Greening the Financial Sector: Evidence from Bank Green Bonds

Turn green or lose ‘licence to operate’—Deutsche Bank CEO Christian Sewing (Olaf Storbeck, “Turn green or lose ‘licence to operate’, says Deutsche Bank chief,” Financial Times, May 20, 2021).

Abstract

Banks are expected to play a key role in assisting the real economy with the green transition process. One of the tools used for this purpose is the issuance of green bonds. We analyze the characteristics of banks that issue green bonds to understand: (i) which banks are more likely to resort to these funding instruments, and (ii) if the issuance of green bonds leads to an improvement in a bank’s environmental footprint. We find that large banks and banks that had already publicly expressed their support for a green transition are more likely to issue green bonds. Conditional on being a green bond issuer, smaller banks tend to resort to green bonds in a more persistent manner and for relatively larger amounts, while larger banks issue green bonds on a more occasional basis and for smaller amounts. This heterogeneity is also reflected in our findings that only banks that issue green bonds more intensively improve their emissions and reduce lending to polluting sectors, thus contributing to the decarbonization of the financial sector.

Appendices

Appendix 1: Event study around green bond announcements

In this appendix we perform an event study to test the stock price reaction of green bond issuers around the announcement of the issuance of green bonds. Our sample covers all green bonds issued by publicly listed banks worldwide between January 1, 2013 and October 31, 2020. The announcement dates of green bond issuances as well as the issuers’ daily stock prices are retrieved from Bloomberg. We use a one-factor market model to estimate the “normal” relation between stock and market returns since previous literature has shown that short-horizon event studies are not sensitive to the benchmark specification (Kothari and Warner, 2007):

$${R}_{it}={\alpha }_{i}+{\beta }_{i}\times {R}_{mt}+{\varepsilon }_{it}$$

where ${R}_{it}$ is the daily return on the common share of bank i on day t and ${R}_{mt}$ is the daily return on the MSCI World total return index on the same day. In line with Tang and Zhang (2020), the market model is estimated from 300 calendar days to 50 calendar days prior to the event date $t=0$, which coincides with the announcement date. We derive daily abnormal stock returns ${AR}_{it}$ as the difference between raw returns and returns estimated from the market model:

$${AR}_{it}={R}_{it}-\left({\widehat{\alpha }}_{i}+{\widehat{\beta }}_{i}\times {R}_{mt}\right)$$

As in Tang and Zhang (2020) and Flammer (2021), we then compute cumulative abnormal returns ${CAR}_{i}$ for bank i around each event date by aggregating ${AR}_{it}$ from 5 days before to 10 days after (5 days before to 5 days after) the announcement date. The table reports the average CARs computed across the two event windows separately for first-time issues and subsequent issues, with t-statistics in parenthesis. Consistently with the findings from previous studies on corporate green bonds in general, we document, also in banks: (i) a positive stock price response following the announcement of the first green bond issuance; (ii) no stock price response following the announcement of subsequent green bond issuances. This is in line with the signaling argument, as the market learns about the bank’s commitment to green lending with the first-time issue, while the information content of subsequent issues is likely to resemble that of conventional bond issues (see the discussion in Tang and Zhang, 2020, and Flammer, 2021).

	First-time issues		Subsequent issues
	[−5, 5]	[−5, 10]	[−5, 5]	[−5, 10]
	I	II	III	IV
CAR	1.276**	1.506*	−0.622	−0.986
	(1.982)	(1.952)	(−1.089)	(−1.412)
Observations	78	78	89	87

