing audit partners. Working Paper.
Kang, J., L. Stice-Lawrence, and Y. Wong. 2020b. The firm next door: Using satellite
images to study local information advantage. Working Paper.
40
Kang, J., C. Williams, and R. Wittenberg-Moerman. 2020c. CDS trading and
nonrelationship lending dynamics. Review of Accounting Studies. Forthcoming.
Katona, Z., M. Painter, P. Patatoukas, and J. Zeng. 2018. On the capital market
consequences of alternative data: Evidence from outer space. Working Paper.
Kim, O., and R. Verrecchia. 1991. Trading volume and price reactions to public
announcements. Journal of Accounting Research 29 (2):302-321.
Kim, O., and R. E. Verrecchia. 1994. Market liquidity and volume around earnings
announcements. Journal of Accounting and Economics 17 (1-2):41-67.
Kyle, A. 1985. Continuous auctions and insider trading. Econometrica: Journal of the
Econometric Society:1315-1335.
Leland, H., and D. Pyle. 1977. Informational asymmetries, financial structure, and
financial intermediation. The Journal of Finance 32 (2):371-387.
Lim, J., B. Minton, and M. Weisbach. 2014. Syndicated loan spreads and the composition
of the syndicate. Journal of Financial Economics 111 (1):45-69.
Loumioti, M. 2019. Direct Lending: The determinants, characteristics and performance of
direct loans. Working Paper.
Maddala, G. 1987. Limited dependent variable models using panel data. Journal of Human
Resources, 22 (3), 307-338.
Massoud, N., D. Nandy, A. Saunders, and K. Song. 2011. Do hedge funds trade on private
information? Evidence from syndicated lending and short-selling. Journal of
Financial Economics 99 (3):477-499.
McMullin, J., and B. Schonberger. 2020. Entropy-balanced accruals. Review of Accounting
Studies:1-36.
McNichols, M., and B. Trueman. 1994. Public disclosure, private information collection,
and short-term trading. Journal of Accounting and Economics 17 (1-2):69-94.
Peyravan, L. 2019. Financial reporting quality and dual-holding of debt and equity. The
Accounting Review. Forthcoming.
Peyravan, L., and R. Wittenberg-Moerman. 2020. Institutional dual-holders and
managers’ earnings disclosure. Working paper.
Rajan, R. 1992. Insiders and outsiders: The choice between informed and arm's ‐length
debt. The Journal of Finance 47 (4):1367-1400.
Standard & Poor’s. 2007. A guide to the loan market. Standard & Poor’s. New York.
Verrecchia, R. 1982. Information acquisition in a noisy rational expectations economy.
Econometrica: Journal of the Econometric Society:1415-1430.
41
Verrecchia, R. 2001. Essays on disclosure. Journal of Accounting and Economics 32 (1-3):97-
180.
Yasuda, A. 2005. Do bank relationships affect the firm's underwriter choice in the
corporate ‐bond underwriting market? The Journal of Finance 60 (3):1259-1292.
Zhu, C. 2019. Big data as a governance mechanism. The Review of Financial Studies 32
(5):2021-2061.
42
APPENDIX A
Variable Definitions
Variable Definition
Age = The number of years since a firm appears in the Compustat
(Compustat).
Altman Z = Altman (1963) Z-score as estimated by the following model: Z=
3.3𝑋 1
+ 0.99 𝑋 2
+ 0.6 𝑋 3
+ 1.2 𝑋 4
+ 1.4 𝑋 5
, where 𝑋 1
is defined as
the ratio of earnings before interest and taxes to total assets, 𝑋 2
is defined as the ratio of total sales to total assets, 𝑋 3
is defined
as the ratio of market value of equity to total liabilities, 𝑋 4
is
defined as the ratio of current assets to total assets, 𝑋 5
is
defined as the ratio of retained earnings to total assets. All
variables are measured in the year preceding the loan’s
issuance (Compustat).
Amounts = The natural logarithm of loan amounts of the largest facility in
the loan package (DealScan).
Assets = The natural logarithm of total assets, measured in the year
preceding the loan’s issuance (Compustat).
Current Ratio = The ratio of current assets to current liabilities, measured in the
year preceding the loan’s issuance (Compustat).
Guarantor = An indicator variable equal to 1 if the loan is guaranteed, and
0 otherwise (DealScan).
Had Inst. Lender = An indicator variable equal to 1 if the loan is issued after the
satellite image data becomes commercially available and the
borrower had lending relationships with institutional lenders
(Inst. Lender) when the satellite image data was not
commercially available, and 0 otherwise (Orbital Insight).
High Perf. Covenants = An indicator variable equal to 1 if the average number of
performance covenants for loans issued by the borrower,
measured in the pre-period before the satellite image data
becomes commercially available, is greater than the sample
median, and 0 otherwise (DealScan).
High Reputation = An indicator variable equal to 1 if the borrower obtains loans
issued by one of the top 5 lead arrangers in the pre-period
before the satellite image data becomes commercially
available, and 0 otherwise (DealScan, Bloomberg).
Inst. Lender Bank
Affiliated
= An indicator variable equal to 1 if the loan is issued with at
least one bank-affiliated-institutional lender but is not issued
with independent institutional lenders, and 0 otherwise
(DealScan).
Inst. Lender = An indicator variable equal to 1 if the loan is issued with at
least one independent institutional lender that engages in
investment businesses, and 0 otherwise (DealScan).
43
APPENDIX A (continued)
Variable Definitions
Variable Definition
Inst. Lender No Investment = An indicator variable equal to 1 if the loan is issued with at
least one independent-institutional lender that does not engage
in investment businesses but is not issued with an independent
institutional lender that engages in investment businesses, and
0 otherwise (DealScan).
Interest Coverage = The ratio of earnings before interest and taxes to interest
expense, measured in the year preceding the loan’s issuance
(Compustat).
