Introduction Hypotheses & Data Empirical Results Conclusion References

Stay Connected, Stay Financed

Evidence from Global Network of Air Links and Syndicated
Loans

Yan Li 1 , Zigan Wang 1 , Qie Ellie Yin 2

1 The University of Hong Kong

2 Hong Kong Baptist University

December 2022

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 1 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Outline

1 Introduction
▶ Motivation
▶ Background

2 Hypotheses & Data

▶ Hypotheses Development
▶ Data Source and Variable Construction

3 Empirical Results
▶ City-Level Analysis of Loan Spread and Number of Loans
▶ Loan Facility-Level Analysis of Spread, Diversification and Amendments
▶ Cross-sectional Tests at Loan Facility Level
▶ Other Syndicated Loan Characteristics: Covenants and Fees

4 Conclusion

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 2 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Motivation

Infrastructure and Economy

▶ Infrastructure development has high fixed costs.

▶ Difficult to recover infrastructure costs in the short run directly.
▶ Some scholars argue that the government should only invest in
infrastructure when there is enough demand.
▶ Some argue that planning infrastructure development must be
forward-looking.
▶ Whether infrastructure has a causal impact on economic growth
is a question.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 3 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Motivation

Infrastructure and Economy

▶ The relationship b/w infrastructure and the economy is endoge-

nous.
▶ It’s possible that a fast-growing economy brings infrastructure
development.
▶ Paved roads, subway, highway, railway, high-speed railway, air-
port, int’l airport, flying cars, crewed spaceflight.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 4 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Motivation

Infrastructure and Economy

▶ Difficult to find an exogenous infrastructure setting to explore

its causal effects.
▶ In this paper, we use the exogenous int’l direct flight setting to
examine the economic effects of airport construction.
▶ More specifically, we examine the effects on corporate financing
through syndicated loans.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 5 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Motivation

Importance of Syndicated Loans

▶ The total proceeds of new syndicated loans amount to $4.6

trillion USD over the globe in 2019, representing 8.25% of total
outstanding debt for publicly listed firms.1
▶ In 2019, the total IPO proceeds or the fund raised from 1,115
IPO cases in the world were 198 billion USD.

1
Total outstanding debt of publicly listed firms is the sum of Total Debt (ITEM3255) in US$ from Worldscope.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 6 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Motivation

Importance of Syndicated Loans

▶ Reuters Loan Pricing Corporation’s DealScan

• 3,756 syndicated loans were borrowed from non-domestic lenders
in 2019.
• Almost six times higher than the number 530 in 1989.

7000 Number Percentage

6000 60%

Percentage of Foreign Loans

Number of Foreign Loans

5000
50%
4000

40%
3000

2000 30%

1000
20%
0
1990 1995 2000 2005 2010 2015 2020
Year

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 7 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Motivation

Introduction of Syndicated Loans

▶ A syndicated loan is a loan issued to a firm jointly by more than

one financial institution.
• Lead arrangers: Establish and maintain a relationship with the
borrower, and take on the primary information collection and
monitoring responsibilities; usually hold a larger share than any
of the participants.
• Participant lenders: Rarely directly negotiate with the borrow-
ing firm, having an “arm’s-length” relationship with the borrow-
ing firm through the lead arranger.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 8 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Motivation

Introduction of Syndicated Loans

1 The lead arranger signs a preliminary loan agreement with the borrower.
The preliminary loan agreement specifies the loan amount, a range of
interest rates, covenants, fees, etc.
2 After signing the preliminary loan agreement, the lead arranger then turns
to potential participant lenders to fund part of the loan.
3 Participants agree to sign the loan agreement and each participant is
responsible for a share of the loan.
4 Lead arranger receives a fee for arranging and managing the syndicated
loan from the borrower.
5 During the life of the loan, the lead arranger monitors the firm, governs
the terms of the loan, calculates interest payments, and enforces financial
covenants.
6 A borrower defaults on a loan if it misses any required interest payment
or if it violates any of the covenants listed in the agreement.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 9 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Motivation

Introduction of Syndicated Loans - Example

FacilityID 305244 308656

PackageID 217969 217969
FacilityStartDate 20130815 20130815
FacilityEndDate 20180815 20180815
Loan Spread 150 150
Maturity 60 60
FacilityAmtUSD 550,000,000 267,173,574
LoanType Revolver/Line ≥ 1 Yr. Revolver/Line ≥ 1 Yr.
LoanPurpose Work. cap. Work. cap.2
Lender Westpac Banking Corp Westpac Banking Corp
LenderRole Arranger Arranger
Leaders 1 1
ParticipantsNum 16 16
LeadersNum 4 4
Lender Country ISO3 AUS AUS
Borrower Huawei Tech Investment Co Ltd Huawei Tech Investment Co Ltd
City Hong Kong Hong Kong
Borrower Country ISO3 HKG HKG

2
Other purposes include debt repayment, repurchases, acquisitions and etc.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 10 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Motivation

Importance of Loan Spread

▶ Interest spread is a key factor under consideration for both

lenders and borrowers.
• Spread ↓ ⇒ bank profit ↓
Claessens et al. (2018): 1% interest rate drop leads to an eight-
basis-point reduction in a bank’s net interest margin.
• Spread ↑ ⇒ firm value ↓
Chen and King (2014): 1% interest spread increase is associated
with about 2% lower Tobin’s Q for US firms.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 11 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Motivation

Determinants of Syndicated Loan Spread and Terms

▶ Literature has discussed many determinants of interest spread

and other terms of syndicated loans.
▶ The distance between international borrowers and lenders
is understudied.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 12 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Background

Global Financial Centers & Air Network Centers

Financial Center3 Air Network4

Rank City Rank City
1 New York 1 London
2 London 2 Pairs
3 Hong Kong 3 Frankfurt
4 Singapore 4 Amsterdam
5 San Francisco 5 Seoul
6 Shanghai 6 Singapore
7 Los Angeles 7 Hong Kong
8 Beijing 8 Zurich
9 Tokyo 9 New York
10 Paris 10 Toronto