1. ∗, ∗∗, and ∗∗∗ denotes significance at the 10, 5, and 1% level

Appendix 2: Variable definitions

Variable	Definition	Source
Brown lending	The proportion of a bank’s credit exposure to carbon-intensive sectors	Carbon-intensive industries are classified by the Inter-governmental Panel on Climate Change (IPCC). Information on lending to such industries is obtained from annual reports and Pillar 3 reports
Buildings	Dummy variable taking the value 1 (0) if green bond proceeds were (not) used for green buildings	Documentation provided on the banks’ websites
Capital ratio	Ratio of equity over total assets	Moody’s analytics bankfocus
CBI Certified GBI	Dummy variable taking the value 1 (0) if an issuer who obtained a third-party certification of compliance with the Climate Bonds Standards of the Climate Bonds Initiative has (has not) already issued a green bond by the current year	Climate bonds initiative
∆CO2 emissions: scope 1 & 2	Yearly log change in reported Scope 1 and 2 emissions	Thomson reuter’s refinitiv
∆CO2 emissions: scope 3	Yearly log change in reported Scope 3 emissions	Thomson reuter’s refinitiv
∆CO2 emissions: total scope	Yearly log change in the sum of reported Scope 1, 2 and 3 emissions	Thomson reuter’s refinitiv
Customer deposit ratio	Ratio of total customer deposits over total assets	Moody’s analytics bankfocus
Emissions score	The bank’s emissions score	Thomson reuter’s refinitiv
Energy	Dummy variable taking the value 1 (0) if green bond proceeds were (not) used for renewable energy purposes	Documentation provided on the banks’ websites
Environmental score	The bank’s environmental score	Thomson reuter’s refinitiv
EPI country score	The Environmental Performance Index score of the country where the bank is headquartered	The EPI is computed by Yale University and available at: https://epi.yale.edu/
Government controlled	Dummy variable taking the value 1 (0) if the bank’s global ultimate owner is (is not) a public authority, state, or government	Moody’s analytics bankfocus
Green bond issuer (GBI)	Dummy variable taking the value 1 (0) if a bank has (has not) issued a green bond during the sample period	Bloomberg
GBI post issuance	Dummy variable taking the value 1 (0) if a GBI has (has not) already issued a green bond by the current year	Bloomberg
G-SIB	Dummy variable taking the value 1 (0) if the bank is (is not) a global systemically important bank	The yearly lists of G-SIBs are obtained from the website of the Financial Stability Board: https://www.fsb.org
Has ESG rating	Dummy variable taking the value 1 (0) if the bank has (has not) an ESG rating	Thomson reuter’s refinitiv
Institutional ownership	Percentage of total institutional ownership of the bank	Bloomberg
Intense GBI	Dummy variable taking the value 1 (0) if a bank has issued in the sample period a proportion of green bonds over total bonds higher (lower) than the median	Bloomberg
Intense GBI post issuance	Dummy variable taking the value 1 (0) if an Intense GBI by number of bonds has (has not) already issued a green bond by the current year	Bloomberg
Listed	Dummy variable taking the value 1 (0) if a bank is (is not) publicly listed	Moody’s analytics bankfocus
Ln(Assets)	Natural log of total assets in million USD	Moody’s analytics bankfocus
Loan ratio	Ratio of gross loans over total assets	Moody’s analytics bankfocus
Long-term funding ratio	Ratio of long-term funding over total assets	Moody’s analytics bankfocus
NDC	Dummy variable taking the value 1 (0) for each year a country in which a bank is headquartered has (has not) a Nationally Determined Contribution in place	https://www.climate-resource.com/tools/ndcs
Other	Dummy variable taking the value 1 (0) if green bond proceeds were (not) used for land, water, or pollution prevention and waste management	Documentation provided on the banks’ websites
ROA	Ratio of net income over total assets	Moody’s analytics bankfocus
Transport	Dummy variable taking the value 1 (0) if green bond proceeds were (not) used for green transport projects	Documentation provided on the banks’ websites
UNEP FI	Dummy variable taking the value 1 (0) if the bank is (is not) a member of the United Nations Environment Programme Finance Initiative	Member lists obtained from: https://www.unepfi.org/members

Appendix 3: t-tests on treated and control samples

	Treated		Control
	Mean	St. dev.	Mean	St. dev.	t-stat.
Environmental score	65.19	25.53	64.59	24.82	− 0.14
Ln(Assets)	19.09	1.58	18.63	1.27	− 1.90
Capital ratio	0.11	0.17	0.11	0.11	− 0.08
ROA	0.01	0.01	0.01	0.03	0.87
Loan ratio	0.50	0.21	0.49	0.23	− 0.25
EPI country score	70.48	13.92	70.09	13.73	− 0.17
Observations	69		69

1. This table reports summary statistics and t-tests for difference in means for treated (green bond issuers) and control (non-green bond issuers) banks from a matched sample. The matching is based on bank characteristics of bond issuers at year-end before the first green bond was issued. All variables are described in Appendix 2.
2. ∗, ∗∗, and ∗∗∗ denotes significance at the 10, 5, and 1% level, respectively.