Interest Spread = The natural logarithm of the all-in-drawn spread of the largest
facility in the package (DealScan).
Leverage = The ratio of long-term debt plus debt in current liabilities to
total assets, measured in the year preceding the loan’s issuance
(Compustat).
Low Press Releases = An indicator variable equal to 1 if the average number of press
releases by the borrower, measured in the pre-period before
the satellite image data becomes commercially available, is
greater than the sample median, and 0 otherwise (RavenPack).
Maturity = The natural logarithm of the loan maturity in months
(DealScan).
MTB = The ratio of market value to book value of equity, measured in
the year preceding the loan’s issuance (Compustat).
No Analyst Coverage = An indicator variable equal to 1 if the borrower does not have
equity analyst coverage, measured in the pre-period before the
satellite image data becomes commercially available, and 0
otherwise (IBES).
No Earnings Forecast = An indicator variable equal to 1 if the borrower does not issue
earnings forecasts, measured in the pre-period before the
satellite image data becomes commercially available, and 0
otherwise (IBES).
Past Return = The accumulated daily stock return measured over 150
calendar days ending 30 days prior to the loan's issuance
(DealScan, CRSP).
Post = An indicator variable equal to 1 if the loan is issued after the
satellite image data becomes commercially available, and 0
otherwise (Orbital Insight).
Post Inst. Lender = An indicator variable equal to 1 if the loan is issued with
institutional lenders (Inst. Lender) after the satellite image data
becomes commercially available, and 0 otherwise (Orbital
Insight).
Post No Inst. Lender = An indicator variable equal to 1 if the loan is not issued with
institutional lenders (Inst. Lender) after the satellite image data
becomes commercially available, and 0 otherwise (Orbital
Insight).
44
APPENDIX A (continued)
Variable Definitions
Variable Definition
ROA = The ratio of net income to total assets, measured in the year
preceding the loan’s issuance (Compustat).
Sales Growth = The ratio of total sales in year t to total sales in year t-1 minus
one, measured in the year preceding the loan’s issuance
(Compustat).
Secured = An indicator variable equal to 1 if the loan is secured, and 0
otherwise (DealScan).
Treatment Firm = An indicator variable equal to 1 if the borrower is covered by
the satellite data after it becomes commercially available
(Orbital Insight).
Treatment Firm High Corr = An indicator variable equal to 1 if the average correlation
between quarterly changes in store level car counts and
quarterly changes in sales for the borrower, measured in the
post-period after the satellite image data becomes
commercially available, is greater than the sample median, and
0 otherwise (Orbital Insight).
Treatment Firm Low Corr = An indicator variable equal to 1 if the average correlation
between quarterly changes in store level car counts and
quarterly changes in sales for the borrower, measured in the
post-period after the satellite image data becomes
commercially available, is less than the sample median, and 0
otherwise (Orbital Insight).
Treatment Firm High SD = An indicator variable equal to 1 if the average standard
deviation of quarterly changes in car counts across stores,
measured in the post-period after the satellite image data
becomes commercially available, is greater than the sample
median, and 0 otherwise (Orbital Insight).
Treatment Firm Low SD = An indicator variable equal to 1 if the average standard
deviation of quarterly changes in car counts across stores,
measured in the post-period after the satellite image data
becomes commercially available, is less than the sample
median, and 0 otherwise (Orbital Insight).
Maturity = The natural logarithm of the loan maturity in months
(DealScan).
MTB = The ratio of market value to book value of equity, measured in
the year preceding the loan’s issuance (Compustat).
No Analyst Coverage = An indicator variable equal to 1 if the borrower does not have
equity analyst coverage, measured in the pre-period before the
satellite image data becomes commercially available, and 0
otherwise (IBES).
45
APPENDIX B
Additional Analyses
This table reports the results of additional analyses. Panel A examines whether the main results
are robust to using continuous dependent variables that capture the extent of institutional lender
participation. Column (1) estimates Tobit model using the dependent variable of Inst. Lender
Proportion which is the proportion (%) of institutional lenders in the loan package. Column (2)
estimates Poisson model using the dependent variable of Inst. Lender Counts which is the number
of institutional lenders in the loan package. Panel B reports the result of analysis whether the
main results are robust to using Post RM which is an indicator variable equal to 1 if the loan is
issued after the satellite image data from either RS Metrics or Orbital Insight become
commercially available, and 0 otherwise. Panel C compares the mean and standard deviation of
the explanatory variables between the treatment and control firms to provide evidence of
covariate balancing in the estimation using an entropy balancing approach. Panel D examines
whether the reduced information demand from institutional lenders affects borrowers’ credit
outcomes using the treatment sample. t-statistics in parentheses are based on standard errors
clustered at the firm level. ***, **, and * indicate significance at the 0.01, 0.05, and 0.10 levels,
respectively. All other variables are defined in Appendix A of the manuscript.