3
Rank by The Global Financial Centres Index (GFCI) index, 2019
4
Rank by International Civil Aviation Organization (ICAO) flights centrality, 1989-2019
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 13 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Background

London as Example

80

60

40

20

−20
Borrower
5500 Miles
−40 6000 Miles
6500 Miles
London
−60
−150 −100 −50 0 50 100 150

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 14 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Hypotheses Development

Research Question

Hypothesis 1.
When a borrower’s headquarters is in a city with higher centrality
in the global network of air links, interest spreads of its syndicated
loans arranged by lead lenders in those flight-connected cities will
be lower.
▶ ↑ Air network centrality ⇒ Information asymmetry ↓
• Information asymmetry ↓ ⇒ Cost of external financing ↓
Myers and Majluf (1984)
• Information asymmetry ↓ ⇒ Loan spread ↓
Agarwal and Hauswald (2010); Knyazeva and Knyazeva (2012)

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 15 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Hypotheses Development

Research Question

Hypothesis 2.
When a borrower’s headquarters is in a city with higher centrality
in the global network of air links, its syndicated loans arranged by
lead lenders in those flight-connected cities have smaller but more
diversified lending syndicates.

▶ ↑ Air network centrality ⇒ Cost of monitoring ↓

• Moral hazard problem ↓ ⇒ Syndicate participants ↑
• Lead arrangers do not need to approach many participants ⇒
Syndicate participants ↓
• Shares of lead arranger ↓
• Loan syndicate diversification ↑

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 16 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Hypotheses Development

Research Question

Hypothesis 3.
When a borrower’s headquarters is in a city with higher centrality
in the global network of air links, its syndicated loans arranged by
lead lenders in those flight-connected cities have fewer and more
flexible covenants.
▶ ↑ Air network centrality ⇒ Cost of monitoring ↓
• Restrictive covenants ↓
Hollander and Verriest (2016)
• Number of financial covenants ↓
• Tightness5 of financial covenants ↑

5
Tightness is a covenant that measures the additional debt that the lender allows the borrower to borrow in the
future. It’s the difference b/w the initial debt/EBITDA covenants and the actual debt/EBITDA ratio, normalized
by the std. dev. of the covenant ratio over the previous twelve quarters.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 17 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Hypotheses Development

Research Question

Hypothesis 4.
When a borrower’s headquarters is in a city with higher centrality
in the global network of air links, lead arrangers in those flight-
connected cities ask for lower fees associated with granted syndi-
cated loans.
▶ ↑ Air network centrality ⇒ Borrowing cost ↓
• Fees account for a substantial portion of borrowing costs.
Berg et al. (2016)
• Syndicated loan fees ↓

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 18 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Data Source and Variable Construction

Data Source

▶ Global flight data

• International Civil Aviation Organization (ICAO)
▶ Number of passenger flights.
▶ Number of cargo flights.
▶ Airport identifier, latitude, and longitude.
▶ Syndicated loan data
• Reuters Dealscan
▶ Borrower and lender information.
▶ Loan characteristics, e.g. loan spread, maturity, facility amount.
▶ Covenants of the loan facility; fees.
▶ Firm fundamental data
• Compustat Global
• Reuters Worldscope
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 19 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Data Source and Variable Construction

Variable Construction - Centrality

▶ Global Air Network

• Direct Flight:
Two cities are “connected” if they have at least one direct flight
every week.
• Closeness Centrality :
Computed as the inverse of the average distance between a city
and every other city in the global flight network, where the dis-
tance is the shortest path connecting a city and another city in
the global air network.
• Degree Centrality :
Computed as the number of cities that the city is connected to,
divided by the number of cities in the global air network.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 20 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Data Source and Variable Construction

Variable Construction - Centrality

▶ Global Air Network

• Betweeness Centrality :
The sum of the fraction of all-pairs shortest paths that pass
through the city in the global air network.
• Eigenvector Centrality :
Calculated based on the centrality of its neighbors connected
in the flight network, following the algorithm introduced by
Bonacich (1987).

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 21 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Data Source and Variable Construction

Variable Construction - Loan Characteristics

▶ Syndicated Loan
• Lead Arranger :
Lenders are classified as lead arranger if the “Lead Arranger
Credit” is marked “Yes” or their roles are one of the following:
“Administrative Agent”, “Agent”, “Arranger”, “Bookrunner”,
“Lead Arranger”, “Lead Bank”, “Lead Manager”.
• Loan Spread:
The all-in-drawn spread over the London Interbank Offered Rate
(LIBOR) charged by the bank for the loan facility in basis points.
• Loan Maturity :
Months to maturity are given to the loan facility.
• Loan Size:
Natural logarithm of facility amount in USD.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 22 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Data Source and Variable Construction

Structure of Syndicated Loan Data - Dealscan Overview

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 23 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Data Source and Variable Construction

Variable Construction - Loan Characteristics

▶ Syndicated Loan
• Bank Term Loan:
“Term Loan A” typically held by banks. Ivashina and Sun (2011)
• Non-Bank Term Loan:
“Term Loan B” typically held by non-bank institutions.
• Revolver Loan:
Revolving loans may “revolve” (i.e., the firm may borrow, repay,
or reborrow) over the contract period.
• Annual Fee:
Fee paid on the entire committed amount, regardless of usage.
• Commitment Fee:
Fee paid on the unused amount of loan commitments.
• Standby Fee:
Fee paid to lender to compensate the syndication.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 24 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Data Source and Variable Construction

Variable Construction - Loan Characteristics

▶ Syndicated Loan
• Initial Covenant:
Initial covenant is the initial debt/EBITDA ratio that the lender
allows the borrower to borrow from all lenders. See Hollander
and Verriest (2016).
• Tightness:
Tightness is a covenant that measures the additional debt that
the lender allows the borrower to borrow in the future. It’s
the difference b/w the initial debt/EBITDA covenants and the
actual debt/EBITDA ratio, normalized by the std. dev. of the
covenant ratio over the previous twelve quarters. See Demiroglu
and James (2010).

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 25 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Data Source and Variable Construction

Variable Construction - Firm Fundamental

▶ Borrower Fundamental
• Assets:
Natural logarithm of total assets in USD.
• Market Book Ratio:
(Market Value of Equity + Book Value of Liabilities) / Total
Assets.
• Profitability :
Earnings Before Interest and Taxes / Total Assets.
• Leverage:
(Long Term Debt + Debt in Current Liabilities) / Total Assets.
• Cash Flow Volatility :
Standard deviation of (Cash Flow from Operating Activities /
Total Assets) over the last five fiscal years.
• Tangibility : Net Property, Plant and Equipment / Total Assets.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 26 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Data Source and Variable Construction

Variable Construction - Sample Construction

1 Sample period 1989 - 2014.

2 Keep loan facilities with only one lead arranger or main lead
arranger’s share ≥ 50%.
3 Cities where borrowers and lead arrangers headquarter had at
least one international flight.
4 Loan facilities that borrowers and lead arrangers are at least
1,000 miles away.
5 Loan facilities amount ≥ 100 million USD.
6 Merge the sample with Compustat Global and Worldscope.6

6
Link table provided by:Chava and Roberts (2008) and WRDS: https://wrds-www.wharton.upenn.edu/pages/
get-data/linking-suite-wrds/dealscan-worldscope-link/
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 27 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Data Source and Variable Construction

Summary Statistics - Facility Level

Our final sample includes 4,887 loans and 1,867 borrowing firms in
263 cities and 59 countries.
Variable Obs Mean Std Dev 25% Median 75%

Loan Spread 4,887 191.056 154.562 75.000 155.000 275.000

Number of Leaders 4,887 1.921 2.823 1.000 1.000 1.000
Number of Participants 4,887 6.836 8.468 0.000 4.000 10.000

Independent Variables

Direct Flight 4,887 0.210 0.407 0.000 0.000 0.000

Closeness Centrality 4,887 38.593 6.461 34.840 39.210 42.777
Degree Centrality 4,887 5.221 5.853 0.984 3.404 7.593
Betweeness Centrality 4,887 2.349 4.302 0.016 0.604 2.881
Eigenvector Centrality 4,887 6.251 6.400 1.471 4.206 8.263

Maturity 4,887 45.994 26.341 24.000 48.000 60.000

LoanSize 4,887 18.634 1.656 17.622 18.812 19.807
Assets 4,887 21.229 2.151 19.752 21.157 22.671
MarketBookRatio 4,887 1.148 1.060 0.472 0.813 1.436
Profitability 4,887 0.114 0.098 0.063 0.112 0.163
Leverage 4,887 0.316 0.214 0.162 0.295 0.442
CashFlowVolatility 4,887 0.059 0.126 0.015 0.031 0.055
Tangibility 4,887 0.368 0.288 0.099 0.313 0.621

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 28 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Data Source and Variable Construction

Summary Statistics - City Level

Variable Obs Mean Std Dev 25% Median 75%

Loan Spread (Average) 1,779 186.57 129.72 78.32 168.33 268.00

Number of Loan 1,779 5.96 7.76 1.00 3.00 7.00
Number of Revolver Loan 1,779 2.38 4.09 0.00 1.00 3.00
Number of Bank Term Loan 1,779 0.28 0.81 0.00 0.00 0.00

Independent Variables

Closeness Centrality 1,779 37.48 5.87 33.88 37.93 41.14

Degree Centrality 1,779 3.95 4.85 0.80 2.40 5.19
Betweeness Centrality 1,779 1.39 3.28 0.00 0.22 1.21
Eigenvector Centrality 1,779 5.38 5.55 1.33 3.69 7.67

Control Variables Obs Mean Std 25% Median 75%

Closeness Centrality (10 Nearest Cities) 1,779 32.86 4.24 30.54 32.96 35.23
Degree Centrality (10 Nearest Cities) 1,779 1.52 1.77 0.39 0.93 2.01
Betweeness Centrality (10 Nearest Cities) 1,779 0.38 0.83 0.00 0.03 0.38
Eigenvector Centrality (10 Nearest Cities) 1,779 2.16 2.36 0.59 1.53 2.78
Maturity (Average) 1,779 51.37 28.73 36.00 49.11 60.00
LoanSize (Average) 1,779 19.08 1.17 18.42 19.15 19.86

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 29 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

The Discontinuity Setting - Introduction

Campante and Yanagizawa-Drott (2018) documents that:

▶ Federal Aviation Authority (FAA) had required since the 1950s
that a two-pilot crew could fly at most 12 hours within a 24-
hour period.7
▶ European regulators adopted in 1991 a daily maximum of 13
hours for a flight crew member’s flight duty period working in
a basic (unaugmented) crew.

Cities are disproportionally more likely to be connected by flights

when the distance between their airports is just under 6,000 miles
than just above 6,000 miles.

7
Flights above that limit require at least three pilots and an additional flight crew member (double
augmentation), as well as “adequate sleeping quarters” on the plane
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 30 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

The Discontinuity Setting - Histogram

All Possible City Pairs Histogram

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 31 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

The Discontinuity Setting - Histogram

City Pairs with Syndicated Loan Histogram

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 32 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

The Discontinuity Setting - London as Example

80

60

40

20

−20
Borrower
5500 Miles
−40 6000 Miles
6500 Miles
London
−60
−150 −100 −50 0 50 100 150

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 33 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

The Discontinuity Setting - Balance Test

Balance Test - Firm Characteristics8

Flight Distance < 6000 Miles Flight Distance > 6000 Miles
Obs = 540 Obs = 91
Control Variable Mean Std Mean Std Diff t-stat

Loan Size 18.259 1.686 18.167 1.437 0.092 0.547

Maturity 46.426 28.723 49.077 26.219 -2.651 -0.876
Assets 21.924 2.217 22.164 1.991 -0.240 -1.040
Profitability 0.112 0.093 0.106 0.083 0.006 0.587
Leverage 0.309 0.203 0.329 0.184 -0.020 -0.926
Market Book Ratio 1.088 1.033 1.106 1.133 -0.019 -0.146
Cash Flow Volatility 0.047 0.067 0.040 0.055 0.006 0.934
Tangibility 0.337 0.285 0.380 0.250 -0.044 -1.500