Panel A: Continuous Dependent Variables
Inst. Lender Proportion Inst. Lender Counts
(1) (2)
Treatment Firm x Post -25.589*** -0.772*
(-3.05) (-1.95)
Assets 0.415 -0.035
(0.25) (-0.22)
Current Ratio 0.806 0.166
(0.40) (1.22)
Leverage -8.005 0.033
(-1.05) (0.08)
MTB 0.287 0.007
(1.50) (0.78)
Sales Growth 3.151 0.330
(0.68) (1.57)
Interest Coverage -0.015* -0.001*
(-1.85) (-1.82)
ROA -45.120*** -0.862
(-3.14) (-1.12)
Altman Z -0.203 -0.041
(-0.26) (-0.84)
AGE -0.005 -0.009
(-0.03) (-0.49)
46
APPENDIX B (continued)
Additional Analyses
Panel A: Continuous Dependent Variables (continued)
Inst. Lender Proportion Inst. Lender Counts
(1) (2)
Past Return -9.600 -0.351
(-1.56) (-1.17)
Amounts 1.664 0.109
(0.77) (1.07)
Maturity 15.954* 1.009**
(1.91) (2.12)
Secured 13.197*** 0.483**
(3.36) (2.09)
Guarantor 3.138 0.345
(0.53) (0.89)
Model Tobit Poisson
Firm FE Yes Yes
Year FE Yes Yes
Year-Quarter FE No No
Credit Rating FE No No
Loan Type FE No No
Observations 2,129 945
Panel B: RS Metrics Data
Inst. Lender
(1) (2) (3)
Post RM -0.928* -0.090** -0.120***
(-1.77) (-2.14) (-2.77)
Model Logit OLS OLS
Controls Yes Yes Yes
Firm FE Yes Yes Yes
Year FE Yes Yes No
Year-Quarter FE No No Yes
Credit Rating FE No No Yes
Loan Type FE No No Yes
Observations 904 2,129 2,129
Adj. (Pseudo) R-squared 0.066 0.228 0.245
47
APPENDIX B (continued)
Additional Analyses
Panel C: Covariate Balancing
Pre-Matching Post-Matching
Treatment
Mean
Control
Mean
Treatment
SD
Control
SD
Diff
Mean
Treatment
Mean
Control
Mean
Treatment
SD
Control
SD
Diff
Mean
Assets 8.174 8.038 1.549 1.839 0.136
8.174 8.173 1.549 1.549 0.001
Current Ratio 1.581 1.602 0.86 1.004 -0.021
1.581 1.58 0.86 0.86 0.001
Leverage 0.311 0.356 0.276 0.264 -0.045***
0.311 0.311 0.276 0.276 0.000
MTB 3.202 3.237 9.83 8.847 -0.035
3.202 3.202 9.83 9.827 0.000
Sales Growth 0.054 0.171 0.142 0.378 -0.117***
0.054 0.054 0.142 0.142 0.000
Interest Coverage 72.63 63.79 268.132 291.148 8.84
72.63 72.63 268.132 268.157 0.000
ROA 0.062 0.019 0.077 0.127 0.043***
0.062 0.062 0.077 0.077 0.000
Altman Z 4.673 3.276 2.277 2.909 1.397***
4.673 4.672 2.277 2.277 0.001
AGE 28.96 23.03 17.433 18.73 5.93***
28.96 28.95 17.433 17.433 0.010
Past Return 0.056 0.055 0.245 0.253 0.001
0.056 0.056 0.245 0.245 0.000
Amounts 19.89 19.8 1.031 1.048 0.09
19.89 19.89 1.031 1.031 0.000
Maturity 3.997 3.95 0.189 0.209 0.047***
3.997 3.997 0.189 0.189 0.000
Secured 0.507 0.496 0.501 0.5 0.011
0.507 0.507 0.501 0.5 0.000
Guarantor 0.102 0.086 0.303 0.28 0.016 0.102 0.102 0.303 0.303 0.000
Panel D: Institutional Lenders’ Demand for Private Information and Borrowing Terms - Treatment Sample
Interest Spread Amounts Maturity
(1) (2) (3) (4) (5) (6)
Post No Inst. Lender -0.166 0.018
0.872*** 1.012***
0.136** 0.148
(-1.40) (0.15)
(4.50) (3.40)
(2.13) (1.53)
Had Inst. Lender -0.428*** -0.118
0.698*** 0.983***
0.187** 0.150
(-2.70) (-0.88)
(2.90) (2.93)
(2.13) (1.42)
Post No Inst. Lender x Had Inst. Lender 0.385** 0.078
-0.925*** -1.188***
-0.208** -0.217*
(2.46) (0.55)
(-2.80) (-2.96)
(-2.01) (-1.87)
Assets -0.008 0.010
0.231* 0.323**
0.004 0.025
(-0.20) (0.26)
(1.85) (2.58)
(0.14) (1.14)
48
APPENDIX B (continued)
Additional Analyses
Panel D: Institutional Lenders’ Demand for Private Information and Borrowing Terms
- Treatment Sample (continued)
Interest Spread Amounts Maturity
(1) (2) (3) (4) (5) (6)
Current Ratio 0.014 0.051
0.080 0.020
-0.002 0.005
(0.33) (1.22)
(0.82) (0.19)
(-0.06) (0.22)
Leverage 0.118 0.305**
-0.036 0.290
-0.169 -0.003
(0.70) (2.52)
(-0.06) (0.53)
(-1.53) (-0.04)
MTB -0.002 -0.002
0.005 0.005
-0.000 -0.000
(-1.50) (-1.11)
(1.34) (1.24)
(-0.09) (-0.44)
Sales Growth -0.240* -0.256**
-0.831** -0.814**
-0.037 -0.217**
(-1.68) (-2.14)
(-2.01) (-2.02)
(-0.34) (-2.24)
Interest Coverage 0.000 0.000
-0.000 -0.000
0.000 0.000
(1.00) (0.35)
(-0.77) (-0.39)
(0.50) (0.77)
ROA -0.263 0.069
2.023** 3.071***
-0.084 0.306
(-0.82) (0.24)
(2.19) (3.84)
(-0.40) (1.65)
Altman Z -0.015 -0.026**
-0.051 -0.073**
-0.008 -0.016
(-1.13) (-2.11)
(-1.52) (-2.41)
(-0.66) (-1.47)
Age 0.002 0.001
-0.016** -0.020***
-0.001 -0.001
(0.46) (0.40)
(-2.32) (-2.80)
(-0.94) (-0.76)
Past Return 0.162*** 0.103*
-0.258 -0.371*
-0.024 -0.021
(3.07) (1.82)
(-1.48) (-1.76)
(-0.61) (-0.53)
Amounts -0.007 -0.041*
0.044** -0.000
(-0.22) (-1.76)
(2.36) (-0.04)
Maturity -0.023 -0.011
0.641** -0.010
(-0.31) (-0.10)
(2.52) (-0.04)
Secured 0.141*** 0.062
0.080 0.121
0.061 0.018
(2.81) (1.30)
(0.41) (0.64)
(1.49) (0.47)
Guarantor -0.162** -0.063
0.198 0.351
0.001 0.051
(-2.57) (-1.56)
(1.34) (1.62)
(0.03) (1.56)
Sales Growth -0.240* -0.256**
-0.831** -0.814**
-0.037 -0.217**
(-1.68) (-2.14)
(-2.01) (-2.02)
(-0.34) (-2.24)
Model OLS OLS OLS OLS OLS OLS
Firm FE Yes Yes
Yes Yes
Yes Yes
Year FE No No
No No
No No
Year-Quarter FE Yes Yes
Yes Yes
Yes Yes
Credit Rating FE No Yes
No Yes
No Yes
Loan Type FE No Yes
No Yes
No Yes
Observations 412 412
412 412
412 412
Adj. (Pseudo)
R-squared
0.614 0.755 0.625 0.670 0.235 0.555
49
FIGURE 1
Parallel Trend of Institutional Lending
This figure plots OLS regression coefficient estimates and two-tailed 90
th
percentile confidence
intervals based on standard errors clustered at the firm level. I replace Treatment Firm x Post
variable in Model (1) with separate interactions between Treatment Firm and trend variables. Each
trend variable is equal to one for every around two-year sample period before and after the
initiation of the satellite data coverage (and zero otherwise). The last two-year period before the
release of the satellite data (from 4
th
quarter in 2013 to 3
rd
quarter in 2015) serves as a benchmark.