8
Our regression discontinuity analysis is on city-pairs level, and the RD setting doesn’t need controls so the obs
in the regressions is larger
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 34 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

The Discontinuity Setting - Regression

We establish the following city-pair regression model:

Ykj = β0 + β1 · 1(Distancekj > 6, 000M iles) + β2 · g(Distancekj ) + ϵkj

where
• Ykj : the average loan spread between borrowers in city k and
lead lenders in city j .
• 1(Distancekj > 6, 000M iles): = 1 if distance measured by direct
fights b/w city k and city j > 6,000 miles, and = 0 o/w.
• g(·):
higher-order polynomials function.
We consider different kernel types following He et al. (2020).9
9
Three choices in RD design: the kernel function, the order of the polynomial, and the bandwidth. See
Cattaneo, M. D., Idrobo, N., & Titiunik, R. (2019). A practical introduction to regression discontinuity designs:
Foundations. Cambridge University Press. Within the bandwidth, it is common to adopt a weighting scheme to
ensure that the observations closer to the cutoff point receive more weight than those further away; a kernel
function determines the weights.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 35 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

The Discontinuity Setting - Regression

Panel A. Uniform Kernel

(1) (2) (3)

Kernel Type Uniform

Flight Distance > 6000 Miles 100.736*** 242.424*** 230.168***

(3.833) (6.978) (5.279)

Cluster Borrower City Borrower City Borrower City

Polynomial 1 2 3

Borrower Lender City-Pairs 300 301 302

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 36 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

The Discontinuity Setting - Regression

Panel B. Triangular Kernel

(1) (2) (3)

Kernel Type Triangular

Flight Distance > 6000 Miles 157.208*** 237.664*** 182.530***

(6.858) (7.628) (4.614)

Cluster Borrower City Borrower City Borrower City

Polynomial 1 2 3

Borrower Lender City-Pairs 300 301 302

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 37 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

The Discontinuity Setting - Regression

Panel C. Epanechnikov Kernel

(1) (2) (3)

Kernel Type Epanechnikov

Flight Distance > 6000 Miles 146.191*** 245.254*** 193.156***

(6.149) (7.766) (4.690)

Cluster Borrower City Borrower City Borrower City

Polynomial 1 2 3

Borrower Lender City-Pairs 300 301 302

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 38 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

The Discontinuity Setting - Figures

(a) Bin Size=20 Miles (b) Global RDD

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 39 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

The Discontinuity Setting - Placebo Test

We randomly select distances between 5,500 and 6,500 miles as

pseudo thresholds for the air-link discontinuity and re-estimate RDD
using these pseudo thresholds 500 times.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 40 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

Baseline Model

We establish the following city-year regression model:

Yc,t+1 = β0 + β1 · Centralityc,t + β2 · Xc,t + F Ecountry∗t + F Ec + ϵc,t

where
• Yc,t+1 :loan characteristics of interest in year t + 1 for borrowers
with headquarters in city c.
• Centralityc,t : city c’s centrality in year t.
• Xc,t : city-year control variables include:
• 10 nearest cities’ average centrality, average loan maturity, and
average loan size.(Lin et al. (2013))
• We include the country-year FE, city FE and cluster the stan-
dard errors at city level.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 41 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

Loan Spread - Result

(1) (2) (3) (4)

Loan Spread

Closeness Centrality -4.918*

(-1.942)
Degree Centrality -7.039*
(-1.963)
Betweeness Centrality -7.395**
(-2.138)
Eigenvector Centrality -8.817**
(-2.136)

10 Nearest Cities’ Centrality 4.322** 14.101** 15.859* 13.320**

(2.548) (2.086) (1.908) (2.198)
Average Maturity 0.664** 0.644** 0.687** 0.618**
(2.549) (2.407) (2.578) (2.319)
Average Loan Size -29.495*** -29.624*** -29.432*** -29.203***
(-4.859) (-4.967) (-4.929) (-4.933)

Country*Year FE Yes Yes Yes Yes

City FE Yes Yes Yes Yes
Cluster City City City City
Observations 1,779 1,779 1,779 1,779
R-squared 0.833 0.832 0.831 0.833

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 42 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

Loan Spread - Placebo Test

(1) (2) (3) (4)

Loan Spread

Closeness Centrality (Cargo) -0.050

(-0.014)
Degree Centrality (Cargo) 0.375
(0.111)
Betweeness Centrality (Cargo) 5.291
(0.588)
Eigenvector Centrality (Cargo) -1.732
(-0.458)

Average Maturity 1.074*** 1.074*** 1.080*** 1.073***

(3.304) (3.309) (3.349) (3.295)
Average Loan Size -27.320*** -27.364*** -27.483*** -27.151***
(-4.184) (-4.171) (-4.189) (-4.159)

Country*Year FE Yes Yes Yes Yes

City FE Yes Yes Yes Yes
Cluster City City City City
Observations 1,518 1,518 1,518 1,518
R-squared 0.862 0.862 0.862 0.862

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 43 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

Number of Loans - Result

(1) (2) (3) (4)

Number of Loans

Closeness Centrality 0.388*

(1.933)
Degree Centrality 1.331***
(3.274)
Betweeness Centrality 1.144***
(4.766)
Eigenvector Centrality 0.823**
(2.170)

10 Nearest Cities’ Centrality -0.251** -1.146*** -1.560*** -0.709**

(-2.510) (-3.229) (-3.324) (-2.316)
Average Loan Spread (100 bps) 0.008*** 0.008*** 0.008*** 0.008***
(3.343) (3.608) (3.424) (3.404)
Average Maturity 0.015 0.017 0.011 0.018
(1.209) (1.545) (0.980) (1.493)
Average LoanSize 0.426 0.452 0.441 0.391
(1.286) (1.464) (1.410) (1.268)

Country*Year FE Yes Yes Yes Yes

City FE Yes Yes Yes Yes
Cluster City City City City
Observations 1,779 1,779 1,779 1,779
R-squared 0.810 0.816 0.812 0.810

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 44 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

Number of Loans - Placebo Test

(1) (2) (3) (4)

Number of Loans

Closeness Centrality (Cargo) -0.024

(-0.094)
Degree Centrality (Cargo) 0.247
(0.557)
Betweeness Centrality (Cargo) 0.715
(1.053)
Eigenvector Centrality (Cargo) 0.129
(0.273)

Average Loan Spread (100 bps) 0.008** 0.008** 0.008** 0.008**

(2.531) (2.536) (2.523) (2.518)
Average Maturity 0.018 0.018 0.019 0.018
(1.046) (1.080) (1.119) (1.053)
Average LoanSize 0.309 0.281 0.281 0.293
(0.849) (0.789) (0.738) (0.834)

Country*Year FE Yes Yes Yes Yes

City FE Yes Yes Yes Yes
Cluster City City City City
Observations 1,518 1,518 1,518 1,518
R-squared 0.836 0.836 0.837 0.836

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 45 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

Number of Revolver Loans - Result

(1) (2) (3) (4)

Number of Revolver Loans

Closeness Centrality 0.050**

(2.251)
Degree Centrality 0.091**
(1.970)
Betweeness Centrality 0.084**
(2.401)
Eigenvector Centrality 0.100*
(1.930)

10 Nearest Cities’ Centrality -0.034* -0.173** -0.290** -0.073*

(-1.883) (-2.567) (-2.583) (-1.960)
Average Loan Spread (100 bps) 0.001 0.001* 0.001 0.001*
(1.631) (1.671) (1.582) (1.671)
Average Maturity 0.003 0.003* 0.003 0.003*
(1.601) (1.790) (1.493) (1.845)
Average LoanSize -0.022 -0.018 -0.019 -0.028
(-0.501) (-0.435) (-0.451) (-0.682)

Country*Year FE Yes Yes Yes Yes

City FE Yes Yes Yes Yes
Cluster City City City City
Observations 1,779 1,779 1,779 1,779
R-squared 0.576 0.578 0.578 0.575
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 46 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

Number of Revolver Loans - Placebo Test

(1) (2) (3) (4)

Number of Revolver Loans

Closeness Centrality (Cargo) 0.011

(0.387)
Degree Centrality (Cargo) 0.027
(0.543)
Betweeness Centrality (Cargo) -0.014
(-0.180)
Eigenvector Centrality (Cargo) 0.034
(0.562)

Average Loan Spread (100 bps) 0.001 0.001 0.001 0.001

(1.247) (1.251) (1.261) (1.267)
Average Maturity 0.003 0.003 0.003 0.003
(1.110) (1.127) (1.110) (1.115)
Average LoanSize -0.058 -0.059 -0.055 -0.059
(-1.145) (-1.147) (-1.080) (-1.163)

Country*Year FE Yes Yes Yes Yes

City FE Yes Yes Yes Yes
Cluster City City City City
Observations 1,518 1,518 1,518 1,518
R-squared 0.641 0.641 0.641 0.641

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 47 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

Number of Bank Term Loans - Result

(1) (2) (3) (4)

Number of Bank Term Loans

Closeness Centrality 0.194**

(1.999)
Degree Centrality 0.722**
(2.261)
Betweeness Centrality 0.503*
(1.914)
Eigenvector Centrality 0.507**
(2.018)

10 Nearest Cities’ Centrality -0.104** -0.494*** -0.616** -0.300*

(-2.043) (-2.668) (-2.059) (-1.883)
Average Loan Spread (100 bps) 0.001 0.001 0.000 0.001
(0.467) (0.686) (0.387) (0.584)
Average Maturity -0.000 0.001 -0.002 0.001
(-0.086) (0.150) (-0.435) (0.316)
Average LoanSize 0.149 0.162 0.154 0.130
(0.901) (1.032) (0.963) (0.825)

Country*Year FE Yes Yes Yes Yes

City FE Yes Yes Yes Yes
Cluster City City City City
Observations 1,779 1,779 1,779 1,779
R-squared 0.793 0.799 0.794 0.795

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 48 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

Number of Bank Term Loans - Placebo Test

(1) (2) (3) (4)

Number of Bank Term Loans

Closeness Centrality (Cargo) 0.004

(0.034)
Degree Centrality (Cargo) 0.084
(0.364)
Betweeness Centrality (Cargo) 0.179
(0.566)
Eigenvector Centrality (Cargo) 0.102
(0.439)

Average Loan Spread (100 bps) 0.001 0.001 0.001 0.001

(0.716) (0.716) (0.704) (0.723)
Average Maturity -0.001 -0.001 -0.001 -0.001
(-0.172) (-0.151) (-0.131) (-0.164)
Average LoanSize 0.117 0.109 0.111 0.107
(0.632) (0.604) (0.587) (0.612)

Country*Year FE Yes Yes Yes Yes

City FE Yes Yes Yes Yes
Cluster City City City City
Observations 1,518 1,518 1,518 1,518
R-squared 0.821 0.821 0.821 0.821

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 49 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

City-Level Analysis of Loan Spread and Number of Loans

Overall Results

1 If a city has more long-distance connections with other major

cities over the globe, firms headquartered in this city obtain
cheaper syndicated loans.
2 This effect is driven by the ease of transportation of people
rather than goods and products.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 50 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Baseline Model

We establish the following facility-level regression model:

Ykc,t+1 = β0 +β1 ·DirectF lightk,t /Centralityc,t +β2 ·Xk +β3 ·Xf,t +FE+ϵk
• Ykc,t+1 : loan characteristics of interest of facility k in city c.
• DirectF lightk /Centralityc,t :
direct flight of facility k from bor-
rower f to lead arranger or city c’s centrality in year t.
• Xk : facility control variables include:
• Loan maturity, loan size
• Xf,t : borrower-year control variables include:
• Assets, MarketBookRatio, Profitability, Leverage, CashFlowVolatil-
ity, Tangibility Lin et al. (2013)
• We include the country-year FE, industry-year FE, lender-year
FE, borrower FE, loan type FE and loan purpose FE.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 51 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Loan Spread - Result