50
TABLE 1
Descriptive Statistics
This table provides descriptive statistics for the main variables of interest. All variables are
defined in Appendix A.
N Mean Median SD
Inst. Lender 2,129 0.165 0.000 0.372
Post 2,129 0.332 0.000 0.471
Treatment Firm 2,129 0.194 0.000 0.395
Post Inst. Lender 2,129 0.115 0.000 0.320
Post No Inst. Lender 2,129 0.217 0.000 0.412
Had Inst. Lender 2,129 0.104 0.000 0.305
Assets 2,129 8.064 7.961 1.787
Current Ratio 2,129 1.598 1.342 0.978
Leverage 2,129 0.348 0.319 0.267
MTB 2,129 3.230 2.506 9.043
Sales Growth 2,129 0.148 0.070 0.348
Interest Coverage 2,129 65.500 8.409 286.798
ROA 2,129 0.027 0.042 0.120
Altman Z 2,129 3.546 3.290 2.851
Age 2,129 24.178 21.000 18.629
Past Return 2,129 0.055 0.040 0.252
No Analyst Coverage 2,129 0.508 1.000 0.500
No Earnings Forecast 2,129 0.307 0.000 0.461
High Reputation 2,129 0.281 0.000 0.450
Amounts 2,129 19.821 19.808 1.045
Maturity 2,129 3.959 4.096 0.206
Secured 2,129 0.498 0.000 0.500
Guarantor 2,129 0.089 0.000 0.284
Interest Spread 2,129 5.273 5.267 0.339
51
TABLE 2
Satellite Image Data and Institutional Lending
This table examines whether the probability that institutional lenders issue a loan is lower after
the satellite data on a borrower becomes commercially available. Panel A reports the result of
univariate analysis. Panel B presents the result of multivariate analysis. Panel C shows the result
of analysis using an entropy balancing approach. In Panel B and C, Column 1 (2 and 3) presents
the results using a logit (OLS) model. t-statistics in parentheses are based on standard errors
clustered at the firm level. ***, **, and * indicate significance at the 0.01, 0.05, and 0.10 levels,
respectively. All variables are defined in Appendix A.
Panel A: Univariate Analysis
Inst. Lender
Treatment Firm=0
(a)
Treatment Firm=1
(b)
Difference
(b) - (a)
Post = 0
(c)
0.163 0.160 -0.003
Post = 1
(d)
0.194 0.087 -0.107***
Difference
(d) - (c)
0.031 -0.073** -0.104**
Panel B: Multivariate Analysis
Inst. Lender
(1) (2) (3)
Treatment Firm x Post -1.123** -0.107*** -0.138***
(-2.43) (-2.82) (-3.48)
Assets -0.311 -0.021 -0.006
(-1.35) (-1.09) (-0.34)
Current Ratio 0.118 0.011 0.012
(0.58) (0.71) (0.77)
Leverage -0.585 -0.043 -0.051
(-0.98) (-0.61) (-0.67)
MTB 0.008 0.001 0.001
(0.79) (1.00) (1.00)
Sales Growth 0.332 0.031 0.033
(0.99) (0.74) (0.79)
Interest Coverage -0.001* -0.000 -0.000*
(-1.82) (-1.56) (-1.72)
ROA -1.790* -0.160 -0.180
(-1.69) (-1.41) (-1.52)
Altman Z -0.083 -0.010 -0.012**
(-1.41) (-1.60) (-2.00)
Age -0.001 -0.000 -0.001
(-0.03) (-0.00) (-0.44)
52
TABLE 2 (continued)
Satellite Image Data and Institutional Lending
Panel B: Multivariate Analysis (continued)
Inst. Lender
(1) (2) (3)
Past Return -0.545 -0.059 -0.079**
(-1.34) (-1.57) (-2.16)
Amounts 0.206 0.023 0.032**
(1.58) (1.60) (2.03)
Maturity 0.685 0.061 0.092
(1.32) (1.18) (1.60)
Secured 0.610** 0.075** 0.063*
(2.21) (2.32) (1.85)
Guarantor 0.607 0.052 0.070
(1.39) (1.00) (1.27)
Model Logit OLS OLS
Firm FE Yes Yes Yes
Year FE Yes Yes No
Year-Quarter FE No No Yes
Credit Rating FE No No Yes
Loan Type FE No No Yes
Observations 904 2,129 2,129
Adj. (Pseudo) R-squared 0.069 0.230 0.247
Panel C: Entropy Balancing Approach
Inst. Lender
(1) (2) (3)
Treatment Firm x Post -1.787*** -0.113*** -0.125***
(-2.94) (-2.82) (-3.16)
Model Logit OLS OLS
Controls Yes Yes Yes
Firm FE Yes Yes Yes
Year FE Yes Yes No
Year-Quarter FE No No Yes
Credit Rating FE No No Yes
Loan Type FE No No Yes
Observations 904 2,129 2,129
Adj. (Pseudo) R-squared N/A 0.483 0.534
53
TABLE 3
Falsification Test
This table reports the results of falsification tests using different types of institutional lenders as
a dependent variable. Panel A presents the result of the falsification test based on OLS and logit
model. Columns 1 and 3 (2 and 4) present the results using a logit (OLS) model. In Columns 1
and 2, the dependent variable is Inst. Lender Bank Affiliated which is equal to 1 if the loan is issued
with at least one bank-affiliated-institutional lender but is not issued with independent
institutional lenders, and 0 otherwise. In Columns 3 and 4, the dependent variable is Inst. Lender
No Investment which is equal to 1 if the loan is issued with at least one independent-institutional