(1) (2) (3) (4) (5)

Loan Spread (100bps)

Direct Flight -0.669**

(-2.241)
Closeness Centrality -0.072**
(-2.423)
Degree Centrality -0.324***
(-3.498)
Betweeness Centrality -0.268***
(-2.777)
Eigenvector Centrality -0.202**
(-2.362)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 4,887 4,887 4,887 4,887 4,887
R-squared 0.952 0.952 0.952 0.952 0.952

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 52 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Loan Spread - Placebo Test

(1) (2) (3) (4) (5)

Loan Spread (100bps)

Cargo Flight -0.578

(-1.099)
Closeness Centrality (Cargo) -0.099
(-1.036)
Degree Centrality (Cargo) 0.067
(0.268)
Betweeness Centrality (Cargo) -0.021
(-0.097)
Eigenvector Centrality (Cargo) 0.303
(0.663)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 4,206 4,206 4,206 4,206 4,206
R-squared 0.951 0.951 0.951 0.951 0.951

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 53 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Loan Spread (Spillover Effects) - Result

For loans of the firms headquartered in neighboring cities without

int’l airports
(1) (2) (3) (4)
Loan Spread (100bps)

Closeness Centrality -0.217**

(-2.245)
Degree Centrality -6.093**
(-2.300)
Betweeness Centrality -11.755*
(-1.744)
Eigenvector Centrality -1.807***
(-3.048)

Control Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes
Cluster City City City City
Observations 683 683 683 683
R-squared 0.930 0.930 0.930 0.930

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 54 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Loan Share HHI - Result

(1) (2) (3) (4) (5)

Loan Share HHI

Direct Flight -0.134***

(-2.661)
Closeness Centrality -0.034***
(-3.188)
Degree Centrality -0.106***
(-4.799)
Betweeness Centrality -0.134***
(-2.831)
Eigenvector Centrality -0.092***
(-3.053)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 2,356 2,356 2,356 2,356 2,356
R-squared 0.978 0.978 0.978 0.978 0.978

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 55 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Loan Share HHI - Placebo Test

(1) (2) (3) (4) (5)

Share Allocation HHI

Cargo Flight -0.050

(-0.490)
Closeness Centrality (Cargo) -0.043
(-1.049)
Degree Centrality (Cargo) -0.018
(-0.266)
Betweeness Centrality (Cargo) 0.052
(0.556)
Eigenvector Centrality (Cargo) -0.017
(-0.351)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 2,036 2,036 2,036 2,036 2,036
R-squared 0.982 0.982 0.982 0.982 0.982

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 56 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Lead Arrangers’ Share - Result

(1) (2) (3) (4) (5)

Leader Share

Direct Flight -11.255**

(-2.568)
Closeness Centrality -2.875***
(-2.620)
Degree Centrality -9.328***
(-5.209)
Betweeness Centrality -10.844***
(-2.679)
Eigenvector Centrality -9.802***
(-3.755)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 2,356 2,356 2,356 2,356 2,356
R-squared 0.975 0.975 0.975 0.975 0.975

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 57 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Lead Arrangers’ Share - Placebo Test

(1) (2) (3) (4) (5)

Leader Share

Cargo Flight -0.073

(-0.009)
Closeness Centrality (Cargo) -3.765
(-1.033)
Degree Centrality (Cargo) -1.175
(-0.193)
Betweeness Centrality (Cargo) 3.660
(0.468)
Eigenvector Centrality (Cargo) -2.987
(-0.666)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 2,036 2,036 2,036 2,036 2,036
R-squared 0.978 0.978 0.978 0.978 0.978

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 58 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Number of Lead Arrangers - Result

(1) (2) (3) (4) (5)

Number of Leaders

Direct Flight -1.585**

(-2.355)
Closeness Centrality -0.192***
(-4.360)
Degree Centrality -0.644***
(-4.453)
Betweeness Centrality -0.300***
(-2.810)
Eigenvector Centrality -0.551***
(-4.367)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 4,887 4,887 4,887 4,887 4,887
R-squared 0.995 0.995 0.995 0.995 0.995

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 59 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Number of Lead Arrangers - Placebo Test

(1) (2) (3) (4) (5)

Number of Leaders

Cargo Flight -0.070

(-0.633)
Closeness Centrality (Cargo) 0.007
(0.114)
Degree Centrality (Cargo) 0.118
(0.955)
Betweeness Centrality (Cargo) 0.089
(1.625)
Eigenvector Centrality (Cargo) -0.063
(-0.461)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 4,206 4,206 4,206 4,206 4,206
R-squared 0.997 0.997 0.997 0.997 0.997

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 60 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Number of Participants - Result

(1) (2) (3) (4) (5)

Number of Participants

Direct Flight -10.208***

(-2.635)
Closeness Centrality -1.298***
(-3.049)
Degree Centrality -3.723***
(-3.838)
Betweeness Centrality -2.522**
(-2.239)
Eigenvector Centrality -2.739**
(-2.505)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 4,887 4,887 4,887 4,887 4,887
R-squared 0.946 0.946 0.946 0.946 0.946

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 61 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Number of Participants - Placebo Test

(1) (2) (3) (4) (5)

Number of Participants

Cargo Flight -0.091

(-0.055)
Closeness Centrality (Cargo) -0.358
(-0.083)
Degree Centrality (Cargo) 0.568
(0.226)
Betweeness Centrality (Cargo) 1.207
(0.581)
Eigenvector Centrality (Cargo) -2.916
(-0.912)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 4,206 4,206 4,206 4,206 4,206
R-squared 0.951 0.951 0.951 0.951 0.951

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 62 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Number of Amendments - Result

(1) (2) (3) (4) (5)

Number of Amendments

Direct Flight -0.598**

(-2.370)
Closeness Centrality -0.050***
(-2.659)
Degree Centrality -0.278***
(-3.490)
Betweeness Centrality -0.139**
(-2.275)
Eigenvector Centrality -0.177**
(-2.494)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 4,887 4,887 4,887 4,887 4,887
R-squared 0.960 0.960 0.961 0.960 0.960

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 63 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Number of Amendments - Placebo Test

(1) (2) (3) (4) (5)

Number of Amendments

Cargo Flight 0.150

(1.019)
Closeness Centrality (Cargo) 0.104
(1.106)
Degree Centrality (Cargo) 0.063
(0.987)
Betweeness Centrality (Cargo) 0.103
(0.726)
Eigenvector Centrality (Cargo) 0.112
(0.792)

Control Yes Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 4,206 4,206 4,206 4,206 4,206
R-squared 0.962 0.962 0.962 0.962 0.962

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 64 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Loan Facility-Level Analysis of Spread, Diversification and Amendments

Overall Results

1 Firms tend to obtain cheaper syndicated loans if they have

direct flights connected with lenders or they’re headquartered
in cities with higher centrality.
2 This effect is driven by ease of transportation of people rather
than goods and products.
3 Firms in neighboring cities without int’l airports also benefit.
4 Loan syndicates are more diversified and lead arrangers have
less incentive to approach many participants.

Support Hypothesis 1 and Hypothesis 2.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 65 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Cross-sectional Tests at Loan Facility Level

Information Asymmetry Proxy

▶ Analyst coverage
• Analyst coverage can reduce information asymmetry and lower
the cost of debt.
Lin et al. (2013); Hallman et al. (2022)
• The original lower information asymmetry implies lower bene-
fits from on-site visits and direct flights. On-site visits are less
important as much information has already been revealed by
analysts.
• The negative impact of centrality on spread should be weaker.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 66 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Cross-sectional Tests at Loan Facility Level

Information Asymmetry Proxy

▶ Earnings management
• Earnings management may indicate worse audit quality, earnings
reporting and less credibility. Li et al. (2021)
• We use discretionary accrual by Jones (1991) to proxy firm earn-
ings management.
• On one hand, worse audit quality implies lower benefits from on-
site visits and direct flights as the manager may hide information
from the visiting lenders.
• The negative impact of centrality on spread should be weaker.
• On the other hand, worse auditing quality may imply greater
importance for lenders to visit on-site because the numbers in
the financial statements are less trustworthy.
• The negative impact of centrality on spread should be stronger.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 67 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Cross-sectional Tests at Loan Facility Level

Information Asymmetry - Baseline Model

We establish the following facility-level regression model:

Ykc,t+1 = β0 + β1 · Centralityc,t #Intc,t + β2 · Xk + β3 · Xf,t + FE + ϵk
• Ykc,t+1 : loan spread of facility k.
• Centralityc,t #Intc,t :
the interaction term of information asym-
metry proxies and city c’s centrality in year t.
• DirectF lightk,t is omitted in regressions with many FEs.
• Xk : facility control variables include:
• loan maturity, loan size
• Xf,t : borrower-year control variables include:
• Assets, MarketBookRatio, Profitability, Leverage, CashFlowVolatil-
ity, Tangibility
• We include country-year FE, industry-year FE, lender-year FE,
borrower FE, loan type FE and loan purpose FE.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 68 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Cross-sectional Tests at Loan Facility Level

Information Asymmetry - Analyst Coverage

(1) (2) (3) (4)

Loan Spread (100bps)

Closeness Centrality * Analysts Number 1.023***

(3.894)
Degree Centrality * Analysts Number 4.521***
(3.935)
Betweeness Centrality * Analysts Number 12.133***
(4.029)
Eigenvector Centrality * Analysts Number 1.896***
(3.902)
Centrality -2.596*** -8.719*** -4.967 -3.948***
(-4.358) (-4.815) (-1.064) (-4.741)
Analysts Number -49.461*** -28.858*** -16.019*** -16.180***
(-3.739) (-3.761) (-3.873) (-3.542)

Control Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes
Cluster City City City City
Observations 2,344 2,344 2,344 2,344
R-squared 0.960 0.960 0.960 0.960

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 69 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Cross-sectional Tests at Loan Facility Level

Information Asymmetry - Earnings Management

(1) (2) (3) (4)

Loan Spread (100bps)

Closeness Centrality * Earnings Management 0.982**

(2.242)
Degree Centrality * Earnings Management 3.239***
(3.405)
Betweeness Centrality * Earnings Management 3.259*
(1.951)
Eigenvector Centrality * Earnings Management 2.386*
(1.928)
Centrality -0.279*** -0.889*** -0.235*** -0.674
(-2.683) (-3.575) (-3.814) (-1.267)
Earnings Management -19.544 10.915*** 13.720*** 9.710***
(-1.359) (3.763) (4.567) (3.755)

Control Yes Yes Yes Yes

Country*Year FE Yes Yes Yes Yes
Lender*Year FE Yes Yes Yes Yes
Industry*Year FE Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes
Cluster City City City City
Observations 2,344 2,344 2,344 2,344
R-squared 0.960 0.960 0.960 0.960

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 70 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Other Syndicated Loan Characteristics: Covenants and Fees

Financial Covenants: Package-level Analysis

(1) (2) (3) (4)

Initial Covenant10

Closeness Centrality 0.053***

(2.824)
Degree Centrality 0.196***
(3.118)
Betweeness Centrality 0.258***
(3.418)
Eigenvector Centrality 0.147***
(2.935)

Country*Year FE Yes Yes Yes Yes

Industry FE Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes
Cluster City City City City
Observations 801 801 801 801
R-squared 0.978 0.978 0.978 0.978

10
Initial covenant is the initial debt/EBITDA ratio that the lender allows the borrower to borrow from all lenders.
See Hollander and Verriest (2016)
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 71 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Other Syndicated Loan Characteristics: Covenants and Fees

Financial Covenants: Package-level Analysis

(1) (2) (3) (4)

Tightness11

Closeness Centrality 0.954***

(3.560)
Degree Centrality 3.328***
(4.480)
Betweeness Centrality 2.512***
(2.795)
Eigenvector Centrality 2.988***
(3.643)