lender that does not engage in investment businesses but is not issued with an independent
institutional lender that engages in investment businesses, and 0 otherwise. Panel B reports the
results of falsification tests using a multinomial logit model. t-statistics in parentheses are based
on standard errors clustered at the firm level. ***, **, and * indicate significance at the 0.01, 0.05,
and 0.10 levels, respectively. All variables are defined in Appendix A.
Panel A: Falsification Test
Inst. Lender
Bank Affiliated
Inst. Lender
No Investment
(1) (2) (3) (4)
Treatment Firm x Post -0.317 -0.024
1.861 0.014
(-0.68) (-0.38)
(1.62) (0.48)
Assets -0.160 -0.035
0.220 0.021
(-0.69) (-1.21)
(0.52) (1.02)
Current Ratio -0.237 -0.016
-1.480*** -0.013*
(-1.23) (-0.87)
(-4.53) (-1.78)
Leverage 0.031 -0.010
-1.239 -0.031
(0.05) (-0.13)
(-0.84) (-0.89)
MTB -0.005 -0.001
0.004 0.000
(-0.51) (-0.52)
(0.20) (0.11)
Sales Growth -0.204 -0.031
-1.033 -0.028*
(-0.59) (-0.74)
(-1.01) (-1.74)
Interest Coverage -0.000 -0.000
0.001 -0.000
(-0.56) (-0.50)
(0.68) (-0.13)
ROA 0.245 0.024
-3.575 -0.104
(0.25) (0.21)
(-1.28) (-1.62)
Altman Z -0.048 -0.009
0.305 0.003
(-0.74) (-1.03)
(1.53) (1.05)
Age -0.014 -0.002
0.012 0.002*
(-0.96) (-0.90)
(0.49) (1.82)
Past Return 0.324 0.050
-0.809 -0.027
(1.02) (1.19)
(-1.40) (-1.17)
Amounts 0.285** 0.030*
0.037 0.013
(2.32) (1.70)
(0.13) (1.35)
54
TABLE 3 (continued)
Falsification Test
Panel A: Falsification Test (continued)
Inst. Lender
Bank Affiliated
Inst. Lender
No Investment
(1) (2) (3) (4)
Maturity 0.642 0.077
-0.766 -0.044
(1.62) (1.23)
(-0.62) (-1.11)
Secured -0.216 -0.023
0.966 0.050***
(-0.78) (-0.63)
(1.29) (2.67)
Guarantor 0.155 0.028
-0.109 -0.010
(0.43) (0.60)
(-0.19) (-0.31)
Model Logit OLS Logit OLS
Firm FE Yes Yes
Yes Yes
Year FE Yes No
Yes No
Year-Quarter FE No Yes
No Yes
Credit Rating FE No Yes
No Yes
Loan Type FE No Yes
No Yes
Observations 1,043 2,129
295 2,129
Adj. (Pseudo) R-squared 0.029 0.248 0.291 0.349
Panel B: Falsification Test - Multinomial Logit Model Analysis
Inst. Lender
Bank Affiliated = 1
Inst. Lender
No Investment = 1
Inst. Lender = 1
(1) (2) (3)
Treatment Firm x Post -0.894 0.807 -2.197***
(-1.36) (0.67) (-2.86)
Assets -0.226 0.312 -0.302
(-0.62) (0.46) (-0.86)
Current Ratio -0.236 -0.604 0.008
(-0.87) (-0.68) (0.03)
Leverage -0.152 -0.781 -0.622
(-0.18) (-0.42) (-0.66)
MTB 0.001 0.021 0.012
(0.10) (0.58) (0.74)
Sales Growth -0.263 -1.593 0.245
(-0.53) (-1.02) (0.54)
Interest Coverage -0.001 0.000 -0.002**
(-1.20) (0.22) (-1.99)
ROA -0.859 -5.589 -2.767
(-0.53) (-1.50) (-1.50)
Altman Z -0.095 0.116 -0.100
(-1.07) (0.43) (-1.20)
55
TABLE 3 (continued)
Falsification Test
Panel B: Falsification Test - Multinomial Logit Model Analysis (continued)
Inst. Lender
Bank Affiliated = 1
Inst. Lender
No Investment = 1
Inst. Lender
= 1
(1) (2) (3)
AGE -0.010 0.005 -0.004
(-0.34) (0.10) (-0.10)
Past Return -0.026 -0.694 -0.822
(-0.05) (-0.86) (-1.37)
Amounts 0.345** 0.264 0.397**
(1.96) (0.67) (2.11)
Maturity 0.711 -0.262 0.799
(1.34) (-0.16) (1.10)
Secured 0.115 1.005 0.758*
(0.28) (1.05) (1.81)
Guarantor 0.558 0.135 1.086
(0.98) (0.16) (1.56)
Model Mlogit Mlogit Mlogit
Controls Yes Yes Yes
Firm FE Yes Yes Yes
Year FE Yes Yes Yes
Year-Quarter FE No No No
Credit Rating FE No No No
Loan Type FE No No No
Observations 2,129 2,129 2,129
Adj. (Pseudo) R-squared 0.617 0.617 0.617
Test: [1]Treatment Firm x Post
> [3]Treatment Firm x Post
p-value: 0.040
Test: [2]Treatment Firm x Post
> [3]Treatment Firm x Post
p-value: 0.011
56
TABLE 4
Borrower Opacity
This table examines whether the effect of the satellite image data coverage on institutional lender
participation is more pronounced when borrowers are opaque. Panel A, B and C report the results
of the analyses in which borrower opacity is measured by a borrower’s equity analyst coverage
(No Analyst Coverage), whether a borrower issues earnings forecasts (No Earnings Forecast) and a
borrower’s press releases (Low Press Releases). In all panels, Columns 1 and 2 (3 and 4) present the
results using a logit (OLS) model. t-statistics in parentheses are based on standard errors clustered
at the firm level. ***, **, and * indicate significance at the 0.01, 0.05, and 0.10 levels, respectively.