Country*Year FE Yes Yes Yes Yes

Industry FE Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes
Cluster City City City City
Observations 490 490 490 490
R-squared 0.846 0.846 0.841 0.846

11
Tightness is a covenant that measures the additional debt that the lender allows the borrower to borrow in the
future. It’s the difference b/w the initial debt/EBITDA covenants and the actual debt/EBITDA ratio, normalized
by the std. dev. of the covenant ratio over the previous twelve quarters. See Demiroglu and James (2010)
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 72 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Other Syndicated Loan Characteristics: Covenants and Fees

Fees: Loan Facility-level Analysis

(1) (2) (3) (4) (5)

Commitment Fee (100bps)12

Direct Flight -0.290***

(-2.800)
Closeness Centrality -1.282***
(-2.790)
Degree Centrality -5.623***
(-5.516)
Betweeness Centrality -4.583***
(-2.809)
Eigenvector Centrality -6.057***
(-5.710)

Country*Year FE Yes Yes Yes Yes Yes

Lender*Year FE Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 1,980 1,980 1,980 1,980 1,980
R-squared 0.980 0.980 0.980 0.980 0.980

12
Fee paid on the unused amount of loan commitments.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 73 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Other Syndicated Loan Characteristics: Covenants and Fees

Fees: Loan Facility-level Analysis

(1) (2) (3) (4) (5)

Standby Fee (100bps)13

Direct Flight -0.791***

(-3.100)
Closeness Centrality -20.542***
(-3.535)
Degree Centrality -21.414***
(-5.029)
Betweeness Centrality -14.514***
(-3.802)
Eigenvector Centrality -27.917***
(-2.869)

Country*Year FE Yes Yes Yes Yes Yes

Lender*Year FE Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 1,081 1,081 1,081 1,081 1,081
R-squared 0.993 0.994 0.993 0.993 0.993

13
Fee paid to lender to compensate the syndication.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 74 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Other Syndicated Loan Characteristics: Covenants and Fees

Fees: Loan Facility-level Analysis

(1) (2) (3) (4) (5)

Annual Fee (100bps)14

Direct Flight -0.210***

(-2.739)
Closeness Centrality -0.822***
(-3.009)
Degree Centrality -3.815***
(-4.459)
Betweeness Centrality -3.371**
(-2.416)
Eigenvector Centrality -2.225***
(-2.718)

Country*Year FE Yes Yes Yes Yes Yes

Lender*Year FE Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes
Borrower FE Yes Yes Yes Yes Yes
Loan Type FE Yes Yes Yes Yes Yes
Loan Purpose FE Yes Yes Yes Yes Yes
Cluster City City City City City
Observations 877 877 877 877 877
R-squared 0.995 0.995 0.995 0.995 0.995

14
Fee paid on the entire committed amount, regardless of usage.
Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 75 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

1 In general, we find that air links facilitate financing through

syndicated loans.
2 When a city has higher centrality in the global network of air
links, firms in this city are likely to obtain more syndicated
loans with lower interest spreads, more diversified lenders,
more flexible financial covenants and lower fees.
3 The effect of air links on syndicated loans is through easy
transportation of people rather than products.
4 Information asymmetry between borrowers and lenders has
effects.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 76 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

Thank You Very Much!

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 77 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

References I

Agarwal, S. and Hauswald, R. (2010). Distance and private information in lending. The
Review of Financial Studies, 23(7):2757–2788.
Berg, T., Saunders, A., and Steffen, S. (2016). The total cost of corporate borrowing in
the loan market: Don’t ignore the fees. The Journal of Finance, 71(3):1357–1392.
Bonacich, P. (1987). Power and centrality: A family of measures. American Journal of
Sociology, 92(5):1170–1182.
Campante, F. and Yanagizawa-Drott, D. (2018). Long-range growth: economic de-
velopment in the global network of air links. The Quarterly Journal of Economics,
133(3):1395–1458.
Chava, S. and Roberts, M. R. (2008). How does financing impact investment? the role
of debt covenants. The journal of finance, 63(5):2085–2121.
Chen, J. and King, T.-H. D. (2014). Corporate hedging and the cost of debt. Journal
of Corporate Finance, 29:221–245.
Claessens, S., Coleman, N., and Donnelly, M. (2018). “low-for-long” interest rates
and banks’ interest margins and profitability: Cross-country evidence. Journal of
Financial Intermediation, 35:1–16.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 78 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

References II

Demiroglu, C. and James, C. M. (2010). The information content of bank loan

covenants. The Review of Financial Studies, 23(10):3700–3737.
Hallman, N., Howe, J. S., and Wang, W. (2022). Analyst coverage and syndicated
lending. Review of Accounting Studies, pages 1–39.
He, G., Wang, S., and Zhang, B. (2020). Watering down environmental regulation in
china. The Quarterly Journal of Economics, 135(4):2135–2185.
Hollander, S. and Verriest, A. (2016). Bridging the gap: the design of bank loan
contracts and distance. Journal of Financial Economics, 119(2):399–419.
Ivashina, V. and Sun, Z. (2011). Institutional demand pressure and the cost of corporate
loans. Journal of Financial Economics, 99(3):500–522.
Jones, J. J. (1991). Earnings management during import relief investigations. Journal
of Accounting Research, 29(2):193–228.
Knyazeva, A. and Knyazeva, D. (2012). Does being your bank’s neighbor matter?
Journal of Banking & Finance, 36(4):1194–1209.
Li, K., Mai, F., Shen, R., and Yan, X. (2021). Measuring corporate culture using
machine learning. The Review of Financial Studies, 34(7):3265–3315.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 79 / 80
Introduction Hypotheses & Data Empirical Results Conclusion References

References III

Lin, C., Officer, M. S., Wang, R., and Zou, H. (2013). Directors’ and officers’ liability
insurance and loan spreads. Journal of Financial Economics, 110(1):37–60.
Myers, S. C. and Majluf, N. S. (1984). Corporate financing and investment decisions
when firms have information that investors do not have. Journal of Financial Eco-
nomics, 13(2):187–221.

Stay Connected, Stay Financed Yan Li , Zigan Wang , Qie Ellie Yin 80 / 80