All variables are defined in Appendix A.
Panel A: No Analyst Coverage
Inst. Lender
No Analyst
Coverage=0
No Analyst
Coverage=1
No Analyst
Coverage=0
No Analyst
Coverage=1
(1) (2) (3) (4)
Treatment Firm x Post -0.426 -2.613***
-0.071 -0.254***
(-0.79) (-2.91)
(-1.42) (-3.79)
Model Logit Logit OLS OLS
Controls Yes Yes
Yes Yes
Firm FE Yes Yes
Yes Yes
Year FE Yes Yes
No No
Year-Quarter FE No No
Yes Yes
Credit Rating FE No No
Yes Yes
Loan Type FE No No
Yes Yes
Observations 492 412
1,047 1,082
Adj. (Pseudo) R-squared 0.105 0.140 0.252 0.251
Test: [1]Treatment Firm x Post
> [2]Treatment Firm x Post
p-value: 0.012
Test: [3]Treatment Firm x Post
> [4]Treatment Firm x Post
p-value: 0.007
57
TABLE 4 (continued)
Borrower Opacity
Panel B: No Earnings Forecast
Inst. Lender
No Earnings
Forecast=0
No Earnings
Forecast=1
No Earnings
Forecast=0
No Earnings
Forecast=1
(1) (2) (3) (4)
Treatment Firm x Post -0.809* -16.590***
-0.113*** -0.349**
(-1.75) (-13.53)
(-2.77) (-2.40)
Model Logit Logit OLS OLS
Controls Yes Yes
Yes Yes
Firm FE Yes Yes
Yes Yes
Year FE Yes Yes
No No
Year-Quarter FE No No
Yes Yes
Credit Rating FE No No
Yes Yes
Loan Type FE No No
Yes Yes
Observations 649 255
1,475 654
Adj. (Pseudo) R-squared 0.066 0.246 0.255 0.222
Test: [1]Treatment Firm x Post
> [2]Treatment Firm x Post
p-value: 0.042
Test: [3]Treatment Firm x Post
> [4]Treatment Firm x Post
p-value: 0.030
Panel C: Low Press Releases
Inst. Lender
Low Press
Releases=0
Low Press
Releases=1
Low Press
Releases=0
Low Press
Releases=1
(1) (2) (3) (4)
Treatment Firm x Post -0.824 -1.997**
-0.091 -0.187***
(-1.54) (-2.33)
(-1.55) (-3.33)
Model Logit Logit OLS OLS
Controls Yes Yes
Yes Yes
Firm FE Yes Yes
Yes Yes
Year FE Yes Yes
No No
Year-Quarter FE No No
Yes Yes
Credit Rating FE No No
Yes Yes
Loan Type FE No No
Yes Yes
Observations 551 353
1,065 1,064
Adj. (Pseudo) R-squared 0.080 0.149 0.197 0.319
Test: [1]Treatment Firm x Post
> [2]Treatment Firm x Post
p-value: 0.097
Test: [3]Treatment Firm x Post
> [4]Treatment Firm x Post
p-value: 0.099
58
TABLE 5
Early Dissemination of Borrower Private Information
This table examines whether early dissemination of borrower private information is important to
the relationship between the satellite image data coverage and institutional lender participation.
Panel A and B report the results of the analyses in which the borrower’s information
dissemination is measured by the number of performance covenants in the loan (High Perf.
Covenants) and the lender reputation (High Reputation), respectively. In all panels, Columns 1 and
2 (3 and 4) present the results using a logit (OLS) model. t-statistics in parentheses are based on
standard errors clustered at the firm level. ***, **, and * indicate significance at the 0.01, 0.05, and
0.10 levels, respectively. All variables are defined in Appendix A.
Panel A: High Performance Covenants
Inst. Lender
High Perf.
Covenants=0
High Perf.
Covenants=1
High Perf.
Covenants=0
High Perf.
Covenants=1
(1) (2) (3) (4)
Treatment Firm x Post -0.301 -1.828***
-0.088 -0.187***
(-0.35) (-3.01)
(-1.24) (-3.82)
Model Logit Logit OLS OLS
Controls Yes Yes
Yes Yes
Firm FE Yes Yes
Yes Yes
Year FE Yes Yes
No No
Year-Quarter FE No No
Yes Yes
Credit Rating FE No No
Yes Yes
Loan Type FE No No
Yes Yes
Observations 381 523
1,030 1,099
Adj. (Pseudo) R-squared 0.090 0.152 0.242 0.256
Test: [1]Treatment Firm x Post
> [2]Treatment Firm x Post
p-value: 0.051
Test: [3]Treatment Firm x Post
> [4]Treatment Firm x Post
p-value: 0.094
59
TABLE 5 (continued)
Early Dissemination of Borrower Private Information
Panel B: Lender Reputation
Inst. Lender
High
Reputation=0
High
Reputation=1
High
Reputation=0
High
Reputation=1
(1) (2) (3) (4)
Treatment Firm x Post -0.663 -2.078***
-0.078 -0.294***
(-1.00) (-3.51)
(-1.59) (-3.76)
Model Logit Logit OLS OLS
Controls Yes Yes
Yes Yes
Firm FE Yes Yes
Yes Yes
Year FE Yes Yes
No No
Year-Quarter FE No No
Yes Yes
Credit Rating FE No No
Yes Yes
Loan Type FE No No
Yes Yes
Observations 586 318
1,531 598
Adj. (Pseudo) R-squared 0.061 0.173 0.258 0.223
Test: [1]Treatment Firm x Post
> [2]Treatment Firm x Post
p-value: 0.057
Test: [3]Treatment Firm x Post
> [4]Treatment Firm x Post
p-value: 0.005
60
TABLE 6
Accuracy of Satellite Image Data
This table examines whether the effect of the satellite image data coverage on institutional lending is
stronger when the data predicts borrowers’ performance more accurately. Panel A and B report the
results of analysis in which the accuracy of the satellite image data is measured by the correlation
between quarterly changes in store car counts and quarterly changes in sales of the borrower
(Treatment Firm High Corr, Treatment Firm Low Corr) and the average standard deviation of quarterly
changes in car counts across stores (Treatment Firm High SD, Treatment Firm Low SD), respectively. In
all panels, Columns 1 (3 and 4) presents the results using a logit (OLS) model. t-statistics in parentheses
are based on standard errors clustered at the firm level. ***, **, and * indicate significance at the 0.01,
0.05, and 0.10 levels, respectively. All variables are defined in Appendix A.
Panel A: Accuracy of the Satellite Image Data - High Correlations
Inst. Lender
(1) (2) (3)
Treatment Firm High Corr x Post -1.852*** -0.176*** -0.194***
(-2.78) (-3.82) (-4.16)
Treatment Firm Low Corr x Post -0.451 -0.067 -0.080
(-0.80) (-1.28) (-1.54)
Model Logit OLS OLS
Controls Yes Yes Yes
Firm FE Yes Yes Yes
Year FE Yes Yes No
Year-Quarter FE No No Yes
Credit Rating FE No No Yes
Loan Type FE No No Yes
Observations 904 2,129 2,129
Adj. (Pseudo) R-squared 0.073 0.239 0.248
Test: Treatment Firm Low Corr x Post
> Treatment Firm High Corr x Post
p-value: 0.042 p-value: 0.038 p-value: 0.029
61
TABLE 6 (continued)
Accuracy of Satellite Image Data
Panel B: Accuracy of the Satellite Image Data - Low Standard Deviation
Inst. Lender
(1) (2) (3)
Treatment Firm High SD x Post -0.654 -0.082* -0.094*
(-1.13) (-1.66) (-1.92)
Treatment Firm Low SD x Post -1.572*** -0.165*** -0.184***
(-2.61) (-3.34) (-3.59)
Model Logit OLS OLS
Controls Yes Yes Yes
Firm FE Yes Yes Yes
Year FE Yes Yes No
Year-Quarter FE No No Yes
Credit Rating FE No No Yes
Loan Type FE No No Yes
Observations 904 2,129 2,129
Adj. (Pseudo) R-squared 0.071 0.238 0.247
Test: Treatment Firm High SD x Post
> Treatment Firm Low SD x Post
p-value: 0.115 p-value: 0.094 p-value: 0.075
62
TABLE 7
Institutional Lenders’ Demand for Private Information and Borrowing Terms
This table examines whether the information demand from institutional lenders affects borrowers’
credit outcomes. In Columns 1 and 2, the dependent variable is Interest Spread. In Columns 3 and
4, the dependent variable is Amounts. In Columns 5 and 6, the dependent variable is Maturity.
The main variable of interest is Treatment Firm x Post No Inst. Lender x Had Inst. Lender which
captures loans issued by treatment borrowers (Treatment Firm = 1) who do not obtain loans from
institutional lenders in the post period (Post No Inst. Lender = 1) but had lending relationships
with institutional lenders in the pre-coverage period (Had Inst. Lender = 1). t-statistics in
parentheses are based on standard errors clustered at the firm level. ***, **, and * indicate
significance at the 0.01, 0.05, and 0.10 levels, respectively. All variables are defined in Appendix A.
Interest Spread Amounts Maturity
(1) (2) (3) (4) (5) (6)
Treatment Firm 0.038 0.029
0.156 0.146
0.001 0.015
x Post No Inst. Lender (1.04) (0.77)
(1.45) (1.32)
(0.04) (0.52)
Treatment Firm
x Post No Inst. Lender
0.348*** 0.233*
-0.580* -0.596*
-0.237** -0.174*
x Had Inst. Lender (2.68) (1.72)
(-1.72) (-1.72)
(-2.35) (-1.90)
Treatment Firm 0.089 0.048
-0.716*** -0.704***
-0.102* -0.092
x Post Inst. Lender (0.99) (0.50)
(-3.80) (-3.42)
(-1.73) (-1.39)
Treatment Firm
x Post Inst. Lender
-0.393*** -0.284**
0.605** 0.639**
0.243*** 0.157*
x Had Inst. Lender (-3.30) (-2.38)
(2.43) (2.49)
(2.97) (1.91)
Post No Inst. Lender -0.075 -0.037
-0.100 -0.132
0.039 0.027
(-1.43) (-0.77)
(-0.93) (-1.33)
(1.19) (0.98)
Had Inst. Lender -0.009 0.016
-0.057 -0.093
-0.028 -0.002
(-0.17) (0.35)
(-0.34) (-0.73)
(-0.77) (-0.06)
Post No Inst. Lender -0.000 -0.022
-0.158 -0.077
0.040 0.000
x Had Inst. Lender (-0.00) (-0.35)
(-0.76) (-0.46)
(0.82) (0.01)
Assets -0.038* -0.020
0.273*** 0.246***
-0.017 -0.018
(-1.92) (-1.10)
(5.32) (4.78)
(-1.33) (-1.35)
Current Ratio -0.002 0.004
-0.031 -0.037
0.013 0.005
(-0.19) (0.37)
(-1.00) (-1.31)
(1.20) (0.53)
Leverage 0.035 0.055
0.075 0.128
-0.065 -0.077*
(0.66) (1.00)
(0.47) (0.93)
(-1.54) (-1.86)
MTB -0.001 -0.001
0.003 0.002
0.000 0.000
(-1.44) (-1.50)
(1.41) (1.08)
(0.30) (0.98)
Sales Growth -0.063** -0.044*
0.025 -0.066
-0.020 -0.029
(-2.36) (-1.83)
(0.34) (-1.02)
(-0.98) (-1.48)
Interest Coverage -0.000 -0.000
-0.000** -0.000*
-0.000 -0.000
(-0.52) (-0.54)
(-2.02) (-1.81)
(-0.59) (-0.54)
ROA -0.160* -0.142
0.828*** 0.832***
0.073 0.062
(-1.78) (-1.64)
(3.36) (3.52)
(0.96) (0.95)
63
TABLE 7 (continued)
Institutional Lenders’ Demand for Private Information and Borrowing Terms
Interest Spread Amounts Maturity
(1) (2) (3) (4) (5) (6)
Altman Z -0.005 -0.007
0.019* 0.023**
-0.003 -0.002
(-0.75) (-1.17)
(1.65) (2.00)
(-0.78) (-0.53)
Age 0.003** 0.002
-0.009** -0.007*
0.001 0.001
(2.02) (1.36)
(-2.41) (-1.73)
(0.38) (0.45)
Past Return 0.050* 0.039
0.013 0.002
0.004 0.009
(1.88) (1.54)
(0.17) (0.03)
(0.18) (0.52)
Amounts -0.033** -0.051***
0.041*** 0.035***
(-2.29) (-4.28)
(4.93) (4.44)
Maturity -0.124*** -0.036
0.494*** 0.509***
(-3.26) (-0.89)
(5.08) (4.53)
Secured 0.120*** 0.068***
0.064 0.009
0.055*** 0.018
(5.09) (3.32)
(1.02) (0.15)
(3.30) (1.10)
Guarantor -0.118*** -0.051*
0.259** 0.187*
0.004 -0.009
(-3.65) (-1.93)
(1.99) (1.67)
(0.14) (-0.36)
Model OLS OLS OLS OLS OLS OLS
Firm FE Yes Yes
Yes Yes
Yes Yes
Year FE No No
No No
No No
Year-Quarter FE Yes Yes
Yes Yes
Yes Yes
Credit Rating FE No Yes
No Yes
No Yes
Loan Type FE No Yes
No Yes
No Yes
Observations 2,129 2,129
2,129 2,129
2,129 2,129
Adj. (Pseudo)
R-squared
0.585 0.681 0.658 0.694 0.263 0.454
Abstract (if available)
Abstract
I explore whether the value of borrowers’ private information is an important determinant of institutional lender participation in syndicated loans. Institutional lenders have been shown to exploit their access to borrowers’ private information by trading on it in financial markets. As a shock to these lenders’ private information advantage, I utilize the release of the satellite image data of car counts in store parking lots of U.S. retail firms. The satellite data provides accurate and near-real-time signals of firm performance, which undermines the value of borrowers’ private information obtained through syndicate participation. I find that once the satellite data becomes commercially available, institutional lenders are less likely to participate in syndicated loans. Consistent with institutional lenders’ information-demand channel, the effect of the satellite data coverage is more pronounced when borrowers are opaque or disseminate private information to their lenders earlier. The satellite data coverage further attenuates institutional lending when the data is more accurate in predicting borrower performance. I also show that institutional lenders’ reduced demand for private information leads to less favorable loan terms for borrowers. Overall, these findings suggest that big data sources can crowd out the value of private information acquired through lending relationships.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Private disclosure contracts within supply chains and managerial learning from stock prices
PDF
Social movements and access to credit
PDF
Architectural innovations for mitigating data movement cost on graphics processing units and storage systems
Asset Metadata
Creator
Kang, Jung Koo
(author)
Core Title
Gone with the big data: institutional lender demand for private information
School
Marshall School of Business
Degree
Doctor of Philosophy
Degree Program
Business Administration
Publication Date
04/24/2021
Defense Date
04/22/2021
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
alternative data,big data,insider trading,institutional lender,lending relationship,OAI-PMH Harvest,private information
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Wittenberg-Moerman, Regina (
committee chair
), Lennox, Clive (
committee member
), Ramcharan, Rodney (
committee member
), Stice-Lawrence, Lorien (
committee member
), Subramanyam, K.R. (
committee member
)
Creator Email
jkgehngis@hotmail.com,jungkook@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c89-449178
Unique identifier
UC11668611
Identifier
etd-KangJungKo-9528.pdf (filename),usctheses-c89-449178 (legacy record id)
Legacy Identifier
etd-KangJungKo-9528.pdf
Dmrecord
449178
Document Type
Dissertation
Rights
Kang, Jung Koo
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
alternative data
big data
insider trading
institutional lender
lending relationship
private information